A CMT Welding Process Component Monitoring System Based On Current And Spectral Synergistic Perception

With the development of the welding manufacturing industry,the traditional post-weld inspection can no longer meet the needs of the process.The electric current and spectrum generated during the welding process contain abundant information related to welding quality,which can provide data support for welding process parameter adjustment and welding quality judgment.This paper proposes an electric current and spectrum collaborative awareness method and designs an collaborative awareness systems based on ZYNQ development board.The main details are as follows:(1)Electric current-based shielding gas flow rate monitoring.In this paper,the Apriori data mining algorithm is applied to monitor the shielding gas flow for the first time.The obtained association rules are transplanted to FPGA through hardware design,which can concisely and efficiently monitor the flow rate of shielding gas online.The module is connected to the welding machine control system and improves the welding quality.(2)Spectrum-based welding wire composition monitoring.In this paper,we revise and optimize the existing convolutional neural network,import the Inception structure,and rebuild a set of convolutional neural networks suitable for embedded development boards,Then use HLS software to design the hardware of the convolutional neural network.We make full use of FPGA parallelism,it can improve the calculation efficiency and speed,which can realize monitor the change of welding wire composition.The module is also connected to the welding machine control system and improves the welding quality.(3)Welding quality monitoring system based on electric current and spectrum synergy.This system is designed based on the ZYNQ development board,which has low power consumption,low cost and small volume.The flow rate of shielding gas and welding wire composition change during the welding process are fed back to the welding machine controller through the IO port and serial port to realize welding process parameters adjustment.Finally,electric current and spectrum data collaborative sensing is proposed.The electric current and spectrum are stitched together,and the fused signal is fed into the convolutional neural network for training.The comparison with the result of electric signal and spectrum,the fused result has better performance.