基于深度神经网络的机器人加工系统模态特性预测

Modal Characteristics Prediction of Robotic Machining Systems Based on Deep Neural Network

串联式工业机器人具有工作空间大且灵活性强的特点,被广泛应用于飞机蒙皮、航空透明件等大型结构件的加工。然而,工业机器人存在刚度弱、动态特性空间分布差异大的问题,导致其铣削稳定性极限低,不同加工区域的铣削性能变化明显,可供选择的工艺参数窗口狭窄的问题。研究机器人铣削系统加工过程中的动态特性,建立位姿相关模态预测模型对提升机器人加工性能有重要意义。本文以ABB机器人加工系统为研究对象,提出了一种基于深度神经网络的模态预测方法。首先,采用多普勒测振仪对机器人加工系统进行了模态试验,对多阶模态的空间变化加以分析。随后,根据机器人实际工作空间,设计测试试验组从而获取位姿相关的频响函数集,并利用有理多项式法准确辨识相关模态参数。在此基础上,采用超参数优化法建立深度神经网络预测模型,最终实现工业机器人工作空间内位姿相关的多阶模态参数准确预测。试验结果表明,该方法预测精度可达80%以上。

Due to large working space and strong flexibility,serial industrial robots are widely used in the machining of large structural parts such as aircraft skin and aviation transparent part.However,the low stiffness of industrial robots and large differences in the spatial distribution of dynamic characteristics lead to low limits of their milling stability,significant variations in milling performance in different machining regions,and narrow windows of available process parameters.It is important to study the dynamic characteristics of the robot milling system during machining and to establish a positional correlation modal prediction model to improve the robot machining performance.In this paper,a modal prediction method based on deep neural network is proposed for an ABB robotic machining system.Firstly,the modal experiment of the robot processing system is carried out by using the Doppler vibrometer,and the spatial variation of each order modal is analyzed.Then,according to the actual working space of the robot,an experiment is designed to obtain the frequency response function set related to the pose,and the related modal parameters are accurately identified by the rational polynomial method.On this basis,the hyperparamter optimization method is used to establish a deep neural network prediction model,which can accurately predict the pose-dependent modal parameters in the robot workspace.Finally,the experimental results show that the prediction accuracy of this method can reach more than 80%.