CN-122016035-A - Intelligent detection monitoring method and system for subway

CN122016035ACN 122016035 ACN122016035 ACN 122016035ACN-122016035-A

Abstract

The invention discloses an intelligent detection monitoring method and system for subways, and belongs to the technical field of detection; the method comprises the steps of obtaining a sensor signal, preprocessing the sensor signal to obtain a preprocessed sensor signal, extracting characteristics of the preprocessed sensor signal to obtain a characteristic value, comparing the characteristic value with a double-layer threshold value to obtain a diagnosis result, and compressing and transmitting the diagnosis result. The invention provides an intelligent vibration noise sensor for complex environments such as subways and a data processing method thereof, which realizes high-fidelity acquisition, intelligent edge-end processing and low-bandwidth reliable transmission of vibration noise data through deep fusion of a multi-mode sensing architecture and a high-efficiency data compression transmission technology.

Inventors

WANG YUHAO
CHEN YINGCHUN
HU NAIXIANG
YAO GUODONG
FU DONGYU
DONG XIANGYU
CUI PENGTAO
YANG JIE

Assignees

北京工业大学

Dates

Publication Date: 20260512
Application Date: 20260228

Claims (10)

1. The subway-oriented intelligent detection and monitoring method is characterized by comprising the following steps of: acquiring a sensor signal; Preprocessing the sensor signal to obtain a preprocessed sensor signal; extracting characteristics of the sensor signals after pretreatment to obtain characteristic values; Comparing the characteristic value with a double-layer threshold value to obtain a diagnosis result; and compressing and transmitting the diagnosis result.
2. The subway-oriented intelligent detection monitoring method according to claim 1, wherein the acquisition of the sensor signal comprises the following steps: continuously monitoring the sensor signal while the main controller is in the deep sleep mode; When the sensor signal is greater than the dynamic threshold, waking up the master control; after the main control is awakened, analyzing the sensor signal, and comprehensively starting when the analysis result is abnormal; and after the full starting, carrying out timing inspection.
3. The subway-oriented intelligent detection monitoring method according to claim 2, wherein the calculation formula of the dynamic threshold is: Wherein: is a dynamic threshold; Is the baseline average of the sensor signal; the standard deviation of noise of the sensor signal; Is a configurable sensitivity coefficient; The calculation formula of the early warning threshold value is as follows: Wherein: is an early warning threshold value; the inspection interval for the timing inspection The calculation formula of (2) is as follows: Wherein: taking the inspection interval as a basis; Adjusting the amplitude coefficient to the maximum; Is a device health index; Is a health safety threshold; The width parameter of the interval is adjusted; is a hyperbolic tangent function used for smooth transitions.
4. The subway-oriented intelligent detection monitoring method according to claim 3, wherein the sensor signal is preprocessed to obtain the preprocessed sensor signal, and specifically comprises the following steps: The sensor signal is preprocessed by an improved adaptive Kalman filtering algorithm: State prediction: Covariance prediction: process noise covariance Adaptive adjustment with signal gradient change: Wherein: Is the initial process noise covariance; a gradient of the signal amplitude at time k; Is a smoothing factor; when the signal changes drastically, decrease More trust in predicted value to suppress transient interference, increase when signal is stable The observation value is more trusted, and the tracking precision is improved.
5. The subway-oriented intelligent detection monitoring method according to claim 4, wherein the feature extraction is performed on the preprocessed sensor signal to obtain a feature value, and specifically comprises the following steps: and carrying out feature extraction on the sensor signal after pretreatment by adopting a multi-scale depth feature fusion algorithm: Time domain enhancement features including root mean square, RMS, kurtosis Waveform index SI, peak index and impact index ; , Frequency domain intelligence feature of FFT transformed spectrum Calculating; the spectrum center of gravity FC reflects the position where the spectrum energy is concentrated; The spectrum variance FV reflects the dispersion degree of the spectrum; After the central control module performs real-time preprocessing on the original data, an enhanced feature extraction algorithm is executed, and besides the effective value of the vibration speed, the acceleration peak value and the noise equivalent sound level are calculated, the method comprises the following steps of: time-frequency domain mixing characteristic, namely extracting wavelet packet energy entropy through wavelet packet transformation Characterizing the non-stationary nature of the signal: , 。
6. the subway-oriented intelligent detection and monitoring method according to claim 5, wherein the frequency spectrum The calculation method of (1) is as follows: For the pre-processed time domain signal x (n), n=0, 1,..l-1, where L is the number of sampling points; a window function w (n) is applied to reduce spectral leakage: The hanning window is selected as the window function: FFT transforming is carried out on the windowed signal xw (n) to obtain a complex frequency spectrum X (k): wherein k is the index of spectral line corresponding to the digital frequency L is the FFT point number; extracting amplitude information from a complex frequency spectrum X (k) to obtain an amplitude spectrum A (k): Wherein: is a modulus of the complex spectrum; converting the spectral line index k into the actual physical frequency fk: fs is sampling frequency, fk is actual frequency corresponding to the kth spectral line; Frequency spectrum sequence of frequency domain feature calculation: Wherein, N=L/2, ak is the amplitude sequence used in the frequency domain feature calculation.
7. The subway-oriented intelligent detection monitoring method according to claim 6, wherein the characteristic value is compared with a double-layer threshold value to obtain a diagnosis result, and specifically comprises the following steps: Comparing the characteristic value with an adaptive physical threshold; The adaptive physical threshold The method comprises the following steps: Wherein: is a threshold value under standard conditions, And Is the reference frequency and temperature of the sample, And Is a correction coefficient; intelligent features in the frequency domain are greater than a threshold Judging that an abnormality or potential fault exists, and performing second-layer diagnosis; Intelligent features in the frequency domain are less than a threshold When the equipment is judged to be in normal running state, the second layer diagnosis is not needed to be carried out, and the diagnosis flow is directly ended; the second layer of diagnosis, namely inputting the characteristic value into the DBN model for diagnosis to obtain a diagnosis result; The DBN model is formed by stacking a plurality of layers of limited Boltzmann machines, and probability distribution of fault types is output through a softmax classifier ; Wherein the output of the jth hidden unit The method comprises the following steps: Wherein, the Is the weight of the connection and the weight of the connection, And Is a bias term; Periodic updating of model parameters : Wherein: is the parameters of the model after the updating, Is an adaptive learning rate and is a loss function.
8. The subway-oriented intelligent detection monitoring method according to claim 7, wherein the diagnosis result is compressed and transmitted, and specifically comprises the following steps: And carrying out predictive differential coding on the vibration acceleration and the sound pressure original time series data: Wherein: sampling value of the original time sequence at the nth moment; the predicted value of the sampling value at the nth moment is obtained by linear combination of the previous P historical sampling values; the predicted value of the sampling value at the kth moment is a general form of a prediction model; historical sampling values at the n-i time; linear predictive coefficient, representing the weight of the ith historical sampling value to the current predictive value, predictive coefficient The prediction error is minimized by solving through a Levinson-Durbin recursion algorithm; the prediction order is as follows; Prediction error; Variable length quantized coding of differential sequences Quantization is performed, and a rate distortion optimization model is built to select the optimal quantization step size Minimizing the distortion D at a given bit rate R; rate-distortion optimized quantization is performed on the differential sequence: Wherein: the rate distortion cost function comprehensively measures the distortion and the code rate cost in the coding process and is an optimized objective function; the distortion generated by quantizing the differential sequence dn at a given bit rate R; R is the coding bit rate, namely the bit number required to be transmitted in unit time or unit data volume; lagrangian multiplier for balancing distortion in rate distortion optimization And the weight of the code rate R, and controlling the trade-off between the compression performance and the transmission efficiency; Compressing the quantized data by Varint codes to obtain processed data; Packaging the processed data and the characteristic parameters into a data frame according to an optimized binary protocol, wherein a frame header comprises a sensor ID, a high-precision time stamp, a data length, a characteristic identification bit and a CRC32 check code; The compressed data stream is organized by an intelligent file block coding mechanism, wherein the system dynamically adjusts the size of a file block according to the channel state, adopts an enhanced TLV structure for storage, and records block index, starting time, sampling rate metadata and compression algorithm version at the head; is the signal-to-noise ratio, BER is the bit error rate, RTT is the round trip delay, Is a weight coefficient; When (when) The channel quality is good, namely, the transmission efficiency is preferential by adopting the LDPC code with high coding rate; When 0.5< In the channel quality less than or equal to 0.8, reed-Solomon codes are adopted to balance efficiency and reliability; When (when) Poor channel quality less than or equal to 0.5, adopting Turbo code, and giving priority to reliability; The data load adopts hybrid coding based on physical characteristics, firstly carries out predictive differential coding and then carries out variable length quantization coding, wherein a message header comprises a device identifier, a message serial number, a time synchronization stamp and a channel state indication, the data is sent to a signal transfer host deployed on site through a wireless network, the transfer host carries out deep analysis, characteristic fusion and intelligent aggregation on the data, and the final data is uploaded to a cloud platform database InfluxDB in batches through HTTPS RESTful API for long-term storage and trend analysis; after the data is successfully received, the cloud platform returns an ACK confirmation instruction, and the system executes an intelligent closing flow after receiving the confirmation; When the transmission condition is met, an optimal transmission strategy is selected based on a real-time LQI evaluation result, and after the file block is read from the memory, the data format is restored through a quick decoding process for local display or analysis.
9. The subway-oriented intelligent detection and monitoring method according to claim 8, further comprising the steps of: Dynamic energy budget allocation: The system manages the total energy budget under the assistance of the cloud And consider energy harvesting ; Wherein, the Is the charging efficiency, the system is based on Dynamically adjusting the working mode and the sampling frequency to ensure continuous working in a task period; Constructing a performance degradation model by utilizing characteristic data of a time sequence at a cloud end to predict the residual service life of equipment; Wherein: the key characteristic reflecting the degradation of the performance of the equipment is a function of time t; judging a threshold value of equipment failure; Failure rate function related to current operating state and environment.
10. An intelligent detection monitoring system facing to a subway is used for realizing the intelligent detection monitoring method facing to the subway according to any one of claims 1-9, and is characterized by comprising a signal transfer control and preprocessing module, an L-shaped support frame, a small portable multi-mode intelligent floating slab state sensing terminal, a broadband anti-interference low-power consumption multi-mode intelligent track vibration monitoring sensor, a tunnel side wall multi-parameter vibration noise intelligent sensing terminal, a support plate, a steel rail and a tunnel; The signal transfer control and pretreatment module is fixed at a reserved position of a tunnel or a temporary position through a strong AB glue; The L-shaped support frame is fixed on the side wall of the tunnel through strong AB glue; The tunnel side wall multi-parameter vibration noise intelligent sensing terminal is adsorbed on the L-shaped support frame through the bottom strong magnetic force module; the supporting plate is fixed at the corresponding positions below the floating plate and the track through strong AB glue; The small portable multi-mode intelligent floating plate state sensing terminal bottom powerful magnetic force module is adsorbed on the supporting plate; the broadband anti-interference low-power consumption multimode intelligent track vibration monitoring sensor is fixed on a support plate at the lower side of the track through the anchor.

Description

Intelligent detection monitoring method and system for subway Technical Field The invention relates to the technical field of detection, in particular to an intelligent detection and monitoring method and system for subways. Background With the deep advancement of urban progress in China, subways are used as main artery of urban public transportation, the operation scale and passenger flow volume of subways continuously climb, and unprecedented high standards and strict requirements are put forward for operation safety, riding comfort and operation and maintenance efficiency. The subway system is a complex dynamic system formed by a track, a tunnel, a vehicle and the like, and is in a high-strength and high-load running state for a long time, and the structural health and the running state of the subway system are directly related to the trip safety of millions of passengers. The traditional operation and maintenance mode mainly depends on manual periodic inspection and off-line selective inspection, has low working efficiency and high cost, has the inherent defects of strong subjectivity, incapability of sensing sudden faults in real time and the like, and is difficult to adapt to the development requirements of modern subway networking and high-density operation. In the current subway operation and maintenance, monitoring on a track, a tunnel structure and a train operation state commonly has the following pain points that firstly, monitoring means are fragmented, key parameters such as vibration, noise, deformation and the like are generally independently collected by single-function sensors which are deployed at different positions and different types, the data are isolated and asynchronous, comprehensive diagnosis conclusion is difficult to form, and the overall health condition of the system cannot be comprehensively reflected. And secondly, the real-time performance and the intelligence of the data processing are insufficient, massive monitoring data mostly need to be transmitted back to the rear end for preliminary analysis, the delay is high, the bandwidth pressure is high, and the real-time edge intelligent research and judgment and early fault early warning of key events such as train passing can not be realized. In addition, the existing sensor network has complex wiring and difficult power supply, particularly in severe environments such as tunnels, the cost for laying cables and power supply lines for a large number of sensors is extremely high, the maintenance is inconvenient, and the density and the coverage range of the monitoring network are restricted. Finally, the data link is not smooth, and the whole process from front end acquisition to back end display has the problems of non-uniform data protocol, low transfer efficiency, disordered storage, unstable transmission and the like, so that a decision maker cannot timely, clearly and accurately acquire key information, and the operation and maintenance efficiency and the train operation safety are affected. Under the background, the intelligent monitoring technology has become a necessary trend for ensuring the safety of subways and realizing predictive maintenance. At present, although some subway lines are tried to introduce online monitoring means, three major pain points generally exist, namely, firstly, the system integration level is low, monitoring of different parameters such as vibration, noise, deformation and the like is generally carried out by adopting independent subsystems, so that the sensor is redundant to be laid, data protocols are different, an 'information island' is formed, multi-source information fusion and comprehensive diagnosis are difficult to carry out, secondly, the data processing capacity is delayed, mass monitoring data mostly need to be transmitted to a central server to be processed, the network bandwidth pressure is extremely high, obvious delay is generated, real-time edge intelligent research and judgment cannot be carried out on trains through key events such as the like, a gold window for fault early warning is missed, thirdly, the power supply and wiring problems of a sensing layer, the tunnel environment is complex, the engineering for laying power supply and communication lines for a large number of wired sensors is huge, the maintenance is extremely difficult, and the density and the long-term stability of a monitoring network are severely restricted. Therefore, the development of a set of highly integrated, intelligent self-consistent and wireless interconnected intelligent detection and monitoring system has important significance for breaking through the bottleneck of the prior art and pushing the subway operation and maintenance mode to digital and intelligent transformation and upgrading. Disclosure of Invention The invention aims to provide an efficient subway-oriented intelligent detection monitoring method and system, and an integrated sensing network for sensing key state parameters suc