A Framework to Improve Existing Heuristic Algorithms for the Non-linear Parameter Estimation: Applied to a Case of Measuring the Critical Dimensions of Semiconductors

In order to solve nonlinear parameter estimation problems, many heuristic algorithms have been developed in a variety of academic and industrial fields. In most cases, the users need to set initial parameter values or a set of initial feasible solutions before executing the heuristic algorithms. However, these settings often affect the quality of their solutions in terms of searching time and estimation accuracy. The framework suggested in this study is designed to search appropriate initial parameter values and the set of initial feasible solutions for such heuristic algorithms, and thus to improve the performance of the existing algorithms. The framework consists of two stages. In the first stage, it searches initial parameter values required by a selected heuristic algorithm and saves the useful information about the solutions obtained during the stage. In the second stage, the framework uses the saved information to efficiently construct the initial set of feasible solutions for the algorithm and then solves the nonlinear parameter estimation problem with the set. The framework is applied to an existing heuristic algorithm to improve its performance and tested for the case of measuring critical dimensions of semiconductors. In the case study, the framework shows that more accurate solution can be derived in shorter time with the less number of calculations when it is compared to the original heuristic algorithm without the framework.