Prediction of remaining useful life of operating mechanism driven by deep feature in classification model

Accurate evaluation of the remaining service life of the operating mechanism plays an important role in ensuring the reliable operation of power transmission and distribution network. To address the drawbacks of high computational complexity and slow convergence speed in existing data-driven regression models, in this research, based on deep feature linear fitting of classification models, a remaining useful life prediction method for operating mechanism is proposed. With the sampled acoustic signals of operating mechanisms, a residual neural network is employed to continuously evaluate the remaining service life grade. For a given sample, with the principal component analysis method, the extracted deep features from the trained network can be further optimized. Finally, with the assistance of linear fitting of a few sample data, the prediction of the remaining service life of the operating mechanism can be achieved. The full life cycle experiment of the operating mechanism verifies the effectiveness of the proposed hybrid model.