With the availability of Big Data in manufacturing, historical data to initially characterize a process is available in abundance. In fact, evaluating and selecting the best-fitted data set replaces data availability as major concern for setting up a short run SPC. We argue that due to the constant rise in computing power, it might not always be necessary to decide on one specific data set for a priori process characterization and modelling, but instead do most of the evaluation a posteriori. Thus, we introduce a new method to combine expert knowledge and Bayesian statistics for short run SPC in data-rich manufacturing environments. After a discussion on the methodology, its applicability and convergence, its application to turbine blade manufacturing is presented.

Full text: IJEIR_2331_FINAL.pdf