PACO: A predictive analysis system for manufacturing test
Timothy Jones, Joseph John Engler · 2010
Testing of products throughout the manufacturing process results in a vast amount of data which is often analyzed in a post process model. Analysis of this data is typically limited to standard Six Sigma or statistical process control methodologies. This form of analysis is susceptible to overlooking correlations between differing test configurations as well as other non-intuitive relationships within the data itself. Often, this analysis is performed too late in the testing process to affect the most beneficial change. Post processing of test data requires that complete testing be performed even though failures may occur during the testing process. This paper presents a predictive analysis system which analyzes test data in a near real-time environment. The presented Predictive Analysis Collaboration Object (PACO) receives real-time data from the testing system and performs data analysis against a constantly evolving rule set to offer rapid feedback to the appropriate personnel. Thus, the user has the ability to stop testing or prevent the unit from entering another test setup phase should faults be detected. Illustrated within this paper is the basic architecture of PACO along with the analysis algorithms which are currently utilized in the analysis. Additionally, a case study of PACO in use on a product under test at Rockwell Collins Inc. is given.