Elevator fault intelligent maintenance processing method based on ai large model elevator industry data

Abstract

The application discloses an elevator fault intelligent maintenance processing method based on an AI large model elevator industry data, and relates to the technical field of elevator maintenance, comprising: constructing a multi-dimensional elevator operation feature data set, establishing an elevator fault prediction model; based on the elevator fault prediction model, predicting the fault type, fault probability and fault time of the elevator in a future period of time; according to the prediction result, establishing a maintenance strategy recommendation model based on the elevator use frequency and the importance factor, and formulating an individualized intelligent maintenance strategy; executing the intelligent maintenance strategy, collecting the elevator operation data after maintenance, and feeding back to the AI large model to optimize and update the elevator fault prediction model. The application predicts the elevator fault through deep learning, intelligently recommends an individualized maintenance strategy, optimizes resource allocation, reduces downtime and invalid maintenance, incrementally learns to continuously optimize the model, improves prediction accuracy, prolongs the service life of the equipment, ensures normal operation, reduces safety hazards and maintenance costs.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of elevator maintenance, in particular to an elevator fault intelligent maintenance processing method based on AI large model elevator industry data. BACKGROUND

[0002] With the acceleration of modern urbanization process, elevators have become an indispensable means of transportation in daily life, especially in high-rise buildings, the use frequency and operation safety of elevators are crucial to people's quality of life. The traditional elevator fault maintenance method usually relies on manual inspection and regular maintenance, which has many shortcomings, for example, manual inspection relies on personnel experience, it is difficult to find potential problems in time and accurately, regular maintenance cannot optimize the maintenance plan targetedly, which is easy to cause resource waste or maintenance delay. Therefore, how to improve the accuracy of elevator fault diagnosis and the efficiency of maintenance has become a technical problem to be solved in the elevator industry.

[0003] In recent years, with the rapid development of artificial intelligence technology, especially the maturity of deep learning and big data analysis technology, AI large model has shown great potential in the field of equipment fault prediction and intelligent maintenance. By using AI large model to analyze elevator operation data in real time, combining historical fault data and elevator usage, the occurrence time, type and possible causes of elevator faults can be accurately predicted, which provides a scientific basis for formulating personalized intelligent maintenance strategies. Through such intelligent maintenance processing method, not only the safety of elevator operation can be improved, but also the operating cost can be significantly reduced, and the service efficiency of elevator can be improved. SUMMARY

[0004] To solve the above technical problems, the elevator fault intelligent maintenance processing method based on AI large model elevator industry data is provided, which solves the above problems.

[0005] To achieve the above purpose, the technical scheme adopted by the present application is:

[0006] The elevator fault intelligent maintenance processing method based on AI large model elevator industry data comprises:

[0007] S1: Obtain a multi-dimensional elevator operation feature data set, perform standardization processing, construct a multi-dimensional elevator operation feature data set, train the multi-dimensional elevator operation feature data set by using a deep learning algorithm in an AI large model, and establish an elevator fault prediction model;

[0008] S2: Based on the elevator fault prediction model, real-time analyze the elevator operation feature data set, predict the fault type, fault probability and fault time of the elevator in the future period of time;

[0009] S3: According to the prediction result, a maintenance strategy recommendation model is established based on the elevator usage frequency and the importance factor, and a personalized intelligent maintenance strategy is formulated based on the output of the maintenance strategy recommendation model;

[0010] S4: The intelligent maintenance strategy is executed, and the elevator operation data after maintenance is collected and fed back to the AI large model to optimize and update the elevator fault prediction model.

[0011] Preferably, the S1 specifically comprises:

[0012] The elevator operation data is collected in real time by the sensors built in the elevator, the historical fault data is retrieved from the elevator management system database, the elevator equipment parameters are obtained, the industry statistical data is collected, the correlation analysis method is used to evaluate the correlation between each feature and the elevator fault, the redundant features are removed, the multi-dimensional original features are obtained, and the normalization method is used for the multi-dimensional original features to map the data to the interval, and a multi-dimensional elevator operation feature data set is constructed;

[0013] The network combining the Transformer model and the multi-layer perception machine is selected to process the multi-dimensional time series data of the elevator and output the prediction probability of the elevator fault type;

[0014] The collected elevator operation feature data is divided into training set, validation set and test set in the ratio of 7:1:2, the historical fault type is used as the label, the model parameters are adjusted by the Adam algorithm, the training set data is input into the model, the fault type prediction probability is obtained by forward propagation, the difference between the prediction result and the true label is calculated by the cross-entropy loss function, the gradient is obtained by the back propagation algorithm, and the network weight is updated by the Adam algorithm. Adjust the parameters repeatedly to minimize the loss function value;

[0015] The model in the training is evaluated by the validation set to obtain the accuracy, recall rate and F1 value, the model is adjusted according to the evaluation result, if overfitting occurs, the model is simplified by the Dropout technology, and if underfitting occurs, the number of Transformer model layers and heads is increased;

[0016] When the model has the best performance on the validation set, the test set is used for final testing, the actual application effect of the model is determined according to the accuracy, recall rate and F1 value on the test set, and the elevator fault prediction model is established.

[0017] Preferably, the network combining the Transformer model and the multi-layer perception machine is selected to process the multi-dimensional time series data of the elevator and output the prediction probability of the elevator fault type specifically comprises:

[0018] The transformer model uses a multi-head attention mechanism to process the multi-dimensional time series data of the elevator, finds the correlation between the data, and then further processes it through a feedforward neural network. The processed data is input into the MLP, which transforms and fuses the features of the input data in different dimensions and levels. After passing through the fully connected layer for non-linear transformation and classification, the output is the predicted probability of the elevator fault type.

[0019] Preferably, the network weights are updated using the Adam algorithm, and the parameters are repeatedly adjusted to minimize the loss function value. Specifically, it includes:

[0020] The Adam algorithm is initialized, the momentum parameter is set, the training set data is input into the built model, and the data is sequentially forwarded through the transformer model and MLP. Finally, the probability distribution of the elevator fault type output by the model is calculated.

[0021] The cross-entropy loss function is used to calculate the difference between the model prediction result and the true target. The loss value reflects the accuracy of the model prediction. The smaller the loss value, the more accurate the prediction.

[0022] Through the backpropagation algorithm, the gradient of the loss function with respect to the model parameters is calculated, and the direction of parameter adjustment is obtained. Based on the direction of parameter adjustment, the Adam algorithm is used to update the parameters.

[0023] Preferably, S2 specifically includes:

[0024] Real-time acquisition of multi-dimensional feature data of the current elevator operation, including real-time running speed, car load change, door machine action time, traction machine current and temperature data;

[0025] Standardize the real-time collected data, and use the same normalization method as in S101 to map to the [0, 1] interval to form a real-time feature data set.

[0026] Divide the real-time feature data set into input sequences according to a 10-minute time window and input it into the elevator fault prediction model.

[0027] The model captures the correlation between real-time data and historical features through the multi-head attention mechanism of the transformer, and outputs the fault type set, the probability of each fault type, and the time interval of possible fault occurrence in the future preset time period through the MLP.

[0028] Threshold filtering is performed on the output results. When the predicted probability of the fault type exceeds 80%, it is marked as a high-risk fault type and triggers an early warning.

[0029] Preferably, S3 specifically includes:

[0030] ​An elevator use frequency evaluation model is constructed, and a use frequency index is obtained by weighted calculation based on the historical running times of the elevator, the running density in peak period and the daily average start-stop times;

[0031] An elevator importance evaluation model is constructed, and an importance index is obtained by an analytic hierarchy process based on the building type where the elevator is located, the served floor number and the historical passenger flow data;

[0032] The failure prediction result, the use frequency index and the importance index are taken as inputs to construct a maintenance strategy recommendation model, and a fuzzy decision algorithm is used to output three types of strategies.

[0033] Preferably, the elevator importance evaluation model is constructed, and the importance index is obtained by an analytic hierarchy process based on the building type where the elevator is located, the served floor number and the historical passenger flow data, and the specific steps include:

[0034] An elevator importance evaluation index system is obtained based on a correlation analysis method, including building type weight, served floor number weight and historical passenger flow weight;

[0035] The building type is graded and valued, and the importance index is calculated by an analytic hierarchy process, wherein the analytic hierarchy process formula is:

[0036]

[0037] In the formula, is the importance index, is the building type score, is the weight coefficient of the building type score, is the floor score, is the weight coefficient of the floor score, is the weight coefficient of the passenger flow score, is the passenger flow score, wherein The value range of is .

[0038] Preferably, the failure prediction result, the use frequency index and the importance index are taken as inputs to construct a maintenance strategy recommendation model, and a fuzzy decision algorithm is used to output three types of strategies, and the specific steps include:

[0039] The failure probability in the failure prediction result, the use frequency index and the importance index are taken as input variables of fuzzy decision, and fuzzy subsets are set;

[0040] The failure probability P: {low, medium, high}, corresponding to the interval [0, 30%], [30%, 70%], [70%, 100%];

[0041] Frequency of use F: {low, medium, high}, corresponding to the interval [0, 0.4], [0.4, 0.7], [0.7, 1];

[0042] Importance I: {low, medium, high}, corresponding to the interval [1, 2], [2, 4], [4, 5]; Establish a fuzzy rule base, including core rules:

[0043] If P = high and I = high, output emergency maintenance;

[0044] If P = medium and F = high, output planned maintenance;

[0045] If P = low or I = low, output routine maintenance;

[0046] Through fuzzy reasoning and defuzzification, generate maintenance strategy, including maintenance priority, on-site time window, required spare parts list and technical personnel level requirements.

[0047] Preferably, the S4 specifically includes:

[0048] According to the individualized intelligent maintenance strategy, generate a work order to obtain maintenance tasks, time nodes and responsible persons, and dispatch maintenance personnel through the maintenance management system;

[0049] When the maintenance personnel perform maintenance operations, record the fault handling process, the model and parameters of the replaced parts, and the elevator operation state data before and after maintenance;

[0050] After maintenance is completed, collect 48 hours of continuous operation data after the elevator resumes operation through sensors, including operation stability indicators, key component temperature changes and energy consumption data;

[0051] Integrate the maintenance record data and operation data, mark as maintained samples, supplement to the original training set at a proportion of 10%, while retaining the original validation set and test set proportion;

[0052] Input the updated data set into the AI large model, retrain the elevator fault prediction model, use incremental learning to freeze the bottom layer network parameters, only update the top layer classifier weights, shorten the model iteration period, and realize dynamic optimization of the model.

[0053] Preferably, in the incremental learning process, set a model performance evaluation threshold, when the F1 value of the updated model on the test set is improved by less than 3% compared with the original model, trigger the model structure optimization mechanism, the structure optimization mechanism specifically includes increasing the number of heads in the multi-head attention mechanism of the Transformer model and adding 10% of the number of neurons in the MLP hidden layer, retrain the model and compare the performance; At the same time, stratified sampling is performed on the maintained samples according to the fault type.

[0054] Compared with the prior art, the present application has the beneficial effects that:

[0055] The present application proposes to analyze elevator multi-dimensional operation data in real time through a deep learning algorithm, predict potential failures, reduce downtime probability, combine factors such as failure prediction, usage frequency and importance, intelligently recommend personalized maintenance strategies, optimize resource allocation, reduce ineffective maintenance, accurately predict failures, reduce maintenance costs, reduce downtime losses, continuously optimize the prediction model through incremental learning and feedback mechanism, and improve accuracy and intelligent level, so as to ensure normal operation of the elevator, prolong the service life, improve user experience and reduce safety hazards. BRIEF DESCRIPTION OF DRAWINGS

[0056] Figure 1 The step flow framework diagram of the present application is shown in the figure;

[0057] Figure 2 The step flow framework diagram of S1 in the present application is shown in the figure;

[0058] Figure 3 The step flow framework diagram of S2 in the present application is shown in the figure;

[0059] Figure 4 The step flow framework diagram of S3 in the present application is shown in the figure;

[0060] Figure 5 The step flow framework diagram of S4 in the present application is shown in the figure. DETAILED DESCRIPTION

[0061] The following description is used to disclose the present application so that those skilled in the art can implement the present application. The preferred embodiments in the following description are only as examples, and other obvious modifications can be thought of by those skilled in the art.

[0062] Referring to Figure 1 As shown, the elevator fault intelligent maintenance processing method based on AI large model elevator industry data comprises:

[0063] S1: Obtain multi-dimensional elevator operation feature data set, perform standardization processing, construct multi-dimensional elevator operation feature data set, train the multi-dimensional elevator operation feature data set by using the deep learning algorithm in the AI large model, and establish an elevator fault prediction model;

[0064] S2: Based on the elevator fault prediction model, real-time analysis of the elevator operation feature data set, prediction of the fault type, fault probability and fault time of the elevator in the future period of time;

[0065] S3: According to the prediction result, based on the elevator usage frequency and importance factor, a maintenance strategy recommendation model is established, and based on the output of the maintenance strategy recommendation model, a personalized intelligent maintenance strategy is formulated;

[0066] S4: Execute intelligent maintenance strategies and collect elevator operation data after maintenance, feeding it back to the AI ​​big data model to optimize and update the elevator fault prediction model.

[0067] Reference Figure 2 As shown, S1 specifically includes:

[0068] S101: Real-time elevator operation data is collected via built-in sensors. Historical fault data is retrieved from the elevator management system database to obtain elevator equipment parameters. Industry statistical data is collected, and correlation analysis is used to assess the degree of association between various features and elevator faults. Redundant features are removed to obtain multi-dimensional raw features. A normalization method is then applied to these multi-dimensional raw features to map the data to... Within a given range, construct a multi-dimensional elevator operation feature dataset;

[0069] The formula for correlation analysis is as follows:

[0070]

[0071] In the formula, It is a correlation index. This represents the actual value at time i in the elevator feature data. This represents the average value of the elevator characteristic data. Elevator fault data The actual value at each moment. is the average value of elevator malfunction data, and n is the total number of time periods;

[0072] in, The value ranges from -1 to 1. When = 1, it indicates that there is a perfectly positive linear correlation between variables x and y. When =−1, it indicates the existence of a completely negative linear correlation. When = 0, it indicates that there is no linear correlation between variables x and y. The closer the absolute value is to 1, the stronger the linear correlation between the elevator's operating characteristics and its malfunctions; the closer it is to 0, the weaker the linear correlation.

[0073] S102: A network combining the Transformer model and the multilayer perceptron is used to process multi-dimensional time series data of elevators and output the predicted probability of elevator fault types.

[0074] S103: Divide the collected elevator operation feature data into training set, validation set and test set in a ratio of 7:1:2. Use historical fault types as labels, adjust the model parameters using the Adam algorithm, input the training set data into the model, obtain the fault type prediction probability through forward propagation, calculate the difference between the prediction result and the true label using the cross-entropy loss function, obtain the gradient through the backpropagation algorithm, update the network weights using the Adam algorithm, and repeatedly adjust the parameters to minimize the loss function value.

[0075] S104: Evaluate the trained model using the validation set to obtain accuracy, recall, and F1 score. Adjust the model based on the evaluation results. If overfitting occurs, simplify the model using Dropout. If underfitting occurs, increase the number of Transformer model layers and heads.

[0076] S105: When the model performs best on the validation set, use the test set for final testing. Based on the accuracy, recall, and F1 score on the test set, determine the actual application effect of the model and complete the establishment of the elevator fault prediction model.

[0077] By analyzing elevator operation data in real time and using the multi-head attention mechanism of the Transformer model to capture complex correlations in time-series data, this method can accurately predict the future fault types and their probability of occurrence of elevators in a short period of time. The core of this method lies in providing dynamic and accurate fault prediction based on the combination of real-time data and historical data, which significantly improves the fault early warning capability.

[0078] S102 specifically includes:

[0079] The Transformer model uses a multi-head attention mechanism to process multi-dimensional time-series elevator data, find the correlation between data, and then further process it through a feedforward neural network. The processed data is input into an MLP, which transforms and fuses the features of the input data at different dimensions and levels. After nonlinear transformation and classification through a fully connected layer, the model outputs the predicted probability of elevator fault types.

[0080] The MLP calculation formula is as follows:

[0081]

[0082] In the formula, For the first in MLP The output of the first hidden layer is the result after the first hidden layer. The feature vector obtained after calculation in each hidden layer It is the ReLU activation function. For the first The weight matrix of each hidden layer For the first The output of each hidden layer For the first The bias vectors of each hidden layer;

[0083] By combining the Transformer model and MLP to process multi-dimensional time-series data of elevators, and using a multi-head attention mechanism to mine long-term dependencies in the data, the accuracy of the model in predicting elevator fault types is further improved. This combination mode enhances the fault prediction model's ability to process complex time-series data.

[0084] S103 specifically includes:

[0085] The Adam algorithm is initialized by setting the momentum parameter. The training set data is input into the built model. The data is then passed through the Transformer model and MLP for forward propagation. Finally, the probability distribution of elevator fault types output by the model is calculated.

[0086] The cross-entropy loss function is used to calculate the gap between the model's prediction and the actual target. The loss value reflects the accuracy of the model's prediction; the smaller the loss value, the more accurate the prediction.

[0087] The gradient of the loss function with respect to the model parameters is calculated using the backpropagation algorithm to obtain the direction of parameter adjustment. Based on the direction of parameter adjustment, the parameters are updated using the Adam algorithm, where the formula for the Adam algorithm is:

[0088]

[0089] In the formula, For first-order moment estimation, Indicates the first First-order moment estimation at the next iteration It is the exponential decay rate estimated by the first moment. It is the gradient of the current iteration. For second-order moment estimation, Indicates the first Second-order moment estimation at the next iteration It is the exponential decay rate estimated by the second moment. This represents the square of the gradient. This is the first-order moment estimate after bias correction. This is the second-order moment estimate after bias correction. It is the current iteration number. For the first Model parameters at the next iteration For learning rate, It is a constant. For the first After the first iteration update, the... Model parameters at the next iteration;

[0090] By reasonable data set division and Adam algorithm, the model training is more efficient. Especially in the training process, the cross entropy loss function is introduced to adjust the model parameters, and the prediction accuracy is further improved through back propagation and parameter update. This innovation optimizes the training process of elevator fault prediction, ensuring the reliability of the model in practical application.

[0091] Referring to Figure 3 S2 specifically includes:

[0092] S201: Real-time acquisition of current running multi-dimensional feature data of the elevator, including real-time running speed, car load change, door machine action time, traction machine current and temperature data;

[0093] S202: Standardization processing is performed on the real-time collected data, and the same normalization method as in S101 is used to map to the [0, 1] interval to form a real-time feature data set;

[0094] S203: The real-time feature data set is divided into input sequences according to a 10-minute time window and input into the elevator fault prediction model;

[0095] S204: The model captures the association between real-time data and historical features through the multi-head attention mechanism of Transformer, and outputs the fault type set in the future preset time period, the occurrence probability of each fault type, and the time interval in which the fault may occur through MLP;

[0096] S205: Threshold screening is performed on the output results, and when the prediction probability of the fault type exceeds 80%, it is marked as a high-risk fault type and triggers an early warning.

[0097] Referring to Figure 4 S3 specifically includes:

[0098] S301: An elevator usage frequency evaluation model is constructed, based on the historical running times of the elevator, the peak period running density, and the daily average start-stop times, the usage frequency index is obtained through weighted calculation, wherein the frequency index calculation formula is:

[0099]

[0100] In the formula, and are weight coefficients , is the daily average start-stop times normalized value, is the peak period running density normalized value;

[0101] S302: Construct an elevator importance evaluation model, combine the building type where the elevator is located, the service floor number and the historical passenger flow data, and obtain the importance index through the analytic hierarchy process;

[0102] S303: Take the fault prediction result, the frequency index F and the importance index I as inputs to construct a maintenance strategy recommendation model, and output three types of strategies by using a fuzzy decision algorithm.

[0103] S302 specifically includes:

[0104] Based on the correlation analysis method, an elevator importance evaluation index system is obtained, including building type weight, service floor number weight and historical passenger flow weight;

[0105] The building type is graded and valued, and the importance index is calculated by the analytic hierarchy process, wherein the analytic hierarchy process formula is:

[0106]

[0107] In the formula, is the importance index, is the building type score, is the weight coefficient of the building type score, is the floor score, is the weight coefficient of the floor score, is the weight coefficient of the passenger flow score, is the passenger flow score, wherein The value range of ;

[0108] The elevator importance evaluation model constructed based on the building type, the service floor number and the historical passenger flow data is to calculate the importance index of the elevator by the analytic hierarchy process. This innovation point combines the environment served by the elevator and the actual use, and provides a scientific basis for the customization of maintenance strategies.

[0109] S303 specifically includes:

[0110] Take the fault probability in the fault prediction result, the frequency index and the importance index as input variables of fuzzy decision, and set fuzzy subsets;

[0111] Fault probability P: {low, medium, high}, corresponding to the interval [0, 30%], [30%, 70%], [70%, 100%];

[0112] Frequency F: {low, medium, high}, corresponding to the interval [0, 0.4], [0.4, 0.7], [0.7, 1];

[0113] Importance I: {low, medium, high}, corresponding interval [1,2], [2,4], [4,5]; Establish a fuzzy rule base, including core rules:

[0114] If P = high and I = high, output emergency maintenance;

[0115] If P = medium and F = high, output planned maintenance;

[0116] If P = low or I = low, output routine maintenance;

[0117] Through fuzzy reasoning and defuzzification, generate maintenance strategy, including maintenance priority, on-site time window, required spare parts list and technical personnel level requirement;

[0118] In the maintenance strategy recommendation process, a fuzzy decision algorithm is used to process the fault type probability, usage frequency index and importance index, and generate the corresponding maintenance strategy. This method can accurately output emergency, planned or routine maintenance plan according to different situations, improving the automation and intelligence of maintenance decision-making.

[0119] Referring to Figure 5 S4 specifically includes:

[0120] S401: Generate a work order according to the individualized intelligent maintenance strategy, obtain the maintenance task, time node and responsible person, and dispatch maintenance personnel through the maintenance management system;

[0121] S402: When the maintenance personnel perform maintenance operations, record the fault handling process, replaced part model and parameters, and elevator operation state data before and after maintenance;

[0122] S403: After maintenance is completed, collect 48 hours of continuous operation data after the elevator resumes operation through sensors, including operation stability indicators, key component temperature changes and energy consumption data;

[0123] S404: Integrate the maintenance record data and operation data, mark it as a maintained sample, supplement it to the original training set at a proportion of 10%, while retaining the original validation set and test set proportion;

[0124] S405: Input the updated data set into the AI large model, retrain the elevator fault prediction model, use incremental learning to freeze the bottom layer network parameters, only update the top layer classifier weights, shorten the model iteration period, and realize dynamic optimization of the model;

[0125] Introduce an incremental learning mechanism to update the elevator fault prediction model after maintenance data feedback, which not only ensures real-time and efficiency, but also avoids time and resource waste of retraining each time. Through sampling and dynamic optimization of the maintained sample, the system can continuously evolve and maintain prediction accuracy.

[0126] In the incremental learning process, a model performance evaluation threshold is set, when the F1 value of the updated model on the test set is less than 3% than the original model, the model structure optimization mechanism is triggered, the structure optimization mechanism specifically includes increasing the number of heads of the multi-head attention mechanism in the Transformer model and adding 10% of the number of neurons in the MLP hidden layer, retraining the model and comparing the performance, at the same time, the maintained samples are stratified sampling according to the fault type.

[0127] In summary, the advantages of the present application are:

[0128] By collecting multi-dimensional running data of the elevator and training an elevator fault prediction model using deep learning algorithm, the running state of the elevator can be analyzed in real time, so as to predict the potential fault type and occurrence time in advance, and reduce the probability of elevator downtime;

[0129] Based on the fault prediction results and multi-dimensional factors such as elevator usage frequency and importance, intelligent maintenance strategy is recommended, and personalized maintenance plan is formulated according to different fault types and elevator running conditions, which not only ensures the timeliness of maintenance, but also maximizes the reduction of invalid maintenance and optimizes resource allocation;

[0130] By accurately predicting the occurrence of elevator faults, unnecessary regular maintenance and excessive maintenance are avoided, the work intensity of maintenance personnel and the cost of equipment replacement are reduced, and intelligent maintenance can reduce downtime losses caused by elevator faults, further saving enterprise operating costs;

[0131] Through incremental learning and feedback mechanism, the running data of the elevator after maintenance is continuously fed back to the AI large model, the elevator fault prediction model is continuously optimized, the accuracy and practicability of the model are improved, in addition, through dynamic optimization of the model structure, the accuracy of elevator fault prediction and the intelligent level of maintenance strategy are continuously improved;

[0132] Intelligent fault prediction and maintenance of the elevator can greatly reduce the probability of elevator failure, ensure the normal operation of the elevator, improve the user experience, at the same time, intelligent maintenance can ensure that the elevator is in the best working condition, prolong the service life of the equipment, and reduce the threat of sudden failure to passenger safety.

[0133] The above shows and describes the basic principles, main features and advantages of the present application. Those skilled in the art should understand that the present application is not limited to the above examples, the above examples and descriptions in the specification are only the principles of the present application, various changes and improvements can be made without departing from the spirit and scope of the present application, and these changes and improvements all fall within the scope of the claimed present application. The scope of protection claimed by the present application is defined by the appended claims and their equivalents.

Claims

1. An elevator failure intelligent maintenance processing method based on AI large model elevator industry data, characterized in that, Comprise: S1: Obtain a multi-dimensional elevator operation feature data set, perform standardization processing, construct a multi-dimensional elevator operation feature data set, train the multi-dimensional elevator operation feature data set by using a deep learning algorithm in an AI large model, and establish an elevator fault prediction model; S2: Based on the elevator fault prediction model, real-time analysis of the elevator operation feature data set, prediction of the elevator fault type, fault probability and fault time in the future period of time; S3: According to the prediction result, based on the elevator use frequency and the important degree factor, a maintenance strategy recommendation model is established, and based on the output of the maintenance strategy recommendation model, a personalized intelligent maintenance strategy is formulated; S4: Execute the intelligent maintenance strategy, and collect the elevator operation data after maintenance, and feed back to the AI large model to optimize and update the elevator fault prediction model; S1 specifically comprises: S101: Real-time collection of elevator operation data through the built-in sensor of the elevator, retrieval of historical fault data from the elevator management system database, acquisition of elevator equipment parameters, collection of industry statistical data, adoption of correlation analysis method to evaluate the correlation degree between each feature and elevator fault, removal of redundant features, obtaining of multi-dimensional original features, and adoption of normalization method for the multi-dimensional original features to map the data to [0-1] interval, and construction of multi-dimensional elevator operation feature data set; S102: Select a network combining Transformer model and multilayer perception to process elevator multi-dimensional time series data and output prediction probability of elevator fault type; S103: Divide the collected elevator operation feature data into training set, validation set and test set according to the ratio of 7:1:2, take historical fault type as label, adjust model parameters by Adam algorithm, input training set data into model, get fault type prediction probability through forward propagation, calculate the difference between prediction result and true label by cross-entropy loss function, then get gradient by back propagation algorithm, update network weight by Adam algorithm, repeatedly adjust parameters to minimize loss function value; S104: Evaluate the model in training by using the validation set to obtain accuracy, recall rate and F1 value, adjust the model according to the evaluation result, if overfitting occurs, simplify the model by using Dropout technology, if underfitting occurs, increase the number of Transformer model layers and heads; S105: When the model has the best performance on the validation set, use the test set for final testing, determine the actual application effect of the model according to the accuracy, recall rate and F1 value on the test set, and complete the establishment of the elevator fault prediction model. 2.The AI large model-based elevator industry data elevator failure intelligent maintenance processing method of claim 1, wherein, S102 specifically comprises: Transformer model uses multi-head attention mechanism to process elevator multi-dimensional time series data, finds out the correlation between data, and further processes through feedforward neural network, inputs the processed data into MLP, transforms and fuses the features of input data in different dimensions and levels, performs nonlinear transformation and classification through full connection layer, and outputs the prediction probability of elevator fault type. 3.The AI large model-based elevator industry data elevator failure intelligent maintenance processing method of claim 2, characterized in that, S103 specifically comprises: The Adam algorithm is initialized, the momentum parameter is set, the training set data is input into the built model, the data sequentially passes through the Transformer model and the MLP for forward propagation, and finally the probability distribution of the elevator fault type output by the model is calculated; The cross-entropy loss function is used to calculate the difference between the model prediction result and the true target, and the loss value reflects the accuracy of the model prediction. The smaller the loss value, the more accurate the prediction. Through the back propagation algorithm, the gradient of the loss function to the model parameters is calculated, the direction of parameter adjustment is obtained, and the parameters are updated based on the direction of parameter adjustment. 4.The AI large model-based elevator industry data elevator failure intelligent maintenance processing method of claim 1, wherein, The S2 specifically comprises: S201: Real-time collection of multi-dimensional feature data of the current elevator operation, including real-time running speed, car load change, door machine action time, traction machine current and temperature data; S202: Standardization processing of the real-time collected data, using the same normalization method as in S101 to map to the [0, 1] interval to form a real-time feature data set; S203: The real-time feature data set is divided into input sequences according to a 10-minute time window and input into the elevator fault prediction model; S204: The model captures the association between real-time data and historical features through the multi-head attention mechanism of Transformer, and outputs the fault type set, the occurrence probability of each fault type and the time interval of possible fault occurrence in the future preset time period through MLP; S205: Threshold screening is performed on the output result, and when the prediction probability of the fault type exceeds 80%, it is marked as a high-risk fault type and an early warning is triggered. 5.The AI large model-based elevator industry data elevator failure intelligent maintenance processing method of claim 1, wherein, The S3 specifically comprises: S301: Building an elevator usage frequency evaluation model, based on the historical running times of the elevator, the running density in peak hours and the daily average start-stop times, a usage frequency index is calculated by weighting; S302: Building an elevator importance evaluation model, combining the building type of the elevator, the number of serviced floors and the historical passenger flow data, an importance index is obtained by the analytic hierarchy process; S303: Taking the fault prediction result, the usage frequency index and the importance index as input, a maintenance strategy recommendation model is built, and a fuzzy decision algorithm is used to output three types of strategies. 6.The AI-based large model elevator industry data-based elevator failure intelligent maintenance processing method of claim 5, wherein The S302 specifically comprises: An elevator importance evaluation index system is obtained based on a correlation analysis method, including building type weight, service floor number weight and historical passenger flow weight; The building type is graded and valued, and the importance index is calculated by the analytic hierarchy process, wherein the analytic hierarchy process formula is: In the formula, is an importance index, is a building type score, is a weight coefficient of the building type score, is a floor score, is a weight coefficient of the floor score, is a weight coefficient of the people flow score, is a people flow score. 7.The AI-based large model elevator industry data-based elevator failure intelligent maintenance processing method of claim 5, wherein The S303 specifically comprises: The fault probability in the fault prediction result, the usage frequency index and the importance index are taken as input variables of fuzzy decision, and a fuzzy subset is set; Through fuzzy reasoning and defuzzification processing, a maintenance strategy is generated, including maintenance priority, on-site time window, required spare parts list and technical personnel level requirement. 8.The AI large model-based elevator industry data elevator failure intelligent maintenance processing method of claim 1, wherein, The S4 specifically comprises: S401: Generating a work order according to the individualized intelligent maintenance strategy, obtaining maintenance tasks, time nodes and responsible persons, and dispatching maintenance personnel through a maintenance management system; S402: When the maintenance personnel performs the maintenance operation, record the fault processing process, the replaced part model and parameters, and the elevator operation state data before and after maintenance; S403: After maintenance is completed, collect 48 hours of continuous operation data after the elevator resumes operation through sensors, including operation stability indicators, key component temperature changes, and energy consumption data; S404: Integrate the maintenance record data and the operation data, mark it as a maintained sample, supplement it to the original training set at a proportion of 10%, and retain the original validation set and test set proportions; S405: Input the updated data set into the AI large model, retrain the elevator fault prediction model, use incremental learning to freeze the bottom layer network parameters, only update the top layer classifier weights, shorten the model iteration period, and realize dynamic optimization of the model. 9.The AI big model-based elevator industry data elevator failure intelligent maintenance processing method of claim 8, characterized in that: In the incremental learning process, a model performance evaluation threshold is set. When the F1 value of the updated model on the test set is improved by less than 3% compared to the original model, the model structure optimization mechanism is triggered. The structure optimization mechanism specifically includes increasing the number of heads in the multi-head attention mechanism of the Transformer model and adding 10% of the number of neurons in the MLP hidden layer, retraining the model and comparing the performance; at the same time, stratified sampling is performed on the maintained samples according to the fault type.

Images