A coal mine underground scraper conveyor power part fault detection method and electronic equipment

Abstract

The application provides a coal mine underground scraper conveyor power part fault detection method and electronic equipment, and belongs to the field of coal mine underground transportation equipment fault detection. The method comprises collecting operation parameters of the scraper conveyor power part, the operation parameters comprising reducer input shaft temperature, reducer output shaft temperature, reducer cooling water temperature, reducer lubricating oil temperature, reducer cooling water pressure, motor bearing temperature and motor winding temperature; a state feature vector of the power part is constructed based on the operation parameters; the state feature vector is input into a pre-trained support vector machine (SVM) fault detection model to obtain an operation state classification result; when the operation state classification result indicates that the power part is in a fault state, fault early warning information is output. The method provided by the application can realize classification identification and fault early warning of the operation state of the scraper conveyor power part.

Description

Technical Field

[0001] This application relates to the field of fault detection technology for underground coal mine transportation equipment, and in particular to a fault detection method and electronic equipment for the power unit of an underground scraper conveyor in a coal mine. Background Technology

[0002] Scraper conveyors are key equipment in fully mechanized coal mining faces for continuous coal transport. Their power unit typically includes a reducer and an electric motor, providing continuous and stable power output to the scraper chain and conveying system. The operating status of the power unit directly affects the transmission efficiency, operational reliability, and the continuity and safety of underground coal mine production. Specifically, the input and output shafts, lubrication and cooling conditions of the reducer, and the temperature rise of the motor's bearings and windings are all important parameters reflecting the health of the power unit. Any abnormality in the power unit can easily lead to decreased conveying efficiency, accelerated equipment wear, and even sudden shutdowns, thereby affecting the normal production of the entire coal mining face.

[0003] In related technologies, fault detection in the power unit of underground scraper conveyors in coal mines typically relies on manual inspections, periodic maintenance, or reactive repairs after obvious equipment anomalies occur. While some detection methods can monitor single operating parameters such as temperature and pressure, most rely on single-parameter threshold judgments or on-site experience for fault identification. This makes it difficult to comprehensively analyze multiple key operating parameters, such as the reducer's input shaft temperature, output shaft temperature, cooling water temperature, lubricating oil temperature, cooling water pressure, and the motor's bearing and winding temperatures. Consequently, it fails to fully reflect the complex operating conditions and fault evolution process of the power unit. Especially in the early stages of equipment failure, changes in relevant parameters often exhibit nonlinear, coupled, and gradual characteristics. Existing technologies are prone to problems such as detection lag, untimely warnings, inaccurate fault location, and missed or false diagnoses, failing to meet the requirements for safe, continuous, and intelligent operation of underground coal mine equipment.

[0004] Therefore, in the fault detection of the power unit of scraper conveyors in underground coal mines, the difficulty in comprehensively characterizing the multi-parameter operating status, the strong reliance on human experience for fault identification, and the difficulty in timely warning of potential faults have become urgent problems to be solved. Summary of the Invention

[0005] This application provides a fault detection method and electronic equipment for the power unit of an underground scraper conveyor in coal mines, aiming to solve the problems of difficulty in comprehensively representing the multi-parameter operating status, strong reliance on human experience for fault identification, and difficulty in timely warning of potential faults in the existing fault detection of the power unit of underground scraper conveyors in coal mines.

[0006] In a first aspect, this application provides a method for detecting faults in the power unit of an underground scraper conveyor in a coal mine, the method comprising: The operating parameters of the scraper conveyor's power unit are collected, including the temperature of the reducer input shaft, the temperature of the reducer output shaft, the temperature of the reducer cooling water, the temperature of the reducer lubricating oil, the pressure of the reducer cooling water, the temperature of the motor bearing, and the temperature of the motor winding. Based on the operating parameters, construct the state feature vector of the power unit; The state feature vector is input into a pre-trained support vector machine (SVM) fault detection model to obtain the operating state classification result; wherein, the operating state classification result is used to indicate whether the power unit is in a normal state or a fault state; When the operation status classification result indicates that the power unit is in a fault state, a fault warning message is output.

[0007] In one possible design, constructing the state feature vector of the power unit based on the operating parameters includes: The temperature of the reducer input shaft, the temperature of the reducer output shaft, the temperature of the reducer cooling water, the temperature of the reducer lubricating oil, the pressure of the reducer cooling water, the temperature of the motor bearing, and the temperature of the motor winding are combined as feature components to obtain a multidimensional feature vector characterizing the operating state of the power unit. The multidimensional feature vector is standardized to obtain the state feature vector.

[0008] In one possible design, the standardization process for the multidimensional feature vector to obtain the state feature vector includes: For each feature component in the multidimensional feature vector, determine the sample mean and sample standard deviation corresponding to the feature component; Based on the sample mean and sample standard deviation corresponding to the feature component, the feature component is dimensionless to obtain the standardized feature value corresponding to the feature component. The state feature vector is obtained based on the standardized feature values ​​corresponding to each of the feature components.

[0009] In one possible design, before inputting the state feature vector into a pre-trained support vector machine (SVM) fault detection model to obtain the running state classification result, the method further includes: Multiple historical running samples are obtained, each of which includes a historical state feature vector and a sample label corresponding to the historical state feature vector; Based on multiple historical operating samples, the initial SVM fault detection model is trained to obtain the SVM fault detection model.

[0010] In one possible design, the sample labels include normal state labels and fault state labels; When all feature parameters corresponding to the historical state feature vector meet the preset safety threshold condition, the sample label corresponding to the historical state feature vector is the normal state label. When at least one feature parameter corresponding to the historical state feature vector does not meet the preset safety threshold condition, and / or when the scraper conveyor actually malfunctions, the sample label corresponding to the historical state feature vector is a fault state label.

[0011] In one possible design, the initial SVM fault detection model is trained based on multiple historical operating samples to obtain the SVM fault detection model, including: Based on multiple historical operation samples, a soft-interval SVM binary classification model is constructed to distinguish between normal and fault states. During the training of the soft-margin SVM binary classification model using multiple historical running samples, slack variables and penalty coefficients are introduced to provide fault tolerance for historical running samples that deviate from the classification boundary and / or misclassified historical running samples, so as to determine the classification boundary used to distinguish between normal and fault states. Based on the classification boundary, the SVM fault detection model is obtained.

[0012] In one possible design, the construction of a soft-margin SVM binary classification model to distinguish between normal and fault states based on multiple historical operating samples includes: Based on multiple historical running samples, a kernel function is determined to characterize the nonlinear relationship between the samples, wherein the kernel function is the radial basis function (RBF). Based on the kernel function, multiple historical running samples are mapped to a high-dimensional feature space; Based on multiple historical running samples mapped to a high-dimensional feature space, a soft-margin SVM binary classification model is constructed to distinguish between normal and fault states.

[0013] In one possible design, the method further includes: For the soft-margin SVM binary classification model, the parameters to be optimized are determined, including the penalty coefficient and the kernel function parameters; Based on the parameters to be optimized, multiple parameter combinations are constructed; Based on multiple parameter combinations, the soft-margin SVM binary classification model is trained and validated to determine the target parameter combination. The combination of target parameters is determined as the target model parameters of the soft-margin SVM binary classification model.

[0014] In one possible design, the step of training and validating the soft-margin SVM binary classification model based on multiple parameter combinations to determine the target parameter combination includes: The historical running samples are divided into multiple training samples and multiple validation samples; For each of the parameter combinations, the soft-margin SVM binary classification model is trained using multiple sets of training samples to obtain the trained soft-margin SVM binary classification model corresponding to the parameter combination. The trained soft-margin SVM binary classification model corresponding to the parameter combination is validated using multiple sets of validation samples to obtain the classification performance of the model corresponding to the parameter combination. The classification performance of the models corresponding to the multiple parameter combinations is compared, and the parameter combination with the best classification performance is determined as the target parameter combination.

[0015] Secondly, this application provides a terahertz-based conveyor belt core electronic device, comprising: a module for performing the aforementioned method embodiment of the first aspect.

[0016] Thirdly, this application provides an electronic device, including: a memory and at least one processor; The memory stores computer-executed instructions; The at least one processor executes computer execution instructions stored in the memory, causing the at least one processor to perform the method described in the first aspect or various possible designs of the first aspect.

[0017] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions, which, when executed, implement the method described in the first aspect or various possible designs of the first aspect.

[0018] Fifthly, this application provides a computer program product, which includes computer program code that, when run on a computer, causes the computer to implement the method described in the first aspect or various possible designs of the first aspect.

[0019] This application provides a method and electronic device for fault detection in the power unit of a scraper conveyor in an underground coal mine. The method collects operating parameters of the scraper conveyor's power unit, which simultaneously cover the temperature of the reducer input shaft, the reducer output shaft, the reducer cooling water temperature, the reducer lubricating oil temperature, the reducer cooling water pressure, the motor bearing temperature, and the motor winding temperature. These parameters reflect the reducer's transmission, cooling, and lubrication states, as well as the mechanical and electrical operating states of the motor. These previously scattered operating parameters are then unified into a state feature vector for the power unit, thus achieving a comprehensive representation of the overall operating state of the power unit. Based on this, the state feature vector is input into a pre-trained SVM fault detection model. This application enables the SVM fault detection model to output operational status classification results based on the combination relationship and state differences between multiple operating parameters. Compared with methods that rely on manual inspection, post-event maintenance, or single parameter threshold judgment, the method of this application can reduce the dependence on human experience and improve the accuracy of identifying the normal and fault states of the power unit. Furthermore, this application outputs fault warning information when the operational status classification result indicates that the power unit is in a fault state. Since the fault determination is triggered directly after the acquisition of operating parameters and status identification, it can provide timely warnings when the power unit is running abnormally. This solves the problems in related technologies, such as the difficulty in comprehensively representing the multi-parameter operational status, the strong dependence of fault identification on human experience, and the difficulty in timely warning of potential faults. Attached Figure Description

[0020] Figure 1 A flowchart illustrating a method for detecting faults in the power unit of an underground scraper conveyor in a coal mine, provided in an embodiment of this application; Figure 2 A visual schematic diagram of the classification results of the operating status of the power unit of a scraper conveyor provided in an embodiment of this application; Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0022] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains; the terminology used herein in the specification of the application is for the purpose of describing particular embodiments only and is not intended to limit the application; the terms “comprising” and “having”, and any variations thereof, in the specification, claims and drawings of this application are intended to cover non-exclusive inclusion.

[0023] The term "embodiment" as used herein means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of the phrase "embodiment" in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0024] In this article, the term "and / or" simply describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can mean: A exists, A and B can exist simultaneously, and B exists. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0025] Furthermore, the terms "first," "second," etc., in the specification and claims of this application or in the aforementioned drawings are used to distinguish different objects rather than to describe a specific order, and may explicitly or implicitly include one or more of the features.

[0026] In the description of this application, unless otherwise stated, "multiple" and "at least two" mean two or more (including two), and similarly, "multiple groups" and "at least two groups" mean two or more (including two groups).

[0027] In the description of this application, it should be noted that, unless otherwise explicitly specified and limited, the terms "connected" and "linked" should be interpreted broadly. For example, "connected" or "linked" can refer not only to a physical connection, but also to an electrical connection or a signal connection. For instance, it can be a direct connection, i.e., a physical connection, or an indirect connection through at least one intermediate component, as long as the circuit is connected. It can also refer to the internal connection between two components. A signal connection can refer not only to a signal connection through a circuit, but also to a signal connection through a medium, such as radio waves. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0028] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. It should be noted that, unless otherwise specified, different technical features in this application can be combined with each other.

[0029] Figure 1 This is a flowchart illustrating a method for detecting faults in the power unit of an underground scraper conveyor in a coal mine, as provided in an embodiment of this application. Figure 1 As shown, the power unit fault detection method provided in this application embodiment specifically includes S101 to S104, and S101 to S104 will be described in detail below.

[0030] It should be noted that the execution subject of the fault detection method for the power unit of an underground scraper conveyor in a coal mine provided in this application embodiment can be an electronic device. The electronic device can be an industrial control computer, a programmable logic controller (PLC), an edge computing gateway, an underground monitoring terminal, a data acquisition and processing all-in-one machine, a server, a host computer, a mine explosion-proof computing device, or other devices with data storage, data processing, and fault detection model calculation capabilities that are matched with the underground scraper conveyor.

[0031] S101. Collect the operating parameters of the scraper conveyor's power unit.

[0032] It should be noted that the power unit of the scraper conveyor includes a reducer and an electric motor. Operating parameters are used to characterize the temperature rise, cooling, and load conditions of the power unit during operation, thus providing a data foundation for subsequently constructing the power unit's state feature vector and performing fault detection.

[0033] The operating parameters include the reducer input shaft temperature, reducer output shaft temperature, reducer cooling water temperature, reducer lubricating oil temperature, reducer cooling water pressure, motor bearing temperature, and motor winding temperature.

[0034] It should be noted that the reducer input shaft temperature refers to the real-time temperature at the reducer input shaft bearing, which can be collected by setting a first temperature sensor at the reducer input shaft bearing.

[0035] The output shaft temperature of the reducer refers to the real-time temperature at the bearing of the reducer output shaft. The output shaft temperature can be collected by setting a second temperature sensor at the bearing of the reducer output shaft.

[0036] The reducer cooling water temperature refers to the real-time temperature at the outlet of the reducer cooling water circuit. The reducer cooling water temperature can be collected by setting a third temperature sensor at the outlet of the reducer cooling water circuit.

[0037] The gearbox lubricating oil temperature refers to the real-time temperature inside the gearbox lubricating oil tank. The gearbox lubricating oil temperature can be collected by installing a fourth temperature sensor inside the gearbox lubricating oil tank.

[0038] The reducer cooling water pressure refers to the real-time pressure in the reducer cooling water circuit, which can be collected by installing a pressure sensor in the reducer cooling water circuit.

[0039] The motor bearing temperature refers to the average real-time temperature at the bearings at both ends of the motor. It can be determined by setting a fifth temperature sensor and a sixth temperature sensor at the bearings at both ends of the motor, and by using the temperature information collected by the fifth and sixth temperature sensors.

[0040] The motor winding temperature refers to the real-time temperature at the stator winding of the motor. It can be collected by installing a seventh temperature sensor at the stator winding of the motor.

[0041] In this embodiment, the input shaft temperature and output shaft temperature of the reducer reflect the heat generation of the reducer's transmission components; the coolant temperature and pressure reflect the operating status of the reducer's cooling system; the lubricating oil temperature reflects the lubrication system and internal friction status; and the motor bearing temperature and winding temperature reflect the mechanical and electrical operating status of the motor. By simultaneously collecting these multiple operating parameters, rather than relying solely on a single temperature or pressure parameter, the completeness of the power unit's operating status characterization can be improved, providing more comprehensive data support for subsequent fault identification.

[0042] Furthermore, after collecting the various operating parameters, the data acquisition unit can aggregate the parameters and store and update them according to a preset sampling period.

[0043] The preset sampling period can be set according to actual application requirements. For example, it can be set to collect data once per second to achieve continuous monitoring of the operating status of the scraper conveyor's power unit. In other embodiments, the preset sampling period can also be set to other values, which are not limited in this application.

[0044] In some embodiments, the collected operating parameters can constitute a set of raw operating data at the same moment, which can be used as input data for subsequently constructing the state feature vector of the power unit. That is, at a certain sampling moment, the sampled values ​​corresponding to the reducer input shaft temperature, reducer output shaft temperature, reducer cooling water temperature, reducer lubricating oil temperature, reducer cooling water pressure, motor bearing temperature, and motor winding temperature can be correlated to form a set of operating parameter data corresponding to that sampling moment. Subsequently, a state feature vector characterizing the operating state of the power unit can be constructed based on this set of operating parameter data.

[0045] S102. Based on the operating parameters, construct the state feature vector of the power unit.

[0046] In this embodiment, a state feature vector is constructed to characterize the operating state of the power unit of the scraper conveyor based on the collected temperature of the reducer input shaft, reducer output shaft, reducer cooling water, reducer lubricating oil, reducer cooling water pressure, motor bearing temperature, and motor winding temperature.

[0047] By combining and processing multiple operating parameters, the operating state of the power unit at the current moment can be transformed into a feature representation that is easy for subsequent fault detection models to identify, thus providing an input basis for subsequent operating state classification.

[0048] S103. Input the state feature vector into the pre-trained support vector machine (SVM) fault detection model to obtain the running state classification result.

[0049] The operational status classification results are used to indicate whether the power unit is in a normal or faulty state.

[0050] It should be noted that the Support Vector Machine (SVM) fault detection model can be pre-trained based on historical operating samples, and the classification boundary used to distinguish between normal and fault states is determined during the training process. Therefore, after inputting the current state feature vector into the SVM fault detection model, the model can output the corresponding operating state classification result based on the positional relationship between the state feature vector and the classification boundary.

[0051] When the state feature vector is located on the side representing the fault state, the operation state classification result indicates that the power unit is in a fault state; when the state feature vector is located on the side representing the normal state, the operation state classification result indicates that the power unit is in a normal state.

[0052] The Support Vector Machine (SVM) fault detection model can comprehensively analyze the combined effects of multiple operating parameters on the operating status of the power unit, thereby improving the accuracy and reliability of identifying power unit fault states.

[0053] S104. When the operation status classification result indicates that the power unit is in a fault state, output fault warning information.

[0054] It should be noted that the fault warning information can be used to indicate the risk of abnormal operation of the current scraper conveyor power unit, so that relevant personnel can take timely measures such as inspection, maintenance, shutdown for inspection or fault handling, thereby reducing the adverse impact of sudden shutdown, overheating damage or transmission abnormality of the power unit on the continuity and safety of underground coal mine production.

[0055] In some embodiments, the fault warning information may include fault prompt information, fault status identification information, and / or alarm information. For example, prompt information such as "power unit fault warning," "reducer abnormality warning," and "motor abnormality warning" may be output; alarm information may also be output through sound, light, electricity, or other means; or the fault warning information may be sent to a host computer, monitoring platform, downhole monitoring terminal, edge computing gateway, or remote operation and maintenance terminal to achieve timely notification of the fault status of the scraper conveyor's power unit.

[0056] In some embodiments, when outputting fault warning information, the operating parameters, state feature vectors, and operating state classification results corresponding to the current sampling time can be associated and stored for subsequent fault tracing, operation analysis, and maintenance decisions.

[0057] Furthermore, when the operating status classification result indicates that the power unit is in a normal state, the method steps shown in S101 to S103 above are continued to be executed to achieve continuous monitoring of the operating status of the power unit.

[0058] This application provides a method for fault detection in the power unit of a scraper conveyor in an underground coal mine. This method collects operating parameters of the scraper conveyor's power unit, which simultaneously cover the temperature of the reducer input shaft, reducer output shaft, reducer cooling water, reducer lubricating oil, reducer cooling water pressure, motor bearing temperature, and motor winding temperature. These parameters reflect the reducer's transmission, cooling, and lubrication states, as well as the mechanical and electrical operating states of the motor. These previously scattered operating parameters are then unified into a state feature vector for the power unit, thus achieving a comprehensive representation of the overall operating state of the power unit. Based on this, the state feature vector is input into a pre-trained SVM fault detection model, enabling the SVM to detect faults in the power unit. The M-fault detection model outputs operational status classification results based on the combination relationship and state differences between multiple operating parameters. Compared with methods that rely on manual inspection, post-event maintenance, or single parameter threshold judgment, the method of this application can reduce the dependence on human experience and improve the accuracy of identifying the normal and fault states of the power unit. Furthermore, this application outputs fault warning information when the operational status classification result indicates that the power unit is in a fault state. Since the fault determination is triggered directly after the acquisition of operating parameters and state identification, it can provide timely warnings when the power unit is running abnormally, thereby solving the problems in related technologies such as the difficulty in comprehensively representing the multi-parameter operational status, the strong dependence of fault identification on human experience, and the difficulty in timely warning of potential faults.

[0059] In one possible embodiment, the method steps shown in S102 can be implemented by S1021 and S1022, which are described in detail below.

[0060] S1021. The temperature of the input shaft of the reducer, the temperature of the output shaft of the reducer, the temperature of the cooling water of the reducer, the temperature of the lubricating oil of the reducer, the pressure of the cooling water of the reducer, the temperature of the motor bearing and the temperature of the motor winding are combined as feature components to obtain a multi-dimensional feature vector characterizing the operating state of the power unit.

[0061] It should be noted that the reducer input shaft temperature, reducer output shaft temperature, reducer cooling water temperature, reducer lubricating oil temperature, reducer cooling water pressure, motor bearing temperature, and motor winding temperature correspond to seven key operating parameters. These parameters respectively reflect the thermal state of the reducer transmission components, the operating state of the cooling system, the operating state of the lubrication system, and the operating state of the mechanical and electrical components of the motor within the power unit. Therefore, by combining these seven operating parameters, the operating information of the power unit at a given sampling moment can be uniformly represented as a multi-dimensional feature vector, thus providing an input basis for subsequent fault detection model identification.

[0062] In some embodiments, for the first For each sample, its corresponding multidimensional feature vector can be represented as: ,in, Indicates the temperature of the reducer input shaft. Indicates the output shaft temperature of the reducer. Indicates the coolant temperature of the reducer. Indicates the temperature of the gearbox lubricating oil. Indicates the coolant pressure of the reducer. This indicates the temperature of the motor bearing. This represents the temperature of the motor windings. Therefore, the original dataset D={( , ), ( , ), ..., ( , )},in, The total number of samples, , For the first The sample label corresponding to each sample.

[0063] S1022. Standardize the multidimensional feature vectors to obtain the state feature vectors.

[0064] It should be noted that, since the SVM model is sensitive to feature scale, it is necessary to standardize the multidimensional feature vectors to eliminate the differences in dimensionality and numerical range between different feature components, thereby obtaining the state feature vector.

[0065] Understandably, since temperature and pressure parameters typically differ in dimensions and value ranges, directly inputting the original multidimensional feature vectors into the SVM fault detection model could lead to feature components with larger numerical ranges having a greater impact on model training and classification results, hindering a balanced representation of the power unit's operating state across all feature components. Therefore, standardizing the multidimensional feature vectors allows different feature components to operate on a relatively uniform numerical scale, thereby improving the accuracy and stability of subsequent fault detection model classification results.

[0066] In one possible embodiment, the method steps shown in S1021 can be implemented by Sa1 to Sa3, which are described in detail below. Sa1. For each feature component in the multidimensional feature vector, determine the sample mean and sample standard deviation corresponding to the feature component.

[0067] In this embodiment, for the first The first sample The characteristic component, the _ ... The original acquired values ​​of each feature component are denoted as follows: ,in, =1, 2, ..., , =1, 2, ..., 7.

[0068] For the The first feature component can be calculated based on the original collected values ​​of all samples. The sample mean corresponding to each feature component and sample standard deviation .

[0069] Among them, the The sample mean corresponding to each feature component It can be represented as: ;No. The sample standard deviation corresponding to each feature component It can be represented as: .

[0070] It should be noted that the sample mean Used to characterize the The average level of each feature component in all samples, and the sample standard deviation. Used to characterize the The degree of dispersion of each feature component across all samples.

[0071] In this embodiment, by determining the sample mean and sample standard deviation corresponding to the seven feature components respectively, a computational basis can be provided for subsequent standardization processing of each feature component.

[0072] Sa2. Based on the sample mean and sample standard deviation corresponding to the feature components, the feature components are dimensionless to obtain the standardized feature values ​​corresponding to the feature components.

[0073] In this embodiment, it can be based on the first The sample mean corresponding to each feature component and sample standard deviation , for the The first sample The dimensionless processing of the characteristic component is performed to obtain the . The first sample Standardized eigenvalues ​​corresponding to each feature component The standardized calculation formula can be expressed as: .

[0074] Through the above processing, the original eigenvalues ​​of each feature component can be converted into corresponding standardized eigenvalues, thereby eliminating the dimensional differences between different feature components.

[0075] Sa3. Based on the standardized eigenvalues ​​corresponding to each feature component, the state feature vector is obtained.

[0076] In this embodiment, after obtaining the standardized feature values ​​corresponding to the seven feature components, a standardized state feature vector can be constructed based on the standardized feature values ​​corresponding to each feature component.

[0077] For the For each sample, its corresponding state feature vector can be represented as: A standardized dataset can be formed from multiple state feature vectors. D ={( , ), ( , ), ..., ( , )}.

[0078] It should be noted that after standardization, the standardized feature values ​​corresponding to each feature component usually have a mean of 0 and a standard deviation of 1. This ensures that the SVM fault detection model assigns relatively fair weights to each feature component and avoids any feature component from having an excessively large value range that could unduly dominate the classification results.

[0079] In this embodiment of the application, by combining and standardizing multiple operating parameters of the power unit, the influence of differences in the dimensions and numerical ranges of different parameters on the state representation can be reduced, thereby improving the consistency and effectiveness of the state feature vector and providing a reliable data foundation for the subsequent fault detection model to accurately identify the operating state of the power unit.

[0080] In one possible embodiment, prior to the method step shown in S103, the method further includes Sc1 and Sc2, which are described in detail below.

[0081] Sc1: Obtain multiple historical running samples. Each historical running sample includes a historical state feature vector and a sample label corresponding to the historical state feature vector.

[0082] In this embodiment, multiple historical running samples can be obtained in advance for subsequent training of the initial SVM fault detection model to obtain the SVM fault detection model.

[0083] The historical state feature vector can be obtained by combining and standardizing historical operating parameters. Sample labels include normal state labels and fault state labels.

[0084] In some embodiments, the historical state feature vector can be categorized based on whether the various operating parameters of the power unit are within a safe range and whether the equipment has actually malfunctioned.

[0085] When all feature parameters corresponding to the historical state feature vector meet the preset safety threshold conditions, the sample label corresponding to the historical state feature vector is the normal state label.

[0086] When at least one feature parameter corresponding to the historical state feature vector does not meet the preset safety threshold condition, and / or when the scraper conveyor actually fails, the sample label corresponding to the historical state feature vector is the fault state label.

[0087] It should be noted that the preset safety threshold conditions refer to each characteristic parameter meeting its corresponding safety threshold range. Specifically, the safety threshold for the reducer input shaft temperature can be no greater than 85℃, the safety threshold for the reducer output shaft temperature can be no greater than 65℃, the safety threshold for the reducer cooling water temperature can be no greater than 60℃, the safety threshold for the reducer lubricating oil temperature can be no greater than 70℃, the safety threshold for the reducer cooling water pressure can be no greater than 2MPa, the safety threshold for the motor bearing temperature can be no greater than 75℃, and the safety threshold for the motor winding temperature can be no greater than 90℃. When all characteristic parameters corresponding to the historical state feature vector meet the above preset safety threshold conditions, the sample label corresponding to the historical state feature vector can be determined as a normal state label; when at least one characteristic parameter corresponding to the historical state feature vector does not meet the above preset safety threshold conditions, and / or when the scraper conveyor actually malfunctions, the sample label corresponding to the historical state feature vector can be determined as a fault state label.

[0088] Furthermore, in some embodiments, the normal state label can be recorded as 0, and the fault state label can be recorded as 1, to form a labeled sample set for training the fault detection model. By labeling the historical state feature vectors with normal / fault labels, the subsequent initial SVM fault detection model can learn the classification boundary between normal state samples and fault state samples based on supervised learning, thereby improving the accuracy of the SVM fault detection model in identifying the fault state of the power unit.

[0089] Sc2. Based on multiple historical running samples, the initial SVM fault detection model is trained to obtain the SVM fault detection model.

[0090] The initial SVM fault detection model is an initial model that has not yet completed parameter learning and classification boundary determination. By using multiple historical running samples for training, the trained SVM fault detection model can be equipped with the ability to distinguish between normal and fault states.

[0091] In one possible embodiment, the method steps shown in Sc2 can be implemented by Sc21 to Sc23, which are described in detail below.

[0092] Sc21. Based on multiple historical operating samples, construct a soft-interval SVM binary classification model to distinguish between normal and fault states.

[0093] In this embodiment, since normal samples and fault samples of the power unit's operating state are usually not strictly linearly separable, a soft-interval SVM binary classification model can be constructed to distinguish between normal and fault states.

[0094] It should be noted that soft-margin SVM, by introducing slack variables into the classification model, allows a small number of samples to deviate from the ideal classification boundary, thereby enhancing the model's fault tolerance and generalization ability for complex working conditions. It is suitable for handling sample situations in the operating state of the power unit of the scraper conveyor in underground coal mines where there is noise, fluctuation and boundary overlap.

[0095] In some embodiments, to unify the mathematical expression in the Support Vector Machine (SVM) model, the normal state label can be mapped from 0 to -1, while the fault state label remains 1, thereby making the sample labels... ∈{-1,1}. Therefore, a binary classification sample set suitable for SVM training can be established based on labeled historical running samples.

[0096] In some embodiments, the primal optimization problem of the soft-margin SVM binary classification model can be expressed as: The constraints are: .

[0097] in, represents the normal vector of the SVM hyperplane, corresponding to the seven normalized features; b represents the intercept of the SVM hyperplane; Indicates the first The slack variable for each sample is used to allow a small number of samples to deviate from the classification boundary; C represents the penalty coefficient, and C > 0, which is used to control the strictness of the soft margin. The larger the penalty coefficient C is, the heavier the penalty for misclassified samples, and the more strict the model is. Indicates the first Standardized feature vectors of each sample; Indicates the first The labels of each sample; express and The inner product of.

[0098] By constructing the soft-interval SVM binary classification model described above, a classification hyperplane for distinguishing between normal and fault states can be determined based on the differences in operating states contained in historical operating samples.

[0099] In one possible embodiment, the method steps shown in Sc21 can be implemented by Sc211 to Sc213, which are described in detail below.

[0100] Sc211. Based on multiple historical running samples, determine the kernel function used to characterize the nonlinear relationship between samples. The kernel function is the radial basis function (RBF).

[0101] In this embodiment, since multiple historical running samples may exhibit nonlinear distribution characteristics in the original feature space, a kernel function can be introduced to characterize the nonlinear relationships between samples. The kernel function is the Radial Basis Function (RBF), which can better handle the nonlinear classification problem between normal state samples and fault state samples, thereby improving the support vector machine (SVM) model's ability to identify complex running states.

[0102] In some embodiments, RBF can be represented as: .

[0103] in, The width parameter of RBF ( >0); The larger the value, the smaller the coverage of the kernel function; the more the SVM fault detection model tends to fit local samples. The smaller the value, the stronger the generalization ability of the SVM fault detection model.

[0104] Represents the standardized feature vector and The square of the Euclidean distance, , Indicates the first The k-th standardized feature value of a sample; Indicates the first The k-th standardized eigenvalue of a feature component.

[0105] Sc212, based on kernel functions, maps multiple historical running samples to a high-dimensional feature space.

[0106] In this embodiment, by using kernel function mapping, normal state samples and fault state samples that are difficult to linearly separate in low-dimensional space can be transformed into a sample distribution form that is easier to separate in high-dimensional space, thereby providing a foundation for the subsequent construction of SVM classification boundaries.

[0107] Furthermore, the SVM fault detection model can directly calculate the similarity between samples in the high-dimensional feature space through kernel functions, without having to explicitly calculate the high-dimensional coordinates after the specific mapping, thereby reducing the computational complexity in the high-dimensional mapping process.

[0108] Sc213. Based on multiple historical running samples mapped to a high-dimensional feature space, a soft-margin SVM binary classification model is constructed to distinguish between normal and fault states.

[0109] In this embodiment, after mapping multiple historical running samples to a high-dimensional feature space, a soft-interval SVM binary classification model can be constructed based on the mapped sample data to distinguish between normal and fault states.

[0110] Furthermore, based on Lagrange duality theory, the original optimization problem of soft-margin SVM can be transformed into a dual optimization problem.

[0111] Specifically, a Lagrange function can be constructed, and Lagrange multipliers can be introduced. ,get: ,in ≥0 is Constraint multipliers ≥ 0.

[0112] right , , Taking the partial derivative and setting it to 0, we simplify to obtain the dual optimization problem: The constraints are , ;in, Indicates the first The Lagrange multiplier for each sample only supports vectors. >0, other samples corresponding =0; The kernel function is represented by RBF in this embodiment, which is used to address the nonlinear relationship between features.

[0113] In this embodiment, by transforming the original optimization problem of soft-margin SVM into a dual optimization problem, the kernel function can be used more efficiently to solve the optimal classification boundary, thereby constructing a soft-margin SVM binary classification model for distinguishing between normal and fault states.

[0114] Sc22. In the process of training the soft-margin SVM binary classification model using multiple historical running samples, slack variables and penalty coefficients are introduced to provide fault tolerance for historical running samples that deviate from the classification boundary and / or misclassified historical running samples, so as to determine the classification boundary used to distinguish between normal and fault states.

[0115] In actual underground coal mine operations, due to factors such as sensor noise, equipment operating condition fluctuations, and blurred sample boundaries, some historical operating samples may not fully meet the ideal linear separability condition. By introducing slack variables, these samples can be allowed to deviate from the classification boundary to some extent; by setting a penalty coefficient, the degree of penalty imposed by the model on these deviating samples can be controlled, thereby achieving a balance between model fitting ability and generalization ability.

[0116] In some embodiments, by solving the optimization problem of the soft-margin SVM binary classification model, the classification boundary used to distinguish between normal and fault states can be determined.

[0117] The classification boundary is the optimal segmentation hyperplane used to separate normal state samples from fault state samples in the high-dimensional feature space.

[0118] In this embodiment, by introducing slack variables and penalty coefficients, not only can the SVM fault detection model be improved to adapt to abnormal and noisy samples, but the SVM fault detection model can also be prevented from overfitting to the training samples, thereby improving the fault identification performance of the model in actual operating scenarios.

[0119] Sc23. Based on the classification boundary, the SVM fault detection model is obtained.

[0120] In some embodiments, after training is complete, the decision function can be used. Determine the running status of the new sample; among which, The standardized feature vector of the newly collected samples; For the set of support vectors, i.e. A sample set > 0; The optimal Lagrange multiplier for the dual problem; The optimal intercept; For a sign function, when When it is determined to be a fault state, when It is judged to be in a normal state at this time.

[0121] After determining the classification boundary, the support vectors, the corresponding optimal Lagrange multipliers, the kernel function, and the optimal intercept together constitute the decision-making mechanism of the Support Vector Machine (SVM) fault detection model. When a new state feature vector is input into the SVM fault detection model, the model can classify the current operating state of the power unit as normal or faulty based on the learned classification boundary.

[0122] In one possible embodiment, after the method steps shown in Sc23, the method may further include Sc24 to Sc27, which are described in detail below.

[0123] Sc24. For the soft-margin SVM binary classification model, determine the parameters to be optimized, including the penalty coefficient and kernel function parameters.

[0124] In this embodiment, in order to improve the classification performance of the soft-spacing SVM binary classification model for the normal and fault states of the power unit, the parameters to be optimized for the soft-spacing SVM binary classification model can be determined.

[0125] The parameters to be optimized may include penalty coefficients. and kernel function parameters Penalty coefficient Kernel function parameters are used to adjust the penalty level for misclassified samples in the model. The penalty coefficient is used to adjust the effective range of RBF; therefore, setting the penalty coefficient appropriately is crucial. and kernel function parameters This is beneficial for improving the classification accuracy and generalization ability of the model.

[0126] In some embodiments, the parameter search range can be preset, for example, the penalty coefficient. The candidate values ​​can be set as: Kernel function parameters The candidate values ​​can be set as: .

[0127] In other embodiments, the penalty coefficient and kernel function parameters The range of candidate values ​​can also be adjusted according to actual application needs, and this application does not limit it.

[0128] Sc25. Based on the parameters to be optimized, construct multiple parameter combinations.

[0129] In this embodiment, multiple parameter combinations can be constructed based on the candidate values ​​of the penalty coefficient and the candidate values ​​of the kernel function parameter determined in Sc24. Each parameter combination includes a penalty coefficient and a kernel function parameter.

[0130] By constructing multiple parameter combinations, different parameter configuration schemes can be provided for the training and validation of the subsequent initial SVM fault detection model, so as to select the target parameter combination that is more suitable for the current power unit fault detection task from multiple parameter combinations.

[0131] Sc26. Based on multiple parameter combinations, the soft-margin SVM binary classification model is trained and validated to determine the target parameter combination.

[0132] In one possible embodiment, the method steps shown in Sc26 can be implemented by Sc261 to Sc264, which are described in detail below.

[0133] Sc261 divides multiple historical running samples into multiple training samples and multiple validation samples.

[0134] In this embodiment, multiple historical running samples can be divided into multiple training samples and multiple validation samples so that model training and model validation can be performed separately for different parameter combinations.

[0135] In a specific example, a 5-fold cross-validation method can be used to divide the standardized dataset D' into 5 groups of samples. Each time, 4 groups are taken as training samples and 1 group is taken as validation samples. This process is repeated 5 times to more comprehensively evaluate the classification performance of the model corresponding to different parameter combinations, thereby improving the reliability of the selection of target parameter combinations.

[0136] Sc262. For each parameter combination, train the soft-margin SVM binary classification model using multiple sets of training samples to obtain the trained soft-margin SVM binary classification model corresponding to the parameter combination.

[0137] In this embodiment, for each parameter combination, the soft-margin SVM binary classification model can be trained using the corresponding multiple sets of training samples to obtain the trained soft-margin SVM binary classification model corresponding to that parameter combination.

[0138] Since different parameter combinations correspond to different penalty coefficients and kernel function parameters, the classification boundary and model performance of soft-margin SVM binary classification models trained under different parameter combinations will also differ.

[0139] Sc263 uses multiple sets of validation samples to validate the trained soft-margin SVM binary classification model corresponding to the parameter combination, and obtains the classification performance of the model corresponding to the parameter combination.

[0140] In some embodiments, the model classification performance can be characterized by the cross-validation accuracy Acc, which can be expressed as: .

[0141] Among them, TP represents a true positive, meaning that the faulty sample was correctly identified as faulty; TN represents a true negative, meaning that the normal sample was correctly identified as normal; FP represents a false positive, meaning that the normal sample was mistakenly identified as faulty; and FN represents a false negative, meaning that the faulty sample was mistakenly identified as normal.

[0142] In some embodiments, one or more of precision (P), recall (R), F1 score, AUC-ROC, and mean squared error (MSE) may be used as evaluation metrics for model classification performance.

[0143] Specifically, the accuracy P can be expressed as: Recall rate R can be expressed as The F1 score can be expressed as: .

[0144] For AUC-ROC, the area under the ROC curve can be used as an evaluation metric for the model's discriminative ability. The ROC curve is further subdivided by the false positive rate. The horizontal axis represents the true positive rate. The vertical axis is denoted by , and AUC is the area under the ROC curve. The AUC value ranges from [0,1], and the closer it is to 1, the better the model performance.

[0145] The mean square error (MSE) can be expressed as: ;in, Indicates the number of samples in the test set; This represents the model output for the i-th sample in the test set, and its value is either 1 or -1. This represents the mapping label of the i-th sample in the test set, with a value of 1 or -1.

[0146] In this embodiment, by validating the trained models corresponding to different parameter combinations and obtaining the corresponding model classification performance, a basis can be provided for selecting the target parameter combination with the best classification performance in the future.

[0147] Sc264 compares the classification performance of the models corresponding to multiple parameter combinations and determines the parameter combination with the best classification performance as the target parameter combination.

[0148] In an embodiment employing 5-fold cross-validation, the average accuracy of the five cross-validations for each parameter combination can be used as the performance metric for that parameter combination, and the parameter combination with the highest average accuracy is selected as the target parameter combination. This allows the determination of the target model parameters for the soft-margin SVM binary classification model.

[0149] In a specific example, the parameter combination that maximizes the average cross-validation accuracy can be determined as the optimal parameter combination. , For example, the optimal parameter could be... =10, =0.1. In other embodiments, the combination of target parameters may vary depending on the actual training samples and application scenarios, and this application does not limit this.

[0150] Sc27. Determine the target parameter combination as the target model parameters for the soft-margin SVM binary classification model.

[0151] In this embodiment, after determining the target parameter combination, the target parameter combination can be determined as the target model parameters of the soft-margin SVM binary classification model, thereby completing the optimized configuration of the support vector machine (SVM) fault detection model parameters.

[0152] In this embodiment, the SVM fault detection model is constructed, trained, and its parameters are optimized based on multiple labeled historical operating samples, thereby obtaining an SVM fault detection model that can distinguish between the normal state and the fault state of the power unit.

[0153] Figure 2 This is a visual schematic diagram illustrating the classification results of the operating status of the power unit of a scraper conveyor, provided as an embodiment of this application. For example... Figure 2 As shown, in this embodiment, the state feature vectors constructed based on the reducer input shaft temperature, reducer output shaft temperature, reducer cooling water temperature, reducer lubricating oil temperature, reducer cooling water pressure, motor bearing temperature, and motor winding temperature are mapped to a two-dimensional plane for visualization after dimensionality reduction processing. In the figure, sample points of different categories represent the sample distribution corresponding to the power unit being in normal and fault states, respectively. The curve boundary line represents the classification boundary learned by the SVM fault detection model, and the circled sample points represent the support vectors that play a major role in forming the classification boundary. Figure 2 It can be seen that the normal state samples and the fault state samples show obvious distribution differences in the feature space. The SVM fault detection model can effectively distinguish between the two types of samples based on the learned classification boundary, thus providing support for the output of the power unit's operating status classification results.

[0154] Next, a specific example of this application is shown.

[0155] First, data collection and preparation: PT100 temperature sensors were installed at the input shaft bearing, output shaft bearing, cooling water outlet, and lubricating oil tank of the reducer, as well as at the bearings at both ends of the motor and the stator winding; a pressure sensor was installed at the cooling water inlet of the reducer. A total of 259,200 sample data points were collected, and some of the raw data are shown in Table 1 below: Table 1

[0156] Normal state ( =0): ≤85、 ≤65、 ≤60、 ≤70、 ≤2、 ≤75、 ≤90; Fault status ( =1): Any feature exceeds the limit.

[0157] The final dataset contains 215,340 normal samples (83.1%) and 43,860 faulty samples (16.9%).

[0158] Second, data preprocessing: Using the above formula The seven characteristics are standardized, and the calculation process is as follows, taking the motor winding temperature xi7 as an example: (1) Calculate the mean and standard deviation :

[0159]

[0160] (2) Standardization Take the fault sample 186320 and find xi7 = 128.32, then substitute it into the formula. : x ′ i7 = (128.32-38.62) / 12.35=7.26, and the standardization results of the other features are similar, finally obtaining the standardized dataset D′.

[0161] Third, SVM model training and optimization (1) Divided into 7:3 ratio: training set mtrain=181440 records (150738 normal records + 30702 fault records), test set mtest=77760 records (64602 normal records + 13158 fault records).

[0162] (2) Using the grid search method, let C∈{2 3,2 1,21,...,211},γ∈{2 9,2 After 5-fold cross-validation, the optimal parameters for the set 7,...,23} are Copt=10 and γopt=0.1 (consistent with the technical solution), and the number of support vectors is S=12860.

[0163] (3) Based on the training results, the optimal decision function is:

[0164] in =2.31 (obtained from the support vector calculation). The optimal Lagrange multiplier corresponding to the support vector (mean is 0.086).

[0165] Fourth, fault detection applications (1) Data from a normal sample at a certain moment (after standardization):

[0166] Substitute into the decision function:

[0167] The system is determined to be in a normal state, and the equipment continues to operate.

[0168] (2) Fault warning and detection data at a certain moment (after standardization):

[0169] Substitute into the decision function:

[0170] This triggers a fault warning, indicating "overheating fault of motor winding and reducer lubricating oil".

[0171] Fifth, result verification Test set detection results statistics: TP=12896 (faulty sample correctly identified), TN=64385 (normal sample correctly identified) FP=217 (false positives for normal samples), FN=262 (false negatives for faulty samples) Substitute into the performance index formula: Accuracy: Acc = (12896 + 64385) / 77760 = 99.38%; Precision: P = 12896 / (12896 + 217) = 98.33%; Recall rate: R = 12896 / (12896 + 262) = 97.98%; F1 score: F1 = (2 × 12896) / (2 × 12896 + 217 + 262) = 98.15%; AUC-ROC: 0.997.

[0172] Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 3 As shown, the electronic device 300 provided in this embodiment includes a memory 301 and a processor 302.

[0173] The memory 301 can be a separate physical unit, connected to the processor 302 via a bus 303. Alternatively, the memory 301 and processor 302 can be integrated and implemented in hardware. The memory 301 stores program instructions, which the processor 302 calls to execute the operations performed by the electronic device in any of the above method embodiments.

[0174] Optionally, when some or all of the methods in the above embodiments are implemented by software, the electronic device 300 may also include only the processor 302. A memory 301 for storing programs is located outside the electronic device 300, and the processor 302 is connected to the memory via circuits / wires to read and execute the programs stored in the memory. The processor 302 may be a central processing unit (CPU), a network processor (NP), or a combination of a CPU and an NP. The processor 302 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The PLD may be a complex programmable logic device (CPLD), a field-programmable gate array (FPGA), a generic array logic (GAL), or any combination thereof.

[0175] The memory 301 may include volatile memory, such as random-access memory (RAM); the memory may also include non-volatile memory, such as flash memory, hard disk drive (HDD) or solid-state drive (SSD); the memory may also include a combination of the above types of memory.

[0176] For example, this application provides a computer-readable storage medium having computer program instructions stored thereon, which are executed by the processor of an electronic device to cause the electronic device to perform the operations performed by the electronic device in the above method embodiments.

[0177] For example, this application provides a computer program product that, when run on an electronic device, causes the electronic device to perform the operations performed by the electronic device in the above method embodiments.

[0178] The above description is merely a specific embodiment of this application, enabling those skilled in the art to understand or implement this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments described herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for detecting faults in the power unit of an underground scraper conveyor in a coal mine, characterized in that, The method includes: The operating parameters of the scraper conveyor's power unit are collected, including the temperature of the reducer input shaft, the temperature of the reducer output shaft, the temperature of the reducer cooling water, the temperature of the reducer lubricating oil, the pressure of the reducer cooling water, the temperature of the motor bearing, and the temperature of the motor winding. Based on the operating parameters, construct the state feature vector of the power unit; The state feature vector is input into a pre-trained support vector machine (SVM) fault detection model to obtain the operating state classification result; wherein, the operating state classification result is used to indicate whether the power unit is in a normal state or a fault state; When the operation status classification result indicates that the power unit is in a fault state, a fault warning message is output.

2. The method according to claim 1, characterized in that, The process of constructing the state feature vector of the power unit based on the operating parameters includes: The temperature of the reducer input shaft, the temperature of the reducer output shaft, the temperature of the reducer cooling water, the temperature of the reducer lubricating oil, the pressure of the reducer cooling water, the temperature of the motor bearing, and the temperature of the motor winding are combined as feature components to obtain a multidimensional feature vector characterizing the operating state of the power unit. The multidimensional feature vector is standardized to obtain the state feature vector.

3. The method according to claim 2, characterized in that, The standardization process for the multidimensional feature vector to obtain the state feature vector includes: For each feature component in the multidimensional feature vector, determine the sample mean and sample standard deviation corresponding to the feature component; Based on the sample mean and sample standard deviation corresponding to the feature component, the feature component is dimensionless to obtain the standardized feature value corresponding to the feature component. The state feature vector is obtained based on the standardized feature values ​​corresponding to each of the feature components.

4. The method according to claim 1, characterized in that, Before inputting the state feature vector into the pre-trained support vector machine (SVM) fault detection model to obtain the running state classification result, the method further includes: Multiple historical running samples are obtained, each of which includes a historical state feature vector and a sample label corresponding to the historical state feature vector; Based on multiple historical operating samples, the initial SVM fault detection model is trained to obtain the SVM fault detection model.

5. The method according to claim 4, characterized in that, The sample labels include normal status labels and fault status labels; When all feature parameters corresponding to the historical state feature vector meet the preset safety threshold condition, the sample label corresponding to the historical state feature vector is the normal state label. When at least one feature parameter corresponding to the historical state feature vector does not meet the preset safety threshold condition, and / or when the scraper conveyor actually malfunctions, the sample label corresponding to the historical state feature vector is a fault state label.

6. The method according to claim 4, characterized in that, The initial SVM fault detection model is trained based on multiple historical operating samples to obtain an SVM fault detection model, including: Based on multiple historical operation samples, a soft-interval SVM binary classification model is constructed to distinguish between normal and fault states. During the training of the soft-margin SVM binary classification model using multiple historical running samples, slack variables and penalty coefficients are introduced to provide fault tolerance for historical running samples that deviate from the classification boundary and / or misclassified historical running samples, so as to determine the classification boundary used to distinguish between normal and fault states. Based on the classification boundary, the SVM fault detection model is obtained.

7. The method according to claim 6, characterized in that, The construction of a soft-margin SVM binary classification model based on multiple historical operating samples to distinguish between normal and fault states includes: Based on multiple historical running samples, a kernel function is determined to characterize the nonlinear relationship between the samples, wherein the kernel function is the radial basis function (RBF). Based on the kernel function, multiple historical running samples are mapped to a high-dimensional feature space; Based on multiple historical running samples mapped to a high-dimensional feature space, a soft-margin SVM binary classification model is constructed to distinguish between normal and fault states.

8. The method according to claim 7, characterized in that, The method further includes: For the soft-margin SVM binary classification model, the parameters to be optimized are determined, including the penalty coefficient and the kernel function parameters; Based on the parameters to be optimized, multiple parameter combinations are constructed; Based on multiple parameter combinations, the soft-margin SVM binary classification model is trained and validated to determine the target parameter combination. The combination of target parameters is determined as the target model parameters of the soft-margin SVM binary classification model.

9. The method according to claim 8, characterized in that, The process of training and validating the soft-margin SVM binary classification model based on multiple parameter combinations to determine the target parameter combination includes: The historical running samples are divided into multiple training samples and multiple validation samples; For each of the parameter combinations, the soft-margin SVM binary classification model is trained using multiple sets of training samples to obtain the trained soft-margin SVM binary classification model corresponding to the parameter combination. The trained soft-margin SVM binary classification model corresponding to the parameter combination is validated using multiple sets of validation samples to obtain the classification performance of the model corresponding to the parameter combination. The classification performance of the models corresponding to the multiple parameter combinations is compared, and the parameter combination with the best classification performance is determined as the target parameter combination.

10. An electronic device, characterized in that, Includes memory and at least one processor; The memory stores computer-executed instructions; The at least one processor executes computer execution instructions stored in the memory, causing the at least one processor to perform the method as described in any one of claims 1 to 9.

Images