Predicting Key Reliability Response with Limited Response Data

ABSTRACT

In this article, real-time process data are aligned in time order with periodic destructive test data on wood composite panel strength for the purpose of predictive modeling. Real-time predictive models of strength properties during product manufacture may reduce rework and scrap and avoid higher costs of unnecessary operating targets between the time periods of destructive strength samples. A modified principal component analysis (PCA) is used to develop an empirical model to predict the internal bond (IB) of particleboard (PB) based on a selected subset of process variables. These process variables are preselected by using variables correlated with IB. Our modified PCA is used on these selected standardized process variables to obtain transformed composite variables or modes. The 10 modes are reduced to three using correlation criteria and the three best modes are used to generate an empirical model to predict IB. Model results for the most often produced PB at the mill test site are presented in the article. The IB model generated from the selected variables has a root mean square error of prediction (RMSEP) for IB of 79.979 kPa, which represents a 20% RMSEP improvement for this manufacturing setting.

KEYWORDS:

• correlation analysis
• internal bond
• modulus of rupture
• prediction

Notes

¹Though categorical predictor variables were not significant in the models, it is possible to assess the significance of such variables (e.g., day/night, crew/shift, etc.) in future models.

Color versions of one or more of the figures in the article can be found online at www.tandfonline.com/lqen.