Elevator fault occurrence probability calculation method and device based on multi-source data

Abstract

The invention relates to an elevator fault occurrence probability calculation method and device based on multi-source data. The invention discloses a multi-source data-based elevator fault occurrence probability calculation method. The method comprises the following steps of: generating elevator maintenance data through a trained XGBoost model; elevator operation data, elevator triggering data and off-line statistical data automatically collected by an elevator and an XGBoost model are used for generating elevator maintenance data to train a long-short term memory neural network model; and then elevator operation data, elevator triggering data and offline statistical data which are automatically collected by an elevator are input into the trained long-short-term memory neural network model, so that the elevator maintenance score HI is automatically obtained, then the elevator fault occurrence probability is obtained through calculation, manual intervention is not needed in the whole process, and the calculation result is rapid and accurate.

Description

Technical Field

[0001] The present invention relates to calculation of elevator failure probability, and in particular to a method and device for calculating elevator failure probability based on multi-source data. Background Art

[0002] Elevators are the most common means of transportation in high-rise buildings, carrying tens of thousands of people every day. Due to the large number and frequency of elevator use, various types of faults are inevitable, including entrapment faults, top and bottom collision faults, and emergency stop faults, which may pose a threat to the personal safety of passengers; while door machine faults and invalid button faults affect the passengers' elevator experience and elevator operation efficiency. At present, elevator faults are still detected by passengers riding the elevator or regular maintenance, which costs a lot of manpower and time. With the development of big data and the Internet of Things, the operation data and trigger data of elevators can be obtained synchronously in real time. At the same time, the rapid upgrade and iteration of algorithms in the field of deep learning and machine learning can be used to mine the corresponding relationship between elevator components and performance status, establish a relationship model between elevator parameters and elevator faults, realize elevator component fault prediction, and provide information decision-making for elevator maintenance.

[0003] In the current field of elevator fault prediction, most existing technologies tend to directly use the remaining life of the elevator or direct parameters of specific fault types (such as the horizontal vibration amplitude of the traction machine) as prediction indicators. Although this method is direct, it lacks flexibility and is difficult to fully cover the complex and changeable elevator types and their diverse fault modes.

[0004] For example, in the patent "An Elevator Fault Prediction System and Method (CN117864892A)", the whole process detection of elevator fault maintenance and processing is realized mainly by constructing a data acquisition module, a data preprocessing module, a fault prediction module and an emergency processing module. In the data acquisition module, the elevator operation data and elevator maintenance data are mainly obtained through the IoT sensing device. The operation data mainly includes sensor data such as up and down signal data, door area signal data, and car switch signal data; the maintenance data includes fault data such as motor contactor contact pressure insufficient fault and SWD electronic board fault. The data preprocessing module analyzes the two types of data to obtain the elevator operation process data that causes the elevator maintenance data to be related to them. The fault prediction module constructs an RBF neural network model for the operation data and maintenance data, in which the nearest neighbor algorithm is used to obtain the weights of the nodes from the hidden layer to the output layer. The elevator operation data is used as input and the elevator maintenance data is used as output. The model results are compared with the elevator data of the same type to determine whether the elevator fault is directly related to the operation data. When the prediction result value is within the range of the elevator maintenance data of the same type, the ARIMA algorithm is used to predict the remaining life of the elevator. The remaining life is updated as the elevator maintenance time, completing the closed loop from elevator failure prediction to on-site elevator maintenance. Summary of the invention

[0005] The present invention generates elevator maintenance data through a trained XGBoost model; uses elevator operation data, elevator trigger data, offline statistical data and the XGBoost model automatically collected by the elevator to generate elevator maintenance data to train a long short-term memory neural network model; then the elevator operation data, elevator trigger data and offline statistical data automatically collected by the elevator are input into the trained long short-term memory neural network model to automatically obtain the elevator maintenance score HI, and then obtain the probability of elevator failure by calculation. The whole process does not require human intervention, and the calculation results are fast and accurate.

[0006] The present invention provides the following technical solution: a method for calculating the probability of elevator failure based on multi-source data, comprising the following steps:

[0007] S1. Collecting first elevator operation data, first elevator trigger data and first offline statistical data;

[0008] S2. Processing the abnormal data in the first elevator operation data, the first elevator trigger data and the first offline statistical data to obtain the second elevator operation data, the second elevator trigger data and the second offline statistical data;

[0009] S3, inputting the second elevator operation data, the second elevator trigger data and the second offline statistical data into the long short-term memory neural network model, thereby obtaining the first elevator maintenance data, where the first elevator maintenance data includes the following types of data: elevator number and elevator maintenance score HI;

[0010] S4. Calculate the probability of failure f(HI) by the following formula:

[0011] f(HI)=x×e -HI / y ;

[0012] Among them, x is the maximum probability of failure, y is the degree of decline in the probability of failure, and e is a natural constant. The values ​​of x and y can be set according to actual conditions. Generally, x is 100 and y is 30.

[0013] The first elevator operation data, the first elevator trigger data and the first offline statistical data all have a corresponding relationship, that is, they are collected based on the same batch of elevators.

[0014] Furthermore, the LSTM neural network model is trained by the following steps:

[0015] (1) collecting first elevator alarm information and second elevator maintenance data, and dividing them into a first training set and a first test set;

[0016] Generally, the ratio of the first training set to the first test set is 7:3, and it can also be divided as needed.

[0017] The first elevator alarm information and the second elevator maintenance data have a corresponding relationship, that is, they are collected based on the same batch of elevators.

[0018] (2) using the first training set to train the initial XGBoost model to obtain a trained XGBoost model; using the first elevator alarm information in the first training set as the input of the XGBoost model, and using the second elevator maintenance data in the first training set as the output of the XGBoost model;

[0019] (3) Using the first test set to test XGBoost, if the accuracy reaches the first threshold or above, the training is completed; if the accuracy is less than the first threshold, the first training set and the first test set are re-divided and re-trained;

[0020] Generally, the first threshold is 80%, and can also be set as required.

[0021] (4) collecting the third elevator operation data, the third elevator trigger data, the third offline statistical data, and the second elevator alarm information, and dividing them into a second training set and a second test set;

[0022] The collected third elevator operation data, the third elevator trigger data, the third offline statistical data, and the second elevator alarm information all have a corresponding relationship, that is, the collection is based on the same batch of elevators.

[0023] Generally, the ratio of the second training set to the second test set is 7:3, and it can also be divided as needed.

[0024] (5) Input the second elevator alarm information in the second training set into the XGBoost model to obtain the third elevator maintenance data;

[0025] (6) using the third elevator operation data, the third elevator trigger data, the third offline statistical data, and the third elevator maintenance data in the second training set to train an initial long short-term memory neural network model to obtain a trained long short-term memory neural network model; using the third elevator operation data, the third elevator trigger data, and the third offline statistical data in the second training set as inputs of the long short-term memory neural network model, and using the third elevator maintenance data as outputs of the long short-term memory neural network model;

[0026] (7) The second test set is used to test the long short-term memory neural network model. If the accuracy reaches a second threshold or above, the training is completed. If the accuracy is less than the second threshold, the second training set and the second test set are re-divided and re-trained.

[0027] Generally, the second threshold is not 80%, and can also be set as needed.

[0028] Further, the first elevator operation data, the second elevator operation data, and the third elevator operation data all include the following types of data: elevator operation speed (divided into x, y, and z axes), operation acceleration (divided into x, y, and z axes), temperature (divided into inside the car, top, and bottom), and humidity (divided into inside the car, top, and bottom);

[0029] The first elevator trigger data, the second elevator trigger data, and the third elevator trigger data all include the following types of data: elevator number, car door open, car door closed, current floor, running contactor energized, and star-blocking contactor released;

[0030] The first offline statistical data, the second offline statistical data, and the third offline statistical data all include the following types of data: elevator number, years of use, days of use, and total historical running distance;

[0031] The second elevator maintenance data and the second elevator maintenance data both include the following types of data: elevator number and elevator maintenance score HI, where the elevator maintenance score HI is a professional assessment of the current situation of the elevator by the operation and maintenance personnel (according to Tables 1 to 4);

[0032] Table 1 Elevator maintenance dimension table

[0033]

[0034]

[0035]

[0036] Table 2 Probability evaluation table

[0037]

[0038] Table 3 Risk Assessment Table

[0039]

[0040] Table 4 Maintenance score evaluation table

[0041]

[0042] The first elevator alarm information and the second elevator alarm information both include the following types of data: the elevator alarm information includes the elevator number, the elevator alarm start time, the elevator alarm end time, the elevator status, the elevator alarm type code and the number of alarm occurrences;

[0043] The elevator alarm type code can be associated through the elevator alarm type code table to obtain the elevator alarm type name, fault type and alarm type; the alarm occurrence times are divided by hours. If the start time and end time of an alarm are within the current one-hour time interval, the alarm occurrence times within the one-hour time period will increase by one; if the start time and end time of an alarm span multiple hour time intervals, the alarm occurrence times within the multiple hour time periods will increase by one respectively;

[0044] The number of alarm occurrences is divided into hours. If the start time and end time of an alarm are within the current one-hour time interval, the number of alarm occurrences within the one-hour time period increases by one; if the start time and end time of an alarm span multiple hour time intervals, the number of alarm occurrences within the multiple hour time periods increases by one respectively.

[0045] Furthermore, the elevator maintenance score HI includes the individual maintenance scores of the six dimensions of car, lobby, car top, pit bottom, machine room, and floor door. The individual maintenance score of each dimension is calculated by the following formula:

[0046] Maintenance score of a single dimension = 100 × (v1 + v2 + ... + vn) / 2n;

[0047] Among them, n is the total number of evaluation items in a single dimension, and v1, v2...vn are the scores of the 1st to nth evaluation items in a single dimension.

[0048] An elevator failure probability calculation device based on multi-source data according to the method of claim 1, comprising at least:

[0049] A first acquisition module, used for acquiring first elevator operation data, first elevator trigger data and first offline statistical data;

[0050] A first processing module is used to process the abnormal data in the first elevator operation data, the first elevator trigger data and the first offline statistical data to obtain the second elevator operation data, the second elevator trigger data and the second offline statistical data;

[0051] A first analysis module is used to input the second elevator operation data, the second elevator trigger data and the second offline statistical data into the long short-term memory neural network model, so as to obtain the first elevator maintenance data, where the first elevator maintenance data includes the following types of data: elevator number and elevator maintenance score HI;

[0052] The first calculation module calculates the probability of failure occurrence f(HI) by the following formula:

[0053] f(HI)=x×e -HI / y ;

[0054] Among them, x is the maximum probability of failure, y is the degree of decrease in the probability of failure, and e is a natural constant.

[0055] An electronic device, comprising:

[0056] one or more processors;

[0057] A memory for storing one or more programs;

[0058] When the one or more programs are executed by the one or more processors, the one or more processors implement the above method.

[0059] A computer-readable storage medium stores computer instructions, which implement the steps of the above method when executed by a processor.

[0060] The beneficial effects of the present invention are:

[0061] Through deep learning models, the health status of elevators can be predicted and converted into the probability of elevator failure. In addition, various elevator health indexes can be constructed for different components or fault types to predict the probability of occurrence of various elevator failures. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 is a flow chart of the method of the present invention;

[0063] Figure 2is a schematic diagram of a device module of the present invention;

[0064] Among them, there are a first acquisition module 1, a first processing module 2, a first analysis module 3, and a first calculation module 4. DETAILED DESCRIPTION

[0065] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation mode, structure, characteristics and effects of the present invention are described in detail below in combination with the accompanying drawings and preferred embodiments.

[0066] This embodiment is based on the real-time operation data, elevator trigger data and offline statistical data of the past 15 days, with hourly granularity, to predict the probability of occurrence of various types of faults every hour in the next day, and the design operation cycle is 6 hours, to achieve automatic learning of the deep correlation between elevator alarm information and maintenance records, and reduce the dependence on expert experience, and accurately predict the various types of faults that occur in the elevator. Specifically:

[0067] 1. Training XGBoost model:

[0068] (1) The first elevator alarm information is collected, which consists of the number of 17 types of elevator alarms and other data of 20 elevators in the records every 24 hours for 15 days, and the second elevator maintenance data is collected, which consists of the maintenance score HI of 20 elevators every 24 hours for 15 days, and is divided into the first training set and the first test set in a ratio of 7:3;

[0069] (2) using the first training set to train the initial XGBoost model to obtain a trained XGBoost model; using the first elevator alarm information in the first training set as the input of the XGBoost model, and using the second elevator maintenance data in the first training set as the output of the XGBoost model;

[0070] The XGBoost model obtains the hyperparameters in the XGBoost model through a conventional grid search method. The specific hyperparameters include learning rate, maximum depth, subsample ratio, feature sampling ratio and minimum subnode weight. The early stopping method is used to prevent the model from overfitting. The XGBoost, an efficient and scalable gradient boosting framework, is used to deeply mine and analyze the massive alarm and maintenance data of elevators. XGBoost can not only effectively process high-dimensional sparse data, but also prevent overfitting and ensure the generalization ability of the model through its built-in regularization terms and pruning strategies. Through XGBoost's automatic feature selection and weight allocation, there is no need to manually define complex indicator rules.

[0071] (3) Use the first test set to test XGBoost. If the accuracy reaches 80% or above, the training is completed. If the accuracy is less than 80%, re-divide the first training set and the first test set and re-train;

[0072] (4) The third elevator operation data is collected, which consists of 12 operation data feature fields of 200 elevators extracted from the records every 24 hours for 15 days; the third elevator trigger data is composed of 5 elevator trigger data feature fields of 200 elevators every 24 hours for 15 days; the third offline statistical data is composed of 3 elevator offline statistical data feature fields of 200 elevators every 24 hours for 15 days; the second elevator alarm information is composed of 17 types of alarm statistical fields of 200 elevators every 24 hours for 15 days, and is divided into the second training set and the second test set according to the ratio of 7:3;

[0073] 2. Training the long short-term memory neural network model:

[0074] (5) Input the second elevator alarm information in the second training set into the XGBoost model to obtain the third elevator maintenance data;

[0075] (6) using the third elevator operation data, the third elevator trigger data, the third offline statistical data, and the third elevator maintenance data in the second training set to train an initial long short-term memory neural network model to obtain a trained long short-term memory neural network model; using the third elevator operation data, the third elevator trigger data, and the third offline statistical data in the second training set as inputs of the long short-term memory neural network model, and using the third elevator maintenance data as outputs of the long short-term memory neural network model;

[0076] (7) The second test set is used to test the long short-term memory neural network model. If the accuracy rate reaches 80% or above, the training is completed. If the accuracy rate is less than 80%, the second training set and the second test set are re-divided and re-trained.

[0077] The HI health index calculated by XGBoost is used as the training label of the bidirectional LSTM neural network model to further improve the accuracy and efficiency of fault prediction. The deep learning model can capture the nonlinear relationship in the data, and combined with the comprehensive indicator of HI, it can more accurately predict the type and time of possible future elevator faults.

[0078] 3. Calculate the probability of elevator failure:

[0079] A method for calculating the probability of elevator failure based on multi-source data comprises the following steps:

[0080] S1. The first elevator operation data collected consists of 12 operation data feature fields for a total of 500 elevators in the record every 24 hours for 15 days, the first elevator trigger data consists of 5 elevator trigger data feature fields for a total of 500 elevators in the record every 24 hours for 15 days, and the first offline statistical data consists of 3 elevator offline statistical data feature fields for a total of 500 elevators in the record every 24 hours for 15 days;

[0081] S2. Process the abnormal data in the first elevator operation data, the first elevator trigger data and the first offline statistical data to obtain the second elevator operation data, the second elevator trigger data and the second offline statistical data. The dimensions of the second elevator operation data, the second elevator trigger data and the second offline statistical data after the abnormal data processing should be the same as those of the corresponding data set a.

[0082] In this embodiment, the abnormal data is processed using a box plot statistical method.

[0083] The specific method is to find the first quartile Q1, median Q2, and third quartile Q3 in the data to be processed, calculate the interquartile range IQR=Q3-Q1, calculate the lower threshold Q1-1.5*IQR and the upper threshold Q3+1.5*IQR, and identify the data that exceeds the upper threshold or is lower than the lower threshold in the data to be processed as abnormal data and delete it.

[0084] S3, inputting the second elevator operation data, the second elevator trigger data and the second offline statistical data into the long short-term memory neural network model, thereby obtaining the first elevator maintenance data, where the first elevator maintenance data includes the elevator maintenance score HI;

[0085] S4. Calculate the probability of failure f(HI) by the following formula:

[0086] f(HI)=x×e -HI / y ;

[0087] Among them, x is the maximum probability of failure, y is the degree of decrease in the probability of failure, x is 100, y is 30, and e is a natural constant.

[0088] The following method is used to verify the method of this application:

[0089] A failure probability greater than or equal to 50% indicates a high probability failure and requires maintenance, while a failure probability less than 50% indicates a low probability failure and does not require maintenance for the time being. The predicted failure probability f(HI) is converted into a binary result and compared with the actual failure situation within the predicted time range to calculate TP (both the prediction and the actual failure), TN (both the prediction and the actual failure), FP (the prediction is a failure, but the actual failure) and FN (the prediction is no failure, but the actual failure) in the confusion matrix, and then calculate the recall rate TPR = (TP / (TP+FN)) and the false positive rate FPR = (FP / (FP+TN)). The false positive rate FPR is arranged in ascending order and then brought into the formula to calculate the AUC value of the model: Where n is the number of 500 elevators predicted in 24 hours, with the hour as the granularity, that is, n = 12000. When the AUC value is greater than 0.7, it means that the failure probability calculation of the present application is accurate.

[0090] An elevator failure probability calculation device based on multi-source data according to the above method at least comprises:

[0091] A first acquisition module, used for acquiring first elevator operation data, first elevator trigger data and first offline statistical data;

[0092] A first processing module is used to process the abnormal data in the first elevator operation data, the first elevator trigger data and the first offline statistical data to obtain the second elevator operation data, the second elevator trigger data and the second offline statistical data;

[0093] A first analysis module is used to input the second elevator operation data, the second elevator trigger data and the second offline statistical data into the long short-term memory neural network model, so as to obtain the first elevator maintenance data, wherein the first elevator maintenance data includes the elevator maintenance score HI;

[0094] The first calculation module calculates the probability of failure occurrence f(HI) by the following formula:

[0095] f(HI)=x×e-HI / y;

[0096] Among them, x is the maximum probability of failure, y is the degree of decrease in the probability of failure, and e is a natural constant.

[0097] An electronic device, comprising:

[0098] one or more processors;

[0099] A memory for storing one or more programs;

[0100] When the one or more programs are executed by the one or more processors, the one or more processors implement the above method.

[0101] In the embodiments provided in the present application, it should be understood that the disclosed method and system can also be implemented in other ways. The method and system embodiments described above are merely schematic. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the method and system, method and computer program product according to multiple embodiments of the present application. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of a code, and the module, a program segment or a part of a code contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or the flowchart, and the combination of boxes in the block diagram and / or the flowchart can be implemented with a dedicated hardware-based system that performs a specified function or action, or can be implemented with a combination of dedicated hardware and computer instructions.

[0102] In addition, the functional modules in the various embodiments of the present application may be integrated together to form an independent part, or each module may exist separately, or two or more modules may be integrated to form an independent part.

[0103] On the other hand, a computer-readable storage medium stores thereon computer instructions, which implement the steps of the above method when executed by a processor. The computer program implements the method of any one of the above first aspects when executed by a processor. If the function is implemented in the form of a software function module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or the part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory 101 (ROM, Read-Only Memory), a random access memory 101 (RAM, Random Access Memory), a disk or an optical disk.

[0104] In the embodiments of the present application, on the one hand, by installing multiple types of sensors in the elevator, the multi-sensor information fusion helps to reduce the environmental interference that a single sensor may be subject to, thereby improving the reliability of the data; on the other hand, by using the XGBoost model and the long short-term memory neural network model method to fuse the characteristic information of the elevator operation status, the accuracy of the calculation of the probability of elevator failure can be improved, so that false alarms and missed alarms can be reduced, and the efficient operation of the calculation of the probability of elevator failure can be ensured, which significantly improves the safety of the elevator and reduces the probability of accidents.

[0105] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Although the present invention has been disclosed as a preferred embodiment as above, it is not used to limit the present invention. Any technical personnel in this field can make some changes or modify the technical contents disclosed above into equivalent embodiments without departing from the scope of the technical solution of the present invention. However, any brief modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. A method for calculating the probability of elevator failure based on multi-source data, characterized in that: The steps include: S1. Collecting first elevator operation data, first elevator trigger data and first offline statistical data; S2. Processing the abnormal data in the first elevator operation data, the first elevator trigger data and the first offline statistical data to obtain the second elevator operation data, the second elevator trigger data and the second offline statistical data; S3, inputting the second elevator operation data, the second elevator trigger data and the second offline statistical data into the long short-term memory neural network model, thereby obtaining the first elevator maintenance data, where the first elevator maintenance data includes the following types of data: elevator number and elevator maintenance score HI; S4. Calculate the probability of failure f(HI) by the following formula: f(HI)=x×e -HI / y ; Among them, x is the maximum probability of failure, y is the degree of decrease in the probability of failure, and e is a natural constant.

2. The method according to claim 1, characterized in that The LSTM neural network model is trained through the following steps: (1) collecting first elevator alarm information and second elevator maintenance data, and dividing them into a first training set and a first test set; (2) using the first training set to train the initial XGBoost model to obtain a trained XGBoost model; using the first elevator alarm information in the first training set as the input of the XGBoost model, and using the second elevator maintenance data in the first training set as the output of the XGBoost model; (3) Using the first test set to test XGBoost, if the accuracy reaches the first threshold or above, the training is completed; if the accuracy is less than the first threshold, the first training set and the first test set are re-divided and re-trained; (4) collecting the third elevator operation data, the third elevator trigger data, the third offline statistical data, and the second elevator alarm information, and dividing them into a second training set and a second test set; (5) Input the second elevator alarm information in the second training set into the XGBoost model to obtain the third elevator maintenance data; (6) using the third elevator operation data, the third elevator trigger data, the third offline statistical data, and the third elevator maintenance data in the second training set to train an initial long short-term memory neural network model to obtain a trained long short-term memory neural network model; using the third elevator operation data, the third elevator trigger data, and the third offline statistical data in the second training set as inputs of the long short-term memory neural network model, and using the third elevator maintenance data as outputs of the long short-term memory neural network model; (7) The second test set is used to test the long short-term memory neural network model. If the accuracy reaches a second threshold or above, the training is completed. If the accuracy is less than the second threshold, the second training set and the second test set are re-divided and re-trained.

3. The method according to claim 1, characterized in that The abnormal data were processed by box plot statistical method.

4. The method according to claim 1 or 2, characterized in that: The first elevator operation data, the second elevator operation data, and the third elevator operation data all include the following types of data: elevator operation speed, operation acceleration, temperature, and humidity; The first elevator trigger data, the second elevator trigger data, and the third elevator trigger data all include the following types of data: elevator number, car door open, car door closed, current floor, running contactor energized, and star-blocking contactor released; The first offline statistical data, the second offline statistical data, and the third offline statistical data all include the following types of data: elevator number, years of use, days of use, and total historical running distance; The second elevator maintenance data and the second elevator maintenance data both include the following types of data: elevator number and elevator maintenance score HI; The first elevator alarm information and the second elevator alarm information both include the following types of data: elevator number, elevator alarm start time, elevator alarm end time, elevator status, elevator alarm type code and alarm occurrence times.

5. The method according to claim 4, characterized in that The elevator maintenance score HI includes the individual maintenance scores of the six dimensions of car, lobby, car top, pit bottom, machine room, and floor door. The individual maintenance scores of each dimension are calculated using the following formula: Maintenance score of a single dimension = 100 × (v1 + v2 + ... + vn) / 2n; Among them, n is the total number of evaluation items in a single dimension, and v1, v2...vn are the scores of the 1st to nth evaluation items in a single dimension.

6. An elevator failure probability calculation device based on multi-source data according to the method of claim 1, characterized in that: At least: A first acquisition module, used for acquiring first elevator operation data, first elevator trigger data and first offline statistical data; A first processing module is used to process the abnormal data in the first elevator operation data, the first elevator trigger data and the first offline statistical data to obtain the second elevator operation data, the second elevator trigger data and the second offline statistical data; A first analysis module is used to input the second elevator operation data, the second elevator trigger data and the second offline statistical data into the long short-term memory neural network model, so as to obtain the first elevator maintenance data, where the first elevator maintenance data includes the following types of data: elevator number and elevator maintenance score HI; The first calculation module calculates the probability of failure occurrence f(HI) by the following formula: f(HI)=x×e -HI / y ; Among them, x is the maximum probability of failure, y is the degree of decrease in the probability of failure, and e is a natural constant.

7. An electronic device, characterized in that: include: one or more processors; A memory for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 6.

8. A computer-readable storage medium having computer instructions stored thereon, characterized in that: When the instruction is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images