Improved Particle Filter-Based Modelling with Robotic GMAW for Weld Forming Prediction

To solve the knowledge modelling problem of strong nonlinear characteristics such as multivariable, high coupling and stochastic interference in GMAW process hull structures, a clustering similarity particle filter (CSPF) method based on consistency principle state trajectory is proposed to establish model for welding dynamics. The mechanism selected construct Hammerstein with uncertain noise relationship between parameters geometric weld pool, identified using step response test constant specification two parameter identification algorithms: recursive least squares final prediction error criterion. Thus, space equation accurate structure provided modelling. Relying algorithm core framework theoretical analysis, SIS filtering xmlns:xlink="http://www.w3.org/1999/xlink">GPF methods are adopted obtain combined formed by current (original) future multistage (modified) spatial information, actual system measured cluster analysis method. A new distribution generated under guidance measurement improve degradation phenomenon, update first-order Markov observation information compensate modify importance weight calculation, replace resampling strategy eliminate depletion problem, then tracking forming prediction. Through simulation experimentation GMAW, it concluded that both training effect accuracy formation can meet requirements ship its improved Meanwhile, integrated convergence theorem CSPF algorithm, compared standard auxiliary filter, better application results have advantages higher accuracy, stronger robustness, timeliness.