Multi-Sensor Data Analytics for Grinding Wheel Redress Life Estimation- An Approach towards Industry 4.0

Abstract

Grinding is an expensive and complex machining process, characterized by cutting grits undergoing non-uniform wear. The worn out grits influence the surface finish of the part, necessitating timely dressing. Conventionally, the dressing interval is decided either based on the wheel life end criteria viz. visual identification of the workpiece burn mark, chatter occurrence and deterioration in the part finish or on the number of parts produced. Improper dressing interval increases auxiliary machining time and grinding wheel wastage. Prevailing demands towards next generation smart manufacturing include product and process related benefits such as low operational cost, better customer service support, operation optimization and control. In the present work, we propose a low cost, process non-intrusive sensor technology with IoT enabled operational intelligence platform to estimate the redress life of grinding wheel based on wheel condition. Traverse grinding tests were carried out in a CNC surface grinding machine installed with Al2O3 wheel against D2 tool steel under wet condition. During experimentation, the spindle motor current, grinding forces and grinding wheel surface images were acquired using the Hall-effect sensor, dynamometer, and CCD camera respectively. Data acquisition, network connectivity, and cloud communication were empowered by serial output. Statistical time, frequency and wavelet domain features signifying the wheel life characteristic were extracted. To show the usefulness of motor current signals, the extracted features thereof were confirmed against grinding forces and wheel surface images. A time series Auto-Regressive Moving Average (ARMA) predictive model was developed to estimate the grinding wheel redress life using the selected root mean square (RMS) feature of a current signal. An android application was also developed for a graphical visualization of dressing time based on the RMS value of the spindle motor current signal. The developed methodology thus, allows operators and machines with sensors to communicate with each other and facilitates real-time traceability, visibility and control over the dressing action to perform automatic dressing before the wheel reaches its end of life.