细纱接头机器人神经网络自适应力跟踪导纳控制

Neural Network Adaptive Force Tracking Admittance Control for Spinning Yarn Piecing Robot

环锭纺细纱工序中断纱自动接头一直是业界难题,细纱强力低易断裂、纱线张力受环境因素影响明显等因素导致机器人接头过程中纱线张力控制困难。为解决接头过程纱线张力控制问题,提出了基于交互力预测的神经网络自适应导纳控制方法。首先设计了导纳控制器参数的神经网络自适应调整策略来解决接头过程环境模型参数动态变化导致恒导纳控制器力跟踪效果差的问题;其次针对现有自适应控制器跟踪时变期望力时由于控制滞后产生的误差突变问题,提出了一种交互力预测方法,通过加入未来控制周期交互力的预测值来完成导纳控制器参数的提前调整,进而避免期望力突变时较大的力跟踪误差产生;最后进行了仿真试验,结果表明神经网络自适应控制器在动态环境下的力跟踪任务中有很好的鲁棒性,基于交互力预测的神经网络自适应控制器在动态环境下的时变期望力跟踪任务中最大误差和总体误差相比未加入交互力预测时分别降低了78.3%和29.7%,证明了所提出方法在机器人接头过程中纱线张力跟踪控制的可靠性。

The automatic joint of broken yarn in ring spinning process has always been a difficult problem in the industry.The low spinning strength is easy to break,and the yarn tension is significantly affected by environmental factors,which makes it difficult to control the yarn tension in the robot joint process.In order to solve the problem of yarn tension control in the process of robot piecing,a neural network adaptive admittance control method based on interactive force prediction is proposed.Firstly,a neural network adaptive adjustment strategy of admittance controller parameters is designed to solve the problem that the dynamic change of the parameters of the joint process environment model leads to the poor force tracking effect of the constant admittance controller.Secondly,in order to solve the problem of the sudden change of error caused by the control lag when the existing adaptive controller tracks the time-varying expected force,an interactive force prediction method is proposed,The parameters of the admittance controller are adjusted in advance by adding the predicted value of the interaction force in the future control period,so as to avoid the large force tracking error when the expected force is abruptly changed;Finally,the simulation test is carried out,the results show that the neural network adaptive controller has good robustness in the force tracking task in dynamic environment.The maximum error and overall error of the neural network adaptive controller based on interactive force prediction in the time-varying expected force tracking task under dynamic environment are reduced by 78.3%and 29.7%respectively compared with those without interactive force prediction,which proves the reliability of the proposed method in the yarn tension tracking control in the process of robot joint.