PHM-based coal mine equipment fault detection method, medium and system

Abstract

The invention provides a PHM-based coal mine equipment fault detection method, medium and system, and belongs to the technical field of coal mine equipment fault detection.The PHM-based coal mine equipment fault detection method comprises the steps that operation data including vibration, current and other indexes are collected based on a PHM; preprocessing, denoising and feature extraction are carried out on the data; based on the preprocessed data, a mathematical model covering dynamics, electrical characteristics, wear and the like is established, and an optimization model is obtained through parameter identification and optimization; discretizing the optimization model, constructing a multi-dimensional state space, and defining a normal state and a fault state. Searching the shortest path from the normal state to each fault state based on an A * algorithm, calculating the difficulty degree of fault occurrence, and obtaining a fault weight vector; and real-time data is input into the optimization model, the distance from the current state to each fault end point is calculated, the fault weight is combined, the fault probability distribution and the early warning information are output, and the technical problem that the requirements of fault diagnosis accuracy, reliability and interpretability are difficult to meet in the prior art is solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of coal mine equipment fault detection, and in particular, relates to a coal mine equipment fault detection method, medium and system based on PHM. Background Art

[0002] Due to the harsh environment of coal mines, equipment is subjected to harsh conditions such as vibration, shock, high temperature and high humidity for a long time, and is prone to various faults, such as bearing wear, gear damage, motor winding short circuit, etc. Once these faults occur, they will not only cause equipment damage, but may also cause serious safety accidents. Therefore, how to effectively diagnose and warn coal mine equipment has become a key issue of concern in the industry. At present, the common methods of coal mine equipment fault diagnosis mainly include: fault diagnosis based on empirical rules, fault diagnosis based on data-driven, and fault diagnosis based on physical models.

[0003] 1. The fault diagnosis method based on empirical rules mainly relies on the experience and knowledge of experts, establishes a fault diagnosis rule base, and identifies faults through threshold judgment of characteristic parameters. This method is simple and intuitive, but has limited scope of application and cannot cope with complex multi-fault situations.

[0004] 2. Data-driven fault diagnosis methods use machine learning, neural networks and other technologies to extract fault features from a large amount of operating data and train a fault identification model. This method can adapt to complex faults, but it requires a large amount of historical data as support, and the diagnostic interpretability of the model is weak.

[0005] 3. Fault diagnosis method based on physical model: establish a mathematical model to describe the operation mechanism of the equipment, and identify faults by comparing measured data with model prediction results. This method has good interpretability, but often requires a lot of parameter testing and model identification, and the modeling process is complicated.

[0006] In general, the existing fault diagnosis methods have their own advantages and disadvantages, and it is difficult to meet the requirements of fault diagnosis accuracy, reliability and explainability in complex coal mine environments. Therefore, it is urgent to develop a new fault diagnosis method that can make full use of equipment operation data, comprehensively apply physical models and data-driven technology, and realize comprehensive diagnosis and prediction of coal mine equipment faults. Summary of the invention

[0007] In view of this, the present invention provides a coal mine equipment fault detection method, medium and system based on PHM, which can solve the technical problem that the existing technology is difficult to meet the requirements of fault diagnosis accuracy, reliability and explainability in complex coal mine environments.

[0008] The present invention is achieved in that:

[0009] A first aspect of the present invention provides a coal mine equipment fault detection method based on PHM, which comprises the following steps:

[0010] S10. Collect the operation data of coal mine equipment, including vibration, temperature, pressure, current, speed, noise, oil analysis data, power consumption, equipment operation time, and ambient humidity;

[0011] S20, preprocessing the collected operation data, including denoising, normalization and feature extraction;

[0012] S30, establishing a mathematical model of the coal mine equipment taking into account the operation data, including a group of dynamic equations, a group of electrical characteristic equations, a group of wear equations, and a group of fault experience equations;

[0013] S40, using the preprocessed data to perform parameter identification and optimization on the mathematical model to obtain an optimized mathematical model;

[0014] S50, discretizing the optimized mathematical model to obtain a multi-dimensional state space, wherein each dimension represents a key parameter, a normal operating state is defined as a starting point, and different types of fault states are defined as end points;

[0015] S60, construct a fault path network, and regard each possible fault evolution process as a path from the starting point to the end point, and each node on the path represents an intermediate state of the device;

[0016] S70, using the A* algorithm to search and optimize each fault path, and calculate the shortest path from the normal state to various fault states, where the path length represents the difficulty of the fault; and calculating the fault weight vector according to the length of each fault path, where the shorter the path, the greater the fault weight;

[0017] S80, inputting the real-time collected equipment operation data into the optimized mathematical model, obtaining the position of the current equipment state in the state space, and calculating the distance from the position to each fault endpoint;

[0018] S90, combining the fault weight vector and the distance from the current state to each fault endpoint, calculating the final fault probability distribution, and outputting the fault diagnosis result and warning information.

[0019] The dynamic equations are used to describe the mechanical motion characteristics of the device, including the relationship between displacement, angle, and angular velocity, which are specifically expressed as follows:

[0020]

[0021] Where m is the mass of the equipment (kg); x is the displacement (m); t is the time (s); c is the damping coefficient (N·s / m); k is the elastic coefficient (N / m); F(t) is the external force (N); I is the moment of inertia (kg·m 2 ), θ is the angular displacement (rad), b is the angular damping coefficient (N·m·s / rad), τ s is the static friction torque (N·m); T(t) is the external torque (N·m); ω is the angular velocity (rad / s); α is the acceleration coefficient (rad / s 2 );β is the velocity attenuation coefficient (1 / rad);∈1,∈2,∈3 are error terms. Parameter acquisition method: m, I are obtained through equipment specification parameters; c, k, b, τ s Obtained through vibration test experiments, the steps include: 1) installing acceleration sensors at different positions of the equipment; 2) applying known excitation forces to the equipment; 3) recording the vibration response of the equipment; 4) fitting the vibration response curve using the least squares method to obtain parameter values. F(t) and T(t) are measured in real time by force sensors and torque sensors. α and β are obtained through angular velocity measurement experiments, the steps include: 1) using a high-precision encoder to measure the angular velocity of the equipment; 2) recording the change of angular velocity over time under different working conditions; 3) fitting the angular velocity curve using a nonlinear regression method to obtain parameter values.

[0022] Furthermore, the electrical characteristic equation group is used to describe the electrical performance of the device, including the relationship between current, voltage, power, and electromagnetic torque, which is specifically expressed as follows:

[0023]

[0024] P = VIcosφ+∈5;

[0025]

[0026] Where, L is inductance (H); i is current (A); R is resistance (Ω); C is capacitance (F); V(t) is voltage (V); P is power (W); I is effective current (A); φ is power factor angle (rad); T e is the electromagnetic torque (N·m); τ is the electromagnetic time constant (s); K t is the torque constant (N·m / A); ∈4, ∈5, ∈6 are error terms. Parameter acquisition method: L, R, C are measured by LCR tester. V(t) and i are measured in real time by voltmeter and ammeter. φ is measured by power analyzer. τ and K t The parameters are obtained through motor parameter test experiments, including: 1) applying a step current input to the motor; 2) recording the response curve of the electromagnetic torque; 3) fitting the response curve using a system identification method to obtain the parameter value.

[0027] Furthermore, the wear equations are used to describe the wear process of key components of the equipment, including the relationship between wear amount, material hardness, and remaining life, which is specifically expressed as follows:

[0028]

[0029] Where, W is the wear amount (mm); K w is the wear coefficient; F is the friction force (N); N is the number of contacts; H is the material hardness (HV); H0 is the initial hardness (HV); K h is the hardness attenuation coefficient; τ h is the hardness decay time constant (s); L is the remaining life (h); L0 is the initial expected life (h); W c is the critical wear amount (mm); ∈7, ∈8, ∈9 are error terms. Parameter acquisition method: K w The K is obtained through wear test, which includes: 1) simulating the working conditions of the equipment on the test bench; 2) measuring the wear amount at different time points; 3) fitting the wear curve using linear regression method to obtain K w Value. F is measured by force sensor. N is recorded by counter. H0, K h , τ h The hardness test is carried out through the following steps: 1) measuring the surface hardness of the component regularly; 2) recording the change of hardness over time; 3) fitting the hardness curve using nonlinear regression method to obtain the parameter value. L0 and W c Determined based on equipment historical data and experience values.

[0030] Furthermore, the fault experience equation group is used to describe the evolution law of common equipment faults, including the relationship between fault probability, mean failure time, and fault characteristic signals, which is specifically expressed as follows:

[0031]

[0032] Where P f (t) is the failure probability; t is the operating time (h); η is the scale parameter; β is the shape parameter; MTBF is the mean time between failures (h); λ1 is the basic failure rate (1 / h); λ2 is the degradation coefficient; γ is the degradation rate; S(t) is the fault characteristic signal strength; S0 is the initial signal strength; K s is the signal growth coefficient; T(τ) is the temperature function; f(T) is the temperature influence function; ∈ 10 ,∈ 11 ,∈ 12is the error term. Parameter acquisition method: η and β are obtained through statistical analysis of failure data, including: 1) collecting historical equipment failure data; 2) using the maximum likelihood estimation method to fit the Weibull distribution and obtain parameter values. λ1, λ2 and γ are obtained through reliability testing, including: 1) conducting accelerated life tests; 2) recording the number of failures at different time points; 3) using nonlinear regression methods to fit the MTBF curve and obtain parameter values. S0, K s and f(T) are obtained through a fault characteristic signal monitoring experiment, the steps of which include: 1) monitoring the fault characteristic signal under different temperature conditions; 2) recording the change of signal intensity with time and temperature; 3) using multiple regression analysis to determine the parameter value and the form of the temperature influence function.

[0033] Furthermore, the optimized mathematical model is discretized to convert the continuous-time mathematical model into a discrete state space representation that can be used in computer programs. The discretization process is as follows:

[0034] (1) Select an appropriate time step Δt, which is usually determined based on the system dynamic characteristics and calculation accuracy requirements.

[0035] (2) Discretize the continuous time equations, such as the displacement equation in the dynamic equations:

[0036]

[0037] Among them, x n represents the displacement at the nth time step, x′ n represents the velocity, which can be approximated by central difference:

[0038]

[0039] (3) Perform similar discretization processing on other equation groups.

[0040] (4) Define the state vector X = [x, x′, θ, θ′, ω, i, W, H, S] T , contains the key variables of the system.

[0041] (5) The discretized system of equations is expressed in state space form:

[0042] X n+1 =f(X n , U n )+w n ;

[0043] Where f is the nonlinear state transfer function, U n is the input vector, w n is the process noise.

[0044] Furthermore, in the step of searching and optimizing each fault path using the A* algorithm and calculating the shortest path from the normal state to various fault states, the process of constructing the fault path network is specifically described as follows:

[0045] (1) Define the normal operation state as the starting point X0 and define different types of fault states as the end points X f，i , where i represents different fault types.

[0046] (2) Discretize the state space into a grid, where each grid point represents a possible intermediate state.

[0047] (3) Define the state transition cost function g(X i , X j ), indicating that from state X i To X j The cost can be determined based on the state equation and physical constraints.

[0048] (4) Define the heuristic function h(X) to estimate the minimum cost from the current state X to the fault state, which can be designed based on the fault experience equation set.

[0049] (5) Use the A* algorithm to search from X0 to each X f，i The shortest path:

[0050] a. Initialize the open list and closed list;

[0051] b. Add the starting point X0 to the open list;

[0052] c. Loop until the target point is found or the open list is empty:

[0053] Select the node X in the open list where f(X)=g(X)+h(X) is the smallest. c ;

[0054] If X c If it is the target point, the path is found and the loop is exited;

[0055] X c Move from the open list to the closed list;

[0056] X c Each adjacent node X n :

[0057] If X n Not in the closed list, calculate f(X n ) and added to the open list;

[0058] If X n Already in the open list, update its f value if the new path is better;

[0059] (6) Repeat step (5) for each fault type to obtain all fault paths.

[0060] (7) Calculate the fault weight vector:

[0061] Among them, L i is the length of the i-th fault path, w i is the corresponding weight.

[0062] Furthermore, the operation data of the coal mine equipment is obtained from the PHM system of the coal mine, or obtained using sensors or instruments.

[0063] Furthermore, the step of outputting the fault diagnosis result and warning information also includes outputting the diagnosis result and warning information to the PHM system in a Json format.

[0064] A second aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores program instructions, and when the program instructions are run in a computer, they are used to execute the above-mentioned coal mine equipment fault detection method based on PHM.

[0065] A third aspect of the present invention provides a PHM-based coal mine equipment fault detection system, which includes the above-mentioned computer-readable storage medium.

[0066] Compared with the prior art, the beneficial effects of the PHM-based coal mine equipment fault detection method, medium and system provided by the present invention are as follows: 1. Comprehensive use of equipment operation data: This method collects rich operation data such as vibration, temperature, pressure, current, speed, noise, oil analysis, etc. of the equipment, providing a solid foundation for subsequent analysis and modeling. Compared with a single fault feature, multi-dimensional operation data can better reflect the actual working status of the equipment.

[0067] 2. Establish a physical mechanism model: This method constructs a comprehensive mathematical model including dynamic equations, electrical characteristic equations, wear equations, and fault experience equations, which can accurately describe the operation law of the equipment and the mechanism of fault evolution. Compared with pure data-driven methods, physical models have stronger interpretability and predictive capabilities.

[0068] 3. Optimize model parameters and fault paths: This method uses parameter identification and model optimization techniques to make the mathematical model better fit the actual equipment characteristics. At the same time, the A* algorithm is used to search and optimize the fault path network to find the shortest path from the normal state to various fault states, providing a quantitative basis for fault diagnosis. These methods greatly improve the accuracy and reliability of the overall diagnosis.

[0069] 4. Real-time monitoring of fault risks: This method inputs the real-time collected equipment operation data into the optimized mathematical model, calculates the distance from the current state to each fault endpoint in a timely manner, and combines the fault path weight to give the fault probability distribution. This can not only detect the current fault in a timely manner, but also predict the trend of future faults, providing effective support for equipment maintenance.

[0070] In summary, the fault detection method of the present invention fully integrates the advantages of physical modeling and data-driven, and realizes comprehensive diagnosis and prediction of coal mine equipment faults. Compared with the existing technology, this method has significantly improved the diagnostic accuracy, reliability and interpretability, and solves the technical problem that the existing technology is difficult to meet the requirements of fault diagnosis accuracy, reliability and interpretability in complex coal mine environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] Figure 1 A flow chart of the method provided by the present invention. DETAILED DESCRIPTION

[0072] In order to make the purpose, technical solution and advantages of the embodiments of the present invention more clear, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0073] like Figure 1 As shown, it is a flow chart of a coal mine equipment fault detection method based on PHM provided by the first aspect of the present invention. The method comprises the following steps: wherein, the following steps are included:

[0074] S10. Collect the operation data of coal mine equipment, including vibration, temperature, pressure, current, speed, noise, oil analysis data, power consumption, equipment operation time, and ambient humidity;

[0075] S20, preprocessing the collected operation data, including denoising, normalization and feature extraction;

[0076] S30, establishing a mathematical model of coal mine equipment taking into account operation data, including a group of dynamic equations, a group of electrical characteristic equations, a group of wear equations, and a group of fault experience equations;

[0077] S40, using the preprocessed data to perform parameter identification and optimization on the mathematical model to obtain an optimized mathematical model;

[0078] S50, discretizing the optimized mathematical model to obtain a multi-dimensional state space, wherein each dimension represents a key parameter, a normal operating state is defined as a starting point, and different types of fault states are defined as end points;

[0079] S60, construct a fault path network, and regard each possible fault evolution process as a path from the starting point to the end point, and each node on the path represents an intermediate state of the device;

[0080] S70, using the A* algorithm to search and optimize each fault path, and calculate the shortest path from the normal state to various fault states, where the path length represents the difficulty of the fault; and calculating the fault weight vector according to the length of each fault path, where the shorter the path, the greater the fault weight;

[0081] S80, inputting the real-time collected equipment operation data into the optimized mathematical model, obtaining the position of the current equipment state in the state space, and calculating the distance from the position to each fault endpoint;

[0082] S90, combining the fault weight vector and the distance from the current state to each fault endpoint, calculating the final fault probability distribution, and outputting the fault diagnosis result and warning information.

[0083] The specific implementation methods of the above steps are described in detail below:

[0084] The specific implementation method of step S10 is to collect the operation data of coal mine equipment. The purpose of this step is to use the PHM system or sensors to collect key parameters in the operation process of coal mine equipment to provide basic data for subsequent mathematical modeling and fault diagnosis. Specifically, this step includes the following sub-steps:

[0085] Install various sensors: Install vibration sensors, temperature sensors, pressure sensors, current sensors, speed sensors, noise sensors, etc. at key parts of coal mine equipment to collect equipment operation data in real time. At the same time, install environmental humidity sensors near the equipment. For important lubricating parts, oil analyzers need to be installed. These sensors should be industrial-grade equipment with high precision and resistance to high temperature and high humidity environments.

[0086] Establish a data acquisition system: Connect the above sensors to industrial control equipment to build a reliable data acquisition system. The system should have functions such as automatic acquisition, real-time storage, and remote monitoring to ensure the integrity and continuity of various operating data. The sampling frequency should be reasonably selected according to the dynamic characteristics of the equipment and should generally be greater than 100Hz.

[0087] Calibrate the sensor: Before the sensor is put into use, it is necessary to calibrate and calibrate each sensor to ensure the accuracy and reliability of the data. Calibration methods include using standard objects, comparative measurements, etc. The appropriate calibration method should be selected according to the sensor type. After the calibration is completed, calibration checks should be performed regularly to ensure data quality.

[0088] Real-time data collection: Once the data collection system is put into use, it will start to collect various operating parameter data of the equipment in real time, including vibration, temperature, pressure, current, speed, noise, oil analysis data, power consumption, equipment running time, ambient humidity, etc. These data will provide a basis for subsequent mathematical modeling and fault diagnosis.

[0089] Through the above sub-steps, the operation data of coal mine equipment can be collected comprehensively and accurately, laying the foundation for subsequent analysis and modeling. The key to this step lies in the selection of sensors, the reliability of the data acquisition system, and the continuity and accuracy of data acquisition.

[0090] Next is step S20, pre-processing the collected operation data:

[0091] The specific implementation of step S20 is to pre-process the collected operating data, including denoising, normalization and feature extraction. The purpose of this step is to eliminate noise interference in the data, so that the data has a better signal-to-noise ratio, and at the same time extract key features that are conducive to fault diagnosis. Specifically, it includes the following sub-steps:

[0092] Data denoising: Filter the collected raw data to remove high-frequency noise and interference generated during the measurement process. Noise suppression can be performed using methods such as Butterworth filter and wavelet transform. The selection of filtering parameters should be optimized according to the spectral characteristics of the signal to ensure that excessive signal distortion is not caused while denoising.

[0093] Data normalization: Unify data of different dimensions into the interval [0,1] to eliminate the influence of dimensions. The minimum-maximum normalization formula can be used: where x n is the normalized data, x is the original data, and x min and x max are the minimum and maximum values ​​of the original data respectively. This method can effectively eliminate the dimensional differences between different parameters.

[0094] Feature extraction: Extract key features that are beneficial to fault diagnosis from preprocessed data. Common feature extraction methods include time domain features (mean, variance, peak factor, etc.), frequency domain features (power spectrum density, peak frequency, etc.), time-frequency features (wavelet coefficients, Hilbert-Huang transform, etc.). Through comparative analysis, select features that are highly correlated with different fault modes.

[0095] Through the above data preprocessing steps, a set of denoised, normalized and feature-extracted data can be obtained, providing better input for subsequent mathematical modeling and fault diagnosis. The key to this step is to select appropriate filtering methods, normalization algorithms and feature extraction indicators to fully explore the fault diagnosis information contained in the data.

[0096] Next is step S30, establishing a mathematical model of coal mining equipment:

[0097] The specific implementation method of step S30 is to establish a mathematical model that takes into account the equipment operation data, including a group of dynamic equations, a group of electrical characteristic equations, a group of wear equations, and a group of fault experience equations. The purpose of this step is to construct a mathematical description that can fully describe the equipment operation characteristics and provide a theoretical basis for subsequent fault diagnosis and prediction. It specifically includes the following sub-steps:

[0098] Establish a set of dynamic equations: The dynamic equations are used to describe the mechanical motion characteristics of the device, including the relationship between key parameters such as displacement, angle, and angular velocity. There are two main equations: one is the mass-spring-damper equation, which describes the linear motion of the device; the other is the moment of inertia-torque equation, which describes the rotational motion of the device. The parameters involved in these equations, such as mass, damping coefficient, elastic coefficient, etc., can be obtained through equipment specifications and experimental tests.

[0099] Establish the electrical characteristic equation group: The electrical characteristic equation group is used to describe the electrical performance of the equipment, including the relationship between parameters such as current, voltage, power, and electromagnetic torque. It mainly includes three equations: one is the inductance-resistance-capacitance equation, which describes the dynamic characteristics of the circuit; the second is the active power equation, which describes the input power; and the third is the electromagnetic torque equation, which describes the torque output of the motor. The parameters involved in these equations, such as inductance, resistance, torque constant, etc., can be obtained through test instruments and parameter identification experiments.

[0100] Establish the wear equation group: The wear equation group is used to describe the wear process of key components of the equipment, including the relationship between parameters such as wear volume, material hardness, and remaining life. It mainly includes three equations: one is the wear rate equation, which describes the change of wear volume over time; the second is the material hardness equation, which describes the attenuation of hardness over time; and the third is the remaining life equation, which describes the remaining use time. The parameters involved in these equations, such as wear coefficient, hardness attenuation coefficient, etc., can be obtained through wear tests and hardness tests.

[0101] Establishment of fault experience equations: Fault experience equations are used to describe the evolution of common equipment failures, including the relationship between parameters such as failure probability, mean failure time, and fault characteristic signals. It mainly includes three equations: one is the Weibull distribution function, which describes the change of failure probability over time; the second is the nonlinear MTBF equation, which describes the change of mean failure time over running time; the third is the fault characteristic signal equation, which describes the change of signal strength over time and temperature. The parameters involved in these equations, such as scale parameters, basic failure rate, etc., can be obtained through reliability tests and signal monitoring experiments.

[0102] By establishing the above four types of equations, a comprehensive and detailed mathematical model is constructed, which can accurately describe the operating characteristics of the equipment, the laws of fault evolution, etc.

[0103] Next is step S40, which is to identify and optimize the parameters of the mathematical model:

[0104] The specific implementation method of step S40 is to use the preprocessed data to perform parameter identification and optimization on the mathematical model to obtain an optimized mathematical model. The purpose of this step is to enable the mathematical model to better fit the operating characteristics of the actual equipment and provide a more accurate basis for subsequent fault diagnosis and prediction. It specifically includes the following sub-steps:

[0105] Model parameter identification: Identify the unknown parameters in each equation of the mathematical model established in step S30. Parameter estimation algorithms such as the least squares method and the maximum likelihood estimation method can be used to identify and optimize the model parameters using the preprocessed operating data. Through parameter identification, parameter values ​​that are closer to the actual equipment characteristics can be obtained.

[0106] Model optimization: Based on parameter identification, the overall performance of the mathematical model can be further optimized. Intelligent optimization algorithms such as genetic algorithms and particle swarm optimization can be used to minimize the error between the model prediction value and the measured value to obtain a more accurate model expression.

[0107] Model validation: Apply the optimized mathematical model to new operating data for validation to check the model's prediction accuracy and robustness. If the model performance does not meet the requirements, the model structure can be appropriately adjusted or the parameter identification and optimization steps can be repeated until the model reaches the ideal state.

[0108] Through parameter identification and model optimization, a mathematical model that is closer to the actual equipment characteristics can be obtained. Such a model can not only more accurately describe the operating laws of the equipment, but also provide a more reliable theoretical basis for subsequent fault diagnosis and prediction. The key to this step is to select appropriate parameter identification and optimization algorithms, as well as in-depth analysis of the model verification results.

[0109] Next is step S50, discretizing the optimized mathematical model:

[0110] The specific implementation of step S50 is to discretize the optimized mathematical model to obtain a multi-dimensional state space. The purpose of this step is to convert the continuous-time mathematical model into a discrete state space representation that can be used in computer programs, laying the foundation for subsequent fault path search. It specifically includes the following sub-steps:

[0111] Time discretization: First, select an appropriate time step Δt, which is usually determined based on the system dynamic characteristics and calculation accuracy requirements. Then, perform differential discretization on the continuous time differential equation, such as the displacement equation in the dynamic equations:

[0112]

[0113] where x n represents the displacement at the nth time step, x′ n represents the velocity, which can be approximated by central difference. Similar discretization is performed on other equations.

[0114] State vector definition: Integrate the discretized equations into state space form and define the state vector X = [x, x′, θ, θ′, ω, i, W, H, S] T , which contains the key variables of the system. This converts the continuous-time mathematical model into a discrete state space representation.

[0115] Establishment of state transfer equation: The discretized equation group is organized into a state transfer equation in the state space form:

[0116] X n+1 =f(X n , U n )+W n ;

[0117] Where f is the nonlinear state transfer function, U n is the input vector, w n is the process noise. This results in a multidimensional state space model that can be used in a computer program.

[0118] Through the above discretization process, the original continuous-time mathematical model is transformed into an operational discrete state space representation. This provides a basis for subsequent state-space-based fault path search and fault diagnosis. The key to this step is to choose an appropriate time step and whether the definition of the state vector can fully describe the key characteristics of the system.

[0119] Next is step S60, constructing a fault path network:

[0120] The specific implementation of step S60 is to construct a fault path network, treating each possible fault evolution process as a path from a normal state to a fault state. The purpose of this step is to establish an intuitive fault evolution model to provide a visual reference for subsequent fault diagnosis and prediction. It specifically includes the following sub-steps:

[0121] Define the state space: First, discretize the multidimensional state space established in step S50 into a grid, where each grid point represents a possible intermediate state. Define the normal operation state as the starting point X0, and define different types of fault states as the end points X1. f，i , where i represents different fault types.

[0122] Determine the state transfer cost: Define the state transfer cost function g(X i , X j ), indicating that from state X i To X j This cost can be determined based on the mathematical model and physical constraints established in step S30, reflecting the difficulty of transition between states.

[0123] Design heuristic function: define a heuristic function h(X) to estimate the minimum cost from the current state X to the fault state. This function can be designed based on the fault experience equation group in step S30 and estimated using parameters such as failure probability and remaining life.

[0124] Search for the shortest fault path: Use the A* algorithm to search and optimize each fault path and calculate the shortest path from the normal state to various fault states. Path length L i It represents the difficulty of a fault occurring. The shorter the path, the easier it is for a fault to occur.

[0125] Calculate the fault weight: According to the length L of each fault path i , calculate the fault weight vector The shorter the path, the greater the fault weight, indicating that the fault is more likely to occur.

[0126] Through the above steps, a visual fault path network model is established. In this network, each path represents a possible fault evolution process, the path length reflects the difficulty of the fault, and the weight represents the relative importance of each fault mode. Such a model provides an intuitive reference for subsequent fault diagnosis and prediction. The key lies in the reasonable definition of state transition cost and heuristic function, as well as the effective application of A* algorithm.

[0127] Next is step S70, using the A* algorithm to optimize the fault path:

[0128] The specific implementation of step S70 is to use the A* algorithm to search and optimize each fault path and calculate the shortest path from the normal state to various fault states. The purpose of this step is to find the optimal evolution process of the fault occurrence and provide a basis for subsequent fault diagnosis. It specifically includes the following sub-steps:

[0129] Initialize the open list and closed list: the open list stores the nodes to be searched, and the closed list stores the nodes that have been searched.

[0130] Add the starting point X0 to the open list: Starting point X0 represents the normal operating state of the device.

[0131] Loop until the target point is found or the open list is empty:

[0132] Select the node X from the open list where f(X)=g(X)+h(X) is the smallest. c As the current node. Here g(X) is the actual cost from the starting point to the current node, and h(X) is the remaining cost estimated by the heuristic function.

[0133] If X c is the target fault state X f，i , the shortest path is found and the loop exits.

[0134] X c Move from the open list to the closed list.

[0135] X c Each adjacent node X n :

[0136] If X n Not in the closed list, calculate f(X n ) and added to the open list.

[0137] If X n Already in the open list, update its f value if the new path is better.

[0138] Repeat step 3 for each fault type to obtain the shortest length L of all fault paths i .

[0139] Compute the fault weight vector: The shorter the path, the greater the fault weight, indicating that the fault is more likely to occur.

[0140] The A* algorithm is an optimal path planning algorithm based on heuristic search. It selects the next search node by evaluating the total cost from the current node to the target node. This algorithm can effectively find the shortest path from the normal state to various fault states, providing a reliable theoretical basis for subsequent fault diagnosis. The key to this step is to reasonably define the state transition cost function g(X) and the heuristic function h(X) to improve the efficiency and accuracy of the search.

[0141] Next is step S80, inputting the real-time data into the optimized mathematical model:

[0142] The specific implementation of step S80 is to input the real-time collected equipment operation data into the optimized mathematical model, obtain the position of the current equipment state in the state space, and calculate the distance from the position to each fault endpoint. The purpose of this step is to use the optimized mathematical model to diagnose and analyze the real-time operation status of the equipment, and provide a basis for subsequent fault warning. It specifically includes the following sub-steps:

[0143] Real-time data input: The real-time operation data of the equipment collected in step S10, such as vibration, temperature, current, etc., are input into the mathematical model optimized in step S40.

[0144] State space positioning: Using a mathematical model, calculate the position X of the current device state in the multidimensional state space established in step S50 c . This position reflects the current operating status of the device.

[0145] Calculate the distance to the fault state: Calculate the current state X c To various fault endpoints X f，i The distance d i =||X c -X f，i ||. This distance reflects the proximity of the current state to various fault states.

[0146] Through this step, the operating status of the equipment can be tracked in real time and its proximity to various fault states can be evaluated. This provides key information for subsequent fault diagnosis and early warning. The key to this step lies in the accuracy of the mathematical model and the reasonable design of the state space to ensure that the current state can be accurately located and the distance to the fault state can be calculated.

[0147] Finally, step S90 calculates the fault probability distribution and outputs the diagnosis result:

[0148] The specific implementation of step S90 is to combine the fault weight vector calculated in step S70 and the distance from the current state to each fault endpoint calculated in step S80, calculate the final fault probability distribution, and output the fault diagnosis result and warning information. The purpose of this step is to comprehensively utilize the mathematical model and fault path network established previously to comprehensively diagnose and predict the current fault state of the equipment and provide decision support for maintenance. It specifically includes the following sub-steps:

[0149] Calculate the failure probability distribution: Based on the current state X c To each fault endpoint X f，i The distance d i , and the fault weight w calculated in step S70 i , calculate the probability of occurrence of each failure mode:

[0150]

[0151] Where a is a parameter that characterizes the width of the probability distribution. In this way, the probability distribution of each fault mode at the current moment is obtained.

[0152] Output diagnostic results: Output the fault diagnostic results of the current device based on the calculated fault probability distribution. Including:

[0153] The most likely failure mode and its probability of occurrence

[0154] Probability of other failure modes

[0155] The current overall health status index of the device

[0156] Provide fault warning: If the probability of a certain fault mode exceeds the preset threshold, a fault warning message will be issued in time to remind maintenance personnel to conduct inspection and maintenance. The threshold can be set based on actual experience, and it is generally recommended to be set between 0.2 and 0.3.

[0157] Through this step, the mathematical model and fault path network established earlier are comprehensively utilized to comprehensively diagnose and predict the current fault status of the equipment, give a specific fault probability distribution, and provide decision support for maintenance. The key to this step is to reasonably design the fault probability calculation formula and reasonably set the fault warning threshold.

[0158] For step S90, the method of outputting diagnostic results and warning information to the PHM system is as follows: convert the calculated fault probability distribution into a standard format (such as JSON or XML), define multi-level warning thresholds, and store the results in a relational database. Use message queues to implement real-time push, and develop a web dashboard for visual display. At the same time, integrate multiple alarm notification methods, generate maintenance suggestions and operation instructions, provide historical query functions, and integrate with the enterprise asset management system. Finally, establish a feedback mechanism to record the comparison between actual fault conditions and predicted results, and support continuous optimization of the model. This method can effectively realize the closed loop of equipment health management and improve the accuracy and timeliness of fault detection.

[0159] In general, this PHM-based coal mine equipment fault detection method makes full use of equipment operation data and realizes comprehensive diagnosis and prediction of equipment faults through a series of steps such as mathematical modeling, fault path search, and real-time status monitoring.

[0160] In order to better understand and implement the present invention, an embodiment of the method provided by the first aspect of the present invention is provided below, and each step in this embodiment is described in detail as follows:

[0161] In step S10, it is necessary to collect the operation data of coal mine equipment. Specifically including:

[0162] 1) Install various sensors: Install vibration sensors, temperature sensors, pressure sensors, etc. at key parts of the equipment to collect the operating parameters of the equipment. Vibration sensors can measure the displacement x(t) and angle θ(t) of the equipment, temperature sensors can measure the temperature T(t) of the equipment, and pressure sensors can measure the pressure p(t) of the equipment. At the same time, current sensors, speed sensors, noise sensors, etc. need to be installed to collect electrical characteristics and environmental parameters. In addition, oil analyzers need to be installed to measure the status of the equipment's lubricating oil, etc. The installation of these sensors should meet the requirements of industrial-grade environments, such as resistance to high temperature and high humidity.

[0163] 2) Establish a data acquisition system: Connect the above sensors to industrial control equipment to build a reliable data acquisition system. The system should have functions such as automatic acquisition, real-time storage, and remote monitoring to ensure the integrity and continuity of various operating data. The sampling frequency should be reasonably selected according to the dynamic characteristics of the equipment, and should generally be greater than 100Hz to meet the needs of subsequent analysis.

[0164] 3) Calibrate the sensor: Before the sensor is put into use, it is necessary to calibrate and calibrate each sensor to ensure the accuracy and reliability of the data. Standard objects can be used for comparison calibration, or input signals with known characteristics can be used for response calibration. After the calibration is completed, calibration checks should be performed regularly to ensure data quality.

[0165] 4) Real-time data collection: Once the data collection system is put into use, it will start to collect various operating parameter data of the equipment in real time, including vibration displacement x(t), angle θ(t), temperature T(t), pressure p(t), current i(t), speed ω(t), noise n(t), oil analysis data, power consumption P(t), equipment operation time t, ambient humidity H(t), etc. These data will provide a basis for subsequent mathematical modeling and fault diagnosis.

[0166] In step S20, the collected operation data needs to be preprocessed. Specifically, it includes:

[0167] 1) Data denoising: Filter the raw data to remove high-frequency noise interference during the measurement process. The Butterworth filter can be used, and its transfer function is:

[0168]

[0169] Where s is the complex frequency, ω c is the cutoff frequency, and n is the filter order. By properly choosing ω c and n, can effectively suppress high-frequency noise while retaining the valid signal.

[0170] 2) Data normalization: Unify data of different dimensions into the interval [0,1] to eliminate the influence of dimensions. The minimum-maximum normalization formula can be used:

[0171]

[0172] Among them, x n is the normalized data, x is the original data, and x min and x max are the minimum and maximum values ​​of the original data respectively. This method can effectively eliminate the dimensional differences between different parameters.

[0173] 3) Feature extraction: Extract key features that are beneficial to fault diagnosis from preprocessed data. Common feature extraction methods include:

[0174] Time domain characteristics:

[0175] Frequency domain characteristics: f p = argmax f P xx (f);

[0176] Time-frequency characteristics:

[0177] Among them, μ is the mean, σ 2 is the variance, γ is the peak factor, P xx (f) is the power spectrum density, f p is the peak frequency, and W(t,f) is the wavelet coefficient. By analyzing these features, the pattern features associated with different fault modes can be identified.

[0178] Through the above data preprocessing steps, a set of denoised, normalized and feature-extracted data is obtained, providing high-quality input for subsequent mathematical modeling and fault diagnosis.

[0179] In step S30, a mathematical model considering the equipment operation data needs to be established. Specifically, it includes:

[0180] 1) Dynamic equations: Dynamic equations are used to describe the mechanical motion characteristics of the device. For linear displacement motion, the mass-spring-damper equation can be established:

[0181]

[0182] Where m is the mass of the equipment (kg), x is the displacement (m), t is the time (s), c is the damping coefficient (N·s / m), k is the elastic coefficient (N / m), F(t) is the external force (N), and ∈1 is the error term.

[0183] For rotational motion, the moment of inertia-torque equation can be established:

[0184]

[0185] Where I is the moment of inertia (kg·m 2 ),θ is the angular displacement (rad), b is the angular damping coefficient (N·m·s / rad), τ s is the static friction torque (N·m), T(t) is the external torque (N·m), and ∈2 is the error term.

[0186] In addition, the angular velocity equation can be established:

[0187]

[0188] Where ω is the angular velocity (rad / s), α is the acceleration coefficient (rad / s 2 ), β is the velocity attenuation coefficient (1 / rad), ∈3 is the error term.

[0189] 2) Electrical characteristic equations: The electrical characteristic equations are used to describe the electrical performance of the device. The inductance-resistance-capacitance equation can be established:

[0190]

[0191] Where L is the inductance (H), i is the current (A), R is the resistance (Ω), C is the capacitance (F), V(t) is the voltage (V), and ∈4 is the error term.

[0192] At the same time, the active power equation can also be established:

[0193] P = VIcosφ+∈5;

[0194] Where P is power (W), I is the effective current (A), φ is the power factor angle (rad), and ∈5 is the error term.

[0195] And the electromagnetic torque equation:

[0196]

[0197] Among them, T e is the electromagnetic torque (N·m), τ is the electromagnetic time constant (s), K t is the torque constant (N·m / A), ∈6 is the error term.

[0198] 3) Wear equations: Wear equations are used to describe the wear process of key components of equipment. The wear rate equation can be established:

[0199]

[0200] Where W is the wear amount (mm), K is w is the wear coefficient, F is the friction force (N), N is the number of contacts, H is the material hardness (HV), and ∈7 is the error term.

[0201] At the same time, the hardness attenuation equation can also be established:

[0202]

[0203] Where H0 is the initial hardness (HV), K h is the hardness attenuation coefficient, τ h is the hardness decay time constant (s), ∈8 is the error term.

[0204] And the remaining life equation:

[0205]

[0206] Where L is the remaining life (h), L0 is the initial expected life (h), W c is the critical wear amount (mm), ∈9 is the error term.

[0207] 4) Fault experience equation group: Fault experience equation group is used to describe the evolution law of common equipment failures. The fault probability equation can be established:

[0208]

[0209] Among them, P f (t) is the failure probability, t is the running time (h), η is the scale parameter, β is the shape parameter, ∈ 10 is the error term.

[0210] At the same time, the mean time between failures (MTBF) equation can also be established:

[0211]

[0212] Where λ1 is the basic failure rate (1 / h), λ2 is the degradation coefficient, γ is the degradation rate, ∈ 11 is the error term.

[0213] And the fault characteristic signal equation:

[0214]

[0215] Among them, S(t) is the fault characteristic signal strength, S0 is the initial signal strength, K s is the signal growth coefficient, T(τ) is the temperature function, f(T) is the temperature influence function, ∈ 12 is the error term.

[0216] By establishing the above four types of equations, a comprehensive and detailed mathematical model is constructed, which can accurately describe the operating characteristics of the equipment, the laws of fault evolution, etc. These equations and their parameters will provide strong support for subsequent fault diagnosis and prediction.

[0217] In step S40, it is necessary to perform parameter identification and optimization on the mathematical model. Specifically, it includes:

[0218] 1) Model parameter identification: Identify the unknown parameters in each equation of the mathematical model established in step S30. The least squares method can be used to find the parameter solution that minimizes the sum of squares of the residuals between the predicted value and the measured value:

[0219]

[0220] Among them, p is the parameter vector to be identified, y i is the ith measured value, is the i-th predicted value. The model parameters are identified and optimized using the preprocessed operating data to obtain parameter values ​​that are closer to the actual equipment characteristics.

[0221] 2) Model optimization: Based on parameter identification, further optimize the overall performance of the mathematical model. Genetic algorithms can be used to iteratively optimize model parameters by simulating natural selection and genetic mechanisms:

[0222] p k+1 =p k +α k d k ;

[0223]

[0224] Among them, p k is the parameter vector of the kth iteration, α k and β k is the step size parameter, is the objective function gradient. By minimizing the error between the model prediction value and the measured value, a more accurate model expression is obtained.

[0225] 3) Model verification: Apply the optimized mathematical model to new operating data for verification to check the prediction accuracy and robustness of the model. If the model performance does not meet the requirements, the model structure can be appropriately adjusted or the parameter identification and optimization steps can be repeated until the model reaches the ideal state.

[0226] Through the above-mentioned parameter identification and model optimization, a mathematical model that is closer to the actual equipment characteristics can be obtained, providing a more reliable theoretical basis for subsequent fault diagnosis and prediction.

[0227] In step S50, the optimized mathematical model needs to be discretized. Specifically, it includes:

[0228] 1) Time discretization: Select an appropriate time step Δt, which is usually determined based on the dynamic characteristics of the system and the requirements for calculation accuracy. Then, the continuous time differential equation is discretized by difference, taking the displacement equation in the dynamic equation group as an example:

[0229]

[0230] Among them, x n represents the displacement at the nth time step, x′ n represents the velocity, which can be approximated by central difference:

[0231]

[0232] 2) State vector definition: Integrate the discretized equations into a state space form and define the state vector X = [x, x′, θ, θ′, ω, i, W, H, S] T , contains the key variables of the system.

[0233] 3) Establishment of state transfer equation: The discretized equation group is organized into a state transfer equation in the state space form:

[0234] X n+1 =f(X n ,U n )+w n ;

[0235] Where f is the nonlinear state transfer function, U n is the input vector, w n is the process noise.

[0236] Through the above discretization processing, the original continuous-time mathematical model is transformed into an operational discrete state space representation, which provides a basis for subsequent state-space-based fault path search and fault diagnosis.

[0237] In step S60, a fault path network needs to be constructed. Specifically, it includes:

[0238] 1) Define the state space: First, discretize the multidimensional state space established in step S50 into a grid, where each grid point represents a possible intermediate state. Define the normal operation state as the starting point X0, and define different types of fault states as the end points X1 and X2. f,i , where i represents different fault types.

[0239] 2) Determine the state transfer cost: Define the state transfer cost function g(X i ,X j), indicating that from state X i To X j This cost can be determined based on the mathematical model and physical constraints established in step S30, reflecting the ease of transition between states. For example:

[0240]

[0241] Among them, ∥·∥ is the L2 norm, and h(X) is the heuristic function.

[0242] 3) Design heuristic function: Define the heuristic function h(X) to estimate the minimum cost from the current state X to the fault state. It can be designed based on the fault experience equation group in step S30, and the parameters such as failure probability and remaining life are used for estimation:

[0243]

[0244] Among them, P f (t) is the failure probability, L(X, X f , i) is from state X to fault endpoint X f，i distance.

[0245] 4) Search for the shortest fault path: Use the A* algorithm to search and optimize each fault path and calculate the shortest path from the normal state to various fault states. Path length L i It represents the difficulty of a fault occurring. The shorter the path, the easier it is for a fault to occur.

[0246] 5) Calculate the fault weight: According to the length L of each fault path i , calculate the fault weight vector:

[0247]

[0248] The shorter the path, the greater the fault weight, indicating that the fault is more likely to occur.

[0249] Through the above steps, a visual fault path network model is established. In this network, each path represents a possible fault evolution process, the path length reflects the difficulty of the fault, and the weight represents the relative importance of each fault mode. Such a model provides an intuitive reference for subsequent fault diagnosis and prediction.

[0250] In step S70, the A* algorithm needs to be used to optimize the fault path. Specifically, it includes:

[0251] 1) Initialize the open list and closed list: the open list stores the nodes to be searched, and the closed list stores the nodes that have been searched.

[0252] 2) Add the starting point X0 to the open list: The starting point X0 represents the normal operating state of the device.

[0253] 3) Loop until the target point is found or the open list is empty:

[0254] a. Select the node X with the smallest f(X) = g(X) + h(X) from the open list. c As the current node. Here g(X) is the actual cost from the starting point to the current node, and h(X) is the remaining cost estimated by the heuristic function.

[0255] b. If X c is the target fault state X f，i , the shortest path is found and the loop exits.

[0256] c. X c Move from the open list to the closed list.

[0257] d. For X c Each adjacent node X n :

[0258] If X n Not in the closed list, calculate f(X n ) and added to the open list;

[0259] If X n Already in the open list, update its f value if the new path is better;

[0260] 4) Repeat step 3 for each fault type to obtain the shortest length L of all fault paths i .

[0261] 5) Calculate the fault weight vector:

[0262]

[0263] The shorter the path, the greater the fault weight, indicating that the fault is more likely to occur.

[0264] The A* algorithm is an optimal path planning algorithm based on heuristic search. It selects the next search node by evaluating the total cost from the current node to the target node. This algorithm can effectively find the shortest path from the normal state to various fault states, providing a reliable theoretical basis for subsequent fault diagnosis.

[0265] In step S80, the real-time data needs to be input into the optimized mathematical model. Specifically, it includes:

[0266] 1) Real-time data input: The real-time operation data of the equipment collected in step S10, such as vibration displacement x(t), angle θ(t), temperature T(t), current i(t), etc., are input into the mathematical model optimized in step S40.

[0267] 2) State space positioning: Using a mathematical model, calculate the position X of the current device state in the multidimensional state space established in step S50 c This position reflects the current operating status of the device and can be expressed as:

[0268] X c =f -1 (X n , U n );

[0269] Among them, f is the state transfer function, X n is the current state vector, U n is the current input vector.

[0270] 3) Calculate the distance to the fault state: Calculate the current state X c To various fault endpoints X f，i The distance d i , the L2 norm can be used:

[0271] d i =||X c -X f，i ||2;

[0272] This distance reflects the proximity of the current state to various fault states.

[0273] Through this step, the operating status of the equipment can be tracked in real time and its proximity to various fault states can be evaluated. This provides key information for subsequent fault diagnosis and early warning.

[0274] In step S90, it is necessary to calculate the fault probability distribution and output the diagnosis result. Specifically, it includes:

[0275] 1) Calculate the failure probability distribution: According to the current state X c To each fault endpoint X f，i The distance d i , and the fault weight w calculated in step S70 i , calculate the probability of occurrence of each failure mode:

[0276]

[0277] Among them, σ is a parameter that characterizes the width of the probability distribution and can be set according to the actual situation. In this way, the probability distribution of each fault mode at the current moment is obtained.

[0278] 2) Output diagnostic results: Output the fault diagnostic results of the current device based on the calculated fault probability distribution. Including:

[0279] a. The most likely failure mode and its probability of occurrence

[0280] b. Probability of other failure modes

[0281] c. The overall health status index of the device can be defined as h = 1-∑ i P f，i ;

[0282] 3) Provide fault warning: If the probability of occurrence of a certain fault mode P f，i Exceeding the preset threshold P t , timely issue fault warning information to remind maintenance personnel to carry out inspection and maintenance. Threshold P t It can be set according to actual experience, and it is generally recommended to be set between 0.2 and 0.3.

[0283] Through this step, the mathematical model and fault path network established previously are comprehensively utilized to conduct a comprehensive diagnosis and prediction of the current fault status of the equipment, give a specific fault probability distribution, and provide decision support for repair and maintenance.

[0284] The second aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores program instructions, and when the program instructions are executed in a computer, they are used to execute the above-mentioned coal mine equipment fault detection method based on PHM; the computer-readable storage medium can be a USB flash drive or a computer hard disk. The third aspect of the present invention provides a coal mine equipment fault detection system based on PHM, comprising the above-mentioned computer-readable storage medium; at the same time, the system also connects to the PHM platform through an API, supports obtaining data from the PHM platform or providing feedback; the system can also be used as a microservice source of PHM, used as a PHM for detecting equipment faults.

[0285] Specifically, the principle of the present invention is: adopting the PHM technical framework, comprehensively applying multiple advanced technical means such as physical model analysis, data-driven modeling and real-time monitoring. The specific principle is as follows:

[0286] 1. Physical model analysis: This method first establishes a set of mathematical models that describe the operating mechanism of the equipment, including dynamic equations, electrical characteristic equations, wear equations, and fault experience equations. These physical models can accurately reflect the working characteristics of each subsystem of the equipment and provide a theoretical basis for subsequent state assessment and fault diagnosis. Compared with pure data-driven methods, physical models have stronger interpretability and generalization capabilities.

[0287] 2. Data-driven modeling: Based on the physical model, this method uses measured equipment operation data and adopts parameter identification and model optimization technology to make the mathematical model better fit the actual equipment characteristics. This method that combines physical modeling and data-driven fully utilizes the advantages of both technologies, while retaining interpretability, it can also improve the overall model accuracy.

[0288] 3. Fault path search: Based on the optimized mathematical model, this method defines the normal operating state and various fault states of the equipment as nodes in the state space, and uses the A* algorithm to search for the shortest path from the normal state to each fault state. The path length reflects the difficulty of the fault and provides a quantitative basis for diagnosis. This fault path analysis method based on state space can intuitively display the mechanism of fault evolution and provide a theoretical basis for fault prediction.

[0289] 4. Real-time monitoring and early warning: This method inputs the real-time collected equipment operation data into the optimized mathematical model, calculates the position of the current state in the state space in real time, and evaluates its distance to each fault endpoint. Combined with the fault path weight obtained in step 3, the probability distribution of each current fault mode can be calculated. Once the probability of a certain fault exceeds the preset threshold, the system immediately issues an early warning message to provide decision support for maintenance. This real-time monitoring and early warning mechanism can effectively prevent the occurrence of equipment failures.

[0290] In summary, the fault detection method of the present invention fully integrates the advantages of physical modeling and data-driven, and adopts advanced technical means in key links such as establishing a mathematical model to describe the equipment mechanism, optimizing model parameters, searching for fault evolution paths, and real-time monitoring of fault risks. This comprehensive and systematic PHM technical framework enables the method to overcome the limitations of existing technologies and truly achieve accurate diagnosis and effective prediction of coal mine equipment faults.

[0291] In order to better understand and implement the present invention, an embodiment of a specific application scenario of the present invention is provided below: a coal mining enterprise frequently encounters coal mining equipment failure problems during the production process, resulting in production interruptions and seriously affecting production efficiency. In order to solve this problem, the enterprise decided to adopt the coal mining equipment failure detection method based on PHM proposed in the present invention to comprehensively monitor and diagnose key equipment.

[0292] 1. Data Collection

[0293] First, the company installed various sensors on the main coal mining equipment, as shown in Table 1:

[0294] Table 1 Sensor installation

[0295] Sensor Type Installation location Measurement parameters Vibration Sensor Motor bearings, gearboxes Displacement x(t), angle θ(t) Temperature Sensor Motor windings, gearboxes Temperature T(t) Pressure Sensors Hydraulic oil circuit Pressure p(t) Current Sensors Motor input Current i(t) Speed ​​sensor Motor shaft Angular velocity ω(t) Noise Sensor Around the equipment Noise n(t) Oil Analyzer Gearbox, hydraulic system Oil status

[0296] These sensors are all industrial-grade high-precision equipment that can withstand the harsh conditions of the coal mine environment. At the same time, the company also installed environmental humidity sensors near the equipment to monitor the working environment of the equipment.

[0297] In order to build a reliable data acquisition system, the company uses industrial control equipment to connect the above sensors, realizing the functions of automatic acquisition, real-time storage and remote monitoring. The sampling frequency of the system is set to 200Hz to ensure that the key dynamic characteristics of the equipment operation can be captured.

[0298] Before the equipment was officially put into use, the company calibrated and calibrated each sensor. Taking the vibration sensor as an example, it was first calibrated at different frequencies using a standard exciter to obtain the frequency response characteristics of the sensor. Then, the response curve of the sensor was recorded while the equipment was running, and the damping coefficient c and elastic coefficient k were obtained using the least squares method. Similar calibration methods were also applied to other sensors to ensure the accuracy and reliability of the data.

[0299] After the above steps, the company has established a complete data acquisition system that can comprehensively and accurately monitor the operating status of coal mine equipment.

[0300] 2. Mathematical modeling

[0301] Based on the collected equipment operation data, the company then established a mathematical model to describe the equipment mechanism, including a set of kinetic equations, a set of electrical characteristic equations, a set of wear equations, and a set of fault experience equations.

[0302] (1) Kinetic equations

[0303] For the linear displacement motion of the device, the mass-spring-damper equation can be established:

[0304]

[0305] Among them, the equipment mass m = 500kg, the damping coefficient c = 5×10 4 N·s / m, elastic modulus k=2×10 6 N / m, and the external force F(t) is measured in real time by the force sensor. By fitting the vibration response of the equipment, the company determined the above dynamic parameters.

[0306] For rotational motion, the moment of inertia-torque equation can be established:

[0307]

[0308] Where, the moment of inertia I = 50 kg·m 2 , angular damping coefficient b = 100 N·m·s / rad, static friction torque τs =500N·m, the external torque T(t) is measured in real time by the torque sensor. These parameter values ​​are also obtained by fitting the response curve.

[0309] In addition, the angular velocity equation can be established:

[0310]

[0311] Where, acceleration coefficient α = 10 rad / s 2 , velocity attenuation coefficient β = 0.05 / rad. These parameters are obtained through angular velocity test experiments.

[0312] (2) Electrical characteristics equations

[0313] For the circuit dynamic characteristics, the inductance-resistance-capacitance equation can be established:

[0314]

[0315] Among them, inductance L = 0.05H, resistance R = 1Ω, capacitance C = 0.01F, voltage V(t) is measured in real time by a voltmeter. These parameters are obtained by measuring with an LCR tester.

[0316] For active power, the power equation can be established:

[0317] P = VIcosφ+∈5;

[0318] Among them, the effective value current I = 50A, the power factor angle φ = 0.8rad, these parameters are measured by a power analyzer.

[0319] For electromagnetic torque, the torque equation can be established:

[0320]

[0321] Among them, the electromagnetic time constant τ = 0.05s, the torque constant K t =1N·m / A. These parameters are obtained through motor parameter test experiments.

[0322] (3) Wear equations

[0323] For the wear process, the wear rate equation can be established:

[0324]

[0325] Among them, the wear coefficient K w =1×10 -6 mm 2 / (N·cycle), friction force F = 500N, number of contacts N is recorded by a counter, and material hardness H = 300HV. These parameters are obtained through wear tests.

[0326] The hardness decay can be described by the following formula:

[0327]

[0328] Among them, the hardness attenuation coefficient K h =50HV, decay time constant τ h =10 4 s. These parameters are obtained through hardness testing experiments.

[0329] Finally, the remaining life can be expressed by the following equation:

[0330]

[0331] Where, initial life expectancy L0 = 10 4 h, critical wear amount W c =1mm. These parameters are determined based on historical equipment data and experience.

[0332] (4) Fault experience equations

[0333] The failure probability can be described by the Weibull distribution equation:

[0334]

[0335] Among them, the scale parameter η = 1000 1 / h, the shape parameter β = 1.5 1 / h. These parameters are obtained by statistical analysis of historical equipment failure data.

[0336] The mean time between failures (MTBF) can be described by the following nonlinear equation:

[0337]

[0338] Among them, the basic failure rate λ1=0.001 1 / h, the degradation coefficient λ2=0.0001 1 / h, and the degradation rate γ=0.00011 / h. These parameters are obtained through accelerated life tests.

[0339] The fault characteristic signal strength can be expressed as follows:

[0340]

[0341] Among them, the initial signal strength S0 = 50, the signal growth coefficient K s=0.1, temperature influence function f(T) = 1 + 0.01 T. These parameters are obtained through the fault characteristic signal monitoring experiment.

[0342] In summary, the company has established a comprehensive mathematical model that can better describe the dynamic characteristics, electrical characteristics, wear process and fault evolution of the equipment. This provides a theoretical basis for subsequent fault diagnosis.

[0343] 3. Model Optimization

[0344] In order to make the mathematical model better reflect the characteristics of the actual equipment, the company optimized it using parameter identification and model optimization technology.

[0345] First, the company uses the pre-processed operating data and the least squares method to identify the unknown parameters in each equation. Taking the displacement equation in the dynamic equation group as an example, the vibration response curve is fitted using the following formula to obtain the optimized m, c and k values:

[0346]

[0347] Similar parameter identification methods are also applied to other sets of equations to make the model parameters closer to reality.

[0348] Next, the company used a genetic algorithm to optimize the overall model. Taking minimizing the error between the predicted value and the measured value as the objective function, the parameters of each equation were iteratively optimized to obtain a more accurate mathematical description. After convergence, the optimized model was applied to new operating data for verification, and it was found that the prediction accuracy and robustness of the model were significantly improved.

[0349] After the above optimization steps, the company established a mathematical model that can more accurately describe the actual characteristics of the equipment, laying the foundation for subsequent fault diagnosis.

[0350] 4. Fault path analysis

[0351] Based on the optimized mathematical model, the company then constructed a fault path network to analyze the evolution of the equipment from normal state to various fault states.

[0352] First, the enterprise discretizes the multidimensional state space established in step 2 into a grid, and each grid point represents a possible intermediate state. The normal operation state of the equipment is defined as the starting point X0, and different types of fault states such as gear damage, bearing wear, and motor failure are defined as the end point X0. f,i .

[0353] Then, the enterprise defines the state transition cost function g(X i ,X j ), indicating that from state X i To Xj This cost function takes into account the physical constraints in the mathematical model, such as:

[0354]

[0355] Among them, ||X j -X i || is the distance between states, P f (t j ) is state X j The failure probability at time, L(X j ,X f,i ) is from state X j To the fault end point X f,i distance.

[0356] The company also designed a heuristic function h(X) to estimate the minimum cost from the current state to the end point of each type of failure:

[0357]

[0358] With the state transition cost and heuristic function, the enterprise uses the A* algorithm to search and optimize each fault path and calculate the shortest path from the normal state to various fault states. i It represents the difficulty of the fault to occur. The shorter the path, the easier it is for the corresponding fault to occur.

[0359] Finally, according to the length L of each fault path i , the enterprise calculated the fault weight vector:

[0360]

[0361] The shorter the path, the greater the fault weight, indicating that the fault mode deserves more attention.

[0362] Through the above steps, the enterprise established a visual fault path network model, which provided a theoretical basis for subsequent fault diagnosis.

[0363] 5. Real-time monitoring and early warning

[0364] With the optimized mathematical model and fault path analysis results, the company began to conduct real-time monitoring and fault warning of equipment.

[0365] First, the enterprise inputs the real-time operation data collected in step 1, such as vibration displacement x(t), angle θ(t), temperature T(t), current i(t), etc., into the optimized mathematical model. The model is used to calculate the position X of the current equipment state in the multidimensional state space. c :

[0366] Xc =f -1 (X n , U n );

[0367] Where f is the state transfer function, X n is the current state vector, U n is the current input vector.

[0368] Next, the enterprise calculates the current state X c To various fault endpoints X f，i The distance d i :

[0369] d i =||X c -X f，i ||2;

[0370] These distances reflect how close the current state is to various fault states.

[0371] Finally, the enterprise combines the fault weight vector w calculated in step 4 i , calculate the occurrence probability distribution of each failure mode:

[0372]

[0373] Among them, σ=0.1 is the width parameter of the probability distribution.

[0374] Based on the calculation results, the company outputs a fault diagnosis report for the current equipment, including:

[0375] The failure mode with the highest probability is gear breakage, with a probability of 28.5%.

[0376] The probabilities of other failure modes are: bearing wear 18.2%, motor failure 12.6%, and pipeline leakage 8.7%.

[0377] The current overall health status index of the device is 0.82, indicating that the overall working condition of the device is good.

[0378] However, if the probability of a certain failure mode exceeds 30%, the company will immediately issue a fault warning message to remind maintenance personnel to conduct inspections and maintenance. This real-time monitoring and early warning mechanism can help companies discover potential equipment hazards in a timely manner and effectively prevent failures.

[0379] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A coal mine equipment fault detection method based on PHM, characterized in that: The following steps are involved: S10. Collect the operation data of coal mine equipment, including vibration, temperature, pressure, current, speed, noise, oil analysis data, power consumption, equipment operation time, and ambient humidity; S20, preprocessing the collected operation data, including denoising, normalization and feature extraction; S30, establishing a mathematical model of the coal mine equipment taking into account the operation data, including a group of dynamic equations, a group of electrical characteristic equations, a group of wear equations, and a group of fault experience equations; S40, using the preprocessed data to perform parameter identification and optimization on the mathematical model to obtain an optimized mathematical model; S50, discretizing the optimized mathematical model to obtain a multi-dimensional state space, wherein each dimension represents a key parameter, a normal operating state is defined as a starting point, and different types of fault states are defined as end points; S60, construct a fault path network, and regard each possible fault evolution process as a path from the starting point to the end point, and each node on the path represents an intermediate state of the device; S70, using the A* algorithm to search and optimize each fault path, and calculate the shortest path from the normal state to various fault states, where the path length represents the difficulty of the fault; The fault weight vector is calculated based on the length of each fault path. The shorter the path, the greater the fault weight. S80, inputting the real-time collected equipment operation data into the optimized mathematical model, obtaining the position of the current equipment state in the state space, and calculating the distance from the position to each fault endpoint; S90, combining the fault weight vector and the distance from the current state to each fault endpoint, calculating the final fault probability distribution, and outputting the fault diagnosis result and warning information.

2. A coal mine equipment fault detection method based on PHM according to claim 1, characterized in that: The dynamic equations are used to describe the mechanical motion characteristics of the device, including the relationship between displacement, angle, and angular velocity.

3. A coal mine equipment fault detection method based on PHM according to claim 2, characterized in that: The electrical characteristic equation group is used to describe the electrical performance of the device, including the relationship between current, voltage, power and electromagnetic torque.

4. A coal mine equipment fault detection method based on PHM according to claim 3, characterized in that: The wear equations are used to describe the wear process of key components of equipment, including the relationship between wear amount, material hardness and remaining life.

5. A coal mine equipment fault detection method based on PHM according to claim 4, characterized in that: The fault experience equation group is used to describe the evolution law of common equipment faults, including the relationship between fault probability, mean failure time, and fault characteristic signals.

6. A coal mine equipment fault detection method based on PHM according to claim 5, characterized in that: Discretizing the optimized mathematical model specifically means converting the continuous-time mathematical model into a discrete state space representation that can be used in computer programs.

7. A coal mine equipment fault detection method based on PHM according to claim 6, characterized in that: The operation data of the coal mine equipment is obtained from the PHM system of the coal mine, or obtained using sensors or instruments.

8. A coal mine equipment fault detection method based on PHM according to claim 7, characterized in that: The step of outputting the fault diagnosis result and warning information also includes outputting the diagnosis result and warning information to the PHM system in a Json manner.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores program instructions, and when the program instructions are executed in a computer, they are used to execute the coal mine equipment fault detection method based on PHM according to any one of claims 1 to 8.

10. A coal mine equipment fault detection system based on PHM, characterized in that: A computer-readable storage medium comprising the computer-readable storage medium of claim 9.

Images