Measurement point deployment and fault diagnosis method for collaborative monitoring of multiple components of electric shovel

Abstract

The present invention relates to a measurement point deployment and fault diagnosis method for collaborative monitoring of multiple components of an electric shovel. The method comprises: first, selecting a key vibration measurement surface by means of dynamic analysis, selecting an optimal measurement point deployment position on the basis of an effective vibration information proportion and a relative effective information amount, and using information entropy to evaluate the rationality of a scheme; next, mounting vibration measurement points on key components such as a drive motor, a disc brake, and a reduction gearbox, comprehensively acquiring a system vibration state, and further constructing a single fault feature extractor; and then, constructing a system-level fault diagnosis model, wherein the model comprises the single fault feature extractor, a component spatial relation graph, a classifier, etc., and the model is trained to implement collaborative fault diagnosis of key components of an electric shovel lifting mechanism transmission system. The present invention improves the accuracy and comprehensiveness of fault diagnosis, reduces downtime and maintenance costs caused by faults, provides effective technical support for safe operation of mining machinery, and has important practical application value.

Description

A method for deploying measuring points and diagnosing faults in collaborative monitoring of multiple components of an electric shovel. Technical Field

[0001] This invention relates to the field of system-level fault diagnosis, and in particular to a method for deploying measuring points and diagnosing faults in the collaborative monitoring of multiple components of an electric shovel. Background Technology

[0002] Electric shovels are key pieces of equipment in mining operations. The lifting mechanism transmission system of an electric shovel is responsible for providing kinetic energy for the bucket's digging operations and is one of the most critical working mechanisms of the shovel. As shown in Figure 2, this system includes key components such as a lifting gearbox, a gearbox support structure 1, a gearbox gear 2, two drive motors 3, a coupling 4, and a disc brake 6. The health of these components directly affects the digging efficiency, operational stability, and operator safety of the electric shovel.

[0003] However, most existing research is limited to fault diagnosis of single components in electric shovels, such as gearbox gear failure, drive motor bearing failure, or disc brake failure. This single-component diagnostic method ignores the interrelationships between components and the fault propagation process, making it difficult to achieve comprehensive fault diagnosis of the entire transmission system. Furthermore, existing technologies lack measurement point deployment schemes for system-level fault diagnosis, resulting in insufficient utilization of sensor information and further limiting the accuracy and efficiency of fault diagnosis.

[0004] Therefore, developing a method for the deployment of measuring points and fault diagnosis for collaborative monitoring of multiple components of electric shovels, which can simultaneously monitor and diagnose faults in multiple key components, is of great significance for improving the operational stability of electric shovels, reducing maintenance costs, and ensuring operational safety. Summary of the Invention

[0005] To address the aforementioned problems in electric shovel monitoring, this invention provides a method for the deployment of measuring points and fault diagnosis of multiple components in the coordinated monitoring of electric shovels. This method integrates the fault diagnosis of key components in the electric shovel's lifting mechanism transmission system, such as the drive motor, lifting gearbox, and disc brake, into a unified fault diagnosis paradigm, enabling more accurate, comprehensive, and flexible fault diagnosis of the entire electric shovel's lifting mechanism transmission system.

[0006] To achieve the above objectives, this invention provides a method for deploying measuring points and diagnosing faults in the collaborative monitoring of multiple components of an electric shovel, comprising the following steps:

[0007] A. The deployment strategy for measuring points in the transmission system of the electric shovel lifting mechanism specifically includes:

[0008] A1. Dynamics analysis of the transmission system of the electric shovel lifting mechanism: Dynamics analysis of the drive motor, lifting gearbox, and disc brake in the transmission system of the electric shovel lifting mechanism;

[0009] A2. Selection of vibration measurement surface: Based on the results of dynamic analysis, the surface with the largest vibration amplitude is selected as the vibration measurement surface for arranging vibration sensors;

[0010] A3. Measurement point deployment scheme: Install vibration sensors in the horizontal and vertical directions of the selected vibration measurement surface, and select the optimal multiple measurement points as the installation positions of the vibration sensors based on the proportion of effective vibration information and the relative amount of effective information.

[0011] A4. Evaluation of Measurement Point Deployment Based on Information Entropy: Introducing information entropy to judge the rationality of vibration measurement schemes until the conditions are met;

[0012] A5. Sensor data acquisition: Vibration sensors are installed on the drive motor, lifting gearbox, and disc brake to acquire their vibration status;

[0013] B. Construct fault diagnosis models for individual components, and build fault diagnosis networks for the drive motor, lifting gearbox, and disc brake, specifically including:

[0014] B1. Dataset Construction: Based on the measurement point deployment plan in step A, collect vibration data when various components such as the drive motor, lifting gearbox, and disc brake malfunction. The fault types of the drive motor include bearing wear and fatigue failure; the fault types of the disc vibrator include normal friction block wear, uneven wear, uniform wear, and spalling; the fault types of the gearbox include gear tooth surface wear and spalling, gear tooth cracks, and broken teeth. Use the vibration data as input and the fault types as label data as output to create dataset D.

[0015] B2 Single Fault Feature Extraction: A single fault feature extractor is constructed by alternately stacking convolutional layers, pooling layers, and batch normalization layers;

[0016] C. Construct a system-level fault diagnosis model, incorporating fault diagnosis of key components into a unified framework, specifically including:

[0017] C1. Generate component spatial relationship diagram: Treat the drive motor, disc brake, and gearbox measurement points as nodes, generate a component spatial relationship diagram through the spatial adjacency relationship between each component, and construct an adjacency matrix;

[0018] C2. Constructing a multi-component collaborative fault diagnosis model based on graph structure features: Constructing a multi-component dynamic coupling graph neural network model based on the adjacency matrix, including graph convolutional layers, fully connected layers, and Softmax classifier layers;

[0019] C3. Multi-component collaborative fault diagnosis: Based on dataset D, a model is trained to achieve collaborative fault diagnosis of multiple components in the electric shovel lifting mechanism, including the drive motor, disc brake, and gearbox.

[0020] Furthermore, the kinetic analysis described in step A1 specifically includes the following steps:

[0021] Dynamic analysis was performed on the transmission system of the electric shovel lifting mechanism, analyzing typical operating conditions of the drive motor, gearbox, and disc brake, including full-load lifting, no-load descent, and emergency braking. Dynamic models of each component were established. The drive motor was based on a rotor-bearing system using a mass-spring-damping model. The gearbox was described using a lumped parameter model to account for the time-varying meshing stiffness, tooth backlash, and bearing support stiffness of the gear pair. A contact dynamic model between the brake disc and friction pads was established for the disc brake. The components were integrated using multibody dynamics, and the motor torque curve, load spectrum, and braking pressure curve were applied as boundary conditions. The dynamic performance of the drive motor, gearbox, and disc brake was simulated and analyzed using numerical methods to obtain the displacement, velocity, and acceleration responses at key points.

[0022] Furthermore, the specific steps for selecting the vibration measurement surface in step A2 are as follows:

[0023] Based on the simulation results of step A1, the vibration measurement surfaces where the maximum amplitude of the drive motor, gearbox, and disc brake is located are extracted.

[0024] Furthermore, the specific steps for optimizing the deployment of vibration measuring points in step A3 are as follows:

[0025] For the drive motor, gearbox, and disc brake, vibration measuring points are installed in the horizontal and vertical directions of the vibration measuring surface, respectively. The collected signals are processed by filtering, Fourier transform, etc., to obtain the signal spectrum.

[0026] The effective vibration information percentage (EM) is defined by the following formula:

[0027] Among them, EM n This represents the percentage of effective vibration information at the nth measuring point. The average amplitude of the fault characteristic frequency is denoted as RMS, and the effective value of the signal is denoted as RMS.

[0028] Calculate the relative information weight P of each measuring point n The formula is as follows:

[0029] Where N is the total number of measuring points, P n To determine the relative information weights of the measuring points, the relative information weights of each measuring point on the drive motor, gearbox, and disc brake of the electric shovel lifting mechanism transmission system are calculated, respectively, as P. n-e P n-r P n-b The four sensors with the largest relative information weights on each vibration measurement surface—drive motor, gearbox, and disc brake—were selected as the installation locations for the vibration sensors.

[0030] Furthermore, the formula for information entropy in step A4 is:

[0031] In the formula, H represents information entropy. When the deployment of measuring points has a high information entropy, the monitoring system can extract richer and more effective information. When the total number of measuring points is fixed, the system has a definite statistical average entropy value, which can be expressed mathematically as follows:

[0032] like The measurement point deployment plan is approved; if Then, low-information-content weighted measurement points need to be removed or replaced, and the entropy value needs to be recalculated until the condition is met.

[0033] Furthermore, the steps for constructing the dataset in step B1 are as follows:

[0034] Based on the original data collected from the vibration measurement points of the drive motor and disc brake of the electric shovel lifting mechanism in the past six months, unnecessary and blank data were removed; data cleaning was performed on the original samples, including removing duplicate data, processing missing data and removing abnormal samples; by looking up the corresponding table of component faults, the normal, wear and fatigue failure of the drive motor bearings were linked with the measurement point data, and the fault data were manually labeled.

[0035] Furthermore, the calculation formula for the feature extractor in the individual fault diagnosis model in step B2 is as follows: H (1) =ET(x)=BatchNorm(Pooling(CNN(x)))

[0036] In the formula, ET represents the feature extraction function, CNN represents the convolutional layer, Pooling represents the pooling layer, BatchNorm represents the batch normalization layer, and H represents the H function. (1) This represents the feature extraction results based on the single-unit fault diagnosis feature extractor.

[0037] Furthermore, the adjacency matrix formula in step C1 is:

[0038] Furthermore, the characteristic feature is that the inter-layer propagation formula for graph convolution in step C3 is:

[0039] In the formula: A is the adjacency matrix, and I represents the identity matrix; represent The degree matrix of diagonal nodes; factors It is a normalized adjacency matrix; H (l) W represents the feature matrix of the l-th layer. (l)The matrix represents the weight matrix of the l-th layer; f(·) is the activation function;

[0040] The specific steps for multi-component collaborative fault diagnosis in step C3 are as follows:

[0041] The dataset obtained in step B1 is normalized, and the node feature extractor constructed in step B2 is used to reduce the dimensionality of the data. The dimensionality-reduced node feature data and the adjacency matrix constructed in step C1 are used as input data for the multi-component collaborative fault diagnosis model. The multi-component collaborative fault diagnosis model is trained by training the single-component fault feature extractor in step B2 and the multi-component collaborative fault diagnosis model in step C2. The multi-component collaborative fault diagnosis model of the electric shovel lifting mechanism transmission system is completed.

[0042] Furthermore, a fully connected layer is used to expand the output of the previous layer into a one-dimensional vector P, and a classifier based on a fully connected neural network is designed to output the fault classification result. U = Classifeier(Flatten(H) (l) ))

[0043] In the formula, Flatten represents the vector flattening operation. Classifeier represents a classifier model based on a fully connected network. U represents the classification result.

[0044] The loss function used is the cross-entropy loss function, as shown in the following formula:

[0045] In the formula, T is the sample size, u i p is the true label of the i-th sample. i It is the probability that the model predicts the i-th sample as positive.

[0046] The advantages of this application compared to existing technologies are:

[0047] 1. Multi-component collaborative fault diagnosis: This invention is no longer limited to a single component, but considers the relationship between components and the fault propagation process, realizing the joint fault diagnosis of multiple components such as gears, support structures, and couplings in the transmission system of the electric shovel lifting mechanism, which greatly improves the comprehensiveness and accuracy of the diagnosis.

[0048] 2. Efficient Measurement Point Deployment Strategy: This invention proposes a measurement point deployment strategy based on information entropy. By scientifically and rationally arranging measurement points, it fully utilizes effective sensor information and improves the utilization rate of sensor data. This strategy not only reduces measurement point costs but also helps improve the monitoring accuracy of electric shovels, providing more reliable data support for fault diagnosis.

[0049] 3. Intelligent and Automated: The system of this invention integrates multiple modules such as data acquisition, processing, analysis, and diagnosis, realizing intelligent and automated fault diagnosis. By monitoring and analyzing the operating status of the electric shovel lifting mechanism transmission system in real time, the system can promptly detect and warn of potential faults, providing strong support for maintenance decisions and reducing downtime and maintenance costs caused by faults. Attached Figure Description

[0050] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof.

[0051] Figure 1 shows the deployment of measuring points and fault diagnosis scheme for collaborative monitoring of multiple components of an electric shovel in the embodiment.

[0052] Figure 2 shows the measurement point deployment scheme of the transmission system of the electric shovel lifting mechanism in the embodiment.

[0053] Figure 3 illustrates the measurement point deployment strategy in this embodiment.

[0054] Figure 4 shows the fault diagnosis framework for multi-component collaboration in the embodiment.

[0055] In Figure 2, the following are labeled: 1. Gearbox support structure; 2. Gearbox gear; 3. Drive motor; 4. Coupling; 5. Lifting drum; 6. Disc brake. Detailed Implementation

[0056] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. Of course, the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0057] This invention provides a system-level fault diagnosis method for the transmission system of an electric shovel lifting mechanism, comprising the following steps:

[0058] A. The deployment strategy for measuring points in the transmission system of the electric shovel lifting mechanism specifically includes:

[0059] A1. Dynamics analysis of the transmission system of the electric shovel lifting mechanism: Dynamics analysis of the drive motor, lifting gearbox, and disc brake in the transmission system of the electric shovel lifting mechanism;

[0060] This step obtains the acceleration response spectrum characteristics of key components through numerical simulation, providing a quantitative basis for the deployment of measurement points and effectively avoiding the problem of missing feature information caused by traditional experience-based point placement.

[0061] A2. Selection of vibration measurement surface: Based on the results of dynamic analysis, the surface with the largest vibration amplitude is selected as the vibration measurement surface for arranging vibration sensors; since the surface with the largest amplitude concentrates the key modal characteristics in the energy transfer path of the system, arranging sensors at this location can significantly improve the signal-to-noise ratio of fault characteristics.

[0062] A3. Measurement point deployment scheme: Install vibration sensors in the horizontal and vertical directions of the selected vibration measurement surface, and select the optimal multiple measurement points as the installation positions of the vibration sensors based on the proportion of effective vibration information and the relative amount of effective information.

[0063] A4. Evaluation of Measurement Point Deployment Based on Information Entropy: Introducing information entropy to judge the rationality of vibration measurement schemes until the conditions are met;

[0064] A5. Sensor Data Acquisition: Vibration sensors are installed on the drive motor, lifting gearbox, and disc brake to acquire their vibration status; comprehensive collection of fault characteristics is achieved through standardized measurement point layout, providing high-quality input data for subsequent diagnostic models.

[0065] B. Construct fault diagnosis models for individual components, and build fault diagnosis networks for the drive motor, lifting gearbox, and disc brake, specifically including:

[0066] B1. Dataset Construction: Based on the measurement point deployment plan in step A, collect vibration data for various components such as the drive motor, lifting gearbox, and disc brake when they malfunction. The drive motor fault types include bearing wear and fatigue failure; disc vibrator faults include normal friction block wear, uneven wear, uniform wear, and spalling; and gearbox faults include gear tooth surface wear and spalling, gear tooth cracks, and broken teeth. Using the vibration data as input and the fault types as label data for output, create dataset D.

[0067] B2 Single Fault Feature Extraction: A single fault feature extractor is constructed by alternately stacking convolutional layers, pooling layers, and batch normalization layers;

[0068] C. Construct a system-level fault diagnosis model, incorporating fault diagnosis of key components into a unified framework, specifically including:

[0069] C1. Generate component spatial relationship diagram: Treat the drive motor, disc brake, and gearbox measurement points as nodes, generate a component spatial relationship diagram through the spatial adjacency relationship between each component, and construct an adjacency matrix;

[0070] C2. Constructing a multi-component collaborative fault diagnosis model based on graph structure features: Constructing a multi-component dynamic coupling graph neural network model based on the adjacency matrix, including graph convolutional layers, fully connected layers, and Softmax classifier layers;

[0071] C3. Multi-component collaborative fault diagnosis: Based on dataset D, a model is trained to achieve collaborative fault diagnosis of multiple components in the electric shovel lifting mechanism, including the drive motor, disc brake, and gearbox.

[0072] Specifically, the kinetic analysis in step A1 includes the following steps:

[0073] This invention first performs a dynamic analysis of the transmission system of the electric shovel lifting mechanism, analyzing typical operating conditions of the drive motor, gearbox, and disc brake, such as full-load lifting, no-load descent, and emergency braking. Next, dynamic models of each component are established. Specifically: for the drive motor, a rotor-bearing system based on a mass-spring-damping model is constructed; for the gearbox, a lumped parameter model describes the time-varying meshing stiffness, tooth backlash, and bearing support stiffness of the gear pair; for the disc brake, a contact dynamic model between the brake disc and the friction pads is established. Subsequently, multibody dynamics is used to integrate the components, and motor torque curves, load spectra, and braking pressure curves are applied as boundary conditions. Finally, numerical methods are used to simulate and analyze the dynamic performance of the drive motor, gearbox, and disc brake, obtaining the displacement, velocity, and acceleration responses at key points. This step provides crucial data support for the subsequent selection of vibration measurement surfaces.

[0074] In detail, the specific steps for selecting the vibration measuring surface in step A2 are as follows:

[0075] Based on the simulation results of step A1, the vibration measurement surfaces where the maximum amplitude of the drive motor, gearbox, and disc brake is located are extracted.

[0076] It should be noted that the specific steps for optimizing the deployment of vibration measuring points in step A3 are as follows:

[0077] For the drive motor, gearbox, and disc brake, vibration measuring points are installed in the horizontal and vertical directions of the vibration measuring surface, respectively. The collected signals are processed by filtering, Fourier transform, etc., to obtain the signal spectrum.

[0078] The effective vibration information percentage (EM) is defined by the following formula:

[0079] Among them, EM n This represents the percentage of effective vibration information at the nth measuring point. The average amplitude of the fault characteristic frequency is denoted as RMS, and the effective value of the signal is denoted as RMS.

[0080] Calculate the relative information weight P of each measuring point n The formula is as follows:

[0081] Where N is the total number of measuring points. P nLet P be the relative information weight of the measuring points. Calculate the relative information weight of each measuring point on the drive motor, gearbox, and disc brake of the electric shovel lifting mechanism's transmission system, respectively. n-e P n-r P n-b Furthermore, the four sensors with the highest relative information weight on each vibration measurement surface—the drive motor, gearbox, and disc brake—were selected as the installation locations for the vibration sensors.

[0082] Specifically, the formula for information entropy in step A4 is:

[0083] In the formula, H represents information entropy. When the deployment of measuring points has a high information entropy, the monitoring system can extract richer and more effective information. When the total number of measuring points is fixed, the system has a definite statistical average entropy value, which can be expressed mathematically as follows:

[0084] like The measurement point deployment plan is approved; if Then, low-information-content weighted measurement points need to be removed or replaced, and the entropy value needs to be recalculated until the condition is met.

[0085] Specifically, the steps for constructing the dataset in step B1 are as follows:

[0086] Based on the original data collected from the vibration measurement points of the drive motor and disc brake of the electric shovel lifting mechanism in the past six months, unnecessary and blank data were removed; data cleaning operations were performed on the original samples, such as removing duplicate data and processing missing data, to remove abnormal samples; by looking up the corresponding table list of component faults, the normal, wear and fatigue failure of the drive motor bearings were linked with the measurement point data, and the fault data were manually labeled.

[0087] It is worth noting that the calculation formula for the feature extractor in the individual fault diagnosis model in step B2 is as follows: H (1) =ET(x)=BatchNorm(Pooling(CNN(x)))

[0088] In the formula, ET represents the feature extraction function, CNN represents the convolutional layer, Pooling represents the pooling layer, and BatchNorm represents the batch normalization layer. (1) This represents the feature extraction results based on the single-unit fault diagnosis feature extractor.

[0089] It should be emphasized that the adjacency matrix formula in step C1 is:

[0090] To further explain, the inter-layer propagation formula for graph convolution in step C3 is:

[0091] In the formula: A is the adjacency matrix, and I represents the identity matrix; represent The degree matrix of diagonal nodes; factors It is a normalized adjacency matrix; H (l) W represents the feature matrix of the l-th layer. (l) The matrix represents the weight matrix of the l-th layer; f(·) is the activation function;

[0092] In addition, a fully connected layer is used to expand the output of the previous layer into a one-dimensional vector P, and a classifier based on a fully connected neural network is designed to output the fault classification results.

[0093] U = Classifeier(Flatten(H) (l) ))

[0094] In the formula, Flatten represents the vector flattening operation. Classifeier represents a classifier model based on a fully connected network. U represents the classification result.

[0095] The loss function used is the cross-entropy loss function, as shown in the following formula:

[0096] In the formula, T is the sample size, u i p is the true label of the i-th sample. i It is the probability that the model predicts the i-th sample as positive.

[0097] Specifically, the steps for multi-component collaborative fault diagnosis in step C3 are as follows:

[0098] The dataset obtained in B1 is normalized, and the node feature extractor constructed in step B2 is used to reduce the dimensionality of the data. The dimensionality-reduced node feature data and the adjacency matrix constructed in step C1 are used as input data for the multi-component collaborative fault diagnosis model. The multi-component collaborative fault diagnosis model is trained by training the single-component fault feature extractor in B2 and the multi-component collaborative fault diagnosis model in C2. The multi-component collaborative fault diagnosis model of the electric shovel lifting mechanism transmission system is completed.

[0099] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for deploying measuring points and diagnosing faults in the collaborative monitoring of multiple components of an electric shovel, characterized in that, Includes the following steps: A. The deployment strategy for measuring points in the transmission system of the electric shovel lifting mechanism specifically includes: A1. Dynamics analysis of the transmission system of the electric shovel lifting mechanism: Dynamics analysis of the drive motor, lifting gearbox, and disc brake in the transmission system of the electric shovel lifting mechanism; A2. Selection of vibration measurement surface: Based on the results of dynamic analysis, the surface with the largest vibration amplitude is selected as the vibration measurement surface for arranging vibration sensors; A3. Measurement point deployment scheme: Install vibration sensors in the horizontal and vertical directions of the selected vibration measurement surface, and select the optimal multiple measurement points as the installation positions of the vibration sensors based on the proportion of effective vibration information and the relative amount of effective information. A4. Evaluation of Measurement Point Deployment Based on Information Entropy: Introducing information entropy to judge the rationality of vibration measurement schemes until the conditions are met; A5. Sensor data acquisition: Vibration sensors are installed on the drive motor, lifting gearbox, and disc brake to acquire their vibration status; B. Construct fault diagnosis models for individual components, and build fault diagnosis networks for the drive motor, lifting gearbox, and disc brake, specifically including: B1. Dataset Construction: Based on the measurement point deployment plan in step A, collect vibration data when various components such as the drive motor, lifting gearbox, and disc brake malfunction. The fault types of the drive motor include bearing wear and fatigue failure; the fault types of the disc vibrator include normal friction block wear, uneven wear, uniform wear, and spalling; the fault types of the gearbox include gear tooth surface wear and spalling, gear tooth cracks, and broken teeth. Use the vibration data as input and the fault types as label data as output to create dataset D. B2 Single Fault Feature Extraction: A single fault feature extractor is constructed by alternately stacking convolutional layers, pooling layers, and batch normalization layers; C. Construct a system-level fault diagnosis model, incorporating fault diagnosis of key components into a unified framework, specifically including: C1. Generate component spatial relationship diagram: Treat the drive motor, disc brake, and gearbox measurement points as nodes, generate a component spatial relationship diagram through the spatial adjacency relationship between each component, and construct an adjacency matrix; C2. Constructing a multi-component collaborative fault diagnosis model based on graph structure features: Constructing a multi-component dynamic coupling graph neural network model based on the adjacency matrix, including graph convolutional layers, fully connected layers, and Softmax classifier layers; C3. Multi-component collaborative fault diagnosis: Based on dataset D, a model is trained to achieve collaborative fault diagnosis of multiple components in the electric shovel lifting mechanism, including the drive motor, disc brake, and gearbox.

2. The method for deploying measuring points and diagnosing faults in collaborative monitoring of multiple components of an electric shovel, as described in claim 1, is characterized in that... The kinetic analysis described in step A1 specifically includes the following steps: Dynamic analysis was performed on the transmission system of the electric shovel lifting mechanism, analyzing typical operating conditions of the drive motor, gearbox, and disc brake, including full-load lifting, no-load descent, and emergency braking. Dynamic models of each component were established. The drive motor was based on a rotor-bearing system using a mass-spring-damping model. The gearbox was described using a lumped parameter model to account for the time-varying meshing stiffness, tooth backlash, and bearing support stiffness of the gear pair. A contact dynamic model between the brake disc and friction pads was established for the disc brake. The components were integrated using multibody dynamics, and the motor torque curve, load spectrum, and braking pressure curve were applied as boundary conditions. The dynamic performance of the drive motor, gearbox, and disc brake was simulated and analyzed using numerical methods to obtain the displacement, velocity, and acceleration responses at key points.

3. The method for deploying measuring points and diagnosing faults in collaborative monitoring of multiple components of an electric shovel, as described in claim 1, is characterized in that... The specific steps for selecting the vibration measurement surface in step A2 are as follows: Based on the simulation results of step A1, the vibration measurement surfaces where the maximum amplitude of the drive motor, gearbox, and disc brake is located are extracted.

4. The method for deploying measuring points and diagnosing faults in a multi-component collaborative monitoring system for electric shovels according to claim 1, characterized in that, The specific steps for optimizing the deployment of vibration measuring points in step A3 are as follows: For the drive motor, gearbox, and disc brake, vibration measuring points are installed in the horizontal and vertical directions of the vibration measuring surface, respectively. The collected signals are processed by filtering, Fourier transform, etc., to obtain the signal spectrum. The effective vibration information percentage (EM) is defined by the following formula: Among them, EM n This represents the percentage of effective vibration information at the nth measuring point. The average amplitude of the fault characteristic frequency is denoted as RMS, and the effective value of the signal is denoted as RMS. Calculate the relative information weight P of each measuring point n The formula is as follows: Where N is the total number of measuring points, P n To determine the relative information weights of the measuring points, the relative information weights of each measuring point on the drive motor, gearbox, and disc brake of the electric shovel lifting mechanism transmission system are calculated, respectively, as P. n-e P n-r P n-b The four sensors with the largest relative information weights on each vibration measurement surface—drive motor, gearbox, and disc brake—were selected as the installation locations for the vibration sensors.

5. The method for deploying measuring points and diagnosing faults in collaborative monitoring of multiple components of an electric shovel according to claim 4, characterized in that, The formula for information entropy in step A4 is: In the formula, H represents information entropy. When the deployment of measuring points has a high information entropy, the monitoring system can extract richer and more effective information. When the total number of measuring points is fixed, the system has a definite statistical average entropy value, which can be expressed mathematically as follows: like The measurement point deployment plan is approved; if Then, low-information-content weighted measurement points need to be removed or replaced, and the entropy value needs to be recalculated until the condition is met.

6. The method for deploying measuring points and diagnosing faults in collaborative monitoring of multiple components of an electric shovel according to claim 1, characterized in that, The steps for constructing the dataset in step B1 are as follows: Based on the original data collected from the vibration measurement points of the drive motor and disc brake of the electric shovel lifting mechanism in the past six months, unnecessary and blank data were removed; data cleaning was performed on the original samples, including removing duplicate data, processing missing data and removing abnormal samples; by looking up the corresponding table of component faults, the normal, wear and fatigue failure of the drive motor bearings were linked with the measurement point data, and the fault data were manually labeled.

7. The method for deploying measuring points and diagnosing faults in collaborative monitoring of multiple components of an electric shovel according to claim 1, characterized in that, The calculation formula for the feature extractor in the single-unit fault diagnosis model in step B2 is as follows: H (1) =ET(x)=BatchNorm(Pooling(CNN(x))) In the formula, ET represents the feature extraction function, CNN represents the convolutional layer, Pooling represents the pooling layer, BatchNorm represents the batch normalization layer, and H represents the H function. (1) This represents the feature extraction results based on the single-unit fault diagnosis feature extractor.

8. The method for deploying measuring points and diagnosing faults in collaborative monitoring of multiple components of an electric shovel according to claim 1, characterized in that, The adjacency matrix formula in step C1 is:

9. The method for deploying measuring points and diagnosing faults in collaborative monitoring of multiple components of an electric shovel according to claim 8, characterized in that, The inter-layer propagation formula for graph convolution in step C3 is: In the formula: A is the adjacency matrix, and I represents the identity matrix; represent The degree matrix of diagonal nodes; factors It is a normalized adjacency matrix; H (l) W represents the feature matrix of the l-th layer. (l) The matrix represents the weight matrix of the l-th layer; f(·) is the activation function; The specific steps for multi-component collaborative fault diagnosis in step C3 are as follows: The dataset obtained in step B1 is normalized, and the node feature extractor constructed in step B2 is used to reduce the dimensionality of the data. The dimensionality-reduced node feature data and the adjacency matrix constructed in step C1 are used as input data for the multi-component collaborative fault diagnosis model. The multi-component collaborative fault diagnosis model is trained by training the single-component fault feature extractor in step B2 and the multi-component collaborative fault diagnosis model in step C2. The multi-component collaborative fault diagnosis model of the electric shovel lifting mechanism transmission system is completed.

10. [Corrected according to detailed rule 91, 03.07.2025] A method for deploying measuring points and diagnosing faults in collaborative monitoring of multiple components of an electric shovel, as described in claim 1, is characterized in that... A fully connected layer is used to expand the output of the previous layer into a one-dimensional vector P, and a classifier based on a fully connected neural network is designed to output the fault classification result. U = Classifeier(Flatten(H) (l) )) In the formula, Flatten represents the vector flattening operation, Classifeier represents the classifier model based on a fully connected network, and U represents the classification result; The loss function used is the cross-entropy loss function, as shown in the following formula: In the formula, T is the sample size, u i p is the true label of the i-th sample. i It is the probability that the model predicts the i-th sample as positive.

Images