Loading...
Estimation of processes with a high-mix of disturbance
Harirchi, Farshad
Harirchi, Farshad
Citations
Altmetric:
Advisor
Editor
Date
Date Issued
2015
Date Submitted
Collections
Research Projects
Organizational Units
Journal Issue
Embargo Expires
Abstract
Semiconductor processing consists of many different unit operations that are combined in a sequence to create the finished product. Many of these unit operations utilize run to run (RtR) control in order to keep the process within the required manufacturing constraints. Typically, the difference between the desired and actual result of processing a particular wafer is called the bias, and it is affected by not only the particular product being produced, but also the prior processing path. Different paths in the factory can be shaped by changing any processing step, e.g., products, tools, etc. These different paths are termed as threads. Because of frequent changes and updates in semiconductor products as well as a large number of product lines, run to run control must deal with a high-mix of paths, hence a large number of bias models, each corresponding to one thread. The nature of disturbance (bias) for each thread is different in these processes. The main challenge of RtR control is to estimate this high-mix of disturbances. There are two main types of bias estimation methods in the literature: Threaded and Non-Threaded estimation. Threaded estimation is widely used in factories and it is first introduced in 90's. However, several authors have discussed a method of describing the bias for a particular thread as a sum of biases corresponding to different sections of each path, and utilize a non-threaded method to estimate all those section biases, which is called context bias. The two issues with previous implementations have been the unobservability of the state realization of the bias model, and the computational cost of the implementation. In this dissertation, we tackle these two problems as well as one of the main challenges of threaded estimation that is initialization of threads. Moreover, we propose a hybrid approach to combine threaded and non-threaded estimation to enhance the estimation performance. The implementation results for all these contributions verify the improved performance of the proposed methods over the prior work on real manufacturing data.
Associated Publications
Rights
Copyright of the original work is retained by the author.