RAPID MODELING, SCHEDULING, AND REAL-TIME CONTROL OF SEMICONDUCTOR FACTORIES

THIS PROPOSAL RESPONDS TO THE STATED GOALS OF DARPA SBIR TOPIC 91-181 BY PROPOSING THREE NOVEL APPROACHES: (A) DEVELOP RAPID MODELING TECHNOLOGY (RMT) TOOLS FOR DESIGN, PLANNING, AND DECISION SUPPORT OF SEMICONDUCTOR FACTORIES, (B) RECOGNIZING THE MULTIPRODUCT COMPLEXITY IN SEMICONDUCTOR MANUFACTURING, DEVELOP SCHEDULING TOOLS BASED ON LAGRANGIAN RELAXATION TECHNIQUES WITH SIGNIFICANT POTENTIAL ADVANTAGES OVER EXISTING HEURISTIC APPROACHES, AND (C) TAKING ADVANTAGE OF NOVEL ON-LINE SENSITIVITY ANALYSIS APPROACHES, DEVELOP REAL-TIME CONTROL MECHANISMS DRIVEN BY ACTUAL REAL-TIME FACTORY DATA; THESE MECHANISMS CAN OBTAIN ANSWERS TO A LARGE NUMBER OF "WHAT-IF" QUESTIONS ALL IN PARALLEL, AND HENCE RAPIDLY IDENTIFY (ON LINE) OPTIMAL PARAMETER SETTINGS OR OPERATING POLICIES. IN ALL CASES, OUR PHASE I GOAL IS TO ESTABLISH FEASIBILITY OF THESE NOVEL APPROACHES THROUGH EXPLICIT DEMONSTRATIONS OF PROTOTYPE SOFTWARE TOOLS. WE ANTICIPATE THAT THE SUCCESSFUL COMPLETION OF THE PROPOSED PHASE I EFFORT WILL (A) LEAD TO THE DEVELOPMENT OF EASY-TO-USE COMMERCIAL SOFTWARE TOOLS FOR RAPID MODELING, DESIGN, AND SCHEDULING OF SEMICONDUCTOR FACTORIES, (B) FIND EXTENSIONS TO OTHER MODERN MANUFACTURING ENVIRONMENTS, (C) INITIATE A RADICAL NEW APPROACH FOR REAL-TIME CONTROL OF FACTORY OPERATIONS THROUGH THE USE OF ACTUAL DATA, (D) LEAD TO A NOVEL WAY OF BUILDING DISCRETE EVENT SIMULATION TOOLS WITH INHERENT MASSIVELY PARALLEL PARAMETRIC ANALYSIS CAPABILITIES. THUS, A BYPRODUCT OF THIS PROJECT WILL BE A CONTRIBUTION TOWARDS SOLVING ONE OF THE MAJOR PROBLEMS IDENTIFIED IN THE 1990 DOD LIST OF CRITICAL TECHNOLOGIES (#5 SIMULATION AND MODELING).