Laser Welding Quality Monitoring Based On Information Fusion

Laser welding is an extremely complicated chemical and physical process influenced by varieties of factors. Due to its complexity, traditional single sensor can hardly meet the requirements to monitor the quality of laser welding. Therefore, this thesis devoted to quality monitoring system of the laser welding process based on multi-sensor data fusion technology.This thesis first introduced the characteristics of laser welding process, based on which two pairs of infrared sensors and visible light sensors were used for acquisition of the optical signal generated by the welding. Then the output signals from sensors were processed by the filter and amplifier through the module of signal processing, which were standardized afterwards. PC acquired signals with a PCI data acquisition card. And based on LabWindows/CVI as virtual instrument designing platform, signals of welding were processed, for ensuring the hardware and software safety during real-time monitoring and the off-line analysis of the laser welding process.In this thesis, the multi-sensor data fusion technique was used for monitoring the quality of the laser welding, due to the complicacy of laser line identification and analysis. The particle swarm optimization algorithm, an intelligent global optimization algorithm, can solve shortcomings of the slow convergence rate and vulnerability to be trapped in local minimum of the traditional neural network. In this thesis, PSO intelligent search algorithm was combined with the traditional BP neural network and utilized to revise the joint coefficients of the network. Meanwhile, standard BP neural network was improved successfully. And the conditions of welding line were classified by modified BP neural network. The final experiment results showed that multi-sensor data fusion algorithm provides an effective method to monitor the quality of the laser welding process.