Branch selection and data optimization for selecting machines for processes in semiconductor manufacturing using AI-based predictions

In the semiconductor industry, the sequence of the manufacturing steps is given by the recipe for each specific device. Whereas only one machine may be available for an individual manufacturing step, there are steps where there exists a choice between machines performing the same task, so that the path for different batches can vary. Although there should not be any difference, in reality, the yield depends on the choice. This paper presents an AI-based strategy for selecting which branch should be taken, whenever there is a choice. This optimized selection will lead to a higher overall yield. In more detail, we will describe our branch selection approach which is based on statistical analysis of existing production data as well as the current process parameters. We will describe the first steps for generating a yield indicator which guides the selection process.