Development of a Reliable Disc Cutter Abnormal Wear Diagnosis Bench Based on Vibration Feature Learning Methods

In this work, a static vibration test bench for diagnosing the abnormal wear states of the Tunnel Boring Machine (TBM) disc-cutters is built, and combined with the Long and Short-Term Memory (LSTM) algorithm model, the abnormal wear states are intelligently diagnosed. The test bench can apply vibration to the disc-cutters, so that the disc cutters in abnormal wear states can produce differentiated vibration response characteristics. After collecting the vibration signals by the sensor and processing the vibration features by the LSTM model, the accuracy of the feature learning model in diagnosing abnormal wear states of cutters exceeds 99.8%. The proposed test bench can be used as a key tool for studying the vibration characteristics of disc-cutters. The abnormal wear diagnosis method of disc cutters can be used in any complex geological conditions, exhibiting high reliability.