工业机器人谐波减速器迟滞特性的神经网络建模

谐波减速器中柔性环节与传动的非线性摩擦,导致谐波传动出现了不可避免地影响传动精度的复杂迟滞特性,为了描述谐波减速器的迟滞特性,本文构建了一个结构简洁的神经网络迟滞混合模型。该模型由类迟滞特性预处理环节和动态RBF神经网络两部分组成:对输入信号进行类迟滞预处理,处理后的信号与输入信号之间具有类迟滞特性;充分利用动态RBF神经网络实现类迟滞到谐波减速器迟滞特性的高精度映射。根据本文搭建的实验平台,在不同实验条件下获得的数据进行建模验证,在不同频率输入信号、不同负载,实现相同建模精度下,神经网络迟滞混合模型的验证精度为0.449 6(MSE),远高于经典RBF神经网络模型的3.032 1(MSE)精度,证明了所构造的神经网络迟滞混合模型的有效性和适应性。

Neural network modeling of hysteresis for harmonic drive in industrial robots

The non-linear friction caused by the flexible link and the transmission process in the harmonic drive leads to the complex hysteresis characteristics of harmonic transmission that inevitably affect the transmission accuracy. In order to describe the hysteresis characteristic of the harmonic drive, a concise neural network hysteresis hybrid model, which is composed of hysteresis-like characteristic preconditioning in series with a dynamic neural network, is constructed in this paper. It inlude:1) the preprocess the input signal to make the processed signal have a hysteresis-like with the input signal; 2) fully utilize dynamic RBF neural network to achieve high-precision approximation from hysteresis-like to the hysteresis characteristics of harmonic drive. According to the experimental platform constructed in this paper, the data obtained under different experimental conditions are modeled and verified: Under the same input accuracy and accuracy with different input signals and loads, the verification accuracy of the neural network hysteresis hybrid model is 0.4496 (MSE), which is much higher than the 3.0321 (MSE) accuracy of the classical neural network model, which proves the effectiveness and adaptability of the constructed neural network hysteresis hybrid model.