The Development Of Resistance Spot Welding Monitor Based On Embedded ARM And Linux

As a high efficiency,simple and low cost joining method,resistance spot welding is widely used in automobile production.But due to the unsuitable selection and fluctuation of welding parameters,as well as the workers’ improper operation,the welding quality is rather unstable.The quality inspection of solder joints often requires destructive testing,which is not only costly but also unrepresentative.Therefore,on-line monitoring of resistance spot welding process is an urgent demand to resistance spot welding producers.Aimed at meeting the demand of actual demand,this thesis firstly introduces the well-performed resistance spot welding monitoring equipment developed by our research group based on STM32F303ZET6 and EasyARM-i MX283 development board as the core of system and a suite of interaction computing core respectively.d The monitoring device is equipped with multi-channel analog signal acquisition section,is able to collect multi-channel welding signal,storage the collecting data and display the welding curve on the TFT screen simultaneously.The monitoring device can also assist in welding production by saving and displaying the process information,such as car number,spot number and group number.The user input the process information before welding production.Each time a single welding process ends,the process information will update.The monitoring device will inform user vocally if weld lacking and repetition occurs.In addition,the system can also handle Ethernet communication.The user can monitor welding status via PC working state,which prepares for network production management.Then,the spot welding monitor was tested at the FAW Hong Qi welding site.It has been proved that the system can stably handle signal acquisition,spot information and welding parameter curve displaying,spot file generation,ethernet communication and so on.As for spot file generation function for DC resistance welding machine,the voltage,resistance,flow ratio and other welding parameter for each half inverting cycle can be stored or welding instantaneous value can be stored at the speed of 80 k Bit/s.For AC resistance welding machine,the welding current,grid voltage and primary voltage can be stored at the speed of 80 k Bit/s.Firing angles and conduction angles of each welding cycle can be stored at the end of the file.At the end of this thesis,the simplified calculation model of mid-frequency DC welding is introduced and the preliminary analysis of welding data are introduced.The formula of current under ideal constant duty cycle is derived from the calculation model.According to the experimental and empirical data,the time when welding current reach steady state is estimated to be about 10 ms.The inductance-elimination dynamic resistance calculation algorithm is derived.According to the analysis and calculation of actual data,the resistance calculation is more accurate compared to traditional average algorithm when current is in unstable state.It also shows that circuit’s inductance remains almost unchanged during welding process.By calculating inductance on different conditions,we find that the inductance is not sensitive to conditions.T he fluctuation around inductance center is about-0.16%~1.52%.When there is ferromagnetic material in the circuit,the inductance value significantly increased.The increasement compared to normal welding condition is about 2.26%~3.76%.As for a preliminary analysis of the data,we compared the characteristics of dynamic resistance curve of different materials.And we choose DP240,whose resistance has obvious peak value characteristic for further analysis.The analysis includes finding welding current,workpiece thickness and number of layers’ influence to welding process.The end of the thesis introduces the calculation method of splash recognition method and splash characteristics value calculation based on dynamic resistance curve.By extracting some characteristics value of dynamic resistance of initial resistance,which includes the initial value,the firs t valley time and the first valley value,using GRNN,PNN,SVM algorithm,we tempt to predict the occurrence of splash.After running different changing training set and test set for 500 times,the result shows that the rate that more than 87.5% judge correct count divided total run count for each algorithm is 92.2%,95% and 89.2% respectively.The result shows that the prediction of welding splash is performed well.