The success of sheet metal stamping relies on die performance because dies directly affect product quality and costs. Die performance is determined by a variety of factors, such as design, setup, and punch condition. Among all these factors, the punch condition often significantly deteriorates during the production process and, thus, becomes the main concern of process monitoring in practice.; This research aims to develop an on-line monitoring system for punch faults, such as wear and breakage, under progressive stamping environments. In these environments, multiple simultaneous operations are involved, so monitored signals reveal an aggregate effect of all these operations. Accordingly, it is challenging to monitor any particular operation of interest on a progressive die by extracting from the sophisticated signals the features contributed only by this particular operation. An algorithm is proposed in this study to eliminate this progressive stamping effect in the monitoring aspect.; Based on force signatures, punch wear monitoring is achieved by applying the multiresolution signal decomposition technique, adopted from image processing, as a feature extractor and the statistical pattern recognition approach as a decision maker. Training data is necessary. Punch breakage detection is, on the other hand, accomplished by utilizing the Generalized Likelihood Ratio Test (GLRT) to catch the statistical discontinuity of monitoring features extracted by multi-scaled Haar wavelets. In contrast to wear monitoring, no training data is entailed.; Experimental results demonstrate a higher than 98% successful detection rate for punch wear. As for punch breakage, the detection performance ranges from 75% for a slightly chipped condition to 100% for a severely chipped condition at the pre-specified false alarm rate of 0.5%.
展开▼
