Conveyor belt fault prediction system and method based on multimodal data analysis

Abstract

The present invention relates to the technical field of conveyor belt failure, and in particular to a conveyor belt failure prediction system and method based on multimodal data analysis. The present invention discloses a conveyor belt failure prediction system based on multimodal data analysis, the system comprising a belt conveyor and a failure prediction module; the belt conveyor comprises a roller frame, rollers, a conveyor belt, a limit bar, a limit protrusion and an L-shaped frame, the roller frame is provided with a plurality of rollers, the conveyor belt is wrapped around the outer walls of the plurality of rollers, the two rollers located at both ends of the conveyor belt are limit rollers, the two ends of the outer wall of the limit roller are symmetrically provided with limit protrusions, the two ends of the inner wall of the conveyor belt are fixedly provided with limit bars, the limit bars are located between the limit protrusions at the two ends, and the limit protrusions limit the position of the conveyor belt through the limit bars. The present invention reduces the technical difficulty of detection and analysis, and improves the accuracy of detecting whether the conveyor belt is offset and predicting deviation faults.

Description

Technical Field

[0001] The present invention relates to the technical field of conveyor belt failure, and in particular to a conveyor belt failure prediction system and method based on multimodal data analysis. Background Art

[0002] Multimodal data analysis refers to the simultaneous consideration of different types of data, such as images, text, and audio, when processing and analyzing data, to help people better understand the data and discover hidden patterns and features.

[0003] The belt conveyor includes a conveyor belt, a roller group, a drive device, a tensioning device, a cleaning device and a protective device. The drive device drives the roller group to rotate, and the roller group drives the conveyor belt to rotate. In order to ensure the stable and safe operation of the conveyor belt, it is necessary to detect the fault of the conveyor belt in time, or to repair and replace the conveyor belt before the fault causes damage to production. Generally, a speed sensor is used to collect data for conveyor belt slip fault detection, a visual sensor is used to collect data for conveyor belt deviation fault detection, and a tension sensor is used to collect data for conveyor belt break fault detection, etc., so as to obtain multimodal data. Then the fault prediction module uses the extracted feature data to establish a machine learning model for fault prediction, trains a classification model based on historical fault data to predict the fault of the conveyor belt, uses cluster analysis to detect abnormal fluctuations, helps identify potential fault states, uses a time series model to analyze the vibration data trend, and outputs fault predictions; formulates corresponding maintenance plans to prevent potential faults and reduce downtime and maintenance costs.

[0004] Among them, the process of detecting conveyor belt deviation faults through visual sensors is relatively complicated. Common methods include building a support frame and a camera directly above the conveyor belt, using deep learning target detection to find the position of the belt rollers, and adaptively determining the maximum displacement edge of the belt for belt conveyors of various specifications. The edge information of the belt is found by image processing methods, and its offset is determined. The fault can only be detected after the belt is offset. The requirements for laser projection, image pixels, conveyor belt color and image extraction models are high, resulting in more interference factors and complex technical processes in the process of conveyor belt deviation fault detection. It is very difficult to determine the edge of the conveyor belt, resulting in the inability to accurately predict the conveyor belt deviation fault. Summary of the invention

[0005] The object of the present invention is to provide a conveyor belt fault prediction system and method based on multimodal data analysis to solve the existing problems raised in the above background technology.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: a conveyor belt fault prediction system based on multimodal data analysis, the system comprising a belt conveyor and a fault prediction module;

[0007] The belt conveyor comprises a roller frame, rollers, a conveyor belt, a limit strip, a limit protrusion and an L-shaped frame, wherein a plurality of rollers are arranged on the roller frame, and the conveyor belt is wrapped around the outer walls of the plurality of rollers, and the two rollers located at both ends of the conveyor belt are limit rollers, and limit protrusions are symmetrically arranged at both ends of the outer walls of the limit rollers, and limit strips are fixedly arranged at both ends of the inner wall of the conveyor belt, and the limit strips are located between the limit protrusions at both ends, and the limit protrusions limit the position of the conveyor belt through the limit strips, and the outer wall of the roller frame is symmetrically provided with an L-shaped frame; the L-shaped frame is located in the middle of the conveyor belt, and a rectangular block 1 and a rectangular block 2 are arranged at the top of the L-shaped frame, and a limit strip, a rectangular block 1 and a rectangular block 2 on the same side form a group;

[0008] The fault prediction module comprises:

[0009] A data collection module, used to collect relevant modal data through a sensor, wherein the sensor includes a light intensity sensor for detecting light intensity;

[0010] The data pre-processing module is used to pre-process the collected modal data to remove the influence of noise on the modal data.

[0011] In the same group, the square block 1 and the square block 2 are respectively attached to the two sides of the limit bar, and the square block 1 and the square block 2 limit the position of the limit bar. The square block 1 and the square block 2 are both provided with grooves on one side close to the limit bar, and the groove includes an upper opening, and the bottom surface a of the groove is higher than the bottom surface b of the limit bar; the groove on the inner wall of the square block 1, the limit bar and the c surface of the conveyor belt constitute a closed cavity 1, and the groove on the inner wall of the square block 2, the limit bar and the c surface of the conveyor belt constitute a closed cavity 2, the inner wall of the closed cavity 1 is provided with a ray lamp, and the inner wall of the groove 2 is provided with a light intensity sensor for detecting the light intensity of the inner wall of the closed cavity 2.

[0012] The sensors also include a vibration sensor for monitoring the vibration characteristics of the conveyor belt, a sound sensor for capturing sound changes during operation, a temperature sensor for detecting the temperature of the conveyor belt and its components, and a current sensor for monitoring current fluctuations of the motor;

[0013] The pre-processing includes filtering, detrending and normalizing. In the filtering process, a low-pass filter or a high-pass filter is used to remove unnecessary high-frequency or low-frequency noise. In the detrending process, the trend is eliminated to analyze periodic changes. In the normalization process, the data are adjusted to the same standard to make effective comparisons.

[0014] The fault prediction module also includes:

[0015] The modal analysis module is used to extract the system dynamic characteristics through the frequency domain analysis unit and the modal shape recognition unit. The frequency domain analysis unit analyzes the system spectrum and identifies the main vibration modes and resonant frequencies through the fast Fourier transform method; the modal shape recognition unit identifies the relevant vibration modes by calculating and comparing the modal shapes;

[0016] A feature extraction module, used to extract feature data using modal analysis results, the feature data including amplitude for indicating vibration intensity, frequency for indicating key frequencies and resonant modes, directionality for indicating vibration direction and mode, and spectrum characteristics for indicating main frequency and bandwidth;

[0017] The fault prediction module is used to use the extracted feature data to build a machine learning model for fault prediction, train a classification model based on historical fault data to predict conveyor belt failures, use cluster analysis to detect abnormal fluctuations to help identify potential fault conditions, use time series models to analyze vibration data trends, and output fault predictions.

[0018] A conveyor belt fault prediction method based on multimodal data analysis, the method comprising:

[0019] S100, the roller frame supports the roller, the roller drives the conveyor belt to rotate, the limit roller limits the position of the limit bar through the limit convex block, the limit bar limits the position of the conveyor belt, so that the conveyor belt cannot deviate left and right during the rotation process, and the data collection module collects relevant modal data through the sensor, and the sensor includes a light intensity sensor for detecting light intensity;

[0020] S200, the data pre-processing module pre-processes the collected modal data to remove the influence of noise on the modal data;

[0021] S300, the modal analysis module extracts the system dynamic characteristics through the frequency domain analysis unit and the modal shape recognition unit. The frequency domain analysis unit analyzes the spectrum of the system through the fast Fourier transform method to identify the main vibration mode and resonant frequency; the modal shape recognition unit identifies the relevant vibration mode by calculating and comparing the modal shapes;

[0022] S400, a feature extraction module extracts feature data using the modal analysis result, wherein the feature data includes an amplitude for representing vibration intensity, a frequency for representing a key frequency and a resonant mode, a direction for representing a vibration direction and a mode direction, and a spectrum feature for representing a main frequency and a bandwidth;

[0023] S500, the fault prediction module uses the extracted feature data to build a machine learning model for fault prediction, trains a classification model based on historical fault data to predict conveyor belt failures, uses cluster analysis to detect abnormal fluctuations to help identify potential fault conditions, uses a time series model to analyze vibration data trends, and outputs fault predictions.

[0024] The sensors in S100 also include a vibration sensor for monitoring vibration characteristics of the conveyor belt, a sound sensor for capturing sound changes during operation, a temperature sensor for detecting the temperature of the conveyor belt and its components, and a current sensor for monitoring current fluctuations of the motor;

[0025] The pre-processing described in S200 includes filtering, detrending and normalizing. In the filtering process, a low-pass filter or a high-pass filter is used to remove unnecessary high-frequency or low-frequency noise. In the detrending process, the trend is eliminated to analyze periodic changes. In the normalization process, the data is adjusted to the same standard to make effective comparisons.

[0026] The relevant modal data collected by the sensor in S100 includes:

[0027] The ray lamp emits rays, which irradiate into a closed cavity;

[0028] In the closed cavity 2, the light intensity sensor obtains the light intensity value in the closed cavity 2 in real time.

[0029] The fault prediction module in S500 performs fault prediction including:

[0030] If the limit bar passing through the first matrix is ​​in normal use, the ray cannot pass through the limit bar to reach the second closed chamber. The light intensity in the second closed chamber is within the standard range. The output conveyor belt usage: the limit bar is not damaged and the conveyor belt is not offset.

[0031] If the limit bar passing through the matrix one is in an abnormal use state, the rays will pass through the limit bar and reach the closed cavity two, thereby enhancing the light intensity in the closed cavity two. The judgment conditions are divided according to the preset light intensity range, and the judgment results corresponding to different light intensity ranges include the limit bar being slightly damaged and the conveyor belt not being offset, the limit bar being moderately damaged and the conveyor belt not being offset, and the limit bar being severely damaged and the conveyor belt being offset. The judgment is output based on the light intensity range corresponding to the collected light intensity, and the offset that will occur in the subsequent time period corresponding to the output of the light damage, moderate damage and severe damage is output to complete the deviation fault prediction.

[0032] The abnormal use state includes the limit strip being broken and the connection between the limit strip and the c-surface being broken.

[0033] The method of extracting characteristic data using the modal analysis results in S400 includes: inputting data W1, W2...Wn After that, the relationship between the data is extracted, and different information layers are established according to different time nodes. 1 , G 2 …G n The first layer of parameters is calculated to form updated data, which is then combined with the original input data W1, W2…W n After the calculation is completed, valid feature data is obtained. The calculation formula is as follows:

[0034]

[0035] Among them, G represents the weight parameter matrix, W represents the vector expression of the input value, and Q represents the predicted output value. Indicates element-by-element multiplication. After completing the n-th layer of calculation, all feature data are output.

[0036] Compared with the prior art, the beneficial effects achieved by the present invention are:

[0037] The present invention does not need to use image processing methods to find the edge information of the conveyor belt to determine whether the conveyor belt is offset. Instead, the position of the conveyor belt is limited by a limit bar, so that the conveyor belt cannot be offset left and right during rotation. The ray lamp emits rays, and the rays are irradiated in the closed cavity one; in the closed cavity two, the light intensity sensor obtains the light intensity value in the closed cavity two in real time, and the fault prediction module performs fault prediction. If the limit bar passing through the matrix one is in an abnormal use state, the ray will pass through the limit bar to reach the closed cavity two, so that the light intensity in the closed cavity two is enhanced; the judgment conditions are divided according to the preset light intensity range, and different light intensity ranges are used. The corresponding judgment results include light damage to the limit bar and no displacement of the conveyor belt, moderate damage to the limit bar and no displacement of the conveyor belt, and severe damage to the limit bar and displacement of the conveyor belt. The judgment is output according to the light intensity range corresponding to the collected light intensity, and the displacement that will occur in the subsequent time period corresponding to the light damage, moderate damage and severe damage is output, which reduces the technical difficulty of detection and analysis, and improves the accuracy of conveyor belt displacement detection and deviation fault prediction; by extracting the relationship between the data, different information layers are established according to different time nodes, and valid feature data is analyzed and the results are output, thereby improving the accuracy of the extracted valid feature data. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention.

[0039] In the attached picture:

[0040] Figure 1 is a side view of the belt conveyor of the present invention;

[0041] Figure 2 is a bottom sectional view of the belt conveyor of the present invention;

[0042] Figure 3 is a front cross-sectional view of the belt conveyor of the present invention;

[0043] Figure 4 The present invention Figure 3 A partial enlarged view of the middle A;

[0044] Figure 5 It is a system composition diagram of the fault prediction module in the present invention;

[0045] In the figure: 1. roller frame; 2. roller; 3. conveyor belt; 4. limit strip; 5. limit convex block; 6. L-shaped frame; 7. rectangular block 1; 8. rectangular block 2; 9. groove; 10. light intensity sensor; 11. X-ray lamp. DETAILED DESCRIPTION

[0046] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0047] See also Figure 1-Figure 5 , the present invention provides a technical solution: a conveyor belt fault prediction system and method based on multimodal data analysis, including a conveyor belt fault prediction system based on multimodal data analysis, the system including a belt conveyor and a fault prediction module;

[0048] The belt conveyor comprises a roller frame 1, rollers 2, a conveyor belt 3, a limiting strip 4, a limiting protrusion 5 and an L-shaped frame 6. A plurality of rollers 2 are arranged on the roller frame 1. The conveyor belt 3 is wrapped around the outer walls of the plurality of rollers 2. The two rollers 2 located at both ends of the conveyor belt 3 are limiting rollers. The two ends of the outer walls of the limiting rollers are symmetrically provided with limiting protrusions 5. The two ends of the inner wall of the conveyor belt 3 are fixedly provided with limiting strips 4. The limiting strips 4 are located between the limiting protrusions 5 at both ends. The limiting protrusions 5 limit the position of the conveyor belt 3 through the limiting strips 4. The outer wall of the roller frame 1 is symmetrically provided with an L-shaped frame 6; the L-shaped frame 6 is located in the middle of the conveyor belt 3. A rectangular block 1 7 and a rectangular block 2 8 are provided at the top of the L-shaped frame 6. A limiting strip 4, a rectangular block 1 7 and a rectangular block 2 8 on the same side form a group.

[0049] The fault prediction module comprises:

[0050] A data collection module, used to collect relevant modal data through a sensor, wherein the sensor includes a light intensity sensor 10 for detecting light intensity;

[0051] The data pre-processing module is used to pre-process the collected modal data to remove the influence of noise on the modal data.

[0052] In the same group, the rectangular block 1 7 and the rectangular block 2 8 are respectively attached to the two sides of the limit bar 4, and the rectangular block 1 7 and the rectangular block 2 8 limit the position of the limit bar 4. The rectangular block 1 7 and the rectangular block 2 8 are both provided with a groove 9 on one side close to the limit bar 4, and the groove 9 includes an upper opening, and the bottom surface a of the groove 9 is higher than the bottom surface b of the limit bar 4; the groove 9 on the inner wall of the rectangular block 1 7, the limit bar 4 and the c surface of the conveyor belt 3 constitute a closed cavity 1, and the groove 9 on the inner wall of the rectangular block 2 8, the limit bar 4 and the c surface of the conveyor belt 3 constitute a closed cavity 2, and the inner wall of the closed cavity 1 is provided with a ray lamp 11, and the inner wall of the groove 9 2 is provided with a light intensity sensor 10 for detecting the light intensity of the inner wall of the closed cavity 2.

[0053] The sensors also include a vibration sensor for monitoring the vibration characteristics of the conveyor belt 3, a sound sensor for capturing sound changes during operation, a temperature sensor for detecting the temperature of the conveyor belt 3 and its components, and a current sensor for monitoring current fluctuations of the motor;

[0054] The pre-processing includes filtering, detrending and normalizing. In the filtering process, a low-pass filter or a high-pass filter is used to remove unnecessary high-frequency or low-frequency noise. In the detrending process, the trend is eliminated to analyze periodic changes. In the normalization process, the data are adjusted to the same standard to make effective comparisons.

[0055] The fault prediction module also includes:

[0056] The modal analysis module is used to extract the system dynamic characteristics through the frequency domain analysis unit and the modal shape recognition unit. The frequency domain analysis unit analyzes the system spectrum and identifies the main vibration modes and resonant frequencies through the fast Fourier transform method; the modal shape recognition unit identifies the relevant vibration modes by calculating and comparing the modal shapes;

[0057] A feature extraction module, used to extract feature data using modal analysis results, the feature data including amplitude for indicating vibration intensity, frequency for indicating key frequencies and resonant modes, directionality for indicating vibration direction and mode, and spectrum characteristics for indicating main frequency and bandwidth;

[0058] The fault prediction module is used to use the extracted feature data to establish a machine learning model for fault prediction, train a classification model based on historical fault data to predict the fault of conveyor belt 3, use cluster analysis to detect abnormal fluctuations to help identify potential fault conditions, use a time series model to analyze the vibration data trend, and output a fault prediction.

[0059] A conveyor belt fault prediction method based on multimodal data analysis, the method comprising:

[0060] S100, the roller frame 1 supports the roller 2, the roller 2 drives the conveyor belt 3 to rotate, the limiting roller limits the position of the limiting bar 4 through the limiting protrusion 5, the limiting bar 4 limits the position of the conveyor belt 3, so that the conveyor belt 3 cannot deviate left and right during the rotation process, and the data collection module collects relevant modal data through the sensor, and the sensor includes a light intensity sensor 10 for detecting light intensity;

[0061] The sensors in S100 also include a vibration sensor for monitoring the vibration characteristics of the conveyor belt 3, a sound sensor for capturing sound changes during operation, a temperature sensor for detecting the temperature of the conveyor belt 3 and its components, and a current sensor for monitoring current fluctuations of the motor;

[0062] The relevant modal data collected by the sensor in S100 includes:

[0063] The ray lamp 11 emits rays, which irradiate the closed chamber 1; in the closed chamber 2, the light intensity sensor 10 obtains the light intensity value in the closed chamber 2 in real time.

[0064] S200, the data pre-processing module pre-processes the collected modal data to remove the influence of noise on the modal data;

[0065] The pre-processing described in S200 includes filtering, detrending and normalizing. In the filtering process, a low-pass filter or a high-pass filter is used to remove unnecessary high-frequency or low-frequency noise. In the detrending process, the trend is eliminated to analyze periodic changes. In the normalization process, the data is adjusted to the same standard to make effective comparisons.

[0066] S300, the modal analysis module extracts the system dynamic characteristics through the frequency domain analysis unit and the modal shape recognition unit. The frequency domain analysis unit analyzes the spectrum of the system through the fast Fourier transform method to identify the main vibration mode and resonant frequency; the modal shape recognition unit identifies the relevant vibration mode by calculating and comparing the modal shapes;

[0067] S400, the feature extraction module extracts feature data using the modal analysis results, the feature data including the amplitude for indicating the vibration intensity, the frequency for indicating the key frequency and the resonant mode, the direction for indicating the vibration direction and the directionality of the mode, and the spectrum characteristics for indicating the main frequency and bandwidth; the feature data extracted using the modal analysis results in S400 includes: input data W1, W2...W n After that, the relationship between the data is extracted, and different information layers are established according to different time nodes. 1 , G 2 …G n The first layer of parameters is calculated to form updated data, which is then combined with the original input data W1, W2…W n After the calculation is completed, valid feature data is obtained. The calculation formula is as follows:

[0068]

[0069] Among them, G represents the weight parameter matrix, W represents the vector expression of the input value, and Q represents the predicted output value. Indicates element-by-element multiplication. After completing the n-th layer of calculation, all feature data are output.

[0070] S500, the fault prediction module uses the extracted feature data to establish a machine learning model for fault prediction, trains a classification model based on historical fault data to predict the fault of conveyor belt 3, uses cluster analysis to detect abnormal fluctuations to help identify potential fault conditions, uses a time series model to analyze vibration data trends, and outputs fault predictions.

[0071] The fault prediction module in S500 performs fault prediction including:

[0072] If the limit strip 4 passing through the square block 1 7 is in normal use, the rays cannot pass through the limit strip 4 to reach the closed cavity 2, the light intensity in the closed cavity 2 is within the standard range, and the use of the output conveyor belt 3 is: the limit strip 4 is not damaged and the conveyor belt 3 is not offset;

[0073] If the limit strip 4 passing through the matrix block 1 7 is in an abnormal use state, the rays will pass through the limit strip 4 and reach the closed cavity 2, thereby enhancing the light intensity in the closed cavity 2. The judgment conditions are divided according to the preset light intensity range, and the judgment results corresponding to different light intensity ranges include that the limit strip 4 is slightly damaged and the conveyor belt 3 is not offset, the limit strip 4 is moderately damaged and the conveyor belt 3 is not offset, and the limit strip 4 is severely damaged and the conveyor belt 3 is offset. The judgment is output based on the light intensity range corresponding to the collected light intensity, and the offset corresponding to the light damage, moderate damage and severe damage in the subsequent time period is output to complete the prediction of the deviation fault.

[0074] The abnormal use state includes the rupture of the limit strip 4 and the rupture of the connection between the limit strip 4 and the c-surface.

[0075] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A conveyor belt fault prediction system based on multimodal data analysis, characterized in that: The system includes a belt conveyor and a fault prediction module; The belt conveyor comprises a roller frame (1), rollers (2), a conveyor belt (3), a limiting strip (4), a limiting protrusion (5) and an L-shaped frame (6); a plurality of rollers (2) are arranged on the roller frame (1); the conveyor belt (3) is wrapped around the outer walls of the plurality of rollers (2); the two rollers (2) located at the two ends of the conveyor belt (3) are limiting rollers; limiting protrusions (5) are symmetrically arranged at the two ends of the outer wall of the limiting roller; limiting strips (4) are fixedly arranged at the two ends of the inner wall of the conveyor belt (3); The limiting strip (4) is located between the limiting protrusions (5) at both ends, and the limiting protrusions (5) limit the position of the conveyor belt (3) through the limiting strip (4). The outer wall of the roller frame (1) is symmetrically provided with an L-shaped frame (6); the L-shaped frame (6) is located in the middle of the conveyor belt (3), and a rectangular block 1 (7) and a rectangular block 2 (8) are provided at the top of the L-shaped frame (6). A limiting strip (4), a rectangular block 1 (7) and a rectangular block 2 (8) on the same side form a group; The fault prediction module comprises: A data collection module, used for collecting relevant modal data through a sensor, wherein the sensor includes a light intensity sensor (10) for detecting light intensity; The data pre-processing module is used to pre-process the collected modal data to remove the influence of noise on the modal data; In the same group, the first square block (7) and the second square block (8) are respectively attached to the two sides of the limit bar (4), the first square block (7) and the second square block (8) limit the position of the limit bar (4), and the first square block (7) and the second square block (8) are both provided with a groove (9) on one side close to the limit bar (4), the groove (9) includes an upper opening, and the bottom surface a of the groove (9) is higher than the bottom surface b of the limit bar (4); the groove (9) on the inner wall of the first square block (7), the limit bar (4) and the c surface of the conveyor belt (3) constitute a closed cavity one, and the groove (9) on the inner wall of the second square block (8), the limit bar (4) and the c surface of the conveyor belt (3) constitute a closed cavity two, the inner wall of the closed cavity one is provided with a ray lamp (11), and the inner wall of the second groove (9) is provided with a light intensity sensor (10) for detecting the light intensity of the inner wall of the closed cavity two; The sensors also include a vibration sensor for monitoring the vibration characteristics of the conveyor belt (3), a sound sensor for capturing sound changes during operation, a temperature sensor for detecting the temperature of the conveyor belt (3) and its components, and a current sensor for monitoring current fluctuations of the motor; The pre-processing includes filtering, detrending and normalizing. In the filtering process, a low-pass filter or a high-pass filter is used to remove unnecessary high-frequency or low-frequency noise. In the detrending process, the trend is eliminated to analyze periodic changes. In the normalizing process, the data are adjusted to the same standard to make effective comparisons. The fault prediction module also includes: The modal analysis module is used to extract the system dynamic characteristics through the frequency domain analysis unit and the modal shape recognition unit. The frequency domain analysis unit analyzes the system spectrum and identifies the main vibration modes and resonant frequencies through the fast Fourier transform method; the modal shape recognition unit identifies the relevant vibration modes by calculating and comparing the modal shapes; A feature extraction module, used to extract feature data using modal analysis results, the feature data including amplitude for indicating vibration intensity, frequency for indicating key frequencies and resonant modes, directionality for indicating vibration direction and mode, and spectrum characteristics for indicating main frequency and bandwidth; The fault prediction module is used to use the extracted feature data to establish a machine learning model for fault prediction, train a classification model based on historical fault data to predict the fault of the conveyor belt (3), use cluster analysis to detect abnormal fluctuations to help identify potential fault conditions, use a time series model to analyze the vibration data trend, and output a fault prediction.

2. A conveyor belt fault prediction method based on multimodal data analysis, applied to the conveyor belt fault prediction system based on multimodal data analysis as claimed in claim 1, characterized in that: The method includes: S100, the roller frame (1) supports the roller (2), the roller (2) drives the conveyor belt (3) to rotate, the limiting roller limits the position of the limiting strip (4) through the limiting protrusion (5), the limiting strip (4) limits the position of the conveyor belt (3), so that the conveyor belt (3) cannot deviate left or right during the rotation process, and the data collection module collects relevant modal data through a sensor, and the sensor includes a light intensity sensor (10) for detecting light intensity; S200, the data pre-processing module pre-processes the collected modal data to remove the influence of noise on the modal data; S300, the modal analysis module extracts the system dynamic characteristics through the frequency domain analysis unit and the modal shape recognition unit. The frequency domain analysis unit analyzes the spectrum of the system through the fast Fourier transform method to identify the main vibration mode and resonant frequency; the modal shape recognition unit identifies the relevant vibration mode by calculating and comparing the modal shapes; S400, a feature extraction module extracts feature data using the modal analysis result, wherein the feature data includes an amplitude for representing vibration intensity, a frequency for representing a key frequency and a resonant mode, a direction for representing a vibration direction and a mode direction, and a spectrum feature for representing a main frequency and a bandwidth; S500, the fault prediction module uses the extracted feature data to establish a machine learning model for fault prediction, trains a classification model based on historical fault data to predict the fault of the conveyor belt (3), uses cluster analysis to detect abnormal fluctuations to help identify potential fault conditions, uses a time series model to analyze vibration data trends, and outputs fault predictions.

3. The conveyor belt fault prediction method based on multimodal data analysis according to claim 2 is characterized in that: The sensors in S100 also include a vibration sensor for monitoring the vibration characteristics of the conveyor belt (3), a sound sensor for capturing sound changes during operation, a temperature sensor for detecting the temperature of the conveyor belt (3) and its components, and a current sensor for monitoring current fluctuations of the motor; The pre-processing in S200 includes filtering, detrending and normalization. In the filtering process, a low-pass filter or a high-pass filter is used to remove unnecessary high-frequency or low-frequency noise; In the process of detrending, the trend is removed so that cyclical changes can be analyzed, and in the process of normalization, the data are adjusted to the same standard so that valid comparisons can be made.

4. The conveyor belt fault prediction method based on multimodal data analysis according to claim 3 is characterized in that: The relevant modal data collected by the sensor in S100 includes: The ray lamp (11) emits rays, which irradiate the closed chamber. In the second closed chamber, the light intensity sensor (10) acquires the light intensity value in the second closed chamber in real time.

5. The conveyor belt fault prediction method based on multimodal data analysis according to claim 2 is characterized in that: The fault prediction module in S500 performs fault prediction including: If the limit strip (4) passing through the square block 1 (7) is in normal use, the rays cannot pass through the limit strip (4) to reach the closed cavity 2, the light intensity in the closed cavity 2 is within the standard range, and the output conveyor belt (3) is in use: the limit strip (4) is not damaged and the conveyor belt (3) is not offset; If the limit bar (4) passing through the matrix block 1 (7) is in an abnormal use state, the ray will pass through the limit bar (4) to reach the closed cavity 2, thereby enhancing the light intensity in the closed cavity 2; the judgment conditions are divided according to the preset light intensity range, and the judgment results corresponding to different light intensity ranges include the limit bar (4) being slightly damaged and the conveyor belt (3) not being offset, the limit bar (4) being moderately damaged and the conveyor belt (3) not being offset, and the limit bar (4) being severely damaged and the conveyor belt (3) being offset; the judgment is output according to the light intensity range corresponding to the collected light intensity, and the offset that will occur in the subsequent time period corresponding to the light damage, moderate damage and severe damage is output, thereby completing the prediction of the deviation fault.

6. The conveyor belt fault prediction method based on multimodal data analysis according to claim 5, characterized in that: The abnormal use state includes the rupture of the limiting strip (4) and the rupture of the connection between the limiting strip (4) and the c-surface.

7. The conveyor belt fault prediction method based on multimodal data analysis according to claim 2, characterized in that: The method of extracting characteristic data using the modal analysis results in S400 includes: inputting data W1, W2...W n After that, the relationship between the data is extracted, and different information layers are established according to different time nodes. 1 , G 2 …G n The first layer of parameters is calculated to form updated data, which is then combined with the original input data W1, W2…W n After the calculation is completed, valid feature data is obtained. The calculation formula is as follows: Among them, G represents the weight parameter matrix, W represents the vector expression of the input value, and Q represents the predicted output value. It means that the corresponding elements are multiplied. After completing the n-th layer of calculation, all the feature data are output.

Images