A statistical quality monitoring method for plastic injection molding using machine built-in sensors

Abstract

Majority of the current quality monitoring for injection molding relies on extra sensors assembled within the mold, which not only is restricted by the mold structure but also requires additional costly investment. A challenge in quality monitoring using machine built-in sensors is identifying couplings among different control trajectories and complicated effects on the product quality. In this paper, a statistical quality monitoring method is proposed, using only hydraulic pressure and screw position data obtained from built-in machine sensors. Statistical variables, which are representatives of the product quality in one batch, are first automatically extracted from the original data acquired. They are subsequently monitored by applying the principal component analysis method. Experimental results show that the rate of successful fault detection of the present method reaches 91.48 % at the confidence level 99 %, compared to 3.93 % using the multi-way principal component analysis (MPCA) method. Statistical variables are proved to be more effective and reliable for quality monitoring and fault detection compared with the whole process control variables, which are used in the classical MPCA, owning to a better compliance with the Gaussian distribution.