Multi-degree cyclic hoist modeling and optimization

抓鉤多元週期建模與優化

Student thesis: Doctoral Thesis

View graph of relations

Author(s)

  • Xin LI

Detail(s)

Awarding Institution
Supervisors/Advisors
Award date15 Jul 2014

Abstract

Automated manufacturing systems, which integrate processing machines and material handling equipment, are commonly used in modern manufacturing industries to improve system productivity and throughput. A good example of an automated manufacturing system is an electroplating line used in the production of printed circuit boards (PCBs)-PCBs are referred to as parts in this thesis. Identical parts go through a number of sequential stages (operations) following a specified process flow. To complete a stage, a part is processed in a specific tank. The loading and unloading tank store raw and completed parts respectively. In an electroplating line, hoists are usually used to transport parts between processing tanks. There are no buffers between processing tanks. Hence, operations in automated electroplating lines can be expressed in terms of hoist moves which transport parts from one tank to the next. Cyclic scheduling is commonly used to sequence hoist moves in PCB (parts) production. After completing a required number of planned hoist moves (a cycle), the line returns to its initial status. The number of parts inserted and completed in the line during a cycle is referred to as the degree of the cycle. The main challenge in an electroplating line is to improve its production throughput done via a more effective planning of its hoist moves. This thesis tackles this challenge by modeling and optimizing multi-degree cyclic hoist moves. This research begins with a basic line, where a single hoist transports parts between successive tanks and there is one-to-one correspondence between stages and tanks. Through a detailed analysis of operations in basic lines, a mixed integer linear programming model is formulated to minimize the cycle time for a line of a given degree. Scenarios in the model are solved using the commercially available software ILOG CPLEX. Experiments on benchmark and randomly generated examples have shown that multi-degree cycles can offer significantly superior throughput. The research then proceeds to consider reentrant lines with a single hoist. Parts visit some tanks more than once in order to complete the process needed to be executed in the reentrant line. In practice, different stages have to be completed in exactly the same processing environment for each of the layers in a multi-layer PCB. The mixed integer linear programming model for basic lines is then extended to deal with reentrant lines. Numerical experiments on reentrant lines are carried out to observe the changes in efficiency achieved by the model. Parallel tanks are usually used for stages with much longer processing times for balancing the line and improving the throughput. Therefore, a staged operation rule accommodating parallel tanks is developed. It is found that the number of feasible schedules decreases when the schedule is optimized by applying the operation rule. An extended mixed integer linear programming model is formulated based on the model in basic lines. The results from the extended model are illustrated via a numerical example. Since all parts are transported by hoists between processing tanks, hoists usually turn out to be the bottlenecks in automated electroplating lines. The problem is usually avoided by using multiple hoists instead of a single hoist. Multiple hoists also help balance the line and improve throughput. When multiple hoists share the same overhead track, collisions among hoists need to be preempted since the hoists are not normally allowed to cross over one another. A hoist assignment principle without overlapping is adopted to avoid collisions in this thesis. A mixed integer linear programming model is formulated to determine the optimal schedules in multi-hoist environments. Results from a number of randomly generated examples have confirmed the benefits of employing multi-degree cycles. Moreover, the general scenario consisting of multi-hoist lines without avoiding overlapping is represented in a mixed integer linear programming model. Computational times associated with the model are found to be substantially longer than those for multi-hoist lines where no overlapping is allowed.

    Research areas

  • Electroplating, Automation