Studies On Quality Information Technology In Aluminum Alloys Resistance Spot Welding

For heightened awareness of energy and environmental concerns, aluminum alloys are recently finding increased application in aerospace, automobile, shipping, train, construction and national defenses industries. Resistance spot welding (RSW) is a major aluminum alloys sheet joining process for its many advantages of lower cost, high production and adaptability for automation. Despite these advantages, spot-welding of aluminum alloys suffers from a major problem of tend to weld defects. This problem results from both higher conductivity for heat and higher conductivity for electric compared to steel. In the past, commonly used quality monitoring method is that depends the operator's experience to determinate whether is a fault spot weld. The method is unscientific and not meets the need of high reliable and low cost to development of industrialization. On the other hand, it is impossible that spot weld quality information could be recordable and traceable required in the modernization quality control system for conventional method. Until recently there is still not a reliable quality monitoring method to check the quality of every spot weld on-line. In order to change the present status, it is necessary to induce the information technology into spot welding process of aluminum alloys and develop a spot welding of aluminum alloys on-line quality monitoring and assessing system. The system should be monitoring every spot weld quality and record related information so as to improve the manufacturer's confidence level and help to reduce the cost of the welded structures. The goal of this dissertation is to develop a quality information system used in blast wave direct current resistance spot welding (BWDCRSW) of aluminum alloys. The research effort could be understand and experience from following three aspects: acquirement technology, analysis technology and application technology of spot welding of aluminum alloys quality information. The data acquisition system based intelligent terminal used in this dissertation measured four variables: tip voltage, welding current, electrode force, and displacement. The intelligent terminal could complete independently data acquisition, A/D transfer, and data local storage. Each intelligent terminal communicates with computer by digital management module. The collected signal that come from data acquisition system based intelligent terminal are real-time, high signal-to-noise ratio and anti-interference. Expulsion and incomplete fusion are major defect in the weld quality with BWDCRSW of aluminum alloys. When expulsion or incomplete fusion occurs, signal such as tip voltage, electrode force and displacement all show abrupt changes. Through the studied the feature of the abrupt changes in signals, the dissertation presents the idea of using the step abrupt change of tip voltage signal and restlessness abrupt change of electrode force signal to identify expulsion, using Expansion Differential Value (EDV) and Maximum Drop Value (MDV) of displacement to identify incomplete fusion, using energy value calculated from tip voltage and welding current to identify expulsion and incomplete fusion simultaneously. The study proved that there is abundant nugget size information in EDV, MDV and energy value. It is show experimentally that criterions built in this dissertation could recognize weld defects and reflect the essence of spot weld quality. According to the feature of abrupt changes in tip voltage and displacement signals, the dissertation presents extracted method of expulsion characteristic information based wavelet transfer singularity detection algorithm. The dissertation also presents the methods to extract the EDV and MDV from displacement signal based on range analysis and time-domain analysis. The energy value was calculated from tip voltage and welding current. So, the six dimension characteristic vector that is used to quality assessment and defect diagnosis was built according above analysis and calculation. The dissertation firstly presents a multiple spot weld defect detection algorithm developed based on a Fuzzy Support Vector Machine (FSVM) theory using tip voltage and electrode force signal. Compared to other algorithm, Support Vector Machine (SVM) is good to solve problem of small sample, nonlinear, and higher dimensional because of its advantages such as independent empirical knowledge, global optimal, and outstanding generalization capability. FSVM could ignore the effect on that isolated point and noise point in training samples influenced. Therefore, the Optimal Hyperplane (OH) trained by FSVM has largest margin so as to enhance the capability of generalization. The FSVM classifier built in this dissertation shows good diagnosis capability and higher accuracy under the relatively limited samples come from production line. The dissertation firstly presents an accurate nugget size estimation model based Artificial Neural Network (ANN). Input parameters of the model are EDV, MDV and energy value. The model was optimized from structure designing, training algorithm and performance evaluating. The test study was carried out with independent test samples. It is shows experimentally that 94.3 percent of test samples estimation error is limited in 2 mm. In view of application, the ANN estimation model has higher accuracy and could meet the need production line. This dissertation developed a quality information system used in BWDCRSW of aluminum alloys. The system could complete functions such as processing parameter collected, data transfer and storage, related information extracted, welding parameter set and preview, data and information display, weld defect diagnosis, nugget size estimation, the quality information statistic analysis, the welding process stability assessment and so on. The system developed this paper realize not only spot weld quality information could be recordable and traceable, but revolution from empirical assessment to scientific diagnosis.