System Development On Classification And Information Management Of Resistance Spot Welding Quality Based On KELM

Resistance spot welding(RSW)is widely used in the industrial manufacturing fields,such as automobile,aerospace and electronics.The quality of the RSW has a great influence on the reliability and performance of the devices.The conventional method to detect quality of RSW is utilizing the sampling,off-line and destructive tests,which not only causes the increase of cost and the difficulties in fully insuring quality,but also doesn't meet the requirement of real-time quality detection in RSW.Meanwhile,with the development of plant information,how to achieve the network management of test results and applications is imperative.Therefore,this paper develops an online classification system that can monitor the welding parameters related with welding quality to recognize the welding quality based on kernel extreme learning machine(KELM),and can achieve the information management of test results in the welding workshop.This paper uses some proper hardware such as the embedded board(PCM3362N)and acquisition card(USB-6009)to develop an on-line detection system of RSW quality based on VB6.0 platform.And the system monitors the real-time signals of current,voltage,dynamic resistance and electrode force while welding,then extracts the characteristic parameters representing the quality of the spot and builds a spot quality detection model based on KELM.After that the quality prediction of the RSW is realized,and the spot information is stored into data base,which is convenient to manage and inquire information.Meanwhile,in the upper machine the information management system provides the two main function modules that are respectively the welding information management module and information transmission between the upper computer and the lower computer,which is in favor of the network of workshop.The main research tasks of this paper are as follows:(1)Build the RSW quality detection system which consists of the mainboard(PCM3362),the acquisition card,sensors and the signal conditioning module to realize the acquisition of real-time welding signals.(2)Utilize the VB6 to program procedures,such as the data acquisition,analysis,storage,inquiring,transmission and self-testing,which can achieve the functions such as the real-time display of the welding signals,the prediction of welding quality and the information storage.(3)Obtain model information related to RSW through the welding tests and tensile tests.Combined with the RSW principle,seven feature parameters are extracted,and then the quality detection model based on KELM is built to realize the online and highly accurate quality detection of RSW.Compared with factual results,it demonstrates that as for the Q235 specimens whose thickness is lmm,the predicted accuracy reaches 97.83%,which can meet the dermand of engineering.(4)Due to the characteristics of multi-variable coupling and nonlinearity in RSW,the extracted features need to be further optimized.Combined with genetic algorithm to optimize the features,the results indicate that the optimized method is effective to increase the classification accuracy of RSW.Moreover,the optimized method needs the fewer training samples.(5)The information management system was designed by the VB6.0 program language,database technology and the Ethernet technology,which can achieve the management of welding information and the real-time information transmission between the upper computer and the lower computer.