Please use this identifier to cite or link to this item:
Title: Metaheuristics for single and multiple objectives production scheduling for the capital goods industry
Authors: Xie, Wenbin
Issue Date: 2011
Publisher: Newcastle University
Abstract: In the capital goods industry, companies produce plant and machinery that is used to produce consumer products or commodities such as electricity or gas. Typical products produced in these companies include steam turbines, large boilers and oil rigs. Scheduling of these products is difficult due to the complexity of the product structure, which involves many levels of assembly and long complex routings of many operations which are operated in multiple machines. There are also many scheduling constraints such as machine capacity as well as operation and assembly precedence relationships. Products manufactured in the capital goods industry are usually highly customised in order to meet specific customer requirements. Delivery performance is a particularly important aspect of customer service and it is common for contracts to include severe penalties for late deliveries. Holding costs are incurred if items are completed before the due date. Effective planning and inventory control are important to ensure that products are delivered on time and that inventory costs are minimised. Capital goods companies also give priority to resource utilisation to ensure production efficiency. In practice there are tradeoffs between achieving on time delivery, minimising inventory costs whilst simultaneously maximising resource utilisation. Most production scheduling research has focused on job-shops or flow-shops which ignored assembly relationships. There is a limited literature that has focused on assembly production. However, production scheduling in capital goods industry is a combination of component manufacturing (using jobbing, batch and flow processes), assembly and construction. Some components have complex operations and routings. The product structures for major products are usually complex and deep. A practical scheduling tool not only needs to solve some extremely large scheduling problems, but also needs to solve these problems within a realistic time. Multiple objectives are usually encountered in production scheduling in the capital goods industry. Most literature has focused on minimisation of total flow time, or makespan and earliness and tardiness of jobs. In the capital goods industry, inventory costs, delivery performance and machine utilisation are crucial competitive. This research develops a scheduling tool that can successfully optimise these criteria simultaneously within a realistic time. ii The aim of this research was firstly to develop the Enhanced Single-Objective Genetic Algorithm Scheduling Tool (ESOGAST) to make it suitable for solving very large production scheduling problems in capital goods industry within a realistic time. This tool aimed to minimise the combination of earliness and lateness penalties caused by early or late completion of items. The tool was compared with previous approaches in literature and was proved superior in terms of the solution quality and the computational time. Secondly, this research developed a Multi-Objective Genetic Algorithm Scheduling Tool (MOGAST) that was based upon the development of ESOGAST but was able to solve scheduling problems with multiple objectives. The objectives of this tool were to optimise delivery performance, minimise inventory costs, and maximise resource utilisation simultaneously. Thirdly, this research developed an Artificial Immune System Scheduling Tool (AISST) that achieved the same objective of the ESOGAST. The performances of both tools were compared and analysed. Results showed that AISST performs better than ESOGAST on relatively small scheduling problems, but the computation time required by the AISST was several times longer. However ESOGAST performed better than the AISST for larger problems. Optimum configurations were identified in a series of experiments that conducted for each tool. The most efficient configuration was also successfully applied for each tool to solve the full size problem and all three tools achieved satisfactory results.
Description: Ph.D thesis
Appears in Collections:Newcastle University Business School

Files in This Item:
File Description SizeFormat 
Xie 11.pdfThesis3.38 MBAdobe PDFView/Open
dspacelicence.pdfLicence43.82 kBAdobe PDFView/Open

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.