Tam Metin PDF - Akademik Platform Mühendislik ve Fen Bilimleri
Transkript
Tam Metin PDF - Akademik Platform Mühendislik ve Fen Bilimleri
Volume: 4 Issue :1 Year : 2016 i Volume 4 / Issue 1 Journal of Engineering and Science Editor in Chief Prof.Dr.Mehmet SARIBIYIK, Sakarya University, Turkey mehmets@sakarya.edu.tr Layout Editors Prof.Dr. Barış Tamer TONGUÇ, Sakarya University, Turkey btonguc@sakarya.edu.tr Assoc. Prof. Dr. Fatih ÇALIŞKAN, Sakarya University, Turkey fcaliskan@sakarya.edu.tr Asst. Prof. Dr. Hakan ASLAN, Sakarya University, Turkey haslan@sakarya.edu.tr Support Lec. Gökhan ATALI, Sakarya University, Turkey gatali@sakarya.edu.tr Members of Advisory Board Prof. Dr. Abdullah Çavuşoğlu, Council of Higher Education, Turkey Prof. Dr. Mehmet Emin AYDIN, University of West of England, England Prof. Dr. Erol ARCAKLIOĞLU, The Scientific and Technological Research Council Prof. Dr. Fahrettin ÖZTÜRK, The Petroleum Institute, The United Arab Emirates Prof. Dr. Ahmet TÜRK, Celal Bayar University, Turkey ISSN: 2147-4575 Contact Academic Platform info@apjes.com http://apjes.com/ ii Journal of Engineering and Science Contents 1-7 Image Damage Analysis With Morphological Image Processing Technique Using Artificial Neural Networks 8-14 Engineering Models of Masonry by Joint Repairing Techniques 15-20 Development and Characterization of Parian Bodies Using Feldspar from Two Selected Deposits in Nigeria Fractional Distribution of Heavy Metals from the Tailings of Itagunmodi Goldmine Site Osun State, Nigeria 21-27 28-37 The Solution of Shift Scheduling Problem by Using Goal Programming 38-47 Design of a Novel Architecture for QPSK Modulator iii Journal of Engineering and Science İçindekiler Morfolojik Görüntü İşleme Tekniği ile Yapay Sinir Ağlarında Görüntü Tahribat Analizi 1-7 8-14 Engineering Models of Masonry by Joint Repairing Techniques 15-20 Development and Characterization of Parian Bodies Using Feldspar from Two Selected Deposits in Nigeria 21-27 Fractional Distribution of Heavy Metals from the Tailings of Itagunmodi Goldmine Site Osun State, Nigeria Hedef Programlama ile Nöbet Çizelgeleme Probleminin Çözümü Yeni Bir QPSK Modülatör Mimarisinin Tasarımı 28-37 38-47 iv G. ATALI/APJES IV-I (2016) 01-07 Morfolojik Görüntü İşleme Tekniği ile Yapay Sinir Ağlarında Görüntü Tahribat Analizi Gökhan Atalı, 2S.Serdar Özkan, 2Durmuş Karayel Sakarya Meslek Yüksekokulu, Mekatronik Programı Sakarya Üniversitesi, Türkiye 2 Teknoloji Fakültesi, Mekatronik Mühendisliği Sakarya Üniversitesi, Türkiye *1 *1 Geliş Tarihi: 2015-12-25 Kabul Tarihi: 2016-04-09 Öz Teknolojinin gelişmesine paralel olarak kamera sistemlerinde yüksek düzeyde kaliteyi öngören lensler tasarlanmıştır. Ancak bu lensler yapısal olarak her ne kadar görüntü kalitesinde başarılı olsalar da üzerinde tahribat meydana gelmiş görüntülerin ayırt edilmesinde herhangi bir ek özellik mevcut değildir. Esasen bu tür tahribatların giderilmesi için yapay zeka teknikleri kullanmak mümkündür. Bu çalışmada, yapay sinir ağları kullanarak üzerinde tahribat meydana gelmiş görüntülerin morfolojik görüntü işleme teknikleri ile birleştirilerek tahribatın derecesine göre orijinal görüntüye yakınsaması ele alınmıştır. Ayrıca bu amaca ulaşmak ve kullanımı kolaylaştırmak amacıyla geliştirilen arayüz sayesinde eğitim ve test verilerinin yanı sıra ağı oluşturmak için kullanılacak parametrelerin kolaylıkla hazırlanan algoritmaya entegrasyonu sağlanmaktadır. Anahtar Kelimeler: Yapay Sinir Ağları, Karakter Tespiti, Morfolojik Görüntü İşleme Image Damage Analysis With Morphological Image Processing Technique Using Artificial Neural Networks *1 Gökhan Atalı, 2S.Serdar Özkan, 2Durmuş Karayel Sakarya Meslek Yüksekokulu, Mekatronik Programı Sakarya Üniversitesi, Türkiye 2 Teknoloji Fakültesi, Mekatronik Mühendisliği Sakarya Üniversitesi, Türkiye 1 Abstract The lenses providing high quality for camera systems are designed as parallel with developing technologies. These lenses do not have any additional functionality to distinguish damaged images while they are successful with regard to image quality. Essentially, artificial intelligence techniques can be used to eliminate such damaged cases. In this study, convergence to original image of the damaged images according to the degree of damage using together artificial neural networks and morphological image processing techniques are discussed. Also, it is provided to be integrated training and test data with used algorithm thanks to developed interfaces to achieve goal and to facilitate use. In addition, this interface is used to be entered the training parameters to the system. Keywords: Artificial Neural Networks (ANN), Character detection, Morphological image processing 1. Giriş Günümüzde birçok alanda morfolojik görüntü işleme teknikleri ile çeşitli analizler yapılmaktadır. Bunlardan en bilinenleri; araç plakası tanıma, bant üzerindeki ürünlerin tanımlanması, cisimlerin çap ve boy uzunlukları için görüntü analizleridir. Görüntü işleme esasen görüntünün sayısallaştırılarak veri setlerine dönüştürülmesi ve çeşitli yöntemlerle işlenmesi olarak tanımlanır. Bu konu ile ilgili yapılan çalışmalar incelendiğinde araştırmacıların; kenar ayrıştırma, Hough dönüşümü, simetri özelliği, renk özelliği, histogram analizi, Gabor süzgeçleri gibi görüntü işleme tekniklerini kullandıkları görülmektedir [1-8]. *Corresponding author: Address: Sakarya Meslek Yüksekokulu, Mekatronik Programı Sakarya Üniversitesi, 54187, Sakarya TÜRKİYE. E-mail address: gatali@sakarya.edu.tr, Telefon: +902642953495 Doi:10.21541/apjes.27271 G. ATALI/APJES IV-I (2016) 01-07 Bu konu hakkında Literatür incelendiğinde; Fatih Kahraman ve arkadaşları aktif görünüm modeline dayalı yüz tanıma isimli çalışmalarında insan yüzünün temel bileşenlerini otomatik olarak saptayan aktif görünüm modeline ve Gabor süzgeçlerine dayalı bir yüz tanıma yöntemi geliştirmişlerdir [9]. C.Tu ve arkadaşları Hough dönüşümünü kullanarak araçların belirli bir rota üzerindeki pozisyonlarının bulunmasını amaçlamış ve bu konuda bir çalışma gerçekleştirmişlerdir [10]. C. Lopez-Molina ve arkadaşları ayrıt saptama yöntemine ait performans analizleri üzerine çalışma yapmışlardır [11]. Görüntü üzerinde yer saptama için en yaygın olarak kullanılan yöntemlerin başında ayrıt saptama ve eşikleme tabanlı yöntemler gelmektedir. Ayrıt saptama ve eşikleme giriş görüntüsünü 1 ve 0 bilgilerinden oluşan ikili resme dönüştürmek için kullanılır. Daha sonra elde edilen bu ikili resmin dikey ve yatay izdüşüm histogramları analiz edilerek resmin üzerinde istenilen bölgeler tespit edilir [12-17]. Diğer bir yöntemde ayrıtlar bulunduktan sonra Hough dönüşümü uygulanarak resmin çevresi bulunmaktadır [18]. Bahsedilen bu yöntemler ile elde edilen ikili moda dönüştürülmüş sayısal görüntü üzerinde morfolojik açma ve kapama gibi işlemler uygulanarak görüntüyü gerçek halinden uzaklaştırmak mümkündür. Esasen burada gerçek görüntüden uzaklaşmadan maksat, görüntü üzerinde tahribat meydana geldiği manasındadır. Çevre şartlarından ve metnin yazılı olduğu zeminden kaynaklı problemler gibi birçok olumsuz etki görüntü üzerine yansımakta ve görüntüyü gerçek halinden uzaklaştırmaktadır. Bu çalışmada bahsedilen bu etkilere benzer şekilde görüntü üzerinde morfolojik açma ve kapama işlemleri uygulanarak görüntü üzerinde tahribat meydana getirilmiştir. Üzerinde değişiklik meydana gelerek farklı bir görüntü haline gelen yeni görüntü, tahribat analizi yaparak doğruya yaklaşımda bulunan bir yapay sinir ağına (YSA) test verisi olarak sunulmuş ve analiz edilmiştir. Bu işlemler sırasında eğitim verileri için hazırlanan resimler üzerinde ayrıt saptama, eşikleme yöntemleri kullanılmış ardından oluşan ikili bilginin dikey yatay izdüşümleri bir veri setine dönüştürülerek MATLAB ortamında yapay sinir ağları tarafından oluşturulan bir ağa eğitim verisi olarak tanıtılmıştır. Ayrıca çalışmada MATLAB ortamında bir arayüz tasarlanarak veri setlerinin eğitimi ve testleri için uygulanacak fonksiyon çeşitleri, ara katman sayısı, ara katmandaki nöron sayısı, açma-kapama miktarı gibi bilgilerin dışarıdan müdahale ile değiştirilebilir olması sağlanmıştır. 2. Yapay Sinir Ağları ve Görüntü İşleme Yapay Sinir Ağları (YSA), insan beyninin bilgi işleme tekniğinden esinlenerek geliştirilmiş bir bilgi işlem teknolojisidir. YSA ile basit biyolojik sinir sisteminin çalışma şekli simüle edilmektedir. Simüle edilen sinir hücreleri nöronlar içerirler ve bu nöronlar çeşitli şekillerde birbirlerine bağlanarak ağı oluştururlar. Bu ağlar öğrenme, hafızaya alma ve veriler arasındaki ilişkiyi ortaya çıkarma kapasitesine sahiptirler. Diğer bir ifadeyle, YSA'lar, normalde bir insanın düşünme ve gözlemlemeye yönelik doğal yeteneklerini gerektiren problemlere çözüm üretmektedir. Bir insanın, düşünme ve gözlemleme yeteneklerini gerektiren problemlere yönelik çözümler üretebilmesinin temel sebebi ise insan beyninin ve dolayısıyla insanın sahip olduğu yaşayarak veya deneyerek öğrenme yeteneğidir. Görüntü işleme ölçülmüş veya kaydedilmiş olan dijital görüntü verilerini, elektronik ortamda çeşitli yazılımlar ile amaca uygun şekilde değiştirmeye yönelik yapılan çalışmaları kapsamaktadır. Görüntü işleme, daha çok, kaydedilmiş olan, mevcut görüntüleri işlemek, yani mevcut resim ve grafikleri, değiştirmek, yabancılaştırmak ya da iyileştirmek için kullanılmaktadır. 3. Morfolojik Görüntü İşleme Matematiksel morfoloji, lineer olmayan komşuluk işlemlerinde güçlü bir görüntü işleme analizidir. Morfolojik görüntü işlemede genişletme ve aşındırma isimli temel iki işlem kullanılmaktadır. Morfolojik görüntü işlemede bilinen açma ve kapama işlemleri gibi diğer tüm yöntemler bu iki işlemi referans alarak gerçekleştirilir. Üzerinde kare ve daire gibi geometrik şekillerle yapısal filtre uygulanan görüntü açma veya kapama gibi morfolojik işlemlere tabi tutulur. Ancak görüntü üzerinde yapısal filtre uygulayarak genişletme veya aşındırma işlemi yapabilmek için görüntü önce binary (ikili) moda çevrilir. 3.1. Genişletme İşlemi İkili moda dönüştürülen görüntü üzerinde büyütme ya da kalınlaştırma işlemlerinin yapıldığı morfolojik işlemleri kapsamaktadır. Sayısal bir resmi genişletmek resmi yapısal elemanla kesiştiği bölümler kadar büyütmek demektir. Kalınlaştırma işleminin nasıl yapılacağını Şekil 3’te örnek verilen yapı elemanları belirler. Şekil 1 de görüldüğü üzere üzerinde genişletme yapılan sayısal görüntüde açma meydana gelmiş ve dolayısıyla görüntüde normalin dışına çıkan bir bozulma gözlenmektedir. G. ATALI/APJES IV-I (2016) 01-07 Açma işlemi: A o B =( A B ) B Kapama işlemi: A ● B =( A B ) Şekil 1. 3x3 yapısal elemanı ile genişletme işlemi Yapısal eleman olarak adlandırılan ifade istenilen boyutlarda ve istenilen şekilde hazırlanmış matris formunda yapıları içermektedir. Yapısal eleman çeşitli geometrik şekillerden biri olabilmektedir; en sık kullanılan yapısal elemanlar kare, dikdörtgen ve daire şeklindedir. Yapısal eleman örnekleri Şekil 3’ te gösterilmiştir. Eğer morfolojik işlem olarak resimdeki nesnelerin keskin hatları silinip yerlerine kavisli veya daha yumuşak hatlar getirilmek isteniyorsa dairesel yapısal eleman kullanılmalıdır. 0 1 0 Şekil 2. 3x3 yapısal elemanı ile aşındırma işlemi 3.2. Aşındırma İşlemi İkili moda dönüştürülen görüntü üzerinde küçültme ya da inceltme işlemlerinin yapıldığı morfolojik işlemleri kapsamaktadır. Aşındırma işlemi bir bakıma genişletme işleminin tersidir. Aşındırma işlemi ile sayısal resim üzerinde inceltme yapılmış dolayısıyla görüntüde tahribat meydana gelmiş olur. Aşındırmadan kaynaklı bu tahribat sonucunda resim içerisindeki nesneler boyutsal olarak daralır, delik varsa genişler ve bağlı nesneler ayrılma eğilimi gösterir. Şekil 2 de aşındırma işlemi için bir örnek görüntü verilmiştir. Eğer sayısal bir görüntüye genişletme ve aşındırma işleminin ardışık olarak uygulanırsa görüntüde açma işlemi meydana gelmektedir. Açma işleminde birbirine yakın iki nesne görüntüde fazla değişime sebebiyet vermeden ayrılmış olurlar. Açmanın tersi olarak sayısal görüntü üzerinde aşındırma ve genişletme işleminin ardışık uygulanmasıyla da kapama işlemi meydana gelmektedir. Dolayısıyla birbirine yakın iki nesne görüntüde fazla değişiklik yapılmadan birbirine bağlanmış olur. Bu işlemlerin matematiksel gösterimi şu şekildedir; Genişletme: A Aşındırma: A B B B 1 1 1 0 1 1 1 0 0 1 0 1 0 1 1 1 1 0 1 0 0 Şekil 3. Yapısal eleman örnekleri 1 0 1 4. Uygulama Çalışması Bu çalışmada yapay sinir ağları ile üzerinde morfolojik görüntü işleme teknikleri uygulanmış sayısal görüntünün daha önceden ağa tanıtılan orijinal resimler ile karşılaştırılması ve ağın doğruya yaklaşımları incelenmiştir. Bu yaklaşımları yapabilmek için Şekil 4'te verilen yol ve yöntemler sırası ile gerçekleştirilmiştir. 4.1. Eğitim verilerinin alınması Geliştirilen bir ara yüz aracılığı ile ağda eğitime tabii tutulacak görüntülerin alınması bu basamakta gerçekleştirilmektedir. Ağda eğitilmesi düşünülen veri setleri 27x27 pixel boyutlarında (yaklaşık 18-24 punto) resimlerden oluşmaktadır. Bu resimler 50x50 boyutlarında sayısal 1 ve 0 değerlerinden oluşan sayısal veri seti haline getirildikten sonra 2500x1 sütun matrisine dönüştürülmektedir. Bu sayede görüntü yapay sinir ağlarında kullanılmak üzere hazır bir eğitim veri setine dönüştürülür. Veri seti olarak aşağıdaki gibi kullanılmıştır. Bu veri setlerinin %20'si eğitim için, diğer %80'i ise test için kullanılmıştır. G. ATALI/APJES IV-I (2016) 01-07 Eğitim verilerinin alınması Oluşturulacak ağ için parametrelerin belirlenmesi Test verilerini oluşturulan ağda eğitimi Test verilerine morfolojik teknik uygulanması Ağın eğitilmesi Test verilerinin alınması Sonuçlar ve karşılaştırma Şekil 4. Uygulamanın gerçekleşme basamakları 4.2. Oluşturulacak ağ belirlenmesi ve ağın eğitimi için parametrelerin Yapay sinir ağı oluşturmak için gerekli eğitim, performans ve transfer fonksiyonları ile ağda ulaşılması hedeflene nihai değer için gerekli parametrelerin girişi bu basamakta sağlanır. Ayrıca ilk katmandaki nöron sayısı ve maksimum epoch (devir) değeri de yine bu basamakta girilen parametreler arasında yer alır. Bu parametreler yapay sinir ağları oluşturulurken esas alınan temel parametrelerdir. Bu çalışmada izlenen yöntem ve teknikler için en uygun parametre dizisi Tablo 1 de verilmiştir. ile uygun matris formuna dönüştürülmüş ve yapay sinir ağına test verisi olarak sunulmuştur. Daha sonra orijinal görüntü ile üzerinde tahribat meydana gelmiş görüntü metinsel ve görüntü olarak incelemeye tabii tutulmuş ve çıkarımda bulunulmuştur. Tablo 1 de belirtilen parametreler dahilinde ağa sunulan eğitim verilerinin eğitimi sonucu Şekil 5 ve Şekil 6 da görüldüğü üzere regresyon değeri 0.999, gradyan değeri ise 148 iterasyonda 0.00010054 olarak saptanmıştır. 4.3. Test verilerinin alınması ve morfolojik teknik uygulanması Eğitilmiş ağda test etmek üzere test verilerinin alınması ve morfolojik teknik uygulayarak görüntü üzerinde tahribat meydana getirilmesi bu basamakta gerçekleştirilmiştir. Test verilerinin ağda eğitilmek üzere dosyadan alınması eğitim verileri için izlenen yöntem ve teknik ile aynıdır. Görüntüler sayısallaştırıldıktan sonra sayısal veri setine 1, 2, 3 ve 4 derecelik dairesel ortalama filtresi ayrı ayrı uygulanmış ve görüntü orijinalliğinden uzaklaştırılarak, görüntüde tahribat meydana getirilmiştir (Şekil 7). Üzerinde değişiklik yapılan görüntü köşe bulma yöntemi ve ayrıt saptama tekniği Şekil 5. Ağın eğitimi G. ATALI/APJES IV-I (2016) 01-07 Şekil 6. Ağın eğitim sonuçları Tablo 1. Belirlenen ağ parametreleri Eğitim fonk. Performans fonksiyonu trainscg mse Transfer fonk. Max. Epoch Hedef İlk katmandaki nöron sayısı Logsig 1000 1e-5 10 Şekil 7. 4 derecelik dairesel ortalama filtresi uygulanmış görüntü 5. Sonuçlar ve Tartışma Görüntüler üzerinde çeşitli etkenlerden dolayı meydana gelen tahribatlar, morfolojik görüntü işleme platformundan yararlanılarak görüntü üzerinde oluşturulmuş ve yapay sinir ağları kullanılarak üzerinde tahribat meydana gelmiş görüntünün gerçeğe yakınlığı test edilmiştir. Görüntüde meydana gelebilecek bu tahribatlar dört farklı derecede dairesel ortalama filtresinin görüntüye uygulanması ile gerçekleşmiş ve sonuç olarak elde edilen değerler Tablo 2 ve Şekil 8’de sunulmuştur. Bu değerlere göre geliştirilen yapay sinir ağında, 2 ve 3 derecelik dairesel ortalama filtresi ile oluşturulan morfolojik görüntüde ortalama yüzde 75 başarım sağlanırken aşınma şayet 4 dereceye çıkarılırsa görüntünün orijinalliğinden oldukça uzaklaşması sebebiyle yaklaşımda da azalmanın görüldüğü sonucuna varılmıştır. Ayrıca tablo 2 de bahsi geçen yaklaşımlar var-yok olarak nitelendirilmiş ve bu doğrultuda yüzde olarak ifade edilmiştir. İlerleyen çalışmalarda yaklaşım değerleri fuzzy lojik ya da neuro fuzzy kullanılarak geniş aralıklarda ifade edilebilir. Meydana gelebilecek tahribatın Gaussian alçak geçiren filtresi, motion G. ATALI/APJES IV-I (2016) 01-07 hareket benzetimi filtresi gibi değişik filtreler altında da incelenmesi sağlanabilir. Ayrıca bu çalışmadakine benzer uygulamaları içeren kamera lensleri tasarlanarak görüntülerin algı esnasında analizleri sağlanabilir. Tablo 2. YSA yaklaşım karşılaştırmaları Morfolojik Test verisi sonuç A A B C C C D D AC AC AD AC ACB ACD BAC CAC CAB CAD CBA CDA ABCD ACCD Ortalama Yaklaşım: Açma Derecesi 1 1 1 1 1 1 1 1 1 1 1 Morfolojik sonuç A C C Test verisi A B C D D AC AC AD AC ACB ACA BAC BAC CAB CAB CBA CBA ABCD ACCD Ortalama Yaklaşım : Açma Derecesi 2 2 2 2 2 2 2 2 2 2 2 Morfolojik sonuç A A D D AC AD ACB BAC CAC CCA ABCA Test verisi A B C D AC AD ACB BAC CAB CBA ABCD Yaklaşım (%) 100 0 100 Açma Derecesi 4 4 4 Morfolojik sonuç A A C Test verisi A B C 100 100 50 66 100 100 100 50 0 100 50 66 66 66 66 50 4 4 4 4 4 4 4 4 C AC AC ACC AAC CAD CDA ADCA D AC AD ACB BAC CAB CBA ABCD 78,727 60,364 Yaklaşım (%) 100 0 100 100 100 50 66 66 66 66 50 69,455 Yaklaşım (%) 100 0 0 100 100 100 100 100 66 66 75 73,364 Açma Derecesi 3 3 3 Yaklaşım (%) 100 0 100 3 3 3 3 3 3 3 3 YSA yaklaşımı Yaklaşım (%) 100 80 69,455 73,364 78,727 60 60,364 40 20 0 1 2 3 4 Açma derecesi Şekil 8. 1-2-3-4 derecelik açma derecelerine karşı YSA yaklaşımı G. ATALI/APJES IV-I (2016) 01-07 Referanslar [1] Barroso, J., Rafael, A., Dagless, E. L., Bulas-Cruz, J., Number plate reading using computer vision, IEEE – International Symposium on Industrial Electronics ISIE’97, Universidade do Minho, Guimarães, 1997. [2] Morphological Segmentation for Textures and Particles, Published as Chapter 2 of Digital Image Processing Methods, E. Dougherty, Editor, MarcelDekker, New York, 1994, Pages 43--102. [3] B. Hongliang and L. Changping. A hybrid license plate extraction method based on edge statistics and morphology.17th International Conference On Pattern Recognition(ICPR’04), 2:831–834, 2004. [4] M. Sarfraz, M. J. Ahmed, and S. A. Ghazi. Saudi arabian license plate recognition system. Proceedings of the 2003 International Conference on Geometric Modeling and Graphics(GMAG’03), pages 36–41, 2003. [5] V. Kamat and S. Ganesan. An efficient implementation of hough transform for detecting vehicle license plate using dsp’s. 1st IEEE Real-Time Technology and Applications Symposium, pages 58– 59, 1995. [6] V. Shapiro, D. Dimov, S. Bonchev, V. Velichkov, and G. Gluhchev. Adaptive license plate image extraction. International Conference on Computer Systems and Technologies, 2003. [7] F. Mart´ın, M. Garc´ıa, and J. L. Alba. New methods for automatic reading of vlps (vehicle license plates). Signal Processing Patten Recognition and Application, 2002. [8] Kahraman F., Gökmen M. “GABOR Süzgeçler Kullanılarak Taşıt Plakalarının Yerinin Saptanması”, 11. sinyal İşleme ve İletişim Uygulamaları Kurultayı, İstanbul, 2003. [9] F. Kahraman, B.Kurt, M.Gökmen "Aktif Görünüm Modeline Dayalı Yüz Tanıma" Signal Processing and Communications Applications Conference, 2005. Proceedings of the IEEE 13th, May 2005, Pages 483486, Print ISBN: 0-7803-9239-6. [10] C.Tu, B.J van Wyk, Y. Hamam, K. Djouani, Shengzhi Du "Vehicle Position Monitoring Using Hough Transform" IERI Procedia Volume 4, 2013, Pages 316–322 2013 International Conference on Electronic Engineering and Computer Science (EECS 2013). [11] C.Lopez-Molina, B. De Baets, H. Bustince "Quantitative error measures for edge detection" Pattern Recognition Volume 46, Issue 4, April 2013, Pages 1125–1139. [12] P. Ponce, S. S. Wang, D. L. Wang, “License Plate Recognition-Final Report”, Department of Electrical and Computer Engineering, Carnegie Mellon University, 2000. [13] M. Yu and Y. D. Kim, ``An Approach to Korean License Plate Recognition Based on Vertical Edge Matching", IEEE International Conference, vol. 4, 2975-2980, 2000. [14] J.R. Parker, P. Federl, ``An Approach To Licence Plate Recognition", The Laboratory For Computer Vision, University of Calgary, 1996. [15] Cui Y., Huang Q., Extracting Characters of License Plates from Video Sequences, Machine Vision and Applications 10, 308-320, 1998. [16] Naito, T., Tsukada, T., Yamada, Yamamoto, S., Robust License-Plate Recognition Method for Passing Vehicles under Outside Environment, IEEE Trans. Vehicular Technology 49, 2309-2319, 2000. [17] Nishiyama, K., Kato, K., Hinenoya, T.: Image processing system for traffic measurement, Proceedings of International Conference on Industrial Electronics, Control and Instrumentation Kobe, Japan, (1991) 1725–1729. [18] Lu, Y., Machine printed character segmenation, Pattern Recognition, vol. 28, n. 1, 67-80, Elsevier Science Ltd, UK, 1995. K. KAPTAN/APJES IV-I (2016) 08-14 Engineering Models of Masonry by Joint Repairing Techniques Kubilay Kaptan Beykent University, Civil Engineering Department, Istanbul, TURKEY Geliş Tarihi: 2015-12-05 Kabul Tarihi: 2016-05-17 Abstract When repointing historic masonry, it is the quality of the bond between mortar and stones that decides on the lifecycle of the structure. Once the composite system or the mortar start cracking, moisture can penetrate into the masonry and destroy the system. What mortar to use for what kind of masonry is normally an empirical decision. But in how far the mortar eventually selected is really suited for the purpose in question will not turn out until several years later. It is with this knowledge in mind that a simple engineering model has been developed, which is easy to use and which is to permit the likelihood of cracks to be assessed quantitatively. The model is based on calculations made for stresses occurring on the surface of the masonry and only requires a few material parameters. A combined, complex research model is being developed, which is to provide for exact structural analysis. For this model, the temperature and moisture transport is calculated with the aid of an FDM program. The temperature and moisture fields thus determined are then transferred to an FEM program which uses the material models of Rots (1997), Lourenço (1996) and Van Zijl (2000) for stress and deformation calculation. Keywords: FDM program, Masonry, Joint Repairing Techniques 1. Introduction Conservation of historic structures normally involves rehabilitation of joints, and the jointing mortar has the function of providing weathering protection. In particular in case of rehabilitation measures extending far into the masonry, the mortar also has to be able to transmit forces. An essential condition for the durability of such repair meas-ures is that the bond between stone and joint mortar is of a good quality and does not show any cracks. The decision as to what kind of mortar to use for joint repair measures in natural stone masonry of historic buildings is usually a question of experience, while trying to give due regard to preservation requirements. Whether or not the masonry mortar or joint mortar chosen is actually suited for the given kind of masonry often does not show until it has been in place for several years. A major criterion is the weather protection for the masonry, i.e. protection against weathering of the stones and mortar destruction, which depends in particular on the crack-free bond between stone and joint mortar. Even if a joint mortar itself has good weather protection properties, the mortar/stone flank bond region is a critical weak spot for the durability of masonry. Since the stones and the mortar in new joints tend to differ in their deformation behaviour (which is the result of differences in their thermal, hygral and mechanical properties), cracks are likely to occur between stones and mortar, or in the mortar itself. Material qualification tests alone do not suffice to predict the occurrence of cracks in the composite stone / mortar system. To be able to assess the risk of cracking, a large number of tests have to be per-formed on composite stone / mortar elements. Since historic buildings are made from a variety of different stones (normally natural stones whose properties tend to vary considerably), the bond characteristics would have to be examined separately for each structure requiring rehabilitation (Grazzini, 2006). This would not only be very costly, but also rather timeconsuming. Another aspect is that different kinds of mortar are generally used in a particular structure. Mortar in the base region will not be the same as that in the ris-ing masonry or on inclined surfaces. This large number of factors would increases the test requirements considerably. However, if it should be possible to use models to predict the durability of new joints in historic masonry for defined boundary conditions, such costly and time consuming tests could be either limited or be avoided altogether. Broadly based parameter analy-ses made before starting any rehabilitation measures will then allow the suitability of a mortar to be reviewed for the application in question. Should the mortar be found to be inadequate, the properties of the mortar can be varied to decide what changes need to be made to produce a joint that is free from cracks. *Corresponding author: Kubilay Kaptan, Address :Beykent University, Civil Engineering Department, Istanbul, TURKEY, E-mail Address: kaptankubilay@gmail.com Doi:10.21541/apjes.61359 K. KAPTAN/APJES IV-I (2016) 08-14 2. Causes of Cracks For the development of the composite structure models below, the cause for cracking must be known. The criteria primarily considered as a first step in developing the model are the mechanical/physical material properties and the residual and the restraint stress resulting from such properties. Stones and mortar are characterised by specific thermal and hygral behaviour. Irregular temperature and moisture distribution (see Fig. 1), which itself is the result of atmospheric conditions, produces constrained thermal and hygric strains. Fig. 2 is a schematic representation of the thermal strains in natural stones at the surface of the masonry, which are produced by changes in ambient temperatures. During summer months, the surface of the natural stone facade heats up considerably due to its exposure to direct sunlight during the day. At night, the surface of the facade cools down to the temperature level of the ambient air. Temperature differences of up to 50 °C at the surface are therefore quite normal. This difference in temperature produces strains in the stones and Figure 1: Thermal and hygric exposure of historical masonry 3. Models for Cracking in New Joints 3.1. Engineering model The engineering model for the durability of the composite system natural stone / mortar joint in the mortar, which because of the mutual deformation restraint in turn gives rise to restraint stress. In winter, the entire facade cools down to very low temperatures. The result are tensile stresses in the mortar and in the stones, and adhesive tensile stress in the bond region. Stresses acting on the bond primarily in nearsurface regions of the masonry are hence a function of moisture and temperature fields and they are subject to stress relaxation. This means they not only region specific but also time specific characteristics (Bocca, et al., 2011). The consequence of restrained deformation normal to the joint flank can be flank failure. Deformation along the joints is limited by internal constraints (Bocca and Grazzini, 2012). The result are residual stresses which can make the mortar crack transverse to the joint. The bond resistance R is determined by the tensile strength of the stone ft,St and of the mortar ft,Mo, and by the adhesive tensile strength ft,a. The lowest value is always the decisive one. The tensile strength is determined by the moisture level and, in the case of the mortar, also by the time. Figure 2: Deformation of natural stones in historical masonry as a result of thermal elongation connection with repointing of historic masonry developed by Schmidt-Döhl and Rosásy (2000) is used as a simple means of modelling the bond behaviour. The engineering model starts from the assumption that stress that can lead to cracks is the result of irregular temperature and moisture K. KAPTAN/APJES IV-I (2016) 08-14 distribution across the masonry cross section. Thermal and hygral strain at the surface is restrained by the inner masonry structure. The basic function of the model is to calculate stresses at the surface, starting from the simplifying assumption of a fully constrained composite stone / mortar element and maximum temperature difference: (1) T S el , pl C 0 (where T = Thermal strain, S = Shrinkage strain, el,pl = Elastic-plastic strain and C = Creep strain) Under conditions of full constraint, the sum total of all strain components has to be 0 at the surface. For the cases flank cracking (crack in parallel with the joint, Fig. 3) and mortar cracking (crack normal to the longitudinal direction of the joint, Fig. 4) the different strain components are examined more closely. Figure 3: Cracking parallel to joint Figure 4: Cracking normal to the joint stone in the composite stone / mortar system. Respecting flank failure, the model starts from series arranged mortar and stone: (4) t el , pl Esec, Mo E l0, Mo (1 l0, Mo ) sec, Mo Esec,St C t ( Ct , Mo l0, Mo EMo Ct , st (1 l0, Mo ) Est ) Creep strain C is calculated from the actual stress t, the creep coefficients Ct of mortar and stone, the modulus of elasticity E of mortar and stone, as well as the area percentage l0 of mortar and stone (Alberto, et al., 2011). Plugging equations 2 to 5 into eq. 1 and solving the equation for the maximum stress the composite stone / mortar system can take, or for the modulus of elasticity of the mortar, yields equations 6 and 7: t l0, Mo (T Tmax, Mo S ,, Mo ) l0, St (T , St Tmax, St S ,, St ) l0, Mo l0, St Esec, Mo 1/ Esec,St Ct , Mo l0, Mo / EMo Ct , St l0, St / ESt Esec, Mo (6) 3.2. Crack initiation parallel to joint (side cracks) Thermal strain T is calculated with the aid of the coefficient of thermal expansion T of mortar and stone, the maximum temperature difference Tmax occurring between mortar and stone or the constraining action of the inside of the masonry (cf. Figs. 1 and 2), and the percentage of area l0 taken up by mortar and stone: T T , Mo Tmax, Mo l0, Mo T , St Tmax, St (1 l0, Mo ) (2) Shrinkage strain S is calculated with the aid of the final degree of shrinkage S, of mortar and stone, and the area percentage l0 of mortar and stone (Collepardi, 1990). The final degree of shrinkage is used for simplification, because it is expected that the relative moisture in the mortar and stone surfaces decisive for cracking will very quickly follow any changes in the relative moisture of the ambient air and that the constraint-induced shrinkage strain will be produced at the surface: S S ,, Mo l0, Mo S ,, St (1 l0, Mo ) (3) Elastic-plastic strain el,pl is the result of the actual stress t and the secant modulus Esec of mortar and stone, and of the area percentage l0 of mortar and (5) EMö (l0, Mo ( T , Mo Tmax, 1 ) l ( Mo S , , Mo 0, St T , St Tmax, St S , , St ) t Ct , St l0, St / ESt l0, St / Esec, l0, Mö Ct , Mö l0, Mö (7) 3.3. Crack initiation normal to joint (mortar cracks) The risk of crack propagation perpendicular to the joint is assessed by connecting mortar and stones in parallel rather than in series. When compared with the residual stress in the mortar, the influence of the stones on crack propagation in the mortar is insignificant. This is why in this case the engineering model is restricted to the mortar and does not consider a composite stone / mortar system. Again, considerations start from a fully restrained system and the maximum temperature difference. The thermal strain is calculated with the aid of the thermal coefficient of expansion T of the mortar and the maximum difference in temperature Tmax between mortar and the restraining masonry: (8) T T , Mo Tmax, Mo S S ,, Mo (9) St K. KAPTAN/APJES IV-I (2016) 08-14 The shrinkage strain corresponds to the relevant final degree of shrinkage S, of the mortar.The elastic-plastic strain follows from the actual stress t and the secant modulus Esec of the mortar el , pl C t (10) Esec, Mo t Ct , Mo (11) EMo The creep strain C can be calculated from the actual stress t, the creep coefficient Ct of the mortar, and the modulus of elasticity E of the mortar. Plugging equations 8 to 11 into eq. 1 and solving the equation for the maximum stress the mortar can take, or for the modulus of elasticity of the mortar, yields equations 12 and 13. t EMo EMo T , Mo Tmax, Mo S , , Mo 1 Ct , Mo t (1 Ct , Mo ) T , Mo Tmax, Mo S , , Mo (12) (13) 3.4. Implementation and application of the engineering model Equations 6 and 7, as well as 12 and 13, form the basis for the engineering model which is applied in the form of a Microsoft Access® database. Respecting the variables in equations 2 to 5 and 8 to 11 the following distinctions can be made: 1. Parameters established on the structure : Area percentage of mortar and stone 2. Parameters established experimentally or from databases coefficient of thermal expansion T of mortar and stone final degree of shrinkage S, of mortar and stone maximum temperature differences T between mortar and stone creep coefficients Ct of mortar and stone modulus of elasticity of stone Should these parameters not be established experimentally, they can be assessed with the aid of the engineering model or they can be imported from the data records in the database. 3. Values established with the engineering model and serving as a basis for mortar selection stress t normal and perpendicular to the joint flank modulus of elasticity EMo of the mortar. Stress t must not be greater than the strength of the mortar, the strength of the stone or the bond strength. It has been developed a graphical user interface using mortar and stone data available from literature and data compiled from our own investigations and analyses. This database can be used for rough parameter studies to be able to select mortars that promise to be a good choice for a given masonry, and it can alternatively be used to determine the requirements the intended mortar has to meet. For verification of the model, the cracking behaviour in the region of the joint of restrained two-stone bodies was examined for constant climatic conditions and for one-sided weather exposure (Schmidt-Döhl and Rosásy 2000).A total of three test series were run, all of them using dolomite rock from the Harz mountains and green sandstone from Rüthen as natural stone, and mortar based on granulated blast-furnace slag and gypsum. Stone and joint deformations, and deformations beyond the joint were measured continuously, as were the temperatures in the joint mortar. The cracking behaviour was assessed weekly, and the moisture content of mortar and stone was determined gravimetrically once at the end of each test series. Changes in the mortar temperature were in addition measured at five points of a masonry section at depths of approx. 2.5, 5, 10, 35 and 60 centimetres. Even though the model only starts from linearelastic material behaviour (while considering timespecific deformation), experimentally determined results could be shown with a high degree of approximation. But the accuracy of the model is limited. Because it has so far been formulated as a deterministic model, it does, for instance, not account for the considerable variation of properties of natural stone (Fassina, et al., 2002). Much thought is at the moment being given to the possibility of automated parameter studies. These would also account for the variation in the mortar and stone properties, provided they have been stored in the database. Another aspect which is at the moment not included in the calculation is the bond shear strength, which is why shear stress perpendicular to the crack front is not accounted for. Neither does the model at the moment consider any chemical degradation processes and frostinduced processes, such as the degradation of mortar properties as a result of weathering. K. KAPTAN/APJES IV-I (2016) 08-14 Figure 5. Calculated Results of Flank Cracking of Three Different Gypsum-Lime-Mortars (R) and Two Different Bricks (SLB = sand-lime brick – CB = clay brick) Fig. 5 shows the results of comparative calculations using the engineering model for flank failure under temperature load case ΔT=5K. In this case, the bond between three different gypsum-lime mortars and calcareous sandstone or highly absorbent bricks is considered (Twelmeier, et al., 2008). Once the maximum stresses exceed the measured bond strength, the flanks will fail. It is evident that the stress-reducing effect of the creep deformation of gypsum mortar has been considered in a very realistic manner. Masonry samples exposed to this temperature load case showed flank failure in the same specimens as had been forecast in the model. 3.5. Research model For the time being no model is available that would be able to describe both heat and moisture transport, and the complex material behaviour of masonry (shrinkage, thermal strain, creep, relaxation, failure patterns) with a high degree of precision. One reason is the highly complex dependence of the material behaviour on moisture and temperature. This dependence pattern produces coupled differential equations that have so far not been solved satisfactorily with the FEM method (Van Zijl 2000). Up to the point at which cracking starts, hygral and thermal transport can be assumed to be a process that is independent of the mechanical condition of the system. This is why a model has been developed which combines the detailed sub-models (Sperbeck 2004). Transport processes are calculated with a program based on the finite-difference method (FDM). This also provides for realistic determination of transport processes under real climatic conditions, including the effects of solar radiation and driving rain. Results of the timespecific thermal and moisture fields are transmitted to an FEM program, which uses the material models of Rots (1997), Lourenço (1996) and Van Zijl (2000) to calculate the resultant deformations, stresses, and cracking, due regard being given to viscous and plastic material behaviour. Figure 6. Geometric model and deformation conditions in Figure 7. Two-dimensional illustration for the the research model wall cross section / Utilization of symmetries K. KAPTAN/APJES IV-I (2016) 08-14 The research model is to serve as a basis for extensive and effective analyses before starting rehabilitation measures, while allowing the number of pre-rehabilitation tests to be reduced substantially. Quantitative determination of the deformation and stress components, sensitivity analyses etc. give more detailed insight into the possible cause of cracks. The research model also permits the moisture distribution to be assessed for the entire cross section as a function of time. So far, the model has been used to describe two-stone bodies (see Fig. 6), in which heat and moisture transport processes were still simulated separately by making use of the symmetry (see Fig. 7). Figure 8. Moisture content and deformation pattern across the cross section 4. Conclusions The simple engineering model offers a tool that permits the likelihood of cracks in new joints to be assessed in a realistic manner. There is good agreement between the results calculated with the engineering model and the results of experimental tests. On the whole, the cracking pattern was forecast correctly. First coupled calculations using the more complex research model also produce plausible results. Model development aims at providing an instrument that permits a better understanding of the failure mechanisms in the bond between natural stone and mortar joint. More broadly based experiments are essential for verification of both models. 5. References [1] Alberto A., Antonaci P., Valente S. 2011. Damage analysis of brick-to-mortar interfaces. In Proceedings of 11th International Conference on Figs. 8 and 9 show the results of moisture distribution, stress distribution and deformations for a two-stone body when dried for 100 days (initial situation: masonry with 90 % rel. air humidity; air with 50 % rel. air humidity). The expected cracking pattern as a result of the high dryness could be approximated with a high degree of precision, which is what measurements during test programmes cannot achieve. Another advantage is that climatic conditions can be simulated at random and that the numerical model can be used for probabilistic analyses. In this way it can also be determined under what conditions the bond between mortar and stone is particularly likely to fail. Figure 9. Detail stone-joint: resulting stress in ydirection and deformation pattern the Mechanical Behavior of Materials, 1151-1156, Como Lake (Italy). [2] Bocca P., Grazzini A., Masera D., Alberto A.,Valente S. 2011. Mechanical interaction between historical brick and repair mortar: experimental and numerical tests. Journal of Physics, 305, 1-10. [3] Bocca P., Grazzini A. 2012. Experimental procedure for the pre-qualification of strengthneing mortars. International Journal of Architectural Heritage, 6 (3): 302-321. [4] Collepardi M. 1990. Degradation and restoration of masonry walls of historical buildings. Materials and Strctures, 23: 81-102. [5] Fassina V., Favaro M., Naccari A. and Pigo M. 2002. Evaluation of compatibility and durability of a hydraulic lime-based plaster applied on brick wall masonry of historical buildings affected by rising damp phenomena. Journal of Cultural Heritage, 3: 45-51. [6] Grazzini, A. 2006. Experimental techniques for the evaluation of the durability of strengthening K. KAPTAN/APJES IV-I (2016) 08-14 works on historical masonry. Masonry International, 19: 113-126. [7] Lourenço, P. B. 1996. Dissertation. Computational strategies for masonry strictures. Delft University of Technology, Netherlands. [8] Rots, J. G. 1997. Structural Masonry: An Experimental/Numerical Basis for Practical Design Rules., Rotterdam, Netherlands. Balkema. [9] Schmidt-Döhl, F. and Rostàsy, F. S. 2000. Abschlussbericht. Ingenieurmodell zur Dauerhaftigkeit des Verbundsystems Naturstein/Mörtelfuge mit Bezug auf die Neuverfugung historischen Mauerwerks. iBMB, TU Braunschweig. [10] Twelmeier, H., Sperbeck, S. T., and Budelmann, H. 2008. Restoration Mortar for Historical Masonry – Durability Prediction by means of numerical and Engineering Models, 14th International Brick and Block Masonry Conference. [11] Van Zijl, G.P.A.G. 2000. Computational Modelling of Masonry Creep and Shrinkage. Meinema BV, Delft, Netherlands. T. OGUNDARE/APJES IV-I (2016) 15-20 Development and Characterization of Parian Bodies Using Feldspar from Two Selected Deposits in Nigeria Toluwalope OGUNDARE1*, Oluwagbenga FATILE2 Olusola AJAYI,3 1,2,3 Department of Glass and Ceramic Technology Federal Polytechnic Ado Ekiti, Ekiti State, Nigeria Geliş Tarihi: 2015-12-10 Kabul Tarihi: 2016-04-01 Abstract Self glazed bodies also called Parian Bodies are chiefly composed of Feldspar which majorly acts as flux in reducing the melting temperature of the particular ceramic body thereby reducing the stress of double firing a ceramic product. In this present research work, Okpella and Ijero-Ekiti feldspar deposits in Nigeria together with other raw materials were utilized to develop parian bodies which are very suitable for making ceramic tiles, dolls and figurines. Chemical analyses were carried out on the raw materials using X-ray fluorescence (XRF) in order to ascertain their suitability for developing parian bodies. Six compositions were made from these two deposits using a standard parian body composition of 70% feldspar. The samples were shaped, dried and sintered between 1050 oC - 1250 oC at the interval of 500C. Flexural strength, fired shrinkage, porosity and water absorption tests were used to characterize the samples. The results showed that samples fired at 12000C and 12500C exhibited technological properties that meet up with ISO standards. The two feldspar deposits were found to be suitable for developing parian bodies. Keywords: Parian, Feldspar, Kaolin, Porcelain, Sintered, Whiteware. 1. Introduction Recently, the development of self glazed porcelain ware is attracting interest from researchers in Ceramic Technology field owing to their technological properties and low cost of production. Porcelain bodies are usually made up of at least three components that play the three fundamental roles for optimum processing, and hence performance of the final products, kaolin or kaolinitic clay for plasticity, feldspar for fluxing and silica as filler for the structure [1]. The thermal, dielectric and mechanical properties of the products can be improved by varying the proportions of the three main ingredients [2]. The main differences between compositions are in the relative amounts and kinds of raw materials used. Most times, it is observed that an increasing amount of feldspar added to porcelain body composition usually results in formation of liquid phase at the eutectic temperature, which increases the degree of vitrification and translucency at lower temperature [3]. As feldspar is replaced by clay, higher temperatures are required for vitrification due to the introduction of a more refractory material, and the firing process becomes more difficult and expensive. However, the forming processes become easier, and the mechanical properties of the resulting body are improved [4]. Porcelain happens to be a class of whiteware, which is distinguished from the other class by its firing temperature, composition and mainly by the lack of open porosities on the fired body, but self glazing effect can be achieved from porcelain bodies with introduction of high quantity of fluxes. This self glazing ceramic ware is known as “Parian” and has been achieved by introducing certain fluxes like feldspar which when fired at high temperatures contributes to the formation of a thin layer of glass fused at the ware surface. This glass film has a very similar composition to the respective ceramic material and prevents the hairline cracks known as crazing. Parian is extremely translucent through a large temperature range and ideally suited for casting to produce figurines, dolls and light forms. This body is chiefly made up of very white feldspar which uses a floatation method of particles distribution and some frothing may occur due to high speed of mixing. Parian was a development of earlier biscuit porcelain, but has higher proportion of feldspar in body composition than the normal porcelain, makers fired it at a lower temperature and the high content of feldspar present in the body will make the fired ware to be more vitrified, thus possessing a colour verging on ivory and having a marble like structure that is smoother than that of biscuit or glazed ware. When parian bodies are made, they come out faint but extremely *Corresponding author: Toluwalope OGUNDARE, Address: Department of Glass and Ceramic Technology Federal Polytechnic Ado Ekiti, Ekiti State, Nigeria, E-mail Address: t_ory2003@yahoo.com, Phone: 2348038058132 Doi: 10.21541/apjes.08877 T. OGUNDARE/APJES IV-I (2016) 15-20 beautiful[5]. Parian is a creamy white, semitranslucent type of porcelain with a granular marble like texture. Parian unique marble- like texture is achieved by using a high percentage of feldspar [6]. Parian when made comes out translucent with a fine granular surface and has a china resembling marble like in texture, Its invention did not come out of thin air however it’s a product of individuals who attempted to find a warmer creamier material more like marble from which they could mould decorative items. A self- glaze clay, so once fired, it has a slight shine/ shimmer to it. It is fine to go straight on to the kiln shelf, however does stick to it is own body if slabs are placed on top of one another during the firing. Self- glaze actions take place when the composition is fired. Parian body is constituted in form of a glass for producing a selfglazed ceramic body which include the step of performing a structure consisting essentially of ceramic powder having a firing temperature with the range from 1150oc – 1250oc consisting essentially 70% of feldspar [6]. Feldspar is a very common mineral found naturally in most primary rock in amount up to 60%. Its function is generally applicable as a flux in clay and glazes many different types of feldspar exist, the main groups are potash and soda. Feldspar can be recognized by its pearly luster and opaque appearance. The crystal has too main cleavages that are nearly at right angles to each other [7]. Basically, the two properties which make feldspar useful for downstream industries are their alkali and alumina content. On those elements we can distinguish three families; feldspathic sand, pegmatite and feldspar. A further distinction can be made between sodium, potassium and mixed feldspars depending on the type of alkali they contain. In glass production feldspar is an important ingredient and an imperative raw material as well because it acts as fluxing agent, reducing the melting temperature of quartz and helping to control the viscosity of glass. The alkali content in feldspar acts as flux thereby lowering the glass batch melting temperature and thus reducing production cost. In ceramics feldspar is the second most important ingredient after clay, feldspar does not have a strict melting point since it melts gradually over a range of temperatures. This greatly facilitates the melting of quartz and clays and through appropriate mixing allows modulation of this important step of ceramic making. Feldspar also improves the strength, toughness and durability of the ceramic body. Feldspar assist the enamel composition assuring the absences of defects and neatness of the end product e.g. enamel frits, ceramic glazes, ceramic tiles glazes, sanitary ware, table ware, electrical porcelain and gift ware. In the flooring sector, feldspar is the main constituent in the body composition. It is used as a flux, lowering the vitrifying temperature of ceramic body during firing and forming a glassy phase. In table ware feldspar gives a good fusibility for a product without defects. In sanitary ware, the use of feldspar within vitreous ceramic bodies facilitates the optimization process [8]. Despite the fact that parian bodies have numerous advantages over non self glazed bodies, insufficient attentions have been devoted to development and characterization of parian bodies. This work is part of current efforts aimed at considering the potentials of numerous feldspar deposits around the world as suitable raw material in making parian bodies for low cost production of ceramic wares. 2. Material and method The raw materials selected for this research work are Feldspar from Okpella in Edo State, Feldspar from Ijero-ekiti in Ekiti State, ball clay, kaolin and flint. All the raw materials were sourced from Nigeria. Chemical analysis was carried out on the raw materials using X-ray fluorescence spectroscopy. The result of the chemical composition is as presented in Table 1. Six different body compositions were formulated by using 70% of Feldspar in each of the samples and varying the percentage of other raw materials as shown in Table 2. 3. Preparation of samples The processed raw materials were weighed according to the composition in table 2 using accurate electronic weighing balance. The weighed raw materials were charged into pot mill containing porcelain grinding media (pebbles) and 35% water of the total charge was added together with sodium tripoliphosphate as deflocculant. The pot mill was allowed to run for two hours before its content was discharged into a dry pan. The pan and its content were dried in electric oven at the temperature of 1100C for two hours. The dried material was then crushed and 7% weight of water was added before passing through a 100 mesh (150μm) to obtain suitable powders for pressing. Each composition were used to produce samples of tiles of size 100mm x 100mm x 10mm by using uniaxial hydraulic pressing machine at the pressure of 50bars. The samples were allowed to dry for 24hours before drying in laboratory oven at 110 0C for two hours after which they were sintered in an electric kiln at different temperature of 10500C, 11000C, 12000C and 12500C C (heating and cooling rates of 50 per minute), and with soaking time of 1 hour at the maximum temperature. T. OGUNDARE/APJES IV-I (2016) 15-20 4. Characterization of the samples The samples produced were characterized using standard methods in order to determine their technical properties. The shrinkage test was carried out on the samples by determining the initial length before firing (Lo) and the length after firing (Lf) using digital venier caliper. The Percentage linear shrinkage was determined using equation (1). % 𝐿𝑖𝑛𝑒𝑎𝑟 𝑆ℎ𝑟𝑖𝑛𝑘𝑎𝑔𝑒 = Table 1: Chemical Composition of Raw Materials (wt. %) Material Oxides (wt. %) SiO2 Al2O3 CaO Na2O Fe2O3 Ijero 66.47 18.20 0.65 2.84 0.33 Feldspar Okpella 65.98 14.62 0.95 3.62 0.76 Feldspar Kaolin 48.21 33.36 0.23 0.13 0.74 Quartz 96.12 1.17 0.11 0.62 0.12 Ball Clay 44.82 37.35 0.07 0.11 1.09 Table 2: Formulation of the Samples (Wt. %) Okpella Feldspar Ijero Feldspar A 70 B 70 C 70 D 70 E 70 F 70 The flexural strength was determined using a universal testing machine (MTS 810.23M), in three-point bending fixture, 70 mm support span and with a crosshead speed of s0.5mm min-1. Porosity and water absorption were determined using boiling method.The specimen were subjected to 1 hour boiling followed by an additional two hour water soaking and then weighed as W sat .The soaked specimen were then suspended from the beam of a balance in a vessel of water in such a way that specimen were completely immersed in the water without touching the side of the vessel. Weights of the suspended specimen were determined as Wsus. The specimen were dried in oven for 24hours and the weight were determined as Wd. The test was carried out on four representative specimens. Percentage porosity and water absorption were calculated using equation 2 and equation 3 respectively. 𝑃𝑜𝑟𝑜𝑠𝑖𝑡𝑦 = Wsat− Wd Wsat− Wsus 𝑊𝑎𝑡𝑒𝑟 𝑎𝑏𝑠𝑜𝑟𝑝𝑡𝑖𝑜𝑛 = x 100 (2) Wsat− Wd Wd x 100 (3) 5. Results and discussion Result of the linear shrinkages after firing at different temperatures is as presented in figure Flint 10 10 10 10 Lo – Lf Lo x 100 (1) K2O 9.61 MgO 0.25 MnO - LOI - 8.87 0.29 - - 0.77 0.20 0.89 0.05 0.07 0.12 - 12.83 3.84 - Kaolin 30 20 Ball Clay 20 30 20 20 1.The linear shrinkage indicates the degree of densification during firing, and it is very important for the dimensional control of the ceramic products. It is observed from figure 1 that linear shrinkage values of the samples were within the range of 2.4%-10.4%. These values are within the safe limits for industrial production of self glazed porcelain products. Generally, it is observed that the linear shrinkage of the samples increased with increase in temperature with samples sintered at 12000C exhibiting the highest linear shrinkage with the exception of sample A and F. It is also observed that sample C exhibited the highest linear shrinkage at all temperatures in comparison with other samples. Figure 2 also shows the results of water absorption tests carried out on the samples. From figure 2, it is observed that the water absorption generally decreased with increase in firing temperature with samples fired at 12500C exhibiting the lowest water absorption values. The decrease in water absorption values at high temperature could be attributed to the formation of more liquid phase that mainly originated from the feldspar. The liquid phase aids sintering which resulted in maximum vitrification at high temperature. This behaviour is similar to that showed by nearly all porcelain bodies [2]. All the samples fired at 12000C and 12500C had water absorption values within the range of 0.1%- T. OGUNDARE/APJES IV-I (2016) 15-20 0.5%.This implies that all the samples fired at 12000C and 12500C meet up with ISO standard which recommended water absorption value of not greater than 0.5% for porcelain tiles. The result is also in agreement with previous work reported by other authors [1, 9]. 12 Linear Shrinkage (%) 10 8 1050 oC 6 1100 oC 1200 oC 4 1250 oC 2 0 A B C D E F Sample Designation Figure 1. Linear shrinkage of samples sintered at different temperature 4 Water Absorption (%) 3,5 3 2,5 1050 oC 2 1100 oC 1,5 1200 oC 1 1250 oC 0,5 0 A B C D E F Sample Designation Figure 2. Water absorption of samples sintered at different temperature T. OGUNDARE/APJES IV-I (2016) 15-20 18 16 14 Porosity (%) 12 10 1050 oC 8 1100 oC 1200 oC 6 1250 oC 4 2 0 A B C D E F Sample Designation Figure 3. Porosity of samples sintered at different temperature 60 Flexural Strength (MPa) 50 40 1050 oC 30 1100 oC 1200 oC 20 1250 oC 10 0 A B C D E F Sample Designation Figure 4. Flexural strength of samples sintered at different temperature The result of porosity tests carried out on the samples is as presented in figure 3. Porosity of the fired samples is associated to other properties such as water absorption and linear shrinkage. The result followed the same trend with that of water absorption rate. From figure 3, it is generally observed that the percentage porosity value of the samples decreased with increase in sintering temperature. Martin-Marquez et al [10] have reported that optimum vitrification range is achieved in porcelain tiles when open porosity reaches a minimum value, tending to be nearly zero, and simultaneously linear shrinkage is maximum. T. OGUNDARE/APJES IV-I (2016) 15-20 The result of bending strength test of the samples sintered between 10500C and 12500C is as presented in figure 4. The result shows that the flexural strength values of the samples are within the range of 20.4Mpa-53.4Mpa. From figure 4, it is generally observed that the flexural strength of the samples increased with increase in firing temperature with sample C exhibiting the highest flexural strength values at all temperature in comparison with other samples. The high flexural strength values exhibited by sample C at all temperature are also consistence with the water absorption values. Boussak et al. [11] have reported that increase in temperature of ceramic compositions containing clays and feldspars result in higher mullite formation, thereby improving mechanical properties. Although, sample C and F contained 70% of Okpella feldspar and Ijero feldspar respectively but chemical analyses of the two feldspar samples showed Okpella feldspar contains more fluxes in comparison with Ijero feldspar. The higher percentage of fluxes in Okpella feldspar contributed to formation of more liquid phase in sample C which enhanced its vitrification and flexural strength in comparison with sample F. According to ISO standard, the minimum flexural strength value recommended for porcelain tiles is 35Mpa. This implies that nearly all the samples fired at 12000C and 12500C meet up with ISO standard with the exemption of sample A and D fired at 12000C. 6. Conclusions The development and characterization of parian body using feldspar from two selected deposits in Nigeria was successfully investigated. The results show that: The percentage linear shrinkage of the samples increased with increase in firing temperature with sample C exhibiting the highest linear shrinkage. All the samples fired at 12000C and 12500C had water absorption values of not greater than 0.5% which is the required standard for porcelain tiles according to ISO. All the samples fired at 12000C and 12500C had flexural strength values of not less than 35Mpa with the exemption of sample A and D. Sample C and F fired at 12000C and 12500C produced the best self glazed body which also meet up with require standards. 7. References [1] Kamseu E., Leonelli C., Boccaccini D.N., Veronesi P., Miselli P., Giancarlo P. and Chinje M.U., Characterisation of Porcelain Compositions Using Two China Clays from Cameroon, Ceramics International, 2007, 33, p. 851-857. [2] Peter W.O., Stefan J. and Joseph K.B., Characterization of Feldspar and Quartz Raw Materials in Uganda for Manufacture of Electrical Porcelains, Journal of Australia Ceramic Society, 2006, 41(1), p. 29-35. [3] Reed, J.S., Principles of Ceramics Processing, New York, Wiley & Sons, 1995. [4]McLaren E. A. and Cao P. T. Ceramics in Dentistry—Part I: Classes of Materials. Inside Dentistry; 2009. [5] Carty, W.M., 2002, Observation on the Glass Phase Composition in Porcelains. Chem. Eng. Sci. Proc., 2002, 23(2), p. 79-94. [6] Holdren, G.R and Berner, R.A, Mechanism of Feldspar Weathering. low Temperature Feldspars in Sedimentary Rocks, American Journal of Science, 1999, p. 279, 435479. [7] Ima (2000). Ceramic Raw Materials. London Pergarmon Press. [8] Nelson,I and Stephen, A (2008): Weathering and Clays minerals. [9] Matthew G.O., and Fatile B.O., Characterization of Vitrified Porcelain Tiles Using Feldspar from Three Selected Deposits in Nigeria. Research Journal of Recent Sciences, 2014, 3(9), p. 67-72. [10] Martín-Márquez J., Rincón M.J. and Romero M., Effect of firing temperature on sintering of porcelain stoneware tiles, Ceramics Internacional, 2008, 34, p. 1867-1873. [11] Boussak H., Chemani H., and Serier A., Characterization of Porcelain Tableware Formulation Containing Bentonite Clay, International Journal of Physical Sciences, 2014, 10(1), p. 38-45. A.O. FATOYE/APJES IV-I (2016) 21-27 Fractional Distribution of Heavy Metals from the Tailings of Itagunmodi Goldmine Site Osun State, Nigeria Abiodun O. FATOYE1,2* and Albert. O ADEBAYO1 Department of Chemistry, Federal University of Technology, Akure, Ondo State, Nigeria. 2 Department of Science Technology, Federal Polytechnic Ado Ekiti, Ekiti State, Nigeria. 1* Geliş Tarihi: 2015-11-19 Kabul Tarihi: 22.03.2016 Abstract The distribution of heavy metals (Cu, Ni, Cd, Cr, Zn and Fe) in goldmine tailing was determined using multi- step sequential extraction. Chemical properties such as pH, conductivity, cation exchange capacity, organic matter, residual humidity and particle size were also analyzed. Similar characteristics distribution patterns were observed for Cd, Cr, Fe and Zn except Cu. The percentage recovery of Ni ranged from 95.25-99.24%. The high Mobility factor and bioaccumulation factor values of soil Ni may be interpreted as symptoms of relatively high liability and biological availability of the metals in soil. Keywords: Tailings, Heavy Metals, Goldmine, Sequential Extraction, Distribution 1. Introduction Mining gives rise to soil erosion and environmental contamination by generating waste during the extraction, beneficiation, and processing of minerals. After closure, mines can still impact the environment by contaminating air, water, soil, and wetland sediments from the scattered tailings, as well as pollution of groundwater by discharged leachate, unless the proper remediation is conducted [1]. Mining activities contribute to heavy metal pollution of the environment [2,3]. Progressive accumulation of heavy metals in soils surrounded by mines, result in increased heavy metal uptake by plants. This is worrisome because of potential health risk to the people leaving in the surrounding areas [4]. Elements like Cd, Cr and Ni, are said to be non biodegradable thus, persist everywhere in the environment and have the ability to be deposited in various body organs which poses a great threat to the human health. Several studies have shown that plants, growing in heavy metal contaminated soils have higher concentrations of heavy metals than those grown in uncontaminated soil [5]. It has been reported that serious health problems may develop as a result of excessive accumulation of heavy metals such as Cd, and Pb in the human body [6]. Despite Zn and Cu being essential elements in the diet, high concentration in plants is of great concern because they are toxic to humans and animals [7]. Pb and Cd metals are believed to be potential carcinogens and are implicated in the ontology of many diseases, especially cardiovascular diseases, kidney, nervous system, blood as well as bone ailments [8]. Heavy metal contamination of agricultural soils and crops surrounding the mining areas is a serious environmental problem in many countries, Nigeria inclusive [9]. It is well known that metals in soil are presented in different chemical forms, which influence their reactivity and hence their mobility and bioavailability. Evaluating metal pollution of soils on the basis of total metal content provides little information on the mobility and bioavailability of heavy metals and thus gives poor guidance for the selection of appropriate remediation strategies for contaminated soil. Recently, pollution of general environment has increasingly gathered a global interest. In this respect, contamination of agricultural soils with heavy metals has always been considered a critical challenge in scientific community [10]. *Corresponding author: Department of Chemistry, Federal University of Technology, Akure, Ondo State, Nigeria, E- mail: abiodun.fatoye@yahoo.com, Phone: +2348036605518 Doi:10.21541/apjes.61578 A.O. FATOYE/APJES IV-I (2016) 21-27 A sequential chemical extraction technique fractionates heavy metals into forms of different solubilities and mobilities, and can therefore furnish potential valuable information for predicting metal availability and metal movement in the soil. and bioavailability of radionuclides [13,14,15]. It is very important to evaluate the mine tailing to actually ascertain the distribution of the heavy metals which is the focus of the research. 2. Material and Method Speciation of metals in soils, sediments and solid wastes is often studied using sequential extraction techniques whereby the target metals are fractionated into several fractions using extractant solutions of increasing strength [11] Several such sequential extraction schemes have been described [11,12]. The technique has been used to study the speciation of heavy metals in soils, street sweepings and urban aquatic sediments, lake sediments, pelagic sediments, semiarid soils, dredged sediment derived surface soils, and solid waste materials [12]. The technique has also been used to study the speciation, mobility Itagunmodi is a small community lies between latitudes 7°30’ and 7°33’ N and between longitudes 4°36’ and 4°39’ E in Atakumosa West Local Government Council southwestern Nigeria (Figure 1). The study area as shown in the map below is a rural community of about 2,400 to 2,600 people that engage predominantly in subsistence farming and cocoa plantation. Itagunmodi is a community with many dilapidated buildings. Figure 1. A map is belonging to Itagunmodi Tailings samples were oven dried at 400C for two weeks. Samples were sieved through a 0.8mm mesh and stored in clean polythene bags for further analysis. The pH was determined by the method [16], particle size analysis by (hydrometer meter), conductivity by (conductivity meter), CEC by ammonium acetate method [17], Organic matter content [18], Residual humidity [19]. Dried and powdered soil sample of 1.2 g was digested with aqua regia (3:1 HCl: HNO3) in 100 ml conical flask on a hotplate and diluted to volume with distilled water. Fe, Cr, Cu, Zn, Cd, and Ni in the digest were determined using 210 VGP (Buck Scientific) atomic absorption spectrophotometer. The detection limit of the atomic absorption spectrophotometer used is 0.01. The sequential extraction of metals was done by the method of Tessier et al,1979 which partitions metals into exchangeable, bound to carbonates, bound to Fe – Mn Oxides, bound to organic matter and residual. 1g of air dried tailing sample were used. A.O. FATOYE/APJES IV-I (2016) 21-27 Table 1. Testing parameters Fraction Extractant F1 1M MgCl2,pH 7 F2 1M NaAC, pH 5 F3 0.04M NH2(OH)Cl/25%CH3COOH F4 0.02M HNO3/30%H2O2, pH2 F5 HClO4/HF Source : Tessier et al, (1979). Shaking Time(hr) 1 5 5 Temp(oC) RM RM 96 Designation Exchangeable Carbonate Fe-Mn oxide 5 85 0rganic/Sulphide Residual Validation of the analytical results was tested by recovery experiments because there was no standard reference material (SRM), which is more preferential or needed to control the accuracy of the method studied, in our laboratory. An important consideration in the reliability of a sequential extraction data is the percentage recovery relative to a single digestion using a mixture of strong mineral acids or generally a mixture of strong acids at the digestion of the residual phase of the sequential extraction protocol employed [20]. Recovery is defined as follows: R= F is the different fractions while SDSA is the single digestion for single acid The distribution of heavy metals in the sample allows us to predict their mobility and bioavailability. The bioavailability factor was expressed as the ratio of the available concentration of a metal in soil to its total concentration. It shows the potentials of a particular metal from the soil matrix to enter the soil solution from which it can be absorbed by plants. Mobility factor was expressed as percentage of the Bioavailability factor X100 3. Results Table 2. Showing the result of physicochemical characteristics of tailing samples pH 1 4.27±0.77 EC (uS/cm) 66.0±6.50 CEC (Meq/100g) 3.16±5.35 OC (%) 0.45±1.13 OM (%) 0.78±1.97 RH (%) 0.30±0.02 Sand (%) 75.0±4.30 Clay (%) 18.0±5.80 Silt (%) 7.0±4.20 Texture 2 4.30±0.77 70.0±6.50 2.30±5.35 0.72±1.13 1.25±1.97 0.22±0.02 78.0±4.30 5.0±5.80 17.0±7.40 Loamy Sand Loamy Sand Sand 3 4.84±0.77 75.0±6.5 3.06±5.35 0.35±1.13 0.60±1.97 0.20±0.02 74.0±4.30 3.0±5.80 23.0±7.40 4 5.61±0.77 79.0±6.50 13.48±5.35 2.49±1.13 4.28±1.97 0.26±0.02 85.0±4.30 10.0±5.80 5.0±7.40 5 5.98±0.77 82.0±6.50 11.70±5.35 2.62±1.13 4.59±1.97 0.25±0.02 78.0±4.30 8.0±5.80 14.0±7.40 Av. 5.0 78.4 6.74 1.33 2.30 0.246 - - - Loamy Sand - Ran 4.27-5.98 66-95 2.30-13.48 0.35-2.62 0.60-4.59 0.20-0.30 - - - - Av: Average, Ran: Range P value=0.834, then P>0.05, there is no significant difference (ANOVA & DUNCAN) EC= Electrical conductivity, CEC= Cation exchange capacity, OC= Organic carbon, OM= Organic matter, RH= Residual humidity Sand A.O. FATOYE/APJES IV-I (2016) 21-27 Table 3. Concentration of Tailing Cu in individual Fractions (mg/kg) Loc F1 F2 F3 F4 F5 Total Cu in original Sample 48.13 30.71 31.80 18.05 65.0 1 0.00a,b 1.00a±0.35 0.00a 6.00a±0.20 36.0b±0.20 43.0 a,b 2 0.00 2.00a±0.20 0.00a 0.00a 28.0b±0.20 30.0 a,b a a a 3 0.00 0.00 0.00 0.00 30.0b±0.10 30.0 4 0.00a,b 1.00a±0.35 0.00a 0.50a±0.10 15.5b±0.20 17.0 5 0.50a,b±0.3 0.00a 0.00a 17.5a±0.20 45.0b±0.25 63.0 Value with different superscript on the same row are significant (p<0.05) %Recovery %BF %MF 99.69 97.68 94.38 94.18 96.77 0.023 0.07 0.00 0.06 0.08 2.3 7.0 0 6.0 0.8 Table 4. Concentration of Tailing Ni in individual Fractions (mg/kg) Loc F1 9.00 a±0.95 8.50 a±2.05 8.00 a±0.95 8.00 a±0.40 7.00 a±0.30 1 2 3 4 5 F2 F3 F4 F5 Total 6.00 a±0.25 7.50 a±1.40 6.50 a±0.45 8.00 a±1.40 0.00a 5.00a±0.10 5.00 a±0.35 0.35 a±0.12 3.00 a±0.65 0.00 a 1.00a±1.20 0.00 a 1.00 a1.0 1.00 a±1.0 17.50 a±0.20 57.5 b±0.1 56.0 b±2.0 68.15b±1.5 35.0 b±0.5 45.0 b±3.2 78.0 77.0 84.0 55.0 55.50 Ni in original Sample 79.91 77.84 84.16 56.15 91.69 % Reco very %BF 98.25 98.92 95.25 97.95 60.53 %MF 0.19 0.21 0.17 0.29 0.13 19 21 17 29 13 Value with different superscript on the same row are significant (p<0.05) Table 5. Concentration of Tailing Zn in individual Fractions (mg/kg) Loc F1 F2 1 2 3 4 5 7.50 a±0.65 2.80 a±1.25 2.00 a±1.65 6.00 a±0.55 1.00 a±0.30 11.50a±3.70 8.00 a±2.25 7.00 a±1.65 7.50 a±1.25 38.0 a±0.80 F3 F4 F5 Total Zn in original Sample %Reco very %BF %MF 12.30a±4.15 9.50 a±1.45 8.50 a±1.50 12.0 a±2.35 39.5 a±0.20 9.00 a±2.25 8.00 a±0.05 9.50 a±2.00 11.5 a±1.30 32.5 a±1.10 28.0a±2.30 26.0a±2.00 27.0 a±1.00 6.00 a±0.50 31.0 a±4.45 68.5 54.5 54.0 43.0 141.5 69.19 55.87 55.30 44.18 142.91 99.00 97.55 97.65 97.33 99.00 0.28 0.19 0.17 0.31 0.28 28 19 17 31 28 Value with different superscript on the same row are significant (p<0.05) Table 6. Concentration of Tailing Cr in individual Fractions (mg/kg) Loc F1 F2 0.00 a 2.25 b±0.25 0.00 a 4.75 b±0.25 a 0.00 6.40 b±0.15 0.00 a 8.05 b±0.25 a 6.15 ±0.65 0.00 b 1 2 3 4 5 F3 F4 F5 0.00 a 0.00 a 0.00 a 1.00 a±0.20 1.85 a±0.20 2.45 b±0.45 4.20 b±0.40 6.20 b±0.60 8.45 b±0.65 8.20 b±0.05 0.40 a±0.04 0.85 a±0.45 0.20 a±0.05 0.00 a 0.15 a±0.50 Cr in \ original Sample Tota l 5.10 9.80 12.8 17.5 16.35 5.58 10.40 13.48 18.55 17.46 %Reco very 91.40 94.23 94.96 94.34 93.64 %BF 0.44 0.48 0.50 0.46 0.38 %MF 44 48 50 46 38 Value with different superscript on the same row are significant (p<0.05) Table 7. Concentration of Tailing Cd in individual Fractions (mg/kg) Loc 1 2 3 4 5 F1 F2 F3 F4 0.00 a 1.10 a±0.50 0.50 a±0.10 0.015a±0.00 0.25 a±0.05 0.25 a±0.05 0.50 a±0.20 0.30 a±0.10 0.25 a±0.00 0.30 a±0.05 0.25a±0.05 0.35 ±0.05 0.15 ±0.10 0.30a±0.00 0.50a±0.15 0.10 a±0.00 0.05 a±0.05 0.85 a±0.05 0.45 a±0.05 0.60 a±0.05 F5 0.10a±0.05 0.00 a 0.40a±0.15 0.35a±0.10 0.20a±0.05 Value with different superscript on the same row are significant (p<0.05) Total 0.85 0.20 2.20 1.37 1.85 Cd in original Sample 1.54 2.17 2.43 1.49 2.00 %Reco very %BF %MF 55.19 92.17 90.95 91.95 92.50 0.29 0.80 0.36 0.19 0.30 29 80 36 19 30 A.O. FATOYE/APJES IV-I (2016) 21-27 Table 8. Concentration of Tailing Fe in individual Fractions (mg/kg) Loc 1 2 3 4 5 F1 10.40a±0.50 5.50 a±0.05 6.25 a±0.10 6.80 a±0.60 4.95 a±0.10 F2 F3 392.15b±0.15 164.90b±0.15 388.95b±0.40 16.25 b±0.15 51.15 b±0.05 446.15d±0.35 435.50d±0.30 441.80d±1.50 477.25d±0.35 484.75d±0.15 F4 345.95b,c±1.75 182.80b,c±0.20 182.20b,c±1.35 276.05b,c±1.55 474.60b,c±1.60 F5 338.25c,d±0.20 425.0 c,d±1.50 472.0 c,d±0.10 476.0 c,d±1.00 486.5 c,d±0.10 Total 1552.90 1213.70 1491.20 1052.35 1401.15 Fe in original Sample 1554.31 1634.98 1506.20 1057.55 1412.15 %Recov ery 100.09 74.24 99.0 99.51 99.28 %BF 0.26 0.14 0.27 0.022 0.040 Value with different superscript on the same row are significant (p<0.05) 4. Discussion Physicochemical properties of the tailings was conducted and the result is presented on Table 2. The pH values of the tailing samples range from 4.27 to 5.98 with a mean value of 5.0 which shows that the tailings are slightly acidic. Residual humidity range from 0.20-0.30% with an average of 0.246. Since residual humidity is proportional to the content of clay and organic matter, tailing has low residual humidity, it indicated less clay but range from loamy sand to sand.. EC range from 66-95 µS/cm with a mean value of 78.4 µS /cm. Furthermore, CEC ranged from 2.30 to 13.48Meq/100g with a mean value of 6.74Meq/100g. The percentage organic carbon (%OC) ranged from 0.35 to 2.62 with a mean value of 6.74. Also, percentage organic matter (%0N) ranged from 0.60-4.59 with a mean value of 2.30. Similar results are reported [1]. The sequential extraction procedures results provided information on the potential mobility and bioavailability of the studied elements in this research. The distribution of heavy metals in the sample allows us to predict their mobility and bioavailability in the tailing. The bioavailability factor was expressed as the ratio of the available concentration of a metal in soil to its total concentration. It shows the potentials of a particular metal from the soil matrix to enter the soil solution from which it can be absorbed by plants. Mobility factor was expressed as percentage of the Bioavailability factor. The speciation of the metals are shown in Table 3-7 for their individual metals. Cu was mainly presented in the residual fraction and organic matter fraction as shown in Table 3. The content of residual fraction ranged from 68%-91% and varied in the range of 5%-27% in organic matter fraction .the level of copper is exchangeable , carbonate fraction and fraction bond to hydrate oxide of iron and manganese was lower. Copper is bound much more tightly by organic matter than Ni and is much less likely to be displaced by the hydroxylamine hydrochloride reagent. Similar results are reported by [21,22]. Copper in the soil is not present in the ionic forms which reduce its toxicity. Ni is mainly found in the fraction bound to residual that it varies from 63%-81% as shown on Table 4. Whereas 0%-27.7% in organic and sulfide fraction. The level of Ni in other fraction is considerably low. Toxicity of nickel is not important because of its low concentration in the mobile and bio available fractions. Heavy metals in residual fraction are not likely to be discharged under normal environment conditions. Ni was mainly associated with residual fractions, so its status in the soil was considered stable [22]. The analysis from Table 5 shows that Zn has been bound to hydrated oxides of iron and manganese fraction and varied in different sample in the range of 15.7%-27.9%. The amount of zinc found in the carbonated fractions varies from 13%-26.9%.Similar results are reported by [24]. Chromium in the tailings sample is mainly found in the organic and sulfide metal fraction with an average of 48% and little lesser in the range of 44%-46% bound to carbonate. Whereas it is smaller in residual ranges from 0%-8.7% as shown on Table 6. Thus, the bioavailability of chromium in the tailing sample is considered insignificant. It is low in exchangeable and iron-manganese oxide metal bound fraction. The analysis of cadmium from Table 7 in different soil samples have been found in organic and sulfide metal fraction range from 2.5%-38.65% and varied in the range of 6.8%-27% in iron manganese oxide metal fraction. The level of cadmium in exchangeable, carbonate bound, and residue fraction was lower absorption by hydrated CaCO3 has been shown to be reversible [23]. %MF 26 14 27 2.2 4.0 A.O. FATOYE/APJES IV-I (2016) 21-27 The results shows that iron occurred more in residual fraction which is varied between 26% and 34.7% in different sample and also 27.5%-45.4% of iron is bound to oxide fraction as presented in Table 8. It is also found in different ranges- carbonate, organic and sulfide metal fraction but minor amount of iron were detected in the exchangeable fractions . Similar results were reported by [25]. The metal concentrations, percentage recovery, mobility and bioavailability factors of all the sequential extraction steps including the residual phase fractions, determined at each extraction step in Tables 3-8, indicate that each metal has a characteristic distribution pattern. Table 4, shows the mobility and bioavailability factors and percentage recovery of Ni for all the sequential extractions steps. The MF of Ni gave average value of 20% while average Bf of Ni was 0.20. The percentage recovery of Ni ranged from 95.25 -99.24%. The high MF and BF values of soil Ni may be interpreted as symptoms of relatively high liability and biological availability of the metals in soil [26,27]. Similar characteristics distribution patterns were observed for Cd, Cr, Fe and Zn except Cu with low Mf of 3.2% and Bf of 0.032(Tables 1 6). The average mobility of Ni, Cu, Cd, Cr, Fe and Zn levels in all the six fractions were in the order: Cr> Cd > Zu > Ni > Fe> Cu. Conclusion The distribution of heavy metals (Cu, Cr, Cd, Ni, Zn, Fe) in the tailings of Itagunmodi Gold mine was assessed. The results obtained in the heavy metal speciation indicated that all metals were mainly associated with the residual and organic& sulfide fraction whereas only small fraction of heavy metals is extracted in exchangeable, carbonate and Fe-Mn oxide fraction. References [1] Ekwue Y.A, Gbadebo A.M, Arowolo T.A, and Adeosun J.K. Assessment of metal contamination in sooil and plants from abandoned secondary, and primary goldmines in osun state, Nigeria. Journal of soil science and environmental mgt 2012, vol3 (11), pp 262-274. [2] Navarro MC, Perez-Sirvent C, Martinez-Sanchez MJ, Vidal J, Tovar PJ, Bech J. Abandoned mine sites as a source of contamination by heavy metals:A case study in a semi-arid zone. J. Geochem. Explor., (2008) 96: 183-193. [3] Singh AN, Zeng DH, Chen FS. Heavy metal concentrations in redeveloping soil of mine spoil under plantations of certain native woody species in drytropical environment, India. J Environ. Sci., (2005) 1: 168-174. [4] Pruvot C, Douay F, Herve F, Waterlot C. Heavy metals in soil, crops and grass as a source of human exposure in the former mining areas. J Soils Sediments, (2006) 6: 215-220. [5] Akan JC, Abdulrahaman FI, Sodipo OA, Lange AG. Physicochemical parameters in soil and vegetable samples from Gongulon Agricultural site, Maiduguri, Borno state, Nigeria. J. Am. Sci., (2010) 6: 12. [6] Zhuang P, McBride MB, Xia H, Li H, Li Z . Heavy metal contamination in soils and food crops around Dabaoshan mine in Guangdong, China: implication for human health. Environ. Geochem. Health, (2008) 31: 707-715. [7] Kabata-Pendias A, Mukherjee AB. Trace elements from soil to human. NewYork: SpringerVerlag (2007). [8] Jarup L . Hazards of heavy metal contamination. Brit. Med. Bull., (2003)68: 167-182. [9] Aslibekian O, Moles R. Environmental risk assessment of metal contaminated soils at silver mines abandoned mine site, co tipperary, Ireland. Environ. Geochem. Health (2003) 25:247-266. (Pub Med) [10] Faruk O, Nazim S, Kara SM . Monitoring of cadmium and micronutrients in spices commonly consumed in Turkey. Res. J. Agric. Biol. Sci. (2006) 2:223-226. [11] Tessier A, Campbell PG, Bisson M . Sequential extraction procedures for the specification of particulate trace metals. Anal. Chem., (1979) 5: 844855.Ma [12] Gibson MJ, Farmer JG . Multi-step sequential chemical extraction of heavy metals from urban soils. Environ. Pollut. Ser. B, (1986)11(2): 117-135. [13] Riise G, Bjornstad HE, Lien HN, Oughton DH, Salbu B. A study on radionuclide association with soil components using a sequential extraction procedure. J. Radioanal. Nuclear Chem., (1990)142(2): 531-538. [14] Salbu, B, Oughton DH, Ratnikov AV, Zhigareva TL, Kruglov SV, Petrov KV,Grebenshakikova NV, Firsakova SK, Astasheva N P, Loshchilov NA, HoveK, Strand P. The mobility of 137Csand 90Sr in agricultural soils inthe Ukraine, Belarus, and Russia, 1991. Health Phys., (1994) 67(5): 518-528. [15] Blanco P, Tome VF, Lozano JC. Sequential extraction for radionuclidefractionation in soil samples: A comparative study. Appl. Radiat. Iso., (2004) 61(2-3): 345-350. A.O. FATOYE/APJES IV-I (2016) 21-27 [16] Carter Mr, ciregorich E G(19940. Soil sampling and method of analysis. P. 99, accessed on www.amazon.com [17] Thomas G W (1982). Exchangeable cations in A. L page et al (ed) methods of soil analysis: part 2. Chemical and microbiological properties. ASA monograph 9 :159-165. [18] Nelson D W sommers L.E(1982). Toltal carbon, organic matter .pp 539-580. In page et al (ed) methods of soil analysis. Part 2 2nd ed . argon monogr.9 ASA and SSSA, Madison, WI. [19] Yobouet Y. A., Adouby K., Trokourey A. and Yao B.Cadmium, Copper, Lead and Zinc speciation in contaminated soils. International Journal of Engineering Science and Technology ,2010,Vol. 2(5), 802-812. [20] Boch, K. Schuster, M., Risse, G., Schwarzer, M. Anal Chem. Acta (2002) 257 – 459. [21] Mclarem R.G and clucas L.M, fractionation of copper, nickel and zinc in metal spiked sewage sludge, journal of environmental quality, vol. 30, no.6.2001 pp 1968-1975. [22] Huang s, Fractional distribution and risk assessment of heavy metals contaminated soil in vicinity of a lead/zinc mine, trans nonferrous met. Soc. China 24(2014) 3324-3331. [23] Sager M. Chemical speciation and environmental mobility of heavy metals in sediments. In Hazardous Metals in the Environment, Stoeppler M, Ed., Elsevier Science Publ., Amsterdam,(1992) pp. 133-174. [24] Kwapulisnski J and Wiechula D., forms of science heavy metals in the bottom sediment of goczalkowice reservoir, chemistry and environmental protection, polythenic university of lubhin, lubhin Poland, 1993. [25] Cotton F A,Wllkinson G(1972c).nadvanced inorganic chemistry; wiley interscience, new York, p. 513. [26] Ma, L.Q and Rao, G.N. Chemical Fractionation of Calcium, Copper, Nickel and Zincin contaminated soils. J. Env. Qual. 1997,26: 259 – 264. [27] Gzyl, J. Lead and Cadmium Contamination of Soil and Vegetables in the Upper Silasia Region of Poland. Sci. Total Env. (1996);119 – 209. [28] Akintola, A.I, Olorunfemi,A.O, Bankole,S.I, Omotoye,S.J and Ajayi, B.O. Petrography and Geochemical evaluation of major and trace elements concentrations in the stream sediments of Itagunmodi and its environs, Southwestern, Nigeria. Journal of Earth Sci & Geochemical Engineering Vol 3 no. 4, 2013, pg 1-24. F.M. ÜNAL/APJES IV-I (2016) 28-37 Hedef Programlama ile Nöbet Çizelgeleme Probleminin Çözümü Fatih Mehmet ÜNAL1 ve Tamer EREN2* Cumhurbaşkanlığı Muhafız Alayı, Tören Tabur Komutanlığı, 06680, Çankaya, Ankara 2* Endüstri Mühendisliği Bölümü, Mühendislik Fakültesi, Kırıkkale Üniversitesi, 71450, Kırıkkale 1 Geliş Tarihi: 2016-01-19 Kabul Tarihi: 2016-04-01 Öz Nöbet çizelgeleme birçok üretim ve hizmet sektöründe kullanılmaktadır. Nöbet çizelgeleme yapılırken birçok kritere dikkat edilmesi ve nöbet tutacak personelin memnuniyetinin aynı anda sağlanmak istenmesi problemi daha karışık hale getirebilmektedir. Bu çalışmada hizmet sektöründeki bir devlet kurumunda nöbet çizelgeleme problemi ele alınmıştır. Ele alınan problemde nöbet tutan personelin istekleri göz önüne alınarak ağırlıklı hedef programlama modeli geliştirilmiştir. Problemin çözümleri GAMS 22.5 paket programı ile gerçekleştirilmiştir. Anahtar Kelimeler: Nöbet Çizelgeleme, Hedef Programlama, Çok Amaçlı Karar Verme. The Solution of Shift Scheduling Problem by Using Goal Programming Fatih Mehmet ÜNAL1 ve Tamer EREN2* Cumhurbaşkanlığı Muhafız Alayı, Tören Tabur Komutanlığı, 06680, Çankaya, Ankara 2* Endüstri Mühendisliği Bölümü, Mühendislik Fakültesi, Kırıkkale Üniversitesi, 71450, Kırıkkale 1 Abstract Shift scheduling is being used in various production and service sectors. It makes it more complicated to consider many aspects while doing shift scheduling and also taking into account the gladness of the shift scheduling. In this study also we are looking at shift scheduling problem of a service sector which belongs to a government association. In this particular problem, goal programming model is developed by considering the wishes of shift scheduling. The solution for the problem is acquired by GAMS 22.5 packet program. Keywords: Shift Scheduling, Goal Programming, Multiple-Objective Decision Making. 1. Giriş Personel çizelgeleme, hastane, havayolu, otel, telekomünikasyon merkezi, çağrı merkezi ve güvenlik sektörü gibi geniş çaplı alanlarda sıkça uygulanmaktadır. Bu tarz hizmet sektörlerinde genellikle yıl boyunca 7 gün 24 saat hizmet esası bulunmaktadır. Bunun için kurumlar personelinden 24 saat esasına göre hizmet istemekte ve personellerini kurumlarının özel ihtiyaçlarına göre çizelgelemektedirler. Çizelgeler günlük, haftalık, aylık gibi birçok şekilde düzenlenmektedir. Personel çizelgeleme, yasal düzenlemelere ve kurumların amaçlarına uygun kısıtların sağlanmasının yanında, personel isteklerini karşılayarak onların memnuniyetlerini arttıracak aynı zamanda hizmet alanların ihtiyaçlarını karşılayacak işgücü oluşturmaktır. planlarını ve çizelgeleri Günümüzdeki personel çizelgeleme problemleri, Edie [1] ve Dantzig [2]’nin 1950’lerde tanımladığı problemlerden çok farklıdır. Personel çizelgeleme problemlerinde çalışanların ihtiyaçlarını karşılamanın önemi artmaktadır. Kurumlar personel çizelgeleme yaparken çalışanların isteklerini dikkate almaktadırlar [3]. Personel çizelgelemede temel amaç; kaynaklardan etkili bir biçimde yararlanmak, dengeli iş yükü dağılımını sağlamak ve mümkün olduğunca bireysel istekleri karşılamaktır. İyi bir çizelge, çalışanların ve hizmet alanların ihtiyaçları ile kurumun görevlerini dengeler, gerekli olan iş yükünü personele adaletli ve uygun bir şekilde planlar. *Corresponding author: Endüstri Mühendisliği Bölümü, Mühendislik Fakültesi, Kırıkkale Üniversitesi, 71450, Kırıkkale, tamereren@gmail.com Doi: 10.21541/apjes.59022 F.M. ÜNAL/APJES IV-I (2016) 28-37 İşgücü, vardiya ve görevlere atanırken çalışan tercihlerine dikkat etmek önemlidir. Çünkü çalışanların tercihlerine dikkat edildiği zaman, personel daha verimli olmakta ve bunun sonucunda hizmet kalitesi artmaktadır. İşletmeler, işgücü yoğunluğunun fazla olduğu hizmet sektörünün artan önemi sonucunda, işgücünün etkin kullanımına daha fazla önem vermeye başlamışlardır [4]. Bu çalışmada, bir devlet kurumunda aylık dönemler halinde nöbet çizelgeleme problemi için ağırlıklı hedef programlama modeli önerilmiştir. Önerilen modelin en önemli özelliği, personele görevi gereği zorunlu olarak tutacağı nöbetlerini seçme imkânı tanımasıdır. Ancak bu hedef aynı zamanda her personele eşit sayıda ve ağırlıkta nöbet planlamak ile bu nöbetlerin mümkün olduğunca aralıklı olmasını sağlamak hedefleri ile birlikte değerlendirilmektedir. Yani personel nöbet tutmak istediği günleri belirtmekte olup, kurum ve diğer çalışanlar zarara uğratılmayacak şekilde, personel mümkün olduğunca kendi belirttiği günlere nöbetçi olarak atanmaktadır. Çalışmanın planı şu şekildedir: İkinci bölümde, personel çizelgeleme üzerine literatürde yapılmış çalışmalar sunulmuştur. Üçüncü bölümde, nöbet çizelgeleme problemlerinin özelliklerinden bahsedilmiştir. Dördüncü bölümde, problemin özellikleri ve çözümü için önerilen hedef programlama modeli üzerinde durulmuştur. Beşinci bölümde örnek uygulama yapılmıştır. Son bölüm olan altıncı bölümde ise yapılan çalışma için genel bir değerlendirme yapılmış, ileride yapılabilecek çalışmalar konusunda önerilerde bulunulmuştur. 2. Literatürde Yapılan Çalışmalar Personel çizelgeleme çok zahmetli ve vakit harcayan bir iştir. Çizelgeleme yapılırken birçok zorunlu kısıt ile personelin ve hizmet alanların memnuniyetinin aynı anda sağlanmak istenmesi problemi daha karışık hale getirebilmektedir. Genellikle el ile hazırlanan ve göz önünde bulundurulacak çok sayıda kısıtın olmasından dolayı hazırlanması uzun ve zahmetli olan çizelgelerin, uygulanan yenilikçi sistematikler sayesinde, çok kısa sürelerde ve daha etkin şekilde oluşturulması sağlanabilmektedir. Gün geçtikçe kurumlar arasında yenilikçi sistematiklerin uygulanması artmaktadır. Ernst vd. [5] yaptıkları çalışmada personel çizelgeleme problemi ile ilgili yapılan çalışmaları, uygulama alanlarına göre sınıflandırarak, bunların modellerini ve çözüm algoritmalarını incelemişlerdir. Bergh vd. [3] yapmış oldukları çalışmada personel çizelgeleme problemleri üzerine yapılan çalışmaları incelemişlerdir. Öncelikle daha önce yapılan inceleme makalelerini inceleyerek sınıflandırma yöntemlerini oluşturmuşlardır. Daha sonra 300'e yakın çalışmayı inceleyerek, oluşturmuş oldukları sınıflandırma yöntemlerine göre bu çalışmaları sınıflandırmışlardır. Ünal [6] hizmet sektöründeki bir devlet kurumunda, yasal ve kurumsal hedeflerin yanında personel isteklerini de ön plana alarak, hedef programlama ve analitik hiyerarşi proses yöntemini birlikte kullanarak personel çizelgeleme problemini incelemiştir. Azaies ve Al-Sharif [7] yaptıkları çalışmada hemşire çizelgeleme problemini ele almışlardır. Burada el ile yapılan çizelgeler yerine 0-1 hedef programlama yaklaşımı ile bir model geliştirmişlerdir. Topaloğlu [8] yaptığı çalışmada hedef programlama ile acil tıp çalışanları çizelgeleme problemini ele almıştır. Chu [9] yaptığı çalışmada hedef programlama ile Hong Kong uluslararası havaalanının bir biriminde personel çizelgeleme problemini incelemiştir. Bağ vd. [10] çalışmalarında hemşire çizelgeleme problemini incelemişlerdir. Problemi çözmek için 0-1 hedef programlama yöntemini kullanmışlardır. Hedeflerinin ağırlıklarının belirlenmesi için ise analitik ağ proses yöntemini kullanmışlardır. HungTsu vd. [11] hedef programlama ile Güney Tayvan'da bir mağazada personel çizelgeleme problemini incelemişlerdir. Bektur ve Hasgül [4] geliştirmiş olduğu hedef programlama modeli ile bir restoranda personel çizelgelenme problemini incelemişlerdir. Louly [12] yaptığı çalışmada bir telekomünikasyon merkezinde vardiya çizelgeleme problemi için hedef programlama modeli geliştirmiştir. Labadi vd. [13] Banka Bilgi Teknolojileri personeli çizelgelemesi için çok amaçlı hedef programlama modeli geliştirmişlerdir. Todovic vd. [14] çalışmalarında hedef programlama modeli ile Bosna Hersek’teki bir polis karakolunda, polis memurların çizelgelenmesini incelemişlerdir. Horn vd. [15] yaptıkları çalışmada Avustralya Kraliyet Donanmasında karakol botları ve mürettebatını çizelgeleme problemini incelemişlerdir. Li vd. [16] birçok zorunlu ve yumuşak kısıttan oluşan hedef programlama ile sezgisel yöntemleri bir arada kullanarak, melez bir yaklaşımla personel çizelgelemesi yapmışlardır. Bard vd. [17] yaptıkları çalışmada Amerika Birleşik Devletleri posta servisinde tur planlama problemini ele almışlar ve problemi tamsayılı doğrusal programlama olarak formüle etmişlerdir. Çetin vd. [18] yaptıkları çalışmada tamsayılı programlama ile uçuş ekibi planlaması üzerinde durmuşlardır. Ekip planlamanın iki aşaması olan, ekip eşleştirme ve ekip atama problemlerini bütünleşik bir yapıda ele almışlardır. Öztürkoğlu ve Çalışan [19] çalışmalarında hemşire çizelgeleme problemi için tam sayılı matematiksel bir model oluşturmuşlardır. Oluşturulan modelde, klasik çizelgeleme F.M. ÜNAL/APJES IV-I (2016) 28-37 modellerinin aksine hemşirelerin işe başlama saatlerine esneklik getirilmiştir. Sungur [20] yaptığı çalışmada bir güzellik salonunda, tur çizelgeleme problemi için karma tamsayılı programlama modeli geliştirmiştir. Fırat ve Hurkens [21] yaptıkları çalışmada karışık tamsayılı programlama ile farklı yetenek isteyen işlere uygun yetenekli teknisyenleri atamışlardır. Güneş vd. [22] yaptıkları çalışmada, genetik algoritma kullanılarak bir yazılım geliştirmiş ve çok zaman alan askeri nöbet çizelgesi hazırlama işlemlerinin; harcanan süre ve sağlanan doğruluk yönü ile en iyilenmesini amaçlamışlardır. Tsai ve Li [23] yapmış oldukları çalışmada genetik algoritma ile hemşire çizelgeleme problemini ele almışlardır. Zolfaghari vd. [24] yapmış oldukları çalışmada perakende sektöründe, personel çizelgeleme problemi için genetik algoritma modeli geliştirmişlerdir. Görüldüğü üzere literatürde personel ve vardiya özelliklerine göre birçok farklı çizelgeleme problemi ele alınmıştır. Oluşturulan çizelgeleme modellerinin, kurumun amaçlarının yanında personel isteklerini ve hizmet alanların memnuniyetini aynı anda sağlaması gerektiği görülmekte olup incelenen problemlerde personel memnuniyeti açısından birçok isteğin sağlandığı görülmektedir. Ancak personelin kendi istek belirttiği günlere nöbetçi olarak atanmasını amaçlayan bir çalışmaya rastlanmamıştır. Aşağıda önerdiğimiz ağırlıklı hedef programlama modeli, personele görevi gereği zorunlu olarak tutacağı nöbetlerini, kurumu ve diğer çalışanları zarara uğratmadan imkânlar dâhilinde seçme imkânı tanımaktadır. Aynı zamanda tüm personele kıdem seviyelerine göre eşit sayıda ve ağırlıkta nöbet planlamayı ve planlanan bu nöbetlerin, personelin yorulmaması için mümkün olduğunca aralıklı olmasını amaçlamaktadır. 3. Nöbet Çizelgeleme Probleminin Özellikleri Bu çalışmada bir devlet kurumunda personel çizelgeleme problemi incelenmiştir. Kurum, görevleri kapsamında çalışanlarından her gün bir kişiyi, normal mesai saatleri dışında nöbetçi olarak görevlendirmektedir. Hâlihazır durumda nöbet çizelgeleri aylık dönemler halinde, görevlendirilen bir personel tarafından el ile hazırlanmaktadır. Görevlendirilen personel bulunulan aylık dönemin son haftası, çizelgeleme yapılacak bir sonraki dönem için personel mazeretlerini toplamakta ve çizelgeleri oluşturmaktadır. Çizelgeler oluşturulurken personel mazeretlerine ve kıdem seviyelerine dikkat edilmekte ve bütün personele hafta sonu ve hafta içi dağılımı adil olmak üzere eşit sayıda nöbet yazılmak istenmektedir. Ancak bütün günlerin aynı ağırlıkta olmaması ve bazı ayların gün olarak fazla olmasından kaynaklı olarak, bu hedefler tam olarak sağlanamayabilmektedir. Böyle durumlarda her personele eşit sayıda ve ağırlıkta nöbet yazıldıktan sonra artık günler öncelikle kıdemsiz personelden kıdemliye doğru yazılmak istenmektedir. Ayrıca personelin gün aşırı veya yakın tarihli nöbetler tutması personeli olumsuz olarak etkilemekte ve nöbetlerin arasının gün olarak uzun olmasının istenmesi problemi daha karmaşık hale getirmektedir. Bu gibi etkenler sonucu el ile hazırlanan çizelgelerin hazırlanma süresi artmakta ve personel isteklerini tam olarak karşılayan çizelgeler hazırlanamamaktadır. Ayrıca çalışanlar birtakım özel sebeplerinden dolayı ayın belirli günlerinde nöbet tutmaya istekli olabiliyorlar. Yani görevi gereği zorunlu olarak tutacağı nöbeti, kurumu ve diğer çalışanları zarara uğratmadan imkânlar dâhilinde kendi belirttiği günlerde tutmak isteyebilmektedirler. Matematiksel bir model ile nöbet çizelgelerinin hazırlanma süresinin azalacağı ve istenilen seviyede çizelgelerin hazırlanacağı öngörülmüştür. Fakat oluşturulacak olan model ile kurumun zorunlu kısıtlarının sağlanmasının yanında tüm personele 1 aylık dönemde eşit sayıda ve ağırlıkta nöbet yazılması ile örtüşen personel istekleri sonucu isteklerin tam olarak sağlanamayacağı değerlendirilmiştir. Bu sebepten dolayı problemin çözümü için hedef programlama yöntemi seçilmiştir. 4. Geliştirilen Hedef Programlama Modeli Zorunlu ve gevşek kısıtlardan oluşan hedef programlama modeli ile personel nöbetlere atanmaktadır. Amaç fonksiyonu zorunlu kısıtların sağlanmasının yanında kıdem seviyelerine göre gevşek kısıtların en küçüklenmesi ve nöbet tutmak için istekli olunan günlere nöbet yazılmasının en büyüklenmesidir. Problemde kullanılan indisler, parametreler, karar değişkenleri ile matematiksel model ve hedefler aşağıda verilmiştir. 4.1. Modelde Kullanılan İndisler i → Personel, i = 1,2, … ,10 ic → Personel, ic ∈ i j → Çizelgeleme yapılan günler, j = 1,2, … ,30 l → Hedefler, l = 1,2,3,41,42 … ,48 4.2. Modelde Kullanılan Parametreler F.M. ÜNAL/APJES IV-I (2016) 28-37 k i → 𝑖. personelin kıdem katsayısı, i = 1,2, … ,10 g j → 𝑗. günün ağırlığı, j = 1,2, … ,30 Wl → 𝑙. hedefin ağırlığı, l = 1,2,3,41,42 … ,48 1, i. personel j. gün mazeret belirtmiş Mij = { 0, d. d. i = 1,2, … ,10 j = 1,2, … ,30 1, i. pers. j. gün nöbet tutmak için istekli Tij = { 0, d. d. i = 1,2, … ,10 j = 1,2, … ,30 4.3. Modelde Kullanılan Karar Değişkenleri +1 vi,ic = 1. değişkeni −1 vi,ic = 1. değişkeni +2 vi,ic = 2. değişkeni −2 vi,ic = 2. değişkeni vi+3 = 3. değişkeni vi−3 = 3. değişkeni +4a vi,j = 4. değişkeni −4a vi,j = 4. değişkeni, a = 1,2, … ,8 x i 1 1 , j 1,2,,30 ,(1) i, j (1) numaralı kısıt her gün sadece bir personelin nöbetçi olarak atanmasını sağlamaktadır. xi , j xi , j 1 1, i 1,2, ,10 , j 1,2,,29 ,(2) (2) numaralı kısıt her personel için ardışık nöbet yazılmasını engellemektedir. Modelde zorunlu olarak ardışık nöbetlere izin verilmemektedir. 10 30 x i 1 j 1 i, j M i , j 0 ,(3) (3) numaralı kısıt ise personelin mazeret belirttiği günlere nöbet yazılmasını engellemektedir. 30 30 j 1 j 1 xi, j xic, j vi,ic1 vi,ic1 0 , i 1,2,,9 , ic (i 1), ,10 ,(4) 1, i. personel j. gün nöbetçi xij = { 0, d. d. i = 1,2, … ,10 j = 1,2, … ,30 Gevşek kısıtın pozitif yönde i = 1,2, … ,10 Gevşek kısıtın negatif yönde i = 1,2, … ,10 Gevşek kısıtın pozitif yönde i = 1,2, … ,10 Gevşek kısıtın negatif yönde i = 1,2, … ,10 Gevşek kısıtın pozitif yönde i = 1,2, … ,10 Gevşek kısıtın negatif yönde i = 1,2, … ,10 Gevşek kısıtın pozitif yönde i = 1,2, … ,10 a = 1,2, … ,8 Gevşek kısıtın negatif yönde i = 1,2, … ,10 10 30 30 j 1 j 1 xi, j g j xic, j g j vi,ic2 vi,ic2 0 sapma ,i 1,2,,9 , ic (i 1), ,10 ,(5) 30 sapma sapma sapma sapma sapma sapma sapma 4.4. Kısıtlar x j 1 Ti , j vi 3 vi3 0 , i 1,2, ,10 ,(6) i, j Kısıt (4), (5), (6) ve (7-14) gevşek kısıtlardır. Kısıt (4) ile her personel ikili olarak karşılaştırılarak aylık toplam nöbet sayıları arasındaki fark bulunmaktadır. Kısıt (5) ile yine her personel ikili olarak karşılaştırılarak aylık toplam nöbet ağırlıkları arasındaki fark bulunmaktadır. Her 2 kısıtta da her personel kıdem olarak kendisinden sonra gelen her personelle karşılaştırılmaktadır. Böylece her personelin gereksiz yere 2 defa kıyaslanması engellenmiştir. Pozitif sapmalar kıdemli personelin +1 fazla nöbet sayısını (vi,ic ) ve fazla nöbet ağırlığını +2 (vi,ic ), negatif sapmalar ise kıdemsiz personelin −1 fazla nöbet sayısını (vi,ic ) ve fazla nöbet ağırlığını −2 (vi,ic ) göstermektedir. Kısıt (6) her personelin nöbet tutmak için istek belirttiği günlere, kaç adet nöbetinin yazıldığını tespit etmek için geliştirilmiştir. Buradaki pozitif sapmalar (vi+3 ) personelin kaç adet nöbetinin personelin istek belirttiği günlere planlandığını göstermektedir. xi , j xi , j 1 xi , j 2 vi, 41j vi, 41j 1 , i 1,2, ,10 , j 1,2, ,28 (7) xi , j xi , j 1 xi , j 2 xi , j 3 vi, 42j vi, 42j 1 , i 1,2, ,10 , j 1,2,,27 (8) xi , j xi , j 1 xi , j 2 xi , j 3 xi , j 4 vi, 43j vi, 43j 1 , i 1,2, ,10 , j 1,2,,26 (9) i 1,2, ,10 , (10) xi , j xi , j 1 xi , j 2 xi , j 3 xi , j 4 xi , j 5 vi, 44j vi, 44j 1 , xi , j xi , j 1 xi , j 2 xi , j 3 xi , j 4 xi , j 5 xi , j 6 vi, 45j vi, 45j 1 , i 1,2, ,10 , j 1,2,,24 46 i, j xi , j xi , j 1 xi , j 2 xi , j 3 xi , j 4 xi , j 5 xi , j 6 xi , j 7 v 46 i, j v 1, (11) F.M. ÜNAL/APJES IV-I (2016) 28-37 i 1,2, ,10 , j 1,2,,23 47 i, j xi , j xi , j 1 xi , j 2 xi , j 3 xi , j 4 xi , j 5 xi , j 6 xi , j 7 xi , j 8 v 47 i, j v (12) 1, i 1,2, ,10 , j 1,2,,22 48 i, j xi , j xi , j 1 xi , j 2 xi , j 3 xi , j 4 xi , j 5 xi , j 6 xi , j 7 xi , j 8 xi , j 9 v 48 i, j v (13) 1, i 1,2, ,10 , j 1,2,,21 Kısıt (7-14) serisi personel nöbetlerinin ne kadar ara ile yazıldığını kontrol etmektedir. Her personelin; kısıt (7) ile her ardışık 3 günde 1’den fazla nöbetinin olup olmadığı, kısıt (8) ile her ardışık 4 günde 1’den fazla nöbetinin olup olmadığı, kısıt (9) ile her ardışık 5 günde 1’den fazla nöbetinin olup olmadığı, kısıt (10) ile her ardışık 6 günde 1’den xi , j 0,1 (14) fazla nöbetinin olup olmadığı, kısıt (11) ile her ardışık 7 günde 1’den fazla nöbetinin olup olmadığı, kısıt (12) ile her ardışık 8 günde 1’den fazla nöbetinin olup olmadığı, kısıt (13) ile her ardışık 9 günde 1’den fazla nöbetinin olup olmadığı, kısıt (14) ile her ardışık 10 günde 1’den fazla nöbetinin olup olmadığı kontrol edilmektedir. i 1,2, ,10 , j 1,2, ,30 (15) (15) numaralı kısıt ile 𝑥𝑖,𝑗 değişkeninin 0 veya 1 değerlerinden herhangi birini alabileceği belirtilmiştir. vi,ic1 , vi,ic1 , vi,ic2 , vi,ic2 , vi3 , vi3 , vi, 41j , vi, 41j , vi, 42j , vi, 42j , vi, 43j , vi, 43j vi, 44j , vi, 44j , vi, 45j , vi, 45j , 46 47 47 48 48 vi, 46 j , vi , j , vi , j , vi , j , vi , j , vi , j 0 (15) Son olarak kısıt (16) ise ilgili karar değişkenlerinin 0’a büyük veya eşit olabileceğini göstermektedir. 4.5 Hedefler Hedef 1: Her personele eşit sayıda nöbet yazılmalıdır. 9 10 9 10 W1 vi,ic1 ki vi,ic1 kic , i 1 ic i 1 i 1 ic i 1 (16) atayabilmektir. Ancak her günün ağırlığının aynı olmaması kaynaklı olarak her personelin toplam nöbet ağırlığını eşitleyememe durumumuz oluşabilmektedir. Böyle durumlarda toplam nöbet ağırlığının öncelikle kıdemli personelden kıdemsize doğru artması için pozitif sapmalar kıdemli personelin kıdem katsayısı, negatif sapmalar ise kıdemsiz personelin kıdem katsayısı ile çarpılmıştır. Denklem (17), kısıt (4)’ün sapmalarını en küçüklemek için geliştirilmiştir. Yani her personelin aylık toplam nöbet sayısı eşitlenmek istenmektedir. Ancak farklı aylarda toplam gün sayısının, toplam personel sayısının tam katları olmayan dönemler ile karşılaşılabilir. Böyle durumlarda her personelin toplam nöbet sayısını eşitlememiz mümkün olmayacaktır. Her personelin toplam nöbet sayısını eşitleyemediğimizde artık nöbetlerin öncelikle kıdemsiz personelden kıdemliye doğru yazılması için pozitif sapmalar kıdemli personelin kıdem katsayısı, negatif sapmalar ise kıdemsiz personelin kıdem katsayısı ile çarpılmıştır. Hedef 2: Her personelin toplam nöbet ağırlığı eşit olmalıdır. Hedef 3: Her personelin nöbetleri, nöbet tutmak için istek belirttiği günlere planlanmalıdır. 9 10 9 10 W2 vi,ic2 k i vi,ic2 k ic , i 1 ic i 1 i 1 ici 1 10 28 , (19) W 41 v i, 41 j ki i 1 j 1 10 27 ,(20) W 42 v i, 42 j ki i 1 j 1 (17) Denklem (18), kısıt (5)’in sapmalarını en küçüklemek için geliştirilmiştir. Burada amaç her personele toplamda eşit ağırlıkta nöbet 10 W3 v i 3 k i , (18) i 1 Denklem (19), kısıt (6)’nın sapmalarını en büyüklenmek için geliştirilmiştir. Buradaki her bir 𝑣𝑖+3 değeri i. personelin kaç tane nöbetinin, personelin nöbet tutmak için istek belirttiği günlere atandığını göstermektedir. Ayrıca 𝑣𝑖+3 değeri ilgili personelin kıdem katsayısıyla çarpılarak öncelikle kıdemli personelden itibaren, kıdemsiz personele doğru istekler karşılanmaktadır. Hedef 4: Her personelin nöbetler arası mesafesi mümkün olduğunca gün olarak uzun olmalıdır. F.M. ÜNAL/APJES IV-I (2016) 28-37 10 26 ,(21) W 43 v i, 43 j ki i 1 j 1 10 25 ,(22) W 44 v i, 44 j ki i 1 j 1 10 24 ,(23) W 45 v i, 45 j ki i 1 j 1 10 23 ,(24) W 46 v i, 46 j ki i 1 j 1 10 22 ,(25) W 47 v i, 47 j ki i 1 j 1 10 21 ,(26) W 48 v i, 48 j ki i 1 j 1 Denklem (20-27) serisi ile (7-14) serisi kısıtların pozitif sapmaları en küçüklenmektedir. Bahse konu nöbet grubunda 10 personel olduğu için 1 aylık çizelgeleme döneminde her personele ortalama 3 nöbet düşmektedir. Bundan dolayı her personelin ortalama 10 günde 1 nöbet tutması hedeflenmiştir. Ancak bu hedefimizi tek kısıt olarak (14) yazmamız halinde, ilgili gevşek kısıtının sağlanamaması durumunda 1 veya 2 gün gibi yakın aralıklı nöbetlerle karşılaşabilme durumumuz oluşabilmektedir. Bunun engellenmesi için bu kısıt kendi içinde kademelendirilmiş ve her personel için nöbetler arasının gün olarak mümkün olduğunca uzun olması sağlanmıştır. 4.6. Amaç Fonksiyonu 9 10 9 10 9 10 9 10 min W1 vi,ic1 k i vi,ic1 k ic W2 vi,ic2 k i vi,ic2 k ic i 1 ici 1 i 1 ici 1 i 1 ici 1 i 1 ici 1 10 28 10 27 42 10 vi , j k i W3 vi3 k i W41 vi, 41 k W j i 42 i 1 i 1 j 1 i 1 j 1 10 26 10 25 44 10 24 45 vi , j k i W43 vi, 43 k W v k W j i 44 i 45 i , j i 1 j 1 i 1 j 1 i 1 j 1 (27) 10 23 10 22 47 10 21 48 W46 vi, 46 j k i W47 vi , j k i W48 vi , j k i i 1 j 1 i 1 j 1 i 1 j 1 Amaç fonksiyonumuz (28), gevşek kısıtlar olan (4), (5) ve (7-14)’ün sapmalarının kıdem seviyelerine göre en küçüklenmesi ile (6) numaralı kısıtların pozitif sapmalarının kıdem seviyelerine göre en büyüklenmesidir. 4.7. Hedef Ağırlıklarının Belirlenmesi Hedeflerin ağırlık değerleri belirlerken öncelikle 4. Hedefin kendi içerisinde kademelendirildiği için bu hedefimizden başlamamız doğru olacaktır. Burada en iyi senaryonun personele 10 günde bir nöbet yazılması olduğundan bu hedefimizin ağırlığı 𝑊48 = 1 olarak kabul edilmiştir. Bu hedefin gerçekleşememesi halinde bir alt kademe olan personele 9 günde 2 nöbet yazılmaya çalışılmalıdır. Bundan dolayı bu hedefimizin değeri bir önceki hedefe göre 2 kat daha fazla bir ağırlığa sahip olduğu düşünülerek 𝑊47 = 2 olarak kabul edilmiştir. Aynı düşünce ile 𝑊46 = 4, 𝑊45 = 8, 𝑊44 = 16, 𝑊43 = 32, 𝑊42 = 64 ve son olarak en istenmeyen durum olan personele 3 günde 2 nöbet yazılması olan hedefin ağırlığı 𝑊41 = 128 olarak belirlenmiştir. Personele planlanacak olan nöbetlerin mümkün olduğunca personelin istek belirttiği günlere planlanmasını sağlayan 3. Hedefin, nöbetler arası mesafelerin mümkün olduğunca gün olarak uzun olmasını sağlayan 4. Hedeften daha büyük bir ağırlığa sahip olması gerektiği görülmektedir. Çünkü personel kendi isteği ile yakın tarihli nöbetler tutmak isteyebilir. Bundan dolayı 3. Hedefin ağırlığı, 4. Hedefin son kademesinin ağırlığından 2 kat daha fazla öneme sahip olduğu düşünülerek 𝑊3 = 256 olarak belirlenmiştir. Diğer iki hedef, personel arasındaki nöbet adaletinin sağlanması için 3. ve 4. Hedefimizden daha büyük bir ağırlığa sahip olmalıdır. Burada en önemli hedef olarak, her personele eşit sayıda nöbet yazılması olan 1. Hedef belirlenmiştir. İkinci olarak da her personelin toplam nöbet ağırlığının eşitlenmesi olan 2. Hedef belirlenmiştir. Sonuç olarak 2. Hedefin ağırlığı 3. Hedefinin ağırlığından 2 kat daha fazla ağırlığa sahip olduğu düşünüldüğünden 𝑊2 = 512 ve aynı ilişki 1. ve 2. Hedeflerin arasında da kurulduğunda 𝑊1 = 1024 olarak belirlenmiştir. Belirlenen ağırlıklar toplu olarak Tablo 1.’de sunulmuştur. F.M. ÜNAL/APJES IV-I (2016) 28-37 Tablo 1. Hedeflerin Ağırlıkları Hedef Ağırlık 𝑊1 1024 𝑊2 512 𝑊3 256 𝑊41 128 𝑊42 64 4.8. Gün Ağırlıklarının Belirlenmesi Hafta içi nöbetleri saat 17:00 ile ertesi sabah saat 08:00 arasını kapsamaktadır. Yani hafta içi tutulan bir nöbet 15 saat sürmektedir. Cuma günü nöbetleri ise saat 17:00’da başlamakta ve ertesi sabah 09:00’da son bulmaktadır. Bu nöbet 16 saatlik bir süreyi kapsamakta ve personelin ilgili hafta sonu tatilini kısıtlamaktadır. Hafta sonu nöbetleri ise sabah 09:00’da başlamakta ve ertesi sabah 09:00’da son bulmakta olup toplam 24 saat sürmektedir. Görüldüğü üzere farklı günlerdeki nöbetlerin ağırlıkları aynı değildir. Bundan dolayı gün ağırlıkları, çalışanların ortak düşünceleri doğrultusunda, günlerin zorluk derecelerine göre birbiriyle karşılaştırılması sonucunda Tablo 2.’de gösterildiği gibi belirlenmiştir. Ayrıca ilgili nöbet döneminde hiçbir günün ağırlığının ay başladıktan sonra değişmediği varsayılmıştır. Tablo 2. Gün Ağırlıkları Günler Ağırlıkları Haftaiçi 3 Cuma 5 Cumartesi 9 Pazar 8 Resmi Bayram 7 Dini Bayram 10 4.9. Personel Kıdem Seviyelerinin Belirlenmesi Kurumda kıdem seviyeleri farklı toplam 10 personel çalışmaktadır. Modelde her personelin kıdem seviyesi, kurumda çalıştığı toplam yıl olarak alınmış olup Tablo 3.’de sunulduğu gibidir. Tablo 3. Personel Kıdem Seviyeleri Personel Kıdem Seviyesi 1, 2 9 3, 4 8 5, 6 7 7, 8 6 9, 10 5 Burada farklı çalışan özelliklerine göre değişik seviyelendirme yapılabilir. Personel sayısının ve kurumda çalışılan toplam yıl sayısının çok artması durumunda, seviyeler 2’şer veya 3’er yıllık dönemler olarak alınabilir. Ancak toplam nöbet dengesinin bozulmaması için çalışanların kıdem 𝑊43 32 𝑊44 16 𝑊45 8 𝑊46 4 𝑊47 2 𝑊48 1 katsayıları arasındaki oranın 2’den küçük olması gerektiği unutulmamalıdır. Ayrıca İlgili nöbet döneminde hiçbir personelin nöbet grubundan ay başladıktan sonra çıkmadığı ve başka hiçbir personelin nöbet grubuna sonradan dâhil olmadığı, ayrıca hiçbir personelin kıdem seviyesinin ilgili dönem içerisinde değişmediği varsayılmıştır. 5. Örnek Uygulama Aylık nöbet çizelgesi hazırlanırken ilk önce her personelden, ilgili nöbet dönemi için mazeretli olduğu günler istenmektedir. Çalışanlar her hangi bir sebepten dolayı mazeretli olarak nöbet tutamayacakları günleri bildirmektedirler. Tablo 4’de sunulmuş olan excel sayfasına personelin mazeret durumları girilmektedir. i. çalışan j. gün mazeret belirtmiş ise Tablo 4’de sunulmuş olan personel mazeret çizelgesinde ilgili değişken 1 değerini almaktadır. Mazeret belirtilmeyen diğer günler ise 0 değerini almaktadır. Çizelgeleme yapılacak örnek ay için personel mazeretleri Tablo 4’de gösterildiği gibidir. İkinci olarak her personelden nöbet tutmak için istekli olduğu günleri belirtmeleri istenmektedir. Tablo 5’te sunulmuş olan excel sayfasına personelin nöbet tutmak için istek belirttiği günler girilmektedir. i. çalışan j. gün nöbet tutmak için istek belirtmiş ise Tablo 5’te sunulmuş olan personel istek çizelgesinde ilgili değişken 1 değerini almaktadır. İstek belirtilmeyen diğer günler ise 0 değerini almaktadır. Personel istekleri Tablo 5’te gösterildiği gibidir. Geliştirilen çok amaçlı ağırlıklı hedef programlama modeli GAMS 22.5 paket programında kodlanarak, ‘Intel® Core™ i3-380M’ işlemcisi, 3 GB belleği olan Windows 8 işletim sistemine sahip bilgisayar ile çözülmüştür. Burada Tablo 4 ve Tablo 5 ile modelin çözümü için gerekli olan 𝑀𝑖𝑗 ve 𝑇𝑖𝑗 parametreleri personelden toplanmış olup; yapılan kodlamalar ile GAMS paket programının bu parametreleri doğrudan tablolar halinde ilgili excel sayfalarından alması sağlanmıştır. Tablo 6. personel mazeret ve isteklerinin de belirtildiği problemimizin optimal çözümüdür. Personel mazeretleri kırmızı, istekleri ise yeşil renk tonları ile gösterilmiştir. i. personelin j. gün nöbetçi olarak atanması halinde ilgili değişken ‘X’ işareti ile belirtilmiştir. F.M. ÜNAL/APJES IV-I (2016) 28-37 Gün/ Personel 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 Tablo 4. Personel Mazeret Çizelgesi 1. Personel 1 1 1 2. Personel 1 1 3. Personel 4. Personel 1 1 1 1 1 7. Personel 1 1 1 1 1 1 1 10. Personel 1 1 1 1 9. Personel 1 1 1 1 1 1 1 1 1 1 1 6. Personel 1 1 1 1 5. Personel 8. Personel 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 Tablo 5. Personel İstek Listesi Gün/ Personel 1. Personel 2. Personel 1 1 1 1 1 1 4. Personel 1 1 6. Personel 1 1 1 1 1 1 1 1 1 1 1 1 1 1 9. Personel 1 1 1 1 8. Personel 1 1 1 1 1 1 7. Personel 1 1 1 10. Personel 1 1 1 3. Personel 5. Personel 1 1 1 1 1 1 1 1 1 1 1 1 Gün/ Personel 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 Tablo 6. Personel Nöbet Listesi 1. Personel X X 2. Personel X 3. Personel X X X X X X X 8. Personel X X X 7. Personel X X X X X 9. Personel 10. Personel X X 6. Personel X X 4. Personel 5. Personel X X X X X X X X F.M. ÜNAL/APJES IV-I (2016) 28-37 Burada modelin zorunlu kısıtların sağlandığı görülmektedir. Her gün bir kişi olacak şekilde ayın bütün günlerine nöbetçi atanmıştır. Hiçbir personele gün aşırı nöbet atanmamıştır. Yine ilgili çizelgede görüldüğü üzere amaç fonksiyonumuzla uyumlu olarak her personelin toplam nöbet sayısı eşitlenmiş olup, toplamda her personel 3 gün nöbetçi olarak atanmıştır. Her günün ağırlığı eşit olmadığı için her personelin nöbet ağırlığı eşitlenememiştir. Fakat toplam nöbet ağırlıkları kıdem seviyelerine göre optimumdur. Her personele sadece 1 hafta sonu nöbeti atandığı görülmekte olup en kıdemli personelin toplam nöbet ağırlığının en az olması model ile uyumludur. Nöbetlerin % 93 oranında personelin istek belirttiği günlere atandığı görülmektedir. İstek dışı atanan 2 nöbet olmakla birlikte 17’nci günde zaten hiçbir personelin istek belirtmeği görülmektedir. Burada istek dışı nöbetlerin öncelikle kıdemsiz personele atanması model ile uyumludur. Nöbetler arası mesafeler, mazeret ve istek günleri ile toplam nöbet sayısını ve ağırlığını eşitleme hedefleriyle birlikte değerlendirildiği için tüm personele 10 günde 1 nöbet yazılamamıştır. Ancak nöbetlerin genel olarak en az 7 gün arayla atandığı, sadece 3 nöbetin 5 gün aralıklı olarak atandığı görülmekte olup sonuçlar çalışanlar için çok iyi seviyede bulunmuştur. 6. Sonuç ve Öneriler Yapılan çalışmada, 24 saat esasına göre hizmet veren bir devlet kurumda personel çizelgeleme problemi incelenmiştir. Literatürde personel çizelgeleme üzerine birçok çalışma yapılmıştır. Ancak yasal ve kurumsal kısıtların sağlanmasının yanında, çalışanlara kıdem durumlarına göre eşit sayıda ve ağırlıkta nöbet yazılması hedeflenirken aynı zamanda ilgili nöbetlerin çalışanların istediği günlere planlamasını hedefleyen bir çalışmaya rastlanmamıştır. Geliştirilen hedef programlama modeli ile personel, kıdem durumlarına göre eşit sayıda ve ağırlıkta olmak üzere, kurumu ve diğer çalışanları zarara uğratmadan, öncelikle kendi belirttiği günlere nöbetçi olarak atanmıştır. Aynı zamanda personeli yormamak için nöbetler arası mesafelerin, mümkün olduğunca fazla olması sağlanmıştır. Model GAMS 22.5 paket programında kodlanarak, farklı kıdem seviyelerinde 10 personelin çalıştığı bir kurumda uygulanmış olup, sonuçlar el ile hazırlanan çizelgelerle karşılaştırılmıştır. Öncelikle daha kısa sürelerde daha etkili çizelgelerin hazırlandığı görülmüştür. Oluşturulan nöbet çizelgesi ise çalışanlar için son derece tatminkâr bulunmuştur. Özellikle personelin görevi gereği zorunlu olarak tutacağı nöbetlerin, imkânlar dâhilinde personelin kendi belirttiği günlere atanması, personel üzerinde çok büyük olumlu etki yaratmıştır. Personelin moral ve motivasyonda artış gözlemlenmiş olup, bu durum kurumsal aidiyet duygusunun gelişmesinde de katkı sağlamıştır. İlerideki çalışmalarda model geliştirilerek iki aşamalı bir yapıya dönüştürülüp, her personelin yıllık toplam nöbet sayısının ve ağırlığının eşitlenmesi amaçlanabilir. Model yeni kısıtlar eklenerek veya mevcut kısıtlar değiştirilerek farklı kurumlar için kullanılabilir. 7. Kaynaklar [1] L.C. Edie, "Traffic delays at toll booths", Operations Research, 2, 107–138 (1954). [2] B.G. Dantzig, "Letter to the Editor—A Comment on Edie’s ‘Traffic Delays at Toll Booths’", Operations Research, 2, 339–341 (1954). [3] J.V. Bergh, J. Beliën, P. Bruecker, E. Demeulemeester, ve L. Boeck, "Personnel scheduling: A literature review", Eur. J. Oper. Res., 226, 367–385 (2013). [4] G. Bektur ve S. Hasgül, "Kıdem seviyelerine göre işgücü çizelgeleme problemi: Hizmet sektöründe bir uygulama", Afyon Kocatepe Üniversitesi İktisadi ve İdari ve Bilimler Fakültesi Dergisi, 15, 385-402 (2013). [5] A.T. Ernst, H. Jiang, M. Krishnamoorthy, ve D. Sier, "Staff scheduling and rostering: a review of applications, methods and models", Eur. J. Oper. Res., 153, 3–27 (2004). [6] Ünal, F.M., Analitik Hiyerarşi Prosesi ve Hedef Programlama ile Nöbet Çizelgeleme Probleminin Çözümü, Yüksek Lisans Tezi, Kırıkkale Üniversitesi Fen Bilimleri Enstitüsü, Kırıkkale, 2015. [7] M.N. Azaiez, ve S.S. Al-Sharif, "A 0–1 goal programming model for nurse scheduling", Computers and Operations Research, 32, 491–507 (2005). [8] S. Topaloğlu, "A multi-objective programming model for scheduling emergency medicine residents", Computers and Industrial Engineering, 51, 375–388 (2006). [9] S.C.K. Chu, "Generating, scheduling and rostering of shift crew-duties: Applications at the Hong Kong International Airport", Eur. J. Oper. Res., 177, 1764–1778 (2007). [10] N. Bağ, N.M. Özdemir ve T. Eren, "0-1 Hedef Programlama ve ANP Yöntemi ile Hemşire Çizelgeleme Problemi Çözümü", International Journal of Engineering Research and Development, 4, 2-6 (2012). [11] L. Hung-Tso, C. Yen-Ting, C. Tsung-Yu ve L. Yi-Chun, "Crew rostering with multiple goals: an empirical study", Computers and Industrial Engineering, 63, 483–493 (2012). F.M. ÜNAL/APJES IV-I (2016) 28-37 [12] M.A. Louly, "A goal programming model for staff scheduling at a telecommunications center", Journal of Mathematical Modelling and Algorithms in Operations Research, 12, 167-178 (2013). [13] M. Labidi, M. Mrad, A. Gharbi, ve M.A. Louly, "Scheduling IT Staff at a Bank: A Mathematical Programming Approach", The Scientific World Journal, Article ID 768374 (2014). [14] D. Todovic, D. Makajic-Nikolic, M. Kostic-Stankovic ve M. Martic, "Police officer scheduling using goal programming", Policing: An International Journal of Police Strategies and Management, 38, 295-313 (2015). [15] M. Horn, H. Jiang ve P. Kilby, "Scheduling patrol boats and crews for the Royal Australian Navy", Journal of the Operational Research Society, 58, 1284-1293 (2007). [16] J. Li, E.K. Burke, T. Curtois, S. Petrovic ve Q. Rong, "The falling tide algorithm: a new multi-objective approach for complex workforce scheduling", Omega, 40, 283–293 (2012). [17] J.F. Bard, C. Binici ve A.H. DeSilva, "Staff scheduling at the United States Postal Service", Comput. Oper. Res., 30, 745–771 (2003). [18] E.İ. Çetin, A. Kuruüzüm ve S. Irmak, "Ekip Çizelgeleme Probleminin Küme Bölme Modeli ile Çözümü", Havacılık ve Uzay Teknolojileri Dergisi, 3, 47-54 (2008). [19] Y. Öztürkoğlu ve F. Çalışkan, "Hemşire Çizelgelemesinde Esnek Vardiya Planlaması ve Hastane Uygulaması", Dokuz Eylül Üniversitesi Sosyal Bilimler Enstitüsü Dergisi, 16, 115-133 (2014). [20] B. Sungur, "Bir güzellik salonunun tur çizelgeleme problemi için karma tamsayılı hedef programlama modelinin geliştirilmesi", İstanbul Üniversitesi İşletme Fakültesi Dergisi, 37, 49-64 (2008). [21] M. Fırat ve C.A.J. Hurkens, "An improved MIP-based approach for a multi-skill workforce scheduling problem", Journal of Scheduling, 15, 363–380 (2011). [22] A. Güneş, A. Kahvecioğlu ve H. Tuncel, "Askeri nöbet çizelgelerinin genetik algoritma kullanılarak en iyilenmesi", Elektrik-ElektronikBilgisayar Mühendisliği 11'inci Ulusal Kongresi ve Fuarı, İstanbul, 470-474 (2005). [23] C.C. Tsai ve S.H.A. Li, "A two-stage modeling with genetic algorithms for the nurse scheduling problem", Expert Systems with Applications, 36, 9506–9512, (2009). [24] S. Zolfaghari, V. Quan, A. El-Bouri ve M. Khashayardoust, "Application of a genetic algorithm to staff scheduling in retail sector", International Journal of Industrial and Systems Engineering, 5, 20–47 (2010). M. SÖNMEZ/APJES IV-I (2016) 38-47 Yeni Bir QPSK Modülatör Mimarisinin Tasarımı Mehmet Sönmez ve 2Ayhan Akbal Fırat Üniversitesi Mühendislik Fakültesi Elektrik-Elektronik Mühendisliği Bölümü 1 1,2 Geliş Tarihi: 2016-01-23 Kabul Tarihi: 2016-03-22 Öz Bu çalışmada kablosuz haberleşme sistemlerinde yaygın olarak kullanılan QPSK (Quadrate Phase Shift Keying: Dördün Faz Kaydırmalı Anahtarlama) modülasyon tekniği için yeni bir mimari önerilmiştir. QPSK modülasyon tekniği, OFDM (Orthogonal Frequency Division Multiplexing: Dikgen Frekans Bölmeli Çoğullama) ve CDMA (Code Division Multiplexing Access: Kod Bölmeli Çoklu Erişim) gibi çoklu kullanıcılara hizmet veren sistemlerde yaygın olarak kullanılmaktadır. Dolayısıyla vericilerin kaynak kullanımının düşük olması oldukça önemlidir. Yeni mimari sayesinde QPSK tekniğinin RAM bit kullanım miktarı %87.5 oranında düşürülmüştür. Ancak sistemin karmaşıklığı arttığı için TLE (Total Logic Element: Toplam Lojik Eleman) ve kaydedici kullanımı da artmıştır. Ayrıca önerilen yöntemin pratik uygulaması da düşük maliyetli Altera DE-0 Nano bord ile gerçekleştirilmiştir. Anahtar Kelimeler: QPSK, FPGA, RAM Bit Kullanımı. Design of a Novel Architecture for QPSK Modulator Mehmet Sönmez and 2Ayhan Akbal Fırat University, Faculty of Engineering, Department of Electrical and Electronics Engineering 1 1,2 Abstract In this study, a new architecture is proposed for QPSK (Quadrate Phase Shift Keying) modulation technique which is widely used in wireless communication systems. QPSK modulation technique is generally utilized in systems serve multiple user such as OFDM (Orthogonal Frequency Division Multiplexing) and CDMA (Code Division Multiplexing Access). Therefore, it is very important that transmitters have low resource utilization. Thanks to the new architecture, RAM bit utilization of QPSK technique is decreased by up to 87.5%. However, the use of TLE (Total Logic Element) and register rise since complexity of the system increases. Besides, practical application of the proposed technique is realized by using low-cost Altera DE-0 Nano Board. Keyword: QPSK, FPGA, RAM Bit Utilization 1. Giriş Günümüzde haberleşme sistemlerinin donanımsal tasarımı oldukça ilgili gören bir konudur. Bunun nedeni daha yüksek hızlarda veri paylaşımı gerçekleştiriliyorken güç tüketiminin düşürülmek istenmesinden dolayıdır. En fazla kullanılan devre yapıları mikrodalga, VLSI (Very-Large-Scale Integration: Çok Geniş Ölçekli Tümleşim) tabanlı ve FPGA tabanlı devrelerdir [1, 2, 3]. Bu devre yapıları kullanılarak birçok modülasyon tekniğinin uygulaması literatürde mevcuttur [4, 5]. Bu modülasyon tekniklerinin arasında kablosuz haberleşme sistemlerinde kullanılan modülasyon tekniklerinden birisi de QPSK modülasyon tekniğidir. QPSK modülasyon tekniğinin FPGA tabanlı uygulaması birçok çalışmada gerçekleştirilmiştir [610 ]. Yapılan çalışmalarda veri iletim oranını artırmak için mikrodalga tabanlı devre yapıları kullanılmıştır. Mikrodalga tabanlı devre yapılarında sinyal üreteci mikrodalga devre kullanılarak gerçekleştirilmiş olup anahtarlama işlevi FPGA ile sağlanmıştır [6]. Bazı *Corresponding author: Address: Fırat Üniversitesi Mühendislik Fakültesi Elektrik-Elektronik Mühendisliği Bölümü, E-mail address : 1msonmeztr@gmail.com Doi:10.21541/apjes.41021 M. SÖNMEZ/APJES IV-I (2016) 38-47 uygulamalarda QPSK modülatörün geleneksel yapısı kullanılmıştır. Yani modülatör şemalarında paralel I ve Q kanal yapılarıyla çarpım blokları kullanılmıştır [7, 8]. Bu çalışmalarda yol uzunluğu mux tabanlı mimarilere oranla daha fazladır ve bu mimarilerde senkronizasyon problemi en büyük sorundur. Bazı çalışmalarda NCO (Numerically Controlled Oscillator: Sayısal Kontrollü Osilatör) yapıları ile sinyal üretimi gerçekleştirilmiştir [9]. NCO yapıları kullanılarak gerçekleştirilen tasarımlarda kaynak kullanım miktarları oldukça yüksektir. Dolayısıyla kaynak kullanımı açısından bu mimariler diğer mimarilere göre daha fazla dezavantaja sahiptirler. Mux tabanlı olarak kullanılan mimarilerde ise QPSK modülatör için dört adet ROM bloğuna ihtiyaç duyulmuştur. QPSK modülasyon tekniğinin demodülasyonu için taşıyıcı senkronizasyonunun başarılmasına yönelik birçok çalışma gerçekleştirilmiştir [11-14]. Bu çalışmalarda önerilen tasarımlar için BER (Bit Error Rate: Bit Hata Oranı) düşürülmesi oldukça önem arz etmektedir. Yapılan çalışmaların benzetim programları yardımıyla BER değişimi izlendikten sonra FPGA üzerinde uygulamaları gerçekleştirilmiştir. Yapılan bir diğer çalışmada adres üreteç blokları kullanılarak BPSK (Binary Phase Shift Keying: İkili Faz Kaydırmalı Anahtarlama), QPSK, 16-QAM (Quadrate Amplitude Modulation, Dördün Genlik Modlasyonu) ve 64 QAM modülasyon teknikleri için mimari tasarımları gerçekleştirilmiştir. Ayrıca önerilen mimarinin temeli mux tabanlı yapılar ile oluşturulmuştur [15]. Bu çalışmada QPSK modülasyonu için yeni bir mimari yapısı önerilerek RAM bit kullanımının düşürülmesi amaçlanmıştır. Bu amaç için I ve Q kanalı sinyallerinin üretimi BPSK olarak düşünülmüş ve ikinci bir mux sayesinde I veya Q kanalı sinyallerinden birisinin seçimi I veya Q kanalı bitlerinin durumları kullanılarak gerçekleştirilmiştir. Sonuç olarak QPSK sinyali başarılı bir şekilde üretilmiş olup uygulanabilir olduğu deneysel sonuçlarla gösterilmiştir. 2. QPSK Modülasyon Tekniği İkili modülasyon teknikleri ile üretilen sinyaller sadece bir bitin değişimine göre durum değiştirmektedirler. Ancak bant genişliliği verimliliğini artırmak için bir bit periyodunun katları süresince belirli sayıdaki bitlerden oluşan bir sembolün gönderilmesini sağlayan modülasyon teknikleri de mevcuttur. QPSK modülasyon tekniği de bu modülasyon tekniklerinden birisidir. QPSK modülasyon tekniğinde bir sembol iki farklı bitten oluşmakta olup sembolün iletilmesi için geçen süre bir bit periyodunun iki katı kadardır. Böylece bit periyodu artırılırken sinyalin bant genişliği de düşürülmüş olur. QPSK modülasyonlu sinyali üretebilmek için Denklem 2.10 kullanılmaktadır [16]. QPSK ( t ) A cos( 2f ct i ) 2i 1 i 4 i 1,2,3,4 (1) (2) Denklem 2.10 ve Denklem 2.11’den görüldüğü gibi QPSK sinyali dört farklı sinyalden oluşmaktadır. Bu sinyallerin frekans ve genlikleri ayı olup fazları farklıdır. Denklem 2.11’den her bir sinyal arasındaki faz farkının 90 derece olduğu görülmektedir. Dolayısıyla θi değerleri sırasıyla, 45, 135, 225 ve 315 derece değerlerindedir. Denklem 2.10 kullanılarak QPSK için daha açık bir ifade Denklem 2.12’de verilmiştir. QPSK ( t ) A cos( i ) cos( 2f ct ) A sin( i ) sin( 2f ct ) (3) QPSK (t ) A 2 I (t ) cos(2f c t ) A 2 Q(t ) sin( 2f c t ) (4) Tablo 2.1’de QPSK modülasyonlu sinyalin faz değişimine göre Denklem 2.12’nin alacağı ifade verilmiştir. Tablo-1. Her sembol için taşıyıcı sinyalin seçimi Sembol QPSK Sinyal Tipi 10 0.707 * A * (cos( w t ) sin( w t )) c c c 00 0.707 * Ac * (cos( wct ) sin( wct )) 01 0.707 * Ac * (cos( wct ) sin( wct )) 11 0.707 * Ac * (cos( wct ) sin( wct )) Tablo 1’den faydalanılarak oluşturulan QPSK sinyal uzayı Şekil 2.24’de verilmiştir. Şekilden görüldüğü gibi birbirlerine göre bir bitlik fark bulunan sembollerin arasında 90 derece fark bulunmaktadır. Ayrıca 10-01 ve 00-11 gibi semboller arasında iki bitlik değişim varsa QPSK sinyalin tam bu noktalardaki geçişlerinde 180 derecelik faz farkı oluşmaktadır. 3. Önerilen QPSK Modülatör Mimarisi QPSK modülasyon tekniği için RAM bit kullanımını düşürmek için bu bölümde yeni bir mimari verilmiştir. Ancak kontrol sinyalleri ile birlikte veri M. SÖNMEZ/APJES IV-I (2016) 38-47 bitlerinin işlemsel bloklardan geçirilmesi ve kullanılan mux sayısındaki artış karmaşıklığın arttığının bir göstergesidir. Daha önceden de açıklandığı gibi QPSK modülasyonlu bir sinyal için dört durum mevcuttur. Önerilen QPSK modülatör şemasında 10-01 sembolleri için ve 00-11 sembolleri için iki ayrı ROM bloğu kullanılmıştır. Dolayısıyla iki ayrı sayaç ve iki ayrı ROM kullanılarak 01-10 ve 1100 sembollerini temsil etmek için QPSK modülasyonlu sinyal üretilebilmektedir. Şekil 1. I1, I2, Q1 ve Q2 bitlerinin dört durum için değişimi Şekil 1’den görüldüğü gibi I1 ve Q1 bitleri, sinyalin alternans değiştirmesi sırasında değerini değiştirmektedir. Ancak I2 ve Q2 bitleri sinyalin genlik değerinin artması ve azalması durumuna göre değişmektedirler. Dolayısıyla I2 ve Q2 bitleri önerilen çalışmada sayacı kontrol eden sinyaller olarak kullanılmıştır. Şekil 1’de ayrıca I2 ve Q2 bitleri, başlangıç durumunda ‘0’ bit değerini almalarına rağmen oluşturdukları sinyallerin örneklerinin genlik değerleri I2 biti için azalırken Q2 biti için artmıştır. Yani sinyaller birbirlerine göre zıt bir davranış göstermişlerdir. Bunu sağlamak için Q2 biti ile I2 bitleri ile dolaylı olarak kontrol edilen ROM bloklarına kaydedilen örneklerin sırası değiştirilmiştir. Yani i kanalı sinyalini oluşturan ROM bloğunun sıfırıncı adresine kaydedilmiş veri ile q kanalı sinyalini oluşturan rom bloğunun son adresine kaydedilmiş veriler eşittir. Dolayısıyla iki sinyal birbirine göre ters bir şekilde oluşturulurlar. Bu mimari ile ilgili bir algoritma aşağıda verilmiştir. Algoritma-1 Giriş: d, I1, I2, Q1, Q2 Değişkenler: i,q,j,s,s1,s2,max,son Çıkış: QPSK Basla: b:=b+1; else QPSK<= i00 end if; end if; if d(i)=’1’ and d(q)=0 then if Q1=’1’ if b<son then if I2=’0’ then s1:=s1+1; else s1:=s1-1; end if; if Q2=’0’ then s2:=s2+1; else s2:=s2-1; end if; s:=s+1; if s=max+1 then s:=0; if i=j then i:=1; q:=0; else i:=i+2; q=q+2; end if; if d(i)=’1’ and d(q)=1 then if I1=’1’ then QPSK<=i11; then QPSK<=i01; else QPSK<= i10 end if; end if; if d(i)=’0’ and d(q)=’1’ then if Q1=’1’ then QPSK<=i10; else QPSK<= i01 end if; end if; if d(i)=’0’ and d(q)=’0’ then if I1=’1’ then QPSK<=i00; else QPSK<= i11 end if; end if; if b=>son then programı sonlandır; end if; GİT Basla M. SÖNMEZ/APJES IV-I (2016) 38-47 Algoritma-1’de verilenlere göre maksimum iletilebilecek bilgi biti sayısına erişildiğinde programdan çıkılacaktır. Bilgi biti sayısı b ile ifade edilirken maksimum gönderilebilecek bilgi biti sayısı da son değişkeni ile gösterilmektedir. Ayrıca bir sembolü oluşturan maksimum bit sayısı da j ile ifade edilmektedir. Eğer i değişkeni (sembolün I kanalından iletilecek bilgi bitinin saklanmasına yardımcı olan değişken) j değişkenine eşit olursa diğer farklı bir örneğin iletileceği bildirilmektedir. BAŞLA Giriş d, I1, I2, Q1, Q2 Değişkenler i, q, s, s1, s2,max,S(t) I2==0 H H E E s1:=s1+1 s=s+1 Q2==0 E s=max+1 s1:=s1-1 s2:=s2+1 s2:=s2-1 s=0 H E i==j H q=0;i=1 q=q+2;i=i+2 b:=b+1 H b==son SON d(i)=1;d(q)=1 I1==1 E S(t)=C11 H H d(i)=1;d(q)=0 d(i)=0;d(q)=1 d(i)=0;d(q)=0 Q1==1 I1==1 H H E E S(t)=C00 S(t)=C10 Q1==1 E S(t)=C01 QPSK=S(t) Şekil 2. Algoritma-1 için oluşturulan akış diyagramı Algoritma-1’de I2 ve Q2 sinyalleri sayaçları kontrol etmek için kullanılmıştır. Algoritmadan görüldüğü gibi her bir durum iç içe iki if döngüsü oluşturmaktadır. İf döngüsü ile birlikte verilen and işlemi bizim algoritmamızda xor bloğu olarak işlem görecektir. Yani iki bitin aynı veya farklı olma durumu diye tanımlanmıştır. Bu algoritmaya ait doğruluk tablosu Tablo 2’de verilmiştir. Tablo 2. Önerilen QPSK için doğruluk tablosu i q I1 Q1 QPSK 0 0 0 X i11 0 0 1 X i00 0 1 X 0 i01 0 1 X 1 i10 1 0 X 0 i10 1 0 X 1 i01 1 1 0 X i00 1 1 1 X i11 M. SÖNMEZ/APJES IV-I (2016) 38-47 Doğruluk tablosundan da görüldüğü gibi i ve q bitlerinin aynı olması durumunda çıkışın I1 biti ile belirlenmesinden dolayı Q1 önemsiz bit olarak tanımlanmıştır. Ancak i ve q bitlerinin farklı olduğu durumlarda ise I1 önemsiz bit olarak görülmektedir. Dolayısıyla verilen algoritmada ilk denetlenmesi gereken özellik i ve q bitlerinin aynı veya farklı olma durumlarıdır. Daha sonra da i biti ile I1 bitinin ve q biti ile de Q1 bitlerinin aynı ve farklı olma durumları ele alınmalıdır. Eğer i ve q bitleri aynıysa ve I1 biti ile de i biti aynıysa çıkış i11 olur. Ancak i ve q bitlerinin durumu değiştirilmeden I1 bitinin durumu değiştirilirse çıkış i00 olur. Aynı şekilde eğer i ve q bitlerinin durumu farklıysa ve Q1 biti ile q bitinin durumu aynıysa çıkış i10 oluyorken Q1 biti ile q bitinin durumu farklıysa çıkış i01 olacaktır. Bu durumda üç adet xor kapısı ve art arda bağlı iki adet mux gereklidir. Birinci mux kapısı i ile I1 ve q ile Q1 blokları için sinyalin üretiminde kullanılıyorken bir sonraki mux kapısı da i ve q bitlerinin aynı veya farklı olma durumlarında çıkışa vereceği sinyale karar vermek için kullanılır. Şekil 3’de önerilen QPSK mimarisi verilmektedir. Şekil 3’de görüldüğü gibi son mux bloğu I kanalı ve Q kanalı sinyallerinden birisini seçmek için kullanılmıştır. Eğer i ve q bitleri aynıysa alt koldan gelen sinyal (Q kanalı sinyali) QPSK sinyalini oluşturuyorken (00 ve 11 durumu) bu bitlerin farklı olması durumunda da üst koldan gelen sinyal (I kanalı sinyali) QPSK sinyalini oluşturur (10 ve 01 durumu). Ayrıca her iki kanal için de farklı ROM kullanıldığı görülmektedir. -1 Çarpıcı Sayaç rom mux K I2 Bilgi Bitleri I1 S/P i q mux QPSK Q1 Q2 K Sayaç rom mux Çarpıcı -1 Şekil 3. Önerilen QPSK modülatör şema Şekilde Q2 ve I2 bitleri I kanalı ve Q kanalı sayaçlarının kontrolünü sağlamaktadır. Önerilen yapının FPGA ortamındaki tasarımı Şekil 4’de görüldüğü gibidir. Şekil 4’de görüldüğü gibi I kanalı ve Q kanalı sinyalleri ayrı ayrı üretilmiştir. Önce sinyallerin üretimi sağlanarak ardından sinyal seçimi yapılmaktadır. Aslında ilk aşamadaki I kanalı ve Q kanalı mux blokları hem üretimi sağlamakta olup hem de I kanalı ve Q kanalı bitlerinin durumuna göre I ve Q kanalı sinyallerini oluşturmaktadır. Yani her bir kol birer BPSK modülatör gibi davranmaktadır. Şekil 4 için elde edilen benzetim sonuçları Şekil 5’de verilmiştir. Şekil 4. Önerilen QPSK modülatör şemasının FPGA ortamındaki tasarımı M. SÖNMEZ/APJES IV-I (2016) 38-47 Şekil 5’de verilen değişimde I kanal sinyali ve Q kanal sinyalleri sırasıyla üst kol ve alt kolda verilen mux bloklarının çıkışlarıdır. Ayrıca I kanal rom ve Q kanal rom sinyalleri ise I kanalı ve Q kanalı için kullanılan ROM bloklarının çıkışlarıdır. Bu ROM bloklarının çıkışları paralel olarak hem bir kaydediciden hem de bir tersleyiciden geçirilmektedir. Tersleyici ve kaydedici çıkışları I1 ve Q1 bitleri sayesinde belirli bir süre aktif ve pasif edilerek önce I kanalı ve Q kanalı sinyalleri oluşturulmuştur. Oluşan bu sinyallerden sonra gelen bilgi bitleri Q kanalında görünen sinyali veya I kanalındaki sinyali aktif ederek daha düşük kaynak kullanımına sahip yeni bir mimarinin oluşturulmasını sağlamıştır. I kanalında ve Q kanalında kullanılan ROM bloklarına kaydedilen örneklerin zamanda sürekli halleri Şekil 6 (a) ve (b)’de görüldüğü gibidir. Şekil 5. Önerilen QPSK modülatör şemasının benzetim sonuçları m B.N 4. B.N 2 3. . du d r du uru um ru m m d 2. 1. du ru m 5. d m um uru ur m d ru 3. du 4. ru m uru 1. du 5. d (a) (b) Şekil 6 (a) Q kanalı ve (b) I kanalı için kullanılan rom bloğuna kaydedilen sinyal uru m M. SÖNMEZ/APJES IV-I (2016) 38-47 Şekil 6 (a) ve (b)’ de görülen sinyallerin örnekleri sırasıyla Q ve I kanalı ROM bloklarına kaydedilmiştir. BN (Başlangıç Noktasından) başlayan sinyal bir periyodu tamamlayabilmek için yine başlangıç noktasında bitmiştir. Şekillerden de görüldüğü gibi 5 durum mevcuttur. Bu durumların kontrolü sayacı kontrol eden I2 ve Q2 bitleri ile sağlanmaktadır. Örneğin 1. durum için bu bitler lojik ‘0’ seviyesindedir. 2.durumda, 3.durumda, 4.durumda ve 5.durumda sırasıyla 1-0-1-0 lojik seviyelerini almaktadır. Karmaşıklığı azaltmak için I ve Q kanalı içerisine kaydedilen örneklerin sırası değiştirilmiştir. I kanalı için sinüs fonksiyonunun çeyrek periyodu kullanılırken Q kanalı için de kosinüs fonksiyonunun çeyrek periyodu kullanılmıştır. iki mimarinin maksimum çalışma frekansları birbirlerine yakın çıkmıştır. Önerilen mimari için elde edilen deneysel sonuçları aşağıda verildiği gibidir. (a) 4. Önerilen QPSK Modülatör için Derleme ve Deneysel Sonuçlar Bu bölümde önerilen QPSK modülatör için deneysel ve derleme sonuçları verilmektedir. Derleme sonuçları Tablo-3’de görülmektedir. Dört ROM bloğu kullanılarak oluşturulan mux tabanlı bir QPSK modülatör ile bu çalışmada önerilen düşük RAM bit kullanımlı modülatörün kaynak kullanım oranları ve maksimum çalışma frekansları verilmiştir. Tablo 3. Önerilen QPSK ve mux tabanlı QPSK için derleme sonuçları Mimari Önerilen QPSK Mux QPSK TLE 181 76 Kaydedici 96 54 RAM bit 320 2560 F0 80 80 173.27 MHz 180.52 MHz Maksimum Frekansı Çalışma tabanlı Tablo 3’de verilen sonuçlara göre bir periyotta kullanılan örnek sayısı (F0=80) eşit olmak şartıyla RAM bit kullanımı önerilen yeni QPSK modülatör yardımıyla düşürülmüştür. Önerdiğimiz mimarinin RAM bit kullanımını düşürmek için diğer lojik elemanlardan faydalandığı da Toplam lojik eleman ve kaydedici sayısındaki artıştan görülmektedir. Ayrıca (b) Şekil 7 (a) I kanalı not_operator bloğu çıkışı (b) I kanalı mux bloğunun seçici biti ile mux bloğu çıkışının değişimi Şekil 7 (a) ve (b), I kanalı sinyallerini göstermektedir. Şekil 7 (a)’dan görüldüğü gibi öncelikle I kanalı not_operator bloğu çıkışından sinyaller üretilmektedir. Bu sinyallerin y eksenine göre simetrik olması genliklerinin mutlak değerinin birbirlerine eşit olduğunu göstermektedir. Yani sinyallerden birisi sinüsün pozitif alternansını gösteriyorken diğer sinyal negatif alternansını göstermektedir. Bu durum Şekil 8 (a)’da da aynıdır. Yani QPSK modülasyonlu sinyali üretebilmek için önerilen algoritmada ilk önce bir sinüs sinyalinin pozitif ve negatif alternanslarının üretimi amaçlanmıştır. Şekil 7 (a)’da görülen mavi veya kırmızı sinyallerin sadece yarım periyotluk örnek değerleri rom bloğuna kaydedilmiştir. Şekil 7 (b)’de ve Şekil 8 (b)’de ise I ve Q kanalı sinyallerinin seçimi için kullanılan mux bloklarının seçici pinlerinin girişi ve mux bloklarının çıkış sinyallerinin zamana göre değişimi verilmiştir. M. SÖNMEZ/APJES IV-I (2016) 38-47 (a) (b) Şekil 8. (a) Q kanalı not_operator bloğu ve (b) Mux bloğunun seçici biti ile Mux bloğu çıkışının değişimi Her iki şekilde de görüldüğü gibi mux bloklarının seçici pin girişleri lojik seviye olarak ‘1’ olduğunda Şekil 7 (a) ve Şekil 8 (a)’da görülen kırmızı renkli sinyaller iletilmektedir. Diğer bir anlamda pozitif alternanslı sinyal mux bloğu çıkışında aktif edilmektedir. Seçici pin girişindeki bit lojik ‘0’ seviyesinde olduğunda ise negatif alternanslı sinyaller iletilmektedir. I kanalı ve Q kanalı iki ayrı BPSK modülatör olarak düşünülürse bu noktaya kadar bütün işlemlerin benzer olduğu görülmektedir. Ama gerçekleştirilen QPSK modülatör mimarisinin karmaşıklığını artıran nokta Şekil 7 (b) ve Şekil 8 (b)’de görülen yarım periyodun yaklaşık olarak 1/5’i kadarlık bir süre boyunca lojik ‘1’ ve lojik ‘0’ bilgi sinyallerinin üretimi ile pozitif ve negatif alternanslı sinyallerin aktif edilmesini sağlamaktır. Bu nokta QPSK modülatörün uygulanması için en büyük senkronizasyon problemini de oluşturmaktadır. Bu nokta bu çalışmada küçük periyot sinyali olarak adlandırılmıştır. Şekil 7 (b) ve Şekil 8 (b), incelenirse küçük periyot sinyali negatif ve pozitif değerde olabilir. Küçük periyot sinyallerinin üretimi için Şekil 7 (a) ve Şekil 8 (a)’da görülen pozitif veya negatif alternanslı sinüs sinyallerinin sadece belirli bir süre aktif edilmesi gerekmektedir. Şekillerden de görüldüğü gibi eğer faz geçişi gerçekleştirilecekse küçük periyot sinyalini aktif eden mux seçici pini sadece o an için değerini değiştirmekte olup, kendinden önce ve kendinden sonra gelen seçici bitlerden farklı olmaktadır. Her bir periyodun üretimi Şekil 5’de detaylı bir şekilde verilmiştir. Küçük periyot sinyalinin üretimi ise 5. duruma denk gelmektedir. Şekil 7 (a) ve Şekil 8 (a)’da verilen mavi renkli sinyaller bilgi bitleri ile I1 ve Q1 sinyallerinin xor işleminden geçtikten sonra mux bloğunun seçici pinine uygulanması ile elde edilmektedirler. Şekil 1’de I1 ve Q1 bitlerinin değişimi detaylı olarak açıklanmaktadır. Şekil 1 incelenirse küçük sinyal periyodunun üretimini sağlayan küçük periyotlu sinyalin gerekliliği görülecektir. Şekil 9’da I kanalından ve Q kanalından oluşturulan pozitif veya negatif sinüs sinyallerinin birbirlerine göre faz durumları verilmiştir. Şekilden de görüldüğü gibi iki sinyal arasında küçük sinyalin periyodu kadarlık bir gecikme mevcuttur. Bunun nedeni küçük sinyal üretimi I kanalı için pozitif veya negatif sinüs sinyallerinin başlangıç (1’e veya -1’e doğru yükselme anı) anında yapılıyorken Q kanalı için ise düşüş (0’a doğru) anında yapılmaktadır. Dolayısıyla sinyalin kontrollü bir şekilde üretilebilmesi için kullanılan kontrol bitleri nedeniyle bu faz farkı oluşmuştur. Şekil 9. I kanalı (kırmızı sinyal) ve Q kanalı (mavi sinyal) not_operator bloklarının çıkışları M. SÖNMEZ/APJES IV-I (2016) 38-47 (a) (b) Şekil 10. (a) I ve (b) Q kanalı mux bloğu çıkışı (kırmızı renkli sinyaller) ile QPSK (mavi renkli sinyal) sinyalin değişimi Şekil 10’da ise kırmızı renkli sinyaller I ve Q kanalında kullanılan mux bloğunun çıkışını gösterirken kırmızı renkli sinyal I kanalında kullanılan mux bloğunun çıkışını göstermektedir. Şekilden görüldüğü gibi QPSK müdülasyonlu sinyal I ve Q kanal bitlerinin durumuna göre ya I kanalındaki sinyali ya da Q kanalındaki sinyali aktif ederek QPSK sinyalini üretmektedirler. (a) (b) Şekil 11 (a)-QPSK sinyali. (b)-I kanalı biti ve Q kanalı biti sinyalleri Şekil 11’de görüldüğü gibi veri bitlerine göre elde edilen QPSK sinyal başarılı bir şekilde elde edilmiştir. Şekil 11’de veri bitlerinin başlangıcı osiloskobun soldan üçüncü karesinden başlanırsa ilk bitler “11” olarak görülecektir. QPSK sinyal de tam olarak soldan ikinci kare içerisinde değişim göstermiştir ve “11” bilgi sinyalleri için başlangıç noktası bu noktadır. Bütün durumlar (dört durum: 11, 01, 10, 00) için QPSK sinyalin oluşumu gösterilmiştir. sinyal) de ‘0’ olduğu görülmektedir. Dolayısıyla bitler farklı olduğundan dolayı I kanalı sinyali son aşamadaki mux bloğu çıkışında aktif edilerek QPSK sinyali oluşturulur (Şekil 10 (a) ve Şekil (b)’de görülen mavi renkli sinyaller). Bu bilgi bitlerinden bir sonraki bitler birbiriyle aynı olduğundan dolayı (“00”) algoritmanın son katmanındaki mux bloğu Q kanalından iletilen sinyali aktif etmiştir. Önerilen QPSK modülatör mimarisi incelenirse I kanalı ve Q kanalı mux bloklarının çıkışları sonucu verecek son bir mux bloğunun girişine uygulanmaktadır. Son aşamadaki mux bloğunun çıkışı ise bilgi sembolünü oluşturan bitlerin aynı ve farklı olmasına göre I kanalı ve Q kanalı sinyallerinden birisini çıkışa aktaracaktır. Dolayısıyla Şekil 10, ve Şekil 11 (b)’den görüldüğü gibi bilgi sembolünü oluşturan bitler farklıysa Şekil 10 (a)’da görülen kırmızı renkli sinyal, aynıysa Şekil 10 (b)’de görülen kırmızı renkli sinyal iletilecektir. Dolayısıyla Şekil 11 (b)’de verilen bilgi bitleri incelenirse ilk durumda I kanalından iletilen bitin (mavi renkli sinyal) ‘1’, Q kanalından iletilen bitin (kırmızı renkli [1] M. Nozaki, S. Yoshizawa, and Hiroshi Tanimoto, "VLSI Design of an Interference Canceller for QPSK OFDM-IDMA Systems", IEEE Asia Pacific Conference on Circuits and Systems (APCCAS), Ishigaki, Japonya, 715-718, (2014). [2] H. M. Thien, N. Sugii, and K. Ishibashi, "A 1.36 μW 312–315MHz synchronized-OOK receiver for wireless sensor networks using 65nm SOTB CMOS technology", Journal of Solid-State Electronics, vol. 117, no. 1, pp. 161-169, Mar. 2016. [3] P. Bhagawat, R. Dash, and C. Gwan, "Architecture for reconfigurable MIMO detector and its FPGA implementation", 15th IEEE International Kaynaklar M. SÖNMEZ/APJES IV-I (2016) 38-47 Conference on Electronics, Circuits and Systems, Denver, ABD, 61-64, (2008). [4] L., Jri, Y. Chen, and Y. Huang, "A low-power low-cost fully-integrated 60-GHz transceiver system with OOK modulation and on-board antenna assembly." IEEE Journal of Solid-State Circuits, vol. 45, no. 2, pp. 264-275, Feb. 2010. [5] D. Denis, and A. P. Chandrakasan. "An energyefficient OOK transceiver for wireless sensor networks." IEEE Journal of Solid-State Circuits, vol. 42, no. 5, pp. 1003-1011, May. 2007. [6] Z. He, W. Wu, J. Chen, Y. Li, D. Stackenas, and H. Zirath, "An FPGA-based 5 Gbit/s D-QPSK modem for E-band point-to-point radios", 41st European Microwave Conference (EuMC), Manchester, İnglitere, 690-692, (2011). [7] S.O. Popescu, A.S.Gontean, and D.Ianchis, "QPSK Modulator on FPGA", IEEE 9th International Symposium on Intelligent Systems and Informatics, Subotica, Sırbistan, 359-364, (2011). [8] W. Song, and Q. Yao, "Design and Implement of QPSK Modem Based on FPGA", IEEE International Conference on Computer Science and Information Technology, Chengdu, Çin, 599-601, (2010). [9] T. Kazaz, M. Kulin, and M. Hadzialic, "Design and Implementation of SDR Based QPSK Modulator on FPGA", MIPRO 2013, Opatija, Hırvatistan, 513518, (2013). [10] X. Xiquan, "Implementation of PSK Digital Demodulator with Variable Rate Based on FPGA", Open Automation and Control Systems Journal, vol. 7, no. 2, pp.1280-1286, Sep. 2015. [11] Y. Linn, "A Self-Normalizing Symbol Synchronization Lock Detector for QPSK and BPSK", IEEE Transactions on Wireless Communıcations, vol. 5, no. 2, pp. 347-353, Feb. 2006. [12] J. Bao, C.Y. Lu, P. D. Graca, S. Zeng, and T. Poon, "A New Timing Recovery Method For DTV Receivers", IEEE Transactions on Consumer Electronics, vol. 44, no. 4, pp. 1243-1249, Nov. 1998. [13] A. Tolmachev, I. Tselniker, M. Meltsin, I. Sigron, D. Dahan, A. Shalom, and M. Nazarathy, "Multiplier-Free Phase Recovery With Polar-domain Multisymbol-Delay-Detector", Journal of Lightwave Technology, vol. 31, no. 23, pp. 3638-3650, Dec. 2013. [14] P. P. Schönwälder, J. B. Dalmau, F. X. Geniz, F. A. López, and R. G. Mas "A Superregenerative QPSK Receiver", IEEE Transactions on Circuıts and Systems—I: Regular Papers, Vol. 61, No. 1, pp. 258265, Jan. 2014 [15] N. K. Venkatachalam, L. Gopalakrishnan, M. Sellathurai, "Low complexity and area efficient reconfigurable multimode interleaver address generator for multistandard radios", IET Computers & Digital Techniques, vol. 10, no. 2, pp. 59-68, Mar. 2016. [16] S. Ertürk, "Sayısal Haberleşme", Birsen Yayınevi, İstanbul, 2005.