Geri Dön

Bulanık kontrolör karar tablosunun genetik algoritmalarla oluşturulması

Genetic algoritthmes for finding fuzzy controller decision table

  1. Tez No: 46432
  2. Yazar: OSMAN KAAN EROL
  3. Danışmanlar: DOÇ.DR. İBRAHİM EKSİN
  4. Tez Türü: Yüksek Lisans
  5. Konular: Bilgisayar Mühendisliği Bilimleri-Bilgisayar ve Kontrol, Computer Engineering and Computer Science and Control
  6. Anahtar Kelimeler: Belirtilmemiş.
  7. Yıl: 1995
  8. Dil: Türkçe
  9. Üniversite: İstanbul Teknik Üniversitesi
  10. Enstitü: Fen Bilimleri Enstitüsü
  11. Ana Bilim Dalı: Belirtilmemiş.
  12. Bilim Dalı: Belirtilmemiş.
  13. Sayfa Sayısı: 44

Özet

ÖZET Bu yüksek lisans tezinde, sayısal optimizasyon yöntemlerinden biri olan genetik algoritmaların yeni bir uygulama alanı olarak bulanık kontrolör karar tablosunun oluşturulmasında kullanımı incelenmiştir. Çalışmada, değişik sistemler ve değişik bulanık kontrolör yapılarında yöntemin etkinliği sınanmış, bulunan kontrolörler, katsayıları röle ve Zeigler-Nichols yöntem leri ile ayrı ayrı bulunmuş klasik PID kontrolörüyle karşılaştırılmalardır. Sonuçlar çalışmanın son bölümünde ayrıntılı olarak verilmiştir.

Özet (Çeviri)

SUMMARY GENETIC ALGORITHMS FOR FINDING FUZZY CONTROLLER DECISION TABLE Using the rules of the nature for optimization has been proposed by Holland to the scientific research area. Since then, this new approach of optimization, named genetic algorithms, has been the subject of many resear ches. The increasing attention of scientists is due to several advantages: 1.) Requiring only few simple operations, it is easy to simulate in computer environment. 2.) It is a global research algorithm. There is no risk of being located on any local optima. 3.) Its efficiency has been proven in the nature, since it has been working for millions of year. The genetic algorithm operators are reproduction, crossover, and mutation. By using these three operators, from iteration to iteration, the improvement on an objec tive function, named fitness function in genetic ter minology can be observed. In this thesis, this algorithm has been used for finding out the decision table of a fuzzy controller. Before being involved in the details, it is useful to explain briefly the genetic operators, and the algorithm itself. The parameters are firstly coded in binary numbers to be handled by genetic algorithms. These parameters are written one after another to form a chromosome structure. Reproduction is used to give a chance to a chromosome to survive in succeeding generations according to its accomplishment. The better the chromosome accomplish the needs, or fits, the more chances to survive, or vice versa. VICrossover exchanges the binary coded data between two chromosomes. Two randomly generated numbers representing the bit positions show where the crossover will be applied. This transfer of data forms the base of the improvement. Mutation introduce pure random search in genetic algorithm. A randomly selected bit is made“1”if it is“0”,or vice versa. Although it seems to have a destruc tive role on preformed chromosomes, it is useful to get rid of local optimas. The calculated chromosomes form again the source operand or the initial population: This is an iteration. It has been continued to iterate until a stopping criteria has been accomplished. Finally the binary coded parameters are decoded to form the original parameters. Fuzzy controllers are attracting more and more attention due to its simplicity of operation. There is no need of any mathematical model of the system to be control led. A skillfull operator describes main points to be known, which form so called“heuristic rules”and then, by using these rules, f uzzif icitaion and def uzzif ication are carried out, as a result, a decision table is found out. The decision table is in the form of a matrix of m X n, where m and n are the number of the quantization levels for the error and the error change respectively. This table is analogous to the fuzzy PD controller. The integral action can be implemented by many ways. The 6 different types of controller are examined in details in Chapter 2. In this study, since the advantages or disad vantages of the mentioned types are of lesser priority only the first and the sixth types are used. The block diagrams belonging these controllers are shown in Figure 1. VIIudl u (a) FUZZY ¥0 ui Fuzzy K P0J (c) "d., u (b) 1j FUZZ* PD ^^ *- razz* pi (d) (e) .2 ?3 ft* ?5 £d Figure 2 - Structure of a chromosome which holds the elements of the fuzzy decision table VIIIIn the beginning, these chromozomes can be randomly formed. But this procedure may cause problems in real-time applications, such as violent changes in control signal, hazardous operation of the system, etc. Therefore, it is advisable to use an approximative model of the system and make the first hundreds of iterations on this model. It is also possible the involve the heuristic rules and to start from preformed decision tables. u Fuzzy PID Controller Plant y *-r* Figure 3 - The block diagram of the system The improvement can be made by using several different objective functions. Depending on the function used, convergence can be obtained in a faster or slower manner. The easiest way to create an objective function is to use the sum of squares of the error. The better the step response, the lesser the sum, but the reciprocal is not always true. Using the sum of the squares of the error may lead to an oscillating response which is not desirable in real-time systems. In this thesis, objective functions based on standart rise time, overshot, settling time and steady-state error are used. The mixed structure of the objective function leads to a better success in control. Simulations have been carried out to prove the efficiency of this new approach and the results have been shown in the section 4. It is also given a comparison with a classical PID controller whose coefficients have been found by the Zeigler-Nichols relay and step response methods separately. IXThe results of this study can be summerized as follows: If the system operates within a specified range, then nonlinear quantization should be prefered. If the error may take any values with an equal probability within that specified range, then the linear quantization should be prefered. If the static gain of the system is known, then the first type of fuzzy controller can be used. If it is not known, or if the load or parameters vary in time, then any other type of fuzzy controller including integral action must be used. The decision table elements may have negative values. This decreases the overshoots but it may slow down the convergence. Since in practical systems, the control signal can generally take positive values, the negative numbers will then be considered as zeros, which may cause problems. A good decision table will be obtained if one runs the method for different reference values. This will increase the effective number of decision table elements used, therefore, each element will take part in the optimization. As it can be observed from computer simulations, the results are encouraging. This thesis may be a stepping point for further researches. X

Benzer Tezler

  1. Tez No

    39286

    Yapay sinir ağlarında öğrenme algoritmalarının analizi

    Analysis of learning algorithms in neural networks

    SEVİNÇ BAKLAVACI

    Yüksek Lisans

    Türkçe

    Türkçe

    1994

    Bilgisayar Mühendisliği Bilimleri-Bilgisayar ve Kontrolİstanbul Teknik Üniversitesi

    DOÇ.DR. LEYLA GÖREN

  2. Tez No

    450942

    An internal model control based tuning method for single input fuzzy PID controller

    Tek girişli bulanık PID kontrolörler için iç model kontrol tabanlı ayarlama yöntemi

    ARDA VAR

    Yüksek Lisans

    İngilizce

    İngilizce

    2017

    Bilgisayar Mühendisliği Bilimleri-Bilgisayar ve Kontrolİstanbul Teknik Üniversitesi

    Kontrol ve Otomasyon Mühendisliği Ana Bilim Dalı

    YRD. DOÇ. DR. TUFAN KUMBASAR

  3. Tez No

    84353

    Fırçasız D.C. motorların bulanık mantık yöntemi kullanılarak ADSP-2101 (digital signal processor) ile kontrolü

    Fuzzy control of brushless D.C. machine by using ADSP-2101 DSP

    KADİR KORKMAZ

    Doktora

    Türkçe

    Türkçe

    1999

    Bilgisayar Mühendisliği Bilimleri-Bilgisayar ve KontrolMarmara Üniversitesi

    Elektronik ve Bilgisayar Eğitimi Ana Bilim Dalı

    PROF. DR. BURHANETTİN CAN

  4. Tez No

    439362

    Modelling and control of the Qball X4 quadrotor system based on pid and fuzzy logic structure

    Qball X4 quadrotor sisteminin modellenmesi ve PID ve bulanık mantık yapısına dayalı kontrolü

    TOLGA BODRUMLU

    Yüksek Lisans

    İngilizce

    İngilizce

    2016

    Elektrik ve Elektronik Mühendisliğiİstanbul Teknik Üniversitesi

    Kontrol ve Otomasyon Mühendisliği Ana Bilim Dalı

    PROF. DR. MEHMET TURAN SÖYLEMEZ

  5. Tez No

    282684

    İnşaat ihalelerinde optimum kâr haddi tespiti için bulanık ? rekabetçi bir teklif stratejisi modeli

    Developing a fuzzified competitive bidding strategy model for optimum mark ? up in construction tendering

    AYSAN URAN

    Doktora

    Türkçe

    Türkçe

    2011

    İnşaat Mühendisliğiİstanbul Üniversitesi

    İnşaat Mühendisliği Ana Bilim Dalı

    PROF. DR. EKREM MANİSALI

Geri Dön