Optimization Model And Algorithms For Imbalance Of Capacity Utilization Problem

Although an increasing in the number of product variety makes manufacturers have more capability to survive in intense market competition, however it does not necessarily lead to higher profitability. The successes strongly depend on the management of its complexity; otherwise, it can even affect overall profit negatively. These make high degree of product customization becomes to a new frontier in businesses and the intelligence system to managing the variety of demand is capitalized as a critical source of competitive advantage.Most of researchers are concerned with the proper manufacturing and marketing activity plan. Nevertheless, there is no strategy yet that reached definitive conclusions on processes and performance. The high degree of product customization configuration and the varieties of products still leaded to continuously changing bottlenecks and capacity balancing problem on the shop floor. This dissertation concluded these all problems in a broad term of imbalance of capacity utilization problem, and divided the problem into two sub-problems; the dynamic bottleneck resources problem and the balancing capacity problem. The research question of how to make the customer demand turns into higher internal flexibility and more efficiency in processes was raised. The study was done under the difficulty of how to coping problem in the web-based intelligence system, where the requisite key of the study is to reduce the cause of the fluctuation of demand and the variation-in-manufacturing-processes by focusing on the batches of online customer demand, through make-to-order manufacturing with multiple-process environment. The main work and contributions can be respectively summarized as the following:To solving the imbalance of capacity utilization problem, the knowledge of treading off between the opportunity of losing and gaining was newly applied with the idea of smoothing the production planning by decreasing inconsistency of production input. The mathematical model is created by merging the original order acceptance model and the original flowshop scheduling model, then enhanced by our introduced policy of trading off between underutilized capacity and over-utilized capacity. In order to maximize the profit by maximize the total profit and minimize the residual leftover capacity cost, the effective of model was approved through brunch and bound algorithm and simulation experiments with the imitated data from the real manufacturing, Thailand’s parasol industry. The results induced to the reduction of production cost and a less significant of variation in production output.For the dynamic bottleneck resources problem, according to the characteristic of the problem and model, four of permutation-oriented estimation of distribution algorithms based on trading off between profit and the underutilization technique was purposed. Algorithms were solved by simultaneously selection and sequencing rule and compared with the newly modified genetic algorithm. The mathematical model was used, however to merging the two optimization objectives become one objective function approach helped reducing the complexity of problem. Numerical results demonstrated that the proposed model reduced the underutilized capacity and increased the total profit in multi-process with dynamic multi-bottleneck environment. The node-based algorithm was more proper with the problem and merging objective technique than the edge-based algorithm, especially when applied it with the incremental learning, its solution was far better than newly modified genetic algorithm’s was.For the balancing capacity problem, the overtime constraint was added into the experiment. The mathematical model in dynamic bottleneck resources problem was changed into three optimization objectives; maximize the total profit, minimize the residual leftover capacity cost, and minimize the gaining capacity cost. In order to maximize total profit by reducing the cost of unnecessary capacity usage*The two best of node-based estimation of distribution algorithms result in dynamic bottleneck resources problem was applied based on trading off the capacity utilization technique. The modified model with the proposed algorithm is remarkable in both solution quality and solving efficiency. NB-COIN, which is use incremental negative learning with node histogram based, could find the set of most profitable orders in overtime period. While the others algorithm suggested that the working in overtime period is unnecessary, their sets of most profitable orders are only found in normal working time period.