Deterioration Score of Cold Forging Dies by Using Acoustic Emission Signals

Abstract

It is generally difficult to estimate the degree of deterioration of forging dies, but it is necessary to prevent a large number of defective products. In this study, we propose a deterioration score in cold lateral forging using acoustic emission (AE) signals. From the analysis of the measured data, the transition of the signal from the initial state to the deteriorated state can be observed, and the transition can be numerically evaluated. In the evaluation, variational auto-encoder (VAE) is used for learning the initial distribution, and the deterioration score is calculated by the degree of deviation from the learned distribution. The AE cumulative maximum amplitude and AE cumulative count during the linearly increasing stress period for each forging shot are given to the input of the VAE encoder, and valid deterioration scores are obtained for multiple actual measurements.