Publication: Sıvı seviye denetimi için bulanık denetleyici parametrelerinin karınca koloni ve genetik algoritma ile optimizasyonu
Abstract
SIVI SEVİYE DENETİMİ İÇİN BULANIK DENETLEYİCİ PARAMETRELERİNİN KARINCA KOLONİ VE GENETİK ALGORİTMA İLE OPTİMİZASYONU Endüstride kullanılan kontrol sistemlerinin parametre optimizasyonunda Genetik Algoritma (GA) ile PID veya bulanık denetleyici yaygın olarak birlikte kullanılmaktadır. Bu çalışmada ise, kontrol alanında henüz yeni olan Karınca Koloni Optimizasyonu (KKO) algoritması ile bulanık denetleyici birlikte kullanılmıştır. Bu tez, Marmara Üniversitesi Elektronik-Bilgisayar Eğitimi Bölümü Sayısal Kontrol Sistemleri Laboratuarında bulunan Gunt RT 512 sıvı seviye sisteminin kontrolü üzerinedir. Bulanık denetleyicinin aksiyon ve sınır değerleri GA ve KKO ile optimize edilerek ayrı ayrı bulunmuş ve her iki algoritmanın performansı simülasyon ve gerçek zamanlı deneyler ile karşılaştırılmıştır. Kontrol edilecek prosesin dinamik modeli yapay sinir ağı (YSA) kullanılarak elde edilmiştir. Dinamik modelinin oluşturulması sırasında, sisteme belli giriş değerleri verilmiş ve bu değerlere karşılık sistem çıkışları elde edilerek giriş-çıkış veri seti oluşturulmuştur. Bu veri seti kullanılarak proses için farklı katman ve nöron sayılı modeller oluşturulmuştur. Bu modeller farklı YSA performans fonksiyonlarına göre değerlendirilmiş ve hata değeri en küçük model, prosesin dinamik modeli olarak seçilmiştir. Seçilen model kullanılarak Bulanık denetleyici çıkış değişkeni aksiyon değerleri ve üyelik fonksiyonu sınır değerleri KKO algoritması ve GA ile optimize edilmiştir. KKO algoritması ve GA teknikleri ile elde edilen sonuçlar, yükselme zamanı, oturma zamanı kriterlerine ve ortalama karesel hataya (Root Mean Square - RMS) göre karşılaştırılmış, her iki optimizasyon algoritmasının performanslarının birbirlerine çok yakın olduğu gözlenmiştir.
PARAMETER OPTIMIZATION OF FUZZY CONTROLLER FOR WATER LEVEL CONTROLLER USING GENETIC AND ANT COLONY OPTIMIZATION ALGORITHMS For the systems used in industry, genetic algorithm is usually used with either PID or fuzzy controllers together. But in this study, ant colony optimization algorithm, which is new for control systems, is used with fuzzy controllers. This thesis describes a real time control algorithm, using genetic algorithm (GA) and ant colony optimization (ACO) algorithm for optimizing Fuzzy controller parameters which are action values and membership function boundary values developed for Gunt RT 512 Water Level Control Trainer System in the Digital Control Systems Laboratory of Technical Education Faculty at Marmara University. Optimized parameters of both algorithms are then compared with real time and simulation applications. The dynamic model of the process to be controlled was obtained using Artificial Neural Network (ANN). To build the model, the system was run with diverse and random input values. Its output values were saved to relate them with the inputs. Using the input-output data set, models with various number of layers and neurons were created for the process. The final model was evaluated based on the performance functions. The model with minimum error was chosen as the dynamic model of the process. Using the ANN model, the parameters of Fuzzy controller were optimized with ACO and GA algorithms. The performances of these three techniques were compared with each other using the criteria of overshoot, rise time, settling time and root mean square (RMS) error of the trajectory. It was observed that the performances of GA and ACO were quite similar.
PARAMETER OPTIMIZATION OF FUZZY CONTROLLER FOR WATER LEVEL CONTROLLER USING GENETIC AND ANT COLONY OPTIMIZATION ALGORITHMS For the systems used in industry, genetic algorithm is usually used with either PID or fuzzy controllers together. But in this study, ant colony optimization algorithm, which is new for control systems, is used with fuzzy controllers. This thesis describes a real time control algorithm, using genetic algorithm (GA) and ant colony optimization (ACO) algorithm for optimizing Fuzzy controller parameters which are action values and membership function boundary values developed for Gunt RT 512 Water Level Control Trainer System in the Digital Control Systems Laboratory of Technical Education Faculty at Marmara University. Optimized parameters of both algorithms are then compared with real time and simulation applications. The dynamic model of the process to be controlled was obtained using Artificial Neural Network (ANN). To build the model, the system was run with diverse and random input values. Its output values were saved to relate them with the inputs. Using the input-output data set, models with various number of layers and neurons were created for the process. The final model was evaluated based on the performance functions. The model with minimum error was chosen as the dynamic model of the process. Using the ANN model, the parameters of Fuzzy controller were optimized with ACO and GA algorithms. The performances of these three techniques were compared with each other using the criteria of overshoot, rise time, settling time and root mean square (RMS) error of the trajectory. It was observed that the performances of GA and ACO were quite similar.