Multivariate control charts of MDF and OSB vertical density profile attributes

The vertical density profile of wood composite panels is influenced by many variables in the manufacturing process and is an important product attribute for composite panel end-users. Monitoring the quality of product attributes, such as the vertical density profile (VDP), is a critical step in ensuring the delivery of product value to the end-user and for maintaining the competitive position of a firm. In this study, multivariate control charting procedures using Hotelling's T-2 statistic for correlated VDP variables were compared with univariate control charts derived from the same VDP variables. Comparisons were based on samples obtained from typical production runs of a manufacturer of medium density fiberboard (MDF) and a manufacturer of oriented strandboard (OSB). The MDF samples were taken from 3/4-inch stock and the OSB samples were taken from 23/32-inch stock. Statistical process control (SPC) is intended to prevent the manufacture of defective product that may otherwise occur using traditional quality-control procedures. The Shewhart control chart is the primary tool of SPC, which separates variation as either "special cause" or "random." Even though Shewhart control charts provide a sound method for detecting problems in manufacturing processes that may otherwise go undetected, such charts are univariate in nature and have limitations for both uncorrelated and correlated variables. The typical analysis of univariate control charts for correlated VDP variables revealed that false signals of statistical control occurred for both the MDF and OSB samples. Multivariate control charts of Hotelling's T-2 statistic provided a more robust (less false-signals) method for control charting of correlated variables.