An intelligent fault prediction model based on machine learning and big data analysis

Abstract

The present invention discloses an intelligent fault prediction model based on machine learning and big data analysis, and relates to the field of data processing technology. An intelligent fault prediction model based on machine learning and big data analysis is provided. First, the data is trained through historical operation and maintenance data and a convolutional neural network model to predict the first operation and maintenance characterization value of each rail transit line; then, combined with real-time monitoring and analysis of multimodal data of rail transit lines, the first operation and maintenance characterization value of each rail transit line is corrected, so as to accurately locate the fault point and issue an early warning. By dynamically acquiring and analyzing historical and real-time data, training a convolutional neural network model and adjusting it, potential faults can be discovered earlier, fault points can be accurately located, and the safety and operation and maintenance efficiency of the rail transit system can be improved, so that the fault warning is not only more timely, but also can be adaptively adjusted according to different conditions.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to an intelligent fault prediction model based on machine learning and big data analysis. Background Art

[0002] Various equipment in rail transit systems, such as trains, signal systems, and track environments, are in complex operating states, and these states are constantly changing. In order to ensure the safe, efficient, and sustainable operation of the system, timely prediction and location of potential faults has become a key task in modern rail transit management.

[0003] Prior art, such as the fault prediction method, model training method, electronic device, and readable storage medium disclosed in the invention patent with publication number: CN118395303A, is applicable to the field of rail transit technology. The fault prediction method includes: obtaining sample feature data of the signal unit in a preset time period before the required prediction month; inputting the sample feature data of the signal unit in a preset time period before the required prediction month into the target fault prediction model to obtain the probability of the signal unit failing in the required prediction month; the target fault prediction model is a model with better performance determined from each optimized candidate fault prediction model. In this embodiment, according to the relationship between the sample feature data of the signal unit and the fault in the original data, a candidate fault prediction model based on different machine learning algorithms is constructed, and a model with better performance is selected as the target fault prediction model. The data format is converted into the universal telemetry data format of the satellite comprehensive measurement system.

[0004] The prior art, such as the invention patent with publication number: CN118133095A, discloses a method for predicting rail transit equipment faults based on a large language model. The method includes the following steps: obtaining equipment operation and maintenance data and preprocessing the original data to construct a labeled text data set; using the operation and maintenance text data set to fine-tune the large language model to obtain a fault prediction model; receiving the real-time operation status and maintenance records of the equipment and inputting them into the large language model to obtain the prediction results and send the results to the operation and maintenance platform, recording the confirmation of the prediction by the operation and maintenance personnel; and regularly updating the model according to historical operation data. This improves the accuracy of fault prediction, reduces the number of operation and maintenance inspections, and improves the operation and maintenance efficiency.

[0005] At present, there are several problems in the process of track fault prediction. On the one hand, track fault prediction usually relies on static data or regular detection data, which cannot adapt to the rapidly changing rail transit operating environment. It may lead to the inability to timely reflect the actual status of the rail transit system, thereby providing the latest data support for the convolutional neural network model. Summary of the invention

[0006] In view of the deficiencies in the prior art, the present invention provides an intelligent fault prediction model based on machine learning and big data analysis, which can effectively solve the problems involved in the above-mentioned background technology.

[0007] To achieve the above objectives, the present invention is implemented through the following technical solutions: an intelligent fault prediction model based on machine learning and big data analysis, including a data acquisition and analysis module, used to obtain the historical operation and maintenance data set of the rail transit network in the big data storage, import the historical operation and maintenance data set of the rail transit network into the convolutional neural network model for training, and predict the first operation and maintenance characterization value of each rail transit line based on the trained convolutional neural network model.

[0008] The rail transit operation and maintenance data analysis module is used to monitor and obtain the multimodal data of each rail transit line in the rail transit network, analyze the multimodal data of each rail transit line, obtain the first operation and maintenance characterization positioning compensation value of each rail transit line, and correct and update the first operation and maintenance characterization value of each rail transit line.

[0009] The fault point prediction and positioning module is used to obtain the second operation and maintenance characterization value of each rail transit line based on the first operation and maintenance characterization positioning compensation value of each rail transit line and the updated first operation and maintenance characterization value of each rail transit line, and locate the predicted fault points of each transportation line of the rail transit network according to the second operation and maintenance characterization value of each rail transit line and issue an early warning.

[0010] The multimodal data of each rail transit line is analyzed to obtain the first operation and maintenance characterization positioning compensation value of each rail transit line. The specific process is: extracting the multimodal data of each rail transit line in the rail transit network, including each train travel data, signal system data and each track environment status data.

[0011] A first operation and maintenance characterization positioning compensation value of each rail transit line is obtained according to the travel data of each train, and the first operation and maintenance characterization positioning compensation value of each rail transit line is used to evaluate the smoothness of train operation of each rail transit line.

[0012] Furthermore, the convolutional neural network model based on the training predicts the first operation and maintenance characterization value of each rail transit line. The specific process is: extracting the historical operation and maintenance data set of the rail transit network, preprocessing the historical operation and maintenance data set of the rail transit network, and importing it into the model for training, and outputting the first operation and maintenance characterization value of each rail transit line.

[0013] Furthermore, the first operation and maintenance characterization value of each rail transit line is corrected and updated, and the specific process is: obtaining the signal compensation value of each rail transit line according to the signal system data, and the signal compensation value of each rail transit line is used to evaluate the reliability of the rail signal system.

[0014] Obtaining a track environment compensation value for each rail transit line according to each track environment status data, wherein the track environment compensation value for each rail transit line is used to evaluate the interference degree of external environmental conditions on the safety of each rail transit line;

[0015] Based on the signal compensation value of each rail transit line and the track environment compensation value of each rail transit line, the first operation and maintenance positive value of each rail transit line is obtained. The first operation and maintenance positive value of each rail transit line is used to evaluate the degree of interference correction of external environmental conditions on the safety of each rail transit line.

[0016] The first operation and maintenance positivity value of each rail transit line is extracted, and compared with the first operation and maintenance positivity threshold value of each rail transit line stored in the database; if the first operation and maintenance positivity value of a rail transit line is higher than or equal to the first operation and maintenance positivity threshold value of each rail transit line, the first operation and maintenance characterization value of the rail transit line and the first operation and maintenance positivity value of each rail transit line are weighted summed to obtain a corrected value of the first operation and maintenance characterization value of the rail transit line, and the corrected value of the first operation and maintenance characterization value of the rail transit line is updated as the first operation and maintenance positivity value of the rail transit line; if the first operation and maintenance positivity value of each rail transit line is lower than the first operation and maintenance positivity threshold value of each rail transit line, the first operation and maintenance characterization value of each rail transit line is directly used as the updated first operation and maintenance characterization value of each rail transit line.

[0017] Furthermore, the first operation and maintenance characterization positioning compensation value of each rail transit line is obtained according to the travel data of each train. The specific process is: preset a monitoring time period, and extract the travel data of each train in each rail transit line during the monitoring time period, including the average operating speed, average acceleration, average traction power and average braking force of each train.

[0018] The average running speed of each train on each rail transit line is extracted, and the track design speed stored in the database is extracted. The average running speed of each train on each rail transit line is processed with the track design speed and then the absolute value is processed to obtain the speed deviation value of each train on each rail transit line.

[0019] The speed deviation value of each train in each rail transit line, the average acceleration, the average traction power and the average braking force of each train are extracted, and the first operation and maintenance characterization positioning compensation value of each rail transit line is obtained after processing.

[0020] Furthermore, the signal compensation value of each rail transit line is obtained by extracting the signal system data of each rail transit line in each monitoring time period, including signal reception strength, change time of signal light status and position coordinates of each train.

[0021] The position coordinates of each train are extracted to obtain the time interval between adjacent trains, and based on the signal reception strength of each rail transit line and the average change time of the signal light status, the signal compensation value of each rail transit line is obtained through processing.

[0022] Furthermore, the track environment compensation value of each rail transit line is obtained according to the track environment status data. The specific process is: extracting the track environment status data of each rail transit line in the monitoring time period, including the average temperature of the track, the average vibration frequency of the track and the distortion of the track, and obtaining the track environment compensation value of each rail transit line after processing.

[0023] Furthermore, the second operation and maintenance characterization value of each rail transit line is obtained, and the specific process is: extract the updated first operation and maintenance characterization value of each rail transit line, and compare it with the set first operation and maintenance characterization positioning compensation threshold; if the first operation and maintenance characterization positioning compensation value of a certain rail transit line is higher than or equal to the first operation and maintenance characterization positioning compensation threshold, then the first operation and maintenance characterization positioning compensation value of the rail transit line is weightedly summed with the first operation and maintenance characterization value of the rail transit line to obtain the second operation and maintenance characterization value of the rail transit line; if the first operation and maintenance characterization positioning compensation value of a certain rail transit line is lower than the first operation and maintenance characterization positioning compensation threshold, then the first operation and maintenance characterization positioning compensation value of the rail transit line is used as the second operation and maintenance characterization value of the rail transit line, thereby obtaining the second operation and maintenance characterization value of each rail transit line.

[0024] Furthermore, the method of locating the predicted fault points of each transportation line in the rail transit network according to the second operation and maintenance characterization value of each rail transit line and issuing an early warning specifically comprises the following process: extracting the second operation and maintenance characterization value of each rail transit line, and comparing it with the second operation and maintenance characterization threshold of each rail transit line stored in the database; if the second operation and maintenance characterization value of a rail transit line is higher than or equal to the second operation and maintenance characterization threshold of each rail transit line, then the rail transit line is marked as a predicted fault point, and an early warning is issued.

[0025] Furthermore, the first maintenance positive value of each rail transit line has the following specific analysis conditions:

[0026] ;

[0027] In the formula, represents the first positive value of maintenance of the ith rail transit line, represents the signal compensation value of the i-th rail transit line, represents the track environment compensation value of the i-th rail transit line, Indicates the weight factor corresponding to the set signal compensation value, It represents the weight factor corresponding to the set track environment compensation value, i represents the number of each rail transit line, , n represents the total number of rail transit lines.

[0028] The present invention has the following beneficial effects:

[0029] (1) The present invention provides an intelligent fault prediction model based on machine learning and big data analysis. First, the data is trained through historical operation and maintenance data and convolutional neural network models to predict the first operation and maintenance characterization value of each rail transit line; then, combined with real-time monitoring and analysis of multimodal data of rail transit lines, the fault point can be accurately located and an early warning can be issued. By dynamically acquiring and analyzing historical and real-time data, training and adjusting the convolutional neural network model, potential faults can be discovered earlier, the fault point can be accurately located, and the safety and operation and maintenance efficiency of the rail transit system can be improved, so that the fault warning is not only more timely, but also can be adaptively adjusted according to different conditions;

[0030] (2) The present invention extracts and preprocesses the historical operation and maintenance data set of the rail transit network, and inputs it into the convolutional neural network model for training, thereby predicting the first operation and maintenance characterization value of each rail transit line. Specifically, the historical data is first preprocessed, cleaned and standardized, and then input into the convolutional neural network model for training, and finally the first operation and maintenance characterization value of each line is output, which helps to improve the accuracy of fault prediction and can also identify possible abnormal situations in advance, thereby providing real-time warnings for operation and maintenance personnel;

[0031] (3) By comprehensively collecting and processing multi-modal monitoring data, the system can accurately analyze the operating status of rail transit lines from different dimensions and identify potential abnormalities or failure risks. These compensation values ​​provide richer input data for the convolutional neural network model, which helps to improve the accuracy and reliability of fault prediction;

[0032] (4) By combining multi-dimensional operation and maintenance data and dynamic compensation mechanisms, the present invention can accurately reflect the health status of rail transit lines in real time and issue early warnings before failures occur. By using the convolutional neural network model, the impact of sudden failures on operations can be minimized, improving the safety, reliability and efficiency of the rail transit system.

[0033] Of course, any product implementing the present invention does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 It is a schematic diagram of the modules of the present invention. DETAILED DESCRIPTION

[0035] 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.

[0036] In the description of the present invention, it should be understood that the terms "opening", "upper", "lower", "thickness", "top", "middle", "length", "inside", "all around" and the like indicating orientation or positional relationship are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the components or elements referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as limiting the present invention.

[0037] See also Figure 1 As shown, an embodiment of the present invention provides an intelligent fault prediction model based on machine learning and big data analysis, including a data acquisition and analysis module, which is used to obtain the historical operation and maintenance data set of the rail transit network in the big data storage, import the historical operation and maintenance data set of the rail transit network into the convolutional neural network model for training, and predict the first operation and maintenance characterization value of each rail transit line based on the trained convolutional neural network model.

[0038] It should be noted that the data acquisition and analysis module also includes a sensor network, which is used to build a sensor network covering key parts of the rail transit network and collect equipment operation status data in real time. Data access is used to support multiple data access methods to achieve real-time data transmission and reception. Data storage is used to use a distributed database to store massive data and improve data storage and access efficiency.

[0039] The rail transit operation and maintenance data analysis module is used to monitor and obtain the multimodal data of each rail transit line in the rail transit network, analyze the multimodal data of each rail transit line, obtain the first operation and maintenance characterization positioning compensation value of each rail transit line, and correct and update the first operation and maintenance characterization value of each rail transit line.

[0040] The fault point prediction and positioning module is used to obtain the second operation and maintenance characterization value of each rail transit line based on the first operation and maintenance characterization positioning compensation value of each rail transit line and the updated first operation and maintenance characterization value of each rail transit line, and locate the predicted fault points of each transportation line of the rail transit network according to the second operation and maintenance characterization value of each rail transit line and issue an early warning.

[0041] Specifically, the first operation and maintenance characterization value of each rail transit line is predicted based on the trained convolutional neural network model. The specific process is: extracting the historical operation and maintenance data set of the rail transit network, preprocessing the historical operation and maintenance data set of the rail transit network, and importing it into the model for training, and outputting the first operation and maintenance characterization value of each rail transit line.

[0042] It should be noted that the specific process of importing the historical operation and maintenance data set of the rail transit network into the convolutional neural network (CNN) model for training includes the following steps: First, collect and organize the historical operation and maintenance data set of the rail transit network. The data set contains multi-dimensional information, such as train operation data (including train speed, acceleration and deceleration, train position data, etc.), track environmental parameters (including track temperature, track curvature and slope, track wear degree, etc.), signal system status (including communication network status, signal system error log, etc.). Then, preprocess these data, including denoising, normalization, filling missing values ​​and other operations to ensure the quality and consistency of the data. Next, convert the preprocessed data into a format suitable for input to CNN, usually converting the data into a two-dimensional or three-dimensional tensor form. Subsequently, the data is input into the convolutional neural network for training. CNN automatically extracts features through multi-layer convolution and pooling operations, and performs final classification or regression prediction through the fully connected layer. During the training process, the error between the predicted result and the actual label is calculated using the preset loss function, and the parameters in the network are adjusted through the back propagation algorithm to gradually optimize the model. Finally, the convolutional neural network model obtained through training.

[0043] It should be noted that the loss function compares the difference between the predicted value output by the model and the true label. In this embodiment, the loss function is the mean square error (MSE), which measures the accuracy of the model by calculating the average of the squared difference between the predicted value and the true value.

[0044] Specifically, the multimodal data of each rail transit line is analyzed, and the specific process is: extracting the multimodal data of each rail transit line in the rail transit network, including each train travel data, signal system data and each track environment status data.

[0045] A first operation and maintenance characterization positioning compensation value of each rail transit line is obtained according to the travel data of each train, and the first operation and maintenance characterization positioning compensation value of each rail transit line is used to evaluate the smoothness of train operation of each rail transit line.

[0046] Specifically, the first operation and maintenance characterization value of each rail transit line is corrected and updated, and the specific process is: the signal compensation value of each rail transit line is obtained according to the signal system data, and the signal compensation value of each rail transit line is used to evaluate the reliability of the rail signal system.

[0047] Obtaining a track environment compensation value for each rail transit line according to each track environment status data, wherein the track environment compensation value for each rail transit line is used to evaluate the interference degree of external environmental conditions on the safety of each rail transit line;

[0048] Based on the signal compensation value of each rail transit line and the track environment compensation value of each rail transit line, the first operation and maintenance positive value of each rail transit line is obtained. The first operation and maintenance positive value of each rail transit line is used to evaluate the degree of interference correction of external environmental conditions on the safety of each rail transit line.

[0049] It should be noted that the first maintenance value of each rail transit line is positive, and the specific analysis conditions are:

[0050] ;

[0051] In the formula, Indicates the first positive value of the maintenance of the i-th rail transit line represents the signal compensation value of the i-th rail transit line, Indicates the track environment compensation value of the i-th rail transit line Indicates the weight factor corresponding to the set signal compensation value, It represents the weight factor corresponding to the set track environment compensation value, i represents the number of each rail transit line, , n represents the total number of rail transit lines.

[0052] It should be noted that the weight factor corresponding to the signal compensation value represents the degree of influence of the signal system compensation value on the safety and operation efficiency evaluation value of the rail transit line. This corresponding relationship is determined by a preset mapping relationship. For example, the actual signal strength of the signal system, the signal light change time, etc. form a mapping set with the preset signal compensation value stored in the database. By inputting the real-time monitored signal compensation value into the mapping set, its corresponding weight factor can be obtained, which is used to further evaluate the stability and response speed of the signal system.

[0053] It should be noted that the weight factor corresponding to the track environment compensation value represents the degree of influence of the track environment state compensation value on the structural health and operational stability assessment value of the rail transit line. This correspondence is determined by a preset mapping relationship. For example, the temperature, vibration frequency, distortion and other data of the track form a mapping set with the preset track environment compensation value stored in the database. By inputting the track environment compensation value monitored in real time into the mapping set, its corresponding weight factor can be obtained, which is used to further analyze the impact of the track environment on the overall operational stability and safety of the system.

[0054] The first operation and maintenance positivity value of each rail transit line is extracted, and compared with the first operation and maintenance positivity threshold value of each rail transit line stored in the database; if the first operation and maintenance positivity value of a rail transit line is higher than or equal to the first operation and maintenance positivity threshold value of each rail transit line, the first operation and maintenance characterization value of the rail transit line and the first operation and maintenance positivity value of each rail transit line are weighted summed to obtain a corrected value of the first operation and maintenance characterization value of the rail transit line, and the corrected value of the first operation and maintenance characterization value of the rail transit line is updated as the first operation and maintenance positivity value of the rail transit line; if the first operation and maintenance positivity value of each rail transit line is lower than the first operation and maintenance positivity threshold value of each rail transit line, the first operation and maintenance characterization value of each rail transit line is directly used as the updated first operation and maintenance characterization value of each rail transit line.

[0055] It should be noted that the specific process of weighted summing the first operation and maintenance characterization value of the rail transit line and the first operation and maintenance proper value of each rail transit line is to extract the weight factors corresponding to the first operation and maintenance characterization value of the rail transit line and the first operation and maintenance proper value of each rail transit line from the database, multiply the first operation and maintenance characterization value of the rail transit line by the corresponding weight factor, and then multiply the first operation and maintenance proper value of each rail transit line by the corresponding weight factor and then add the two values.

[0056] It should be noted that the weight factor corresponding to the first operation and maintenance punctuality of the rail transit line is determined by a preset mapping relationship. For example, the first operation and maintenance punctuality of the rail transit line and the preset correction value stored in the database form a mapping set. By inputting the first operation and maintenance punctuality of each rail transit line obtained in real time into the mapping set, the weight factor corresponding to the first operation and maintenance punctuality of each rail transit line can be obtained through the preset weight mapping relationship.

[0057] It should be noted that the weight factor corresponding to the first operation and maintenance characterization value of the rail transit line is determined by a preset mapping relationship. For example, the first operation and maintenance characterization value of the rail transit line and the preset operation and maintenance characterization value stored in the database form a mapping set. By inputting the first operation and maintenance characterization value of the rail transit line obtained in real time into the mapping set, the weight factor corresponding to the first operation and maintenance characterization value of the corresponding rail transit line can be obtained through the preset weight mapping relationship.

[0058] Specifically, the first operation and maintenance characterization positioning compensation value of each rail transit line is obtained according to the travel data of each train. The specific process is: preset a monitoring time period, and extract the travel data of each train in each rail transit line during the monitoring time period, including the average operating speed, average acceleration, average traction power and average braking force of each train.

[0059] It should be noted that the average running speed of each train is obtained by collecting the train's location information (such as GPS coordinates or track measurement point data) during the monitoring period, obtaining the train's travel distance at different time points, and calculating the speed according to the time interval. Finally, the average value of all instantaneous speeds in the time period is taken. The average acceleration is calculated based on the real-time speed data of the train, the speed change between each two measurement points, and then divided by the time interval to obtain the instantaneous acceleration. Then, the average value of all accelerations in the entire monitoring period is taken. The traction power is determined by the product of the traction force and the train speed. First, the power of the traction system is determined by the current and voltage. The power of the traction system is obtained by multiplying the current of the traction motor with the voltage of the traction motor and the correction factor of the motor efficiency and power transmission loss stored in the database. Finally, the average value of all traction powers in the time period is taken. The average braking force is obtained by monitoring the deceleration process of the train, multiplying the mass of the train by the deceleration (i.e., negative acceleration), and obtaining the braking force at each moment in the monitoring period. Finally, the average braking force is calculated.

[0060] The average running speed of each train on each rail transit line is extracted, and the track design speed stored in the database is extracted. The average running speed of each train on each rail transit line is processed with the track design speed and then the absolute value is processed to obtain the speed deviation value of each train on each rail transit line.

[0061] The speed deviation value of each train in each rail transit line, the average acceleration, the average traction power and the average braking force of each train are extracted, and the first operation and maintenance characterization positioning compensation value of each rail transit line is obtained after processing.

[0062] It should be noted that the first operation and maintenance characterization positioning compensation value of each rail transit line has the following specific analysis conditions:

[0063] ;

[0064] In the formula, represents the first operation and maintenance characterization positioning compensation value of the i-th rail transit line, represents the speed deviation of the jth train in the i-th rail transit line, represents the average acceleration of the jth train in the i-th rail transit line, represents the average traction power of the jth train in the i-th rail transit line, represents the average braking force of the jth train in the i-th rail transit line, Indicates the set reference speed deviation value. Indicates the set reference acceleration, Indicates the set reference traction power, Indicates the set reference braking force, Indicates the correction factor corresponding to the set driving speed deviation value, Indicates the correction factor corresponding to the set average acceleration, Indicates the correction factor corresponding to the set average traction power, It indicates the correction factor corresponding to the set average braking force, i indicates the number of each rail transit line, , n represents the total number of rail transit lines, j represents the number of each train, , m represents the total number of trains.

[0065] In a specific embodiment, the speed deviation value, average acceleration, average traction power and average braking force of each train in each rail transit line do not exist in isolation. First, the speed deviation value of the train directly affects the average acceleration. When the actual speed of the train deviates from the predetermined speed, it usually means that the speed needs to be adjusted by acceleration or braking, resulting in a change in acceleration. If the speed deviation is large, the train may need a higher acceleration to make up for the speed difference, which will increase the demand for traction power. Secondly, the average acceleration is closely related to the traction power. When the train accelerates, the traction system needs to provide a larger traction force, and the magnitude of the traction force is proportional to the traction power. Therefore, a larger acceleration is usually accompanied by a higher traction power. If the acceleration process of the train is too intense, the traction power may increase significantly, resulting in increased energy consumption, which in turn affects the efficiency of the entire rail transit system. Furthermore, the traction power and the braking force may show a certain inverse relationship in some cases. For example, in the acceleration stage of the train, the traction power increases, while the braking force is basically at a low level; while in the deceleration stage of the train, the traction power will decrease, but the braking force will increase. Under normal operating conditions, these two parameters are complementary: traction power ensures the acceleration and continuous operation of the train, while braking force ensures the safe deceleration of the train. Therefore, their coordination is very important. Unbalanced traction power and braking force may cause unstable train operation or increase energy consumption.

[0066] In a specific embodiment, the correction factor corresponding to the speed deviation value generally ranges from 0 to 1. When in use, the correction factor corresponding to the speed deviation value can be directly obtained from the database. The correspondence between the speed deviation value and its corresponding correction factor is determined by a pre-set mapping table. For example, a mapping table between the speed deviation value and the correction factor is constructed. The train speed deviation value detected in real time can be input into the mapping table to quickly obtain the corresponding correction factor, thereby helping to evaluate the running status of the train and determine whether operational adjustment or maintenance is required.

[0067] In a specific embodiment, the correction factor corresponding to the average acceleration generally ranges from 0 to 1. When in use, the correction factor corresponding to the average acceleration can be directly obtained from the database. The correspondence between the average acceleration and its corresponding correction factor is determined by a pre-set mapping table. For example, by constructing a mapping table between the average acceleration and the correction factor. The real-time monitored train acceleration value can be input into the mapping table to quickly obtain the correction factor corresponding to the average acceleration, so as to evaluate the acceleration process of the train, optimize energy efficiency, and determine whether there is an abnormal situation.

[0068] In a specific embodiment, the correction factor corresponding to the average traction power generally ranges from 0 to 1. When in use, the correction factor corresponding to the average traction power can be directly obtained from the database. The correspondence between the average traction power and its corresponding correction factor is determined by a pre-set mapping table. For example, by analyzing the energy efficiency, driving stability and power system health of the train at different traction power levels, a mapping table between the average traction power and the correction factor is constructed. The real-time measured train traction power value can be input into the mapping table to quickly obtain the correction factor corresponding to the average traction power, thereby helping to evaluate the working efficiency of the train traction system and determine whether energy efficiency optimization or system maintenance is required.

[0069] In a specific embodiment, the correction factor corresponding to the average braking force generally ranges from 0 to 1. When in use, the correction factor corresponding to the average braking force can be directly obtained from the database. The correspondence between the average braking force and its corresponding correction factor is determined by a pre-set mapping table. For example, by analyzing the deceleration performance of the train under different braking forces, the temperature rise and wear of the braking system, a mapping table between the average braking force and the correction factor is constructed. The train braking force value monitored in real time can be input into the mapping table to quickly obtain the correction factor corresponding to the average braking force, thereby helping to evaluate the performance of the braking system and determine whether calibration or maintenance is required.

[0070] Specifically, the signal compensation value of each rail transit line is obtained, and the specific process is: extracting the signal system data of each rail transit line in each monitoring time period, including signal reception strength, change time of signal light status and position coordinates of each train.

[0071] The position coordinates of each train are extracted to obtain the time interval between adjacent trains, and based on the signal reception strength of each rail transit line and the average change time of the signal light status, the signal compensation value of each rail transit line is obtained through processing.

[0072] It should be noted that signal reception strength refers to the strength of radio signals or electromagnetic waves received by trains or signal receiving equipment on rail transit lines. Special signal receiving devices (such as radio receivers, GPS receivers, etc.) are installed on rail transit lines. These devices can regularly monitor and record signal strength. Receivers are usually connected to trains or fixed track facilities. For the entire track section, the signal reception strength of a rail transit line can usually be obtained by taking the standard deviation of multiple points in the area.

[0073] It should be noted that the time interval between adjacent trains is obtained by determining the position information of the train relative to a fixed starting point through the track circuit or sensor's odometer. The position information of each train (usually obtained through the GPS system, track circuit, or signal system) indicates the specific position of the train at a certain moment, expressed in longitude and latitude or mileage on the track (such as kilometers from the starting point). By comparing the mileage difference between two adjacent trains, the distance between them is obtained, and the difference processing is performed based on the position information and the timestamp to obtain the time interval between adjacent trains.

[0074] It should be noted that adjacent trains usually refer to adjacent trains on the same track. In a rail transit system, trains travel on the same track in sequence, so "adjacent trains" refer to the adjacent positions of two trains on the track, and in this embodiment, refers to two trains that are relatively close in time.

[0075] It should be noted that the signal compensation value of each rail transit line has the following specific analysis conditions:

[0076] ;

[0077] In the formula, represents the signal compensation value of the i-th rail transit line, represents the time interval between the kth adjacent trains in the i-th rail transit line, represents the signal reception strength of the i-th rail transit line, represents the average change time of the signal light status in the i-th rail transit line, Indicates the reference time interval between set trains, Indicates the set reference signal reception strength. Indicates the change time of the set reference signal light state. Indicates the correction factor corresponding to the set time interval between trains, Indicates the correction factor corresponding to the set signal reception strength. It represents the correction factor corresponding to the average change time of the set signal light state, i represents the number of each rail transit line, , n represents the total number of rail transit lines, K represents the number of adjacent trains, , b represents the total number of adjacent trains.

[0078] In a specific embodiment, the time interval between adjacent trains, the signal reception strength of each rail transit line, and the average change time of the signal light state do not exist in isolation, but are interrelated and affect each other. First, the time interval between adjacent trains directly affects the operating efficiency and safety of the train. When the time interval between two trains is short, it means that the distance between them is close, which may cause the signal system to need to be adjusted more frequently, increasing the demand for signal reception strength. At the same time, a shorter time interval may also require the signal light state to change more frequently to ensure the smooth operation of the train. Secondly, the signal reception strength is closely related to the average change time of the signal light state. When the signal reception strength is high, the train can receive the signal more quickly and accurately, thereby reducing the delay in the change of the signal light state and improving the response efficiency of the signal light state. Stronger signal reception strength means that the train can obtain the signal state ahead earlier and respond in time to avoid shortening the driving interval or sudden braking operation due to signal delay, and maintain a reasonable time interval between trains. Furthermore, there is a certain influence relationship between the average change time of the signal light state and the time interval between trains. When the signal light changes for a long time, the train may encounter a longer waiting time when approaching the signal light, resulting in an increase in the time interval, which affects the scheduling and operation rhythm of the trains. When the signal light changes for a short time, the train can pass the signal point more quickly, thus maintaining a shorter time interval and improving operation efficiency.

[0079] In a specific embodiment, the correction factor corresponding to the time interval between trains generally ranges from 0 to 1. When in use, the correction factor corresponding to the time interval between trains can be directly obtained from the database. The relationship between the time interval and its corresponding correction factor is determined by a pre-set mapping table. For example, by constructing a mapping table between the time interval and the correction factor. The time interval value between trains monitored in real time can be input into the mapping table to quickly obtain the correction factor corresponding to the time interval.

[0080] In a specific embodiment, the correction factor corresponding to the signal reception strength generally ranges from 0 to 1. When in use, the correction factor corresponding to the signal reception strength can be directly obtained from the database. The relationship between the signal reception strength and its corresponding correction factor is determined by a pre-set mapping table. For example, by constructing a mapping table between the signal reception strength and the correction factor. The signal reception strength value monitored in real time can be input into the mapping table to quickly obtain the correction factor corresponding to the signal reception strength, thereby helping to evaluate the reliability and stability of the signal system.

[0081] In a specific embodiment, the correction factor corresponding to the average change time of the signal light state generally ranges from 0 to 1. When in use, the correction factor corresponding to the average change time of the signal light state can be directly obtained from the database. The relationship between the average change time of the signal light state and its corresponding correction factor is determined by a pre-set mapping table. For example, by constructing a mapping table between the signal light state change time and the correction factor. The signal light state change time value monitored in real time can be input into the mapping table to quickly obtain the correction factor corresponding to the average change time of the signal light state, thereby helping to evaluate the performance of the signal light control system.

[0082] Specifically, the track environment compensation value of each rail transit line is obtained according to the track environment status data. The specific process is: extracting the track environment status data of each rail transit line in the monitoring time period, including the average temperature of the track, the average vibration frequency of the track and the distortion of the track, and obtaining the track environment compensation value of each rail transit line after processing.

[0083] It should be noted that temperature sensors, accelerometers and track deviation meters, such as thermocouples or RTDs (platinum resistance temperature sensors), are installed at different locations on the track (such as the track surface or the bottom) to obtain the temperature monitored by each temperature sensor of each rail transit line, the average vibration frequency of the track and the distortion of the track. The average temperature usually refers to the average value of the temperature monitored by each temperature sensor of each rail transit line during the monitoring period. The distortion of the track refers to the degree of deviation of the track from the normal geometric shape, which is usually described by the vertical and horizontal deviations of both sides of the track relative to the center line. The distortion of the track usually refers to the average value of the track deviation measured by the distortion sensor or deviation meter of each rail transit line during the monitoring period. The average vibration frequency of the track usually refers to the average value of the vibration frequency monitored by the vibration sensor of each rail transit line during the monitoring period.

[0084] It should be noted that the specific analysis conditions for the track environment compensation value of each rail transit line are as follows:

[0085] ;

[0086] In the formula, represents the track environment compensation value of the i-th rail transit line, represents the average temperature of the track in the i-th rail transit line, represents the average vibration frequency of the ith rail transit line, represents the twisting degree of the track of the ith rail transit line, represents the reference average temperature of the set track, represents the average vibration frequency of the set reference orbit, Indicates the distortion of the set reference track, Indicates the correction factor corresponding to the average temperature of the set track, Indicates the correction factor corresponding to the set vibration frequency, It indicates the correction factor corresponding to the set track distortion, i indicates the number of each rail transit line, , n represents the total number of rail transit lines.

[0087] It should be noted that, in a specific embodiment, the average temperature of the track, the average vibration frequency of the track, and the distortion of the track do not exist in isolation, but are mutually influential and closely related. First, when the track temperature rises, the rail will expand thermally, and when the temperature drops, it will shrink. This temperature change may cause the deformation of the track, which in turn affects the distortion of the track. Secondly, the distortion of the track is closely related to the average vibration frequency. When the track is distorted due to temperature changes, the wheel-rail contact force will change when the train passes through these areas, causing uneven vibration. If the track distortion is large, the wheel-rail interaction force during the train operation will also become uneven, which may cause the track vibration frequency to increase, and even induce resonance, affecting driving comfort and safety. Furthermore, the average vibration frequency of the track will in turn affect the temperature change of the track. Continuous high-frequency vibration may intensify the friction between the track and the wheel-rail, resulting in local temperature rise, especially in sections with frequent braking or high-load lines, where this effect is particularly significant.

[0088] In a specific embodiment, the correction factor corresponding to the average temperature of the track generally ranges from 0 to 1. When in use, the correction factor corresponding to the average temperature of the track can be directly obtained from the database. The correspondence between the average temperature of the track and its corresponding correction factor is determined by a pre-set mapping table. For example, by analyzing the expansion and contraction of the track under different temperature conditions, the thermal stress changes of the rail material, and the deformation trend of the track, a mapping table between the average temperature of the track and the correction factor is constructed. The track temperature value monitored in real time can be input into the mapping table to quickly obtain the correction factor corresponding to the average temperature of the track.

[0089] In a specific embodiment, the correction factor corresponding to the average vibration frequency of the track generally ranges from 0 to 1. When in use, the correction factor corresponding to the track vibration frequency can be directly obtained from the database. The correspondence between the track vibration frequency and its corresponding correction factor is determined by a pre-set mapping table. For example, by analyzing the structural fatigue of the track at different vibration frequencies, the degree of wear of the track contact surface, and the resonance effect that may be caused, a mapping table between the track vibration frequency and the correction factor is constructed. The track vibration data monitored in real time can be input into the mapping table to quickly obtain the correction factor corresponding to the average vibration frequency of the track.

[0090] In a specific embodiment, the correction factor corresponding to the torsion of the track generally ranges from 0 to 1. When in use, the correction factor corresponding to the torsion of the track can be directly obtained from the database. The correspondence between the torsion of the track and its corresponding correction factor is determined by a pre-set mapping table. For example, by analyzing the train running stability, wheel-rail contact pressure distribution and the influence of track deformation on train vibration under different torsion conditions, a mapping table between the torsion of the track and the correction factor is constructed. The track torsion data monitored in real time can be input into the mapping table to quickly obtain the correction factor corresponding to the torsion of the track.

[0091] Specifically, the second operation and maintenance characterization value of each rail transit line is obtained. The specific process is: extract the updated first operation and maintenance characterization value of each rail transit line, and compare it with the set first operation and maintenance characterization positioning compensation threshold; if the first operation and maintenance characterization positioning compensation value of a certain rail transit line is higher than or equal to the first operation and maintenance characterization positioning compensation threshold, then the first operation and maintenance characterization positioning compensation value of the rail transit line is weightedly summed with the first operation and maintenance characterization value of the rail transit line to obtain the second operation and maintenance characterization value of the rail transit line; if the first operation and maintenance characterization positioning compensation value of a certain rail transit line is lower than the first operation and maintenance characterization positioning compensation threshold, then the first operation and maintenance characterization positioning compensation value of the rail transit line is used as the second operation and maintenance characterization value of the rail transit line, thereby obtaining the second operation and maintenance characterization value of each rail transit line.

[0092] It should be noted that when the model is trained or predicted, it will generate the corresponding first operation and maintenance characterization positioning compensation value of each rail transit line based on the input data (such as train operation parameters, track environment data, etc.).

[0093] It should be noted that the specific process of weighted summing the first operation and maintenance characterization positioning compensation value of the transportation line and the first operation and maintenance characterization value of the rail transit line is to extract the weight factors of the first operation and maintenance characterization positioning compensation value of the transportation line and the first operation and maintenance characterization value of the rail transit line from the database respectively, multiply the first operation and maintenance characterization positioning compensation value of the transportation line by the weight factor of the extracted first operation and maintenance characterization positioning compensation value of the transportation line, and then multiply the first operation and maintenance characterization value of the rail transit line by the weight factor of the first operation and maintenance characterization value of the rail transit line and the first operation and maintenance characterization value of the rail transit line and then add them together.

[0094] It should be noted that the weight factor corresponding to the first operation and maintenance characterization positioning compensation value of the traffic line is determined by a preset mapping relationship. For example, the first operation and maintenance characterization positioning compensation value of the traffic line and the preset compensation value stored in the database form a mapping set. By inputting the first operation and maintenance characterization positioning compensation value of the traffic line obtained in real time into the mapping set, the weight factor corresponding to the first operation and maintenance characterization positioning compensation value of the corresponding traffic line can be obtained through the preset weight mapping relationship.

[0095] It should be noted that the weight factor corresponding to the first operation and maintenance characterization value of the rail transit line is determined by a preset mapping relationship. For example, the first operation and maintenance characterization value of the rail transit line and the preset operation and maintenance characterization value stored in the database form a mapping set. By inputting the first operation and maintenance characterization value of the rail transit line obtained in real time into the mapping set, the weight factor corresponding to the first operation and maintenance characterization value of the corresponding rail transit line can be obtained through the preset weight mapping relationship.

[0096] Specifically, the predicted fault points of each transportation line in the rail transit network are located and an early warning is issued according to the second operation and maintenance characterization value of each rail transit line. The specific process is: the second operation and maintenance characterization value of each rail transit line is extracted, and compared with the second operation and maintenance characterization threshold of each rail transit line stored in the database. If the second operation and maintenance characterization value of a rail transit line is higher than or equal to the second operation and maintenance characterization threshold of each rail transit line, the rail transit line is marked as a predicted fault point, and an early warning is issued.

[0097] It should be noted that early warning prompts include automatically generating fault prediction reports, including equipment status analysis, fault trend prediction, maintenance suggestions, etc.

[0098] It should be noted that an intelligent fault prediction model based on machine learning and big data analysis also includes a database for storing track design speed, second operation and maintenance characterization thresholds of each rail transit line, etc.

[0099] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0100] The preferred embodiments of the present invention disclosed above are only used to help illustrate the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to the specific implementation methods described. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. An intelligent fault prediction system based on machine learning and big data analysis, characterized in that: include: A data acquisition and analysis module is used to obtain the historical operation and maintenance data set of the rail transit network in the big data storage, import the historical operation and maintenance data set of the rail transit network into the convolutional neural network model for training, and predict the first operation and maintenance representation value of each rail transit line based on the trained convolutional neural network model; A rail transit operation and maintenance data analysis module, used to monitor and obtain multimodal data of each rail transit line in the rail transit network, analyze the multimodal data of each rail transit line, obtain a first operation and maintenance characterization positioning compensation value of each rail transit line, and correct and update the first operation and maintenance characterization value of each rail transit line; A fault point prediction and positioning module is used to obtain a second operation and maintenance characterization value of each rail transit line based on the first operation and maintenance characterization positioning compensation value of each rail transit line and the updated first operation and maintenance characterization value of each rail transit line, and locate the predicted fault points of each transportation line of the rail transit network according to the second operation and maintenance characterization value of each rail transit line and issue an early warning; The multimodal data of each rail transit line is analyzed to obtain the first operation and maintenance characterization positioning compensation value of each rail transit line. The specific process is as follows: Extract multimodal data of each rail transit line in the rail transit network, including each train running data, signal system data and each track environment status data; A first operation and maintenance characterization positioning compensation value of each rail transit line is obtained according to the travel data of each train, and the first operation and maintenance characterization positioning compensation value of each rail transit line is used to evaluate the smoothness of train operation of each rail transit line.

2. The intelligent fault prediction system based on machine learning and big data analysis as claimed in claim 1, characterized in that: The first operation and maintenance characterization value of each rail transit line is predicted based on the trained convolutional neural network model. The specific process is as follows: The historical operation and maintenance dataset of the rail transit network is extracted, and the historical operation and maintenance dataset of the rail transit network is preprocessed and imported into the model for training, and the first operation and maintenance representation value of each rail transit line is obtained as output.

3. The intelligent fault prediction system based on machine learning and big data analysis as claimed in claim 1, characterized in that: The specific process of correcting and updating the first operation and maintenance characterization value of each rail transit line is as follows: Obtaining a signal compensation value for each rail transit line according to the signal system data, wherein the signal compensation value for each rail transit line is used to evaluate the impact of the operating status of the signal system in the rail transit line on the travel of the train; Obtaining a track environment compensation value for each rail transit line according to each track environment state data, wherein the track environment compensation value for each rail transit line is used to evaluate the impact of external environmental conditions on the travel of trains in the rail transit line; Based on the signal compensation value of each rail transit line and the track environment compensation value of each rail transit line, a first operation and maintenance positive value of each rail transit line is obtained, wherein the first operation and maintenance positive value of each rail transit line is used to evaluate the degree of interference correction of external environmental conditions on the safety of each rail transit line; The first operation and maintenance positivity value of each rail transit line is extracted, and compared with the first operation and maintenance positivity threshold value of each rail transit line stored in the database; if the first operation and maintenance positivity value of a rail transit line is higher than or equal to the first operation and maintenance positivity threshold value of each rail transit line, the first operation and maintenance characterization value of the rail transit line and the first operation and maintenance positivity value of each rail transit line are weighted summed to obtain a corrected value of the first operation and maintenance characterization value of the rail transit line, and the corrected value of the first operation and maintenance characterization value of the rail transit line is updated as the first operation and maintenance positivity value of the rail transit line; if the first operation and maintenance positivity value of each rail transit line is lower than the first operation and maintenance positivity threshold value of each rail transit line, the first operation and maintenance characterization value of each rail transit line is directly used as the updated first operation and maintenance characterization value of each rail transit line.

4. The intelligent fault prediction system based on machine learning and big data analysis as claimed in claim 1, characterized in that: The first operation and maintenance characterization positioning compensation value of each rail transit line is obtained according to the travel data of each train, and the specific process is as follows: Preset a monitoring time period, and extract the travel data of each train in each rail transit line during the monitoring time period, including the average running speed, average acceleration, average traction power and average braking force of each train; Extract the average running speed of each train on each rail transit line, and extract the track design speed stored in the database, perform difference processing on the average running speed of each train on each rail transit line and the track design speed, and then perform absolute value processing to obtain the running speed deviation value of each train on each rail transit line; The speed deviation value of each train in each rail transit line, the average acceleration, the average traction power and the average braking force of each train are extracted, and the first operation and maintenance characterization positioning compensation value of each rail transit line is obtained after processing.

5. The intelligent fault prediction system based on machine learning and big data analysis as claimed in claim 3, characterized in that: The specific process of obtaining the signal compensation value of each rail transit line is as follows: Extract signal system data of each rail transit line in each monitoring time period, including signal reception strength, signal light status change time and position coordinates of each train; The position coordinates of each train are extracted to obtain the time interval between adjacent trains, and based on the signal reception strength of each rail transit line and the average change time of the signal light status, the signal compensation value of each rail transit line is obtained through processing.

6. The intelligent fault prediction system based on machine learning and big data analysis as claimed in claim 3, characterized in that: The track environment compensation value of each rail transit line is obtained according to each track environment state data, and the specific process is as follows: During the monitoring period, the environmental status data of each track in each rail transit line are extracted, including the average temperature of the track, the average vibration frequency of the track and the distortion of the track, and the track environmental compensation value of each rail transit line is obtained after processing.

7. The intelligent fault prediction system based on machine learning and big data analysis as claimed in claim 1, characterized in that: The specific process of obtaining the second operation and maintenance characterization value of each rail transit line is as follows: The updated first operation and maintenance characterization value of each rail transit line is extracted and compared with the set first operation and maintenance characterization positioning compensation threshold. If the first operation and maintenance characterization positioning compensation value of a rail transit line is higher than or equal to the first operation and maintenance characterization positioning compensation threshold, the first operation and maintenance characterization positioning compensation value of the rail transit line is weightedly summed with the first operation and maintenance characterization value of the rail transit line to obtain the second operation and maintenance characterization value of the rail transit line. If the first operation and maintenance characterization positioning compensation value of a rail transit line is lower than the first operation and maintenance characterization positioning compensation threshold, the first operation and maintenance characterization positioning compensation value of the rail transit line is used as the second operation and maintenance characterization value of the rail transit line, thereby obtaining the second operation and maintenance characterization value of each rail transit line.

8. The intelligent fault prediction system based on machine learning and big data analysis as claimed in claim 1, characterized in that: The specific process of locating the predicted fault points of each transportation line of the rail transit network and issuing an early warning according to the second operation and maintenance characterization value of each rail transit line is as follows: The second operation and maintenance characterization value of each rail transit line is extracted and compared with the second operation and maintenance characterization threshold of each rail transit line stored in the database. If the second operation and maintenance characterization value of a rail transit line is higher than or equal to the second operation and maintenance characterization threshold of each rail transit line, the rail transit line is marked as a predicted fault point and an early warning prompt is issued.

9. The intelligent fault prediction system based on machine learning and big data analysis as claimed in claim 3, characterized in that: The first maintenance positive value of each rail transit line, the specific analysis conditions are: ； In the formula, Indicates the first positive value of the maintenance of the i-th rail transit line represents the signal compensation value of the i-th rail transit line, Indicates the track environment compensation value of the i-th rail transit line Indicates the weight factor corresponding to the set signal compensation value, It represents the weight factor corresponding to the set track environment compensation value, i represents the number of each rail transit line, , n represents the total number of rail transit lines.

Images