Online diagnosis algorithm for random drift error in fiber optic current sensor

Abstract

Focusing on the problem of random drift error in Fiber Optic Current Sensor (FOCS), a random drift error extraction algorithm of FOCS based on optimal wavelet packet and Long Short-Term Memory (LSTM) and Support Vector Machine (SVM) is proposed. Firstly, the characteristic of FOCS signal is analyzed, and the wavelet packet multi-resolution error signals are extracted. Secondly, since the time domain span of the random drift error is large and the short-term characteristics are not obvious, the output signal of the random drift error is extracted by the span interval. The low-frequency mean sequence is constructed by selecting mean parameters according to relevant evaluation indicators. Thirdly, segmentation sequence to construct signal database. The error signal prediction model based on LSTM is constructed by using historical data. Finally, current working state of the device is judged by the model prediction signal and SVM, then, random drift error prediction function is achieved. Experiment results show that the wavelet packet decomposition can extract equipment fault signal effectively. The prediction effect of LSTM model is better than BP neural network model. The LSTM-SVM model can be used to realize the state monitoring and fault warning for the random drift error of FOCS.