Melez Üretim Sisteminde Conwıp Kontrolü Ve Parti Bölmesinin Birlikte Modellenmesi

Abstract

Akış tipi yerleşime sahip üretim hücrelerinin ve fonksiyonel yerleşime sahip üretim birimlerinin bir arada bulunduğu, melez üretim sistemlerinde, yalnızca çekme üretim kontrolüne dayalı kanban kontrolünün uygulanması mümkün değildir. Fonksiyonel birimlerde uygulanacak olan kanban kontrolü, ürünlerin izledikleri rota karmaşıklığından, talebin değişkenliğinin fazla olmasından ötürü mümkün olmamaktadır. Böyle durumlarda uygulanması önerilen CONWIP (ing. Constant Work in Process) kontrolü ile birlikte, sipariş listesi oluşturulmakta ve süreç içi stokun sabit kalması sağlanmaktadır. Çekme kontrolünün uygulandığı üretim hücrelerinde, ürünlerin akışı karmaşık olmadığı için, tek parça akışının uygulanması mümkün olmaktadır. Ancak melez üretim ortamlarında ürünler, hücrelerin yanı sıra fonksiyonel alanlarda da işlem gördüğü için, tek parça akışının uygulanması gereksiz malzeme taşımaya sebep olacağından, tek parça akışı yerine parti bölmesi önerilmektedir. Tez çalışmasında, melez üretim sistemi için önerilen CONWIP üretim kontrolü ve parti bölmesi matematiksel modelleri geliştirilmiştir. Geliştirilen modeller GAMS optimizasyon yazılımı kullanılarak, çözülmüştür. Modellerin çıktısı, optimal alt parti büyüklükleri, ürün sıralaması ve sistemdeki yarı mamul sayısıdır. Geliştirlen matematiksel modeller, işlem süresi değişkenliği (DK), talep büyüklüğü ve hazırlık süresi azaltma oranının (DHS, YHS) farklı seviyelerinde incelenerek, CONWIP üretim kontrolünün ve parti bölmesinin ayrı ayrı ve birlikte ortalama temin süresi üzerindeki etkileri incelenmiştir. Geliştirilen matematiksel modeller NP-zor sınıfı modeller olduğu için, problem boyutunun büyümesi durumunda yararlanılabilecek, ortalama temin süresini minimize eden tek bloke mekanizmalı sezgisel bir yöntem önerilmiştir. Matematiksel modeller ve tek bloke mekanizmalı sezgisel yöntem, hipotetik ve gerçek örnekler üzerinde uygulanmıştır. Hipotetik örnekten elde edilen sonuçlara göre parti bölmesinin ve CONWIP üretim kontrolünün bir arada bulunduğu durumlar, tüm senaryolarda ortalama temin süresi açısından, diğer durumlara göre üstünlük sağlamıştır. Matmematiksel modellerin farklı durumlarda incelenmesi ile birlikte işlem süresi değişkenliğinin etkisi, incelenen diğer parametrelere göre ortalama temin süresi üzerindeki etkisinin fazla olduğu gözlemlenmiştir. Önerilen tek bloke mekanizmalı sezgisel yöntem ile matematiksel modeller karşılaştırılmış, sonuç olarak önerilen sezgisel yöntem ile matematiksel model arasındaki farkın kabul edilebilir olduğu görülmüştür.

Customer demand for ‘high-volume and low-variety’ products is shifting to customised products, and competition is getting more intense. As the number of product types increases, meeting demand promptly requires more flexibility. Thus, standardisation of operations is an important point to respond quickly to changes. The standardisation of operations requires dedicating different machines to similar product types. That is what cellular manufacturing means. In cellular manufacturing, the equipment required for all products are known in advance. However, for customised products, the design, volume and routes of products are highly variable. Thus, in addition to cellular layouts, the functional layouts in which all the equipment that functions similarly are close to each other also takes place in real manufacturing environment. In hybrid manufacturing environments where flow shop production units and functional departments exist together, a production control mechanism which relies purely on pull principles is not possible. The kanban production control on functional departments would cause lead time increses because of route complexity and demnd variability. CONWIP (CONstant Work In Process) production control need to be applied in hybrid environments. By applying CONWIP production control, the production sequence would be found and the WIP (work In Process) would be controlled. One-piece flow helps to eliminate work-in-process (WIP) inventory in the system and reveals manufacturing-related problems and provides opportunities for kaizen (continuous improvement) activities. Pull principles and one piece flow can be applied in production cells since there is no route complexity. However in hybrid production systems, products have processes in both production cells and functional departments, applying on piece flow in these environments would cause unnecessary handling of mateirals. In hybrid environments, lot splitting can be applied instead of one piece flow. Applying one-piece flow in cellular layouts is possible due to the proximity of equipment, low set-up and low operation cycle time variation. Both CONWIP production control and lot splitting support the JIT philosophy. The underlying idea in both approaches is to eliminate the non-value-added activities in the value stream through less WIP and finally reduce the lead times via small lot production. Both approaches support each other in terms of small lot production. However, there is a research gap in the literature where these approaches are investigated together. The combination of these approaches is feasible to implement in real manufacturing environments. The basic motivation behind this study is to investigate the combined effect of both CONWIP production control and lot splitting. We investigate the CONWIP production control, lot splitting and both approaches together to understand their stand-alone and combined effects. A mathematical programming approach is utilised under CONWIP production control and lot splitting. Sequence-dependent set-up times and time length of set-up times are also investigated. Developed mathematical models are solved for different levels of processing time variability, demand and set up time percent reduction. The stand alone and combined effects of CONWIP production control and lot splitting on average lead time are investigated. Sequencing models are NP-hard class problems. For larger size problem settings the solution time would increase exponentially and finding the optimal values gets harder. To cope with this disadvantage of the mathematical model, single blocking mechanism heuristic is also proposed in the study. The heuristic method is based on system blocking because of number of WIP travelling in the system as in CONWIP production control. The developed heuristic method also considers lot splitting. The mathematical models and single blocking mechanism heuristic are applied on hypothetic and real cases. The hypothetic case, there are 3 parallel cells and a functional department. 8 types of products are produced. 4 products follow cell 1 and the functional department. One product follows cell 2 and functional department and last three products follow cell 3 and the functional department. All products have uni-directional flow. In real case study, an engine seal company is investigated. The models and the heuristic methods are applied to the Multi Layer System product families. For mathematical models 7 product types are considered and 15 product types are considered for heuristic method. Results show that in both cases, when CV value gets larger average flow time increases. In LCV and low demand, applying lot splitting in a pull-controlled environment provides less improvement than a system that does not limit the amount of WIP as the proposed mathematical model proves. However, in case of HCV and high demand, lot splitting provides a remarkable improvement in CONWIP control in terms of average flow time. In all demand settings, if set-up times are reduced, average flow time tends to be lower in LCV settings than HCV settings in each of the production control scenarios. When CONWIP control is applied to push production control and lot splitting scenario in HCV and low demand settings, the improvement effect becomes larger than LCV settings. In lot-splitting scenarios, the most significant improvement is observed on LCV settings. The results indicate that both CONWIP control and lot splitting improve average flow time when the CV is low and demand is high. Although variability decreases system performance, applying CONWIP control and lot splitting in a highly variable environment decreases this negative effect on system performance. Results show that lot splitting under CONWIP production control provides better performance in terms of average flow time in a pure push production control environment. CV has a prominent effect on the average flow time performance of the production system. In LCV setting, applying CONWIP control decreases the average flow time apparently. The optimal sublot sizes show that in low set-up times setting, lot splitting is applied more in SUR settings. Thus, the importance of set-up times studies become apparent. To understand the value of the mathematical model proposed in this study, the optimum solution derived from the mathematical models is compared with two different heuristics in the related literature. The first heuristic is called flow time multiple insertion heuristic (FTMIH) that considers sequence-dependent set-up times in a flexible flow shop where at least one job does not need to be processed on any machine in the system. Kurz and Askin (2004) developed an integer programming model and proposed four different heuristics to solve the problem. FTMIH is a multiple insertion heuristic that aims to minimise the sum of flow times at every stage of the production. It is a multiple machine-multiple stage adaptation of the insertion heuristic for the Travelling Sales Person (TSP). FTMIH considers the modified processing times for each stage and after assigning parts to machines, the true processing times and sequence-dependent set-up times are considered. The FTMIH is adapted in the case considered in this study. One-piece flow structure is kept in GT cells and lot manufacturing is kept in functional department. For lot-splitting case, the heuristic developed by Kalir and Sarin (2003) that considers lot splitting in a flow shop environment is utilised. The aim of the heuristic is to find the sublot sizes and the corresponding sequence. Kalir and Sarin (2003) consider equal sublot size concept for a two-machine flow shop environment and the optimum solutions in case of a single lot and machines case, is found. Kalir and Sarin (2003) also proposed a solution procedure for when there are m machines. Developed single blocking mechanism heuristic method and the heuristics from the literature are compared with the results of the mathematical models and the difference between these methods, is considered as reasonable. Unidirectional flow system configuration can be extended to backward movement cases and also pure functional layout and/or pure GT cell layout can be evaluated by modifying the model developed. In addition, stochastic nature of the problem considered can be investigated via simulation which fits real-life settings better. The impact of concepts such as set-up times, control mechanisms and lot-splitting concepts can also be investigated.