An Optimization Client for Design Automation--Simulation Technology & Applied Research, 11520 North Port Washington Road, Suite 101B, Mequon, WI 53092-1195; (414) 240-0291
Dr. John F. DeFord, Principal Investigator
Dr. John F. DeFord, Business Official
DOE Grant No. DE-FG02-98ER82669
Amount: $74,392

Because of constraints imposed by manufacturing processes and intended use, the design of electromagnetic hardware, such as accelerator components, often requires substantial iteration before acceptable performance is achieved. Design iteration is typically very expensive and error prone even when done on a computer, since the complex process must be managed manually by an engineer. Lack of automation in the design process leads to long design times and sub-optimal hardware. This project will construct a design automation tool that will implement a two-step design process: design-of-experiment to identify critical design parameters, followed by a functional optimization to determine an optimum design. An architecture based on executable software component technology will enable the tool to effectively use the distributed computational power of networked computers, easily integrate with other products, and be user-extensible. The design automation software will be developed utilizing a client/server architecture. A Design Optimization Client will obtain services from three software objects, each of which may reside either on the local machine or on a networked machine: a design-of-experiment server implementing the Taguchi method, a functional optimization server implementing a local search algorithm, and an analysis server. The analysis server will be interfaced to the GOOFEE finite element code from the Stanford Linear Accelerator Center, and the complete package will be demonstrated on a design optimization problem involving an accelerator cavity on a heterogeneous network of Windows NT and UNIX computers.

Commercial Applications and Other Benefits as described by the awardee: Electromagnetic design automation software has the potential to reduce costs and improve product performance in the communications and computer industries, where shortened product life-cycles and other market pressures require better products in less time and for less money. The technology developed in this project should fit naturally into the commercial design process and effectively utilize networked computational resources. The modern client/server architecture of the software will allow in-house customization for more complete integration with existing technologies.