Please use this identifier to cite or link to this item:
|Title:||Hipotetik Bir Tekstil Atölyesinin Dinamik Çizelgelenmesinde Yollama Kurallarının Benzetim Tekniğiyle Analizi|
|Other Titles:||Dynamic Scheduling In Hypothetic Textile Shop For Analyzing Dispatching Rules Via Simulation Technique|
|Authors:||Bozdağ, Cafer Erhan|
|Keywords:||ayrık olay benzetimi|
discrete event simulation
design of experiment
|Publisher:||Fen Bilimleri Enstitüsü|
Institute of Science and Technology
|Abstract:||Dinamik ve rassal gelişlere sahip olan atölyelerde statik çizelge metotları işe yaramamaktadır. Bu yüzden dinamik çizelgeleme yaklaşımı kullanmak gerekir. Dinamik çizelgelemede kontrol sistemlerine dayanan ve duruma göre sisteme müdahale edip yeniden çizelgeleme yapan çeşitli yaklaşımlar söz konusudur. Bu yaklaşımlar da genelde uzman sistemlere dayanır. Fakat sürekli yeniden çizelgeleme yapmak zor olduğundan böyle atölyelerde yollama kurallarının uygulanması daha uygundur. İlgili yollama kurallarının uygulanmasında da benzetim yaygın olarak kullanılan bir tekniktir. Bu sebeple literatürden belli alt atölyelerde paralel tezgahlar ve hazırlık süresi gerektiren boya proseslerini içeren tekstil atölyesi seçilmiştir. Ancak sisteme gelen dinamik işler, sıra-bağımlılık ve esnek makinalar dikkate alınarak atölye sistemi biraz değiştirilmiş, yollama kuralları bu hipotetik atölyede benzetim tekniğiyle analiz edilmiştir. Uygulanacak kurallar için sistemin ortalama akış süresi, akış süresi varyansı gibi üretim sisteminin kalitesini ve performansını belirleyen, bunun yanında teslim gecikmesi süresi, teslim gecikmesi yüzdesi ve teslim gecikmesi varyansı ile de müşteri teslim performansını doğrudan ölçen performans ölçütleri ortaya konmuştur. Literatürde ilgili performans ölçütlerinde en iyi performans gösteren statik, dinamik ve yazarların geliştirdiği kompozit kurallar bunun yanında bu çalışmada önerilen altı kural da olmak üzere toplam 17 kural benzetimle denenmek üzere seçilmiştir. Atölye yüklemesini ayarlayan gelişler arası süre ve işlerin teslim gününü belirleyen teslim günü sıkılık katsayısı gibi iki deney parametresi, yollama kurallarının çeşitli şartlarda analizi için tespit edilmiştir. Her deney parametresi iki farklı durum içerdiğinden ve her yollama kuralı on benzetim koşumuna tabi tutulduğundan her yollama kuralı için (10x2x2) 40, toplamda (40x17) 680 benzetim koşumu yapılmıştır. Her bir performans ölçütüne göre yollama kurallarının hangi deney parametreleri kombinasyonunda iyi sonuç verdiği tek yönlü varyans analizi ve ikili istatistiksel karşılaştırmalarla analiz edilmiştir. Çalışmada önerilen kurallardan olan WINQRPT, deney parametrelerinin ikili kombinasyonlarıyla yapılan koşumlardan elde edilen ve yukarıda tanımlanan beş performans ölçütünün istatistiksel sonuçları dikkate alındığında, literatürdeki COVERT kuralıyla birlikte en iyi sınıfta yer almıştır. Böylece literatüre önemli bir yollama kuralı kazandırılmıştır. Ayrıca önerilen birbirlerine benzer kural performanslarının istatistiksel olarak birbirlerinden anlamlı derecede farklı olmadığı hatta türetildikleri kurallardan daha düşük olduğu sonucuna varılmıştır. Son olarak deney parametrelerinin kendi aralarındaki ilişkinin ve parametrelerin yollama kuralları performansına etkileri çift yönlü varyans analiziyle incelenmiştir. Yollama kuralları arasındaki ilişki ve teslim günü sıkılık etkisinin istisnalar dışında genelde teslim gecikmesi yüzdesi performans ölçütüyle sınırlı kaldığı, gelişler arası süre faktörünün yollama kuralları performansını etkileyen asıl faktör olduğu gözlenmiştir.|
Manufacturing firms have started to design their manufacturing system thanks to increasing competition which arises from globalization. However manufacturing becomes more complex than before via growth of product types. Static scheduling which depends on deterministic variables like integer programming and not take into account of shop status can no longer deal with problems. Thus, dynamic scheduling has gained an importance to succeed in ascending complexity in manufacturing and deal with issues like continuous order arrivals, process delays and machine breakdowns. There are two ways to do dynamic scheduling. Rescheduling is the first way that not only includes shop status, shop objectives and dynamic shop information but also afected from internal changes such as machine breakdowns and external changes like urgent orders. Second way is to use dispatching rules. Dispatching rules are used to determine which job is to be done next in a machine buffer. Dispatching rules are preferred on account of easy execution, short computation and no need of rescheduling while reacting random events. Dynamic scheduling problems are able to be solved in different approaches. Various approaches are combined in literature studies but classification is made with regard to major approaches. Four approaches are encountered during literature survey: Heuristic methods, agent based systems, mathematical modelling and simulation. Heuristics propose alternative approaches. Agent based systems are composed of generic shop which includes combination of different units that communicates with each other. Mathematical modelling is developed to find a solution to different kind of objective functions. Simulation, which is a last method for dynamic scheduling, is the most used methodology due to execution of different scenerios and inclination to dynamic modelling. Complex events can be modelled easily thanks to this method. Simulation method is examined in two topics according to usage areas. First one is expert scheduling system which reacts to changes in the system and takes dynamic scheduling as a root. These systems can be called production control system that grounds on simulation. Examination of static, dynamic and composite dispatching rules under different conditions and scenerios is second and base thing for simulation technique. There have been numerous studies about this topic and several rules have been found as a result of many works. In addition to this, simulation with sequence-dependant setup is also a critical issue that is worked in the literature. In this study, various dispatching rules are selected to be experimented due to the fact that hypothetic textile shop required to be experimented according to different kinds of performance measures and dispatching rules are the easiest and widely-used way to do it. Static, dynamic and composite rules are investigated for under which performance measures they perform well. During selection of dispatching rules for application most notable performance measures, results of dispatching rules with regard to these measures and hypothetic shop framework in which rules are going to be applied are taken into consideration. Besides, six dispatching rules are developed in this study that derivation method of rules are similar to counterparts in the literature survey. Experimental factors are sought for throughout the literature to try these dispatching rules in different conditions. As a result of research, frequent and rare interarrival times are given to arrange shop load. In addition to this, due date tightness coefficient , which is composed of tight and loose variables, is suggested to regulate internal due dates according to TWK rule in order to test due date based performance measures. The textile shop is taken from literature survey to make a trial with these factors. This textile shop consists of four sub-shops and applies multi-stage manufacturing system. Sub-shops have parallel machines and dye processes that requires setup. Nevertheless, the textile shop is changed slightly with regard to dynamic arrivals, sequence-dependant setup, lot sizes, order numbers and flexible machines. Performance measures such as mean flow time, percent tardy of jobs, mean tardiness, tardiness variance and flow time variance which both measure customer delivery performance and determine quality of production system for process performance is introduced to asses a performance of dispatching rules under various experimental factors. Simulation is selected to make an experiment with these dispatching rules on account of widely used technique according to literature survey. Simulations can not be done without assumptions and determining warm-up periods. Some assumptions are made according to system which is experimented. In order to specify warm-up period, plots are drawed for each dispatching rule and a warm-up period is specified according to max warm-up period time of dispatching rules related to flow time performance measure. 17 different simulation model is build to test each dispatching rules. Each experiment parameter has two variation and each dispatching rule undergoes ten simulation run so that for each dispatching rule (10x2x2) 40 and in total (40x17) 680 simulation runs are done via ARENA 11.0. One way variance tests and Tukey pairwise comparison tests are done through Minitab 16.0 to find out which dispatching rules perform well under experimental factors in combination with each performance measure. Dispatching rules are grouped according to the number of getting location in top statistical classes which are comprised of combination of five performance measures and four experimental factors that equal to (5x4) 20 Tukey test results. When results of this clustering are taken into consideration WINQRPT rule which is one of the proposed rules in this study is located in top class with COVERT. Hence, an important dispatching rule is intoduced to literature. Other dispatching rules are splitted into four groups due to how much they are located in the top statistical classes. Furthermore, similar rules which are derived from developed rules show no signifficantly difference from each other and their performance are lower than rules which are derived from. Finally, it is known that each experimental factor may have different effects on performance of dispatching rules. Apart from that, there may be some relations between two experimental factors. In order to test whether there is a relation or not, relationship between them and impacts of these factors on dispatching rules are examined through two way variance test in Minitab. These tests are executed considering all combinations of experimental factors in which dispatching rules perform under each performance measure. As a result, although relationship between experimental factors influences performance of most of dispatching rules under percent tardy of jobs and other few rules under other performance measures, it is observed that interarrival times is the main factor that has impact on performance of dispatching rules. On the other hand, due date tightness coefficient plays no efficient role in affecting performance of dispatching rules.
|Description:||Tez (Yüksek Lisans) -- İstanbul Teknik Üniversitesi, Fen Bilimleri Enstitüsü, 2012|
Thesis (M.Sc.) -- İstanbul Technical University, Institute of Science and Technology, 2012
|Appears in Collections:||Endüstri Mühendisliği Lisansüstü Programı - Yüksek Lisans|
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.