Quality Monitoring And Control Of Laser Spot Welding

Recent developments in high-power CO2 lasers have accelerated the application of laser beam welding to vehicle structure fabrication in the automotive industry. As a novel spot welding method, laser spot welding (LSW) has been shown to offer many advantages such as low heat input, minimal heat affected zone, great precision and flexibility. Especially LSW can be used as a single side, non-contact process. As a result, LSW is a very attractive joining method and can be an alternative to conventional resistance spot welding (RSW).In this paper, the relationships between the welding parameters and the geometry of laser spot welds are investigated. A study is conducted on the possibility of implementing different sensors for monitoring laser spot welding process. Artificial neural network is used for predicting the laser spot weld quality. Furthermore, a control system of laser spot welding process is established.The influence of welding parameters including laser power, welding time, defocusing distance and gap on weld dimensions and shape is investigated using 1mm thickness mild steel and stainless steel. Laser spot welding of aluminum alloy is difficult for its high optical reflection and thermal conductivity. According to this, different laser pulse shapes is carried out, and the relationship between the pulse shape and the quality of laser spot welds is studied. The experimental results show that the repetition of laser spot weld is much better using the enhanced spike pulse shape, and the quality of spot weld is improved using saw-like pulse shape, which combine the advantages of cool-down and ramp-up pulse shapes. Furthermore, an artificial neural network and a weld shape model are developed for the analysis and simulation of the correlation between the LSW parameters and weld shapes.In order to substitute LSW for RSW in manufacturing, understanding strength and failure behaviour of spot welds under different loading conditions is important. In this paper, ultimate strength and failure mechanism of laser spot welds under lap shear loading are investigated. A simple stress solution related to the far field load is conducted and the critical weld nugget diameter to ensure pullout failure mode is estimated. It is observed that the critical nugget diameter of LSW is related to the shear strength of weld interface and the tensile strength of the material at the HAZ/WM boundary.A visual sensing system based on CCD (Charge Coupled Device) camera and a temperature monitoring system using a point infrared sensor are designed to detect the diameter of weld pool surface and surface temperature during LSP. The experimental data is measured under normal conditions as well as under conditions where several interferences occur. The dynamic response characters of weld pool and surface temperature to welding parameters are investigated. With analysis of the variation of weld pool diameter and surface temperature during LSP, the sum of diameters of weld pool surface, maximum diameter, peak temperature and cooling duration are proposed as the four characterized parameters to describe the features of two sensor outputs.According to linear regression analysis technique, the single regression model and multiple regression model between characterized parameters and nugget diameter are developed. The results show that the linear relations of various models are all obvious by statistic test and the determination coefficients of the multiple regression models is 89.9%, which is apparently higher than that of the single regression models. It shows that the multi-sensor system is more efficient to monitor the weld quality.For describing the complex nonlinear relationship between the sensor information and weld quality, the artificial neural network is applied. The test results show that the related prediction error is 7% and the local minimum is the main problem to further decrease the prediction error. According to this, genetic algorithm is proposed for its strong nonlinear searching ability and the related prediction error is decreased to 3%.An ARX model and dynamic neural network are established and the dynamic response characters of weld surface diameter to laser power are investigated. The results suggest that laser spot welding process is a complex nolinear system and dynamic neural network is more suitable to describle this process. At last, fuzzy controller is designed for LSW and simulation is conducted with the dynamic neural network. Close loop controlling experiments under the disturbance of defocusing distance and pre-heating is implemented. The results show that the fuzzy controller is activated and the actural weld surface diameter curves closely match the desired curve.