A conveying belt tear evolution state intelligent diagnosis and trend prediction system

Abstract

The present application relates to the technical field of intelligent monitoring and fault diagnosis of industrial conveying equipment, in particular to a kind of conveying belt tearing evolution state intelligent diagnosis and trend prediction system, comprising: data acquisition center, for obtaining real-time stress data;Physical prediction unit, for dynamic evolution analysis on the system state of last time, to obtain prior state estimation;Neural observation unit, for obtaining observed state;Fusion correction unit, for obtaining optimal posterior state estimation;Trend diagnosis unit, for fracture risk evaluation feedback analysis on crack parameter in optimal posterior state estimation, to obtain safety margin coefficient, discriminant processing is carried out to safety margin coefficient, to obtain maintenance warning signal, emergency shutdown signal or normal operation signal;The present application solves the problem that the prior art lacks deep understanding of crack evolution trend, and provides solid theoretical data support for predicting potential failure.

Description

Technical Field

[0001] This invention relates to the field of intelligent monitoring and fault diagnosis technology for industrial conveying equipment, specifically to an intelligent diagnosis and trend prediction system for the tear evolution state of conveyor belts. Background Technology

[0002] In the field of heavy-duty industrial conveying, conveyor belts are key load-bearing equipment, and their structural integrity directly determines the continuity and safety of the production line. During long-term high-load operation, belts are subject to alternating stress and material impact, making them prone to cracking and eventually longitudinal tearing. To avoid the risk of sudden shutdowns, companies typically use various monitoring methods to inspect the surface and internal condition of the belts. However, traditional monitoring methods often focus on post-event alarms or single-dimensional condition characterization, lacking the ability to dynamically perceive the entire process of crack evolution.

[0003] While existing technologies employ detection schemes based on physical models or machine vision, these methods generally have limitations. Pure physical models struggle to accurately adapt to complex and variable operating parameters, while single data-driven models lack robustness in the face of nonlinear noise interference. Furthermore, current technologies struggle to achieve deep integration of prior physical knowledge with multi-source observation data, resulting in low estimation accuracy for key evolution parameters such as crack propagation rate and energy release rate, and a lack of quantitative assessment and graded early warning mechanisms for system fracture risk trends. Therefore, providing a system that integrates physical dynamics evolution with neural observation features to achieve accurate estimation and quantitative diagnosis of conveyor belt tear evolution is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides an intelligent diagnostic and trend prediction system for the tear evolution state of conveyor belts. Specifically, the technical solution of this invention includes:

[0005] The data acquisition center is used to retrieve the basic operating data of the conveyor belt and send the basic operating data to the physical prediction unit for stress field calculation and analysis to obtain real-time stress data;

[0006] The physical prediction unit is used to perform dynamic evolution analysis on the system state at the previous moment to obtain a priori state estimate, which includes crack evolution parameters derived from the physical model.

[0007] The neural observation unit is used to perform nonlinear mapping analysis on the acquired belt surface images and acoustic emission features to obtain the observation state, which includes crack characterization parameters identified based on deep learning.

[0008] The fusion correction unit is used to perform optimal estimation weighted analysis on the prior state estimate and the observed state to obtain the optimal posterior state estimate, which is the system state after closed-loop fusion of data and physical model.

[0009] The trend diagnosis unit is used to perform fracture risk assessment and feedback analysis on the crack parameters in the optimal posterior state estimate, obtain the safety margin coefficient, and perform discrimination processing on the safety margin coefficient to obtain maintenance warning signal, emergency shutdown signal or normal operation signal.

[0010] Preferably, the stress field calculation and analysis process is as follows:

[0011] Collect the real-time longitudinal tension and effective cross-sectional area of ​​the conveyor belt;

[0012] Divide the real-time longitudinal tension by the effective cross-sectional area of ​​the belt to obtain the real-time stress data;

[0013] Real-time stress data characterizes the average tensile stress inside the belt at the current moment.

[0014] Preferably, the dynamic evolution analysis process is as follows:

[0015] Obtain the system state at the previous moment, which includes the equivalent crack length, crack propagation rate, and energy release rate;

[0016] Based on a preset sampling time interval, the linear cumulative value of the equivalent crack length and crack propagation rate is calculated, and the linear cumulative value is set as the prior crack length.

[0017] Set the crack propagation rate to the prior crack propagation rate;

[0018] Obtain real-time stress data and the elastic modulus of the conveyor belt cover rubber;

[0019] Fracture mechanics energy is calculated based on real-time stress data, prior crack length, and elastic modulus to obtain the prior energy release rate.

[0020] The prior crack length, prior crack propagation rate, and prior energy release rate are combined to generate a prior state estimate.

[0021] Preferably, the nonlinear mapping analysis process is as follows:

[0022] Acquire the belt surface image matrix and acoustic radio frequency domain feature vector collected by the data acquisition center;

[0023] The deep neural network model is invoked, and the belt surface image matrix and the acoustic radio frequency domain feature vector are input into the deep neural network model for feature extraction and regression calculation to obtain the output vector;

[0024] Extract the components corresponding to crack length, crack propagation rate, and energy release rate from the output vector;

[0025] The extracted components are combined to generate the observation state, which includes observation noise.

[0026] The preferred, optimal estimation weighted analysis process is as follows:

[0027] Obtain the prior error covariance matrix and the current observation noise covariance matrix at the current time.

[0028] Calculate the Kalman gain based on the prediction error covariance matrix and the observation noise covariance matrix;

[0029] Calculate the difference between the observed state and the prior state estimate, and set the difference as the innovation residual;

[0030] The correction term is obtained by multiplying the Kalman gain by the innovation residual.

[0031] The prior state estimate is added to the correction term to generate the optimal posterior state estimate.

[0032] Preferably, it also includes an energy mutation monitoring unit, and the analysis process of the energy mutation monitoring unit is as follows:

[0033] Extract the observed energy release rate component from the observed state;

[0034] Extract the prior energy release rate component from the prior state estimate;

[0035] Calculate the absolute value of the difference between the observed energy release rate component and the prior energy release rate component, and set the absolute value as the energy mutation factor;

[0036] The energy mutation factor is compared and analyzed with a preset statistical threshold.

[0037] If the energy mutation factor is greater than the preset statistical threshold, a tearing acceleration abnormal signal is generated;

[0038] If the energy mutation factor is less than or equal to the preset statistical threshold, a stable operation signal is generated.

[0039] Preferably, the fracture risk assessment feedback analysis process is as follows:

[0040] Extract the corrected crack length from the optimal posterior state estimate;

[0041] Obtain real-time stress data;

[0042] Real-time stress intensity factor is calculated based on real-time stress data and corrected crack length.

[0043] Obtain the fracture toughness threshold of the material;

[0044] The result is obtained by subtracting the ratio of the real-time stress intensity factor to the fracture toughness threshold from the calculated value 1.

[0045] Set the calculation result as the safety margin factor.

[0046] Preferably, the discrimination process is as follows:

[0047] The safety margin coefficient is compared and analyzed with the preset alarm threshold and the preset safety threshold.

[0048] If the safety margin coefficient is less than or equal to the preset alarm threshold, an emergency stop signal is generated. The emergency stop signal is used to trigger the braking operation of the conveyor system.

[0049] If the safety margin coefficient is greater than the preset alarm threshold but less than the preset safety threshold, a maintenance warning signal is generated.

[0050] If the safety margin coefficient is greater than or equal to the preset safety threshold, a normal operation signal is generated.

[0051] Compared with the prior art, the present invention has the following beneficial effects:

[0052] 1. This invention combines the prior state of the physical prediction unit with the observation state of the neural observation unit by integrating the correction unit, thereby achieving the complementary advantages of physical model-driven and data-driven approaches. By using optimal estimation weighted analysis, closed-loop correction is performed on the crack evolution parameters based on dynamic deduction and the characterization parameters based on deep learning recognition. This effectively overcomes the problems of poor adaptability of pure physical models under complex working conditions and insufficient noise resistance of single data models, and significantly improves the accuracy and robustness of state estimation.

[0053] 2. This invention establishes a comprehensive dynamic evolution analysis mechanism, which calculates fracture mechanics energy based on real-time stress data and material elastic modulus. The system can extrapolate the equivalent crack length, crack propagation rate, and energy release rate in real time, realizing dynamic and continuous perception of the entire process of belt tearing evolution. This dynamic extrapolation based on physical mechanisms solves the problem of existing technologies lacking a deep understanding of crack evolution trends, and provides solid theoretical data support for predicting potential failures.

[0054] 3. This invention introduces an energy mutation monitoring unit, which calculates the energy mutation factor by comparing the difference between the observed energy release rate and the prior energy release rate. This mechanism can keenly capture abnormal energy fluctuations caused by nonlinear disturbances or internal defects during the tearing process and promptly identify the accelerated propagation state of the crack. By comparing with statistical thresholds, the system can issue abnormal signals before the crack enters the unstable stage, effectively improving the early warning capability for sudden tearing failures.

[0055] 4. This invention establishes a fracture risk assessment and feedback analysis system. It uses real-time stress intensity factor and material fracture toughness threshold to calculate safety margin coefficient, thereby realizing a quantitative assessment of belt fracture risk. The system classifies and distinguishes according to the safety margin coefficient, and intelligently outputs normal operation, maintenance warning or emergency shutdown signals. This graded warning mechanism not only avoids production losses caused by blind shutdown, but also triggers braking when the risk is critical, thus balancing production efficiency and equipment safety. Attached Figure Description

[0056] The present invention will be further explained below with reference to the accompanying drawings and embodiments:

[0057] Figure 1 This is a structural diagram of the system of the present invention. Detailed Implementation

[0058] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0059] Example 1:

[0060] Please see Figure 1 A smart diagnostic and trend prediction system for the tear evolution state of a conveyor belt, comprising:

[0061] The data acquisition center is used to retrieve the basic operating data of the conveyor belt and send the basic operating data to the physical prediction unit for stress field calculation and analysis to obtain real-time stress data;

[0062] The physical prediction unit is used to perform dynamic evolution analysis on the system state at the previous moment to obtain a priori state estimate, which includes crack evolution parameters derived from the physical model.

[0063] The neural observation unit is used to perform nonlinear mapping analysis on the acquired belt surface images and acoustic emission features to obtain the observation state, which includes crack characterization parameters identified based on deep learning.

[0064] The fusion correction unit is used to perform optimal estimation weighted analysis on the prior state estimate and the observed state to obtain the optimal posterior state estimate, which is the system state after closed-loop fusion of data and physical model.

[0065] The trend diagnosis unit is used to perform fracture risk assessment and feedback analysis on the crack parameters in the optimal posterior state estimate, obtain the safety margin coefficient, and perform discrimination processing on the safety margin coefficient to obtain maintenance warning signal, emergency shutdown signal or normal operation signal.

[0066] A smart diagnostic and trend prediction system for the tear evolution state of a conveyor belt includes a data acquisition center, a physical prediction unit, a neural observation unit, a fusion correction unit, and a trend diagnosis unit. The data acquisition center retrieves basic operating data of the conveyor belt and sends it to the physical prediction unit for stress field calculation and analysis to obtain real-time stress data. The physical prediction unit performs dynamic evolution analysis on the system state at the previous moment to obtain a priori state estimate, which includes crack evolution parameters derived from a physical model. The neural observation unit performs nonlinear mapping analysis on the acquired belt surface images and acoustic emission characteristics to obtain the observed state, which includes crack characterization parameters identified based on deep learning. The fusion correction unit performs optimal estimation weighted analysis on the priori state estimate and the observed state to obtain an optimal posterior state estimate, which is the system state after closed-loop fusion of data and the physical model. The trend diagnosis unit performs fracture risk assessment feedback analysis on the crack parameters in the optimal posterior state estimate to obtain a safety margin coefficient, and performs discrimination processing on the safety margin coefficient to obtain maintenance warning signals, emergency shutdown signals, or normal operation signals.

[0067] Example 2:

[0068] The real-time stress data calculation and analysis process is as follows:

[0069] Collect the real-time longitudinal tension and effective cross-sectional area of ​​the conveyor belt;

[0070] Divide the real-time longitudinal tension by the effective cross-sectional area of ​​the belt to obtain the real-time stress data;

[0071] Real-time stress data characterizes the average tensile stress inside the belt at the current moment.

[0072] The real-time net cross-sectional stress calculation and analysis process is as follows; this embodiment focuses on monitoring transverse propagation cracks in conveyor belts; real-time longitudinal tension of the conveyor belt is collected. ; Obtain the crack length at the previous moment as fed back by the optimal posterior state estimate. Because the sampling interval is extremely short, it is used as a priori reference value for the crack length at the current moment; based on the original width of the belt. and thickness Calculate the current effective cross-sectional area If the current effective cross-sectional area Less than the preset minimum cross-section threshold Then the real-time net cross-sectional stress data will be used. Set to the preset fault limit stress value; otherwise, apply real-time longitudinal tension. Divide by the current effective cross-sectional area Obtain real-time net cross-sectional stress data The calculation formula is: Real-time stress data It represents the average tensile stress inside the belt at the current moment.

[0073] Example 3:

[0074] The dynamic evolution analysis process is as follows:

[0075] Obtain the system state at the previous moment, which includes the equivalent crack length, crack propagation rate, and energy release rate;

[0076] Based on a preset sampling time interval, the linear cumulative value of the equivalent crack length and crack propagation rate is calculated, and the linear cumulative value is set as the prior crack length.

[0077] Set the crack propagation rate to the prior crack propagation rate;

[0078] Obtain real-time stress data and the elastic modulus of the conveyor belt cover rubber;

[0079] Fracture mechanics energy is calculated based on real-time stress data, prior crack length, and elastic modulus to obtain the prior energy release rate.

[0080] The prior crack length, prior crack propagation rate, and prior energy release rate are combined to generate a prior state estimate.

[0081] The dynamic evolution analysis process is as follows: obtain the system state at the previous moment. The system state includes the equivalent crack length. Crack propagation rate and energy release rate Based on a preset sampling time interval Calculate the equivalent crack length With crack propagation rate The linear cumulative value is set as the prior crack length. The calculation formula is: ; Crack propagation rate Set as the prior crack propagation rate ,Right now Obtain the real-time stress data calculated by the steps in Example 2. The elastic modulus of the conveyor belt cover rubber, which has been pre-determined and stored through standard tensile testing. ;

[0082] Based on real-time stress data Prior crack length and elastic modulus The prior energy release rate was obtained by performing fracture mechanics energy calculations. The calculation follows the theory of linear elastic fracture mechanics, and the formula is as follows: ; the prior crack length Prior crack propagation rate and prior energy release rate Combinatorial generation of prior state estimates Simultaneously, based on the posterior error covariance matrix of the previous time step... With the preset process noise covariance matrix The matrix It is obtained by pre-calculating the variance of the predicted residual sequence of the statistical physical evolution model on a historical standard dataset, and is used to characterize the uncertainty of the physical model. Among them, the prior state estimation The first two terms are updated based on linear kinematic equations, and the third term is updated nonlinearly based on the aforementioned fracture mechanics energy formula. When predicting the error covariance matrix, a linearization approximation is used for covariance prediction and updating, resulting in the prior error covariance matrix. The calculation formula is: ,in This is the state transition matrix; considering the crack length... With crack propagation rate It follows a linear kinematic relationship, while the energy release rate Although calculated using a nonlinear formula, it can be considered a linear transfer over a very short time; the state transition matrix... Defined as follows Matrix, corresponding to the state vector :

[0083]

[0084] Note: Regarding energy release rate The nonlinear evolution error is mainly manifested through the process noise covariance matrix. The third diagonal element is compensated.

[0085] Example 4:

[0086] The nonlinear mapping analysis process is as follows:

[0087] Acquire the belt surface image matrix and acoustic radio frequency domain feature vector collected by the data acquisition center;

[0088] The deep neural network model is invoked, and the belt surface image matrix and the acoustic radio frequency domain feature vector are input into the deep neural network model for feature extraction and regression calculation to obtain the output vector;

[0089] Extract the components corresponding to crack length, crack propagation rate, and energy release rate from the output vector;

[0090] The extracted components are combined to generate the observation state, which includes observation noise.

[0091] The nonlinear mapping analysis process is as follows: The data acquisition center retrieves the basic operating data of the conveyor belt, which includes the belt surface image matrix. With acoustic radio frequency domain feature vector In order to overcome the difficulties of long-distance monitoring of conveyor belts and the noise interference from the production environment, the acoustic emission radio frequency domain feature vector was developed. The data acquisition adopts a fixed-point near-field monitoring method, that is, a resonant acoustic emission sensor array is installed in the easily torn critical section of the conveyor belt, such as the buffer idler frame below the material drop outlet. Utilizing the frequency domain difference between the high-frequency characteristics of the acoustic emission signal and the low-frequency characteristics of the environmental mechanical noise, a pre-pass hardware bandpass filter with a passband range of 30kHz to 150kHz is configured to extract high-frequency data and construct the acoustic emission radio frequency domain feature vector. This allows for the extraction of effective crack features from complex noise as the conveyor belt passes the acquisition point; acoustic emission radio frequency domain feature vector To characterize the frequency domain data of internal crack evolution parameters of the belt, a deep neural network model was invoked. This model employs a two-stream architecture, including a matrix for processing the belt surface image. Convolutional neural network branches and features for processing acoustic radio frequency domain feature vectors The fully connected network branches are fused through a splicing layer; the specific process is as follows: the belt surface image matrix is... Visual feature vectors are extracted from the input convolutional neural network branches. Simultaneously, the acoustic emission radio frequency domain feature vector Acoustic feature vectors are extracted from the fully connected network branches. Subsequently, the visual feature vectors are processed through a splicing layer. Acoustic eigenvectors By performing channel-level connections, the observation state can be obtained. The corresponding fusion feature vector The calculation formula is expressed as follows: The observed state is represented in vector form. , where z a Represents the observed crack length, z v z represents the observed crack propagation rate. G This represents the observed energy release rate and the observation status. Includes observation noise The observation noise Follows a normal distribution ,in This is the observation noise covariance matrix; The distribution of predicted residuals on the validation set is determined in advance using a statistical neural network.

[0092] Example 5:

[0093] The optimal estimation weighted analysis process is as follows:

[0094] Obtain the prior error covariance matrix and the current observation noise covariance matrix at the current time.

[0095] Calculate the Kalman gain based on the prediction error covariance matrix and the observation noise covariance matrix;

[0096] Calculate the difference between the observed state and the prior state estimate, and set the difference as the innovation residual;

[0097] The correction term is obtained by multiplying the Kalman gain by the innovation residual.

[0098] The prior state estimate is added to the correction term to generate the optimal posterior state estimate.

[0099] The optimal estimation weighted analysis process is as follows: obtain the prior error covariance matrix at the current time. With the current observation noise covariance matrix Based on the prediction error covariance matrix Covariance matrix of observation noise Calculate Kalman gain The calculation formula is: ; Calculate the observed state Compared with prior state estimation The difference is set as the new information residual. ,Right now ; Calculate Kalman gain With the new residual The product yields the correction term; the prior state estimate... Adding the correction term to generate the optimal posterior state estimate The calculation formula is: .

[0100] Example 6:

[0101] This system also includes an energy mutation monitoring unit, and the analysis process of the energy mutation monitoring unit is as follows:

[0102] Extract the observed energy release rate component from the observed state;

[0103] Extract the prior energy release rate component from the prior state estimate;

[0104] Calculate the absolute value of the difference between the observed energy release rate component and the prior energy release rate component, and set the absolute value as the energy mutation factor;

[0105] The energy mutation factor is compared and analyzed with a preset statistical threshold.

[0106] If the energy mutation factor is greater than the preset statistical threshold, a tearing acceleration abnormal signal is generated;

[0107] If the energy mutation factor is less than or equal to the preset statistical threshold, a stable operation signal is generated.

[0108] The analysis process of the energy mutation monitoring unit is as follows: extracting the observation status. The observed energy release rate component Extracting prior state estimates The prior energy release rate component ; Calculate the observed energy release rate components With the prior energy release rate component The absolute value of the difference is set as the energy mutation factor. The formula is ; energy mutation factor Compared with the preset statistical threshold Perform comparative analysis; statistical thresholds The method for determining the value is as follows: collect historical residual data of the system under normal operating conditions, and calculate the standard deviation of the historical data. ,set up If energy mutation factor Greater than the preset statistical threshold Generates tearing acceleration abnormal signals; if energy mutation factor Less than or equal to the preset statistical threshold Generate a stable operating signal.

[0109] Example 7:

[0110] The fracture risk assessment feedback analysis process is as follows:

[0111] Extract the corrected crack length from the optimal posterior state estimate;

[0112] Obtain real-time stress data;

[0113] Real-time stress intensity factor is calculated based on real-time stress data and corrected crack length.

[0114] Obtain the fracture toughness threshold of the material;

[0115] The result is obtained by subtracting the ratio of the real-time stress intensity factor to the fracture toughness threshold from the calculated value 1.

[0116] Set the calculation result as the safety margin factor.

[0117] The fracture risk assessment feedback analysis process is as follows: Extracting the optimal posterior state estimate. Corrected crack length Obtain the real-time stress data calculated by the steps in Example 2. Based on real-time stress data With modified crack length Calculate the real-time stress intensity factor The calculation formula is: Obtain the fracture toughness threshold of the material. The threshold This is a constant obtained through standard material fracture toughness testing; the calculated value is 1 minus the real-time stress intensity factor. With fracture toughness threshold The ratio is used to obtain the calculation result; the calculation result is set as the safety margin factor. The formula is .

[0118] Example 8:

[0119] The discrimination and processing procedure is as follows:

[0120] The safety margin coefficient is compared and analyzed with the preset alarm threshold and the preset safety threshold.

[0121] If the safety margin coefficient is less than or equal to the preset alarm threshold, an emergency stop signal is generated. The emergency stop signal is used to trigger the braking operation of the conveyor system.

[0122] If the safety margin coefficient is greater than the preset alarm threshold but less than the preset safety threshold, a maintenance warning signal is generated.

[0123] If the safety margin coefficient is greater than or equal to the preset safety threshold, a normal operation signal is generated.

[0124] The discrimination process is as follows: The safety margin coefficient is... Compared with the preset alarm threshold and preset security thresholds Comparative analysis was performed; preset alarm thresholds were set. In this embodiment, a safety threshold of 0.15 is set based on typical test data of steel cord conveyor belts. The threshold is set to 0.40. In practical applications, the threshold can be calibrated based on the specific material fracture toughness of the conveyor belt and historical failure data; if the safety margin coefficient... Less than or equal to the preset alarm threshold An emergency stop signal is generated, which is used to trigger the braking operation of the conveyor system; if the safety margin factor Greater than the preset alarm threshold And less than the preset safety threshold Generate maintenance early warning signals; if the safety margin coefficient Greater than or equal to the preset safety threshold Generate normal operation signals.

[0125] 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.

Claims

1. A system for intelligent diagnosis and trend prediction of the evolution of a conveyor belt tear, characterized in that, include: The data acquisition center is used to retrieve the basic operating data of the conveyor belt and send the basic operating data to the physical prediction unit for stress field calculation and analysis to obtain real-time stress data; The physical prediction unit is used to perform dynamic evolution analysis on the system state at the previous moment to obtain a priori state estimate, which includes crack evolution parameters derived from the physical model. The neural observation unit is used to perform nonlinear mapping analysis on the acquired belt surface images and acoustic emission features to obtain the observation state, which includes crack characterization parameters identified based on deep learning. The fusion correction unit is used to perform optimal estimation weighted analysis on the prior state estimate and the observed state to obtain the optimal posterior state estimate, which is the system state after closed-loop fusion of data and physical model. The trend diagnosis unit is used to perform fracture risk assessment and feedback analysis on the crack parameters in the optimal posterior state estimation, obtain the safety margin coefficient, and perform discrimination processing on the safety margin coefficient to obtain maintenance warning signal, emergency shutdown signal or normal operation signal. The dynamic evolution analysis process is as follows: Obtain the system state at the previous moment, which includes the equivalent crack length, crack propagation rate, and energy release rate; Based on a preset sampling time interval, the linear cumulative value of the equivalent crack length and crack propagation rate is calculated, and the linear cumulative value is set as the prior crack length. Set the crack propagation rate to the prior crack propagation rate; Obtain real-time stress data and the elastic modulus of the conveyor belt cover rubber; Fracture mechanics energy is calculated based on real-time stress data, prior crack length, and elastic modulus to obtain the prior energy release rate. The prior crack length, prior crack propagation rate, and prior energy release rate are combined to generate a prior state estimate. The optimal estimation weighted analysis process is as follows: Obtain the prior error covariance matrix and the current observation noise covariance matrix at the current time. Calculate the Kalman gain based on the prediction error covariance matrix and the observation noise covariance matrix; Calculate the difference between the observed state and the prior state estimate, and set the difference as the innovation residual; The correction term is obtained by multiplying the Kalman gain by the innovation residual. The prior state estimate is added to the correction term to generate the optimal posterior state estimate; It also includes an energy mutation monitoring unit, and the analysis process of the energy mutation monitoring unit is as follows: Extract the observed energy release rate component from the observed state; Extract the prior energy release rate component from the prior state estimate; Calculate the absolute value of the difference between the observed energy release rate component and the prior energy release rate component, and set the absolute value as the energy mutation factor; The energy mutation factor is compared and analyzed with a preset statistical threshold. If the energy mutation factor is greater than the preset statistical threshold, a tearing acceleration abnormal signal is generated; If the energy mutation factor is less than or equal to the preset statistical threshold, a stable operation signal is generated; The fracture risk assessment feedback analysis process is as follows: Extract the corrected crack length from the optimal posterior state estimate; Obtain real-time stress data; Real-time stress intensity factor is calculated based on real-time stress data and corrected crack length. Obtain the fracture toughness threshold of the material; The result is obtained by subtracting the ratio of the real-time stress intensity factor to the fracture toughness threshold from the calculated value 1. Set the calculation result as the safety margin factor.

2. The intelligent diagnosis and trend prediction system for the tear evolution state of a conveyor belt according to claim 1, characterized in that, The stress field calculation and analysis process is as follows: Collect the real-time longitudinal tension and effective cross-sectional area of ​​the conveyor belt; Divide the real-time longitudinal tension by the effective cross-sectional area of ​​the belt to obtain the real-time stress data; Real-time stress data characterizes the average tensile stress inside the belt at the current moment.

3. The intelligent diagnosis and trend prediction system for the tear evolution state of a conveyor belt according to claim 1, characterized in that, The nonlinear mapping analysis process is as follows: Acquire the belt surface image matrix and acoustic radio frequency domain feature vector collected by the data acquisition center; The deep neural network model is invoked, and the belt surface image matrix and the acoustic radio frequency domain feature vector are input into the deep neural network model for feature extraction and regression calculation to obtain the output vector; Extract the components corresponding to crack length, crack propagation rate, and energy release rate from the output vector; The extracted components are combined to generate the observation state, which includes observation noise.

4. The intelligent diagnosis and trend prediction system for the tear evolution state of a conveyor belt according to claim 3, characterized in that, The discrimination and processing procedure is as follows: The safety margin coefficient is compared and analyzed with the preset alarm threshold and the preset safety threshold. If the safety margin coefficient is less than or equal to the preset alarm threshold, an emergency stop signal is generated. The emergency stop signal is used to trigger the braking operation of the conveyor system. If the safety margin coefficient is greater than the preset alarm threshold but less than the preset safety threshold, a maintenance warning signal is generated. If the safety margin coefficient is greater than or equal to the preset safety threshold, a normal operation signal is generated.

Images