Semiconductor production planning using simulation optimization
Abstract
One problem in semiconductor manufacturing optimization is to find an optimal release rate of raw materials (silicon wafers) so that the final amount of finished products (integrated circuits) can meet a specified demand on time. The underlying challenge is the existence of a nonlinear relationship between the expected output, expected work-in-progress inventory and the release rates. Traditional techniques typically use mathematical models to approximate this relationship and then use deterministic optimization techniques to solve the planning problem. In practice, however, most of the processes in an industrial setting are too complicated to be accurately modeled by a mathematical model. The main contribution of this research is to treat the underlying relationship discussed above as being unknown and then to use stochastic simulation optimization to solve the semiconductor optimization problem. An existing simulation model of the semiconductor fabrication lab is used to generate observations of the underlying unknown relationship between the expected output and input release rates which are then used in the objective function. Based on these observations, an iterative optimization called STRONG is employed to estimate the optimal release rates of the silicon wafers. The efficacy of the proposed technique is validated by performing a case study which considers varying demand, planning horizons, objective function parameters and levels of simulation randomness. It is observed that in all cases, the difference between the finished products and demand is always within 7-9% of the actual demand value. Imposing an additional penalty in the case that 95% of the demand is not met and making the overage and underage costs equal leads to a better optimization performance. The optimization process performs well also when randomness is included both in the demand and simulation.
Degree
M.S.I.E.
Advisors
Wan, Purdue University.
Subject Area
Industrial engineering
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.