Optical cable external damage vibration sensing and early warning device

Abstract

The application discloses an optical cable external damage vibration sensing early warning device, which comprises a sensing and collecting component, the sensing and collecting component comprises a passive fiber grating sensing array and a composite coupling substrate collecting mechanism, the composite coupling substrate collecting mechanism embraces and covers the outer wall of the optical cable, the passive fiber grating sensing array is packaged in the composite coupling substrate collecting mechanism, and a single sensing optical fiber is in series connection with a plurality of fiber grating measuring points. The passive fiber grating is adopted as the vibration collecting core component, can cover 0.1Hz-10kHz full-band vibration signals, can capture early weak external force disturbance such as artificial light planing to avoid missed early warning, can synchronously collect temperature data to offset signal deviation caused by environmental temperature change, the sensing link does not need a battery and a wireless transmission circuit, does not exist the operation and maintenance problems of channel wireless signal shielding and regular replacement of the sensor battery, and the embracing type substrate with a damping layer can collect optical cable vibration in all directions and eliminate the monitoring blind area.

Description

Technical Field

[0001] This invention relates to the field of online monitoring and early warning of external force damage to optical cable channels, specifically to a vibration sensing and early warning device for external force damage to optical cables. Background Technology

[0002] Urban communication and power optical cables are extensively laid using direct burial, cable trenches, and ductwork methods. External forces such as construction excavation, mechanical cutting, theft, and heavy vehicle traffic can easily cause cable breakage, leading to communication outages, power transmission failures, and significant economic losses. The current mainstream monitoring solution is a distributed piezoelectric wireless vibration monitoring device, corresponding to the cable external force damage early warning and diagnostic device described in patent CN110443968A. This device relies on distributed piezoelectric ceramic vibration sensors attached to the cable to collect vibration signals. After local sampling and noise reduction by the ARM unit, the signals are uploaded to a host computer via a wireless module. The host computer uses a fixed amplitude threshold combined with fast Fourier transform spectrum matching to complete fault identification and early warning.

[0003] However, piezoelectric ceramics rely on the piezoelectric effect to convert vibrations into voltage signals. Their standard response frequency band has a high lower limit, making it impossible to collect weak low-frequency vibration signals in the 0.1Hz to 500Hz range. Low-frequency vibrations caused by early external forces such as manual planing and slow squeezing will be missed. The alarm can only be triggered when the optical cable is about to break, thus losing the early warning capability. Piezoelectric ceramic materials will undergo polarization aging and sensitivity drift when subjected to alternating vibration and channel temperature change cycles for a long time. The device does not have a hardware temperature compensation structure, and changes in ambient temperature will continuously amplify the signal acquisition error. Summary of the Invention

[0004] The purpose of this invention is to provide a vibration sensing and early warning device for external force damage to optical cables, so as to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] A vibration sensing and early warning device for external force damage to optical cables includes:

[0007] The sensing and acquisition component includes a passive fiber grating sensing array and a composite coupling substrate acquisition mechanism. The composite coupling substrate acquisition mechanism surrounds and covers the outer wall of the optical cable. The passive fiber grating sensing array is encapsulated inside the composite coupling substrate acquisition mechanism. Several fiber grating measurement points are arranged in series on a single sensing fiber.

[0008] The vibration monitoring host includes a high-speed tunable laser demodulation host and an edge AI early warning host computer. The high-speed tunable laser demodulation host is connected to the passive fiber optic grating sensor array via a sensing fiber to collect the wavelength shift signal generated by the strain of the fiber optic grating under external force, and demodulates the vibration time domain data and ambient temperature data. The edge AI early warning host computer is electrically connected to the high-speed tunable laser demodulation host computer. The host computer hardware has a built-in data processing unit, which integrates an adaptive dual threshold judgment unit, a wavelet packet time-frequency decomposition unit, a Mel spectrum feature extraction unit, a K-means dynamic clustering recognition unit, and a bidirectional grating time difference differential positioning unit.

[0009] The cloud-based incremental training platform communicates with the edge AI early warning host computer via network, receives unknown vibration feature samples uploaded by the edge computer, and sends out updated recognition models.

[0010] Furthermore, the composite coupling substrate acquisition mechanism is a split arc-shaped elastic clamping structure, with a damping noise reduction layer set in the inner layer of the substrate, and fiber optic grating measurement points fixed by slotting in the middle of the substrate; fiber optic grating measurement points are densely deployed at optical cable joints, wellheads, and high-risk sections at bends, and fiber optic grating measurement points are deployed at equal intervals in direct burial and ordinary trench sections.

[0011] Furthermore, the passive fiber optic grating sensing array operates based on the fiber wavelength modulation principle, with a sensing response frequency band covering 0.1Hz-10kHz, and simultaneously collects two physical quantities: optical cable vibration strain and ambient temperature.

[0012] Furthermore, the adaptive dual-threshold determination unit pre-collects a 72-hour baseline of normal optical cable vibration, and updates the dynamic amplitude threshold and vibration duration threshold in real time by combining synchronously collected temperature and soil environmental parameters; only when the vibration amplitude exceeds the dynamic baseline threshold and the vibration duration exceeds the construction characteristic time window, a suspicious vibration signal is output.

[0013] Furthermore, the wavelet packet time-frequency decomposition unit performs three-layer wavelet packet hierarchical decomposition on the suspected vibration time-domain signal, and synchronously outputs the time-domain impact timing characteristics and frequency-domain energy characteristics of each frequency band, separating the low-frequency excavation component, the medium-frequency mechanical impact component, and the high-frequency environmental noise component.

[0014] Furthermore, the Mel spectrum feature extraction unit receives the time-frequency data output by wavelet packet layering, extracts multidimensional Mel spectrum feature vectors, and transmits them to the K-means dynamic clustering recognition unit.

[0015] Furthermore, the K-means dynamic clustering identification unit has a built-in local fault cluster; when the feature vector matches an existing cluster, it outputs the fault type: manual pickaxe digging, excavator digging, theft, and vehicle crushing; when there is no matching cluster, it generates a temporary cluster sample and uploads it to the cloud incremental training platform.

[0016] Furthermore, the bidirectional grating time difference differential positioning unit uses the vibration signals synchronously demodulated at both ends of the sensing fiber to calculate the time difference of the wavelength signal reaching the adjacent fiber optic grating measurement points under the same external force excitation, and combines the grating calibration spacing to calculate the location of the external force damage point, with a positioning accuracy of ±1m.

[0017] Furthermore, the high-speed tunable laser demodulation host and the passive fiber optic grating sensor array form a complete passive optical transmission link, and the link transmits all sensor data in a fiber wavelength modulation manner.

[0018] Furthermore, the edge AI early warning host computer is equipped with an audio-visual linkage alarm unit. The host computer locally stores all historical vibration waveforms, temperature data, and early warning records, and supports fault tracing and retrieval.

[0019] Compared with the prior art, the beneficial effects of the present invention are:

[0020] 1. This invention uses a passive fiber optic grating as the core component for vibration acquisition, which can cover the full frequency band of vibration signals from 0.1Hz to 10kHz. It can capture early and weak external force disturbances such as manual light scratching to avoid missing early warnings, and can also collect temperature data simultaneously to offset signal deviations caused by environmental temperature changes. The sensing link does not require batteries or wireless transmission circuits, and there are no maintenance problems such as wireless signal shielding in the channel or regular replacement of sensor batteries. With the addition of a ring-shaped substrate with a damping layer, it can collect optical cable vibration from all directions and eliminate monitoring blind spots.

[0021] 2. This invention abandons the traditional FFT single frequency domain analysis method, and fully preserves the vibration time sequence and frequency band characteristics through wavelet packet time-frequency joint processing, eliminating the identification error caused by spectrum leakage. It generates an adaptive judgment threshold based on a 72-hour environmental baseline, filters instantaneous interference such as vehicles and rainwater to reduce false alarms, and automatically updates fault features by combining cloud incremental training. At the same time, it achieves meter-level fault location by relying on grating time difference calculation, which solves the defects of the original algorithm such as low identification accuracy, inability to learn autonomously, and lack of location function.

[0022] 3. This invention uses a single optical fiber to connect multiple sensing points in series. The device can be expanded to adapt to various laying scenarios such as direct burial, cable trench, and multi-branch long-distance optical cable. The device comes with on-site audible and visual alarms and long-term waveform storage functions. When an alarm is triggered, it can alert construction personnel on-site. It can completely retain monitoring data for easy post-fault review. The overall laying and subsequent maintenance costs are far lower than traditional piezoelectric monitoring solutions, and it has better adaptability for large-scale deployment. Attached Figure Description

[0023] Figure 1 This is a schematic diagram of the overall four-layer architecture of the present invention;

[0024] Figure 2 This is a schematic diagram of the assembly of the composite coupling substrate and the fiber grating measurement point according to the present invention;

[0025] Figure 3 This is a simplified diagram of the internal optical path topology of the demodulation host of the present invention;

[0026] Figure 4 This is a flowchart of the edge AI data serial processing flow of the present invention;

[0027] Figure 5 This is a complete power-on operating timing flowchart of the present invention. Detailed Implementation

[0028] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0029] Please see Figure 1 This invention provides a vibration sensing and early warning device for external force damage to optical cables. The device is divided into four layers: a sensing and acquisition layer, an optical signal demodulation layer, an edge AI processing layer, and a cloud model training layer. The four layers interact with each other according to the logic of optical transmission, electrical signal transmission, and network transmission. The whole machine is equipped with an industrial-grade DC regulated power supply to power the demodulation host, the edge host computer, and the audible and visual alarm unit. The input voltage specification is DC24V.

[0030] Please see Figure 2 In one embodiment of the present invention, the sensing acquisition layer consists of a passive fiber optic grating sensing array and a composite coupling substrate acquisition mechanism. The composite coupling substrate acquisition mechanism adopts a split arc-shaped elastic 304 stainless steel clamping structure. The thickness of a single substrate is 1.2 mm. The inner layer of the substrate is bonded with a 2 mm thick nitrile rubber damping noise reduction layer. A rectangular slot with a width of 5 mm and a depth of 3 mm is reserved in the middle of the substrate. The bare fiber optic grating measurement points are fixed inside the slot with high-temperature resistant epoxy glue. M4 stainless steel fastening bolts are set at both ends of the substrate. The bolts, together with rubber buffer washers, realize stress-free circumferential clamping of the optical cable sheath, avoiding rigid compression that could damage the optical cable.

[0031] As a preferred embodiment of the present invention, fiber optic grating measurement points are densely arranged at 2m intervals at all optical cable joints, entrance and exit wells, and 90-degree bends of optical cables within the optical cable laying channel. Fiber optic grating measurement points are arranged at 10m intervals in ordinary direct-buried sections and straight sections of cable trenches. A single 10km long single-mode G.652D sensing fiber can connect 100 sets of fiber optic grating measurement points in series. The reflection wavelength range of the measurement points is distributed between 1520nm and 1560nm, with a wavelength interval of 4nm between adjacent gratings. The reflection wavelengths of each grating do not overlap, avoiding crosstalk of demodulation signals. The fiber optic gratings are encapsulated on a metal substrate with a thickness of 0.8mm, and the encapsulation temperature range covers -40℃ to 150℃.

[0032] Furthermore, the working mechanism of the passive fiber grating sensing array is that the optical cable is subjected to external force to generate mechanical strain. The strain is uniformly transmitted to the fiber grating through the stainless steel substrate and damping layer. The micron-level change in the grating period causes the center wavelength of the reflected light to shift. The high-speed tunable laser demodulation host emits a broadband laser into the sensing fiber. The fiber optic circulator separates the incident laser from the grating reflected light. The photodetector collects the reflected wavelength shift to complete the physical quantity conversion. The array has no active circuits, energy storage elements, or wireless transmission devices. Data transmission is completed entirely by optical signals, and the total loss of a single-channel optical path is controlled within 8dB.

[0033] Furthermore, the passive fiber optic grating sensor array has a nominal response frequency band of 0.1Hz-10kHz, which can completely collect low-frequency continuous vibrations from manual picks, medium-frequency impact vibrations from excavators, high-frequency cutting vibrations from cutting machines, and wide-frequency vibrations from vehicle rolling. It can simultaneously output vibration amplitude data corresponding to strain and ambient temperature wavelength data. The temperature acquisition resolution is 0.1℃, and the strain measurement resolution is 0.1με, making it suitable for all working conditions in underground passages, including extremely cold and high-temperature environments.

[0034] Please see Figure 3 Preferably, the optical signal demodulation layer is equipped with a high-speed tunable laser demodulation host. The internal optical path of the demodulation host integrates a 1×4 optical coupler, a four-way fiber optic circulator, a narrowband tunable laser, a high-speed photodetector, and a signal conversion circuit. The laser scanning wavelength range is 1510nm-1570nm, the wavelength demodulation resolution is 0.1pm, and the single complete scan cycle is 0.5ms. It can simultaneously acquire wavelength data of all series fiber optic grating measurement points. The host has a built-in moving average filtering preprocessing unit, which performs a 5-point moving average operation on the original wavelength sequence to remove random noise in the optical path. Then, the wavelength offset is converted into a vibration time-domain waveform sequence and ambient temperature value through wavelength-strain and wavelength-temperature dual calibration curves. The standardized floating-point digital signal is continuously sent to the edge AI early warning host computer through the RJ45 gigabit port. A single demodulation host supports four independent fiber optic channels for parallel demodulation, and the maximum number of series fiber optic grating measurement points per channel is 100.

[0035] Please see Figure 4 In another embodiment of the present invention, the edge AI early warning host computer hardware adopts an industrial-grade ARM Cortex-A76 processing unit with 8GB DDR4 local high-speed storage and 128GB solid-state drive. The built-in data processing unit of the hardware is divided into five independent operating sub-units. Each sub-unit completes signal processing in sequence according to the serial data flow logic. The first level runs an adaptive dual threshold judgment unit, the second level runs a wavelet packet time-frequency decomposition unit, the third level runs a Mel spectrum feature extraction unit, the fourth level runs a K-means dynamic clustering recognition unit, and the fifth level runs a bidirectional grating time difference differential positioning unit. The host computer reserves an RS485 switch output interface, a 4G industrial gateway network port, and a local touch display screen.

[0036] Specifically, the adaptive dual-threshold judgment unit operates by continuously collecting the vibration amplitude baseline of all measuring points under 72 hours of no external disturbance after the device is powered on. Every hour, it synchronously collects soil moisture and ambient temperature data within the channel, establishing a baseline correction mapping relationship. This mapping relationship uses a linear fitting formula to complete baseline compensation. It updates the dynamic amplitude threshold and vibration duration threshold in real time. The dynamic amplitude threshold is adaptively adjusted according to the range of 1.8 to 3 times the baseline amplitude. The construction characteristic time window is divided into two standards: a 0.3s threshold for manual excavation and a 1s threshold for heavy machinery construction. Only when the vibration amplitude exceeds the dynamic amplitude threshold and the vibration duration exceeds the corresponding time window threshold is the current waveform determined to be a suspicious vibration signal. Complete time-domain waveform data is then transmitted to the next-level wavelet packet time-frequency decomposition unit. Normal data that does not trigger the threshold is uploaded only once every 5 minutes as a simplified status message, reducing network transmission load.

[0037] Optionally, the wavelet packet time-frequency decomposition unit uses a three-layer db4 wavelet packet basis function to perform hierarchical decomposition on the suspected vibration time-domain waveform. Before decomposition, the waveform is subjected to min-max normalization processing to uniformly map the amplitude to the interval. After decomposition, four sets of output data are obtained: low-frequency sub-band, mid-low-frequency sub-band, mid-high-frequency sub-band, and high-frequency sub-band. The low-frequency sub-band corresponds to the continuous vibration component of slow manual digging, the mid-low-frequency and mid-high-frequency sub-bands correspond to the impact components of excavators and picks, and the high-frequency sub-band corresponds to environmental noise components such as rainwater erosion and light wind. The unit synchronously retains the time-domain impact time-series coordinates and frequency-domain energy amplitudes corresponding to each sub-band, fully preserving the time-frequency joint features. This avoids the defect of fast Fourier transform, which only retains frequency-domain information and loses time-domain features. At the same time, it eliminates the spectral leakage error caused by non-integer period signal truncation. The wavelet packet decomposition threshold is automatically calculated using unbiased risk estimation threshold, and soft threshold noise reduction and reconstruction are performed on the high-frequency noise components.

[0038] In the above scheme, the Mel spectrum feature extraction unit receives the four-dimensional time-frequency dataset output by wavelet packet layering, constructs a 24th-order Mel filter according to the Mel filter bank mapping rule, with a filter cutoff frequency range of 0Hz-10kHz, and extracts a 24-dimensional Mel spectrum feature vector. The feature vector contains four core feature parameters: mean energy, variance energy, impulse time interval, and peak duration for each frequency band. Z-score standardization is performed on the feature vector to eliminate dimensional differences. The standardized 24-dimensional feature vector is then transmitted to the K-means dynamic clustering identification unit to complete the fault matching operation.

[0039] Based on this, the K-means dynamic clustering identification unit pre-stores four types of standard fault cluster centers locally, corresponding to manual pickaxe digging, excavator digging, fiber optic cable theft, and heavy vehicle crushing, respectively. The unit uses the Euclidean distance calculation formula to complete the feature vector matching operation:

[0040]

[0041] In equation (1), For the input feature vector and the th Euclidean distance between fault cluster centers; The input is a 24-dimensional Mel feature vector; For the first Pre-stored cluster center vectors for each type of fault; This is the feature dimension number.

[0042] The unit sets a distance judgment threshold of 0.08. When the Euclidean distance between the vector and a certain cluster center is less than the threshold, the corresponding fault type is determined and an early warning message is output to the host computer. When the Euclidean distance between the input feature vector and all local cluster centers exceeds the threshold, the group of feature vectors is marked as a temporary unknown sample. The corresponding original vibration waveform, ambient temperature, and measurement point location data are packaged and uploaded to the cloud incremental training platform through the 4G industrial gateway.

[0043] Based on this, the cloud-based incremental training platform is equipped with a lightweight K-means clustering iterative training model. The platform receives unknown vibration samples uploaded by multiple edge host computers in batches, and performs an iterative update of cluster centers every 50 new samples. The iteration termination condition is set to the change in cluster centers between two consecutive iterations being less than 1%. After iteration, an optimized global fault clustering model file is generated. After training, it is distributed to all on-site edge AI early warning host computers via the network. The host computers automatically cover the original local clustering centers, realizing the fault identification model's autonomous iterative optimization without human intervention, and continuously improving the fault identification accuracy under complex working conditions.

[0044] This invention provides a vibration sensing and early warning device for optical cable damage caused by external force. A bidirectional grating time-difference differential positioning unit retrieves the same vibration excitation waveform data synchronously acquired by demodulation hosts at both ends of the sensing optical fiber. A generalized cross-correlation function is used to calculate the time difference of the wavelength signal arriving at adjacent fiber optic grating measurement points. The displacement calculation formula is then applied, combining the factory-calibrated grating spacing parameters.

[0045]

[0046] Mode middle, The straight-line distance between the point of external force damage and the front-end grating measuring point; Let be the speed of light in a single-mode fiber, taken as a constant value. ; The signal time difference from the arrival of the same vibration excitation at two sets of grating measuring points is used to calculate the damage point. After the unit is completed, the mileage coordinates of the optical cable channel corresponding to the damage point are output. The coordinate positioning error is controlled within ±1m. The host computer interface synchronously displays the section to which the point belongs and the distance value from the nearest wellhead / joint.

[0047] In one embodiment of the present invention, the edge AI early warning host computer is connected to an external active sound and light linkage alarm unit. The alarm unit operates at DC24V, with a buzzer loudness of 110dB and a red flashing warning light wavelength of 620nm. After the host computer determines that there is an external force damage fault, it synchronously outputs a passive switch signal to trigger the alarm unit to sound a high-decibel buzzer and flash the red light, thereby realizing on-site warning and dissuasion of construction behavior in the passage. After the fault is eliminated and there are no abnormal vibration waveforms for 5 consecutive minutes, the host computer automatically disconnects the switch output, and the alarm unit stops the sound and light output. The host computer's local solid-state hard drive continuously stores vibration waveforms, temperature data, and early warning event records of all measuring points for 30 consecutive days. The stored files are hierarchically classified according to the measuring point number and timestamp, and support maintenance personnel to enter time intervals and measuring point numbers to retrieve historical monitoring data, complete the post-fault tracing and review. The data storage format is a general CSV waveform file, which can be directly imported into third-party data analysis software.

[0048] As a preferred embodiment of the present invention, the entire passive optical transmission link of the device has no active sensing nodes throughout, eliminating the need for on-site power supply lines, battery compartments, and RF antennas. A single 10km optical cable channel only requires the deployment of one high-speed tunable laser demodulation host and one edge AI early warning host computer. Compared with existing distributed piezoelectric wireless sensing solutions, the on-site laying workload is reduced by 65%, and the annual sensor battery replacement and gateway maintenance costs are reduced by 80%. It is suitable for monitoring applications in all scenarios, including urban underground communication optical cables, power cable trenches, and long-distance direct-buried optical cables. Multi-branch optical cable channels can be extended to multiple sensing optical fibers through the four-way optical couplers of the demodulation host, and a single demodulation host can cover a maximum optical cable monitoring mileage of 40km.

[0049] Please see Figure 5 In another embodiment of the present invention, the entire machine operates in a complete closed-loop power-on sequence. After power-on, the demodulation host performs laser wavelength calibration and photodetector zero-point calibration, with a calibration time of 60 seconds. After calibration, the edge host computer starts a 72-hour normal baseline acquisition process. During the baseline acquisition phase, only basic amplitude data is stored, and no warnings are triggered. After baseline acquisition, the system automatically switches to real-time monitoring mode and continuously performs wavelength demodulation, signal preprocessing, threshold determination, time-frequency decomposition, feature extraction, cluster recognition, and point location calculations. If a fault warning is triggered, an audible and visual alarm is activated simultaneously, local data is encrypted and stored, and cloud samples are uploaded. After maintenance personnel handle and eliminate external disturbances on-site, the system continuously monitors for 5 minutes and automatically resets the alarm if no abnormal waveform is detected. The device automatically performs a small baseline update once a day at 2:00 AM and simultaneously uploads all warning statistics for the day to the cloud platform. The cloud platform generates a statistical report on the risk of external damage to the regional optical cable channel for maintenance planning.

[0050] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A vibration sensing and early warning device for external force damage to optical cables, characterized in that, include: The sensing and acquisition component includes a passive fiber grating sensing array and a composite coupling substrate acquisition mechanism. The composite coupling substrate acquisition mechanism surrounds and covers the outer wall of the optical cable. The passive fiber grating sensing array is encapsulated inside the composite coupling substrate acquisition mechanism. Several fiber grating measurement points are arranged in series on a single sensing fiber. The vibration monitoring host includes a high-speed tunable laser demodulation host and an edge AI early warning host computer. The high-speed tunable laser demodulation host is connected to the passive fiber optic grating sensor array via a sensing fiber to collect the wavelength shift signal generated by the strain of the fiber optic grating under external force, and demodulates the vibration time domain data and ambient temperature data. The edge AI early warning host computer is electrically connected to the high-speed tunable laser demodulation host computer. The host computer hardware has a built-in data processing unit, which integrates an adaptive dual threshold judgment unit, a wavelet packet time-frequency decomposition unit, a Mel spectrum feature extraction unit, a K-means dynamic clustering recognition unit, and a bidirectional grating time difference differential positioning unit. The cloud-based incremental training platform communicates with the edge AI early warning host computer via network, receives unknown vibration feature samples uploaded by the edge computer, and sends out updated recognition models.

2. The optical cable external force damage vibration sensing and early warning device according to claim 1, characterized in that: The composite coupling substrate acquisition mechanism is a split arc-shaped elastic clamping structure. A damping and noise reduction layer is set in the inner layer of the substrate, and a slot is opened in the middle of the substrate to fix the fiber optic grating measurement points. Fiber optic grating measurement points are densely deployed in high-risk sections such as optical cable joints, wellheads, and bends, and fiber optic grating measurement points are deployed at equal intervals in direct burial and ordinary trench sections.

3. The optical cable external force damage vibration sensing and early warning device according to claim 1, characterized in that: The passive fiber optic grating sensing array operates based on the fiber wavelength modulation principle, with a sensing response frequency band covering 0.1Hz-10kHz, and simultaneously collects two physical quantities: optical cable vibration strain and ambient temperature.

4. The optical cable external force damage vibration sensing and early warning device according to claim 1, characterized in that: The adaptive dual-threshold judgment unit pre-collects a 72-hour normal vibration baseline of the optical cable, and updates the dynamic amplitude threshold and vibration duration threshold in real time by combining synchronously collected temperature and soil environmental parameters; only when the vibration amplitude exceeds the dynamic baseline threshold and the vibration duration exceeds the construction characteristic time window, a suspicious vibration signal is output.

5. The optical cable external force damage vibration sensing and early warning device according to claim 1, characterized in that: The wavelet packet time-frequency decomposition unit performs three-level wavelet packet hierarchical decomposition on the suspected vibration time-domain signal, and synchronously outputs the time-domain impact timing characteristics and frequency-domain energy characteristics of each frequency band, separating the low-frequency excavation component, the medium-frequency mechanical impact component, and the high-frequency environmental noise component.

6. The optical cable external force damage vibration sensing and early warning device according to claim 1, characterized in that: The Mel spectrum feature extraction unit receives the time-frequency data output by wavelet packet layering, extracts multidimensional Mel spectrum feature vectors, and transmits them to the K-means dynamic clustering recognition unit.

7. The optical cable external force damage vibration sensing and early warning device according to claim 1, characterized in that: The K-means dynamic clustering identification unit has a built-in local fault cluster; When the feature vector matches an existing cluster, the fault type is output as manual pickaxe digging, excavator digging, theft, or vehicle crushing; when there is no matching cluster, a temporary cluster sample is generated and uploaded to the cloud incremental training platform.

8. The optical cable external force damage vibration sensing and early warning device according to claim 1, characterized in that: The bidirectional grating time difference differential positioning unit uses the vibration signals synchronously demodulated at both ends of the sensing fiber to calculate the time difference of the wavelength signal reaching the adjacent fiber optic grating measurement points under the same external force excitation. Combined with the grating calibration spacing, it calculates the location of the external force damage point with a positioning accuracy of ±1m.

9. The optical cable external force damage vibration sensing and early warning device according to claim 1, characterized in that: The high-speed tunable laser demodulation host and the passive fiber optic grating sensor array form a complete passive optical transmission link, which transmits all sensor data in fiber wavelength modulation mode.

10. The optical cable external force damage vibration sensing and early warning device according to claim 1, characterized in that: The edge AI early warning host computer is equipped with an audio-visual linkage alarm unit. The host computer locally stores all historical vibration waveforms, temperature data, and early warning records, and supports fault tracing and retrieval.

Images