基于内含传感器的两电机调速系统神经网络左逆张力辨识

Assumed internal sensor-based neural network left-inverse for tension identification of two-motor speed regulation system

针对两电机调速系统高精度张力传感器价格昂贵、安装要求高、材料和环境限制多的问题,提出基于"内含传感器"的神经网络左逆张力软测量方法。为实现两电机调速系统张力辨识,基于"内含传感器"的概念,建立"内含传感器"张力子系统,并证明其数学模型的左可逆性。考虑到逆系统模型较为复杂,存在参数时变的特点,采用BP神经网络精确逼近张力左逆模型,并串联在原系统之后,实现张力的辨识。基于两电机实验平台,对张力辨识效果进行仿真及实验验证,研究结果表明,该策略能够快速、精确地跟踪张力实际值,且易于实现。

In order to solve the problem of high cost, high installation requirements, material and environ-ment constraints of high precision tension sensor of two-motor drive system, a tension identification metli-od based on assumed internal sensor neural network left inverse was proposed. To identify the tension oftwo-motor speed regulation system, the tension subsystem of two-motor drive sthe existence of its left-inverse was proved based on the named inherent sensor. Considerinplexity of t!ie mat!iematic model and time-variant system parameter in the left-inverse system, a novel i-dentification strategy based on neural network left inverse （ NNLI） was proposed, in which the back prop-agation neural network （ BPNN） was used to approximate the left-inverse system of tension. Then,it isconnected in series wit!i the original system to realize the estimation of tension. Timental results of two-motor drive system are given,verifying that the proposed strategy can identify theactual tension quickly and accurately.