Research On Ultrasonic Metal Welding Quality Evaluation Method Based On Process Information

Ultrasonic metal welding is widely used in the manufacturing of lithium battery. In the production process, welding joint quality is usually influenced by process parameters, interface condition and many other factors. In order to ensure the stability of product, it is necessary to realize welding process monitoring and welded joints quality evaluation. In this paper, various welding process information were acquired and analyzed to select 4 process signal features which reflect welding quality to realize the evaluation of welding quality.In order to obtain the welding process information, the ultrasonic metal welding process signal acquisition and analysis system was established for M5000 ultrasonic metal welding platform. Data acquisition part was developed by LabVIEW data acquisition software, data acquisition board and different sensors, it realized accruing, displaying and recording of the welding information; Data analysis part realized the calculation of ultrasonic power and ultrasonic frequency, wavelet packet time-frequency analysis of ultrasonic current signal based on MATLAB. The tensile testing machine was used to test the mechanical property of the welded joint, and the maximum lap sheer load of the joint was used to evaluate the quality of the welded joint.A large amount of welding test was conducted to deepen the understanding of welding mechanism and then the connection between process information and forming process of welding joint was established. By analyzing the influence of different welding pressure, welding time and interface conditions on maximum lap sheer load and failure mode of welded joints, this paper studied the reason why welding quality is affected by different welding conditions. The variation and characteristics of ultrasonic power, ultrasonic current and welding tip displacement of different welding quality process was analyzed, and according to the analysis results, 11 signal features which related to welding quality were extracted from those three process signals.By analyzing the connection between extracted features and the quality of welded joint, four features such as total input energy E, power difference ?P, average downward speed of welding tip Vm, main frequency band energy proportion of ultrasonic current w51 were finally selected to evaluate the welding quality. An RBF neutral network was set up to predict welding quality, using the four selected features as inputs and the strength of welding joint as the output. The relationship between process signal features and welding quality is acquired by training and testing the network and the evaluation of welded joint strength was eventually realized.