Elevator traction machine fault diagnosis algorithm based on multi-source information fusion

Through the multi-source information fusion algorithm, combined with adaptive filters and empirical modal decomposition technology, multi-scale entropy and principal component analysis, and combined with support vector machine for fault diagnosis, the limitations of single signal direction diagnosis are solved and the elevator drag is improved The accuracy and efficiency of induction machine fault diagnosis enhance the scientific nature of fault prevention and maintenance.

CN119059386APending Publication Date: 2024-12-03SHANDONG JIANZHU UNIV
View PDF 0 Cites 6 Cited by

Patent Information

Application Number
CN202411566414.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-11-05
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

The prior art is difficult to diagnose faults in a single vibration signal direction, especially under complex faults and environmental noise interference, which is difficult to fully reflect the operating status of the elevator traction machine, and a single sensor cannot effectively distinguish the fault types.

Method used

Using multi-source information fusion algorithm, data is collected through vibration sensors in multiple directions, improved adaptive filter denoising, empirical mode decomposition, multi-scale entropy calculation and principal component analysis, and pattern recognition and classification are combined with support vector machines. , form comprehensive health indicators.

Benefits of technology

It significantly improves the accuracy and efficiency of fault diagnosis of elevator traction machines, enhances the scientificity and practicality of fault prevention and maintenance, and can more accurately identify high-frequency and low-frequency fault characteristics, reducing the impact of noise interference.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119059386A_ABST
    Figure CN119059386A_ABST
Patent Text Reader

Abstract

The invention discloses an elevator traction machine fault diagnosis algorithm based on multi-source information fusion, and the algorithm comprises the following steps: collecting the vibration signal data of a multi-source sensor under a specific fault and a normal working condition, and carrying out the denoising processing through an improved adaptive filter, so as to improve the data quality; decomposing the denoised vibration signal into a plurality of intrinsic mode functions by using an empirical mode decomposition technology; the method comprises the following steps: constructing a multi-scale time sequence, calculating a sample entropy of each scale, and obtaining a multi-scale entropy vector of a single intrinsic mode function; on this basis, calculating a covariance matrix of a multi-scale entropy vector, extracting principal components, and selecting the first several principal components with the highest contribution rate as a fusion result; and inputting the tractor signal data with the label into a support vector machine, carrying out supervised learning training, and carrying out pattern recognition and classification on the fused signal data by utilizing a trained model. Through effective fusion of multi-source information, the precision and efficiency of fault diagnosis of the elevator traction machine are remarkably improved, and the method has wide application prospects and important practical significance.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the field of equipment fault diagnosis, and in particular relates to an elevator traction machine fault diagnosis algorithm based on multi-source information fusion. Background Art

[0002] The elevator traction machine is the core power transmission equipment in the elevator system and is widely used in high-rise residential buildings, commercial buildings, industrial facilities and other places. The traction machine is usually under high load and frequently starts and stops. The occurrence of a fault usually causes the elevator to stop operating, passengers to be stranded, and creates safety hazards and high maintenance costs. Therefore, the fault diagnosis of the elevator traction machine is particularly important.

[0003] As the core power component of the elevator system, the traction machine usually shows slight changes in its operating state through vibration signals. When a fault occurs, the waveform of the vibration signal in the time domain is difficult to clearly identify the operating state of the mechanical system. Through signal feature extraction technology, characteristic parameters that reveal the abnormal conditions of mechanical components can be extracted. The vibration signal transmitted through the base contains vibration characteristics of multiple parts from the motor, transmission system to the bearing. The vibration signal of the traction machine can be divided into three directions: horizontal radial, vertical radial and axial. The horizontal vibration is affected by the traction machine drive motor, coupling and other components, reflecting the unbalanced movement of the traction machine's transmission system, bearings and other components or the alignment problem of the transmission system. The vertical vibration signal is affected by the force conditions of the traction machine bearing and base, which can help identify problems such as bearing wear, bearing seat abnormality or uneven load. The axial vibration signal mainly involves problems such as bearing and axis centerline alignment, and can reflect the relative movement of the inner and outer rings of the bearing, the bending or offset of the shaft, etc.

[0004] However, due to the limitations of the vibration signal itself, the vibration signal in a single direction can often only reveal a certain component or a certain type of fault, and is easily disturbed by factors such as environmental noise and vibration wave propagation path, making it difficult to fully reflect the operating status of the entire traction machine system. In addition, some complex faults may manifest similarly in different directions, and a single sensor cannot effectively distinguish these types of faults. In order to overcome these limitations, vibration sensors in multiple directions need to be used in conjunction, and a multi-source information fusion algorithm is used to integrate multi-dimensional vibration data from different sensors. Through data processing technology, the vibration signals in each direction are deeply analyzed and comprehensively evaluated, and the characteristic parameters of the signals in each direction are integrated into a comprehensive health indicator to provide an accurate judgment of the overall condition of the traction machine.

[0005] In summary, the elevator traction machine fault diagnosis algorithm based on multi-source information fusion not only improves the accuracy and efficiency of elevator traction machine fault diagnosis, but also greatly enhances the scientificity and practicality of fault prevention and maintenance strategies by realizing more complex data processing and intelligent decision-making. Summary of the invention

[0006] In view of the deficiencies in the prior art, the present invention integrates the signal data of multi-source sensors and utilizes a multi-source information fusion strategy to integrate the monitoring data features of various sensors, thereby supplementing the information omissions that may occur during the fault detection process when relying on a single data source.

[0007] The present invention discloses an elevator traction machine fault diagnosis algorithm based on multi-source information fusion, the method comprising:

[0008] S1, collecting vibration signal data of multi-source sensors under specific fault conditions and normal working conditions and using improved adaptive filters for denoising;

[0009] S2, using empirical mode decomposition on the denoised data to decompose the vibration signal in each direction into multiple intrinsic mode functions;

[0010] S3, by constructing a multi-scale time series, calculating the sample entropy of each scale, and calculating the multi-scale entropy vector of a single intrinsic mode function;

[0011] S4. Calculate the covariance matrix of the multi-scale entropy vector, extract the principal components, select the first few principal components with the highest contribution rate as the fusion result, fuse the intrinsic mode functions in different directions, and linearly combine the extracted principal components to form a new fusion vector.

[0012] S5. Input the labeled traction machine signal data into the support vector machine, use supervised learning for training, and use the trained model to perform pattern recognition and classification on the fused signal data.

[0013] Furthermore, the elevator traction machine multi-source sensor data includes: a vertical radial vibration signal of the base, a horizontal radial vibration signal of the base, and an axial vibration signal of the base.

[0014] Furthermore, the specific faults include bearing damage, magnetic steel demagnetization, encoder failure, and control system abnormality.

[0015] Furthermore, the use of an adaptive filter to remove noise is specifically as follows: under the real-time input of the error signal, the recursive least squares algorithm is applied to dynamically adjust the weight of the filter, the algorithm parameters are automatically optimized by real-time monitoring and analysis of the change trend and performance indicators of the signal, and the iteration is terminated by setting the number of iterations or the condition that the error signal drops below a predetermined threshold. The filtering process can be expressed as:

[0016]

[0017] in, is the coefficient vector of the filter, is the output vector of the filter. is the input vector to the filter.

[0018] Furthermore, the use of empirical mode decomposition to decompose the vibration signal in each direction into multiple intrinsic mode functions is specifically as follows: First, the local maximum and minimum values ​​of the denoised signal are identified. The maximum points are used for cubic spline interpolation to construct the upper envelope, and the minimum points are used to construct the lower envelope. Next, the mean of the upper envelope and the lower envelope is calculated to obtain the mean envelope to form a candidate intrinsic mode function. This candidate intrinsic mode function must meet the following conditions: first, in the entire data set, the number of zero-crossing points and the number of extreme points are equal or differ by at most one; second, at any point, the mean of the upper envelope and the lower envelope is zero. If not satisfied, this candidate function is regarded as new data and the above process is repeated.

[0019] After several iterations, if the intrinsic mode function conditions are met, the first-order intrinsic mode function component is obtained. . Subtract it from the original signal to form a new signal , and repeat the same process on the new signal to extract the second intrinsic mode function component , this process continues until the preset order is reached or the residual signal is smaller than the set threshold.

[0020] Finally, the original signal can be expressed as a superposition of multiple intrinsic mode functions through empirical mode decomposition, namely:

[0021]

[0022] in, is the residual signal.

[0023] Furthermore, we calculate the multi-scale entropy vector of a single intrinsic mode function. First, we need to construct a multi-scale time series. , construct its multi-scale version. At each scale, the sample entropy is used to quantify the complexity of the time series. Lower sample entropy indicates that the time series has higher predictability. The final multi-scale entropy is the collection of sample entropies at different scales.

[0024] Furthermore, principal component analysis is used to reduce the dimension of high-dimensional entropy data while retaining the most critical information for fusing intrinsic mode functions in different directions. The multi-scale entropy vectors corresponding to the intrinsic mode functions of the same sequence number k in each direction are used as the input of principal component analysis. The covariance matrix of these multi-scale entropy vectors is calculated to extract the principal components. The first few principal components with the highest contribution rate are selected as the fusion result.

[0025] Furthermore, the fault mode is identified and classified by using support vector machine. The fusion vector and the corresponding fault type label are used as data. The dataset is randomly divided into training set, validation set and test set in a ratio of 3:1:1.

[0026] The goal of a support vector machine is to find a hyperplane that maximizes the margin between different classes. The best classification hyperplane is determined by solving the following optimization problem:

[0027]

[0028] in, is the normal vector of the hyperplane, is the bias term, is the regularization parameter, is a slack variable used to handle nonlinear separability. The constraints are:

[0029]

[0030] in, It is a sample Tags, It is a sample The feature vector of .

[0031] Furthermore, an elevator traction machine fault diagnosis algorithm based on multi-source information fusion includes: a data acquisition module, a signal processing module, a feature extraction module, a data fusion module, and a classification and diagnosis module.

[0032] The data acquisition module is used to collect vibration signal data of the elevator traction machine under specific fault conditions and normal working conditions from multi-source sensors, and use an improved adaptive filter to remove noise;

[0033] The signal processing module is used to decompose the vibration signal in each direction using the empirical mode decomposition technology on the denoised data to obtain multiple intrinsic mode functions;

[0034] The feature extraction module is used to calculate the sample entropy of each scale by constructing a multi-scale time series, and calculate a multi-scale entropy vector for each intrinsic mode function;

[0035] The data fusion module is used to calculate the covariance matrix of the multi-scale entropy vector, extract the main components through principal component analysis, select several principal components with the highest contribution rate, and fuse them into fusion vectors from the intrinsic mode functions in different directions;

[0036] The data classification and diagnosis module is used to input the fused signal data with labels into the support vector machine for supervised learning training, and use the trained model to perform pattern recognition and classification on new fused signal data.

[0037] According to the third aspect of the present specification, there is provided an elevator traction machine fault diagnosis operation device based on multi-source information fusion, comprising a memory and one or more processors, wherein the memory stores executable code, and when the processor executes the executable code, the elevator traction machine fault diagnosis algorithm based on multi-source information fusion is implemented.

[0038] The beneficial effect of the technical solution provided by the present invention is that the present invention realizes an elevator traction machine fault diagnosis algorithm based on multi-source information fusion, which effectively solves the limitations of relying on a single signal source in fault detection, which is single and susceptible to noise, and the problem of high risk of error propagation in multi-source information fusion. By combining the vibration signal data of multiple sensors and using improved adaptive filters and empirical mode decomposition techniques, the present invention can extract more detailed intrinsic mode functions in all directions. Further, by constructing a multi-scale time series and calculating its sample entropy, the present invention not only enhances the model's ability to identify high-frequency and low-frequency fault features, but also optimizes the integration of each signal feature through principal component analysis, significantly improving the decision-making accuracy and generalization ability of the diagnostic system. Through the comprehensive analysis of multi-dimensional signals and feature fusion strategies, the ability to accurately diagnose traction machine transmission system faults is enhanced, which has important theoretical and practical application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 This is a flow chart of an elevator traction machine fault diagnosis algorithm based on multi-source information fusion;

[0040] Figure 2 It is the principle diagram of adaptive filter;

[0041] Figure 3 This is the effect diagram of preprocessing the vibration signal;

[0042] Figure 4 This is the flow chart of the empirical mode decomposition algorithm;

[0043] Figure 5 This is the result diagram of empirical mode decomposition;

[0044] Figure 6 A schematic diagram of an elevator traction machine fault diagnosis algorithm operation device based on multi-source information fusion provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0045] In order to make the objectives, technical solutions and advantages of the present invention more clear, the embodiments of the present invention are described in further detail below.

[0046] like Figure 1As shown, this embodiment proposes an elevator traction machine fault diagnosis algorithm based on multi-source information fusion, which is used to realize the fault diagnosis of the elevator traction machine, and solves the problem that it is difficult to accurately identify the fault type by monitoring a single sensor signal processing in engineering practice. See the following description for details:

[0047] S1, such as Figure 2 As shown, the signal denoising algorithm:

[0048] This embodiment proposes an adaptive filter denoising algorithm. An adaptive filter is capable of dynamically adjusting parameters according to the characteristics of the input signal. Compared with traditional filters, adaptive filters can adapt to environmental changes in real time and are particularly suitable for processing non-stationary signals of elevator traction machines. The core idea is to optimize the output of the signal to track the statistical characteristics of the changes by continuously updating the coefficients of the filter. In each iteration, the least squares problem is solved recursively using the sampled values ​​of the input signal. This process uses an exponentially weighted recursive least squares algorithm to update the filter coefficients by minimizing the expected value of the square of the error, and its cost function is as follows:

[0049]

[0050] in, is the cost function, which means The weighted sum of squared errors at iterations. This function measures the difference between the filter output and the desired signal. is the forgetting factor. The algorithm applies greater weight to newer errors and smaller weight to older errors, which helps to ignore old data, enables the filter to adapt to non-stationary environments, and better track the statistical characteristics of current data. Indicated in The target signal at the iteration is the ideal signal that the filter needs to output. Represents the filter coefficient vector, which is calculated in each iteration according to the cost function Updated, it contains all the parameters of the filter in its current state. A feature vector representing the input signal, typically including current and past input signal samples. Indicated in The error at the iteration is defined as the difference between the expected output and the current output of the filter, expressed as:

[0051] S2, such as Figure 4 As shown, the empirical mode decomposition algorithm:

[0052] Empirical mode decomposition decomposes the vibration signal in each direction into multiple intrinsic mode functions. Specifically, all local maxima and minima are identified in the denoised signal. The maximum points are used for interpolation to form an upper envelope:

[0053]

[0054] in, is the original signal, express The The time position of the maximum point, is a local maximum. represents the cubic spline fitting function passing through the maximum point, is the number of maximum points.

[0055] Find all the minimum points: Similarly, use the minimum points to interpolate and form a lower envelope:

[0056]

[0057] in represents the cubic spline fitting function passing through the minimum point, is the number of minimum points.

[0058] Calculate the mean of the upper envelope and the lower envelope to get the mean envelope .

[0059]

[0060] Subtract the mean envelope from the original signal sequence , and obtain a candidate intrinsic mode function .

[0061]

[0062] This candidate intrinsic mode function must meet two conditions: first, in the entire data set, the number of zero-crossing points and the number of extreme points must be equal or differ by at most one; second, at any point, the mean of the upper envelope and the lower envelope is zero. If the candidate intrinsic mode function does not meet the above conditions, it is regarded as new data and the calculation is repeated. Assume that after After that, If the conditions of the intrinsic mode function are met, then the original signal The first-order intrinsic mode function component of is :

[0063]

[0064] Use original signal minus , and get a new signal without high-frequency components :

[0065]

[0066] right Repeated The second intrinsic mode function component is obtained by , and repeat this process until the The intrinsic mode function components or the remainder Less than the preset value; or when the residual When is a monotonic function or a constant, the EMD process stops.

[0067] Finally, the original signal can be expressed as follows after empirical mode decomposition:

[0068]

[0069] S3, multi-scale entropy vector calculation method:

[0070] To calculate the multi-scale entropy vector of a single intrinsic mode function, we first need to construct a multi-scale time series. , construct its multi-scale version. The multi-scale time series is realized by coarsening the original time series. The specific formula is as follows:

[0071]

[0072] in, is the scale factor, is in scale The next data points, It is the original IMF data points.

[0073] At every scale Next, we use sample entropy to quantify the time series The complexity of the sample entropy is lower. The lower the sample entropy, the more predictable the time series is. In The length of the template vector

[0074] For each vector , calculated with all other vectors The Euclidean distance of is defined as the distance less than the threshold The vector pairs of are similar. , calculated in and The ratio of similar vectors under dimension is denoted as and .

[0075] The calculation formula of sample entropy is:

[0076]

[0077] in yes Length, is the length of the template vector, is the similarity threshold.

[0078] The final multiscale entropy is at different scales Sample entropy under can be used to describe the complexity and information content of intrinsic mode functions on multiple time scales.

[0079] S4, principal component analysis algorithm:

[0080] The high-dimensional multi-scale entropy data is reduced in dimension through principal component analysis to retain key information for fusion of IMFs in different directions. The multi-scale entropy value vector corresponding to the IMF in different directions is used as the input of principal component analysis. For all directions The matrix composed of multi-scale entropy value vectors of IMF is as follows:

[0081]

[0082] in, Indicates In all directions The multi-scale entropy value of the IMF. Calculation The covariance matrix of is:

[0083]

[0084] in, yes The covariance matrix of yes The transpose of is the total number of directions.

[0085] calculate According to the eigenvalue size, select the first few principal components with the highest contribution rate.

[0086]

[0087] in, It is The IMF corresponds to principal components. yes No. feature vectors.

[0088]

[0089] S5, Support Vector Machine:

[0090] Furthermore, the fault mode is identified and classified by using support vector machine. First, the fusion vector obtained by S4 and the corresponding fault type label are used as data. The data set is randomly divided into training set, validation set and test set in a ratio of 3:1:1.

[0091] The goal of SVM is to find a hyperplane that maximizes the margin between different classes. The best classification hyperplane is determined by solving the following optimization problem:

[0092]

[0093] in, is the normal vector of the hyperplane, is the bias term, is the regularization parameter, is a slack variable used to handle nonlinear separability. The constraints are:

[0094]

[0095] in, It is a sample Tags, It is a sample The feature vector of .

[0096] The training data set is used to train the SVM model, and the trained SVM model is used to classify the fused signal data and predict the label of each test sample. Finally, the accuracy, precision, recall, and F1 score of the model are evaluated by comparing the model prediction results with the actual labels.

[0097] The intelligent fault diagnosis method based on support vector machine proposed in the present invention is applicable to elevator traction machines based on multi-source information fusion. The method aims to achieve high-precision identification of transmission system faults through effective feature extraction and classification.

[0098] During the training process, supervised learning techniques are first used to extract features from the input data to minimize the difference between the input data and the reconstructed output and construct a feature space. The training set consists of labeled samples, 60% of which are divided into a training set, 20% as a validation set, and the remaining 20% ​​as a test set.

[0099] In the process of model validation, the validation set is used to evaluate the model performance, mainly through cross-validation of indicators such as accuracy, recall, precision and F1 score. Through the feedback of these indicators, the model parameters (such as learning rate, kernel function type and its parameters, number of support vectors, etc.) are tuned to improve the classification performance.

[0100] The tuned support vector machine diagnosis model is used to perform fault diagnosis on the fused feature vector. When the output of the model does not meet the accuracy requirements, repeated training is required to improve its performance. Specifically, the training set can be further subdivided into several training subsets, and multiple support vector machines are trained in parallel. After the training is completed, the diagnostic result with the largest number of repeated results is selected through a voting mechanism, thereby improving the accuracy and reliability of fault detection.

[0101] In summary, the elevator traction machine fault diagnosis algorithm based on multi-source information fusion provided by the present invention successfully overcomes the overfitting problem caused by traditional monitoring of single signal processing, improves the accuracy of fault type recognition, and demonstrates superior performance and broad application prospects.

Claims

1. An elevator traction machine fault diagnosis algorithm based on multi-source information fusion, characterized in that: The method includes: S1, collecting vibration signal data of multi-source sensors under specific fault conditions and normal working conditions and using improved adaptive filters for denoising; S2, using empirical mode decomposition on the denoised data to decompose the vibration signal in each direction into multiple intrinsic mode functions; S3, by constructing a multi-scale time series, calculating the sample entropy of each scale, and calculating the multi-scale entropy vector of a single intrinsic mode function; S4, calculate the covariance matrix of the multi-scale entropy vector, extract the principal components, select the first few principal components with the highest contribution rate as the fusion result, fuse the intrinsic mode functions in different directions, and linearly combine the extracted principal components to form a new fusion vector; S5. Input the labeled traction machine signal data into the support vector machine, use supervised learning for training, and use the trained model to perform pattern recognition and classification on the fused signal data.

2. The elevator traction machine fault diagnosis algorithm based on multi-source information fusion according to claim 1 is characterized in that: The vibration signal data of the elevator traction machine multi-source sensor include: a vertical radial vibration signal of the base, a horizontal radial vibration signal of the base, and an axial vibration signal of the base.

3. The method for traction machine transmission system fault diagnosis based on multi-source information fusion according to claim 1 is characterized in that: The specific faults include bearing damage, magnetic steel demagnetization, encoder failure, and control system abnormality.

4. The elevator traction machine fault diagnosis algorithm based on multi-source information fusion according to claim 1 is characterized in that: In S1, the use of an adaptive filter to remove noise is specifically as follows: under the real-time input of the error signal, the recursive least squares algorithm is applied to dynamically adjust the weight of the filter, and the algorithm parameters are automatically optimized by real-time monitoring and analyzing the change trend and performance indicators of the signal, and the iteration is terminated by setting the number of iterations or the condition that the error signal drops below a predetermined threshold. The filtering process can be expressed as: , in, is the coefficient vector of the filter, is the output vector of the filter. is the input vector to the filter.

5. The method for traction machine transmission system fault diagnosis based on multi-source information fusion according to claim 1, characterized in that: In S2, empirical mode decomposition is used to decompose the vibration signal in each direction into multiple intrinsic mode functions. Specifically, first, the local maximum and minimum values ​​of the denoised signal are identified. The maximum points are used for cubic spline interpolation to construct the upper envelope, and the minimum points are used to construct the lower envelope. Next, the mean of the upper envelope and the lower envelope is calculated to obtain the mean envelope to form a candidate intrinsic mode function. This candidate intrinsic mode function must meet the following conditions: first, in the entire data set, the number of zero-crossing points and the number of extreme points are equal or differ by at most one; second, at any point, the mean of the upper envelope and the lower envelope is zero. If not satisfied, the candidate function is treated as new data and the above process is repeated. After several iterations, if the intrinsic mode function conditions are met, the first-order intrinsic mode function component is obtained. . Subtract it from the original signal to form a new signal , and repeat the same process on the new signal to extract the second intrinsic mode function component This process continues until the preset order is reached or the residual signal is less than the set threshold. Finally, the original signal can be expressed as a superposition of multiple intrinsic mode functions through empirical mode decomposition, that is: , in, is the residual signal.

6. The method for traction machine transmission system fault diagnosis based on multi-source information fusion according to claim 1, characterized in that: In S3, the multi-scale entropy vector of a single intrinsic mode function is calculated. First, a multi-scale time series needs to be constructed. For the original , construct its multi-scale version. At each scale, the sample entropy is used to quantify the complexity of the time series. Lower sample entropy indicates that the time series has higher predictability. The final multi-scale entropy is the collection of sample entropies at different scales.

7. The method for traction machine transmission system fault diagnosis based on multi-source information fusion according to claim 1, characterized in that: In S4, the high-dimensional entropy data is reduced in dimension by principal component analysis, while retaining the most critical information for fusing the intrinsic mode functions in different directions. The multi-scale entropy vectors corresponding to the intrinsic mode functions of the same sequence number k in each direction are used as the input of the principal component analysis. The covariance matrix of these multi-scale entropy vectors is calculated to extract the principal components. The first few principal components with the highest contribution rate are selected as the fusion result.

8. The method for traction machine transmission system fault diagnosis based on multi-source information fusion according to claim 1 is characterized in that: In S5, the fault mode is identified and classified by using a support vector machine. The fusion vector and the corresponding fault type label are used as data. The data set is randomly divided into a training set, a validation set, and a test set in a ratio of 3:1:

1. The goal of the support vector machine is to find a hyperplane to maximize the interval between different categories. The best classification hyperplane is determined by solving the following optimization problem: , in, is the normal vector of the hyperplane, is the bias term, is the regularization parameter, is a slack variable used to handle nonlinear separability. The constraints are: , in, It is a sample Tags, It is a sample The feature vector of .

9. An algorithm for implementing the method according to any one of claims 1 to 8, characterized in that: The algorithm includes: data acquisition module, signal processing module, feature extraction module, data fusion module, classification and diagnosis module; The data acquisition module is used to collect vibration signal data of the elevator traction machine under specific fault conditions and normal working conditions from multi-source sensors, and use an improved adaptive filter to remove noise; The signal processing module is used to decompose the vibration signal in each direction using the empirical mode decomposition technology on the denoised data to obtain multiple intrinsic mode functions; The feature extraction module is used to calculate the sample entropy of each scale by constructing a multi-scale time series, and calculate a multi-scale entropy vector for each intrinsic mode function; The data fusion module is used to calculate the covariance matrix of the multi-scale entropy vector, extract the main components through principal component analysis, select several principal components with the highest contribution rate, and fuse them into fusion vectors from the intrinsic mode functions in different directions; The data classification and diagnosis module is used to input the fused signal data with labels into the support vector machine for supervised learning training, and use the trained model to perform pattern recognition and classification on new fused signal data.

10. An elevator traction machine fault diagnosis operation device based on multi-source information fusion, comprising a memory and one or more processors, wherein the memory stores executable code, characterized in that: When the processor executes the executable code, an elevator traction machine fault diagnosis algorithm based on multi-source information fusion as described in any one of claims 1-8 is implemented.

Citation Information

Cited By

  • Numerical control machine tool fault analysis method and terminal

    CN119717687A

  • Aviation converter fault feature extraction and diagnosis method based on multi-source information

    CN119740195A

  • Fault Feature Extraction and Diagnosis Method for Aero-Converter Based on Multi-Source Information

    CN119740195B

  • Unmanned aerial vehicle positioning method and device based on multi-modal fusion, and medium

    CN119919499A

  • Fault detection method, device and equipment based on mechanical vibration and medium

    CN120429836A