Semiconductor device manufacturing system, processing result estimation method, search method, and server

Abstract

The purpose of this invention is to provide a technique that, even under processing conditions different from those used as training data in terms of the number of steps or step structure, can estimate processing results by inputting parameters into a model corresponding to each step. The semiconductor device manufacturing system of this invention is executed by an application: a learning step, using at least one step model, an integrated model that integrates the step models, and training data to learn the model parameters of the integrated model; an estimation step, using the learned integrated model to estimate information about the processing results obtained under the desired processing conditions; and an output step, outputting the estimated processing results; the integrated model takes the parameter values ​​corresponding to the steps of the step model as input values ​​and outputs information about the processing results, including information about processing results obtained from two or more steps.

Images