Mechanical Engineering Building
Industry supported AMC Project is Shaping the Next Generation Maintenance Models for Semiconductor Manufacturing
AUSTIN, TEXAS—April 21, 2008
In capital-intensive industries such as semiconductor manufacturing, maintenance is key for maintaining machine uptime and the effective utilization of resources. Although there are relatively mature processes and techniques for manufacturing scheduling and control in such industries, there is still a need to address the unique challenges in planning and scheduling maintenance tasks, especially in terms of minimizing the disruptive effects of maintenance tasks on actual production and making sure that production targets are achieved while undertaking maintenance tasks. Most conventional maintenance models are restricted to time-based (e.g., every 3 months) or count-based (e.g., every 1000 lots processed) maintenance decision rules, regardless of the condition of the machine. In this sense, periodic or preventive maintenance can be labor intensive, expensive, and ineffective in identifying machine problems. An Advanced Manufacturing Center (AMC) project supported as part of the center's new industry affiliate program is developing the next generation maintenance models that monitor the machine conditions to determine if, when, and what type of maintenance actions should be taken. Different from periodic maintenance, so-called predictive maintenance (or condition-based maintenance) has the advantage of performing maintenance only when necessary due to observed or predicted machine conditions. This can be done on a subset of machines in the factory to tightly control the frequency of maintenance on these monitored machines. Conducted in close collaboration with the AMC industry affiliate member company AMD and its Operations Research Center, the project aims to develop underlying mathematical and simulation models for predictive maintenance, investigate solution techniques for such models, and develop scalable algorithms and rule-based practices to be used in actual systems.
Towards this goal, AMC-affiliated ME professors Erhan Kutanoglu and John J. Hasenbein have recruited Yiwei Cai, a Ph.D. candidate in the Operations Research and Industrial Engineering program, to work on the project. Cai was a summer intern at AMD twice and had just finished working on the integration of advanced process control and production scheduling as part of a project jointly supported by the National Science Foundation (NSF) and the Semiconductor Research Consortium (SRC). The researchers state that a side benefit of such tight control on maintenance actions through predictive maintenance is the possibility of taking into account the effect of maintenance on production scheduling and vice versa. For example, predictive maintenance may try to capture the dependency between the product mix that passes through the machines and the varying degradation rates of the machines. Therefore, the research team plans to study the relationship between predictive maintenance and production scheduling, two areas that are typically addressed separately both in research and in practice. However, maintenance decisions, predictive or periodic, affect the production schedule due to delays and lost production time which impact different products differently. On the other hand, the schedule and its anticipated product mix affect the machine conditions, often leading to different degradation rates, which in turn may prompt alternative maintenance decisions. For example, a production schedule that delays the deterioration of a machine, and which thus reduces the frequency of maintenance tasks may be reasonable at times. Supported with business insight, real-life data, and feedback from AMD, the team plans to showcase the progress of the project in an upcoming AMC workshop on lean manufacturing planned for Spring 2008.
About AMC and the Industry Affiliate program:
AMC is a cross-disciplinary research unit of the Cockrell School of Engineering at UT Austin and its mission is to initiate, support and coordinate research and education in advanced manufacturing and materials processing, to disseminate the results of this research to potential users, and to promote and provide resources for education in this field. AMC's newest initiative, the industry affiliate program, aims to foster industry-motivated research projects by bringing in companies as the center's industry affiliate members to work with the AMC faculty members and students. Member companies have a chance to shape the research conducted within the center by identifying important research-oriented industrial questions, find faculty and students genuinely interested in influencing practice through research, and have a chance to obtain and apply the center's cutting-edge research findings first-hand. The program's current focus is lean manufacturing, the collection of activities that aim to improve the operational efficiency and productivity through analytical decision making modeling, computer simulations and algorithms, and optimization, in areas that include (but not limited to) predictive maintenance, cycle time reduction, inventory control, coordination and integration of decision silos, small lot manufacturing, and design for manufacture.