ANN Based Tool Condition Monitoring and Prediction of Tool Status

摘要

Industry demands optimal performance from the machineries. The performance of manually operated and human monitored systems are mainly dependent on the operating personnel. Production process and optimal performance of automated process can further be improved if human intelligence and observations can be inbuilt into a machine system. In automated machine tools, condition monitoring of cutting tools is very important to determine the expected tool life. Tool monitoring is becoming important as most of the times reduced performance is attributed to worn out tools. Acoustic Emission (AE) signal analysis has emerged as a promising tool condition monitoring technique. AE Signal parameters such as Ring Down Count (RDC), Rise Time (RTT), Event Duration (EDT), Energy (ENT) and Peak Amplitude (PA) have been widely used by various researchers in tool status monitoring. Artificial Neural Networks (ANN) have emerged as powerful pattern classifiers because of their better generalization and fault tolerance. Artificial Neural Networks(ANN) have high discriminatory power to handle variations in the input features. In this research work, we have implemented a incremental learning Radial Basis Function (RBF) called Resource Allocating Network (RAN) neural network trained using the Acoustic Emission Parameters as input features. The trained neural network is capable of handling input features with noise and predict the condition of tool.