Identification Research Of Active Balancing Reducer Experimental System

Active balancing reducer(ABR)is a new flexible adjusting device for dynamic performance,which integrates the functions of active balancing and speed reducing.The experimental system was built to verify the dynamic performance of ABR.The precise mathematical model is very important for the dynamic performance analysis.There existed some deviations between the results of the simulation and experiments while using the theoretical model of the system.System identification is a method to set up system mathematical model based on measured experimental data.In order to obtain a more precise system model of the test bed,the system identification method was used in this paper.Neural network system identification method was seleced to set up the identification model of the ABR experimental system based on an amount of data.The dynamic performance of the ABR experimental system was analyzed based on the system theory model and identification model through simulation and experimental method,respectively.The results showed that the identification model was more accurate to the actual system,and the dynamic performance of the system based on the identification model was improved.Specific work includes:(1)The data collection system of the ABR experimental system was set up based on the existing test bed.The experimental scheme was designed for further study of the system identification.The experimental data under different experimental conditions was collected and saved.(2)A kind of data processing method was proposed according to the characteristics of experimental data,which consisted of elimination of trend terms,wavelet threshold denoising and frequency domain integration based on low-frequency filtering.The experimental data were processed by the proposed method and the qualified system vibration response data were obtained.(3)On account of the non-linear characteristics of the experimental system,several algorithms based on BP neural network were studied and the identification effects by using those algorithms were analyzed through simulation.Among them,Levenberg-Marquardt algorithm had the best identification effect.The identification process was carried out by using the selected Levenberg-Marquardt algorithm and the model of the ABR experimental system was obtained.(4)The theoretical model of the ABR experimental system was deduced.The simulation and experimental study on the vibration respond of the system was carried out based on the theoretical model and the identification model,respectively.The results showed the identification model of the system was more approximate to the real system.At the same time,it was shown that the optimal design of the control parameters of the ABR based on the identification model had better balancing effect on the dynamic performance of the system.