Process-Oriented Basis Representations for Multivariate Process Diagnosis and Control

Barton 9322840 The award is for research involving the use of multivariate process measurements to identify most likely causes of variations in product characteristics by linking symmetry patterns in the multivariate data with symmetry patterns associated with certain kinds of production problems. Process measurement vectors will be represented using a process oriented basis that would be developed by process engineers. The representation of multivariate bias will identify a small set of potential causes. The basis developed during this research can also be used to assess impact of changes in process control variables. Such information may be useful in real-time feedforward correction of production problems in multi-stage production lines. Such a development would complement current efforts to combine process monitoring and automatic control.