Improving Maintenance Scheduling through Knowledge Based Simulation (KBS Improve)

Context

Maintenance is rapidly becoming the largest indirect operating cost for manufacturing organisations. The trend is set to continue as 'low cost' manufacturing methods place ever greater emphasis on automation and reduced work-in-progress. A key element of the maintenance process is the scheduling of maintenance tasks, which if performed effectively reduces the loss of throughput caused by machine failures and reduces the need for work-in-progress (WIP) to buffer the process from such interruptions. Maintenance scheduling, however, proves to be an extremely complex task that continues to be largely within the remit of a maintenance supervisor. A means for improving the performance of supervisors in maintenance scheduling would undoubtedly lead to reduced manufacturing costs and improved throughput.

Aims and Objectives

The objectives are:

to develop a mechanism for determining the scheduling strategies of maintenance supervisors

to develop a means for determining the effect of alternative decision making strategies on key performance measures of the manufacturing system, particularly maintenance costs, WIP and throughput

to develop a means for improving maintenance scheduling by comparing alternative decision making strategies

Method

The research is to be based on a simulation modelling approach linked to artificial intelligence (AI) systems that represent the maintenance supervisors' decision making strategies. Initially a detailed model of a specific Ford facility is to be developed, through which maintenance supervisors are to be presented with typical maintenance scenarios. These scenarios and the resulting decisions will be recorded and used to train the AI systems (neural networks and rule based expert systems). The AI systems will then be used to develop an understanding of the supervisors' decision making strategies, and in co-operation with the simulation used to determine the effectiveness of the various strategies and to look for improvements in those strategies.

Benefits

The development of a methodology for improving maintenance supervisor decisions, leading to increased throughput and reduced maintenance costs as well as a reduced need for work-in-progress. A key advantage of the proposed approach is that simulation models can be developed prior to the commissioning of a new manufacturing facility and, therefore, maintenance supervisors could be trained prior to start-up. The approach also has benefits for the wider community by providing a means for modelling human decision-makers and a method for improving their decision making strategies.

Partners

Ford Motor Company

University of Warwick

Aston University

Lanner Group

 

 

Return to Stewart Robinson's Home Page