Method for predicting optical fiber vibration signal danger level based on deep learning

Abstract

The invention discloses a method for predicting an optical fiber vibration signal danger level based on deep learning, and the method comprises the following steps: time sequence sample collection, single-frame sample data extraction, data preprocessing, feature extraction, SVM single-frame sample addition of danger level labels, SVM single-frame sample training, SVM template generation, LSTM model construction, and LSTM model generation; acquisition of a real-time signal sequence, preprocessing of real-time signal sequence data, extracting of characteristics of the real-time signal sequence,prediction of a data sequence by using an LSTM template, and prediction of a single-frame characteristic data sequence; and danger level classification on the predicted single-frame feature data sequence by the SVM algorithm. According to the invention, real-time alarm can be realized, the danger level of the current signal sequence can be predicted, and emergency processing of on-site conditionsis facilitated; and the time sequence is analyzed, both the instantaneous signal and the forward signal are considered, the backward signal is predicted, and overall analysis is carried out on the danger level of the signal.

Description

technical field [0001] The invention relates to the technical field of optical fiber detection, in particular to a method for predicting the danger level of optical fiber vibration signals based on deep learning. Background technique [0002] The existing techniques for classifying signals of vibration-type optical fiber sensing products include: vibration signal amplitude and intensity judgment, multi-point feature supervised pattern recognition algorithm, image pattern recognition after one-dimensional signal is converted to two-dimensional image, vibration signal frequency domain and phase spectrum analysis. [0003] The main drawbacks of existing vibratory fiber optic sensing products for signal classification technology: [0004] 1) Insufficient representation of amplitude and intensity features [0005] A single vibration amplitude value and strength value can only represent the energy strength of the current signal, not the type of the signal. Different signal type...

AI Technical Summary

Problems solved by technology

[0006] 2) The multi-point feature supervised pattern recognition algorithm is only related to the current signal

[0009] Conventional pattern recognition algorithms cannot predict and classify time series

[0010] 4) Time correlation is lost when converting one-dimensional signal to two-dimensional image

[0012] 5) Frequency domain and phase spectrum analysis are difficult

[0013] Algorithms based on frequency domain and phase spectrum analysis, due to the limited data carried by one-dimensional vibration signals, it is often difficult to analyze, not only time-consuming, but also the classification effect is not good

[0014] 6) Not having the ability to predict the level of danger

For the time being, there is no prediction of the danger level of the signal, such as: the danger level of only one knock and continuous knocking cannot be distinguished

Images