基于BP神经网络的风机塔筒法兰螺栓紧固力预测

BP Neural Network-based Prediction of Bolt Tightening Force for Wind Turbine Tower Flange

为了解决对风机螺栓进行全覆盖实时监测而引起的成本高和监测效率低的问题,提出采用BP神经网络模型预测风电塔筒法兰螺栓紧固力。该网络由四层网络结构构成,以风电塔筒法兰上各个监测点螺栓紧固力值作为网络的输入,通过前向传播与反向传播依次迭代,最终完成模型训练过程。应用仿真数据的验证结果表明,与插值算法相比,使用BP神经网络预测风电塔筒法兰上螺栓紧固力的相关性提高6.73%,绝对系数提高15.53%,均方根误差下降59.95%。

The present work aimed to addressing high cost and low efficiency in full-coverage real-time monitoring of wind turbine bolts,and proposed a back propagation(BP)neural network model to predict the bolt tightening force of wind turbine tower flange.The network consisted of a four-layer network structure,and took the bolt tightening force value of each monitoring point on the wind turbine flange as the input of the network,iterated in turn through forward propagation and reverse propagation,and thereby completed the model training process.The proposed BP neural network-based predictive model was verified by simulation data to achieve an increase in correlation,an increase in absolute coefficient,and a reduction in root mean square error by 6.73%,15.53%,and 59.95%,respectively,compared with the conventional interpolation algorithm when predicting bolt tightening force on wind tower flange.