Monitoring and early warning system based on array type optical fiber sensing technology

By arranging fiber optic sensors under the railway tracks, using array fiber optic sensing technology to receive and process real-time data, identify abnormal points and calculate risk scores, the problems of abnormal points data accuracy and real-time transmission rate in the existing technology are solved, and the problems of abnormal points data accuracy and real-time transmission rate are improved. The reliability of monitoring and the response speed of the early warning system are provided.

CN118968733APending Publication Date: 2024-11-15贺雨轩
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202411037829.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-31
Publication Date
2024-11-15

AI Technical Summary

Technical Problem

In the monitoring and early warning, existing fiber optic sensing technology has problems such as difficulty in determining the accuracy of abnormal points data and untimely real-time data transmission rates, which affects timely early warning in the monitoring area.

Method used

Using a monitoring and early warning system based on array fiber sensor technology, by laying fiber sensors under the railway tracks, real-time data reception, preprocessing, feature extraction and signal compensation, identifying abnormal points, calculating risk scores and generating early warning signals, including layout Module, reception module, data analysis module, signal compensation module, accuracy improvement module and intelligent early warning module.

Benefits of technology

It significantly improves the system's robustness and detection accuracy in complex environments, reduces false alarm rates and missed alarm rates, and enhances the reliability of monitoring and the response speed and accuracy of the early warning system.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118968733A_ABST
    Figure CN118968733A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of optical fiber sensing, in particular to a monitoring and early warning system based on an array type optical fiber sensing technology, and aims to realize multi-point and multi-parameter real-time monitoring and early warning. Firstly, optical fiber sensors are arranged below a rail in a railway monitoring area in an array mode so as to receive real-time data of the surrounding environment of the rail; through preprocessing, feature extraction and real-time monitoring of the received real-time data, abnormal points on the rail are identified. Secondly, determining the type and position of the abnormal point according to the characteristics of the abnormal point, quantifying the loss degree of the abnormal point, and performing signal compensation on the corresponding position of the optical fiber sensor on the rail; and key features of the compensated signal are further extracted and enhanced. And finally, risk score calculation and risk grade evaluation are carried out on the abnormal compensation point based on the key features, and an early warning signal is generated. According to the invention, early warning information with higher pertinence is provided, and the overall safety monitoring and early warning capability is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of optical fiber sensing technology, in particular to a monitoring and early warning system based on array optical fiber sensing technology. Background Art

[0002] Fiber optic sensing technology is an advanced technology that is widely used in various fields that require real-time monitoring and early warning, such as structural health monitoring, safety monitoring, and environmental monitoring. Fiber optic sensing technology uses the characteristics of optical fiber to transmit light signals, and senses changes in environmental parameters by measuring changes in light signals during transmission in the optical fiber (such as intensity, phase, frequency, and polarization). Its main advantages include high sensitivity, anti-electromagnetic interference, corrosion resistance, and adaptability to harsh environments.

[0003] Array fiber optic sensing technology is based on fiber optic sensing technology, using multiple optical fibers or fiber optic networks to form an array, which can achieve multi-point and multi-parameter monitoring. By arranging multiple fiber optic sensors at different locations, more comprehensive and detailed monitoring information can be obtained. This technology is particularly suitable for monitoring needs in large-scale and complex environments. In terms of structural health monitoring, fiber optic sensors can monitor parameters such as stress, strain and vibration of infrastructure such as bridges, tunnels, and large buildings in real time, and discover potential safety hazards in advance. In terms of safety monitoring, fiber optic sensing technology can monitor abnormal conditions such as leakage, damage and displacement of key facilities such as oil and gas pipelines, power lines and railways, and issue early warnings to avoid accidents. The main advantages of array fiber optic sensing technology include high sensitivity and high precision, long-distance monitoring, multi-point monitoring, as well as intelligence and data analysis.

[0004] Railway transportation is a field that is highly dependent on safety and reliability. Through the array fiber optic sensing monitoring system, the status of railway tracks and structures can be monitored in real time. Fiber optic sensing technology can cover the entire length of the railway line without the need to deploy a large number of sensor devices; it can achieve comprehensive monitoring of long-distance railways, including monitoring of track displacement, temperature changes, stress and strain parameters; it can achieve multi-point simultaneous monitoring, adapt to complex monitoring needs, and can realize automated monitoring and intelligent early warning, improving the response speed and accuracy of the system.

[0005] At present, there are still some defects in the existing literature on the monitoring and early warning of fiber optic sensing technology. For example, the literature ("Demands and Challenges of Distributed Fiber Optic Sensing Technology in Coal Mine Safety Monitoring and Early Warning", "Acta Optica Sinica", Zhang Xuping et al., 2023-12-01) uses distributed fiber optic sensing technology to monitor underground structural deformation, gas leakage and temperature changes in coal mines, and evaluates its performance and reliability. For another example, the literature ("A Review of Rock and Soil Deformation Monitoring Based on Distributed Fiber Optic Sensing", "Progress in Laser and Optoelectronics", Cheng Gang et al., 2021-10-11) points out that distributed fiber optic sensing technology can realize real-time, long-distance and high-precision monitoring of rock and soil deformation, and has excellent spatial resolution and sensitivity. The above existing technologies all use fiber optic sensing technology for monitoring, but there are also problems such as the difficulty in determining the accuracy of abnormal point data and the slow real-time data transmission rate during the monitoring process, which is not conducive to timely warning of abnormal signals in the monitoring area.

[0006] Therefore, a monitoring and early warning system based on array optical fiber sensing technology is proposed. Summary of the invention

[0007] The purpose of the present invention is to provide a monitoring and early warning system based on array optical fiber sensing technology. Optical fiber sensors 2 are arranged under the rails 1 in the railway monitoring area, and the layout is in the form of an array to receive real-time data of the environment around the rails 1; the received real-time data is preprocessed, feature extracted and monitored in real time to identify abnormal points on the rails 1. The type and location of the abnormal point are determined by the characteristics of the abnormal point, the loss degree of the abnormal point is quantified, and the signal compensation is performed at the corresponding position of the optical fiber sensor 2 on the rail 1; the key features of the compensated signal are further extracted and enhanced. Based on the key features, the risk score calculation and risk level assessment of the abnormal compensation point are performed, and an early warning signal is generated.

[0008] To achieve the above object, the present invention provides the following technical solutions:

[0009] A monitoring and early warning system based on array optical fiber sensing technology, comprising:

[0010] Layout module, receiving module, data analysis module, signal compensation module, accuracy improvement module and intelligent warning module;

[0011] The layout module is used to arrange the optical fiber sensor 2 in the soil 3 below the two rails 1 of the railway and parallel to the two rails 1;

[0012] The receiving module is used to receive the real-time rail data transmitted from the monitoring point 4;

[0013] The data analysis module is used to pre-process, extract features and monitor the real-time data in real time to obtain abnormal points on the rail 1;

[0014] The signal compensation module is used to quantify the degree of loss of the abnormal point on the rail 1, perform signal compensation on the abnormal point, and obtain the abnormal compensation point;

[0015] The precision improvement module is used to extract features passing through the abnormal compensation point to obtain key features of the abnormal compensation point on the rail 1;

[0016] The intelligent early warning module performs risk score calculation and risk level assessment on the abnormal compensation point based on the key features, and generates an early warning signal;

[0017] The key features include abnormal time, abnormal location and abnormal frequency;

[0018] The risk levels include low risk, medium risk and high risk;

[0019] The warning signal includes a first warning signal, a second warning signal and a third warning signal;

[0020] The first warning signal is used to warn the high risk level; the second warning signal is used to warn the medium risk level; and the third warning signal is used to warn the low risk level.

[0021] Preferably, the layout module is used to arrange multiple groups of optical fiber sensors 2 on two rails 1 of the railway to perform multi-point and multi-parameter real-time monitoring.

[0022] Preferably, the data analysis module includes preprocessing the real-time data on the rail 1 in three steps: Butterworth low-pass filter, wavelet transform and Kalman filter to obtain processed data.

[0023] Preferably, the data analysis module includes performing feature extraction on the rail processing data through wavelet transform and Fourier transform to obtain a feature vector of the rail data, and the specific expression formula is:

[0024]

[0025] Among them, x(t) is the track processing data, D j (t) is the detail coefficient of the jth layer, A j (t) is the approximation coefficient of the jth layer, n is the number of decomposition layers, X(f) is the frequency component of the rail, f is the frequency, e -i2πft is a complex exponential function, where i is an imaginary unit, representing the rotation of the signal at frequency f;

[0026] The rail processing data is subjected to n-layer wavelet decomposition, and the obtained detail coefficients and approximation coefficients are D1, D2, ..., D n and A1,A2,...,An ; The mean, standard deviation and energy of the detail coefficients are expressed as:

[0027]

[0028] in, is the set of detail coefficient mean, standard deviation and energy of the rail processing data, is the mean value of detail coefficient, is the standard deviation of the detail coefficient, is the energy of the detail coefficient;

[0029] The mean, standard deviation and energy of the approximation coefficients of the rail processing data are expressed as:

[0030]

[0031] in, is the set of the mean, standard deviation and energy of the approximation coefficients of the rail processing data, is the mean of the approximation coefficients, is the standard deviation of the approximation coefficient, is the energy of the approximation coefficient;

[0032] The amplitude and phase of the feature extraction of the rail processing data after the Fourier transform are expressed as:

[0033] F f =[X(f),∠X(f)];

[0034] Among them, F f is the set of amplitude and phase extracted through Fourier transform, |X(f)| is the amplitude, ∠X(f) is the phase;

[0035] The characteristics of the rail processing data are:

[0036]

[0037] Wherein, F is the feature vector of the rail processing data.

[0038] Preferably, the data analysis module includes establishing a monitoring model to analyze and monitor the characteristic vector of the rail processing data to identify the abnormal points on the rail 1; the monitoring model is a Gaussian mixture model.

[0039] Preferably, the data analysis module includes comparing the identified abnormal point on the rail 1 with historical data to determine whether it is a real abnormal point. The specific formula is:

[0040]

[0041] Among them, S is the similarity between the abnormal point on track 1 and the historical abnormal point, F cur,i is the set of abnormal points on rail 1, F his,i is a set of historical outlier points;

[0042] The smaller S is, the greater the probability that the abnormal point on the rail 1 is a real abnormal point.

[0043] Preferably, the signal compensation module uses a regression model to quantize the linear compensation model for the characteristic vector of the abnormal point on the rail 1. The specific calculation method is:

[0044] L i =f(F cur );

[0045] Among them, L i is the degree of loss, f is the mapping function of the regression model, F cur is the abnormal point set on rail 1;

[0046] According to the quantitative loss level L i And the linear compensation model is used to calculate the compensation coefficient. The specific model is:

[0047] C i =αL i ;

[0048] Among them, C i is the compensation coefficient of the abnormal point on the rail 1, and α is the parameter of the linear compensation model;

[0049] The signal of the abnormal point on the rail 1 is compensated using the compensation coefficient C i The signal of each abnormal point is restored; the signal of the abnormal compensation point on the rail 1 is specifically:

[0050] x comp (t) = x(t)·C(t)+x(t);

[0051] Among them, x comp (t) is the signal of the abnormal compensation point on the rail 1, and C(t) is the time-dependent compensation coefficient.

[0052] Preferably, the accuracy improvement module further extracts key features based on the abnormal compensation points on the rail 1 through wavelet transform and Fourier transform.

[0053] Preferably, the intelligent early warning module obtains the information of the abnormal compensation point on the rail 1 based on the key features of the abnormal compensation point on the rail 1; the information includes abnormal time, abnormal position and abnormal frequency;

[0054] The risk score is calculated based on the information, and the specific calculation formula is:

[0055]

[0056] Where R is the risk score, ω i is the weight of each factor, f i is the evaluation value of each factor of information.

[0057] Preferably, the intelligent early warning module further includes dividing the risk level into three levels according to the risk score R, specifically including:

[0058] Low risk level: R<T1;

[0059] Medium risk level: T1≤R≤T2;

[0060] High risk level: R>T2;

[0061] Among them, T1 and T2 are quantiles obtained by analyzing the distribution of each risk level in historical data, with values ​​of 25% and 75%;

[0062] Generate a corresponding early warning signal according to the risk score and the risk level;

[0063] The warning signal includes a first warning signal, a second warning signal and a third warning signal;

[0064] The first warning signal is used to warn the high risk level; the second warning signal is used to warn the medium risk level; the third warning signal is used to warn the low risk level;

[0065] The warning information will be recorded in the log for easy subsequent analysis and tracking, including warning time, risk level, and details of abnormal events.

[0066] Compared with the prior art, the present invention has the following beneficial effects:

[0067] 1. The present invention arranges multiple optical fiber sensors 2 on the monitored object to form an array to achieve real-time monitoring of multiple points and multiple parameters; in the data analysis process, after the Gaussian mixture model is used to initially identify the abnormal points, historical data is further used for verification to ensure the authenticity and accuracy of the abnormal points. By identifying and confirming the abnormal points multiple times, the robustness and detection accuracy of the system in complex environments are significantly improved, the false alarm rate and missed alarm rate are reduced, and the reliability of monitoring is enhanced.

[0068] 2. The present invention analyzes the preprocessed data and establishes a mathematical model of signal loss using a regression model. It realizes quantitative analysis of the identified abnormal points, and uses a linear compensation model to calculate the compensation coefficient to compensate the abnormal point signal. Through the signal compensation module, the noise and loss in the transmission process of the abnormal point signal are reduced, the overall quality of the monitoring data is improved, and a clearer and more reliable data basis is provided for real-time monitoring.

[0069] 3. The present invention proposes a type of intelligent early warning system that uses anomaly detection algorithms to analyze data in real time and identify potential abnormal situations. Based on factors such as abnormal time, location, and frequency, the risk score is calculated, risk assessment is performed, and the risk level is determined. By setting a reasonable risk threshold, different levels of risks are distinguished, and corresponding early warning signals are issued, the response speed and accuracy of the early warning system are effectively improved. By refining the risk level and the corresponding early warning form, the system can provide more targeted early warning information and improve the overall safety monitoring and early warning capabilities. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] Figure 1 A schematic diagram of an implementation process of a monitoring and early warning system based on array optical fiber sensing technology provided by an embodiment of the present invention;

[0071] Figure 2 A schematic diagram of the structure of a monitoring and early warning system based on array optical fiber sensing technology provided by an embodiment of the present invention;

[0072] Figure 3 A schematic diagram of the locations of monitoring points for the optical fiber sensing technology provided in an embodiment of the present invention.

[0073] In the figure: 1. Railroad track; 2. Fiber optic sensor; 3. Soil; 4. Monitoring point. DETAILED DESCRIPTION

[0074] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0075] The present invention provides a monitoring and early warning system based on array optical fiber sensing technology. This method is applied to a monitoring and early warning system based on array optical fiber sensing technology. The flowchart of the specific method and system is shown in FIG. Figure 1 and Figure 2 .

[0076] Embodiment 1

[0077] As an embodiment of the present invention, refer to Figure 1 In this embodiment, the optical fiber sensor 2 is laid parallel to the two rails 1 on the rails 1 in the railway area A to form an array for real-time monitoring of multiple points and multiple parameters; by receiving the real-time rail data transmitted from the monitoring point 4;

[0078] Furthermore, the optical fiber sensor array is arranged in the railway area A to accurately sense the changes in environmental parameters and obtain the real-time data on the rail 1 through the monitoring point 4. Figure 1 S10 and Figure 3 Optical fibers are arranged below two rails 1, and the optical fibers are in the soil 3 below the rails 1, and monitoring points 4 are evenly arranged on the optical fibers.

[0079] The real-time data is shown in the following table:

[0080] Table 1 Real-time data example table

[0081] Sensor location time Real-time data Sensor1 08:00:00 250 Sensor2 08:00:01 255 Sensor3 08:00:02 260 Sensor4 08:00:03 245 Sensor5 08:00:04 248 Sensor6 08:00:05 253 Sensor7 08:00:06 258 Sensor8 08:00:07 242 Sensor9 08:00:08 247 Sensor10 08:00:09 251

[0082] In this embodiment, by arranging an array of optical fiber sensors 2 on the rails 1, real-time monitoring of changes in environmental parameters around the rails 1 is achieved, providing high-precision data support to ensure the timeliness and accuracy of monitoring. The environmental parameters include temperature, humidity, pressure, train operation, and intrusion events.

[0083] Reference Figure 1 S20 in the figure is applied to a data analysis module in a monitoring and early warning system based on array optical fiber sensing technology, and is used to preprocess, extract features and monitor in real time the real-time data to obtain abnormal points on the rail 1.

[0084] Further, acquiring the real-time data from the optical fiber sensor array and preprocessing the real-time data;

[0085] The real-time data obtained from the optical fiber sensor 2 is first processed by a Butterworth low-pass filter; the specific formula of the real-time Butterworth low-pass filter is expressed as:

[0086] y pass =filtfilt(b,a,y(t));

[0087] Among them, y pass is the real-time rail data processed by the Butterworth low-pass filter, filtfilt() is a type of function used to eliminate the phase delay introduced by signal filtering, b, a are filter coefficients, and y(t) is the real-time rail data;

[0088] The filter coefficients b, a can be obtained by the following formula:

[0089] b,a=butter(N,W n ,btype='low');

[0090] Where N is the filter order, butter() is a function used to design a Butterworth low-pass filter, and W n is the normalized cutoff frequency;

[0091] Normalized cutoff frequency W n It is calculated by:

[0092]

[0093] Among them, f c is the cut-off frequency, f s is the sampling rate.

[0094] The rail data y processed by the Butterworth low-pass filter pass Performing wavelet transform processing to capture the time-frequency characteristics and multi-scale information of the signal on the rail 1;

[0095] Table 2 Preprocessed data table

[0096] Sensor location time Preprocessed data Sensor1 08:00:00 251 Sensor2 08:00:01 253 Sensor3 08:00:02 254 Sensor4 08:00:03 248 Sensor5 08:00:04 249 Sensor6 08:00:05 251 Sensor7 08:00:06 253 Sensor8 08:00:07 247 Sensor9 08:00:08 248 Sensor10 08:00:09 250

[0097] y pass The output of wavelet decomposition is expressed as:

[0098] cA,cD=wavedec(y pass (t),wavelet,level=T);

[0099] Among them, cA is the approximation coefficient, wavedec() is the wavelet decomposition function, and cD is the detail coefficient.

[0100] The rail data y processed by the Butterworth low-pass filter pass The approximate coefficient cA after wavelet transformation is applied to the Kalman filter for state estimation and data optimization, and the output data is x(t).

[0101] The data after preprocessing by Butterworth low-pass filter, wavelet transform and Kalman filter refer to Table 2;

[0102] In this embodiment, the data quality and applicability obtained by the optical fiber sensor 2 on the rail 1 are improved through the preprocessing of the three parts of Butterworth low-pass filter, wavelet transform and Kalman filter. Before preprocessing, the optical fiber sensor 2 collects real-time data on the rail 1. First, the Butterworth low-pass filter effectively eliminates the high-frequency noise in the data, making the data smooth and stable. Secondly, the wavelet transform enhances the ability to extract data features by capturing the time-frequency characteristics of the data. Finally, after these preprocessing steps, the Kalman filter further eliminates possible measurement errors and system noise by estimating and optimizing the state of the data, ensuring the accuracy and stability of the data. Through these three preprocessing steps, the data quality of the optical fiber sensor 2 on the rail 1 is significantly improved, meeting the needs of high-precision monitoring and analysis.

[0103] The pre-processed data on the rail 1 is subjected to feature extraction through wavelet transform and Fourier transform, and the specific expression formula is:

[0104]

[0105] Wherein, x(t) is the processed data, D j (t) is the detail coefficient of the jth layer, A j (t) is the approximation coefficient of the jth layer, n is the number of decomposition layers, X(f) is the frequency component, f is the frequency, e -i2πft is a complex exponential function, where i is an imaginary unit, representing the rotation of the signal at frequency f;

[0106] Where J is calculated by log2x(t), |x(t)| is the length of the data;

[0107] The rail processing data is subjected to n-layer wavelet decomposition, and the obtained detail coefficients and approximation coefficients are D1, D2, ..., D n and A1,A2,...,A n ; The mean, standard deviation and energy of the detail coefficients are expressed as:

[0108]

[0109] in, is the set of detail coefficient mean, standard deviation and energy of the rail processing data, is the mean value of detail coefficient, is the standard deviation of the detail coefficient, is the energy of the detail coefficient;

[0110] The mean, standard deviation and energy of the approximation coefficients of the rail processing data are expressed as:

[0111]

[0112] in, is the set of the mean, standard deviation and energy of the approximation coefficients of the rail processing data, is the mean of the approximation coefficients, is the standard deviation of the approximation coefficient, is the energy of the approximation coefficient;

[0113] The specific calculation formulas for the mean, standard deviation and energy of the detail coefficient and approximation coefficient are:

[0114]

[0115]

[0116] The Fourier transformed features of the rail processing data can be used to extract amplitude and phase, which can be specifically expressed as:

[0117] F f =[X(f),∠X(f)];

[0118] Among them, F f is the set of amplitude and phase extracted through Fourier transform, |X(f)| is the amplitude, ∠X(f) is the phase;

[0119] Among them, the specific expression of the amplitude |X(f)| is:

[0120]

[0121] Where Re is the real part of X(f), Im is the imaginary part of X(f);

[0122] The feature vector of the rail processing data is:

[0123]

[0124] Wherein, F is the feature vector of the rail processing data, as shown in Table 3;

[0125] Table 3 Extracted feature vector data table

[0126] Sensor location <![CDATA[μ A ]]> <![CDATA[σ A ]]> <![CDATA[E A ]]> <![CDATA[μ D ]]> <![CDATA[σ D ]]> <![CDATA[E D ]]> X(f) ∠X(f) Sensor1 102.5 10.5 43500 0.5 2.5 35 0.3 0.8 Sensor2 112.0 9.8 45000 -0.7 3.2 40 0.4 0.9 Sensor3 122.0 11.0 47000 0.9 2.8 38 0.5 0.7 Sensor4 88.0 8.5 40000 -0.5 2.0 30 0.3 0.6 Sensor5 94.5 7.2 41000 0.3 1.5 28 0.4 0.5 Sensor6 106.0 9.0 44500 -0.8 3.0 37 0.2 0.4 Sensor7 116.5 10.2 46000 0.6 2.6 39 0.1 0.3 Sensor8 84.5 7.5 39500 -0.3 1.8 32 0.2 0.2 Sensor9 86.5 8.0 40050 0.4 1.9 33 0.3 0.1 Sensor10 91.0 8.7 41500 0.1 2.3 36 0.5 0.0

[0127] In this embodiment, the signal features can be extracted by performing wavelet transform and Fourier transform on the rail processing data. The multi-scale analysis and time-frequency localization capabilities provided by wavelet transform, and the frequency domain analysis and periodic feature extraction capabilities provided by Fourier transform, make feature extraction more comprehensive and accurate. This combined method can not only improve the sensitivity of rail 1 anomaly detection, but also enhance the flexibility of signal analysis, laying a solid foundation for subsequent signal processing and analysis.

[0128] Use the data from the historical dataset to train the Gaussian mixture model;

[0129] The comprehensive feature vector F is used to calculate the probability density of each data point through the Gaussian mixture model, and the outliers are output; the output outlier point set is [Sensor1, Sensor4].

[0130] The extracted features are used to extract anomalies through a Gaussian mixture model, which can significantly improve the accuracy and robustness of anomaly detection on rail 1. By inputting the extracted features into a Gaussian mixture model, the probability distribution of the data can be modeled, thereby more accurately capturing the potential patterns and anomalies of the data and improving the overall effect of data analysis.

[0131] In this embodiment, the abnormal points of the rail 1 include excavator construction, manual excavation, cracks and fractures, track misalignment, corrugation, loose bolts, track sinking, and vehicle driving.

[0132] Furthermore, by comparing the identified abnormal point with the historical data, it is determined whether it is the real abnormal point. The specific formula is:

[0133]

[0134] Among them, d(F cur ,F his ) is the Euclidean distance between the abnormal point on rail 1 and the historical abnormal point, S is the similarity between the abnormal point on rail 1 and the historical abnormal point, F cur is the set of abnormal points on rail 1, F his is a set of historical outlier points;

[0135] The smaller S is, the greater the probability that the abnormal point is a real abnormal point;

[0136] By using the data from different optical fiber sensors 2 and comparing the results from different data sources, it is confirmed that the abnormal point is a real abnormal point. The specific formula is:

[0137]

[0138] Among them, F com is the average value of multiple data sources at the same abnormal point on rail 1, is the data from multiple data sources for the same abnormal point on the rail 1; if F cur With F com The small difference indicates that the abnormal point is a real abnormal point.

[0139] In this embodiment, the abnormal point data [Sensor1, Sensor4] on rail 1 is compared with the historical data and different data sources are confirmed; the results show that [Sensor1, Sensor4] is a real abnormal point.

[0140] By comparing with historical data and multi-source data to confirm abnormal points, more accurate anomaly detection and more comprehensive fault diagnosis can be achieved, improving the overall reliability and practicality of the monitoring system.

[0141] Reference Figure 1 S30 in the embodiment is applied to a signal compensation module in a monitoring and early warning system based on array optical fiber sensing technology, and is used to quantify the degree of loss of the abnormal point on the rail 1, perform signal compensation on the abnormal point, and obtain the abnormal compensation point.

[0142] Furthermore, through the above-mentioned abnormal point, a support vector machine is used to identify whether the abnormal point is signal attenuation, noise interference or other types of loss;

[0143] The regression model is used for the characteristic vector of the abnormal point on the rail 1 to quantify the linear compensation model. The specific calculation method is as follows:

[0144] L i =f(F cur );

[0145] Among them, L i is the degree of loss, f is the mapping function of the regression model;

[0146] According to the quantitative loss level L i And the linear compensation model is used to calculate the compensation coefficient. The specific model is:

[0147] C i =αL i +β;

[0148] Among them, C i is the compensation coefficient, α and β are parameters determined by experiments or calibration data;

[0149] The signal of the abnormal point on the rail 1 is compensated using the compensation coefficient C i The signal of each abnormal point is restored; the signal of the abnormal compensation point on the rail 1 is specifically:

[0150] x comp (t) = x(t)·C(t)+x(t);

[0151] Among them, x comp (t) is the signal of the abnormal compensation point on the rail 1, and C(t) is the time-dependent compensation coefficient.

[0152] In this embodiment, signal compensation is performed on the abnormal point [Sensor1, Sensor4] on the rail 1 identified in the previous step, and the signal of the abnormal point displayed by the optical fiber sensor 2 at the corresponding position on the rail 1 is compensated by calculating the loss degree and the compensation coefficient. In this way, the influence of the signal attenuation and noise interference of the optical fiber sensor 2 on the monitoring system can be significantly reduced, and the monitoring accuracy and early warning capability of the system for the rail 1 can be enhanced.

[0153] Reference Figure 1 S40 in the embodiment is applied to an accuracy improvement module in a monitoring and early warning system based on array optical fiber sensing technology, and is used to extract features of the abnormal compensation points passing through the abnormal compensation points to obtain key features of the abnormal compensation points on the rail 1.

[0154] Furthermore, the accuracy improvement module further extracts key features based on the abnormal compensation points on the rail 1 through wavelet transform and Fourier transform.

[0155] By extracting key features from abnormal points on the rail 1 after compensating the signal, it can be ensured that the features extracted from the abnormal points are more accurate and representative, thereby improving the effectiveness of subsequent data analysis and processing.

[0156] Reference Figure 1 S50 in the embodiment, S50 is applied to an intelligent early warning module in a monitoring and early warning system based on array optical fiber sensing technology, which performs risk score calculation and risk level assessment on the abnormal compensation point based on the key features, and generates an early warning signal.

[0157] Furthermore, the further extracted features are used to perform the abnormal point detection through the Gaussian mixture model, and based on the key features of the abnormal compensation points on the rail 1, the abnormal compensation point information on the rail 1 is obtained; the information includes abnormal time, abnormal position and abnormal frequency;

[0158] The risk score is calculated based on the information, and the specific calculation formula is:

[0159]

[0160] Where R is the risk score, ω i is the weight of each factor, f i is the evaluation value of each factor of information.

[0161] In this embodiment, the weights of time, location and frequency are proposed to be 0.378, 0.365 and 0.257 respectively; the evaluation values ​​of time, location and frequency of the abnormal point Sensor1 are 10, 1 and 5 respectively; the evaluation values ​​of time, location and frequency of the abnormal point Sensor2 are 6, 1 and 4 respectively; then the calculated risk score is R Sensor1=5.43 and R Sensor4 =3.661.

[0162] The calculation of the risk score for abnormal point detection on Rail 1 involves comprehensive consideration of multiple factors to reflect their importance in the overall risk. Risk scores are used to help predict and prevent potential risks and improve the scientificity and efficiency of decision-making. By calculating the risk score for each abnormal point, high-risk areas can be identified more accurately, and maintenance and repair can be prioritized to ensure the safe operation and stability of the railway system. This risk-scoring-based approach not only improves the sensitivity of detection, but also enhances the pertinence and effectiveness of Rail 1 maintenance.

[0163] According to the risk score R, the risk level is divided into multiple levels, including:

[0164] Low risk level: R<T1;

[0165] Medium risk level: T1≤R≤T2;

[0166] High risk level: R>T2;

[0167] Among them, T1 and T2 are quantiles obtained by analyzing the distribution of each risk level in historical data, with values ​​of 25% and 75%;

[0168] Generate corresponding early warning signals according to the risk score and the risk level; including: high-risk events, emergency early warning signals; medium-risk events, warning signals; low-risk events, reminder signals;

[0169] The warning information will be recorded in the log for easy subsequent analysis and tracking, including warning time, risk level, and details of abnormal events.

[0170] By mapping the numerical risk scores to predefined risk levels, this classification method not only helps to prioritize high-risk rail1 anomalies, but also effectively allocate resources and formulate corresponding risk response strategies. The generated warning signals are classified according to the risk level, specifically in the form of emergency warnings, warnings and reminders. This differentiated early warning system ensures timely action at critical moments to minimize possible losses or impacts. All warning information is recorded in detail in the log, which not only helps post-analysis and tracking, but also supports continuous improvement of risk management.

[0171] In this embodiment, the data receiving and preprocessing stage ensures the quality and accuracy of the raw data obtained from the rail 1 monitoring object. By arranging multiple optical fiber sensors 2 and effective data acquisition equipment, the system can obtain physical quantity data on the rail 1 in real time and at multiple points. The data preprocessing stage further improves the quality of the data through denoising and error correction processing, reduces the impact of noise on subsequent analysis, and ensures the reliability and accuracy of the data. The data processing and analysis stage uses advanced mathematical models and neural network technology to extract the key spatial and temporal features of the abnormal points of the rail 1 from complex data, and identifies and responds to potential abnormal situations in real time, thereby improving the monitoring system's perception and response capabilities to various complex situations. The signal loss compensation stage effectively improves the quality and stability of the sensor signal by establishing a signal loss model and applying a deep learning model to predict and compensate for signal attenuation. The intelligent early warning stage can detect and identify potential abnormal situations of the rail 1 in a timely manner and quickly generate early warning information through trained monitoring models and real-time data analysis. This improves the system's response speed and accuracy to abnormal events and effectively reduces the risk of railway accidents and losses.

[0172] Embodiment 2

[0173] As an embodiment of the present invention, refer to Figure 1 In this embodiment, the optical fiber sensor 2 is laid on the rails 1 in the railway area B in parallel with the two rails 1 to form an array for real-time monitoring of multiple points and multiple parameters; by receiving the real-time rail data transmitted from the monitoring point 4;

[0174] Furthermore, the optical fiber sensor array is arranged in the railway area B to accurately sense the changes in environmental parameters and obtain the real-time data on the rail 1 through the monitoring point 4. Figure 1 S10 and Figure 3 .

[0175] The real-time data is shown in the following table:

[0176] Table 4 Real-time data example table

[0177] Sensor location time Real-time data Sensor1 09:00:00 120 Sensor2 09:00:01 115 Sensor3 09:00:02 130 Sensor4 09:00:03 140 Sensor5 09:00:04 125 Sensor6 08:00:05 135 Sensor7 09:00:06 118 Sensor8 09:00:07 128 Sensor9 09:00:08 138 Sensor10 09:00:09 123

[0178] Reference Figure 1 S20 in the figure is applied to a data analysis module in a monitoring and early warning system based on array optical fiber sensing technology, and is used to preprocess, extract features and monitor in real time the real-time data to obtain abnormal points on the rail 1.

[0179] Due to changes in the accurate environmental parameters of fiber optic sensing technology, such as human factors causing signal interference;

[0180] Further, acquiring the real-time data from the optical fiber sensor array and preprocessing the real-time data;

[0181] The pre-processed data on the rail 1 is subjected to feature extraction through wavelet transform and Fourier transform, and the specific expression formula is:

[0182]

[0183] Wherein, x(t) is the processed data, D j (t) is the detail coefficient of the jth layer, A j (t) is the approximation coefficient of the jth layer, n is the number of decomposition layers, X(f) is the frequency component, f is the frequency, e -i2 π ft is a complex exponential function, where i is an imaginary unit, representing the rotation of the signal at frequency f;

[0184] The rail processing data is subjected to n-layer wavelet decomposition, and the obtained detail coefficients and approximation coefficients are D1, D2, ..., D n and A1,A2,...,A n ; The mean, standard deviation and energy of the detail coefficients are expressed as:

[0185]

[0186] in, is the set of detail coefficient mean, standard deviation and energy of the rail processing data, is the mean value of detail coefficient, is the standard deviation of the detail coefficient, is the energy of the detail coefficient;

[0187] The mean, standard deviation and energy of the approximation coefficients of the rail processing data are expressed as:

[0188]

[0189] in, is the set of the mean, standard deviation and energy of the approximation coefficients of the rail processing data, is the mean of the approximation coefficients, is the standard deviation of the approximation coefficient, is the energy of the approximation coefficient;

[0190] The Fourier transformed features of the rail processing data can be used to extract amplitude and phase, which can be specifically expressed as:

[0191] F f =[X(f),∠X(f)];

[0192] Among them, Ff is the set of amplitude and phase extracted through Fourier transform, |X(f)| is the amplitude, ∠X(f) is the phase;

[0193] Among them, the specific expression of the amplitude |X(f)| is:

[0194]

[0195] Where Re is the real part of X(f), Im is the imaginary part of X(f);

[0196] The feature vector of the rail processing data is:

[0197]

[0198] Wherein, F is the feature vector of the rail processing data, as shown in Table 5;

[0199] The comprehensive feature vector F is used to calculate the probability density of each data point through the Gaussian mixture model, and the outliers are output; the output outlier point set is an empty set, with no outliers.

[0200] Furthermore, by comparing the identified abnormal point with the historical data, it is determined whether it is the real abnormal point. The specific formula is:

[0201]

[0202] Among them, d(F cur ,F his ) is the Euclidean distance between the abnormal point on rail 1 and the historical abnormal point, S is the similarity between the abnormal point on rail 1 and the historical abnormal point, F cur is the set of abnormal points on rail 1, F his is a set of historical outlier points;

[0203] The smaller S is, the greater the probability that the abnormal point is a real abnormal point;

[0204] Table 5 Extracted feature vector data table

[0205] Sensor location <![CDATA[μ A ]]> <![CDATA[σ A ]]> <![CDATA[E A ]]> <![CDATA[μ D ]]> <![CDATA[σ D ]]> <![CDATA[E D ]]> X(f) ∠X(f) Sensor1 2.1 0.1 13689 117.0 0.5 4.41 0.3 0.8 Sensor2 1.8 0.2 12544 112.0 0.4 3.24 0.4 0.9 Sensor3 2.5 0.3 15625 125.0 0.3 6.25 0.5 0.7 Sensor4 2.8 0.4 18225 135.0 0.2 7.84 0.3 0.6 Sensor5 2.0 0.1 14884 122.0 0.3 4.00 0.4 0.5 Sensor6 2.6 0.3 17161 131.0 0.2 6.76 0.2 0.4 Sensor7 1.9 0.1 13225 115.0 0.4 3.61 0.1 0.3 Sensor8 2.3 0.2 15376 124.0 0.3 5.29 0.2 0.2 Sensor9 2.4 0.2 17956 134.0 0.4 5.76 0.3 0.1 Sensor10 2.2 0.3 14400 120.0 0.5 4.84 0.5 0.05

[0206] By using the data from different optical fiber sensors 2 and comparing the results from different data sources, it is confirmed that the abnormal point is a real abnormal point. The specific formula is:

[0207]

[0208] Among them, F com is the average value of multiple data sources at the same abnormal point on rail 1, is the data from multiple data sources for the same abnormal point on the rail 1; if F cur With F com The small difference indicates that the abnormal point is a real abnormal point.

[0209] Reference Figure 1 S30 in the embodiment is applied to a signal compensation module in a monitoring and early warning system based on array optical fiber sensing technology, and is used to quantify the degree of loss of the abnormal point on the rail 1, perform signal compensation on the abnormal point, and obtain the abnormal compensation point.

[0210] Furthermore, through the above-mentioned abnormal point, a support vector machine is used to identify whether the abnormal point is signal attenuation, noise interference or other types of loss;

[0211] The regression model is used for the characteristic vector of the abnormal point on the rail 1 to quantify the linear compensation model. The specific calculation method is as follows:

[0212] L i =f(F cur );

[0213] Among them, L i is the degree of loss, f is the mapping function of the regression model;

[0214] According to the quantitative loss level L i And the linear compensation model is used to calculate the compensation coefficient. The specific model is:

[0215] C i =αL i +β;

[0216] Among them, C i is the compensation coefficient, α and β are parameters determined by experiments or calibration data;

[0217] The signal of the abnormal point on the rail 1 is compensated using the compensation coefficient C i The signal of each abnormal point is restored; the signal of the abnormal compensation point on the rail 1 is specifically:

[0218] x comp (t) = x(t)·C(t)+x(t);

[0219] Among them, x comp (t) is the signal of the abnormal compensation point on the rail 1, and C(t) is the time-dependent compensation coefficient.

[0220] Reference Figure 1 S40 in the embodiment is applied to an accuracy improvement module in a monitoring and early warning system based on array optical fiber sensing technology, and is used to extract features of the abnormal compensation points passing through the abnormal compensation points to obtain key features of the abnormal compensation points on the rail 1.

[0221] Furthermore, the accuracy improvement module further extracts key features based on the abnormal compensation points on the rail 1 through wavelet transform and Fourier transform.

[0222] Reference Figure 1 S50 in the embodiment is applied to an intelligent early warning module in a monitoring and early warning system based on array optical fiber sensing technology.

[0223] Reference Figure 1 S50 in the embodiment, S50 is applied to an intelligent early warning module in a monitoring and early warning system based on array optical fiber sensing technology, which performs risk score calculation and risk level assessment on the abnormal compensation point based on the key features, and generates an early warning signal.

[0224] Furthermore, the further extracted features are used to perform the abnormal point detection through the Gaussian mixture model, and based on the key features of the abnormal compensation points on the rail 1, the abnormal compensation point information on the rail 1 is obtained; the information includes abnormal time, abnormal position and abnormal frequency;

[0225] The risk score is calculated based on the information, and the specific calculation formula is:

[0226]

[0227] Where R is the risk score, ω i is the weight of each factor, f i is the evaluation value of each factor of information.

[0228] According to the risk score R, the risk level is divided into multiple levels, including:

[0229] Low risk level: R<T1;

[0230] Medium risk level: T1≤R≤T2;

[0231] High risk level: R>T2;

[0232] Among them, T1 and T2 are quantiles obtained by analyzing the distribution of each risk level in historical data, with values ​​of 25% and 75%;

[0233] Generate corresponding early warning signals according to the risk score and the risk level; including: high-risk events, emergency early warning signals; medium-risk events, warning signals; low-risk events, reminder signals;

[0234] The warning information will be recorded in the log for easy subsequent analysis and tracking, including warning time, risk level, and details of abnormal events.

[0235] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A monitoring and early warning system based on array optical fiber sensing technology, characterized in that: Including layout module, receiving module, data analysis module, signal compensation module, accuracy improvement module and intelligent warning module; The layout module is used to arrange the optical fiber sensors (2) in the soil (3) below two rails (1) of the railway, and in parallel with the two rails (1), in an array form; The receiving module is used to receive the real-time rail data transmitted from the monitoring point (4); The data analysis module is used to pre-process, extract features and perform real-time monitoring on the real-time data to obtain abnormal points on the rail (1); The signal compensation module is used to quantify the degree of loss of the abnormal point on the rail (1) through a regression model, perform signal compensation on the abnormal point, and obtain an abnormal compensation point; The accuracy improvement module is used to extract features of the abnormal compensation points to obtain key features of the abnormal compensation points on the rail (1); The intelligent early warning module is used to calculate the risk score and assess the risk level according to the key features, and generate an early warning signal.

2. A monitoring and early warning system based on array optical fiber sensing technology according to claim 1, characterized in that: The layout module is used to arrange multiple groups of optical fiber sensors (2) on two rails (1) of a railway to perform real-time monitoring of multiple points and multiple parameters.

3. The monitoring and early warning system based on array optical fiber sensing technology according to claim 1 is characterized in that: The data analysis module comprises preprocessing the real-time data on the rail (1) in three steps: Butterworth low-pass filter, wavelet transform and Kalman filter, to obtain processed data.

4. The monitoring and early warning system based on array optical fiber sensing technology according to claim 3 is characterized in that: The data analysis module includes performing feature extraction on the rail processing data through wavelet transform and Fourier transform to obtain the feature vector of the rail data. The specific expression formula is: Among them, x(t) is the track processing data, D j (t) is the detail coefficient of the jth layer, A j (t) is the approximation coefficient of the jth layer, n is the number of decomposition layers, X(f) is the rail frequency component, f is the frequency, e -i2 π ft is a complex exponential function, where i is an imaginary unit, representing the rotation of the signal at frequency f; The rail processing data is subjected to n-layer wavelet decomposition, and the obtained detail coefficients and approximation coefficients are D1, D2, ..., D n and A1,A2,...,A n ; The mean, standard deviation and energy of the detail coefficients are expressed as: in, is the set of detail coefficient mean, standard deviation and energy of the rail processing data, is the mean value of detail coefficient, is the standard deviation of the detail coefficient, is the energy of the detail coefficient; The mean, standard deviation and energy of the approximation coefficients of the rail processing data are expressed as: in, is the set of the mean, standard deviation and energy of the approximation coefficients of the rail processing data, is the mean of the approximation coefficients, is the standard deviation of the approximation coefficient, is the energy of the approximation coefficient; The amplitude and phase of the feature extraction of the rail processing data after the Fourier transform are expressed as: F f =[|X(f),∠X(f)]; Among them, F f is the set of amplitude and phase extracted through Fourier transform, |X(f)| is the amplitude, ∠X(f) is the phase; The characteristics of the rail processing data are: Wherein, F is the feature vector of the rail processing data.

5. The monitoring and early warning system based on array optical fiber sensing technology according to claim 4 is characterized in that: The data analysis module includes establishing a monitoring model, processing data feature vectors according to the rail, performing analysis and monitoring, and identifying abnormal points on the rail (1); the monitoring model is a Gaussian mixture model.

6. The monitoring and early warning system based on array optical fiber sensing technology according to claim 5 is characterized in that: The data analysis module includes comparing the identified abnormal point on the rail (1) with historical data to determine whether it is a real abnormal point. The specific formula is: Where S is the similarity between the abnormal point on the track (1) and the historical abnormal point, and F cur,i is the set of abnormal points on the rail (1), F his,i is a set of historical outlier points; The smaller S is, the greater the probability that the abnormal point on the rail (1) is a real abnormal point.

7. The monitoring and early warning system based on array optical fiber sensing technology according to claim 1 is characterized in that: The signal compensation module uses a regression model to quantify the linear compensation model based on the characteristic vector of the abnormal point on the rail (1). The specific calculation method is as follows: L i =f(F cur ); Among them, L i is the degree of loss, f is the mapping function of the regression model, F cur is the set of abnormal points on the rail (1); According to the quantitative loss level L i And the linear compensation model is used to calculate the compensation coefficient. The specific model is: C i =αL i ; Among them, C i is the compensation coefficient of the abnormal point on the rail (1), and α is the parameter of the linear compensation model; By compensating the signal of the abnormal point on the rail (1), the compensation coefficient C i The signal of each abnormal point is restored; the signal of the abnormal compensation point on the rail (1) is specifically: x comp (t)=x(t)·C(t)+x(t); Among them, x comp (t) is the signal of the abnormal compensation point on the rail (1), and C(t) is the time-dependent compensation coefficient.

8. The monitoring and early warning system based on array optical fiber sensing technology according to claim 1 is characterized in that: The accuracy improvement module further extracts key features through wavelet transform and Fourier transform based on the abnormal compensation points on the rail (1).

9. The monitoring and early warning system based on array optical fiber sensing technology according to claim 1, characterized in that: The intelligent early warning module obtains information of the abnormal compensation point on the rail (1) based on the key features of the abnormal compensation point on the rail (1); the information includes abnormal time, abnormal position and abnormal frequency; The risk score is calculated based on the information, and the specific calculation formula is: Where R is the risk score, ω i is the weight of each factor, f i is the evaluation value of each factor of information.

10. The monitoring and early warning system based on array optical fiber sensing technology according to claim 1, characterized in that: The intelligent early warning module further includes dividing the risk level into three levels according to the risk score R, specifically including: Low risk level: R<T1; Medium risk level: T1≤R≤T2; High risk level: R>T2; Among them, T1 and T2 are the distribution quantiles of each risk level in the historical data. and A corresponding warning signal is generated according to the risk score and the risk level; the warning signal includes a first warning signal, a second warning signal and a third warning signal; the first warning signal is used to warn the high risk level; the second warning signal is used to warn the medium risk level; the third warning signal is used to warn the low risk level; and the warning information is recorded in a log for subsequent analysis and tracking; including warning time, risk level, and details of abnormal events.

Citation Information

Cited By

  • Aluminum bar casting safety monitoring method, monitoring system and readable storage medium

    CN119387525A