Electronic Nose, Data Evaluation for Cleanliness and Contamination Control in Cleanrooms, Machine Learning for decision making in production

Abstract

The paper reports on a feasibility study concerning the detection of offgassing substances from assemblies that could deposit on optical systems or electronics of spacecrafts in the cleanroom and influence their performance dramatically. The sensitivity of an electronic nose is sufficient to detect these very small gas concentrations. The obtained gas signal carries both the absolute information on gas molecule content and the change of newly produced gas intensity over time. Employing a principal component model based evaluation, the smell of the material and the change in gas intensity can be visualised. Using the obtained gas kinetics for a batch analysis, a machine learning approach becomes feasible thus enabling a prediction and classification of unknown materials.