OTC Projekt

Online Tool Controlling Projekt

Outline
The aim of the project is quality control and assurance of the fabrication process in the semiconductor industry. We have developed software to analyse machine parameters and wafer data. Two goals are to be met with this system: Firstly, a correlation between typical machine malfunction and the parameters causing this is to be found. Secondly, production halts caused by machine malfunction are to be predicted.
Procedure
We receive the data (Process Control Monitoring data measured on wafers, and various machine or chamber dependent parameters called EPT data) from our industry project partners. This data is analysed with artificial neural networks (ANN) and machine learning methods. We investigate the advantage of using feature selection methods such as the Sequential Floating Forward Selection in the context of PCM data. Thereafter, either a Fuzzy-ARTMAP or Support Vector Machine (SVM) classifier is used, depending on the task. Fuzzy-ARTMAP is used to find the correlation between production parameters and machine malfunction. The influence of parameters on the malfunction can be studied by extracting Sugeno Fuzzy-rules. On the other hand, machine malfunctions or stoppages during production are predicted by an SVM, as its strength lies in the analysis of high dimensional data (such as time series).