A NEURAL MONITORING SCHEME FOR ON-LINE QUALITY ASSESSMENT OF PRODUCTS IN MANUFACTURING INDUSTRY

Abstract

The viability of most manufacturing systems is often benchmarked on low cost of operation, availability, low defective output ratio, and low life-cycle cost. Output quality depends largely on system's efficiency. Efficiency is predicated on overall machinery reliability, via orchestration of effective subsystems and minimization of disturbances. The task of tailoring product quality therefore consists of functionality assurance of the production facility at the micro and macro levels. Consequently, this work is motivated by the need to assure output quality and systems reliability by continuous monitoring of operational, structural and environmental disturbances. Thus, we present herein a composite but continuous neural monitoring scheme for output quality assurance and reliability of manufacturing systems. Considering the diversity of inputs into a typical manufacturing set up, our scheme starts with the reduction of large systems using the significant parameters identification approach as incorporated into Artificial Neural Network (ANN) architecture. Subsequently, the feature extraction attributes of the Kolmogorov 's ANN is employed to monitor the health status of products and machinery. Instrumentation data for training our ANN monitoring scheme was obtained from a local food processing plant.