Research On Process Monitoring And Quality Prediction Of Resistance Spot Welding

Resistance spot welding technology is widely used in public transportation,household appliances,automobiles and ships,aerospace and other manufacturing fields because of its high efficiency,fast and consumable free welding characteristics.Because of many factors that directly affect the weld joint formation in the closed environment,it is impossible to predict the weld joint formation.Traditional evaluation methods include nondestructive testing and destructive testing.Nondestructive testing is the use of ultrasonic flaw detection or acoustic emission signal inspection,which is time-consuming and laborious and has poor anti-interference ability.Destructive test is to measure the diameter of solder joint or tensile tensile shear force.The detection cost of this method increases linearly with the number of inspection solder joints.Aiming at the prediction of solder joint quality,this paper studies the establishment of solder joint quality prediction model by collecting welding process signals.Aiming at the research of welding process signal acquisition,a process monitoring platform based on stm32h743iit6 is built in this paper.The platform has four functions:accurately collecting welding current,electrode voltage,welding power and dynamic resistance;Real time display of signal waveform through human-computer interaction;Touch to enlarge or reduce the waveform;Acquisition parameter setting;Signal data transmission;Abnormal data alarm and other functions.In terms of hardware,the platform studies and designs: Rogowski coil cooperates with integral circuit and conditioning circuit to realize welding current signal acquisition,twisted pair matching and conditioning circuit to realize electrode voltage signal acquisition,RS485 serial port and Bluetooth communication circuit to realize signal data transmission,TFTLCD LCD screen interface circuit to realize human-computer interaction,external storage and NAND FLASH to realize signal data storage,Expand SDRAM to realize data cache.In the software development of the platform,due to the need to realize a large number of complex functions mentioned above,the traditional front and rear systems are no longer applicable.Therefore,this paper studies the development of real-time operating system using Free RTOS.In the operating system,the underlying drivers of current sensor,voltage sensor,ADC module,TFTLCD LCD screen,serial port,external storage and cache are encapsulated and made into API interface functions.When the system creates tasks such as signal acquisition and storage,human-computer interaction waveform display,signal pre alarm and data transmission,the corresponding functional tasks are realized by calling the corresponding API functions.Because the monitoring platform needs to store and display the waveform in real time during the welding process,it has high requirements for the data transmission speed.In the software development,this paper studies the use of DMA to store the collected data into NAND FLASH,uses stemwin to develop the underlying GUI of human-computer interaction,and uses dma2 d to read the signal data cached in SDRAM to draw the waveform directly,so as to realize the real-time display of process signal waveform.After completing the software and hardware design of the welding process monitoring platform,the platform is used to conduct spot welding test on 1mm thick dx51 d + Z hotdip galvanized cold-formed steel plate.Taking the tensile shear force of solder joint as the measurement index of solder joint quality,the prediction problem is studied.The control variable method is used in the test.First,190 groups of welding tests are carried out by changing the welding current and welding time,then RS485 communication is carried out between PC and MCU,the signal data is obtained from the monitoring platform,FFT filtering is carried out,and 22 characteristic quantities are selected from the welding power curve and dynamic resistance curve for Pearson correlation analysis,Finally,six characteristic quantities related to the strong tensile shear force of the welding joint are extracted: the time of the inflection point of the welding power,the initial value of the power decline before the end of the welding,the decline rate from the initial peak point of the welding power to the inflection point,the variation coefficient of the welding power waveform,the decline amplitude of the dynamic resistance from the initial peak point to the inflection point,and the decline rate from the initial peak point of the dynamic resistance to the inflection point.Then,combined with the two important influencing factors of welding current and welding time,the BP neural network prediction model of8-10-1 structure is established.The standard deviation of the prediction error of the model is 3.88%.In order to improve the prediction accuracy of the model,the genetic algorithm is used to optimize the initial weight and threshold of the network,and the GA-BP neural network prediction model is established.The standard deviation of the prediction error of the optimized model is 0.68%,which shows that the optimized model can accurately predict the solder joint quality.Figure [62] table [15] reference [93]...