为了解决实际工程中带减振器斜拉索索力测试难度大、精度低的问题，提出一种基于IWPA-LKCNN-LSTM的带减振器斜拉索索力智能识别方法。对实际工程中的带减振器斜拉索开展了动态响应试验，基于试验所得数据开发了一种可以智能化识别带减振器斜拉索索力的深度学习模型。模型以斜拉索索力、长度、线密度、频率和阶次作为特征输入，首先采用改进狼群算法（Improve Wolf Pack Algorithm，IWPA）对LSTM神经网络中的超参数进行自适应寻优，然后利用LKCNN-LSTM（Large convolutional Kernel Convolutional Neural Network-Long and Short-Term Memory）进行训练，从而实现对带减振器斜拉索索力的智能识别。训练后的网络在测试集上识别的索力值与实际索力值间的平均误差仅为2.024%，均方误差值仅为0.0994%，决定系数为0.9806，索力误差均小于5%。最后又与索力计算公式和其它机器学习算法进行对比论证，结果表明该方法可实现带减振器斜拉索索力的智能化精准识别，拥有广阔的应用前景。

In order to solve the problem of difficulty and low accuracy in testing the cable tension with damper in practical engineering, an intelligent dentification method of the cable tension with damper based on IWPA-LCNN-LSTM is proposed. The dynamic response test of the cable with damper in practical engineering is carried out. Based on the data obtained from the test, a deep learning model that can intelligently identify the cable tension with damper is developed. The model takes the cable tension, length, line density, frequency, and order as the feature inputs. First, the hyperparameters in the LSTM neural network are adaptive optimized by using the IWPA. Then LKCNN-LSTM is used for training. The intelligent recognition of the cable tension with damper is realized. The average error between the recognized cable tension value on the test set and the actual cable tension value is only 2.024%, the mean square error value is only 0.0994%, the coefficient of determination is 0.9806, and the cable tension error is less than 5%. Finally, it is compared with the formula of cable tension calculation and other machine learning algorithms. The results show that this method can realize the intelligent and accurate recognition of the cable tension with damper and has a broad application prospect.

国家重点发展研究计划发展计划(973)项目(2015CB057702)；国家自然科学(项目编号:52078059)