Research On Quality Inspection Of Resistance Spot Welding Based On Multichannel

Resistance spot welding is widely used in various industries due to its advantages of high automation,high production efficiency and convenient operation,especially in the automobile industry,such as body bottom plate,frame,roof,door and side wall,etc.However,resistance spot welding is full of uncertainty,high non-linearity and interference because of many factors such as the short welding time and the influence of process site.In spot welding quality evaluation,while joint performance is mainly evaluated by the diameter and tensile strength of nugget,and the nugget formation is a closed and unobtrusive process.This thesis aimed at the detection time of nugget diameter damage is long,high cost,and its low efficiency,expulsion,incomplete fusion resistance spot welding electrical signal curve evaluation characteristics,distribution,small margin,surface cleaning quality study is less,working condition of actual welding resistance electrical characteristics,such as selection method of the characteristics of a single,builds multi-channel resistance spot detection system,designs and implement welding process test and bad working condition test,collects voltage signal,current signal and displacement signal in the welding process and calculates power signal.The characteristic of electrical signal curve of resistance spot welding is analyzed according to the different influencing factors of welding quality.The characteristic quantities related to the nugget diameter were selected from the electrical signal features,and the Coefficient of Variation and Grey Relation Analysis(CV-GRA)were used to analyze and select the strong correlation quantity of the nugget diameter.First,the traditional algorithms such as BP and SVR were used to predict the nugget diameter,and then the SAE-SVR algorithm was overlapping according to the shortcomings of the two algorithms.It solves the problems such as excessive parameters set by the traditional BP neural network,the inability to directly select the optimal parameters,and the inability to automatically reduce the dimension of SVR,etc.,which can more effectively predict the diameter of nugget and provide a basis for further research on welding quality.The main research contents of this paper are as follows:(1)Build a multi-channel electrical signal detection system for resistance spot welding.A resistance spot welding multi-channel electrical signal detection system is built based on Hall sensor,laser displacement sensor,twisted cable and data acquisition device,with data acquisition,analysis and filtering software,which can realize the synchronous measurement and storage of current,voltage and displacement signals.(2)Design batch welding process test and electrical signal acquisition test.The process window test of welding time,welding current and electrode pressure was designed and floated in a wide range,which is above and below the welding process specification,to study the influence law of single welding process parameter on the diameter of nugget and the characteristics of electric signal curve.Resistance spot welding test was designed under bad welding conditions,including shunt,small edge distance and poor surface quality,to study the characteristics of electric signal curve.(3)Characteristic analysis of electrical signals in multi-channel resistance spot welding process.The curve characteristics of single welding process parameters and electrical signals under bad working conditions were analyzed.The difference of voltage,current,displacement and power signal characteristic curves in welding process under normal solder joint,splash and nonfusion conditions were contrasted and analyzed.The analysis shows that the rising rate,voltage peak value and inflection point of voltage curve and power curve all increase with the increase of current.With the increase of welding time,voltage curve and power curve have little change trend.The difference is only between voltage peak and power peak.With the increase of welding pressure,the peak values of voltage curve and power curve decrease.Electrode voltage and welding power decrease with the increase of shunt degree.The smaller the solder joint edge distance,the greater the voltage peak,the greater the power peak;The voltage and the power increase when the welding surface quality is poor.(4)Determination conditions of expulsion and non-fusion signal characteristics.According to the process test of welding condition,the welding quality is divided into normal solder joint,expulsive solder joint and incompletely fusion.The judgment conditions of signal characteristic of expulsion and incomplete fusion are obtained.By analyzing the characteristic of electric signal curve of three kinds of solder joints the analysis shows that the voltage curve and power curve of stainless steel have a step drop when expulsion occurs.There is no obvious difference between the voltage curve and the power curve,only the amplitude is different.Different from the voltage curve,the power value at the inflection point of the power curve increases with the increase of the current,and the difference value is large.Therefore,the expulsion defect can be determined by the sudden change of voltage curve and power curve,and the incompletely fusion defect can be distinguished by the peak value of voltage curve and the peak value of power and the power value at inflection point.Different from stainless steel,when the voltage curve rose after the first drop aluminum alloy spatter,this is because aluminum alloy welding using double pulse welding,the normal solder joint voltage curve has a sudden drop already.Therefore,the condition of stainless steel expulsion can not be used to judge the aluminum alloy.After differential of the voltage curve of aluminum alloy,the curve oscillates and the expulsion of solder joint can be determined.(5)Study on prediction model of nugget diameter.Nine characteristic quantities related to the diameter of nugget were extracted by comparing and analyzing the electrical signal curves of different welding qualities.Five characteristic quantities strongly correlated with the diameter of nugget were selected by combining Coefficient of Variation and Grey Relation Analysis(CV-GRA).The prediction model of nugget diameter based on BP neural network is established,and the prediction accuracy of the model is 98.92%.The prediction model of nugget diameter based on SVR neural network is established,and the prediction accuracy of the model is 99.20%.Prediction model of resistance welding nugget diameter based on SAE-SVR algorithm was established to solve the problems such as too many parameters set by the traditional BP neural network,unable to directly select the optimal parameters,and the failure of dimension-reducing processing of highdimensional data in the calculation of SVR model.The prediction accuracy of the model was99.34%,which provided a basis for further research on welding quality.