9624402 Shi Most existing quality control techniques are off-line, i.e., they depend on the occurrence of defects that are removed or reworked to improve the outgoing quality level. To improve quality during the manufacturing process involves emphasizing defect prevention rather than defect removal, on-target production with minimum variance rather than within tolerance, and continuous improvement rather than acceptable quality. To achieve these goals, this research develops statistical process control algorithms for multivariate, correlated processes that are then integrated with on-line automatic process control algorithms. The central ideas are to develop engineering models and anticipate fault patterns from product/process knowledge, to apply advanced statistics to extract features/indices of the process characteristics based on in-process sensoring, to identify the inherent relationship between the patterns from the engineering model and the indices obtained from the statistics, and to use this knowledge in automatic diagnosis, predictive maintenance and automatic compensation of process changes. Automotive body manufacturing will be used as the application area of the research. Two new courses are to be developed, and results from this effort will be incorporated into existing quality control courses. Manufacturing quality is a significant factor in global market competition, particularly in the automotive sector, a vital part of our industrial infrastructure. The variability inherent in equipment status contributes substantially to poor quality and productivity. This work provides the technical basis for integrating on-line sensor information with operational control, and it has the potential to lead to improved equipment maintenance strategies that result in significant improvements in both product yield and equipment availability.
