Fault Detection and Diagnosis of Industrial Robot Based on Power Consumption Modeling

Fault detection via power consumption monitoring of industrial robots is a substantial problem considered in this article, in which the healthy measurements of power consumption and encoders data for a prespecified task are employed as a reference for comparison to diagnose the potential failures or excessive degradation in the robot joints. Since most electrical and mechanical faults directly affect the consumed energy, the proposed solution analyzes the comparison outcomes between the healthy reference data with that monitored in a real time for each individual task. To integrate the power measurements with a base station, a ZigBee-based wireless data acquisition circuit has been developed to process the joints data. This article suggests a measurement-based mathematical model called Bode equations vector fitting as a robust fitting method to estimate such power consumption patterns. The achieved estimates allow a clear distinction for the potential failures in the robot joints that affect the power rate patterns even when involving sharp fluctuations. A table-based neural network classifier is presented to indicate the faulty joint or encoder according to the time intervals that divided for the executed task. The experimental results demonstrate the performance verification and feasibility of the proposed approach in ABB-IRB-1200 robot manipulator. Note to Practitioners- Industrial machines are seeking to achieve energy optimization to verify the sustainability demand goal. Currently, many industrial robotic systems are not effectively monitored and modeled mathematically toward detecting the potential faults. In this context, a faults diagnosis method with an accurate mathematical model based on reference power patterns is proposed for monitoring the performance of that system. The proposed energy-based diagnosis technique can be readily integrated with the existent industrial robots supply and can be monitored remotely. Furthermore, no significant changes in the machine's hardware, but a reference pattern of a power consumption per each individual task per each robot, are required.