Study On Control Strategy And Application For Full Digital Pulse Inverter Welding Power Source

The digitalization of inverter power sources is an effective way to improve welding quality. Welding power sources work in a harsh environment with high voltage, heavy current and strong interference. It is difficult to acquire useful signals for the digital control system, especially the pulse welding method has high requirements for processing rate and reliability of control strategies. This paper studies key problems and difficulties deeply and systematically in the process of digitalization.First, two digitalization methods are analyzed for the welding system, which are respectively based on primary current feedback and secondary current feedback. Then, a new double close-loop digitalization idea is proposed that uses secondary current as the feedback signal for PWM generator and welding current controller, and uses primary current together to produce PWM to restrain magnetic bias. A new program is designed, which uses FPGA as the main controller. At last, the dissertation describes design methods of the whole idea, and points out the system study direction.According to actual parameters of components, welding system physical models are established that include the input rectifier filter circuit, the inverter, and the output rectifier filter circuit. The nonlinear interface model connects physical models and digital control system to realize hybrid modeling and simulation. System dynamic response mathematical models are established to quantitatively analyze the change mechanism of every physical quantities and effect of control parameters to understand current and voltage control. The simulation effectively predicts the work process of digital pulse MIG welding power sources, and has an important guidance for further study.A new filtering idea is proposed after characteristic analysis of welding feedback current signal and formation mechanism of noise, that current signal inμs display should be first de-noised in essence to obtain the useful signal and information from the whole waveform. First, improved IIR Butterworth filter is designed. Then, wavelet de-noising method is studied, which is based on threshold and modulus maximum. Welding experiments show that the IIR Butterworth filter can be achieved when welding current is small. When current increases, system interference becomes larger to decrease filtering effect. Although the wavelet de-noising method can achieve the desired signal feature, it has poor real-time capability. For this reason, according to signal and noise characteristics, Kalman prediction method is applied in digital welding power sources at the first time. Kalman minimum variance estimation equation group is built by measuring residual value to determine pulse time and disturbance amplitude. Threshold value is set as the critical point of residual signal. Simulation and welding experiments show that the method can effectively reduce the adverse impact of process and measurement noise to meet the need of digital PWM generator and welding current real-time control, and the digital pulse MIG welding power source can work stably and reliably.Because existent control strategies have problems of parameter adjustment inconvenience and poor control performance for welding power sources, high-speed intelligent algorithms are used to adjust PID parameters automatically on line. First, standard genetic algorithm is improved from points of parameter encoding, decoding, fitness function and so on. Square waveform welding experiments show that although the improved chaos adaptive genetic algorithm is more simple and rapid than the common adjustment method, the PID parameters are not optimal because of the inaccurate model and other aspects. Then, through optimization and improvement of back propagation artificial neural network (ANN), this paper studies expert adaptive ANN PID algorithm for inverter power sources, which could avoid local minimum, improve convergence speed, and resolve control difficulty for the complex nonlinear welding system. Welding experiments show that the current controller based on EAANN-PID control strategy can realize fine control for pulse welding current to produce droplet transition.After study of arc-length regulation and droplet transition theories, a new control strategy on arc and molten droplet transition is proposed in combination with some forward simulation conclusions, which could resolve problems of less susceptibility of welding process to interference, and ensure fast stability of arc and consistency of droplet transition. A double-loop control strategy is designed for the short-circuit problem. A Fuzzy controller is realized with two inputs and three outputs in FPGA, which has better control performance than PID controller. Welding experiments show that the welding process is stable, the droplet transition has good consistency, the weld seam is beautiful, and the welding quality is perfect.According to pulse welding technology and system design method, a welding system is realized, including a main-loop circuit based on hard-switching inverter, a FPGA-based main control system, and a host system based on 32-bit MCU. After characteristic analysis of control strategies proposed, a new realization idea is presented, which adopts pipeline optimal technique and parallel operation to achieve high speed performance, and uses sequential logic to ensure system stability. The system digitalization idea is also verified.A full digital pulse MIG welding power source is designed using the research achievements, which meets the national standard on welding power sources. Pulse MIG welding and quality detection experiments show that the full digital intelligent pulse welding power source has better performance than the analog one in any aspect. It is applied successfully in gas transmission and aluminum yacht construction projects.