Transfer Learning-Based Intelligent Fault Detection Approach for the Industrial Robotic System

With increasing customer demand, industry 4.0 gained a lot of interest, which is based on smart factories. In factories, robotic components are vulnerable to failure due various industrial operations such as assembly, manufacturing, and product handling. Timely fault detection diagnosis (FDD) important keep the operation smooth. Previously, only unloaded-based FDD algorithms were considered for system. environment, robot working under conditions speeds, loads, motions. Hence, reduce domain discrepancy between lab scale real we conducted experimentations conditions. For that purpose, an extensive experimental setup prepared perform series experiments mimicking environmental condition. addition, in previous research work, machine learning (ML) deep (DL) approaches proposed arm component detection. However, issues related DL ML approaches. The models problem-specific, complex computations. model needs huge amount data. composed layers have not been thoroughly explored; result, lacks comprehensive explanation. To overcome these issues, transfer (TL) with diverse scenarios. main contribution increase generalization capabilities PHM context previously available work. VGG16 used because its autonomous feature extractions classification. data collected variety different operating loadings, motion patterns. 1D signal converted 2D (scalogram) TL model. approach shows effective performance has variable