Intelligent Diagnosis of Forging Die based on Deep Learning

Abstract

Estimating the states of a cold-forging die is challenging when trying to extract the features from load-stroke signals during the long-term process. To solve this problem, this paper proposes a diagnosis learning model with an autoencoder and a support vector machine to distinguish the die states from the load-stoke signals. The autoencoder is utilized to extract the encoded features that serve as the input of the support vector machine to cluster die states. With timely indication of the current die state estimated by the learning model, the proposed die diagnosis system is promising to achieve the goal of reducing time for die maintenance.