Ultrasonic sensors for process control in powder injection molding.

摘要

Ultrasonic sensors show great promise in defect detection and classification when monitoring the Powder Injection Molding (PIM) manufacturing process. A complete ultrasonic sensor based system has been developed that collects ultrasonic data in-situ, processes the data in a real time fashion, uses the processed data to determine part quality, and tunes manufacturing parameters to improve part quality. Neural network techniques are used to build classifiers that use the features extracted from ultrasonic amplitude profiles as input data. The features that are selected serve as guidelines for determining future part quality. A simplex method is developed that automatically tunes the manufacturing system operating conditions based on in-situ ultrasonic sensor readings. Experimental results obtained while manufacturing two specimen geometries are presented. Two classification data sets are presented as well as data obtained from two process parameter tuning experiments.