Application study on SMC filtering techniques based on model identification

Aiming at the actual problems of the combination weigher hopper door switching disturbances and inaccurate physical parameters, a new sequential Monte Carlo (SMC) data processing method for dynamic weighing based on the Gaussian sum particle filter is presented. Through analyzing the spectrum of the weighing signal, the flaw of the physical modeling method is pointed out. The negative-step data of the dynamic calibration experiment are used to identify and obtain the object model; the asymmetric trailing characteristic with gamma distribution is adopted to model the low-frequency noise of the hopper door switching disturbances, and the noise model is obtained. On the basis of the identification model with gamma noise characteristic, a new SMC method based on the Gaussian sum particle filter is selected to process the dynamic weighing data. The simulation and experiment study results show that the Gaussian sum particle filter can effectively filter the door switching disturbances, so as to significantly improve the accuracy and speed of the dynamic weighing, and is better than traditional KF and PF.