Mine disc brake wear detection method based on intelligent sensor

Abstract

This invention discloses a wear detection method for mining disc brakes based on intelligent sensors, belonging to the field of mechanical health monitoring technology. The method includes: unfolding a wrapped phase into a continuous phase; converting the continuous phase into a three-dimensional shape; calculating the surface roughness within a specified window based on the three-dimensional shape to generate an optical feature vector; using the Sobel operator to calculate the gradient of the continuous phase to generate a phase gradient flow field; using discrete cosine transform to calculate low-frequency spectral components; calculating the information flow intensity based on the low-frequency spectral components; constructing a diffusion kernel matrix; calculating the diffusion distance; and using the Euclidean distance formula to calculate the wear index for wear state judgment. This invention improves the comprehensiveness and robustness of wear state assessment by combining multimodal data fusion with nonlinear feature analysis, and enhances the accuracy of three-dimensional shape reconstruction and vibration feature extraction by using a pyramid phase unfolding algorithm combined with fractional Fourier transform.

Description

Technical Field

[0001] This invention relates to the field of mechanical health monitoring technology, and in particular to a wear detection method for mining disc brakes based on intelligent sensors. Background Technology

[0002] With the continuous improvement of the automation level of mining equipment, the safety and reliability of mining braking systems have received increasing attention. Disc brakes are widely used in mining transportation equipment due to their advantages such as simple structure, large braking torque, and fast response speed. However, due to factors such as high braking frequency and harsh working environment (such as high temperature, high dust, and high vibration), the friction pair of disc brakes (brake disc and brake pads) is prone to wear, deformation, and cracking. In severe cases, this can lead to brake failure and threaten the safety of mine workers.

[0003] Current methods for detecting wear on mining disc brakes still have shortcomings. The degree of multimodal data fusion is low. Traditional methods usually only process optical, vibration, or temperature data independently, lacking the organic integration of cross-modal features, making it difficult to fully characterize the complexity of wear conditions. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a wear detection method for mining disc brakes based on intelligent sensors, which solves the problem of low multimodal data fusion. Traditional methods usually only process optical, vibration or temperature data independently, lacking organic integration of cross-modal features and making it difficult to fully characterize the complexity of wear state.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides a wear detection method for mining disc brakes based on intelligent sensors, which includes the following steps:

[0008] Collect and preprocess multimodal data, calculate the wrapping phase using a four-step phase shift algorithm, unfold the wrapping phase into a continuous phase, convert the continuous phase into a three-dimensional shape, calculate the surface roughness within a specified window based on the three-dimensional shape, and generate optical feature vectors using feature vector construction.

[0009] Vibration data is extracted from multimodal data, fractional Fourier transform is performed to generate fractional frequency domain signals, frequency band energy is calculated and defined as FrFT energy features, temperature data is extracted and defined as temperature features, gradient calculation is performed on continuous phase using the Sobel operator to generate phase gradient flow field, consistency coefficient is calculated, denoised optical feature vectors are calculated based on consistency coefficients, and comprehensive feature vectors are generated using feature vector construction.

[0010] The low-frequency spectral components are calculated using discrete cosine transform. Based on the low-frequency spectral components, the information flow intensity is calculated, a diffusion kernel matrix is ​​constructed, the diffusion distance is calculated, the wear index is calculated using the Euclidean distance formula, the wear state is judged based on the diffusion distance, the morphology difference is calculated using the difference method, the maximum value of the morphology difference is selected as the maximum wear depth, and the wear area is calculated.

[0011] Build a visual interface to display wear status and wear area, and store the collected and analyzed multimodal data.

[0012] As a preferred embodiment of the wear detection method for mining disc brakes based on intelligent sensors described in this invention, the step of unfolding the enclosed phase into a continuous phase, converting the continuous phase into a three-dimensional morphology, and based on the three-dimensional morphology, includes:

[0013] Phase shift images are extracted from the normalized multimodal data, and the wrap-around phase is calculated using a four-step phase shift algorithm.

[0014] The pyramid phase unfolding algorithm is used to unfold the wrapped phase into a continuous phase, and then convert the continuous phase into a three-dimensional shape.

[0015] The sliding window method is used to set a specified window, and the surface roughness within the specified window is calculated based on the three-dimensional topography.

[0016] Optical feature vectors are generated using feature vector construction.

[0017] As a preferred embodiment of the wear detection method for mining disc brakes based on intelligent sensors described in this invention, the step of extracting vibration data from multimodal data, performing fractional Fourier transform to generate fractional frequency domain signals, and calculating frequency band energy, defined as FrFT energy features, includes:

[0018] Vibration data are extracted from the normalized multimodal data, and fractional Fourier transform is performed to generate fractional frequency domain signals.

[0019] Based on fractional-order frequency domain signals, frequency band energy is calculated and defined as FrFT energy characteristics.

[0020] Temperature data is extracted from the normalized multimodal data and defined as temperature features.

[0021] As a preferred embodiment of the wear detection method for mining disc brakes based on intelligent sensors described in this invention, the method includes: using the Sobel operator to calculate the gradient of the continuous phase to generate a phase gradient flow field, calculating the consistency coefficient, calculating the denoised optical feature vector based on the consistency coefficient, and using the feature vector to construct and generate a comprehensive feature vector, including:

[0022] The Sobel operator is used to calculate the gradient of the continuous phase to generate a phase gradient flow field.

[0023] Collect historical phase gradient flow field data of the brake and calculate the mean value, which is defined as the standard gradient flow field. Calculate the consistency coefficient.

[0024] Based on the consistency coefficient, calculate the denoised optical feature vector;

[0025] The mean of the denoised optical feature vector is calculated to obtain the global optical feature vector, and the mean of the three-dimensional topography and the mean of the surface roughness are extracted from the global optical feature vector.

[0026] Use the feature vectors to construct and generate a comprehensive feature vector.

[0027] As a preferred embodiment of the wear detection method for mining disc brakes based on intelligent sensors described in this invention, the steps of calculating the information flow intensity, constructing the diffusion kernel matrix, calculating the diffusion distance, and calculating the wear index using the Euclidean distance formula include:

[0028] Calculate the low-frequency spectral components using discrete cosine transform;

[0029] Calculate the information flow intensity based on low-frequency spectral components;

[0030] Collect historical comprehensive feature vectors, construct a diffusion kernel matrix, and calculate the diffusion mapping;

[0031] Calculate the diffusion distance based on the diffusion map;

[0032] The wear index is calculated using the Euclidean distance formula.

[0033] As a preferred embodiment of the wear detection method for mining disc brakes based on intelligent sensors described in this invention, the step of judging the wear state based on diffusion distance includes:

[0034] Based on the diffusion distance, a dynamic threshold is set using statistical analysis.

[0035] The dynamic threshold is compared with the wear index. When the wear index is less than or equal to the dynamic threshold, it is judged as a normal state; otherwise, it is judged as a worn state.

[0036] As a preferred embodiment of the wear detection method for mining disc brakes based on intelligent sensors described in this invention, the method for calculating the wear area by selecting the maximum value of the morphological difference as the maximum wear depth includes:

[0037] Extract the three-dimensional morphology corresponding to the wear state, calculate the morphology difference using the difference method, and select the maximum value of the morphology difference as the maximum wear depth.

[0038] The percentile method is used to set the classification threshold and calculate the wear area.

[0039] As a preferred embodiment of the wear detection method for mining disc brakes based on intelligent sensors described in this invention, the step of collecting and preprocessing multimodal data includes:

[0040] Use smart sensors to collect multimodal data and perform time synchronization, noise reduction and normalization processing;

[0041] The smart sensor includes a vibration sensor, a CMOS camera, and a temperature sensor.

[0042] The multimodal data includes vibration, phase shift images, and temperature data.

[0043] As a preferred embodiment of the wear detection method for mining disc brakes based on intelligent sensors described in this invention, the step of constructing a visual interface to display the wear state and wear area includes:

[0044] A visual interface is built using the front-end framework React.js to visualize the wear status and wear area.

[0045] Users who have passed real-name verification are allowed to view this information.

[0046] As a preferred embodiment of the wear detection method for mining disc brakes based on intelligent sensors described in this invention, the method for storing, collecting, and analyzing the generated multimodal data includes:

[0047] The collected multimodal data and the wear area generated by the analysis are stored in a central database, and secure access measures are set. The central database backs up the stored data to the cloud and performs integrity checks on the stored data and backup data regularly. After the checks are completed, integrity check records are generated and stored synchronously in the central database.

[0048] The beneficial effects of this invention are as follows: This invention improves the comprehensiveness and robustness of wear condition assessment by combining multimodal data fusion with nonlinear feature analysis, and enhances the accuracy of three-dimensional morphology reconstruction and vibration feature extraction by using the pyramid phase expansion algorithm combined with fractional Fourier transform. Attached Figure Description

[0049] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the 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.

[0050] Figure 1 This is a flowchart illustrating the operation of the wear detection method for mining disc brakes based on intelligent sensors in Example 1. Detailed Implementation

[0051] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0052] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0053] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0054] Example 1, referring to Figure 1 This is the first embodiment of the present invention, which provides a wear detection method for mining disc brakes based on intelligent sensors, including the following steps:

[0055] S1. Collect multimodal data and preprocess it. Use the four-step phase shift algorithm to calculate the wrapping phase. Unfold the wrapping phase into a continuous phase. Convert the continuous phase into a three-dimensional shape. Based on the three-dimensional shape, calculate the surface roughness within a specified window. Use feature vectors to construct and generate optical feature vectors.

[0056] Specifically, collecting and preprocessing multimodal data includes:

[0057] Use smart sensors to collect multimodal data and perform time synchronization, noise reduction and normalization processing;

[0058] The smart sensor includes a vibration sensor, a CMOS camera, and a temperature sensor.

[0059] The multimodal data includes vibration, phase shift images, and temperature data.

[0060] Collecting multimodal data enables the collaborative acquisition and unified scale representation of multidimensional physical features, enhancing the system's comprehensive perception of target states in complex scenarios. Normalization avoids computational biases caused by differences in data scales, ensuring the stability and accuracy of subsequent algorithm calculations, thus achieving the basic construction effect of multimodal fusion.

[0061] Furthermore, the wrapped phase is unfolded into a continuous phase, and the continuous phase is converted into a three-dimensional topography. Based on the three-dimensional topography, the following are included:

[0062] Extract the phase shift image from the normalized multimodal data, and calculate the wrap-around phase using a four-step phase shift algorithm. The formula is as follows:

[0063] ,

[0064] in To wrap the phase, we represent the phase information of the actuator surface, where x and y are the horizontal and vertical pixel coordinates on the image plane, respectively. The intensity of the phase-shifted image;

[0065] The pyramid phase unwrapping algorithm is used to unwrap the wrapped phase into a continuous phase, and then convert the continuous phase into a three-dimensional shape. The formula is as follows:

[0066] ,

[0067] in The topography represents the height distribution on the brake surface in three dimensions. For continuous phase, P is the fringe period, set using a projector calibration. The projector's focal length can be obtained from the projector's technical manual.

[0068] Using the sliding window method, a specified window is defined. Based on the 3D topography, the surface roughness within the specified window is calculated using the following formula:

[0069] ,

[0070] in For surface roughness, To specify the number of pixels in the window, To specify the 3D shape of pixels within a window, and These are the horizontal and vertical relative coordinates of the pixels within the specified window, representing the horizontal and vertical offsets relative to the image pixels within the specified window, where W is the specified window.

[0071] Using feature vectors to construct and generate optical feature vectors, the formula is as follows:

[0072] ,

[0073] in This is the optical feature vector.

[0074] The four-step phase shift algorithm has good anti-interference performance and measurement accuracy, and can effectively suppress noise interference on phase values, thus ensuring the stability of phase calculation results. The pyramid method has strong anti-edge distortion ability and low computational complexity, and is particularly suitable for phase unfolding tasks under the influence of noise disturbance or occlusion. Spatial reconstruction of the micro-geometric features of the target surface improves the measurement system's ability to detect microstructures. The quantitative characterization of the micro-structure state of the target surface is helpful for quality inspection and defect localization. Roughness, as an indicator strongly correlated with mechanical properties such as wear and friction coefficient, is represented by optical feature vectors, which realize the standardized expression of multi-dimensional optical features, making it convenient for machine learning classification, anomaly detection, or subsequent control decisions.

[0075] S2. Extract vibration data from multimodal data, perform fractional Fourier transform to generate fractional frequency domain signal, calculate frequency band energy and define it as FrFT energy feature, extract temperature data and define it as temperature feature, use Sobel operator to calculate gradient of continuous phase to generate phase gradient flow field, calculate consistency coefficient, calculate denoised optical feature vector based on consistency coefficient, use feature vector to construct and generate comprehensive feature vector.

[0076] Specifically, vibration data is extracted from multimodal data, fractional Fourier transform is performed to generate fractional frequency domain signals, and frequency band energy is calculated, defined as FrFT energy features, including:

[0077] Vibration data are extracted from the normalized multimodal data, and a fractional Fourier transform is performed to generate a fractional frequency domain signal, as shown in the formula:

[0078] ,

[0079] ,

[0080] ,

[0081] in Let u be a fractional-order frequency domain signal, and u be a fractional-order frequency domain variable. and These are the upper and lower limits of integration, indicating that the integration range covers the entire time axis. The vibration data are for time t, where t is time. p is the rotation angle, p is the FrFT order, and j is the virtual unit;

[0082] Based on fractional-order frequency domain signals, frequency band energy is calculated and defined as the FrFT energy characteristic, with the following formula:

[0083] ,

[0084] in For FrFT energy characteristics, For frequency mapping, The upper limit frequency of the frequency band is set through signal frequency analysis;

[0085] Temperature data is extracted from the normalized multimodal data and defined as temperature features.

[0086] Traditional Fourier transform is only applicable to periodic or steady-state signals, while FrFT can extract stable features for complex periodic or frequency-drift signals. Frequency band energy changes often precede macroscopic structural damage, and FrFT features have early warning value at the edge of drastic changes in morphology.

[0087] Furthermore, the Sobel operator is used to calculate the gradient of the continuous phase to generate a phase gradient flow field. A consistency coefficient is calculated, and based on this coefficient, a denoised optical eigenvector is calculated. Using these eigenvectors, a comprehensive eigenvector is generated, including:

[0088] The Sobel operator is used to calculate the gradient of the continuous phase to generate the phase gradient flow field, as shown in the formula:

[0089] ,

[0090] in For the phase gradient flow field, represents the phase gradient in the x and y directions. and These are the partial derivatives of the continuous phase in the x and y directions, respectively;

[0091] Collect historical phase gradient flow field data of the brake and calculate the mean, which is defined as the standard gradient flow field. Calculate the consistency coefficient using the following formula:

[0092] ,

[0093] in The consistency coefficient, The consistency analysis window is set using the sliding window method. Let i and o represent the gradient flow field of pixels within the consistency analysis window, respectively, and represent the horizontal and vertical relative coordinates of the pixels within the window, indicating the horizontal and vertical offsets relative to the image pixels within the window. For standard gradient flow fields;

[0094] Based on the consistency coefficient, the denoised optical feature vector is calculated using the following formula:

[0095] ,

[0096] in For denoising optical feature vectors, The standard optical feature vector is set based on the mean of historical optical feature vectors;

[0097] The mean of the denoised optical feature vector is calculated to obtain the global optical feature vector, and the mean of the three-dimensional topography and the mean of the surface roughness are extracted from the global optical feature vector.

[0098] Using feature vectors, a comprehensive feature vector is generated using the following formula:

[0099] ,

[0100] in For the comprehensive feature vector, The average value of the three-dimensional morphology. This represents the average surface roughness. It is a temperature characteristic.

[0101] The gradient flow field will show a significant deviation, which can help to judge cracks, wear or foreign object adhesion. The gradient is less susceptible to interference from spot jitter or reflection distortion than the original phase. The consistency coefficient can significantly reduce the misjudgment of morphology caused by high-frequency disturbances, improve the consistency of feature distribution between different samples, and is conducive to large-scale industrial deployment. It can construct standardized input vectors, which can be used as input for machine learning models or for abnormal pattern recognition.

[0102] S3. Calculate the low-frequency spectral components using discrete cosine transform. Based on the low-frequency spectral components, calculate the information flow intensity, construct the diffusion kernel matrix, calculate the diffusion distance, calculate the wear index using the Euclidean distance formula, determine the wear state based on the diffusion distance, calculate the morphology difference using the difference method, select the maximum value of the morphology difference as the maximum wear depth, and calculate the wear area.

[0103] Specifically, the process involves calculating the information flow intensity, constructing a diffusion kernel matrix, calculating the diffusion distance, and using the Euclidean distance formula to calculate the wear index, including:

[0104] The formula for calculating low-frequency spectral components using discrete cosine transform is as follows:

[0105] ,

[0106] in Let v be the v-th low-frequency spectral component, representing the frequency component of the eigenvector, where K is the dimension of the eigenvector. , where k is the component index of the comprehensive feature vector, and v is the low-frequency component;

[0107] The information flow intensity is calculated based on the low-frequency spectral components using the following formula:

[0108] ,

[0109] in Information flow intensity reflects the deviation between current characteristics and health status. The standard low-frequency spectral components are set based on the average of historical low-frequency spectral components, and V is the number of low-frequency spectral components.

[0110] Collect historical comprehensive feature vectors and construct a diffusion kernel matrix, using the following formula:

[0111] ,

[0112] in The first of the diffusion kernel matrix The element represents the similarity between the historical comprehensive feature vectors b and c. and These are the b-th and c-th standard comprehensive feature vectors, respectively, set based on the mean of historical comprehensive feature vectors. The Gaussian kernel value is set based on the median of the historical comprehensive feature vector;

[0113] The diffusion map is calculated using the following formula:

[0114] ,

[0115] in Let be the diffusion mapping value of the b-th sample, representing the projection of the feature into the diffusion space. This represents the number of historical comprehensive feature vectors;

[0116] Based on the diffusion map, the diffusion distance is calculated using the following formula:

[0117] ,

[0118] in The diffusion distance reflects the distributional differences between current characteristics and health status. The standard diffusion map value for the b-th sample is set based on the mean of historical diffusion map values.

[0119] The wear index is calculated using the Euclidean distance formula, which is:

[0120] ,

[0121] Where H is the wear index.

[0122] DCT exhibits excellent energy concentration, focusing the main information contained in the original feature vectors into the low-frequency part, effectively removing high-frequency noise, and retaining the principal component features reflecting the overall operating trend of the equipment. After DCT, the feature dimension can be compressed to only the low-frequency part, reducing computational complexity while enhancing the contrast between different samples. The information flow intensity reflects the overall deviation between the current state and the historical health state, providing a quantitative indicator for subsequent anomaly identification. As a result of frequency domain features, the information flow intensity can be fused with optical features, thermal features, vibration features, etc., to improve the comprehensive judgment ability. Diffusion mapping is a nonlinear dimensionality reduction method that can preserve the inherent manifold structure between the original data and avoid the simplification of complex feature relationships by linear methods (such as PCA). Through the diffusion kernel, the distance metric between the original feature vectors is more structurally informative, making subsequent similarity assessment more robust and unaffected by local perturbations. The feature distribution in the diffusion space is more suitable for revealing the nonlinear, multi-stage evolution path in the wear process, providing a theoretical basis for the classification of complex wear states.

[0123] Furthermore, based on the diffusion distance, wear condition is determined, including:

[0124] Based on the diffusion distance, a dynamic threshold is set using statistical analysis.

[0125] The dynamic threshold is compared with the wear index. When the wear index is less than or equal to the dynamic threshold, it is judged as a normal state; otherwise, it is judged as a worn state.

[0126] As a continuous variable, the wear index better reflects the phased changes in the wear process compared to traditional binary classification diagnostic methods. The dynamic threshold set through statistical analysis not only considers the feature deviation of the current sample itself, but also combines the volatility of historical samples, making the judgment more sensitive and reliable. The wear index and the dynamic threshold constitute an adaptive diagnostic mechanism, which can be extended to online monitoring and remote condition assessment scenarios of industrial equipment.

[0127] Furthermore, the maximum value of the morphological difference is selected as the maximum wear depth, and the wear area is calculated, including:

[0128] Extract the three-dimensional morphology corresponding to the wear state, and calculate the morphology difference using the interpolation method. The formula is as follows:

[0129] ,

[0130] in The morphological difference value represents the wear depth. The standard topographic height is set based on the average of historical 3D topographic values. The three-dimensional morphology corresponding to the wear state;

[0131] The maximum value of the morphological difference is selected as the maximum wear depth;

[0132] The percentile method is used to set the classification threshold, and the wear area is calculated using the following formula:

[0133] ,

[0134] Where A is the wear area. for, For indicator functions, if If it is 1, then it is 0; otherwise, it is 0. s is the area of ​​a single pixel.

[0135] By analyzing the morphological differences of each pixel, wear patterns such as local depressions, erosion, and cracks can be accurately determined, providing a basis for subsequent repair or scrapping decisions. By using percentiles instead of fixed thresholds to delineate wear areas, the algorithm's adaptability to abnormal noise and extreme points is enhanced, avoiding misjudgments.

[0136] S4. Construct a visual interface to display wear status and wear area, and store the collected and analyzed multimodal data;

[0137] Specifically, a visual interface will be built to display the wear status and wear area, including:

[0138] A visual interface is built using the front-end framework React.js to visualize the wear status and wear area.

[0139] Users who have passed real-name verification are allowed to view this information.

[0140] Operators, maintenance personnel, and quality inspection engineers can log in to the system in real time to view the results, avoiding misoperations caused by delayed transmission. Different data access permissions can be assigned according to user roles (e.g., ordinary operators can only view wear levels, engineers can access complete data, and auditors can download reports).

[0141] Furthermore, the storage, collection, and analysis of the resulting multimodal data includes:

[0142] The collected multimodal data and the wear area generated by the analysis are stored in a central database, and secure access measures are set. The central database backs up the stored data to the cloud and performs integrity checks on the stored data and backup data regularly. After the checks are completed, integrity check records are generated and stored synchronously in the central database.

[0143] Cloud databases allow data sharing and access across different devices, facilitating the construction of general wear and tear assessment models and enabling large-scale applications. Data corruption or loss will significantly affect the accuracy of analysis results, and integrity checks ensure a reliable closed loop in the data chain.

[0144] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A wear detection method for mining disc brakes based on intelligent sensors, characterized in that: Includes the following steps: Collect and preprocess multimodal data, calculate the wrapping phase using a four-step phase shift algorithm, unfold the wrapping phase into a continuous phase, convert the continuous phase into a three-dimensional shape, calculate the surface roughness within a specified window based on the three-dimensional shape, and generate optical feature vectors using feature vector construction. Vibration data is extracted from multimodal data, fractional Fourier transform is performed to generate fractional frequency domain signals, frequency band energy is calculated and defined as FrFT energy features, temperature data is extracted and defined as temperature features, gradient calculation is performed on continuous phase using the Sobel operator to generate phase gradient flow field, consistency coefficient is calculated, denoised optical feature vectors are calculated based on consistency coefficients, and comprehensive feature vectors are generated using feature vector construction. The low-frequency spectral components are calculated using discrete cosine transform. Based on the low-frequency spectral components, the information flow intensity is calculated, a diffusion kernel matrix is ​​constructed, the diffusion distance is calculated, the wear index is calculated using the Euclidean distance formula, the wear state is judged based on the diffusion distance, the morphology difference is calculated using the difference method, the maximum value of the morphology difference is selected as the maximum wear depth, and the wear area is calculated. Build a visual interface to display wear status and wear area, and store the collected and analyzed multimodal data.

2. The wear detection method for mining disc brakes based on intelligent sensors as described in claim 1, characterized in that: The process of unfolding the wrapped phase into a continuous phase, converting the continuous phase into a three-dimensional shape, and based on the three-dimensional shape includes: Phase shift images are extracted from the normalized multimodal data, and the wrap-around phase is calculated using a four-step phase shift algorithm. The pyramid phase unfolding algorithm is used to unfold the wrapped phase into a continuous phase, and then convert the continuous phase into a three-dimensional shape. The sliding window method is used to set a specified window, and the surface roughness within the specified window is calculated based on the three-dimensional topography. Optical feature vectors are generated using feature vector construction.

3. The wear detection method for mining disc brakes based on intelligent sensors as described in claim 2, characterized in that: The process involves extracting vibration data from multimodal data, performing a fractional Fourier transform (FFT) to generate a fractional frequency domain signal, calculating the frequency band energy, and defining it as the FrFT energy feature, including: Vibration data are extracted from the normalized multimodal data, and fractional Fourier transform is performed to generate fractional frequency domain signals. Based on fractional-order frequency domain signals, frequency band energy is calculated and defined as FrFT energy characteristics. Temperature data is extracted from the normalized multimodal data and defined as temperature features.

4. The wear detection method for mining disc brakes based on intelligent sensors as described in claim 3, characterized in that: The process involves using the Sobel operator to calculate the gradient of the continuous phase, generating a phase gradient flow field, calculating a consistency coefficient, and then calculating a denoised optical eigenvector based on the consistency coefficient. Using eigenvector construction, a comprehensive eigenvector is generated, including: The Sobel operator is used to calculate the gradient of the continuous phase to generate a phase gradient flow field. Collect historical phase gradient flow field data of the brake and calculate the mean value, which is defined as the standard gradient flow field. Calculate the consistency coefficient. Based on the consistency coefficient, calculate the denoised optical feature vector; The mean of the denoised optical feature vector is calculated to obtain the global optical feature vector, and the mean of the three-dimensional topography and the mean of the surface roughness are extracted from the global optical feature vector. Use the feature vectors to construct and generate a comprehensive feature vector.

5. The wear detection method for mining disc brakes based on intelligent sensors as described in claim 4, characterized in that: The calculation of information flow intensity, construction of a diffusion kernel matrix, calculation of diffusion distance, and calculation of the wear index using the Euclidean distance formula include: Calculate the low-frequency spectral components using discrete cosine transform; Calculate the information flow intensity based on low-frequency spectral components; Collect historical comprehensive feature vectors, construct a diffusion kernel matrix, and calculate the diffusion mapping; Calculate the diffusion distance based on the diffusion map; The wear index is calculated using the Euclidean distance formula.

6. The wear detection method for mining disc brakes based on intelligent sensors as described in claim 5, characterized in that: The wear condition determination based on diffusion distance includes: Based on the diffusion distance, a dynamic threshold is set using statistical analysis. The dynamic threshold is compared with the wear index. When the wear index is less than or equal to the dynamic threshold, it is judged as a normal state; otherwise, it is judged as a worn state.

7. The wear detection method for mining disc brakes based on intelligent sensors as described in claim 6, characterized in that: The maximum value of the selected morphological difference is used as the maximum wear depth to calculate the wear area, including: Extract the three-dimensional morphology corresponding to the wear state, calculate the morphology difference using the difference method, and select the maximum value of the morphology difference as the maximum wear depth. The percentile method is used to set the classification threshold and calculate the wear area.

8. The wear detection method for mining disc brakes based on intelligent sensors as described in claim 1, characterized in that: The collection and preprocessing of multimodal data includes: Use smart sensors to collect multimodal data and perform time synchronization, noise reduction and normalization processing; The smart sensor includes a vibration sensor, a CMOS camera, and a temperature sensor. The multimodal data includes vibration, phase shift images, and temperature data.

9. The wear detection method for mining disc brakes based on intelligent sensors as described in claim 7, characterized in that: The construction of a visual interface to display wear status and wear area includes: A visual interface is built using the front-end framework React.js to visualize the wear status and wear area. Users who have passed real-name verification are allowed to view this information.

10. The wear detection method for mining disc brakes based on intelligent sensors as described in claim 7, characterized in that: The multimodal data collected and analyzed includes: The collected multimodal data and the wear area generated by the analysis are stored in a central database, and secure access measures are set. The central database backs up the stored data to the cloud and performs integrity checks on the stored data and backup data regularly. After the checks are completed, integrity check records are generated and stored synchronously in the central database.

Images