Processing technologies use of data in the new millennium and the evolution of process knowledge
Young, T. M.  2014.  PTF BPI 2014. 3rd Intern. Conf. Processing Tech. for the Forest and Bio-based Products Industries, Salzburg/Kuchl, Austria, 24-26 September 2014. p. 21-29.  (invited)

The manuscript assesses the use of real-time process data by forest products manufacturers as related to data mining and knowledge discovery. Forest products companies have intellectual latency con-cerning process variation and such latency leads to higher than necessary costs of manufacturing, underdeveloped product quality, and inaccurate prediction of key product attributes. Strength properties from the laboratory are unreliable for real-time prediction given the time lag between sampling from the process and development of the product quality attribute statistics. Real-time process data when aligned with destructive test data from the laboratory improves manufacturing latency, identifies sources of process variation, and improves the predictability of the process. This discovery of sources of process and product variation is essential for process improvement and low costs of manufacturing. The manuscript outlines the importance of the real-time relational database and data quality verification as essential steps in continuous improvement. The proper fusion of process data with product attributes from the destructive test laboratory is a key step in continuous improvement. Statistical methods for inductive data mining are discussed, e.g., multiple linear regression analysis (MLR), regression trees (RT), boosted regression trees, Bayesian adaptive regression trees (BART), principal component analysis (PCA), and quantile regression. The strengths and weaknesses of each method are discussed.