关节式机器人应用于各类生产环节，对负载进行实时监测是确保机器人安全运行的前提。但在某些特殊场景下无法直接测量负载，通常使用动力学方法间接求解，由于其非线性特性明显且模型参数的不确定性，负载识别的精度与效率一直不高。因此本文基于傅里叶神经网络提出了一种改进模型来实现负载辨识，以提高系统负载参数的预测精度与时效性。所提方法利用傅里叶神经网络中的卷积与频域截断机制快速获取特征信号，与前馈神经网络的输出结果进行数据融合得到辨识结果。所提方法相比动力学模型求解方法精度更高、计算速度更快，仅需学习预测范围内几个相间的样本集，就可识别预测范围内的任意结果，泛化能力好。同时进行网络敏感参数的分析，并与成熟神经网络算法进行性能比较。该方法将两种神经网络模型进行协同配合，能有效识别高维数据中的不同特征集，从而实现参数辨识，为复杂非线性系统的参数识别提供参考。

The joint robot is used in all kinds of production links, real-time monitoring of the load is the premise to ensure the safe operation of the robot. However, in some special scenarios, the load cannot be measured directly, and the dynamic method is usually used to solve the problem indirectly. Due to its obvious nonlinear characteristics and the uncertainty of model parameters, the accuracy and efficiency of load identification are not high. Therefore, this paper proposes an improved model based on Fourier neural network to realize load identification, so as to improve the prediction accuracy and timeliness of system load parameters. The proposed method uses the convolution and frequency domain truncation mechanism of Fourier neural network to obtain the characteristic signal quickly, and fuses the output result of feedforward neural network to obtain the identification result. Compared with the dynamic model solving method, the proposed method has higher precision and faster calculation speed. It only needs to learn several interphase sample sets within the prediction range to identify any result within the prediction range, and has good generalization ability. At the same time, the network sensitive parameters are analyzed, and the performance is compared with the mature neural network algorithm. In this method, two kinds of neural network models cooperate with each other, which can effectively identify different feature sets in high-dimensional data, so as to realize parameter identification, and provide reference for parameter identification of complex nonlinear systems.