Real-time parameter adjustment and fault detection of remote laser welding by using ANN

Abstract

Remote Laser Welding (RLW) has been considered as a new and promising green technology for sheet metal assembly in automotive industry because of several benefits, such as reduced processing time, decreased factory floor footprint, flexible process base for future model introduction or product change, as well as reduced environmental impact through reduction in energy consumption. However, the recent RLW systems are limited in their applicability due to lack of systematic control methodologies. Therefore, this study aims to develop a control module to obtain good quality joints of RLW by using Artificial Neural Network (ANN) model consisting of two stages, fault detections and parameter adjustments. A certain combination of parameters value, such as melting temperature, part type and thickness, laser power, and welding speed, is used as an input for the network in the first stage. The first network can recognize the fault patterns and gives an estimated faults type as an output. Then, the second stage performs sensitivity analysis of output faults and generation of adjustment rules, resulting in parameters adjustment rules as the final output. The proposed module will provide a systematic control of RLW joints during the production and facilitate acceptable faults detection to reduce defectives.