基于RBF神经网络凿岩台车钻臂逆解分析

Analysis on inverse kinematics solution of drilling arm of rock drilling jumbo based on RBF neural network

凿岩台车钻臂运动学逆解是钻臂轨迹规划的关键,采用人工神经网络是实现钻臂运动学逆解的有效途径。通过D-H法对钻臂进行正运动学分析,建立各级连杆坐标系,求解钻头位置相对于钻臂底座的位姿关系矩阵,基于蒙特卡洛法利用MATLAB计算出钻臂工作空间及掌子面覆盖范围,采用RBF神经网络逼近钻臂姿态与钻头位置的复杂映射关系,利用梯度下降法计算构建神经网络参数,完成钻臂逆解。结果表明:RBF神经网络计算转动关节角度最大误差为0.082°,误差均方根为0.0072,最大误差占比为0.91%;伸缩关节位移最大误差为1.07 mm,误差均方根为0.083,最大误差占比为0.042%,能够较为精准地计算凿岩台车钻臂逆解,为规划钻臂轨迹实现自动化与智能化凿岩提供理论基础。

The inverse kinematics solution of the drilling arm of the rock drilling jumbo is the key to plan the trajectory of the drilling arm,and the artificial neural network is an effective way to realize the inverse kinematics solution of the drilling arm.D-H method was applied to conduct the forward kinematics analysis on the drilling arm,and the coordinate system of linkage at various levels was established to solve the matrix indicating the posture relationship of the drilling bit and the pedestal of the drilling arm.In addition,the workspace of the drilling arm and the coverage scope of heading face were calculated by MATLB based on Monte Carlo method,and RBF neural network was applied to fit the complex mapping relationship between the posture of the drilling arm and the position of the drilling bit.After that,the gradient descent method was applied to calculate the parameters of constructing the neural network,so as to complete the inverse kinematics solution of the drilling arm.Results showed:the maximum error of rotational joint angle predicted by the constructed RBF neural network was 0.082°,the maximum error MES was 0.0072,and the maximum error proportion was 0.91%;the maximum error of the displacement of the telescopic joint was 1.07 mm,the error MES was 0.083,and the maximum error proportion was 0.042%.The inverse kinematics solution of the drilling arm of the rock drilling jumbo was accurately achieved,which offered theoretical basis for planning the trajectory of the drilling arm and realizing automatic and intelligent rock drilling.