Home
Prospective Students
Support Fulton
News & Publications
Events Calendar
Engineering Technical Services
Contact Us
Resources
Newsletter Archives
Contact
Edward L. Hall,
Executive Associate Dean of Research
Phone: 480.965.2964
FAX: 480.965.0439
Location: 699 S. Mill Ave.
Tempe, Arizona 85281
research news
Research Newsletter - Summer 2006
Spotlight on Industrial Engineering (IE) Department
IE Techniques Improve Cost and Efficiency of Semiconductor Manufacturing
Dr. John Fowler
Today’s semiconductor companies have a dual challenge on their hands: Design and develop cutting-edge computer chip technology and at the same time stay ahead of the competition in cost and speed.
“You can’t just have a good product idea any more, but you also have to make sure your cost is not higher than what your competitor can deliver,” says IE professor, Dr. John Fowler, who researches the application of production and supply chain techniques to reduce manufacturing cost and cycle times and improve on-time delivery of products to customers.
“Up until the very late 1980s, how efficiently one could produce didn’t matter very much,” says Fowler. “It was what product you could develop. If you could develop the right product, people would buy it regardless because of limited competition. The emphasis then was on more on product development, less on the production process.”
In the past, electrical and chemical engineers traditionally ran these factories, explains Fowler, where the focus was on keeping the processes under control. “Now we are in the mode where almost all semiconductor industries use the same processes. They run the same equipment from the same suppliers—but the question now is ‘how can I use the equipment more efficiently than my competition.’”
“If we can make even a 1% improvement, tremendous gains in cost savings can be realized,” he adds.
The semiconductor industry is not necessarily different from other industries in how processes are applied to improve delivery time and cost of production. “While there are some additional complexities, we can certainly build upon the principles that industrial engineers have developed for other industries over the years,” says Fowler.
New scheduling approach
A recent project supported by the Factory Operations Research Center (FORCe) that is jointly funded by the Semiconductor Research Operation and International SEMATECH focused on developing a new deterministic scheduling approach for wafer fabrication operations.
This research was conducted in collaboration with several ASU colleagues (Drs. Esma Gel, Michele Pfund, and George Runger), a former ASU student now at the University of Arkansas (Dr. Scott Mason), and three colleagues from Germany (Drs. Lars Mönch, Oliver Rose, and Roland Sturm).
In a typical wafer fab, there are dozens of process flows with each flow containing 200-500 processing steps and requiring more than 100 machines. These machines are expensive, and there is a need to reduce capital by all jobs sharing a particular processing operation at different stages in the manufacturing cycle.
As a result, wafers at different stages have to compete with each other for the same machine. The manner in which this competition is resolved has a clear impact on plant performance measures. In most wafer fabs today, the competition is resolved using dispatching techniques, i.e., when a machine completes a job, a simple rule is applied to determine which job to process next.
In this project, Fowler and his colleagues developed a deterministic scheduling methodology as an alternative to dispatching. In their approach, the status of all jobs and machines in the fab is collected at the start of each production shift, and a schedule for the entire shift is generated. While the approach ultimately decomposes the problem into individual machine groups, the system-wide view of the method allows for better use of the machines and more on- time delivery of jobs.
To test the validity of this technique, the research team developed a prototype proof-of-concept model using C++ that could be commercialized into a scheduling system. The goal was to generate the schedule fast enough in order to be effective and implementable since previous approaches to this problem often required many hours to generate a schedule. The team ultimately developed an approach in which a schedule was generated for half a shift in a matter of just a few minutes.
Generating cycle times
Currently, Fowler and a team of researchers, including IE colleague, Dr. Gerald Mackulak, and Drs. Barry Nelson and Bruce Ankenman from Northwestern University, are focused on generating simulation based cycle time-throughput curves. These curves are an important analytical tool used to assess operating policies in manufacturing systems and help quantify the relationship of cycle times to throughput rates.
The goal of this project is to provide simulation-on-demand results related to product cycle times (mean, standard deviation and percentiles) as a function of the release rates of the various products or product families into the fab.
The philosophy of “simulation on demand” is to provide simulation quality results—meaning results obtained from a detailed simulation model that represents all of the relevant features of the problem at hand—in a real-time, dynamic setting that facilitates interactive “what if” analyses.
Simulation on demand is accomplished by exercising the simulation model in advance of decision making to produce a model structure consisting of a collection of metamodels that can be interpolated to answer questions of interest in real time.
The project was initially funded by the National Science Foundation and then received funding from the Factory Operations Research Center. The plan is to disseminate this into the semiconductor industry by way of student interns working at semiconductor manufacturing companies.
Simulation modeling
A third concurrent project focuses on building a simulation model of the semiconductor manufacturing supply chain. Here, the goal is to build a model of an entire supply chain by building a simulation model of each factory in the supply chain and having the separate components communicate with each other to evaluate various scenarios for the entire supply chain.
Currently, a prototype is operational that has modules running simulations in Arizona, Singapore, Germany, and Atlanta with inter-model communication via the Internet. The modules communicate in compressed time to answer ‘what if questions’ about the supply chain. This project is in collaboration with IE colleague Dr.Teresa Wu and has received funding from IBM.
Fowler was employed by SEMATECH, a semiconductor R & D consortium in the 1990s. His research centered on building simulation models for the semiconductor industry that could mimic factory operations. He was one of the founding members of an operational modeling group whose goal was to bring industrial engineering and operations research into the mainstream of semiconductor production processes.
Today, the semiconductor industry is increasingly hiring industrial engineers for this purpose, says Fowler. “The semiconductor manufacturing landscape is so competitive that hiring industrial engineers to build and exercise operational models is catching on,” he adds.