The Quality Prediction Model For Stainless Steel Spot Welding Based On The Theory Of Regression Analysis

Resistance spot welding has the advantage of high efficiency, especially applied to large-scale industrial production. Nowadays the Automotive Industry, the Railway Vehicle Industry and so on has widely applied to the method of spot welding. But resistance spot welding itself has certain limitations, for example there is no reliable non-destructive detection to examine the defects in the nugget, as well as to detect the strength of welded joints. The method which was widely used still based on the method of single feature in the field of quality monitoring and controlling for spot welding, like constant current control method, energy controlling. As the spot welding process is complex, the factors that affect the growth of nugget are not clear, there are some deficiencies for the method of traditional quality monitoring, which is not able to monitor the quality of spot welding joints accurately. In order to monitor the quality of spot welding accurately, the detection that based on multi-parameter monitoring emerged. Nowadays there are some methods for the quality of spot welding, like statistical methods, numerical simulation methods and artificial intelligence methods and so on. With regard to the method of mathematical statistics for predicting the quality of spot welding, establishing the regression model between the input parameter and the nugget diameter or the welded joint shear strength based on the theory of regression analysis, using this model to predict the quality of welding. At present, for this research home or abroad has already yielded some progress, most of which aimed at extracting the welding parameter of low-carbon steel to control the quality of welding. But to the dynamic parameters of stainless steel spot welding, the study of establishing the regression model about the quality of spot welding was rare. This paper studied the relationship between the parameters of stainless spot welding and the quality of welding, establishing the corresponding mathematical model to predict the quality of welding.It used the data acquisition system for spot welding process to gather the signal data of electrode voltage, welding current and electrode force for the process of stainless steal spot welding synchronously, putting the data of electrode voltage and welding current into the computer to get the information of dynamic resistance for stainless steal welding. The experiment showed the relationship between nugget diameter and welding current, welding time, electrode force and so on. It analyzed the characteristics of the dynamic resistance curve for stainless steel, discussed the relationship between the spot welding parameters and dynamic resistance. The analyzing results showed that the welding current and electrode force having great impact on the dynamic resistance and there was close relationship among them.The researching result showed that there was significant linear relationship between the nugget diameter of stainless steel spot welding with welding current and welding time under the determined conditions. To establish the linear regression model for two elements about the nugget diameter together with welding current and welding time under the conditions of different electrode force. As each linear regression model for two elements has great diversity, the quality can not be predicted in this model, we need to consider the factor of electrode. To establish the linear regression model for the elements of welding current, welding time and electrode force. The experiments results showed that the prediction accuracy was not satisfactory when using this kind of model to predict the quality of spot welding. Because of the characteristic for spot welding itself, the given static parameters can not fully reflect the growth of nugget. Dynamic resistance is a dynamic parameter, which can reflect the whole process for the growth of nugget. So taking into account the factor of electrode force, to extract the characteristic parameters of dynamic resistance curve,current curve and electrode force in the time of power, verifying the degree that these characteristic parameters associated with the nugget diameter. If the value of the relevant coefficient to each variable is small, this variable will be removed from the model, using this method to establish the linear regression model for multiple elements. As inputting more parameters, there exists multicollinearity among each parameters of the regression equation, the regression equation coefficients can not be a significant test. So the whole model needed to be optimized. All the regression coefficients were significant by the test after optimizing the regression model, whose prediction accuracy had been greatly improved. We can use this regression model to predict and control the quality of stainless steel spot welding.