| Spot welding inverter power, as one of the important developing trends whichreliability, stability and control characteristics still need improving, so it'sindustrialization progress slowly far behind the developed countries. Thesoft-switching has the advantages of low switching loss, little electromagneticcompatibility and small sensitivity of the parasitic capacitance, therefore the spotwelding power with soft-switching can achieve high reliability, stability and efficiency.Intelligent control has predominance to solve the serious nonlinear and uncertainsystem resulting in improvement of the control characteristics of the spot weldinginverter power, this application of spot welding inverter power becomes one of thedeveloping trends. In order to improve the reliability and control characteristics of thespot welding inverter power, a novel soft-switching technology and fuzzy neuralnetwork control are studied in detail.The current transition ways and zero-voltage-on process of the main circuit forfull bridge inverter with soft-switching are studied thoroughly. The topologies towiden the zero-voltage-switching range of the lagging leg are compared and analyzed.According to the characteristics of the spot welding inverter power, a novelsoft-switching topology for the spot welding inverter power is proposed. In thetopology the leak conductor is used and there is an assistant net in parallel with thelagging leg. The proposed topology has the advantages of wide soft-switching rangeand low duty ratio loss. Based on the theory of circuit analysis and the state analysisof the switching power, the main circuit state of the new topology for each operationstate and the relationship between the principal circuit parameters are obtained, theequations to calculate the energy, the dead time and the duty ratio are built in order toachieve the zero voltage switch also.The soft-switching process and the waveforms of the power transformer are simulated and analyzed by use of the circuit simulation software as Pspice andSimplorer. The effects of transformer leak inductor, resonant capacitor, equivalentinductor of secondary and assistant net parameters on the characteristics of thesoft-switching and the power are the study focuses. The studies show that suitabletransformer leak inductor is good for soft-switching, the effect of the transformer leakinductor for output and efficiency of the power can be neglected while the leakinductor is less than 0.01mH. The resonant capacitor has effect on the zero voltage onand off of the IGBT. The equivalent inductor is benefit to widen the soft-switchingrange of the leading leg, however it narrows the soft-switching range of the laggingleg. The bigger the assistant inductor is, the narrower the soft-switching range of thelagging leg becomes. The bigger the assistant capacitor is, the wider thesoft-switching range of the lagging leg becomes. There is little effect of the assistantcapacitor and assistant inductor on the load current and the duty ratio loss.According to the results of the theory analysis and simulation, the middlefrequency transformer, the resonant capacitor and parameters of the assistant net aredesigned and calculated, the experiment system using the UC3875 as the control chipfor the soft-switching spot welding inverter with assistant net is constructed. Theexperiments show that when the IGBT of the lagging leg is hard on, there is adecreasing surge caused by the equivalent input capacitor of IGBT, the distributedinductor and the wire resistance, this decreasing surge is bad to the normal progress ofthe inverter and the life of IGBT. With the assistant net, during the current transitionof the soft-switching, there is a current spike owing to the leak inductor, the abruptchange of the magnetic flux and the nonidentity of the leak inductors for the two partsof the secondary of the transformer. And there is a non symmetric between thepositive wave and the negative wave of the secondary voltage caused by the voltagedrop on the resistance of the coil and the connector. With the assistant net, the dutyratio loss are about 13μs and 10.5μs, measured and calculated respectively. Theerror is about 19.2 percentage. Fuzzy logic control and neural network control have their respective advantagesand disadvantages. Based on the studies of the factors affecting the quality for spotwelding inverter and of the general control methods, the adaptive constant currentcontrol for spot welding inverter is put forward by use of fuzzy neural network, thefuzzy neural network model for spot welding inverter is built. The control programconsists of some functions in MATLAB linked with matrixes to build the fuzzy neuralnetwork and the BP algorithm is adopted.The way to get the training examples of the fuzzy neural network for spotwelding inverter is proposed, which is building the model of the peak value for outputcurrent and the duty ratio for inverter. This way reduced greatly the cost of getting thetraining examples. The rationality of the control logic and the precision of finaltraining are achieved by making use of the constant input training after the sinusoidalinput training and adjusting the learning ratio adaptively.By simulating, it is found that the quantification factor and the proportion factorplay an important role in the fuzzy neural network for spot welding inverter. Adjustingthe quantification factor and the proportion factor, the average control error is 2.28percentages and the overshoot is about 3.35 percentages, while there are 20percentage changes for resistance and inductance and -20 percentage changes for thesource voltage. This shows that the fuzzy neural network for spot welding inverterhave the generalization ability and the adapting capability for the varying of thesystem parameters. The study of the fuzzy neural network for spot welding inverter isthe foundation to improve the control characteristics of spot welding inverter further.
|
PDF Full Text Request