Fault Diagnosis In The Processing Of Resistance Spot Welding Based On Electrode Displacement Curve

The resistance spot welding is a kind of highly efficient welding method, which is easy to realize automation. In the resitance spot welding, the time of the engendering nugget is short, and the distributing range of temperature is narrow, there are a lot of factors affect the quality, so only through stabilizing parameter of arts and crafts, which can not avoid completely producing welding faults and emergence of failed joint. It is necessary that we carry the test after welding, but it hasn't the character of real time. Especially when we using resistance spot welding, a method of welding process fault diagnosis and quality of real-time monitoring is set up to replace or reduce crack detection after welding and devastating spot test, to improve productivity and reduced cost, and to ensure quality of welding is of great significance. In welding process, the fault factors such as workpiece warping, untreated surface, voltage flunction, electrode axial misalignment and etc, which are directly and indirectly contained in the change of welding current, welding voltage and electrode displacement signals. In this paper, displacement electrode signal as the main source of information, through the analysis of time-domain and waveform characteristics, a new method, which can diagnose the faults in welding process, based on signal real-time monitoring and feature extraction is explored. The work was done as follows:1) A data gathering system based on the core of AC6115 (A/D) card was developed, by which electrode voltage, welding current and electrode displacement signals were synchronously gathered. Through the above-mentioned signals processing and analysis show that: welding signals gathered can be used as information to classify and on-line diagnosis faults in welding process.2) Fault factors such as workpiece warping, untreated surface, voltage flunction, electrode axial misalignment and etc, which may arise in the actual product environment, are simulated, and then compare electrode displacement signal in fault with electrode displacement signal in normal process. The preliminary analysis showed that all faults arises signals' strange change and quality change. The signals' strange change can be used to diagnosis faults in welding process.3) Utilized the electrode displacement signals' parameter to devide the different phases of the nugget forming, and extracted characteristic parameters of electrode displacement curve, which could describe fault state of singularity changes, and by which a factor vector token as fault state could be get.4) A prediction model based on RBF neural network is established to predict electrode displacement curve, in which the welding current, welding time and electrode pressure are as the input vectors, and the rising slope V₁, rising slope V₂, declining slope V₃, the peak displacement S₁ and the end displacement S₂ are as the output vectors. Judging by the characteristic parameters extracted, faults classification rules in welding process have established.5) One kind of faults diagnosis method baced on case-based reasoning in welding process has probed. For several kinds' representative faults: electrode axial misalignment, voltage flunction, untreated surface and etc, a case retrieval model baced on fuzzy similar ratio is established, after validity checkout, the model diagnosis rightness rate's turn to be 94%.