Method for monitoring and classifying road impact events based on distributed fiber optic acoustic sensing

Abstract

The application provides a pavement impact event monitoring and classification method based on distributed optical fiber acoustic sensing, and relates to the technical field of road engineering. The method comprises the following steps: collecting phase signals of optical fibers along a line to construct a dynamic strain response matrix; dividing a pavement into a plurality of road section units and binding a road section parameter vector; decomposing the matrix to obtain sparse impact components, constructing an impact saliency field to screen impact candidate events and positioning; determining an impact contact force time history of the events, obtaining impact physical quantities through inversion, and constructing a pavement impact damage index; constructing a propagation mechanism dictionary to demix mixed signals; obtaining an impact footprint topological feature vector according to the saliency field; constructing a pavement physical graph state space model to analyze pavement instantaneous impact and long-term health status; and determining a risk level and grading early warning according to the impact physical quantities and the pavement impact damage index. The application can realize real-time monitoring, accurate positioning, physical quantification, type identification and risk grading of pavement impact events.

Description

Technical Field

[0001] This invention relates to the fields of road engineering and fiber optic sensing technology, and in particular to a method for monitoring and classifying road impact events based on distributed fiber optic acoustic sensing. Background Technology

[0002] my country's total highway mileage has exceeded 5.49 million kilometers, with over 1.1 million highway bridges. This infrastructure needs to withstand complex traffic loads and harsh natural environments over long periods. During its service life, road pavement structures, in addition to conventional fatigue damage, are frequently subjected to sudden impact events. The sources of these impacts can be mainly divided into three categories: 1. Natural factors: falling rocks and debris from mountainous roadside slopes onto the road surface. 2. Traffic factors: cargo (steel coils, tires, machinery) and detached parts from vehicle chassis falling from heavy vehicles. 3. Human factors: construction debris and hard foreign objects scattered on the road surface.

[0003] These types of impact events are characterized by their suddenness, high energy, and short duration. At best, they cause potholes, cracks, and pavement damage; at worst, they trigger tire blowouts, loss of control, and other traffic accidents. Long-term accumulation can also cause hidden damage to bridge structures. Statistics show that the direct economic losses caused by road damage and traffic accidents due to foreign objects and sudden impacts exceed 2 billion yuan annually.

[0004] Existing road surface monitoring technologies have significant shortcomings, failing to meet the needs for comprehensive, real-time, and accurate monitoring of sudden impact events: 1. Video surveillance: Greatly affected by environmental factors such as rain, fog, nighttime lighting, and road surface obstructions, resulting in numerous monitoring blind spots. It cannot achieve effective all-weather monitoring of the entire road network and struggles to capture millisecond-level transient impact signals. 2. Manual inspection: Long inspection cycles and significant delays; it can only detect existing road surface defects, failing to capture instantaneous impact events and hidden structural damage, making it difficult to provide early warnings and trace the source of the event. 3. Vehicle vibration response monitoring: Only reflects the vibration felt by the vehicle itself, a passive monitoring method. It cannot independently perceive impact events across the entire road surface, severely limiting its monitoring range and accuracy.

[0005] Distributed Acoustic Sensing (DAS) technology is a rapidly developing new sensing paradigm in recent years. Its core principle is to utilize ordinary single-mode communication optical fibers as a continuous, highly sensitive array of acoustic sensors. By demodulating the phase changes of the backscattered Rayleigh light in the fiber using a Φ-OTDR (Phase-sensitive Optical Time Domain Reflectometer), the dynamic strain field along the fiber can be sensed in real time. DAS technology has demonstrated outstanding performance in perimeter security intrusion detection, railway safety monitoring, and oil and gas pipeline leak early warning, offering unique advantages such as long-distance operation (30-70 km at one end), all-weather operation, resistance to electromagnetic interference, and spatial resolution down to the meter level.

[0006] In the prior art, the technical solutions closest to this invention mainly include the following two categories: 1. Vibration monitoring technology for road and bridge locations based on MEMS (Micro-Electro-Mechanical Systems) accelerometer sensors.

[0007] Technical Content: MEMS accelerometer sensing and piezoelectric vibration monitoring are standard deployment solutions for structural monitoring of bridges, high slopes, and tunnels in China, representing typical point-based monitoring technologies. In industry engineering applications, they are only deployed at key discrete points such as piers, expansion joints, and slope anchor points to collect steady-state vibration data and trigger alarms for exceeding thresholds. They only serve the safety monitoring of key structures and are not suitable for long-distance, full-coverage road surface monitoring. Existing drawbacks: 1. Fragmented monitoring coverage, unavoidable blind spots: MEMS sensors can only sense vibration signals within a few meters of the installation point, failing to form a continuous sensing field along the road. Achieving dense, full-coverage linear long-distance highway monitoring would result in exponentially increasing costs for equipment, power supply, networking, and maintenance, making it impractical. 2. Limited field deployment and poor long-term stability: Discrete sensors rely on independent power supply and wireless transmission, making large-scale deployment difficult in long tunnels and remote mountainous areas without power grids. Road surface embedded sensors are subject to long-term vehicle traffic, temperature fluctuations, and construction milling disturbances, resulting in low project survival rates and high maintenance frequency, making them unsuitable for long-term online road surface monitoring. 3. Limited identification dimensions and high false alarm rate: Existing commercial systems mostly determine anomalies based on a single vibration amplitude threshold, failing to distinguish between multiple vibration signals from the same source, such as rockfall impacts, falling heavy objects, heavy vehicle crushing, and construction disturbances. Industry test data shows that the false alarm and missed alarm rates for vibration events are high under the pure threshold discrimination mode, lacking the ability to distinguish fine features and classify events.

[0008] 2. Strain monitoring technology for road and bridge structures based on FBG fiber grating.

[0009] Technical Content: FBG (Fiber Bragg Grating) is the most mature fiber optic sensing solution in the field of road and bridge structural health monitoring in China, and has been widely deployed in bridge main bodies, bridge deck pavement, and long-term deformation monitoring of roadbeds. Commercial equipment in the industry generally operates in the low-frequency acquisition range, with conventional demodulators sampling frequencies ≤1kHz, suitable for monitoring slow variables such as structural static strain, temperature stress, and long-term deflection, providing data support for long-term gradual structural damage assessment. Existing limitations: 1. Discrete grating measurement points, unable to provide continuous monitoring across the entire area: FBG is a quasi-point sensing method, with a typical engineering deployment spacing of 5~10m. It can only acquire local data at grating points, completely obscuring the gaps between gratings, and cannot acquire information on the spatiotemporal propagation of impacts along the entire road surface, failing to meet the requirements for continuous monitoring. 2. Insufficient sampling bandwidth, unable to capture transient impacts: Road impacts are millisecond-level high-speed transient events with extremely short energy release times. Existing mainstream commercial FBG demodulation equipment has insufficient sampling frequency to fully reproduce key transient characteristics such as impact peak value, rise time, and decay waveform, making it unable to support impact energy quantification and event identification. 3. Poor adaptability to road conditions: The rigid and fragile packaging structure of FBG sensors cannot adapt to the harsh service conditions of high-temperature construction, high-frequency vehicle rolling, freeze-thaw corrosion, and repeated deformation of asphalt pavements, resulting in insufficient long-term service stability of the road surface. 4. Lack of event classification and grading capabilities: FBG only outputs basic physical quantities such as strain and temperature, lacking a dedicated feature system for road impacts. It cannot distinguish between temperature stress, vehicle braking, conventional loads, and sudden impacts, and cannot achieve refined classification and identification of impact type and impact energy.

[0010] In summary, current mainstream highway monitoring technologies all suffer from shortcomings in scenario adaptability: MEMS point-based monitoring has large blind spots and low event identification accuracy; FBG fiber optic technology has insufficient bandwidth and is unsuitable for transient impact conditions. Existing technologies cannot simultaneously meet the core engineering requirements of highway pavement monitoring, including long-distance continuous sensing across the entire area, all-weather anti-interference, millisecond-level transient capture, impact energy classification and quantification, and long-term stable operation. Therefore, there is an urgent need to propose an intelligent monitoring and classification scheme that integrates impact mechanics modeling, multi-dimensional specific feature extraction, and multi-modal deep learning fusion to fill the technological gap in the industry. Summary of the Invention

[0011] To address the aforementioned technical problems in existing technologies, this invention provides a method for monitoring and classifying road impact events based on distributed fiber optic acoustic sensing. The technical solution is as follows: On the one hand, a method for monitoring and classifying road impact events based on distributed fiber optic acoustic sensing is provided. This method is implemented by a road impact event monitoring and classification device, and includes: S1. The back Rayleigh scattering phase signal along the optical fiber is collected by a phase-sensitive optical time-domain reflectometry device. The back Rayleigh scattering phase signal is subjected to phase expansion, differential demodulation and abnormal jump correction to obtain the dynamic strain response matrix. The road surface is divided into several road segment units along the optical fiber direction, and the corresponding road segment parameter vector is bound to each road segment unit. The optical fiber is laid along the monitoring area.

[0012] S2. The dynamic strain response matrix is ​​decomposed using a robust background manifold decomposition method to obtain sparse impact components. An impact saliency field is constructed based on the sparse impact components. Based on whether the impact saliency field meets the preset judgment conditions within a preset window, impact candidate events are selected from the dynamic strain response matrix.

[0013] S3. Based on the elastic wave front arrival time, non-Euclidean equivalent propagation distance, and robust optimization model, the impact candidate event is located and the location result is obtained; the calibration source is obtained, and the road surface fiber coupling coefficient, propagation attenuation coefficient, and road segment elastic wave equivalent propagation speed in the road segment parameter vector are adaptively calibrated according to the calibration source.

[0014] S4. Determine the observation signal representation of each fiber channel within the window corresponding to the impact candidate event based on the road segment parameter vector. Invert the impact contact force time history in the observation signal representation using a constrained optimization model. Calculate the impact physical quantity based on the inversion result and construct the road impact damage index.

[0015] S5. Construct a propagation mechanism dictionary. Based on the propagation mechanism dictionary and the road segment parameter vector, determine the spatiotemporal response representation within the window corresponding to the impact candidate event. Use a non-negative group sparse algorithm to perform mixed signal demixing on the spatiotemporal response representation to obtain the physical mechanism of the impact candidate event. Perform multi-threshold binarization on the impact saliency field to obtain the impact footprint topological feature vector and determine the impact event type.

[0016] S6. Take each road segment unit as a node, and take the impact physical quantity, road impact damage index and impact footprint topological feature vector as node features to construct a road physical map state space model and set physical rule constraints. Analyze the instantaneous impact and long-term health status of the road surface based on the road physical map state space model.

[0017] S7. Determine the risk level based on the impact physical quantity and road impact damage index, and conduct graded early warning based on the risk level.

[0018] Optionally, S1 also includes: Combining Hertzian contact theory, elastic wave propagation law and fiber optic sensing principle, a physical transmission model of impact signal is constructed. The physical transmission model of impact signal includes a road impact mechanics time history model, a pavement elastic wave propagation model and a fiber optic phase response model.

[0019] Among them, the road impact mechanics time history model uses Hertz contact theory to describe the impact force time history function, as shown in the following equation (1): (1) In the formula, Indicates any moment during the impact process The corresponding instantaneous impact force Indicates the instantaneous time during the impact process. Indicates peak impact force. Indicates the waveform shape index. Indicates the duration of impact contact.

[0020] The propagation model of elastic waves on the pavement is shown in equation (2) below: (2) In the formula, Indicates the radial distance of wave propagation Place Road surface response at all times This represents the radial distance from the observation point to the impact point on the surface. Indicates the peak strain at the impact point. Indicates the reference distance. This represents the attenuation coefficient of asphalt mixture pavement for elastic waves. Function representing the impact waveform. This indicates the Rayleigh wave velocity of asphalt pavement.

[0021] The fiber optic phase response model is used to describe the relationship between the phase change rate and the road surface strain, as shown in equation (3) below: (3) In the formula, This indicates the phase of the backscattered Rayleigh light in the optical fiber. Represents the longitudinal coordinate along the optical fiber. Indicates the refractive index of the fiber core. Indicates the laser wavelength. This represents the fiber strain coefficient. Indicates the position along the longitudinal direction of the optical fiber. place, time Axial strain acting on the optical fiber.

[0022] Optionally, the road segment parameter vector in S1 includes: road surface structure parameters and fiber optic coupling parameters.

[0023] Among them, the pavement structure parameters and optical fiber coupling parameters include pavement layer thickness, optical fiber burial depth, lateral distance from the optical fiber to the potential impact area, pavement temperature, pavement optical fiber coupling coefficient, and equivalent propagation speed of elastic waves in the road section.

[0024] Optionally, S3 includes: For each fiber channel within the window corresponding to the impact candidate event, the arrival time of the elastic wave front of the fiber channel is calculated based on the arrival time scoring function. The arrival time relationship between the impact point and the fiber channel is obtained based on the arrival time of the elastic wave front and the non-Euclidean equivalent propagation distance. The location result of the impact candidate event is obtained based on the arrival time relationship and the robust optimization model. The location result includes the impact occurrence time, impact location, wave velocity estimate, and location confidence.

[0025] The time relationship between the impact point and the optical fiber channel is expressed as shown in equation (4): (4) In the formula, Indicates the first The arrival time of the elastic wavefront at each fiber optic channel. This indicates the spatial index of the fiber optic channel, corresponding to the [number]th [channel] along the fiber optic deployment direction. Each sampling point Indicates the actual start time of the impact event. Indicates the road segment parameters The corrected equivalent propagation distance of the road surface. Indicates the first The spatial coordinates of the optical fiber corresponding to each optical fiber channel. This represents the spatial coordinates of the point where the impact event occurred. This represents the equivalent wave velocity of the corresponding road segment. This indicates measurement error.

[0026] Alternatively, the observation signal representation for each fiber channel in S4 is as shown in equation (5) below: (5) In the formula, Indicates the first The observation signal representation of each fiber optic channel, Represents a time variable. Indicates the fiber optic coupling coefficient of the road surface. Indicates the propagation from the point of impact to the... The road surface propagation kernel function for each channel. Indicates the spatial location of the impact source. This indicates the equivalent propagation speed of elastic waves in the road section. Indicates the propagation attenuation coefficient. Indicates the thickness of the paving layer. This represents the convolution operation. Represents the impact contact force time history. Indicates noise.

[0027] Optionally, the impact physical quantities in S4 include: peak impact force, impulse, impact energy, and contact duration.

[0028] The road impact damage index is used to quantitatively assess the impact intensity and structural damage risk, as shown in equation (6) below: (6) In the formula, Indicates the road surface impact damage index. Indicates the weighting coefficient. Indicates peak impact force. This indicates the peak impact force reference safety threshold. Indicates the weighting coefficient. Indicates impact energy. This indicates the reference safety threshold for impact energy. Indicates the weighting coefficient. Indicates impulse. This indicates the impulse reference safety threshold. Indicates the weighting coefficient. This represents the elastic wave propagation attenuation coefficient of the corresponding road segment under the current condition. This represents the attenuation coefficient under the historical health conditions of a road segment. Indicates the weighting coefficient. This represents the change in wave speed. This indicates the wave velocity during the historical healthy state of the road segment.

[0029] Optionally, the topological feature vector of the impact footprint in S5 is shown in equation (7) below: (7) In the formula, Represents the topological feature vector of the impact footprint. Indicates the number of major connected components. Indicates the length of the space extension. Indicates the duration of time. Indicates the curvature of the propagation ridge. This indicates the asymmetry of left-right propagation. Indicates the lifecycle of the connected component. This indicates the complexity of the impact footprint hierarchy.

[0030] On the other hand, a road impact event monitoring and classification device based on distributed fiber optic acoustic sensing is provided. This device is applied to a road impact event monitoring and classification method based on distributed fiber optic acoustic sensing. The device includes: The data acquisition module is used to collect the back Rayleigh scattering phase signal along the optical fiber using a phase-sensitive optical time-domain reflectometry device. The back Rayleigh scattering phase signal is then subjected to phase expansion, differential demodulation, and anomalous jump correction to obtain the dynamic strain response matrix. The road surface is divided into several road segment units along the optical fiber direction, and a corresponding road segment parameter vector is bound to each road segment unit. The optical fiber is laid along the monitoring area.

[0031] The impact candidate event screening module is used to decompose the dynamic strain response matrix using a robust background manifold decomposition method to obtain sparse impact components, construct an impact saliency field based on the sparse impact components, and screen impact candidate events from the dynamic strain response matrix based on whether the impact saliency field meets preset judgment conditions within a preset window.

[0032] The positioning module is used to locate impact candidate events based on the elastic wave front arrival time, non-Euclidean equivalent propagation distance, and robust optimization model to obtain positioning results; it also acquires calibration sources and uses these sources to adaptively calibrate the pavement fiber coupling coefficient, propagation attenuation coefficient, and equivalent propagation speed of elastic waves in the road segment parameter vector.

[0033] The inversion module is used to determine the observation signal representation of each fiber channel within the window corresponding to the impact candidate event based on the road segment parameter vector. The impact contact force time history in the observation signal representation is inverted through a constrained optimization model. The impact physical quantities are calculated based on the inversion results, and the road impact damage index is constructed.

[0034] The unmixing and feature extraction module is used to construct a propagation mechanism dictionary, determine the spatiotemporal response representation within the window corresponding to the impact candidate event based on the propagation mechanism dictionary and the road segment parameter vector, and perform mixed signal unmixing on the spatiotemporal response representation through a non-negative group sparse algorithm to obtain the physical mechanism of the impact candidate event; and perform multi-threshold binarization on the impact saliency field to obtain the impact footprint topological feature vector to determine the impact event type.

[0035] The graph state-space model construction module is used to construct a road physical graph state-space model by taking each road segment unit as a node, and taking the impact physical quantity, road impact damage index and impact footprint topological feature vector as node features. It sets physical rule constraints and analyzes the instantaneous impact and long-term health status of the road surface based on the road physical graph state-space model.

[0036] The output module is used to determine the risk level based on the impact physical quantity and the road impact damage index, and to provide graded early warnings based on the risk level.

[0037] On the other hand, a road impact event monitoring and classification device is provided, the road impact event monitoring and classification device comprising: a processor; a memory, the memory storing computer-readable instructions, which, when executed by the processor, implement any of the methods described above for road impact event monitoring and classification based on distributed fiber optic acoustic sensing.

[0038] On the other hand, a computer-readable storage medium is provided, wherein at least one instruction is stored in the storage medium, the at least one instruction being loaded and executed by a processor to implement any of the above-described methods for monitoring and classifying road impact events based on distributed fiber optic acoustic sensing.

[0039] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: Full-area, all-weather continuous monitoring with a wide coverage: a single-end device can achieve long-distance continuous monitoring of 70km of roads, completely making up for the blind spots of video surveillance and manual inspection, and is not affected by environmental factors such as night, rain, fog, or obstruction, so as to achieve uninterrupted monitoring of the road network around the clock.

[0040] Driven by a full-link physical model, it has strong adaptability: It constructs a complete physical transmission system of "impact mechanics - elastic wave propagation - fiber phase response", combined with road segment unit modeling and parameter self-calibration, to accurately match different road surfaces and fiber optic deployment conditions, and to distinguish between impact and conventional disturbances from the mechanism level.

[0041] Multi-level anti-interference design with high robustness: It adopts technologies such as robust background manifold decomposition, propagation mechanism unmixing, and topology feature screening to peel away traffic, wind and rain, and equipment noise layer by layer, which greatly reduces the false alarm and missed alarm rates.

[0042] Multi-dimensional quantitative analysis enables damage prediction: It completes the inversion of physical parameters such as impact force, energy, and attenuation coefficient, and constructs an impact damage index, which can not only identify events but also assess the risk of hidden damage to pavement structures.

[0043] Hierarchical reasoning + unknown event rejection for high classification accuracy: It combines physical evidence, topological features, and graph state space reasoning for hierarchical classification, and actively prompts manual review for abnormal unknown events, adapting to complex road surface events.

[0044] Adaptive online updates and stable long-term operation: The design incorporates drift suppression and parameter gating update mechanisms to adapt to long-term operating conditions such as road aging and temperature changes, ensuring the long-term monitoring accuracy of the system.

[0045] Tiered early warning and linkage support refined maintenance: Based on the impact risk classification, multi-level early warning is set up and linked with maintenance and traffic control platforms to achieve an upgrade from passive emergency repair to proactive prediction and differentiated maintenance, thereby reducing road operation and maintenance costs. Attached Figure Description

[0046] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0047] Figure 1 This is a flowchart of a method for monitoring and classifying road impact events based on distributed fiber optic acoustic sensing, provided by an embodiment of the present invention. Figure 2 This is a schematic cross-sectional view of the road fiber optic cable installation provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the overall architecture provided in an embodiment of the present invention; Figure 4 This is a block diagram of a road impact event monitoring and classification device based on distributed fiber optic acoustic sensing provided in an embodiment of the present invention; Figure 5 This is a schematic diagram of the structure of a road impact event monitoring and classification device provided in an embodiment of the present invention. Detailed Implementation

[0048] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0049] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0050] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0051] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0052] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0053] This invention provides a method for monitoring and classifying road impact events based on distributed fiber optic acoustic sensing. This method can be implemented using a road impact event monitoring and classification device, which can be a terminal or a server. Figure 1 The flowchart shown is for a road impact event monitoring and classification method based on distributed fiber optic acoustic sensing. The processing flow of this method may include the following steps: S1. The back Rayleigh scattering phase signal along the optical fiber is collected by a phase-sensitive optical time-domain reflectometry device. The back Rayleigh scattering phase signal is subjected to phase expansion, differential demodulation and abnormal jump correction to obtain the dynamic strain response matrix. The road surface is divided into several road segment units along the optical fiber direction, and the corresponding road segment parameter vector is bound to each road segment unit. The optical fiber is laid along the monitoring area.

[0054] In one feasible implementation, the present invention first utilizes sensing optical fibers laid along the monitoring area to collect the back Rayleigh scattering phase signal along the optical fiber in real time using phase-sensitive optical time-domain reflectometry, and constructs a spatiotemporal response matrix.

[0055] Specifically, such as Figure 2 As shown, standard single-mode optical fibers are shallowly buried along the road shoulder or median strip, with a burial depth controlled between 0.3 and 0.6 meters. Existing communication optical cables along the road can also be reused, reducing construction costs. The fiber phase signal is acquired in real time using a Φ-OTDR device. Core device parameters are configured as follows: pulse width 50-100 ns, repetition frequency 2-10 kHz, and spatial resolution 5-10 meters. The spatiotemporal response matrix of the road surface impact is then obtained. ,in The spatial distance along the fiber optic cable. To monitor the time, a complete physical transmission model of the impact signal is constructed by combining Hertzian contact theory, elastic wave propagation laws, and fiber optic sensing principles: (1) Time history model of road impact mechanics: simplifying road impact into mass Speed ​​of movement A rigid object impacts the pavement structure perpendicularly; the impact force time history function is described using Hertzian contact theory.

[0056] In the formula, Indicates any moment during the impact process The corresponding instantaneous impact force Indicates the instantaneous time during the impact process. Indicates peak impact force. This represents the waveform shape index, describing the sharpness of the impact force waveform. According to Hertzian contact theory, for an elastic sphere impacting a plane, For actual road impacts (such as falling rocks or steel coils), it is recommended to take... The larger the value, the steeper the peak impact.

[0057] In the formula, the peak impact force is:

[0058] In the formula, , These are the elastic constants of the impacting object and the pavement material, respectively. Let be the equivalent radius of curvature of the impacting object.

[0059] Impact contact duration:

[0060] (2) Pavement Elastic Wave Propagation Model: Road impact energy will excite volume waves (P-waves, S-waves) and surface waves (Rayleigh waves) in the pavement structure. Among them, Rayleigh waves propagate along the road surface, have slow geometric attenuation, and travel long distances, making them the main source of fiber optic sensing signals. (Distance along the road surface) The formula for the axial strain amplitude of the optical fiber at the specified location is:

[0061] In the formula, Indicates the radial distance of wave propagation Place Road surface response at all times This represents the radial distance from the observation point (where the optical fiber is laid) to the impact point on the surface. Indicates the peak strain at the impact point. This indicates the reference distance, i.e., the origin reference position. This represents the attenuation coefficient of elastic waves on asphalt mixture pavement (unit: m). -1 The typical value at 100Hz is 0.03~0.08m. -1 , This represents the impact waveform function, which is dimensionless and describes the change in impact strain over time. This indicates the Rayleigh wave velocity of asphalt pavement (typically 500~1200m / s).

[0062] (3) Fiber phase response model: The change in pavement strain will cause the fiber phase shift, and the relationship between the phase change rate and the pavement strain is as follows:

[0063] In the formula, Represents the phase of the backscattered Rayleigh light in an optical fiber (unit: rad). Represents the longitudinal coordinate along the optical fiber (unit: m). Indicates the refractive index of the fiber core. nm represents the laser wavelength. This represents the fiber strain coefficient. Indicates the position along the longitudinal direction of the optical fiber. place, time Axial strain acting on the optical fiber.

[0064] This invention adopts a three-layer physical modeling system: it establishes a complete quantitative transmission link of Hertzian contact impact force model + road surface Rayleigh wave propagation model + fiber optic phase response model, and combines road segment unit modeling and parameter vector construction to provide a theoretical basis for signal processing, feature extraction and parameter inversion, thus overcoming the shortcoming of existing DAS technology that lacks a dedicated physical model for road impact.

[0065] The backscattered Rayleigh phase signal along the optical fiber was continuously acquired using a phase-sensitive optical time-domain reflectometry device. The acquired raw signal is represented as follows:

[0066] in, This represents the i-th fiber spatial sampling point. This indicates the j-th sampling time. This represents the phase response value at that location and time.

[0067] Furthermore, the original phase signal is subjected to phase expansion, differential demodulation, and anomalous jump correction to obtain the dynamic strain response matrix:

[0068] in, This indicates the phase expansion operation. This indicates time difference operations.

[0069] Optionally, the road segment parameter vector in S1 includes: road surface structure parameters and fiber optic coupling parameters.

[0070] Among them, the pavement structure parameters and optical fiber coupling parameters include pavement layer thickness, optical fiber burial depth, lateral distance from the optical fiber to the potential impact area, pavement temperature, pavement optical fiber coupling coefficient, and equivalent propagation speed of elastic waves in the road section.

[0071] In one feasible implementation, the road is divided into several road segment units along the fiber optic direction:

[0072] In the formula, This indicates the number of road segment units.

[0073] Each road segment unit The corresponding road structure parameters and fiber optic coupling parameters are bound together, including pavement layer thickness, road surface type, fiber optic burial depth, fiber optic distance from lane centerline, ambient temperature, historical traffic background intensity, construction joint location, and bridge expansion joint location, to form a road segment parameter vector:

[0074] in, Indicates the thickness of the paving layer. Indicates the burial depth of the optical fiber. This indicates the lateral distance from the optical fiber to the area where impact is likely to occur. Indicates the road surface temperature. Indicates the fiber optic coupling coefficient of the road surface. Indicates the propagation attenuation coefficient. This indicates the equivalent propagation speed of elastic waves in that road section.

[0075] The purpose of this step is to transform the DAS signal from a simple one-dimensional vibration sequence into a spatiotemporal response field of road impact with road structure attributes, spatial location attributes, and material propagation attributes, thus providing a basis for subsequent physical inversion and classification.

[0076] S2. The dynamic strain response matrix is ​​decomposed using a robust background manifold decomposition method to obtain sparse impact components. An impact saliency field is constructed based on the sparse impact components. Based on whether the impact saliency field meets the preset judgment conditions within a preset window, impact candidate events are selected from the dynamic strain response matrix.

[0077] One feasible implementation involves extracting impact candidate events based on a robust background manifold.

[0078] Specifically, due to the persistent presence of vehicle traffic, wind and rain disturbances, bridge micro-vibrations, and equipment noise in road scenarios, directly applying ordinary filtering to the signal can easily weaken the rising edge of the impact. This invention employs a robust background manifold decomposition method to separate stable background, sparse impact components, and random noise components from the continuous spatiotemporal response matrix: .in, This represents the background manifold composed of normal traffic loads, environmental vibrations, and low-frequency road surface background. It indicates short-term, strong transient components such as sudden impacts, falling object impacts, and rockfall impacts; This represents random noise and equipment error.

[0079] Background manifold decomposition is achieved through the following objective function:

[0080] in, Used to constrain the sparsity of impact signals, This is a constraint on the total variation in spatial directions. To constrain the total variation in the time direction, the background field is made to change slowly in both space and time. Indicates control of transient components of sudden impact Weighting coefficients of sparse constraint strength Represents the control background manifold Weighting coefficient of spatial smoothness intensity Represents the control background manifold Weighting coefficients for time-direction smoothness intensity Indicates control of random noise and equipment error The weighting coefficient of constraint strength.

[0081] In obtaining rarefaction impact components Then, construct the impact significance field:

[0082] in, Indicates the intensity of abrupt changes in the time direction. This represents the gradient propagating in spatial directions. This indicates the phase consistency between adjacent spatial channels. Indicates background confidence level. These are the weighting coefficients.

[0083] When the impact saliency field simultaneously meets the four conditions of "high amplitude, strong rising edge, spatial coherent propagation, and background inconsistency" within a local spatiotemporal window, it is determined to be an impact candidate event. This step is not a simple threshold alarm, but rather eliminates normal vehicle crushing, continuous traffic vibration, and isolated noise through background manifold stripping and spatiotemporal coherence constraints.

[0084] This invention employs robust background manifold decomposition combined with impact saliency field detection: for complex mixed interference on roads, manifold decomposition is used to separate background, impact, and noise components, and a saliency field is constructed by combining spatiotemporal multi-dimensional indicators to achieve accurate extraction of impact candidate events and effectively preserve the transient rising edge characteristics of impact.

[0085] S3. Based on the elastic wave front arrival time, non-Euclidean equivalent propagation distance, and robust optimization model, the impact candidate event is located and the location result is obtained; the calibration source is obtained, and the road surface fiber coupling coefficient, propagation attenuation coefficient, and road segment elastic wave equivalent propagation speed in the road segment parameter vector are adaptively calibrated according to the calibration source.

[0086] Optionally, S3 includes: For each fiber channel within the window corresponding to the impact candidate event, the arrival time of the elastic wave front of the fiber channel is calculated based on the arrival time scoring function. The arrival time relationship between the impact point and the fiber channel is obtained based on the arrival time of the elastic wave front and the non-Euclidean equivalent propagation distance. The location result of the impact candidate event is obtained based on the arrival time relationship and the robust optimization model. The location result includes the impact occurrence time, impact location, wave velocity estimate, and location confidence.

[0087] The arrival time relationship represents the time delay relationship between the impact point and each fiber optic measuring point.

[0088] One feasible implementation method is impact localization based on non-Euclidean wavefront arrival time.

[0089] Specifically, for candidate events, this invention does not directly use the point of maximum amplitude as the impact location, but instead constructs a shock wavefront model based on the elastic wave arrival time. For each fiber spatial sampling point within the candidate event window... Calculate its wavefront arrival time:

[0090] in, The arrival time scoring function is composed of impact significance, phase consistency, and local energy mutation.

[0091] in, Indicates the first fiber optic spatial sampling points exist The significance value of the impact at time. This represents the cross-correlation function, used to calculate the waveform similarity between two time-domain signals. After decomposition of the background manifold, the first... Each fiber optic channel The sparse impact component signal at time, Indicates the first Each neighborhood channel, after propagation delay The sparse impact component signal after time-domain shift This represents the time variable, corresponding to the sampling time of the signal. Indicates the first The set of neighboring channels of each channel. This indicates the propagation delay between different channels.

[0092] Considering that roads are not ideally homogeneous media, road surface joints, different paving materials, and fiber optic bending paths can all alter wave propagation paths. Therefore, this invention employs non-Euclidean propagation distance for positioning. Impact point. With the The arrival time relationship between the fiber optic channels is expressed as follows:

[0093] In the formula, Indicates the first The arrival time of the elastic wavefront at each fiber optic channel. This indicates the spatial index of the fiber optic channel, corresponding to the [number]th [channel] along the fiber optic deployment direction. Each sampling point Indicates the actual start time of the impact event. Indicates the road segment parameters The corrected equivalent propagation distance of the road surface. Indicates the first The spatial coordinates of the optical fiber corresponding to each optical fiber channel. This represents the spatial coordinates of the point where the impact event occurred. This represents the equivalent wave velocity of the corresponding road segment. This indicates measurement error.

[0094] The impact location was determined using the following robust optimization model:

[0095] in, This is a robust loss function used to mitigate the impact of abnormal channels, fiber breakage points, and localized loosening points on the positioning results. This represents the equivalent propagation speed variable of the elastic wave to be optimized.

[0096] This step outputs the impact time, impact location, wave velocity estimate, and location confidence level. Compared to methods based on single-channel peak values ​​or ordinary spatial cross-correlation, this method can adapt to heterogeneous road materials, non-linear fiber optic deployments, and complex paving structures.

[0097] Optionally, S3 further includes: acquiring a calibration source, and adaptively calibrating the pavement fiber coupling coefficient, propagation attenuation coefficient, and equivalent propagation velocity of elastic waves in the road segment parameter vector based on the calibration source.

[0098] In one feasible implementation, the fiber optic coupling parameters of the road surface are self-calibrated.

[0099] Specifically, in order to avoid different DAS responses to the same impact due to differences in burial depth, compaction degree, and paving materials in different road sections, this invention introduces a self-calibration mechanism for road fiber optic coupling parameters.

[0100] In the initial stage of system operation, using low-risk known events or stable wheel loads in natural traffic as calibration sources, the response amplitude, propagation velocity, and attenuation curve of each road segment unit are extracted to establish the coupling parameters of the road segment unit:

[0101] in, Indicates the first The actual observed response amplitude of each road segment This indicates the reference response amplitude under standard road sections or standard calibration sources.

[0102] Furthermore, the attenuation coefficient is estimated based on the amplitude attenuation relationship of the shock wave in adjacent channels:

[0103] in, Indicates the first Peak response of each channel, Indicates the initial amplitude of the impact source. Indicates the point of impact up to the 1st The equivalent propagation distance of each channel, To avoid near-field singularity correction distance, This represents the propagation attenuation coefficient of that road section.

[0104] By minimizing the difference between the observed amplitude and the model amplitude, the inversion is obtained for each road segment. , , :

[0105] The above parameters are updated online over time, but only in the high-confidence, low-risk, non-abnormal event window to prevent real road damage events from being incorrectly absorbed into the background model.

[0106] This invention employs non-Euclidean wavefront positioning combined with self-calibration of coupling parameters: impact positioning is achieved based on the arrival time of elastic waves and the equivalent propagation distance of non-uniform road surfaces; adaptive calibration of fiber optic coupling, attenuation, and wave velocity parameters across the entire road section is achieved by relying on a standard calibration source, eliminating system errors caused by differences in road sections.

[0107] S4. Determine the observation signal representation of each fiber channel within the window corresponding to the impact candidate event based on the road segment parameter vector. Invert the impact contact force time history in the observation signal representation using a constrained optimization model. Calculate the impact physical quantity based on the inversion result and construct the road impact damage index.

[0108] Among them, the observed signal represents the time-domain signal of dynamic strain of the impact elastic wave after the structural and fiber coupling parameters of the road section are corrected.

[0109] One feasible implementation involves impact contact process inversion and energy quantification.

[0110] Specifically, this invention models sudden road impacts as a continuous physical process of "external impact force – propagation through pavement structure – fiber optic coupling response". For the first... Each fiber optic channel has an observed signal represented as follows:

[0111] in, Indicates the propagation from the point of impact to the... The road surface propagation kernel function for each channel. Represents the impact contact force time history. This represents the convolution operation. Indicates noise.

[0112] Impact contact force time history Expanding using asymmetric positive basis functions:

[0113] in, It is an asymmetric B-spline basis function used to characterize complex contact processes such as rapid rise, slow decay, or secondary rebound of impact force.

[0114] The impact contact process is inverted using the following constrained optimization model:

[0115] in, , Used to constrain finite variations in the impact force curve, Used to suppress irrational oscillations. This represents prior constraints on physical parameters.

[0116] After completing the inversion, calculate the following impact physical quantities:

[0117] in, Peak impact force To boost volume, To impact energy, For the duration of contact, This is the equivalent contact speed.

[0118] Constructing the road impact damage index:

[0119] In the formula, Indicates the road surface impact damage index. Indicates the weighting coefficient. Indicates peak impact force. This indicates the peak impact force reference safety threshold. Indicates the weighting coefficient. Indicates impact energy. This indicates the reference safety threshold for impact energy. Indicates the weighting coefficient. Indicates impulse. This indicates the impulse reference safety threshold. Indicates the weighting coefficient. This represents the elastic wave propagation attenuation coefficient of the corresponding road segment under the current condition. This represents the attenuation coefficient under the historical health conditions of a road segment. Indicates the weighting coefficient. This represents the change in wave speed. This indicates the wave velocity during the historical healthy state of the road segment.

[0120] This step not only determines whether an impact has occurred, but also further determines how strong the impact is, whether it may cause damage to the pavement layer, and whether maintenance is required.

[0121] This invention employs impact contact process inversion and damage quantification: by constraining and optimizing the impact contact force time history, physical indicators such as peak force, impulse, and impact energy are quantified to construct a road impact damage index, thereby achieving a quantitative assessment of impact intensity and structural damage risk.

[0122] S5. Construct a propagation mechanism dictionary. Based on the propagation mechanism dictionary and the road segment parameter vector, determine the spatiotemporal response representation within the window corresponding to the impact candidate event. Use a non-negative group sparse algorithm to perform mixed signal demixing on the spatiotemporal response representation to obtain the physical mechanism of the impact candidate event. Perform multi-threshold binarization on the impact saliency field to obtain the impact footprint topological feature vector and determine the impact event type.

[0123] In one feasible implementation, event demixing based on a propagation mechanism dictionary is used to purify the feature input.

[0124] Specifically, a road scene may simultaneously contain multiple signals such as vehicle running over, falling object impact, continuous dragging, raindrop impact, and construction impact. To avoid classifying different sources together, this invention constructs a propagation mechanism dictionary to physically demix candidate events.

[0125] Define mechanism dictionary: .in, This represents a vertically rigid impact propagation template. Indicates the template for vehicle rolling load. This indicates dragging and rubbing the template. This indicates raindrops or small particles hitting the template. Indicates construction vibration formwork. This represents a template for equipment noise.

[0126] The spatiotemporal response within the candidate event window is represented as follows:

[0127] in, The first after the road segment parameters are corrected Class propagation mechanism template, The activation intensity of the corresponding mechanism, It represents the residual.

[0128] The components of each mechanism are obtained by sparse solving of non-negative groups:

[0129] like If the activation intensity is high, and its wavefront arrival time, energy decay, and contact time all meet the physical constraints of sudden impact, it is considered a valid road impact event; if or If it is dominant, it is considered a normal vehicle disturbance or towing disturbance; if If the disturbance is dominant and has low spatial coherence, it is considered an environmental disturbance.

[0130] Furthermore, the topological features of the impact footprint are constructed.

[0131] Specifically, to further enhance the ability to distinguish complex shock events, this invention constructs the topological features of the shock footprint from the shock saliency field, propagation wavefront, and mechanism unmixing results, rather than using general time-frequency maps or image coding.

[0132] First, the impact significance field is binarized at multiple threshold levels:

[0133] With threshold The impact footprint topological description vector is obtained by recording the number of connected regions, duration, spatial extension length, number of main propagation ridges, and disappearance order of the impact region, from high to low.

[0134] In the formula, Represents the topological feature vector of the impact footprint. Indicates the number of major connected components. Indicates the length of the space extension. Indicates the duration of time. Indicates the curvature of the propagation ridge. This indicates the asymmetry of left-right propagation. Indicates the lifecycle of the connected component. This indicates the complexity of the impact footprint hierarchy.

[0135] The topological patterns of the footprints differ significantly for different events such as rockfalls, falling steel coils, vehicle debris, tire blowouts, and construction impacts. For example, rockfall impacts typically exhibit high peak values, short contact times, and rapid decay in multiple directions; vehicle debris impacts are usually accompanied by a main impact and subsequent rolling / dragging components; construction impacts may show periodic, repeating footprints; and raindrop noise presents numerous low-energy, weakly coherent, and small connected domains without stable wavefronts.

[0136] This invention employs a propagation mechanism dictionary demixing + impact footprint topology features: establishing multi-class disturbance propagation templates to complete signal demixing; and extracting impact topology features based on multi-threshold binarization to distinguish the morphological differences of different types of impact events.

[0137] S6. Take each road segment unit as a node, and take the impact physical quantity, road impact damage index and impact footprint topological feature vector as node features to construct a road physical map state space model and set physical rule constraints. Analyze the instantaneous impact and long-term health status of the road surface based on the road physical map state space model.

[0138] One feasible implementation is a physical constraint graph state-space reasoning model.

[0139] Specifically, after obtaining the impact physical quantities, mechanism unmixing coefficients, and impact footprint topological features, this invention constructs a road physical map state space model to perform spatiotemporal reasoning on events.

[0140] Each road segment unit As a graph node, the node characteristics are:

[0141] Edge weights are constructed based on road connectivity, wave propagation time, fiber optic cable routing path, and material similarity:

[0142] in, Represents road segment unit and The strength of the propagation correlation between them.

[0143] Establish the graph state-space equations:

[0144] in, The hidden state of the m-th road segment at time t is represented, which includes the background disturbance level, structural health memory, impact residual effects, and sensor coupling drift. This is the state transition matrix for this road segment. The propagation matrix of adjacent road segments, Input matrix for events, This is state noise.

[0145] To ensure that the model conforms to the physical laws of road impact, the following constraints are set:

[0146] That is, the wave velocity must be within the allowable range of the road material, the attenuation coefficient must be non-negative, the propagation energy must not increase unnecessarily when there is no new external impact, and the response amplitude should decrease as a whole with the propagation distance.

[0147] The model's function is to link the results of a single impact with the long-term condition of the road, enabling it to identify instantaneous events and determine whether there are abnormal coupling changes, local loosening, pavement damage, or sensor fiber failure risks in the road section.

[0148] This invention employs a physical constraint graph state-space reasoning model: using road segments as graph nodes to fuse multi-dimensional features, adding physical rule constraints, associating instantaneous events with the long-term health status of the road, and identifying hidden structures and sensor faults.

[0149] S7. Determine the risk level based on the impact physical quantity and road impact damage index, and conduct graded early warning based on the risk level.

[0150] One feasible implementation involves hierarchical classification based on evidence reasoning and rejection of unknown events.

[0151] Specifically, this invention uses a hierarchical evidence reasoning mechanism to output event categories, rather than directly providing a single classification result through a fully connected layer.

[0152] The first layer determines whether an event is a valid impact: following the output of robust background manifold separation and propagation mechanism dictionary unmixing: with the impact mechanism activation intensity as the core, combined with the spatiotemporal coherence of the impact saliency field and background confidence, regular traffic disturbances, environmental noise, and equipment anomalies are eliminated, and only events that conform to the physical propagation law of impact are retained, thus completing the initial screening.

[0153]

[0154] The second layer determines the specific type of effective impact: Based on the determination of an effective impact, it integrates the physical parameters of the impact contact process inversion, the component ratio of the propagation mechanism unmixing, the topological features of the impact footprint, and the road health background of the graph state space to match the physical prototypes of different impact types such as falling rocks, vehicle debris, and heavy object falling, and completes fine classification; at the same time, it realizes the rejection of unknown events through physical rule verification.

[0155]

[0156] The third level determines the impact risk level:

[0157] For each type of event, construct a physical prototype:

[0158] in, This represents the mean of the physical characteristics of this type of event. Represents the characteristic covariance. This represents the set of physical rules that this type of event should satisfy.

[0159] For the feature vector of the event to be identified Calculate its relationship to category Strength of evidence:

[0160] in, This represents a physical rule consistency function; it is used when an event characteristic violates the necessary physical conditions for that category. reduce.

[0161] Normalize the strength of various types of evidence into basic probability assignments:

[0162] in, This indicates evidence of an unknown event. When the maximum category of evidence is below a set threshold, or the evidence of an unknown event is above a set threshold, the system does not force a classification, but instead outputs "Unknown impact event" or "Event requiring manual review".

[0163] Risk level based on impact damage index Impact energy Peak impact force Scope of influence Overall judgment: .

[0164] in, Indicates a minor impact. Indicates a moderate impact. This indicates a severe impact. It indicates a critical or sudden crisis.

[0165] Furthermore, online self-updating and drift suppression.

[0166] Specifically, considering that temperature, humidity, road surface aging, fiber optic coupling status, and traffic background will continuously change during the long-term service of roads, this invention sets up an online self-updating mechanism.

[0167] For ordinary traffic events with high confidence and classified as non-damaging, the system uses them to update the background manifold and road segment coupling parameters; for high-risk impact events, unknown events, and equipment malfunction events, the system prohibits them from entering the background update queue to prevent the malfunctions from being absorbed as normal background.

[0168] Background updates employ a forgetting factor mechanism: .

[0169] in, For adaptive coefficient updates. When temperature changes are stable and traffic conditions are stable, Increase; when continuous anomalies, strong impacts, or equipment drift are detected, Decrease.

[0170] Road segment parameter updates use confidence gating: .

[0171] in, To update the gating coefficient: .

[0172] The system only allows road segment parameters to be updated when the probability of an event belonging to a normal background is higher than the threshold, the damage index is lower than the threshold, or the probability of an unknown event is lower than the threshold.

[0173] This invention employs hierarchical evidence reasoning classification and online drift suppression: a three-level hierarchical judgment mechanism is used to classify events and rate risks, and to support the rejection of unknown events; a gating update and forgetting factor mechanism is designed to achieve adaptive iteration and drift suppression of the model.

[0174] Furthermore, the output results are linked with early warning systems.

[0175] Specifically, the system ultimately outputs the following information: impact time; impact spatial location and location reliability; impact event type; peak impact force, impulse, contact duration and impact energy; road impact damage index; risk level; whether manual verification is required; and maintenance and treatment recommendations for the corresponding road section.

[0176] When the risk level is minor, the system only records the event and includes it in the road health record; when the risk level is moderate, the system sends inspection suggestions to the road maintenance platform; when the risk level is severe, the system triggers key inspections and video linkage; when the risk level is critical, the system triggers emergency alarms and outputs the impact location and suggested closure range to the traffic control platform.

[0177] Through the above technical solutions, this invention forms a complete technical path that differs from general vibration classification algorithms: such as Figure 3As shown, starting from the spatiotemporal phase response of DAS, through robust background manifold separation, non-Euclidean wavefront localization, road fiber-optic coupling self-calibration, impact contact force inversion, propagation mechanism dictionary unmixing, impact footprint topology modeling, physical graph state space reasoning, and evidence-based hierarchical classification, real-time monitoring, precise localization, physical quantification, type identification, and risk classification of road impact events are achieved.

[0178] To address the problems existing in the prior art, this invention establishes a complete physical transfer model from road impact mechanical parameters to DAS fiber strain signals, realizing the quantitative mapping between impact physical characteristics and sensing signals. To address complex road surface background interference, environmental and traffic noise are stripped away, and sudden impact signals are accurately extracted, improving signal recognition and anti-interference capabilities. By combining the propagation law of elastic waves and the coupling characteristics of optical fibers, impact location, parameter self-calibration, impact force and energy inversion are completed, and the impact intensity and road damage risk are quantified. Based on physical propagation mechanisms, topological features, and evidence reasoning, the system enables accurate classification and risk grading of various impact events, and completes online adaptive updates and graded early warnings to support differentiated and refined road maintenance decisions.

[0179] This invention deploys sensing optical fibers along the shoulder of the road surface, acquires backscattered Rayleigh phase signals using a Φ-OTDR device, constructs a DAS spatiotemporal response matrix, and performs segment unitization of the road, binding segment structure and fiber coupling parameters to complete segment unitization modeling. Phase expansion, differential demodulation, and anomalous jump correction are performed on the original phase signals to obtain a dynamic strain response matrix. Robust background manifold decomposition is used to split the response matrix into background manifold, sparse impact components, and random noise components. An impact saliency field is constructed based on the impact components, and candidate impact events are screened using spatiotemporal multi-dimensional judgment rules. Based on the elastic wavefront arrival time, non-Euclidean equivalent propagation distance, and a robust optimization model, the impact event is accurately located, and the impact location, time, wave velocity, and location confidence are output. Online self-calibration of the fiber coupling coefficient, propagation attenuation coefficient, and elastic wave velocity for the entire road segment is achieved using a standard calibration source. An impact contact process inversion model is constructed to solve the impact contact force time history, calculate physical quantities such as peak impact force, impulse, impact energy, and contact duration, and establish a pavement impact damage index to assess damage risk. A dictionary of multiple disturbance propagation mechanisms is constructed, and a non-negative group sparse algorithm is used to demix mixed signals, distinguishing different signal sources such as impact, traffic, environment, and construction. Multi-threshold binarization is applied to the impact saliency field to extract the topological feature vector of the impact footprint, enabling differentiation of different impact event morphologies. A physical constraint graph state-space model is constructed using road segment units as graph nodes, fusing multi-dimensional features and combining them with physical rule constraints to analyze the instantaneous impact and long-term health status of roads. A hierarchical evidence reasoning mechanism is employed to classify major event categories, determine sub-categories of impacts, and rate risk levels. Unknown impact events are rejected and prompted for manual review. An online self-updating mechanism with confidence gating and a forgetting factor is designed to distinguish between normal and risky events, achieving adaptive model updates and signal drift suppression. Monitoring results are output based on the impact risk level, linking maintenance and traffic platforms to achieve tiered early warning and emergency response.

[0180] This invention provides a method for real-time detection, location, and intelligent classification of sudden impact events (such as rockfalls, falling objects from vehicles, and falling heavy objects) on highway and bridge pavements using distributed fiber optic acoustic sensing technology. An algorithm system specifically designed for sudden impact events on road surfaces is established, based on the framework of "road impact physical mechanism – elastic wavefront propagation – road fiber optic coupling inversion – impact mechanism evidence reasoning." The core idea is to first recover the shock wavefront, propagation path, and energy attenuation patterns from the DAS spatiotemporal phase response; then invert the impact contact parameters and road damage risk; and finally, classify, locate, and issue graded early warnings based on physical evidence rather than simply image features.

[0181] Figure 4This is a block diagram illustrating a road impact event monitoring and classification device based on distributed fiber optic acoustic sensing, according to an exemplary embodiment. The device is used in a method for monitoring and classifying road impact events based on distributed fiber optic acoustic sensing. (Refer to...) Figure 4 The device includes a data acquisition module 310, an impact candidate event screening module 320, a localization module 330, an inversion module 340, a demixing and feature extraction module 350, a graph state space model construction module 360, and an output module 370. Among them: The data acquisition module 310 is used to acquire the back Rayleigh scattering phase signal along the optical fiber through a phase-sensitive optical time-domain reflectometry device, and to obtain the dynamic strain response matrix by performing phase expansion, differential demodulation and abnormal jump correction on the back Rayleigh scattering phase signal; the road surface is divided into several road segment units along the optical fiber direction, and a corresponding road segment parameter vector is bound to each road segment unit; wherein, the optical fiber is laid along the monitoring area.

[0182] The impact candidate event screening module 320 is used to decompose the dynamic strain response matrix using a robust background manifold decomposition method to obtain sparse impact components, construct an impact saliency field based on the sparse impact components, and screen impact candidate events from the dynamic strain response matrix based on whether the impact saliency field meets preset judgment conditions within a preset window.

[0183] The positioning module 330 is used to locate the impact candidate event based on the elastic wave front arrival time, non-Euclidean equivalent propagation distance and robust optimization model, and obtain the positioning result; it also acquires the calibration source and performs adaptive calibration of the road surface fiber coupling coefficient, propagation attenuation coefficient and road segment elastic wave equivalent propagation speed in the road segment parameter vector based on the calibration source.

[0184] The inversion module 340 is used to determine the observation signal representation of each fiber channel within the window corresponding to the impact candidate event based on the road segment parameter vector, perform inversion on the impact contact force time history in the observation signal representation through a constrained optimization model, calculate the impact physical quantity based on the inversion result, and construct the road impact damage index.

[0185] The unmixing and feature extraction module 350 is used to construct a propagation mechanism dictionary, determine the spatiotemporal response representation within the window corresponding to the impact candidate event based on the propagation mechanism dictionary and the road segment parameter vector, and perform mixed signal unmixing on the spatiotemporal response representation through a non-negative group sparse algorithm to obtain the physical mechanism of the impact candidate event; and perform multi-threshold binarization on the impact saliency field to obtain the impact footprint topological feature vector to determine the impact event type.

[0186] The graph state space model construction module 360 ​​is used to construct a road physical graph state space model by taking each road segment unit as a node, and taking the impact physical quantity, road impact damage index and impact footprint topological feature vector as node features, and setting physical rule constraints. Based on the road physical graph state space model, the instantaneous impact and long-term health status of the road surface are analyzed.

[0187] The output module 370 is used to determine the risk level based on the impact physical quantity and the road impact damage index, and to provide graded early warning based on the risk level.

[0188] This invention provides a method for real-time detection, location, and intelligent classification of sudden impact events (such as rockfalls, falling objects from vehicles, and falling heavy objects) on highway and bridge pavements using distributed fiber optic acoustic sensing technology. An algorithm system specifically designed for sudden impact events on road surfaces is established, based on the framework of "road impact physical mechanism – elastic wavefront propagation – road fiber optic coupling inversion – impact mechanism evidence reasoning." The core idea is to first recover the shock wavefront, propagation path, and energy attenuation patterns from the DAS spatiotemporal phase response; then invert the impact contact parameters and road damage risk; and finally, classify, locate, and issue graded early warnings based on physical evidence rather than simply image features.

[0189] Figure 5 This is a schematic diagram of the structure of a road impact event monitoring and classification device provided in an embodiment of the present invention, as shown below. Figure 5 As shown, road impact event monitoring and classification equipment may include the above-mentioned... Figure 4 The illustrated pavement impact event monitoring and classification device is based on distributed fiber optic acoustic sensing. Optionally, the pavement impact event monitoring and classification device 410 may include a first processor 2001.

[0190] Optionally, the road impact event monitoring and classification device 410 may also include a memory 2002 and a transceiver 2003.

[0191] The first processor 2001, memory 2002, and transceiver 2003 can be connected via a communication bus.

[0192] The following is combined Figure 5 A detailed description of each component of the road impact event monitoring and classification equipment 410 is provided below: The first processor 2001 is the control center of the road impact event monitoring and classification device 410. It can be a single processor or a collective term for multiple processing elements. For example, the first processor 2001 can be one or more central processing units (CPUs), application-specific integrated circuits (ASICs), or one or more integrated circuits configured to implement embodiments of the present invention, such as one or more digital signal processors (DSPs), or one or more field-programmable gate arrays (FPGAs).

[0193] Optionally, the first processor 2001 can perform various functions of the road impact event monitoring and classification device 410 by running or executing software programs stored in the memory 2002 and calling data stored in the memory 2002.

[0194] In a specific implementation, as one example, the first processor 2001 may include one or more CPUs, for example... Figure 5 CPU0 and CPU1 are shown in the diagram.

[0195] In a specific implementation, as one example, the road impact event monitoring and classification device 410 may also include multiple processors, for example... Figure 5 The first processor 2001 and the second processor 2004 are shown in the diagram. Each of these processors can be a single-core processor or a multi-core processor. Here, a processor can refer to one or more devices, circuits, and / or processing cores used to process data (such as computer program instructions).

[0196] The memory 2002 is used to store the software program that executes the present invention, and is controlled by the first processor 2001 to execute it. The specific implementation method can be referred to the above method embodiment, and will not be repeated here.

[0197] Optionally, the memory 2002 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto. The memory 2002 may be integrated with the first processor 2001 or may exist independently, and may be connected via the interface circuit of the road impact event monitoring and classification device 410. Figure 5 (Not shown in the image) is coupled to the first processor 2001, and this embodiment of the invention does not specifically limit this.

[0198] The transceiver 2003 is used to communicate with network devices or with terminal devices.

[0199] Alternatively, transceiver 2003 may include a receiver and a transmitter. Figure 5 (Not shown separately). The receiver is used to implement the receiving function, and the transmitter is used to implement the transmitting function.

[0200] Optionally, the transceiver 2003 can be integrated with the first processor 2001, or it can exist independently and be connected to the interface circuit of the road impact event monitoring and classification device 410. Figure 5 (Not shown in the image) is coupled to the first processor 2001, and this embodiment of the invention does not specifically limit this.

[0201] It should be noted that, Figure 5 The structure of the road impact event monitoring and classification device 410 shown in the figure does not constitute a limitation on the router. The actual knowledge structure identification device may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0202] Furthermore, the technical effect of the road impact event monitoring and classification device 410 can be referred to the technical effect of the road impact event monitoring and classification method based on distributed fiber optic acoustic sensing described in the above method embodiments, and will not be repeated here.

[0203] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for monitoring and classifying road impact events based on distributed fiber optic acoustic sensing, characterized in that, The method includes: S1. The backscattered Rayleigh light phase signal along the optical fiber is acquired by a phase-sensitive optical time-domain reflectometry device. The backscattered Rayleigh light phase signal is subjected to phase expansion, differential demodulation and abnormal jump correction to obtain a dynamic strain response matrix. The road surface is divided into several road segment units along the direction of the optical fiber, and a corresponding road segment parameter vector is bound to each road segment unit. The optical fiber is laid along the monitoring area. S2. The dynamic strain response matrix is ​​decomposed using the robust background manifold decomposition method to obtain sparse impact components. An impact saliency field is constructed based on the sparse impact components. Based on whether the impact saliency field meets the preset judgment conditions within a preset window, impact candidate events are selected from the dynamic strain response matrix. S3. Based on the elastic wave front arrival time, non-Euclidean equivalent propagation distance and robust optimization model, the impact candidate event is located and the location result is obtained; the calibration source is obtained, and the road surface fiber coupling coefficient, propagation attenuation coefficient and road segment elastic wave equivalent propagation speed in the road segment parameter vector are adaptively calibrated according to the calibration source. S4. Determine the observation signal representation of each fiber channel within the window corresponding to the impact candidate event based on the road segment parameter vector. Invert the impact contact force time history in the observation signal representation using a constrained optimization model. Calculate the impact physical quantity based on the inversion result and construct the road impact damage index. S5. Construct a propagation mechanism dictionary. Based on the propagation mechanism dictionary and the road segment parameter vector, determine the spatiotemporal response representation within the window corresponding to the impact candidate event. Use a non-negative group sparse algorithm to perform mixed signal demixing on the spatiotemporal response representation to obtain the physical mechanism of the impact candidate event. Perform multi-threshold binarization on the impact saliency field to obtain the impact footprint topological feature vector and determine the impact event type. S6. Take each road segment unit as a node, and take the impact physical quantity, road impact damage index and impact footprint topological feature vector as node features to construct a road physical map state space model and set physical rule constraints. Analyze the instantaneous impact and long-term health status of the road based on the road physical map state space model. S7. Determine the risk level based on the impact physical quantity and road impact damage index, and conduct graded early warning based on the risk level.

2. The method for monitoring and classifying road impact events based on distributed fiber optic acoustic sensing according to claim 1, characterized in that, The S1 further includes: Combining Hertzian contact theory, elastic wave propagation law and fiber optic sensing principle, a physical transmission model of impact signal is constructed. The physical transmission model of impact signal includes a road impact mechanics time history model, a pavement elastic wave propagation model and a fiber optic phase response model. Among them, the road impact mechanics time history model uses Hertz contact theory to describe the impact force time history function, as shown in the following equation (1): （1） In the formula, Indicates any moment during the impact process The corresponding instantaneous impact force Indicates the instantaneous time during the impact process. Indicates peak impact force. Indicates the waveform shape index. Indicates the duration of impact contact; The propagation model of elastic waves on the pavement is shown in equation (2) below: （2） In the formula, Indicates the radial distance of wave propagation Place Road surface response at all times This represents the radial distance from the observation point to the impact point on the surface. Indicates the peak strain at the impact point. Indicates the reference distance. This represents the attenuation coefficient of asphalt mixture pavement for elastic waves. Function representing the impact waveform. Indicates the Rayleigh wave velocity of asphalt pavement; The fiber optic phase response model is used to describe the relationship between the phase change rate and the road surface strain, as shown in equation (3) below: （3） In the formula, This indicates the phase of the backscattered Rayleigh light in the optical fiber. Represents the longitudinal coordinate along the optical fiber. Indicates the refractive index of the fiber core. Indicates the laser wavelength. This represents the fiber strain coefficient. Indicates the position along the longitudinal direction of the optical fiber. place, time Axial strain acting on the optical fiber.

3. The method for monitoring and classifying road impact events based on distributed fiber optic acoustic sensing according to claim 1, characterized in that, The road segment parameter vector in S1 includes: road surface structure parameters and optical fiber coupling parameters; Among them, the pavement structure parameters and optical fiber coupling parameters include pavement layer thickness, optical fiber burial depth, lateral distance from the optical fiber to the potential impact area, pavement temperature, pavement optical fiber coupling coefficient, and equivalent propagation speed of elastic waves in the road section.

4. The method for monitoring and classifying road impact events based on distributed fiber optic acoustic sensing according to claim 1, characterized in that, The S3 includes: For each fiber channel within the window corresponding to the impact candidate event, the arrival time of the elastic wave front of the fiber channel is calculated based on the arrival time scoring function. The arrival time relationship between the impact point and the fiber channel is obtained based on the arrival time of the elastic wave front and the non-Euclidean equivalent propagation distance. The location result of the impact candidate event is obtained based on the arrival time relationship and the robust optimization model. The location result includes the impact occurrence time, impact location, wave velocity estimate, and location confidence. The time relationship between the impact point and the optical fiber channel is expressed as shown in equation (4): （4） In the formula, Indicates the first The arrival time of the elastic wavefront at each fiber optic channel. This indicates the spatial index of the fiber optic channel, corresponding to the [number]th [channel] along the fiber optic deployment direction. Each sampling point Indicates the actual start time of the impact event. Indicates the road segment parameters The corrected equivalent propagation distance of the road surface. Indicates the first The spatial coordinates of the optical fiber corresponding to each optical fiber channel. This represents the spatial coordinates of the point where the impact event occurred. This represents the equivalent wave velocity of the corresponding road segment. This indicates measurement error.

5. The method for monitoring and classifying road impact events based on distributed fiber optic acoustic sensing according to claim 1, characterized in that, The observation signal representation for each fiber optic channel in S4 is shown in equation (5) below: （5） In the formula, Indicates the first The observation signal representation of each fiber optic channel, Represents a time variable. Indicates the fiber optic coupling coefficient of the road surface. Indicates the propagation from the point of impact to the... The road surface propagation kernel function for each channel. Indicates the spatial location of the impact source. This indicates the equivalent propagation speed of elastic waves in the road section. Indicates the propagation attenuation coefficient. Indicates the thickness of the paving layer. This represents the convolution operation. Represents the impact contact force time history. Indicates noise.

6. The method for monitoring and classifying road impact events based on distributed fiber optic acoustic sensing according to claim 1, characterized in that, The impact physical quantities in S4 include: peak impact force, impulse, impact energy, and contact duration; The road impact damage index is used to quantitatively assess the impact intensity and structural damage risk, as shown in equation (6) below: （6） In the formula, Indicates the road surface impact damage index. Indicates the weighting coefficient. Indicates peak impact force. This indicates the peak impact force reference safety threshold. Indicates the weighting coefficient. Indicates impact energy. This indicates the reference safety threshold for impact energy. Indicates the weighting coefficient. Indicates impulse. This indicates the impulse reference safety threshold. Indicates the weighting coefficient. This represents the elastic wave propagation attenuation coefficient of the corresponding road segment under the current condition. This represents the attenuation coefficient under the historical health conditions of a road segment. Indicates the weighting coefficient. This represents the change in wave speed. This indicates the wave velocity during the historical healthy state of the road segment.

7. The method for monitoring and classifying road impact events based on distributed fiber optic acoustic sensing according to claim 1, characterized in that, The impact footprint topological feature vector in S5 is shown in the following equation (7): （7） In the formula, Represents the topological feature vector of the impact footprint. Indicates the number of major connected components. Indicates the length of the space extension. Indicates the duration of time. Indicates the curvature of the propagation ridge. This indicates the asymmetry of left-right propagation. Indicates the lifecycle of the connected component. This indicates the complexity of the impact footprint hierarchy.

8. A road impact event monitoring and classification device based on distributed fiber optic acoustic sensing, wherein the road impact event monitoring and classification device based on distributed fiber optic acoustic sensing is used to implement the road impact event monitoring and classification method based on distributed fiber optic acoustic sensing as described in any one of claims 1-7, characterized in that, The device includes: The data acquisition module is used to acquire the backscattered Rayleigh light phase signal along the optical fiber using a phase-sensitive optical time-domain reflectometry device, and to obtain the dynamic strain response matrix by performing phase expansion, differential demodulation and abnormal jump correction on the backscattered Rayleigh light phase signal; the road surface is divided into several road segment units along the direction of the optical fiber, and a corresponding road segment parameter vector is bound to each road segment unit; wherein, the optical fiber is laid along the monitoring area. The impact candidate event screening module is used to decompose the dynamic strain response matrix using the robust background manifold decomposition method to obtain sparse impact components, construct an impact saliency field based on the sparse impact components, and screen impact candidate events from the dynamic strain response matrix based on whether the impact saliency field meets preset judgment conditions within a preset window. The positioning module is used to locate impact candidate events based on the elastic wave front arrival time, non-Euclidean equivalent propagation distance, and robust optimization model to obtain positioning results; and to acquire calibration sources to achieve adaptive calibration of the road surface fiber coupling coefficient, propagation attenuation coefficient, and equivalent propagation speed of elastic waves in the road segment parameter vector based on the calibration sources. The inversion module is used to determine the observation signal representation of each fiber channel within the window corresponding to the impact candidate event based on the road segment parameter vector, invert the impact contact force time history in the observation signal representation through a constrained optimization model, calculate the impact physical quantity based on the inversion result, and construct the road impact damage index. The unmixing and feature extraction module is used to construct a propagation mechanism dictionary, determine the spatiotemporal response representation within the window corresponding to the impact candidate event based on the propagation mechanism dictionary and the road segment parameter vector, and perform mixed signal unmixing on the spatiotemporal response representation through a non-negative group sparse algorithm to obtain the physical mechanism of the impact candidate event; and perform multi-threshold binarization on the impact saliency field to obtain the impact footprint topological feature vector to determine the impact event type. The graph state space model construction module is used to construct a road physical graph state space model by taking each road segment unit as a node, and taking the impact physical quantity, road impact damage index and impact footprint topological feature vector as node features, and setting physical rule constraints. Based on the road physical graph state space model, the instantaneous impact and long-term health status of the road surface are analyzed. The output module is used to determine the risk level based on the impact physical quantity and the road impact damage index, and to provide graded early warnings based on the risk level.

9. A road impact event monitoring and classification device, characterized in that, The road impact event monitoring and classification equipment includes: processor; A memory storing computer-readable instructions that, when executed by the processor, implement the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium contains program code that can be invoked by a processor to execute the method as described in any one of claims 1 to 7.

Images