Signal Based On The Resistance Spot Welding Process Weld Quality Monitoring Study

Resistance spot welding is widely used for joining sheets such as automobile body, aerospace, household appliances and furniture. Formation of the nugget is in invisible, which makes quality parameters of joints such as diameter and strength of nugget can not be observed directly, so quality monitoring and controlling of resistance spot welding are carried out hardly. Because of its more and more wider used for different fields, quality controlling of resistance spot welding has become hot point in correlative apartments. Old quality controlling ways mostly build up models based on special hypothetic situation, but results diverge indeed objects far away because of complexity of welding course, so satisfied results can be gained difficultly. Purposed on on-line monitoring and controlling of resistance spot welding quality, this thesis selects welding current, voltage and displacement as subject investigated, adopts artificial intelligent methods such as neural network and fuzzy controlling to build up process model of resistance spot welding. The work was done as follows:1) Developing of sampling system. A signal collection system was developed with KS2062, which can carry out the task of sampling welding voltage, current and electrode displacement signals during resistance spot welding.2) Characteristic extracting of signals. Linear equations are used for depicting variety trend of welding parameters, characteristic parameters and arithmetic are given subsequently.3) Prediction model based on RBF neural network. Generalized time series of cycle parameters during resistance spot welding are used for network inputs, corresponding strength of joints are used for network outputs, the Radial Basis Function (RBF) neutral network is selected for constructing reflecting relations between inputs and outputs. In the end, depends on signals during resistance spot welding, quality of joints can be known directly.4) Expulsion recognizing system. Fuzzy clustering and fuzzy inference methods are used for recognizing expulsion phenomena during resistance spot welding.5) Fuzzy inference of quality. Resistance spot welding courses are classed base on results of expulsion recognizing system, if expulsion is recognized, dynamic resistance parameters are used for build up fuzzy inference system of quality; if no expulsion is recognized electrode displacement parameters are used.6) Results of RBF neural network predicting and fuzzy inference are compared, results show that inference results are more precise.