Fault diagnosis method and system for elevator driving host

Abstract

The invention discloses a fault diagnosis method and system for an elevator driving host, and relates to the technical field of fault diagnosis of the elevator driving host, and the method comprises the steps: obtaining the operation data of a to-be-detected elevator driving host in various working states, and extracting characteristic parameters; preprocessing the extracted characteristic parameters based on a time sequence; correlation analysis is conducted on the characteristic parameters subjected to time sequence preprocessing, and weight parameters influencing the elevator driving host are comprehensively analyzed; and a fault diagnosis model is constructed based on the weight parameters obtained through comprehensive analysis, and real-time fault diagnosis is conducted on the elevator driving host to be detected through the fault diagnosis model. According to the method, the fault diagnosis model based on the random forest model is constructed, the random forest model is optimized through the probability search algorithm, the matching degree of the decision trees is calculated, and the decision trees with the best matching degree are selected to be combined into the new random forest model, so that the detection precision of the model can be improved, and the detection success rate is increased.

Description

Technical Field

[0001] The present invention relates to the technical field of elevator drive host fault diagnosis, and more particularly to an elevator drive host fault diagnosis method and system. Background Art

[0002] Elevators are a fast and convenient means of transportation in buildings, which improves people's quality of life. With the annual increase in the number of elevators, the demand for elevator detection, fault diagnosis, and maintenance has also increased. At present, the reform of the elevator inspection and testing system and the means of smart supervision are constantly improving. The elevator maintenance mode has gradually changed from traditional regular maintenance to on-demand maintenance. In order to improve the pertinence and effectiveness of elevator maintenance, it is necessary to propose an intelligent early warning algorithm for elevator drive system faults to replace traditional manual detection.

[0003] The main method of elevator drive host detection, fault diagnosis, maintenance and repair is to rely on the experience and knowledge of detection personnel and maintenance workers to make judgments. The level of relevant personnel is uneven, and potential faults of elevator drive host cannot be eliminated well. In addition, the compatibility of elevators of different brands is not good, and the speed of elevator product update and upgrade is also fast. The detection personnel and maintenance personnel have different levels of understanding of elevators, which hinders the reliability of related fault detection and diagnosis. Developing a system for remote real-time monitoring of elevator drive host, combined with big data analysis technology, can improve the efficiency and pertinence of elevator maintenance and lay the foundation for on-demand maintenance. In addition, the existing elevator drive host (traction machine) detection and fault diagnosis technology does not monitor dynamic parameters enough, and it is not convenient to conduct a comprehensive evaluation of the performance of the drive host under various working conditions. The detection results are not comprehensive and accurate enough, and it is impossible to monitor the operating status of the elevator drive host in real time, which makes it difficult to discover and prevent potential faults in advance, increasing the safety risks of the elevator drive host in the actual operation process.

[0004] Therefore, how to propose a fault diagnosis method and system for the elevator drive host, conduct a comprehensive evaluation of the drive host performance under various working conditions based on timing, monitor the operating status of the elevator drive host in real time, and reduce the safety risks of the elevator drive host during actual operation is an urgent problem that technical personnel in this field need to solve. Summary of the invention

[0005] In view of this, the present invention provides a fault diagnosis method and system for an elevator drive host, which comprehensively evaluates the performance of the drive host under various working conditions based on the time sequence, monitors the operating status of the elevator drive host in real time, reduces the safety risks of the elevator drive host during actual operation, and improves the comprehensiveness and accuracy of detection. In order to achieve the above purpose, the present invention adopts the following technical solutions:

[0006] A fault diagnosis method for an elevator drive host, comprising:

[0007] Obtain the operating data of the elevator drive host under various working conditions and extract characteristic parameters;

[0008] Perform time series-based preprocessing on the extracted feature parameters;

[0009] Conduct correlation analysis on the characteristic parameters that have been preprocessed in time series, and comprehensively analyze the weight parameters that affect the elevator drive host;

[0010] A fault diagnosis model is constructed based on the weight parameters obtained through comprehensive analysis, and real-time fault diagnosis is performed on the elevator drive host to be tested through the fault diagnosis model.

[0011] Optionally, the characteristic parameters include voltage, current, power, rotation speed, three-phase vibration acceleration, traction wheel verticality, noise, braking time and braking distance.

[0012] Optionally, the performing time series-based preprocessing on the extracted characteristic parameters includes:

[0013] Obtaining a preset operation plan of the elevator drive host to be tested, and obtaining a preset operation time period of the elevator drive host to be tested according to the preset operation plan of the elevator drive host to be tested;

[0014] The preset operation time period is divided into a plurality of preset sub-operation time periods, and when the elevator drive host to be tested is working, its actual parameters in the preset sub-operation time periods are obtained.

[0015] Optionally, the method further includes: screening the actual parameters again;

[0016] Construct a decision tree model to classify the actual parameters and obtain a subset of the actual parameters of the elevator drive host to be tested based on the timing;

[0017] The PAC algorithm is used to reduce the dimension of each actual parameter subset, and the Dunn index is used to evaluate the classification effect of the actual parameter subset after dimension reduction.

[0018] The actual parameter subset with abnormal classification as the evaluation result is corrected to obtain the actual parameter subset with normal classification; all the actual parameter subsets with normal classification are aggregated into the final actual parameter subset.

[0019] Optionally, the performing correlation analysis on the characteristic parameters that have undergone time series preprocessing includes:

[0020] The feature parameters that have been preprocessed by time series are randomly divided into training sets and prediction sets according to a certain ratio;

[0021] Normalize the training set and prediction set data to a certain range;

[0022] The Pearson correlation coefficient was used to perform correlation analysis on different characteristic parameters of the data, and the characteristic parameters with low correlation were deleted to select the key characteristic parameters.

[0023] Optionally, the comprehensive analysis of weight parameters affecting the elevator drive host includes:

[0024] Input an initial sample set containing key characteristic parameters of the elevator drive host under various working conditions;

[0025] All key feature parameters of the initial sample set are classified into levels to obtain several sample subsets containing different key feature parameters;

[0026] Perform loop processing on each sample subset to obtain the principal component coefficient matrix;

[0027] Rearrange the principal component coefficient matrix, and transform the initial sample set based on the principal component coefficient matrix to generate a new key feature parameter sample;

[0028] The evaluation model is trained based on the new key characteristic parameters to obtain a sub-model for evaluating the performance of the elevator drive host, and new key characteristic parameter samples are cyclically generated to obtain an integrated evaluation model;

[0029] The initial sample set is evaluated using an integrated evaluation model, and all evaluation results are voted on to determine the final weight parameters of the elevator drive host to be tested.

[0030] Optionally, constructing a fault diagnosis model based on weight parameters obtained through comprehensive analysis includes:

[0031] Input the weight parameters of the elevator drive host to be tested, preset the height of the decision tree, and initialize the random forest model;

[0032] According to the weight parameters of the elevator drive host to be tested, multiple decision trees are constructed to form an initial random forest model;

[0033] The key feature parameters are used as training sets to train the initial random forest model, and the trained fault diagnosis model is used for fault diagnosis.

[0034] Optionally, the method also includes optimizing the initial random forest model, calculating the accuracy value of each decision tree in the initial random forest model, and calculating the difference values ​​between the decision trees; selecting decision trees with a higher matching degree from the initial forest model to form an optimized random forest model; and predicting whether the elevator drive host is normal based on the input performance parameters by optimizing the random forest model.

[0035] Optionally, the step of selecting decision trees with a higher matching degree from the initial forest model to form an optimized random forest model includes:

[0036] Initialize the probability search algorithm, set the initial temperature and initial solution;

[0037] Randomly perturb the set initial solution to generate a new solution;

[0038] If the matching degree of the new solution is higher than that of the initial solution, the new solution is accepted as the initial solution. Otherwise, the acceptance probability of the new solution is calculated according to the Metropolis rule. If the acceptance probability is greater than the random number, the new solution is accepted as the initial solution. Otherwise, the initial solution is retained.

[0039] If the set termination condition is met, the output initial solution is the optimal solution;

[0040] Several decision trees that meet the preset matching degree are selected from the initial forest model and combined into an optimized random forest model.

[0041] Optionally, a fault diagnosis system for an elevator drive host comprises:

[0042] Acquisition module: used to obtain the operating data of the elevator drive host under various working conditions and extract characteristic parameters;

[0043] Preprocessing module: used to perform time series-based preprocessing on the extracted feature parameters;

[0044] Parameter screening module: used to perform correlation analysis on characteristic parameters that have been preprocessed by time series, and comprehensively analyze the weight parameters that affect the elevator drive host;

[0045] Fault diagnosis module: It is used to build a fault diagnosis model based on the weight parameters obtained by comprehensive analysis, and perform real-time fault diagnosis on the elevator drive host to be tested through the fault diagnosis model.

[0046] It can be seen from the above technical solutions that, compared with the prior art, the present invention discloses a fault diagnosis method and system for an elevator drive host, which has the following beneficial effects:

[0047] The present invention proposes a fault diagnosis method for an elevator drive host, including: obtaining the operating data of the elevator drive host under various working conditions to be tested, extracting characteristic parameters; performing time series preprocessing on the extracted characteristic parameters; performing correlation analysis on the characteristic parameters after time series preprocessing, and comprehensively analyzing the weight parameters affecting the elevator drive host; constructing a fault diagnosis model based on the weight parameters obtained by the comprehensive analysis, and performing real-time fault diagnosis on the elevator drive host to be tested through the fault diagnosis model. The present invention constructs a fault diagnosis model so that abnormal conditions of the elevator drive host can be detected, and by constructing multiple decision trees and training the model to determine whether the parameters of the elevator drive host are normal, the fault point of the elevator drive host can be effectively discovered. At the same time, the random forest model can be optimized through a probability search algorithm, and the decision tree with the best matching degree is selected by calculating the matching degree of the decision tree to form a new random forest model, thereby improving the detection accuracy of the model and the detection success rate. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0049] Figure 1 A schematic flow chart of a fault diagnosis method for an elevator drive host provided by the present invention.

[0050] Figure 2 A structural framework diagram of a fault diagnosis system for an elevator drive host provided by the present invention. DETAILED DESCRIPTION

[0051] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0052] The embodiment of the present invention discloses a fault diagnosis method for an elevator drive main unit, such as Figure 1 As shown, including:

[0053] Obtain the operating data of the elevator drive host under various working conditions and extract characteristic parameters;

[0054] Perform time series-based preprocessing on the extracted feature parameters;

[0055] Conduct correlation analysis on the characteristic parameters that have been preprocessed in time series, and comprehensively analyze the weight parameters that affect the elevator drive host;

[0056] A fault diagnosis model is constructed based on the weight parameters obtained through comprehensive analysis, and real-time fault diagnosis is performed on the elevator drive host to be tested through the fault diagnosis model.

[0057] Furthermore, the characteristic parameters include voltage (power supply voltage, output voltage), current, power, speed, three-phase vibration acceleration, traction wheel verticality, noise, braking time, braking distance, current, motor speed, torque, efficiency, temperature and drive host status.

[0058] Furthermore, the time-series-based preprocessing of the extracted characteristic parameters includes:

[0059] Obtaining a preset operation plan of the elevator drive host to be tested, and obtaining a preset operation time period of the elevator drive host to be tested according to the preset operation plan of the elevator drive host to be tested;

[0060] The preset operation time period is divided into a plurality of preset sub-operation time periods, and when the elevator drive host to be tested is working, its actual parameters in the preset sub-operation time periods are obtained.

[0061] Furthermore, the method further includes: screening the actual parameters again;

[0062] Construct a decision tree model to classify the actual parameters and obtain a subset of the actual parameters of the elevator drive host to be tested based on the timing;

[0063] The PAC algorithm is used to reduce the dimension of each actual parameter subset, and the Dunn index is used to evaluate the classification effect of the actual parameter subset after dimension reduction.

[0064] The actual parameter subset with abnormal classification as the evaluation result is corrected to obtain the actual parameter subset with normal classification; all the actual parameter subsets with normal classification are aggregated into the final actual parameter subset.

[0065] In a specific implementation, it also includes: obtaining historical operating parameters of the elevator drive host to be tested, building an LSTM model based on the historical operating parameters, and importing the preset operating plan of the elevator drive host to be tested into the LSTM model for prediction, so as to obtain a preset parameter set based on time series for the elevator drive host to be tested within a preset operating period.

[0066] In a specific implementation, obtaining the preset parameter set specifically includes:

[0067] Obtaining historical operating parameters of the elevator drive host to be tested, analyzing data characteristics of the historical operating parameters, and dividing the historical operating parameters of the elevator drive host to be tested into a plurality of discrete intervals according to the data characteristics, each discrete interval representing a state;

[0068] According to the historical operating parameters of each discrete interval, the Markov chain is introduced, and the number of transitions from one state to another is counted in combination with the maximum likelihood method to obtain the transition probability between the states, and the state transition probability matrix is ​​constructed according to the transition probability;

[0069] Building an LSTM model based on a deep convolutional neural network, importing the state transition probability matrix into a convolutional layer for convolution processing and extracting local features of the state transition probability matrix, generating a sequence relationship and a long-term dependency relationship of the LSTM structure according to the local features, so as to determine the number of layers of the LSTM model and the dimension of the hidden state of each LSTM layer, and outputting initial model parameters;

[0070] Through the forward propagation and back propagation algorithms, the initial model parameters are adjusted so that the LSTM model can learn the temporal relationship and long-term dependency of the sequence, embed the temporal relationship and long-term dependency into the fully connected layer, output the feature sequence and the target sequence, and finally classify through Softmax, save the final model parameters, and obtain the trained LSTM model;

[0071] Obtain a preset operation plan of the elevator drive host to be tested, import the preset operation plan into the trained LSTM model, so as to perform time series-based prediction on the operation parameters of the elevator drive host to be tested and obtain a prediction result; and obtain a preset parameter set of the elevator drive host to be tested based on time series within a preset operation period according to the prediction result.

[0072] In a specific implementation, it also includes: obtaining the preset sub-operation time period to which the final actual parameter subset belongs, dividing the preset parameter set of the elevator drive host to be tested according to the preset sub-operation time period to obtain the preset parameter subset of the elevator drive host to be tested in the preset sub-operation time period;

[0073] Each final actual parameter subset is compared with the corresponding preset parameter subset to determine the first operating state of the elevator drive host to be tested; corresponding control measures are determined according to the operating state of the elevator drive host to be tested, and feedback is provided based on the control measures.

[0074] Furthermore, the correlation analysis of the characteristic parameters after time series preprocessing includes:

[0075] The feature parameters that have been preprocessed by time series are randomly divided into training sets and prediction sets according to a certain ratio;

[0076] Normalize the training set and prediction set data to a certain range;

[0077] The Pearson correlation coefficient was used to perform correlation analysis on different characteristic parameters of the data, and the characteristic parameters with low correlation were deleted to select the key characteristic parameters.

[0078] In a specific implementation, the selecting of key characteristic parameters includes:

[0079] Assume that the multivariate regression model is: L = XW + ε; where x = [b1, b2, …, b m ] T , X∈R m×n is the sign variable, L=[11,12,…,1 m ] T , L∈R m×1 is the discriminant label, ε∈R m×1 is a random error, E=[ω1,ω2,…,ω n ] T ,ω∈R n×1 is the regression coefficient vector;

[0080] The parameter a is adjusted according to the optimization function Q(ω) to achieve sparse dimension reduction of the target variable. The optimization function expression is: Q(ω) = argmin{ / / L-Xω|| 2 +λ1|ω|+λ2||ω||2}, where λ1 and λ2 are penalty coefficients, let a=λ1 / λ1+λ2, λ=λ1+λ2, and get the following formula: Q(ω)=argmin{||L-Xω / / 2 +λ[a|ω|+(1-a)||ω / / 2]};

[0081] Delete the sign vectors corresponding to the regression coefficient of 0 and keep the remaining key parameters.

[0082] Furthermore, the comprehensive analysis of weight parameters affecting the elevator drive host includes:

[0083] Input an initial sample set containing key characteristic parameters of the elevator drive host under various working conditions;

[0084] All key feature parameters of the initial sample set are classified into levels to obtain several sample subsets containing different key feature parameters;

[0085] Perform loop processing on each sample subset to obtain the principal component coefficient matrix;

[0086] Rearrange the principal component coefficient matrix, and transform the initial sample set based on the principal component coefficient matrix to generate a new key feature parameter sample;

[0087] The evaluation model is trained based on the new key characteristic parameters to obtain a sub-model for evaluating the performance of the elevator drive host, and new key characteristic parameter samples are cyclically generated to obtain an integrated evaluation model;

[0088] The initial sample set is evaluated using an integrated evaluation model, and all evaluation results are voted on to determine the final weight parameters of the elevator drive host to be tested.

[0089] In a specific implementation, performing a loop process on each sample subset to obtain a principal component coefficient matrix includes:

[0090] The sample subset is resampled using the bootstrap method to obtain a new key parameter sample subset;

[0091] Perform principal component analysis on the new key parameter sample subset to obtain the principal component coefficient vector related to the performance of the elevator drive host;

[0092] Repeat to obtain several groups of principal component coefficients, and merge the coefficients into a principal component coefficient matrix.

[0093] Furthermore, the construction of the fault diagnosis model based on the weight parameters obtained by comprehensive analysis includes:

[0094] Input the weight parameters of the elevator drive host to be tested, preset the height of the decision tree, and initialize the random forest model;

[0095] According to the weight parameters of the elevator drive host to be tested, multiple decision trees are constructed to form an initial random forest model. Specifically, a number of time series points are obtained according to the preset sub-operation period to which the time series data of the weight parameters belong;

[0096] For each time series point, a decision tree node matching the time series point is generated to obtain a plurality of parallel decision trees;

[0097] Calculating the node weight corresponding to each decision tree node according to the distribution characteristics of the decision tree nodes in the plurality of parallel decision trees; obtaining the initial node weight of each decision tree node according to the characteristic attributes of the decision tree nodes in the plurality of parallel decision trees; determining the weighted change value matching each decision tree node according to the distribution characteristics of the decision tree nodes in the plurality of parallel decision trees; calculating the node weight of each decision tree node according to the initial node weight of each decision tree node and the weighted change value matching each decision tree node;

[0098] Generate the next batch of decision tree nodes for each decision tree node according to the preset integration principle and the node weight corresponding to each decision tree node, select the nodes corresponding to each decision tree node whose node weight is less than the first preset weight as error nodes according to the preset integration principle, and select the nodes corresponding to each decision tree node whose node weight is greater than the second preset weight as correct nodes according to the preset integration principle; calculate the difference between the node weight of the error node and the first preset weight as the first weight difference, and calculate the difference between the node weight of the correct node and the second preset weight as the second weight difference; reduce the node weight of the error node according to the first weight difference to obtain a first integrated weight, and increase the node weight of the correct node according to the second weight difference to obtain a second integrated weight; generate the next batch of decision tree nodes for each decision tree node according to the first integrated weight and the second integrated weight;

[0099] Determine whether the number of data types corresponding to the next batch of decision tree nodes is less than or equal to the preset number of data types; if so, stop building the decision tree and obtain the target decision tree; if not, execute the step of determining the next batch of decision tree nodes as new decision tree nodes, and execute the step of generating the next batch of decision tree nodes for each decision tree node according to the preset integration principle and the node weight corresponding to each decision tree node;

[0100] The key feature parameters are used as training sets to train the initial random forest model, and the trained fault diagnosis model is used for fault diagnosis.

[0101] In a specific implementation, the preset integration principle is the Boost integration principle, which is to generate the decision tree nodes of the next tree in the order of the previous tree. Therefore, the next batch of decision tree nodes of the current tree node can be generated according to the Boost integration principle and the node weight corresponding to each decision tree node. When the generated decision tree is the target decision tree, the data corresponding to each time series point can be analyzed according to the target decision tree, and then the future change trend of the data can be predicted, thereby enhancing the stability of the analysis results obtained by analyzing the data with time series characteristics.

[0102] In a specific implementation, a trained fault diagnosis model is used to perform fault diagnosis to obtain a second operating state, which is coupled with the first operating state for comprehensive judgment. When the first operating state is a fault and the second operating state is normal, the elevator drive host to be tested is judged to be a low-probability fault and requires manual inspection. When the first operating state is normal and the second operating state is a fault, the elevator drive host to be tested is judged to be a fault. When the first operating state is a fault and the second operating state is a fault, the elevator drive host to be tested is judged to be a fault.

[0103] Furthermore, it also includes optimizing the initial random forest model, calculating the accuracy value of each decision tree in the initial random forest model by the leave-one-out method, and calculating the difference value between the decision trees by the statistical method; using a probability search algorithm to select decision trees with a higher matching degree from the initial forest model to form an optimized random forest model; and predicting whether the elevator drive host is normal based on the input performance parameters by optimizing the random forest model.

[0104] Furthermore, the selecting decision trees with higher matching degree from the initial forest model to form an optimized random forest model includes:

[0105] Initialize the probability search algorithm, set the initial temperature and initial solution;

[0106] Randomly perturb the set initial solution to generate a new solution;

[0107] If the matching degree of the new solution is higher than that of the initial solution, the new solution is accepted as the initial solution. Otherwise, the acceptance probability of the new solution is calculated according to the Metropolis rule. If the acceptance probability is greater than the random number, the new solution is accepted as the initial solution. Otherwise, the initial solution is retained.

[0108] If the set termination condition is met, the output initial solution is the optimal solution. The termination condition is that the new solution is not accepted in several consecutive Metropolis chains or reaches the end temperature. Otherwise, the temperature is decayed by the decay function and the set initial solution is randomly perturbed to generate a new solution.

[0109] The calculation formula of the attenuation function is:

[0110]

[0111] Among them, R t It represents the temperature value at the ηth step, and R0 represents the initial temperature.

[0112] Several decision trees that meet the preset matching degree are selected from the initial forest model and combined into an optimized random forest model.

[0113] In a specific implementation, a fault diagnosis system for an elevator drive host, such as Figure 2 As shown, including:

[0114] Acquisition module: used to obtain the operating data of the elevator drive host under various working conditions and extract characteristic parameters;

[0115] Preprocessing module: used to perform time series-based preprocessing on the extracted feature parameters;

[0116] Parameter screening module: used to perform correlation analysis on characteristic parameters that have been preprocessed by time series, and comprehensively analyze the weight parameters that affect the elevator drive host;

[0117] Fault diagnosis module: It is used to build a fault diagnosis model based on the weight parameters obtained by comprehensive analysis, and perform real-time fault diagnosis on the elevator drive host to be tested through the fault diagnosis model.

[0118] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0119] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A fault diagnosis method for an elevator drive main unit, characterized in that: include: Obtain the operating data of the elevator drive host under various working conditions and extract characteristic parameters; Perform time series-based preprocessing on the extracted feature parameters; Conduct correlation analysis on the characteristic parameters that have been preprocessed in time series, and comprehensively analyze the weight parameters that affect the elevator drive host; A fault diagnosis model is constructed based on the weight parameters obtained through comprehensive analysis, and real-time fault diagnosis is performed on the elevator drive host to be tested through the fault diagnosis model.

2. A fault diagnosis method for an elevator drive main unit according to claim 1, characterized in that: The characteristic parameters include voltage, current, power, rotation speed, three-phase vibration acceleration, traction wheel verticality, noise, braking time and braking distance.

3. A fault diagnosis method for an elevator drive main unit according to claim 1, characterized in that: The time-series-based preprocessing of the extracted characteristic parameters includes: Obtaining a preset operation plan of the elevator drive host to be tested, and obtaining a preset operation time period of the elevator drive host to be tested according to the preset operation plan of the elevator drive host to be tested; The preset operation time period is divided into a plurality of preset sub-operation time periods, and when the elevator drive host to be tested is working, its actual parameters in the preset sub-operation time periods are obtained.

4. A fault diagnosis method for an elevator drive main unit according to claim 3, characterized in that: It also includes: screening the actual parameters again; Construct a decision tree model to classify the actual parameters and obtain a subset of the actual parameters of the elevator drive host to be tested based on the timing; Perform dimensionality reduction processing on each actual parameter subset, and evaluate the classification effect of the actual parameter subset after dimensionality reduction; The actual parameter subset with abnormal classification as the evaluation result is corrected to obtain the actual parameter subset with normal classification; all the actual parameter subsets with normal classification are aggregated into the final actual parameter subset.

5. A fault diagnosis method for an elevator drive main unit according to claim 1, characterized in that: The correlation analysis of the characteristic parameters after time series preprocessing includes: The feature parameters that have been preprocessed by time series are randomly divided into training sets and prediction sets according to a certain ratio; Normalize the training set and prediction set data to a certain range; The Pearson correlation coefficient was used to perform correlation analysis on different characteristic parameters of the data, and the characteristic parameters with low correlation were deleted to select the key characteristic parameters.

6. A fault diagnosis method for an elevator drive main unit according to claim 5, characterized in that: The comprehensive analysis of weight parameters affecting the elevator drive host includes: Input an initial sample set containing key characteristic parameters of the elevator drive host under various working conditions; All key feature parameters of the initial sample set are classified into levels to obtain several sample subsets containing different key feature parameters; Perform loop processing on each sample subset to obtain the principal component coefficient matrix; Rearrange the principal component coefficient matrix, and transform the initial sample set based on the principal component coefficient matrix to generate a new key feature parameter sample; The evaluation model is trained based on the new key characteristic parameters to obtain a sub-model for evaluating the performance of the elevator drive host, and new key characteristic parameter samples are cyclically generated to obtain an integrated evaluation model; The initial sample set is evaluated using an integrated evaluation model, and all evaluation results are voted on to determine the final weight parameters of the elevator drive host to be tested.

7. A fault diagnosis method for an elevator drive main unit according to claim 1, characterized in that: The construction of the fault diagnosis model based on the weight parameters obtained by comprehensive analysis includes: Input the weight parameters of the elevator drive host to be tested, preset the height of the decision tree, and initialize the random forest model; According to the weight parameters of the elevator drive host to be tested, multiple decision trees are constructed to form an initial random forest model; The key feature parameters are used as training sets to train the initial random forest model, and the trained fault diagnosis model is used for fault diagnosis.

8. A fault diagnosis method for an elevator drive main unit according to claim 7, characterized in that: It also includes optimizing the initial random forest model, calculating the accuracy value of each decision tree in the initial random forest model, and calculating the difference values ​​between the decision trees; selecting decision trees with higher matching degrees from the initial forest model to form an optimized random forest model; and predicting whether the elevator drive host is normal based on the input performance parameters by optimizing the random forest model.

9. A fault diagnosis method for an elevator drive main unit according to claim 8, characterized in that: The step of selecting decision trees with a higher matching degree from the initial forest model to form an optimized random forest model includes: Initialize the probability search algorithm, set the initial temperature and initial solution; Randomly perturb the set initial solution to generate a new solution; If the matching degree of the new solution is higher than that of the initial solution, the new solution is accepted as the initial solution. Otherwise, the acceptance probability of the new solution is calculated. If the acceptance probability is greater than the random number, the new solution is accepted as the initial solution. Otherwise, the initial solution is retained. If the set termination condition is met, the output initial solution is the optimal solution; Several decision trees that meet the preset matching degree are selected from the initial forest model and combined into an optimized random forest model.

10. A fault diagnosis system for an elevator drive main unit, characterized in that: include: Acquisition module: used to obtain the operating data of the elevator drive host under various working conditions and extract characteristic parameters; Preprocessing module: used to perform time series-based preprocessing on the extracted feature parameters; Parameter screening module: used to perform correlation analysis on characteristic parameters that have been preprocessed by time series, and comprehensively analyze the weight parameters that affect the elevator drive host; Fault diagnosis module: It is used to build a fault diagnosis model based on the weight parameters obtained by comprehensive analysis, and perform real-time fault diagnosis on the elevator drive host to be tested through the fault diagnosis model.

Images