System and method to predict the state of a process controller in a semiconductor manufacturing facility

Abstract

The disclosed system and method relates to the prediction of processing tool control parameters, i.e. controller state, for a particular processing tool, which has little or no utilization history, i.e. is data starved or has not gone through the learning curve, for a given process, or has undergone an event for which the current controller state has been reset or is otherwise now sub-optimal. The prediction is based on the processing tool control parameters of a substantially similar processing tool, being used in a substantially similar fashion to the given situation, which has significant utilization history. The processing tool having significant utilization history may be the same processing tool as the processing tool with little or no processing history where a manufacturing event disrupts the operations thereof. In this case, the pre-event control parameters and utilization history may be used, according to the disclosed embodiments, to predict the post-event controller state. Effectively, the disclosed embodiments provide for the processing tool with little or no utilization history to inherit the controller state, i.e. the evolved control parameters, of the processing tool with significant utilization history. Thereby, the processing tool with little or no utilization history is spared having to go through the learning curve, and the associated costs in delay and resources, to arrive at a particular controller state, i.e. the processing tool does not have to go through the iterative process-evaluate-adapt procedure to refine its control parameters to achieve results within the desired specifications.