Siparişe dayalı üretim sistemlerinde sipariş kabul ve çizelgeleme kararları

In this thesis, we examine simultaneous order acceptance and scheduling decisions in amake-to-order system. We model the manufacturing environment as a single machineenvironment, with a set of orders coming from customers. In this pool of orders, of which weknow the release dates, due dates, processing times, deadlines, sequence dependent setuptimes and revenues, manufacturer has to decide on the subset of orders to select, consideringthe limited production capacity in order to maximize the profit. The tardiness of orders ispenalized and revenue gained from an order decreases with tardiness, until the deadline, afterwhich no revenue can be obtained. The problem generalizes some well-known schedulingproblems with the objective of minimizing total tardiness and has the sequence dependentsetup times as a further complicating property, as well as the order acceptance decisions. Onespecial case of the problem is total weighted tardiness with sequence dependent setups, whichis proven to be strongly NP-hard. As a result, the OAS problem is strongly NP-hard. In thethesis, first, we give an MILP model for the problem, and solve it with a commercial solverover a range of generated data sets. We draw some conclusions on the factors making theproblem harder. Next, we propose a heuristic algorithm called ISFAN to solve the large-sizeproblems. We compare the performance of the algorithm with respect to the optimal solution,when the problem can be solved optimally, and with several constructive heuristics wedesigned. We use upper bounds generated by LP relaxation of the MILP model forcomparison when optimal objective values are not available. LP relaxation bound isstrengthened with valid inequalities that are effective for some instances. By extensivecomputational tests, the proposed algorithms are found to provide high quality feasiblesolutions, even for large-scale instances, with modest computational effort.