Rail structure damage identification method and system based on combined light-emitting intelligent skin

Abstract

The application discloses a track structure damage identification method and system based on a combination type light-emitting intelligent skin, and the method comprises the following steps: acquiring a plurality of unit area images of a track structure covered by the intelligent skin; screening out a damaged unit area image from the plurality of unit area images through a clustering algorithm; performing deep identification on the damaged unit area image through a pre-obtained adaptive multi-task sparse digital image correlation model to obtain a damage sensing position and a corresponding light-emitting value of the damaged unit area image; inputting the damage sensing position and the corresponding light-emitting value of the damaged unit area image into a pre-obtained intelligent skin calibration function model to calculate strain data of a damage position of the track structure; and analyzing the strain data to obtain a damage detection result of the track structure. The application realizes track structure damage detection and early warning based on the intelligent skin, and can conveniently and effectively detect track structure damage.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of structural damage perception, in particular to a rail structure damage identification method and system based on a combined light-emitting smart skin. BACKGROUND

[0002] The smart skin is a powerful tool for stress distribution perception and early damage occurrence and development of vulnerable structures. As a new and important technology for structural health monitoring, the smart skin has the characteristics of light weight, wide coverage and good flexibility. The research on smart skin is very rich, but mainly focuses on large-scale circuit and lead design, manufacturing and performance improvement of electrical smart skin. The traditional smart skin mainly relies on the design and manufacturing of large-scale circuit and lead, which greatly reduces the applicability of such smart skin for dynamically loaded railway facilities such as trains and rails.

[0003] Therefore, how to conveniently detect the damage of rail transit structure based on the smart skin is a problem to be solved at present. SUMMARY

[0004] The main purpose of the present application is to provide a rail structure damage identification method and system based on a combined light-emitting smart skin, which aims to solve the problem of how to conveniently detect the damage of rail transit structure based on the smart skin.

[0005] To achieve the above-mentioned purpose, the present application provides a rail structure damage identification method based on a combined light-emitting smart skin, which is applied to a rail structure damage identification system based on a combined light-emitting smart skin, the rail structure damage identification system based on a combined light-emitting smart skin at least includes a smart skin of a force-induced light-emitting material and a periodic grid unit structure, a high-speed camera and a damage identification module, and the rail structure damage identification method based on a combined light-emitting smart skin includes the following steps:

[0006] Obtain a plurality of unit area images of the rail structure covered by the smart skin;

[0007] Filter out a damage unit area image from the plurality of unit area images through a clustering algorithm;

[0008] Perform deep identification on the damage unit area image through a pre-obtained adaptive multi-task sparse digital image correlation model to obtain a damage sensing position and a corresponding light-emitting value of the damage unit area image;

[0009] Input the damage sensing position and the corresponding light-emitting value of the damage unit area image into a pre-obtained smart skin calibration function model to calculate strain data of a damage position of the rail structure;

[0010] analyzing the strain data to obtain a damage detection result of the track structure;

[0011] In the adaptive multi-task sparse digital image correlation model, A is a reference crack feature matrix, γ is a feature matrix of the damage unit area image, C is a sparse correlation mapping matrix, containing the correlation coefficient of the reference crack feature matrix A and the feature matrix γ of the damage unit area image, λ is an adaptive adjustment factor, and k is an iteration step number. The adaptive multi-task sparse digital image correlation model is as follows:

[0012]

[0013] s.t.C k ≥0

[0014]

[0015] In the intelligent skin calibration function model, T is the environmental temperature, P is the background light intensity, ΔL k is the light intensity change amount of the kth step, Δt k is the time increment of the kth step, and Δξ k is the strain data change amount of the kth step. The intelligent skin calibration function model is as follows:

[0016] Δξ k =C1(T,P)(Δt k +log||ΔL k +Δξ (k-1) |-log|Δξ (k-1 )|)+C2(T,P)(ΔL k +log|Δt k |+C3(T,P)(ξ (k-1) -ξ (k-1) ΔL k -log|ξ (k-1) ΔL k |)+C4(T,P)。

[0017] Optionally, the step of screening the damage unit area image from the plurality of unit area images by using the clustering algorithm comprises the following steps.

[0018] According to the unit area image, a unit area image feature is extracted.

[0019] The unit area image features are clustered into a plurality of different classes of unit area image feature sets by using a clustering algorithm.

[0020] The damage unit area image is screened from the plurality of different classes of unit area image feature sets.

[0021] Optionally, before the step of inputting the damage-induced position of the unit area image of the damage and the corresponding luminescence value into the pre-obtained intelligent skin calibration function model to calculate the strain data of the damage position of the track structure, a laboratory calibration is further required, and the laboratory calibration step comprises:

[0022] obtaining the luminescence value corresponding to the damage-induced position under different preset temperatures and different preset background light intensities;

[0023] selecting any one group of data from the different preset temperatures, different preset background light intensities and corresponding luminescence values to establish a first functional relationship between the temperature, background light intensity and corresponding luminescence value of the damage-induced position;

[0024] calibrating the first functional relationship in sequence according to the different preset temperatures, different background light intensities and corresponding luminescence values to obtain an intelligent skin calibration function model.

[0025] Optionally, the step of analyzing the strain data to obtain the damage detection result of the track structure comprises:

[0026] performing damage and trend change analysis on the strain data to obtain an analysis result of the strain data;

[0027] comparing the analysis result with a risk evaluation index threshold to obtain the damage detection result of the track structure.

[0028] Optionally, after the step of comparing the analysis result with the risk evaluation index threshold to obtain the damage detection result of the track structure, the method further comprises:

[0029] performing corresponding over-limit damage early warning according to the damage detection result of the track structure.

[0030] Optionally, the step of obtaining a plurality of unit area images of the track structure covered by the intelligent skin comprises:

[0031] obtaining a plurality of unit area images of the track structure covered by the intelligent skin through the high-speed camera.

[0032] Optionally, the step of establishing the first functional relationship between the temperature, background light intensity and corresponding luminescence value of the damage-induced position comprises:

[0033] establishing the first functional relationship between the temperature, background light intensity and corresponding luminescence value of the damage-induced position through regression modeling and / or curve fitting.

[0034] Optionally, the smart skin comprises a base layer, the base layer being a flexible material with a periodic partition; a surface sealing layer, the surface sealing layer being arranged on one side of the base layer and enclosing a cavity with the base layer; and a sensing layer, the sensing layer being arranged in the cavity and being a mechanoluminescent material.

[0035] The application also provides a track structure damage identification system based on the combined light-emitting smart skin, which comprises a smart skin of a mechanoluminescent material and a periodic grid unit structure, a high-speed camera, and a damage identification module.

[0036] The high-speed camera is configured to acquire a plurality of unit region images of a track structure covered by the smart skin.

[0037] The damage identification module is configured to filter out a damage unit region image from the plurality of unit region images by using a clustering algorithm, to perform deep identification on the damage unit region image by using a pre-obtained adaptive multi-task sparse digital image correlation model, to obtain a damage sensing position and a corresponding light-emitting value of the damage unit region image, to input the damage sensing position and the corresponding light-emitting value of the damage unit region image into a pre-obtained smart skin calibration function model, to calculate strain data of a damage position of the track structure, and to analyze the strain data to obtain a damage detection result of the track structure.

[0038] In the adaptive multi-task sparse digital image correlation model, A is a reference crack feature matrix, Y is a feature matrix of the damage unit region image, C is a sparse correlation mapping matrix, which contains a correlation coefficient of the reference crack feature matrix A and the feature matrix Y of the damage unit region image, λ is an adaptive adjustment factor, and k is an iteration step number. The adaptive multi-task sparse digital image correlation model is shown in the following formula.

[0039]

[0040] s.t.C k ≥0

[0041]

[0042] In the smart skin calibration function model, T is an environmental temperature, P is a background light intensity, ΔL k is a light intensity change amount of the kth step, Δt k is a time increment of the kth step, and Δξ k is a strain data change amount of the kth step. The smart skin calibration function model is shown in the following formula.

[0043] Δξ k =C1(T,P)(Δtk + log | AL k + Δξ (k-1) - log | Δξ (k-1) |) + C2(T, P)(AL k + log | At k |) + C3(T, P)(ξ (k-1) - ξ (k-1) AL k - log | ξ (k-1 ) AL k |) + C4(T, P).

[0044] The embodiment of the present application also provides a terminal device, which comprises a memory, a processor and a track structure damage identification program based on a combined light-emitting smart skin stored in the memory and running on the processor, and the track structure damage identification program based on the combined light-emitting smart skin realizes the steps of the track structure damage identification method based on the combined light-emitting smart skin when executed by the processor.

[0045] The embodiment of the present application also provides a storage medium, which stores a track structure damage identification program based on a combined light-emitting smart skin, and the track structure damage identification program based on the combined light-emitting smart skin realizes the steps of the track structure damage identification method based on the combined light-emitting smart skin when executed by the processor.

[0046] The track structure damage identification method based on the combined light-emitting smart skin provided by the present application is applied to a track structure damage identification system based on the combined light-emitting smart skin, which at least comprises a smart skin of a force-induced light-emitting material and a periodic grid unit structure, a high-speed camera and a damage identification module. The track structure damage identification method based on the combined light-emitting smart skin comprises the following steps: acquiring a plurality of unit region images of a track structure covered by the smart skin; screening a damage unit region image from the plurality of unit region images through a clustering algorithm; performing deep identification on the damage unit region image through a pre-obtained adaptive multi-task sparse digital image correlation model to obtain a damage sensing position and a corresponding light-emitting value of the damage unit region image; inputting the damage sensing position and the corresponding light-emitting value of the damage unit region image into a pre-obtained smart skin calibration function model to calculate strain data of a damage position of the track structure; and analyzing the strain data to obtain a damage detection result of the track structure. Based on the present application, track structure damage detection and early warning based on the smart skin are realized, and early damage of the track structure can be conveniently and effectively detected. BRIEF DESCRIPTION OF DRAWINGS

[0047] Figure 1A functional module schematic diagram of a terminal device to which the track structure damage identification device based on the combination type light-emitting intelligent skin of the present application belongs;

[0048] Figure 2 A flowchart schematic diagram of the first exemplary embodiment of the intelligent skin damage detection method of the present application;

[0049] Figure 3 A flowchart schematic diagram of the second exemplary embodiment of the intelligent skin damage detection method of the present application;

[0050] Figure 4 A flowchart schematic diagram of the third exemplary embodiment of the intelligent skin damage detection method of the present application;

[0051] Figure 5 A flowchart schematic diagram of the fourth exemplary embodiment of the intelligent skin damage detection method of the present application.

[0052] The implementation, functional features and advantages of the present application will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION

[0053] It should be understood that the specific implementation cases described herein are only used to explain the present application, and are not used to limit the present application.

[0054] The main solution of the embodiments of the present application is to obtain a plurality of unit area images of a track structure covered by the intelligent skin; to filter out a damage unit area image from the plurality of unit area images through a clustering algorithm; to obtain a damage sensing position and a corresponding light-emission value of the damage unit area image through a pre-obtained adaptive multi-task sparse digital image correlation model for deep identification of the damage unit area image; to input the damage sensing position and the corresponding light-emission value of the damage unit area image into a pre-obtained intelligent skin calibration function model to calculate strain data of a damage position of the track structure; and to analyze the strain data to obtain a damage detection result of the track structure. Based on the present solution, track structure damage detection and early warning based on the intelligent skin are realized, and early damage of the track structure is conveniently and effectively detected.

[0055] Specifically, referring to Figure 1 , Figure 1 A functional module schematic diagram of a terminal device to which the track structure damage identification device based on the combination type light-emitting intelligent skin of the present application belongs. The track structure damage identification device based on the combination type light-emitting intelligent skin is a device based on a terminal device, which can conveniently detect object damage and thus warn relevant personnel, and can be carried on the terminal device in the form of hardware or software.

[0056] In the embodiment, the terminal device of the track structure damage identification device based on the combined light-emitting intelligent skin at least includes an output module 110, a processor 120, a memory 130, and a communication module 140.

[0057] The memory 130 stores an operating system and a track structure damage identification program based on the combined light-emitting intelligent skin. The track structure damage identification device based on the combined light-emitting intelligent skin can obtain a plurality of unit area images of the track structure covered by the intelligent skin; filter out damaged unit area images from the plurality of unit area images through a clustering algorithm; perform deep identification on the damaged unit area images through a pre-obtained adaptive multi-task sparse digital image correlation model, obtain the damage sensing position and the corresponding light-emitting value of the damaged unit area images; input the damage sensing position and the corresponding light-emitting value of the damaged unit area images into a pre-obtained intelligent skin calibration function model, calculate the strain data of the damage position of the track structure; analyze the strain data to obtain the damage detection result of the track structure, and store the information in the memory 130; and the output module 110 can be a display screen or the like. The communication module 140 can include a WIFI module, a mobile communication module, a Bluetooth module, and the like, and communicate with external terminal devices or servers through the communication module 140.

[0058] The track structure damage identification program based on the combined light-emitting intelligent skin in the memory 130 realizes the following steps when the track structure damage identification program based on the combined light-emitting intelligent skin is executed by the processor:

[0059] Obtain a plurality of unit area images of the track structure covered by the intelligent skin;

[0060] Filter out damaged unit area images from the plurality of unit area images through a clustering algorithm;

[0061] Perform deep identification on the damaged unit area images through a pre-obtained adaptive multi-task sparse digital image correlation model, obtain the damage sensing position and the corresponding light-emitting value of the damaged unit area images;

[0062] Input the damage sensing position and the corresponding light-emitting value of the damaged unit area images into a pre-obtained intelligent skin calibration function model, calculate the strain data of the damage position of the track structure;

[0063] Analyze the strain data to obtain the damage detection result of the track structure;

[0064] A is a reference crack feature matrix, Y is a feature matrix of the damaged unit area image, C is a sparse correlation mapping matrix, which contains the correlation coefficient of the reference crack feature matrix A and the feature matrix Y of the damaged unit area image, lambda is an adaptive adjustment factor, k is an iteration step number, and the adaptive multi-task sparse digital image correlation model formula is as follows:

[0065]

[0066] st.C>=0

[0067]

[0068] In the intelligent skin calibration function model, T is the environmental temperature, P is the background light intensity, and Delta L k is the light intensity change amount of the kth step, Delta t k is the time increment of the kth step, Delta xi k is the strain data change amount of the kth step, and the intelligent skin calibration function model formula is as follows:

[0069] Delta xi k =C1(T,P)(Delta t k +log|Delta L k +Delta xi (k-1) |-log|Delta xi (k-1) |)+C2(T,P)(Delta L k +log Delta t k |)+C3(T,P)(xi (k-1) -xi (k-1) Delta L k -log|xi (k-1) Delta L k |)+C4(T,P)

[0070] The intelligent skin includes a base layer made of flexible material with periodic partitions, a surface sealing layer arranged on one side of the base layer and enclosing a cavity with the base layer, and a sensing layer arranged in the cavity and made of mechanoluminescent material.

[0071] Further, the track structure damage identification program based on the combined light-emitting intelligent skin in the memory 130 is further implemented when the processor executes the following steps:

[0072] According to the unit area image, the unit area image features are extracted;

[0073] The unit area image features are clustered into several different unit area image feature sets through a clustering algorithm;

[0074] Screening the damaged unit area image from the several different types of unit area image feature sets.

[0075] Further, the track structure damage identification program based on the combined light-emitting intelligent skin in the memory 130 is also implemented when the processor is executed to realize the following steps:

[0076] Obtain the light-emission value corresponding to the damage sensing position under several groups of different preset temperatures and different preset background light intensities.

[0077] Select any group of data from the several groups of different preset temperatures, different preset background light intensities, and corresponding light-emission values to establish the first functional relationship between the temperature, background light intensity, and corresponding light-emission value of the damage sensing position.

[0078] According to the several groups of different preset temperatures, different background light intensities, and corresponding light-emission values, the first functional relationship is calibrated in sequence to obtain the intelligent skin calibration function model.

[0079] Further, the track structure damage identification program based on the combined light-emitting intelligent skin in the memory 130 is also implemented when the processor is executed to realize the following steps:

[0080] Perform damage and trend change analysis on the strain data to obtain the analysis result of the strain data.

[0081] Compare the analysis result with the risk evaluation index threshold to obtain the damage detection result of the track structure.

[0082] Further, the track structure damage identification program based on the combined light-emitting intelligent skin in the memory 130 is also implemented when the processor is executed to realize the following steps:

[0083] According to the damage detection result of the track structure, perform corresponding over-standard damage early warning.

[0084] Further, the track structure damage identification program based on the combined light-emitting intelligent skin in the memory 130 is also implemented when the processor is executed to realize the following steps:

[0085] Obtain several unit area images of the track structure covered by the intelligent skin through the high-speed camera.

[0086] Further, the track structure damage identification program based on the combined light-emitting intelligent skin in the memory 130 is also implemented when the processor is executed to realize the following steps:

[0087] Establish the first functional relationship between the temperature, background light intensity, and corresponding light-emission value of the damage sensing position through regression modeling and / or curve fitting.

[0088] The method embodiments of the present application are based on, but not limited to, the terminal device architecture described above.

[0089] Referring to Figure 2 , Figure 2 FIG. 1 is a flowchart of a first exemplary embodiment of a track structure damage identification method based on a combination-type luminescent smart skin. The track structure damage identification method based on the combination-type luminescent smart skin is applied to a track structure damage identification system based on the combination-type luminescent smart skin, which at least includes a smart skin of a force-induced luminescent material and a periodic grid unit structure, a high-speed camera, and a damage identification module. The track structure damage identification method based on the combination-type luminescent smart skin includes the following steps:

[0090] Specifically, the smart skin includes a base layer made of a flexible material with a periodic partition, a surface sealing layer arranged on one side of the base layer and enclosing a cavity with the base layer, and a sensing layer arranged in the cavity and being a force-induced luminescent material, which is a sensor. The stress distribution smart skin based on the force-induced luminescent material and the periodic grid unit structure has strong weather resistance. The periodic grid unit structure can be a honeycomb grid unit, and the boundaries of the grid unit help to perceive and distinguish signals, locate damage positions of the track structure, and enhance the extensibility of the smart skin. The periodic grid unit structure and the surface sealing layer are designed to enhance the practicability and weather resistance of the smart skin. The smart skin can effectively avoid the design and manufacturing problems of large-scale signal leads and can realize early identification of railway track stress distribution changes, fatigue cracks, and tracking and reproduction of expansion deterioration. The high-speed camera includes a near-infrared camera, a solar-blind violet camera, and a hyperspectral camera, which supports daytime and nighttime smart skin luminescence shooting. The near-infrared camera and the solar-blind violet camera have more advantages than ordinary cameras in daytime shooting.

[0091] A vibration sensor can be arranged on the camera mounting device. When the track is subjected to different train loads, the vibration sensor detects the track vibration, and when the vibration exceeds a threshold value, it indicates that the train is about to pass through the measured area, and the high-speed camera starts shooting. The smart skin in the measured area is subjected to different stress and strain, and emits light of different wavelengths and intensities. At this time, several unit area images of the track structure covered by the smart skin can be shot, so as to determine the damage sensing position and the damage degree according to the luminescence intensity of different positions of the unit area images in the subsequent process; optionally, the measured object is generally a steel rail structure.

[0092] In step S120, a damage unit area image is selected from the several unit area images by a clustering algorithm.

[0093] Specifically, clustering algorithms can classify images of unit regions. Based on the different image features, images with similar features are grouped together. Since the image features of damaged unit region images and undamaged unit region images are different, clustering algorithms can effectively distinguish them.

[0094] Step S130: Using a pre-obtained adaptive multi-task sparse digital image correlation model, perform depth recognition on the image of the damaged unit region to obtain the damage sensing location and corresponding luminous value of the image of the damaged unit region.

[0095] Specifically, since the smart skin uses a honeycomb-like periodic grid cell structure as its substrate, the cell region image can be divided into periodic cells, and the corresponding image feature data can be extracted from each cell. Because the strain at the location of structural damage to the smart skin varies, it will emit light of different wavelengths. Therefore, the damage sensing location corresponding to the damage feature data in each periodic cell can be identified based on the image feature data. After merging, the damage sensing location and corresponding emission value of the entire detection cell region image can be finally determined. Furthermore, the periodic cells of the cell region image can be divided into more fine grids, allowing for a more precise determination of the damage sensing location and corresponding emission value of the entire detection cell region image.

[0096] In the adaptive multi-task sparse digital image correlation model, A is the baseline crack feature matrix, Y is the feature matrix of the damaged unit region image, C is the sparse correlation mapping matrix, which contains the correlation coefficient between the baseline crack feature matrix A and the feature matrix Y of the damaged unit region image, λ is the adaptive adjustment factor, and k is the iteration step number. The formula of the adaptive multi-task sparse digital image correlation model is as follows:

[0097]

[0098] st.C≥0

[0099]

[0100] The objective is to minimize the first and second terms of the objective function by solving the sparse correlation mapping matrix Ck. Due to the sparsity of the crack across the entire tested image, only a portion of the image feature matrix of the obtained unit region image has a strong correlation with the benchmark crack feature library, leading to Ck... k The corresponding element takes the larger value. (C) k The initial value of C is a 0 matrix. When the orbital structure is damaged or the image feature matrix of the acquired unit region image is abnormal, C... k The element corresponding to the location of the crack or breakage receives a large value, while other elements remain at 0 or close to 0.

[0101] In the present scheme, the optimization problem of sparse regularization needs to be solved, but the solution of the sparse regularization problem often has no analytical solution, and the solution is more difficult than that of non-sparse regularization. The method for solving the optimal value of C in the present scheme can be a fast iterative soft threshold algorithm or a prediction correction original dual path tracking algorithm, which can efficiently obtain accurate and robust damage identification results. k+1 -c k |<v is a convergence condition, where v is a user-set convergence threshold. When the convergence condition is met, the iteration process is stopped, and the optimal solution of C k is obtained.

[0102] When C and A are not matrices but single column vectors, the above formula is a single-task damage identification objective function. When multiple column vectors are arranged side by side to form a matrix, it becomes a multi-task damage identification objective function. Without sparse constraints, the multi-task damage identification objective function is prone to false positives when solving, i.e., identifying a larger damage at a position in a column as smaller damages at multiple positions in adjacent columns, which is reflected in C as a large value at a position in a column being incorrectly solved as smaller values at multiple positions in adjacent columns.

[0103] In traditional methods, lambda is a fixed value set or selected. In the present scheme, an adaptive adjustment factor is designed to automatically update through a mathematical expression, which is used to balance the first term in the objective function, i.e., the reconstruction term, and the second term in the objective function, i.e., the sparse constraint term. If the reconstruction term is too large relative to the sparse term, noise or interference information may be reconstructed, resulting in elements in C that should be 0 or close to 0 becoming large values, leading to false positives. Conversely, if the sparse constraint term is too large, the solution may be too sparse, i.e., elements in C that should be large values become 0 or close to 0, leading to false negatives. By designing such an adaptive adjustment factor, a balance is achieved between the two terms, resulting in a more accurate solution for C. Through the above model, the positions of local damages can be accurately and efficiently identified.

[0104] In step S140, the damage-induced position and the corresponding light-emitting value of the damage unit area image are input into the pre-obtained intelligent skin calibration function model, and the strain data of the damage position of the track structure is calculated.

[0105] Specifically, the collected data of the damage sensing position are analyzed to obtain an intelligent skin calibration function model, and the collected data at least include background light intensity of the damage sensing position captured by a high-speed camera when the intelligent skin detects an object, an environmental temperature when detection is performed, and a luminous value of the intelligent skin. According to the intelligent skin calibration function model, the known damage sensing position and the luminous value are brought into the above model to obtain a damage degree and a damage trend of the object, and according to a judgment standard of different structural damages in an industrial field, it is determined whether the measured region needs to be warned, and if so, the damage position of the measured region is warned.

[0106] The intelligent skin calibration function model is a function with the luminous value, the strain value at a previous moment, the temperature and the background light as independent variables and the strain data at a current moment as a dependent variable. The strain value can be calculated by inputting the independent variables such as the luminous value and the temperature, so that the strain data of the measured structural damage position are detected, early crack identification, expansion deterioration tracking and reproduction are realized, the damage severity is quantitatively evaluated, and hidden dangers for the continuous safe operation of the measured object due to further damage are avoided.

[0107] In the intelligent skin calibration function model, T is the environmental temperature, P is the background light intensity, and ΔL k is the light intensity change amount of the kth step, Δt k is the time increment of the kth step, and Δξ k is the strain data change amount of the kth step. The intelligent skin calibration function model formula is as follows:

[0108] Δξ k = C1(T, P)(Δt k + log |ΔL k + Δξ (k-1) | - log |Δξ (k-1) |) + C2(T, P)(ΔL k + log |Δt k |) + C3(T, P)(ξ (k-1) - ξ (k-1) ΔL k - log |ξ (k-1) ΔL k |) + C4(T, P).

[0109] In step S150, the strain data are analyzed to obtain a damage detection result of the track structure.

[0110] Specifically, by perceiving and recording changes of the strain data of the damage position, reproduction and analysis of the strain data of the generation, development and deterioration of the track structure crack are realized, and the damage degree of the track structure is determined.

[0111] The embodiment obtains a plurality of unit region images of a track structure covered by the intelligent skin through the above scheme; filters out damaged unit region images from the plurality of unit region images through a clustering algorithm; performs deep recognition on the damaged unit region images through a pre-obtained adaptive multi-task sparse digital image correlation model, to obtain damage sensing positions and corresponding luminous values of the damaged unit region images; inputs the damage sensing positions and corresponding luminous values of the damaged unit region images into a pre-obtained intelligent skin calibration function model, to calculate strain data of damage positions of the track structure; and analyzes the strain data, to obtain a damage detection result of the track structure. Based on the scheme, the luminous image of the intelligent skin detection region is obtained and damage recognition is performed, the strain data of the detected region is analyzed, and thus the detection result is obtained, the expansion and deterioration of the fatigue crack are warned, and further damage is avoided, thereby leaving hidden dangers for the continuous and safe operation of the measured object. Compared with the existing damage detection region technology, the intelligent skin in the embodiment has the advantages of no need of large-scale lead and circuit design and manufacturing, and compared with the hard sensor and the sensor network, has the advantages of high installation position requirement, heavy additional weight, large volume, and the like.

[0112] Referring to Figure 3 , Figure 3 FIG. 2 is a flowchart of a second exemplary embodiment of a track structure damage recognition method based on a combined luminous intelligent skin. The step of filtering out damaged unit region images from the plurality of unit region images through a clustering algorithm includes the following steps.

[0113] In step S1201, unit region image features are extracted according to the unit region images.

[0114] Specifically, since the intelligent skin includes a mechanoluminescent material and a periodic grid unit structure, the unit region image features are extracted from the unit region images based on the periodic grid unit structure, to obtain the unit region image features. The wavelength and intensity of the light emitted by the intelligent skin are different under different mechanical stresses and strains in the detection region, and the corresponding extracted unit region image features are also different.

[0115] In step S1202, the unit region image features are clustered into a plurality of different classes of unit region image feature sets through a clustering algorithm.

[0116] Specifically, according to the different wavelengths and intensities of the light, the unit region image features are divided through the clustering algorithm, to obtain a plurality of different classes of unit region image features. The division facilitates subsequent identification of abnormal unit region image features.

[0117] Step S1203, screening the damage unit area image from the several different types of unit area image feature sets.

[0118] Specifically, the light emission intensity is different, the light wave length is different, and the corresponding extracted image features are also different, so by matching the pre-set damage image features with the several different types of unit area image features obtained in the above step S1202, the abnormal unit area image features are identified. At this time, the identified abnormal area features are several separate unit area image features divided according to the periodic grid unit structure.

[0119] The embodiment extracts unit area image features according to the unit area image, aggregates the unit area image features into several different types of unit area image feature sets through a clustering algorithm, and screens the damage unit area image from the several different types of unit area image feature sets. Based on the scheme, the unit area image features are extracted and the abnormal unit area image features are identified, so as to obtain the damage position and the corresponding light emission value, which provides data support for determining whether the damage position needs to be warned in the future.

[0120] Reference Figure 4 , Figure 4 The flowchart of the third exemplary embodiment of the track structure damage identification method based on the combined light-emitting intelligent skin is shown in the figure. Before the step of inputting the damage induction position and the corresponding light emission value of the damage unit area image into the pre-obtained intelligent skin calibration function model to calculate the strain data of the damage position of the track structure, laboratory calibration is also needed. The laboratory calibration step includes:

[0121] Step S1401, obtaining the light emission value corresponding to the damage induction position under several groups of different pre-set temperatures and different pre-set background light intensities;

[0122] Specifically, both temperature and background light intensity can affect the luminescence value of the collected smart skin, so different temperatures and background light intensities are arranged in combination, the temperature is first set to a fixed temperature value, and the background light intensity is changed, and the luminescence value of the smart skin is collected under different background light intensities; then, the above temperature is changed and set to a fixed value, and the background light intensity is changed again in the same way, the test steps are repeated to ensure the accuracy and comprehensiveness of the collected data. The data of the damage sensing position includes but is not limited to the stress and strain amount, amplitude of the mechanical structure, and the same initial damage is set at the same position of all the same structure test pieces, the smart skin is covered on the initial damage and placed at the same position of all the structure test pieces, when the smart skin emits light of different intensities, a plurality of unit area images of the track structure covered by the smart skin are obtained by the high-speed camera, the adaptive multi-task sparse digital image correlation model is obtained in advance, the damage unit area image is deep identified, and the damage sensing position and the corresponding luminescence value of the damage unit area image are obtained. The adaptive multi-task sparse digital image correlation model is the function in step S130.

[0123] Step S1402, selecting any one group of data from the plurality of groups of different preset temperatures, different preset background light intensities and corresponding luminescence values, establishing a first functional relationship between the temperature, background light intensity and corresponding luminescence value of the damage sensing position;

[0124] Specifically, the ambient temperature and the background light intensity of the smart skin can affect the light intensity of the smart skin luminescence layer collected by the camera, in order to ensure the reliability of the detection result, a fixed temperature needs to be taken, which can be determined by experimental data to ensure stable luminescence effect, and the first luminescence value of the known damage sensing position is calculated according to the luminescence intensity under the first temperature.

[0125] Step S1403, according to the plurality of groups of different preset temperatures, different background light intensities and corresponding luminescence values, the first functional relationship is calibrated in turn to obtain a smart skin calibration function model.

[0126] Specifically, by acquiring the luminescence values of the damage sensing position under different preset temperatures and background light intensities for multiple times, a first functional relationship of the three factors is established by regression modeling and / or curve fitting. Specifically, to ensure the accuracy of the first functional relationship, the temperature is changed, the intelligent skin covers the same initial damage at different temperatures and is placed at the same position of all structural specimens, and when the luminescence layer of the intelligent skin emits light of different intensities, the corresponding functional relationship is determined using the damage sensing position and the corresponding luminescence value in step S1401. By detecting the data of the luminescence values at multiple different temperatures, the first functional relationship is verified, the parameters in the function are optimized, the relationship between the parameters in the function and the temperature is established, the current function is calibrated, and finally the intelligent skin calibration function model is obtained. The intelligent skin calibration function model here is the model in step S130.

[0127] The embodiment specifically obtains the luminescence values of the damage sensing position under different preset temperatures and different preset background light intensities. Any one group of data is selected from the different preset temperatures, different preset background light intensities, and corresponding luminescence values to establish a first functional relationship of the temperature, background light intensity, and corresponding luminescence value of the damage sensing position. The first functional relationship is calibrated in turn according to the different preset temperatures, different background light intensities, and corresponding luminescence values to obtain an intelligent skin calibration function model. Based on the scheme, any one group of data is selected to establish a corresponding first functional relationship, which is then calibrated to obtain an intelligent skin calibration function model, providing a reliable calculation basis for detecting damage data in unknown areas by using an intelligent skin.

[0128] Referring to Figure 5 , Figure 5 The flowchart of the fourth exemplary embodiment of the track structure damage identification method based on the combined luminescent intelligent skin is shown. The step of analyzing the strain data to obtain the damage detection result of the track structure includes:

[0129] In step S1501, the strain data is analyzed for damage and trend change to obtain the analysis result of the strain data.

[0130] Specifically, the strain data is the strain data corresponding to the damage position of the track structure, reflecting the damage degree of the measured object. By analyzing the strain data of the damage position at different time points, the current damage degree and strain data of the measured structure can be obtained, the fatigue cracks can be identified, and the expansion and deterioration of the fatigue cracks can be tracked and reproduced according to the strain data of the damage position at different time points, so that the specific situation of the damage can be accurately analyzed.

[0131] Step S1502, comparing the analysis result with the risk evaluation index threshold value to obtain the damage detection result of the track structure.

[0132] Specifically, the risk evaluation index threshold value can be an industrial standard risk evaluation index threshold value or a strain trend change threshold value or distribution feature calibrated through measured data. By comparing the current analysis result with the risk evaluation index threshold value, the risk degree of the current measured object can be determined to obtain the damage detection result.

[0133] Step S1503, performing corresponding over-limit damage early warning according to the damage detection result of the track structure.

[0134] Specifically, the damage detection result of the track structure is analyzed, and over-limit damage early warning is performed and relevant personnel are informed if the risk threshold value is exceeded, so as to avoid further damage and leave hidden dangers for the continuous safe operation of the measured object. The specific early warning mode is not limited here.

[0135] Through the above scheme, specifically, the strain data is analyzed for damage and trend change to obtain the analysis result of the strain data; the analysis result is compared with the risk evaluation index threshold value to obtain the damage detection result of the track structure. Based on the scheme, the strain data of the damage position is analyzed to determine the damage degree of the measured structure.

[0136] In addition, the application also provides a track structure damage identification system based on a combined light-emitting intelligent skin, which comprises a force-induced light-emitting material and an intelligent skin of a periodic grid unit structure, a high-speed camera, and a damage identification module.

[0137] The high-speed camera is used to acquire a plurality of unit region images of the track structure covered by the intelligent skin.

[0138] The damage identification module is used to filter out a damage unit region image from the plurality of unit region images through a clustering algorithm; perform deep identification on the damage unit region image through a pre-obtained adaptive multi-task sparse digital image correlation model to obtain a damage sensing position and a corresponding light-emitting value of the damage unit region image; input the damage sensing position and the corresponding light-emitting value of the damage unit region image into a pre-obtained intelligent skin calibration function model to calculate strain data of the damage position of the track structure; and analyze the strain data to obtain a damage detection result of the track structure.

[0139] A is a reference crack feature matrix, Y is a feature matrix of the damage unit area image, C is a sparse correlation mapping matrix, including the correlation coefficient of the reference crack feature matrix A and the feature matrix Y of the damage unit area image, lambda is an adaptive adjustment factor, k is an iteration step number, and the adaptive multi-task sparse digital image correlation model is as follows:

[0140]

[0141] s.t.C k ≥0

[0142]

[0143] In the intelligent skin calibration function model, T is the environmental temperature, P is the background light intensity, Delta L k is the light intensity change amount of the kth step, Delta t k is the time increment of the kth step, Delta xi k is the strain data change amount of the kth step, and the intelligent skin calibration function model is as follows:

[0144] Delta xi k =C1(T,P)(Delta t k +log|Delta L k +Delta xi (k-1) |-log|Delta xi (k-1) |)C2(T,P)(Delta L k +log|Delta t k |+C3(T,P)(xi (k-1) -xi (k-1) Delta L k -log|xi (k-1) Delta L k |)+C4T,P)。

[0145] In addition, the application also proposes a terminal device, which comprises a memory, a processor, and a combination-type light-emitting intelligent skin-based track structure damage identification program stored in the memory and running on the processor. When the combination-type light-emitting intelligent skin-based track structure damage identification program is executed by the processor, the steps of the combination-type light-emitting intelligent skin-based track structure damage identification method are realized.

[0146] Since the combination-type light-emitting intelligent skin-based track structure damage identification program is executed by the processor, all the technical solutions of the foregoing embodiments are adopted, and at least all the beneficial effects brought by all the technical solutions of the foregoing embodiments are achieved, which will not be repeated here.

[0147] In addition, the application also provides a readable storage medium, and the readable storage medium stores a track structure damage identification program based on the combined light-emitting intelligent skin.

[0148] When the track structure damage identification program based on the combined light-emitting intelligent skin is executed by the processor, all the technical solutions of all the foregoing embodiments are adopted, and thus all the beneficial effects brought by all the technical solutions of all the foregoing embodiments are achieved, which will not be repeated here.

[0149] It should be noted that, in this document, the terms "comprising", "containing" or any other variant thereof are intended to cover non-exclusive inclusion, so that processes, methods, articles or systems including a series of elements not only include those elements, but also include other elements not explicitly listed or inherent to such processes, methods, articles or systems. Without more limitations, the element defined by the statement "comprising a" does not exclude the presence of other identical elements in the process, method, article or system including the element.

[0150] The serial numbers of the above embodiments of the application are only for description, and do not represent the advantages and disadvantages of the embodiments.

[0151] Through the above description of the embodiments, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be realized by means of software and necessary general hardware platforms, of course, they can also be realized by hardware, but in many cases, the former is a better embodiment. Based on such understanding, the technical solutions of the application can be embodied in the form of a software product, which is stored in a storage medium (such as a ROM / RAM, a magnetic disk, an optical disk) as described above, and includes a plurality of instructions for making a terminal device (which can be a mobile phone, a computer, a server, a controlled terminal, or a network terminal device, etc.) execute the method of each embodiment of the application.

[0152] The above is only the preferred embodiment of the application, and does not limit the patent scope of the application, and any equivalent structure or equivalent process transformation based on the content of the specification and drawings, or direct or indirect application in other related technical fields, are also included in the patent protection scope of the application.

Claims

1. A method for identifying track structure damage based on a combined luminescent smart skin, characterized in that, An application is made to a track structure damage identification system based on a combined luminescent smart skin. The track structure damage identification system based on the combined luminescent smart skin includes at least a smart skin with a mechanoluminescent material and a periodic grid cell structure, a high-speed camera, and a damage identification module. The track structure damage identification method based on the combined luminescent smart skin includes the following steps: Acquire several unit region images of the track structure covered by the intelligent skin; Damaged unit region images are selected from the plurality of unit region images using a clustering algorithm; Using a pre-obtained adaptive multi-task sparse digital image correlation model, depth recognition is performed on the image of the damaged unit region to obtain the damage sensing location and corresponding luminescence value of the image of the damaged unit region. The damage sensing location and corresponding luminescence value of the damaged unit region image are input into a pre-obtained smart skin calibration function model to calculate the strain data of the damage location of the track structure. The strain data is analyzed to obtain the damage detection results of the track structure; In the adaptive multi-task sparse digital image correlation model, A is the baseline crack feature matrix, Y is the feature matrix of the damaged unit region image, and C is the sparse correlation mapping matrix, which contains the correlation coefficients between the baseline crack feature matrix A and the feature matrix Y of the damaged unit region image. Here, k is the adaptive adjustment factor, and k is the iteration step number. The formula for the adaptive multi-task sparse digital image correlation model is as follows: ； In the intelligent skin calibration function model, T is the ambient temperature and P is the background light intensity. It is the change in light intensity at step k. It is the time increment at step k. This represents the change in strain data at step k. The formula for the intelligent skin calibration function model is shown below: 。 2. The method for identifying track structure damage based on a combined luminescent smart skin according to claim 1, characterized in that, The step of selecting damaged unit region images from the plurality of unit region images using a clustering algorithm includes: Extract the features of the unit region image based on the unit region image; The image features of the unit region are clustered into several different classes of unit region image feature sets using a clustering algorithm. The damaged unit region image is selected from the feature sets of the several different types of unit region images.

3. The method for identifying track structure damage based on a combined luminescent smart skin according to claim 1, characterized in that, Before the step of inputting the damage sensing location and corresponding luminescence value of the damaged unit region image into the pre-obtained smart skin calibration function model to calculate the strain data of the damage location of the track structure, laboratory calibration is required. The laboratory calibration steps include: Obtain the luminescence values ​​corresponding to the damage sensing locations under several sets of different preset temperatures and different preset background light intensities; Select any set of data from the several sets of different preset temperatures, different preset background light intensities and corresponding luminous values, and establish a first functional relationship between the temperature, background light intensity and corresponding luminous value of the damage sensing location; Based on the aforementioned sets of different preset temperatures, different background light intensities, and corresponding luminous values, the first functional relationship is calibrated sequentially to obtain the intelligent skin calibration function model.

4. The method for identifying track structure damage based on combined luminescent smart skin according to claim 1, characterized in that, The step of analyzing the strain data to obtain the damage detection results of the track structure includes: Damage and trend change analysis is performed on the strain data to obtain the analysis results of the strain data; The analysis results are compared with the risk assessment index threshold to obtain the damage detection results of the track structure.

5. The method for identifying track structure damage based on a combined luminescent smart skin according to claim 4, characterized in that, After the step of comparing the analysis results with the risk assessment index threshold to obtain the damage detection results of the track structure, the method further includes: Based on the damage detection results of the track structure, a corresponding warning of excessive damage is issued.

6. The method for identifying track structure damage based on a combined luminescent smart skin according to claim 1, characterized in that, The step of acquiring several unit region images of the track structure covered by the smart skin includes: The high-speed camera acquires several unit area images of the track structure covered by the smart skin.

7. The method for identifying track structure damage based on a combined luminescent smart skin according to claim 3, characterized in that, The step of establishing the first functional relationship between the temperature, background light intensity, and corresponding luminance value at the damage sensing location includes: A first functional relationship between the temperature, background light intensity, and corresponding luminous value at the damage-sensing location is established through regression modeling and / or curve fitting.

8. The method for identifying track structure damage based on a combined luminescent smart skin according to claim 1, characterized in that, The smart skin includes a base layer, which is a flexible material with periodic partitions; a surface sealing layer, which is located on one side of the base layer and encloses the base layer to form a cavity; and a sensing layer, which is located inside the cavity and is a mechanoluminescent material.

9. A track structure damage identification system based on a combined luminescent smart skin, characterized in that, The track structure damage identification system based on a combined luminescent smart skin includes a mechanoluminescent material and a smart skin with a periodic grid cell structure, a high-speed camera, and a damage identification module, wherein: The high-speed camera is used to acquire several unit area images of the track structure covered by the smart skin; The damage identification module is used to filter out damaged unit region images from the plurality of unit region images using a clustering algorithm; perform depth recognition on the damaged unit region images using a pre-obtained adaptive multi-task sparse digital image correlation model to obtain the damage sensing location and corresponding luminescence value of the damaged unit region images; input the damage sensing location and corresponding luminescence value of the damaged unit region images into a pre-obtained intelligent skin calibration function model to calculate the strain data of the damage location of the track structure; and analyze the strain data to obtain the damage detection result of the track structure. In the adaptive multi-task sparse digital image correlation model, A is the baseline crack feature matrix, Y is the feature matrix of the damaged unit region image, and C is the sparse correlation mapping matrix, which contains the correlation coefficients between the baseline crack feature matrix A and the feature matrix Y of the damaged unit region image. Here, k is the adaptive adjustment factor, and k is the iteration step number. The formula for the adaptive multi-task sparse digital image correlation model is as follows: ； In the intelligent skin calibration function model, T is the ambient temperature and P is the background light intensity. It is the change in light intensity at step k. It is the time increment at step k. This represents the change in strain data at step k. The formula for the intelligent skin calibration function model is shown below: 。

Images