RBF-NN和改进SA的压力传感器标定及其系统辨识的研究

Research on Calibration and System Identification of Pressure Sensor Based on RBF-NN and Improved SA

针对液压机械无级变速器(HMCVT)中电磁阀-湿式离合器-压力传感器组成的系统,为较精确地建立带有压力传感器系统的动态响应模型,并且提高在特定工况下压力传感器的标定精度,分别基于改进的模拟退火(SA)算法和径向基函数神经网络(RBF-NN)提出对该系统进行辨识以及压力传感器精确标定的方法。试验结果表明:该系统的动态响应与二阶系统响应相吻合,其固有频率为13.12 Hz,阻尼比系数为0.38,系统灵敏度为0.45;系统辨识精度较高,平均误差仅为0.07;利用零点处的测量数据,仅需2个工况下的样本作为训练样本,并结合RBF-NN能很好地对输入阶跃信号的压力传感器进行精确标定。

Aiming at the system composed of solenoid valve,wet clutch and pressure sensor in hydro-mechanical continuously variable transmission(HMCVT),in order to establish the dynamic response model of the pressure sensor system more accurately,and to improve the calibration accuracy of the pressure sensor under specific operating conditions,based on the improved simulated annealing(SA)algorithm and the radial basis function neural network(RBF-NNN),the methods to identify the system and calibrate the pressure sensor are proposed.The experimental results show that the dynamic response of the system is consistent with that of the second order system.The natural frequency is 13.12 Hz,the damping ratio is 0.38,and the system sensitivity is 0.45.The identification accuracy of the system is high,and the average error is only 0.07.According to the measurement data at the zero point,only the samples under 2 working conditions are needed as the training sample,and the pressure sensor under the input step signal condition can be calibrated accurately with the RBF-NN.