Process optimization using a fuzzy logic response surface method

Abstract
A new response surface method using fuzzy logic models (FL-RSM) has been proposed. The algorithm starts with a fuzzy logic model (FLM) constructed on the experimental data obtained with design of experiments (DOE). The gradient search method is used with a specified step size, and a confirming experiment is conducted at each step. The search continues until no further improvement in the objective function is observed in that gradient direction. The FLM is trained with the new experimental data combined with the old DOE data, and a new gradient is evaluated. The process is repeated until the working point is close to the optimum, as indicated by a marginal improvement in the objective function. Then the algorithm switches to the optimum search mode. It calculates the optimum based on the model, and a confirming experiment is conducted at the suggested optimum settings. The procedure is repeated until the exit criterion is satisfied. The optimization procedure has been applied to a vertical chemical vapor deposition (CVD) process with various noise levels. The results demonstrate the effectiveness of the proposed FL-RSM. It is similar to the existing regression-model-based RSM approaches. The main difference is that it uses one self-adjusted FLM to replace the combination of linear and nonlinear regression models. As a result, FL-RSM can be more user friendly and efficient in many applications.<>