Oil derrick damage identification based on wavelet packet analysis and deep learning algorithm

Aiming at the disadvantages of traditional oil derrick damage identification, such as troublesome operation and cumbersome calculation process, combined with the advantages of wavelet packet analysis for non-stationary and nonlinear signal processing and the ability of bidirectional long short-term memory (BLSTM) network to model time series, a damage identification method of oil derrick based on wavelet packet analysis and bidirectional long short-term memory network is proposed. Firstly, use wavelet packet analysis to extract the energy features of the acceleration response signal of the oil derrick after damage; Secondly, combined with the BLSTM deep learning algorithm for supervised learning of the damage location; Finally, the feature information is mapped to the corresponding damage location through the fully connected layer and the softmax classification layer, and the recognition results are obtained. The method is applied to the damage identification of ZJ70 derrick model in the laboratory. The results show that the method can quickly and accurately identify the damage location, and has high damage classification accuracy and anti-interference ability.