Railway facility state prediction method based on working condition identification and multi-task learning

Abstract

The invention provides a railway facility state prediction method based on working condition identification and multi-task learning, and belongs to the technical field of railway safe operation, and the method comprises the steps: collecting multi-source data in the railway facility operation process through a sensor network and monitoring equipment, and carrying out the preprocessing; constructing a working condition feature extraction model in a mode of combining a convolutional neural network and a long-short-term memory network, and performing feature extraction on the preprocessed data to obtain working condition features; performing association mining on the working condition features, determining the number of task branches and a state prediction target of each task branch, and constructing a multi-task learning neural network model; inputting railway facility operation data collected in real time and current working condition information into the trained multi-task learning prediction model, and outputting state prediction results of railway facilities in different directions; and according to actual working condition changes and prediction effect feedback, structures and parameters of the working condition feature extraction model and the multi-task learning model are adjusted, and the accuracy of state prediction is ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of railway safe operation, and in particular to a railway facility status prediction method based on working condition identification and multi-task learning. Background Art

[0002] In the railway transportation system, the proper operation of railway facilities is crucial for ensuring transportation safety and efficiency. Traditional methods for predicting the status of railway facilities rely primarily on manual inspections and simple data analysis. Under manual inspections, personnel regularly conduct visual inspections and measure basic parameters of facilities such as track geometry and signal equipment voltage and current. This approach has significant drawbacks. Firstly, manual inspections are inefficient. Limited by manpower and time, they make it difficult to conduct frequent and comprehensive inspections of railway facilities, hindering the timely detection of potential minor damage or performance degradation. Secondly, manual judgments are subjective, and differences in experience and standards among inspectors can lead to inconsistent inspection results, compromising prediction accuracy. Furthermore, analytical methods based on simple statistical data can only process a single type of data and are unable to explore the complex correlations between facility operating data. They are also unable to adapt to the volatile operating conditions of railway facilities and cannot effectively predict changes in facility status under different operating conditions. With the development of railway transportation towards higher speeds and heavier loads, the operating environment of railway facilities has become increasingly complex. Traditional prediction methods are no longer able to meet the requirements for ensuring safe and efficient railway operations. A new prediction method is urgently needed to improve the accuracy and reliability of railway facility status prediction.

[0003] Therefore, the present invention proposes a railway facility status prediction method based on working condition identification and multi-task learning. Summary of the Invention

[0004] The present invention provides a railway facility status prediction method based on working condition identification and multi-task learning to solve the above-mentioned technical problems.

[0005] The present invention proposes a railway facility status prediction method based on working condition identification and multi-task learning, comprising:

[0006] Step 1: Use sensor networks and monitoring equipment to collect multi-source data during railway facility operation and perform pre-processing;

[0007] Step 2: Use a combination of convolutional neural network and long short-term memory network to build a working condition feature extraction model, and extract features from the preprocessed data to obtain working condition features;

[0008] Step 3: Perform association mining on working condition features, determine the number of task branches and the state prediction target for each task branch, and build a multi-task learning neural network model;

[0009] Step 4: Input the real-time collected railway facility operation data and current working condition information into the trained multi-task learning prediction model, and output the status prediction results of the railway facilities in different directions;

[0010] Step 5: Adjust the structure and parameters of the working condition feature extraction model and multi-task learning model based on actual working condition changes and prediction effect feedback.

[0011] Preferably, sensor networks and monitoring equipment are used to collect multi-source data during the operation of railway facilities and perform pre-processing, including:

[0012] Develop a sensor network deployment plan based on the distribution and operational characteristics of railway facilities;

[0013] The sensor network is activated, and each sensor collects railway facility operation data in real time at a set sampling frequency. Onboard sensors continuously collect train speed, load, and route data during train operation, and transmit this data to a ground-based data aggregation node via a dedicated railway communication network. The data aggregation node then performs preliminary integration and encoding of the received multi-source data.

[0014] The data collected by the data aggregation node is transmitted to the data processing center, and the collected data is cleaned using the data cleaning algorithm;

[0015] For problems where different types of data have different dimensions and numerical ranges, a normalization algorithm is used to process the cleaned data, and a one-hot encoding method is used to convert categorical data;

[0016] Through statistical analysis of data, missing values ​​in the data are identified and filled.

[0017] Preferably, a sensor network deployment plan is developed based on the distribution characteristics and operational characteristics of railway facilities, including:

[0018] Collect basic information on railway facilities and use data analysis tools to conduct in-depth analysis of the basic information to determine the key monitoring points and data requirements of each part of the railway facilities;

[0019] Screen matching sensor types based on key monitoring points and data requirements, and in combination with the operating environment characteristics of each key monitoring point;

[0020] Based on the distribution and monitoring needs of railway facilities, geographic information system technology is used to conduct preliminary planning of the sensor network on the electronic map of the railway line. According to the key parts of the track and the characteristics of operation data, the deployment locations of sensors matching the key monitoring points involved in the key parts are determined to form a sensor network deployment plan.

[0021] Preferably, step 2 includes:

[0022] Construct a CNN network to extract local features of the preprocessed data to obtain preliminary feature representation;

[0023] The preliminary feature representation is used as the input of the long short-term memory network and learned to mine the dynamic change patterns of the preprocessed data in the time dimension, and to extract the temporal features and contextual information from the preprocessed data;

[0024] The features output by the LSTM network are fused with some local features extracted by the CNN network in the middle layer to form a complete working condition feature vector containing the data's spatiotemporal information and dynamic change rules, and the working condition feature vector is output as the final extraction result.

[0025] Preferably, step 3 includes:

[0026] Use association rule mining algorithms to analyze working condition characteristics, calculate the support and confidence between different working condition characteristics, and find feature combinations with strong correlations;

[0027] Counting the number of combinations of the feature combinations, and determining the number of task branches and the state prediction target of each task branch in combination with the actual needs of railway facility state prediction;

[0028] Extract common features from all working condition features based on a shared underlying network;

[0029] Connecting the task branch to the shared underlying network;

[0030] The branch network of each task branch is screened according to the state prediction target, the general features are processed, the specific features related to the corresponding task branch are extracted, and the state prediction task is completed.

[0031] Preferably, determining the number of task branches includes:

[0032] Count the total number of single working condition features that have strong correlation;

[0033] Counting the occurrence frequency and strong correlation coefficient of each operating condition feature with a strong correlation relationship, and setting a first coefficient for the corresponding operating condition feature;

[0034] Assign a representation target to the corresponding feature combination according to the feature representation of each feature combination, and perform cluster analysis on all representation targets to obtain the number of clusters;

[0035] Based on all first coefficients and a single total quantity, a comparative quantity is calculated;

[0036]

[0037] Among them, Nb represents the number of comparisons; N max Indicates the maximum frequency of occurrence; N max-1 Indicates the second most frequent occurrence; Indicates the rounding symbol;

[0038] If the comparison number is less than or equal to the cluster number, setting the cluster number as the task branch number;

[0039] If the comparison number is greater than the cluster number, the accuracy of the cluster analysis is refined, and cluster analysis is performed again on all representation targets, and the newly obtained number of clusters is set as the number of task branches.

[0040] Preferably, the calculation formula for setting the first coefficient to the corresponding operating condition characteristic is as follows:

[0041]

[0042] Where X represents the first coefficient of the corresponding working condition feature; Nc represents the frequency of occurrence of the corresponding working condition feature; Nd represents the total number of single i1 Indicates the strong correlation coefficient of the feature combination corresponding to the i1th occurrence of the working condition feature.

[0043] Preferably, the multi-source data includes: vibration data, temperature data and geometric dimension data of railway tracks, stress and strain data and displacement data of bridges, current and voltage data and operating frequency data of signal equipment, train running speed, load and travel route.

[0044] Preferably, the basic information includes: topographic data of the railway line, track design parameters, bridge structure data and signal equipment layout diagram.

[0045] Preferably, based on the key monitoring points and data requirements, and in combination with the operating environment characteristics of each key monitoring point, matching sensor types are screened, including:

[0046] Analyze the type of data and accuracy requirements that need to be monitored at each key monitoring point;

[0047] Conduct field surveys and data collection on the operating environment of each key monitoring point, analyze the environmental characteristics of the corresponding key monitoring point, and collect historical environmental data of each key monitoring point to analyze the patterns of environmental changes;

[0048] The fuzzy comprehensive evaluation method is introduced to deal with the uncertainty factors in the evaluation indicators. According to the data requirements, environmental characteristics and environmental change laws of each key monitoring point, the sensor types that meet the basic measurement requirements are screened from the sensor product library.

[0049] Compared with the prior art, the present invention has the following advantages:

[0050] Preprocessing ensures data accuracy, providing a reliable data foundation for subsequent working condition feature extraction and state prediction. Working condition features containing spatiotemporal information and dynamic change laws are extracted from preprocessed multi-source data, which can more comprehensively and accurately capture the key characteristics of the operating conditions of railway facilities. The multi-task learning neural network model can make full use of the correlation between working condition features and realize multi-task collaborative learning. By feedbacking errors, the working condition feature extraction model and the multi-task learning model can adapt to the dynamic changes of actual working conditions and continuously optimize model performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the technical solutions in the present invention or the prior art, a brief introduction is given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0052] Figure 1 This is a flowchart of a railway facility status prediction method based on working condition identification and multi-task learning provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0053] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0054] The present invention proposes a railway facility status prediction method based on working condition identification and multi-task learning, such as Figure 1 Shown, including:

[0055] Step 1: Use sensor networks and monitoring equipment to collect multi-source data during railway facility operation and perform pre-processing;

[0056] Step 2: Use a combination of convolutional neural network and long short-term memory network to build a working condition feature extraction model, and extract features from the preprocessed data to obtain working condition features;

[0057] Step 3: Perform association mining on working condition features, determine the number of task branches and the state prediction target for each task branch, and build a multi-task learning neural network model;

[0058] Step 4: Input the real-time collected railway facility operation data and current working condition information into the trained multi-task learning prediction model, and output the status prediction results of the railway facilities in different directions;

[0059] Step 5: Adjust the structure and parameters of the working condition feature extraction model and multi-task learning model based on actual working condition changes and prediction effect feedback.

[0060] Preferably, the multi-source data includes: vibration data, temperature data and geometric dimension data of railway tracks, stress and strain data and displacement data of bridges, current and voltage data and operating frequency data of signal equipment, train running speed, load and travel route.

[0061] In this embodiment, vibration sensors and temperature sensors are installed at regular intervals along the railway tracks, and stress, strain, and displacement sensors are installed at key stress points, such as bridge piers and main beams. These sensors are connected wirelessly or wired to form a network capable of real-time data collection. Monitoring equipment is used to assist sensors in data collection or to monitor specific facilities. For example, a track gauge can measure track geometry, and a signal monitor can collect current, voltage, and operating frequency data from signal equipment.

[0062] Multi-source data refers to data from multiple sources and types. For example, for railway facilities, this includes vibration data (reflecting track vibrations when a train passes through), temperature data (reflecting changes in track temperature under different ambient temperatures), and geometric data (such as track gauge and track height); stress and strain data (indicating the stress on the bridge structure) and displacement data (monitoring bridge displacement and deformation); current and voltage data on signal equipment (determining whether the signal equipment is properly powered) and operating frequency data (ensuring stable signal transmission); train speed and load (understanding the impact of train operating conditions on facilities), and travel routes (analyzing the status of facilities under different route conditions).

[0063] Operating condition characteristics are key features that reflect the operating conditions of railway facilities. Examples include abnormal vibration frequency characteristics extracted from track vibration data, stress mutation characteristics extracted from bridge stress and strain data, and comprehensive operating condition characteristics derived from factors such as train speed and load. These characteristics can characterize the current operating status of railway facilities.

[0064] In this embodiment, association mining is the process of discovering potential correlations between features from a large amount of operating condition feature data. For example, analysis revealed a significant correlation between the vibration frequency characteristics of the track and the stress and strain characteristics of bridge piers when a train passes through a railway curve at high speed and with a heavy load. This type of association mining helps to gain a deeper understanding of the interactions between different parts of railway infrastructure.

[0065] Task branches are independent task units within a multi-task learning neural network model, divided for different prediction objectives. For example, track wear prediction can be one task branch, and bridge structural health prediction can be another. Each branch focuses on predicting the condition of a specific facility or aspect.

[0066] In this embodiment, the operating condition information is information reflecting the current operating condition of the railway facilities, including the current operating speed, load, route of the train, and environmental factors (such as weather conditions), etc. This information is closely related to the status of the facilities.

[0067] Condition prediction results include, for example, predicting the degree of track wear in the next week, the stress changes in the bridge structure when the next train passes, and the possibility of signal equipment failure in the next month.

[0068] In this embodiment, prediction effectiveness feedback is obtained by comparing the model's predictions with actual events, providing feedback on the accuracy and effectiveness of the predictions. For example, if the model predicts 5mm of wear on a certain track section for one week, but the actual wear measured is 8mm, the difference between the two is part of the prediction effectiveness feedback.

[0069] For the working condition feature extraction model, adjust the network structure of CNN and LSTM (such as increasing the number of convolutional layers, adjusting the number of LSTM units) or parameters (such as learning rate, convolution kernel size); for the multi-task learning model, modify the structure of the shared underlying network and task branch network, readjust the task branch weights, and retrain the model using a new dataset.

[0070] The beneficial effects of the above technical solution are: ensuring the accuracy of data through preprocessing, providing a reliable data basis for subsequent working condition feature extraction and state prediction, extracting working condition features containing spatiotemporal information and dynamic change laws from preprocessed multi-source data, and being able to more comprehensively and accurately capture the key features of the operating conditions of railway facilities. The multi-task learning neural network model can make full use of the correlation between working condition features and realize multi-task collaborative learning. By feedbacking errors, the working condition feature extraction model and the multi-task learning model can adapt to the dynamic changes of actual working conditions and continuously optimize model performance.

[0071] The present invention proposes a railway facility status prediction method based on operating condition identification and multi-task learning. It uses sensor networks and monitoring equipment to collect multi-source data during the operation of railway facilities and performs preprocessing, including:

[0072] Develop a sensor network deployment plan based on the distribution and operational characteristics of railway facilities;

[0073] The sensor network is activated, and each sensor collects railway facility operation data in real time at a set sampling frequency. Onboard sensors continuously collect train speed, load, and route data during train operation, and transmit this data to a ground-based data aggregation node via a dedicated railway communication network. The data aggregation node then performs preliminary integration and encoding of the received multi-source data.

[0074] The data collected by the data aggregation node is transmitted to the data processing center, and the collected data is cleaned using the data cleaning algorithm;

[0075] For problems where different types of data have different dimensions and numerical ranges, a normalization algorithm is used to process the cleaned data, and a one-hot encoding method is used to convert categorical data;

[0076] Through statistical analysis of data, missing values ​​in the data are identified and filled.

[0077] In this embodiment, railway facilities include railway tracks, bridges, tunnels, signal equipment, power supply equipment, etc., and the distribution characteristics refer to the layout of railway facilities in geographical space.

[0078] Operational characteristics involve the impact of factors such as train speed, load, operating frequency, and line slope on railway facilities.

[0079] In this embodiment, onboard sensors are sensors installed on the train, used to collect the train's operating parameters. For example, speed sensors measure the train's speed; load sensors detect the weight of the train's cargo; and positioning sensors acquire information about the train's route.

[0080] In this embodiment, the railway dedicated communication network is a communication network specially built for railway data transmission, which can adopt wireless communication (such as 4G / 5G private network) or wired communication (such as optical fiber). The data aggregation node is a device used to receive and aggregate data transmitted by multiple sensors.

[0081] In this embodiment, the data cleaning algorithm is an algorithm used to identify and process problems such as noise, outliers, and duplicate data in the data.

[0082] Categorical data: data representing different categories or attributes, such as train line names (e.g., Beijing-Shanghai Line, Beijing-Guangzhou Line), signal equipment types (e.g., signal lights, switch controllers), etc.

[0083] One-hot encoding: A method of converting categorical data into binary vectors. For example, if there are train routes "Beijing-Shanghai Line," "Beijing-Guangzhou Line," and "Beijing-Kowloon Line," after one-hot encoding, "Beijing-Shanghai Line" can be represented as [1,0,0], "Beijing-Guangzhou Line" as [0,1,0], and "Beijing-Kowloon Line" as [0,0,1].

[0084] The beneficial effects of the above technical solution are: the sensor network deployment solution can ensure that sensors fully cover key railway facilities and parts, collect key data reflecting the operating status of railway facilities, and after preliminary integration and encoding, ensure that the data format is unified and the structure is clear, so that the stability and efficiency of data transmission are improved. After data cleaning, outliers, duplicate data and noise in the data are effectively removed, and the accuracy and reliability of the data are significantly improved. After normalization and one-hot encoding processing, different types of data have a unified format and scale, eliminating the impact of dimensional and numerical range differences on data analysis, and filling in missing values ​​in the data to restore the integrity of the data.

[0085] This paper proposes a railway facility status prediction method based on working condition identification and multi-task learning. According to the distribution characteristics and operating characteristics of railway facilities, a sensor network deployment plan is formulated, including:

[0086] Collect basic information on railway facilities and use data analysis tools to conduct in-depth analysis of the basic information to determine the key monitoring points and data requirements of each part of the railway facilities;

[0087] Screen matching sensor types based on key monitoring points and data requirements, and in combination with the operating environment characteristics of each key monitoring point;

[0088] Based on the distribution and monitoring needs of railway facilities, geographic information system technology is used to conduct preliminary planning of the sensor network on the electronic map of the railway line. According to the key parts of the track and the characteristics of operation data, the deployment locations of sensors matching the key monitoring points involved in the key parts are determined to form a sensor network deployment plan.

[0089] The basic information includes: topographic data of the railway line, track design parameters, bridge structure data and signal equipment layout diagram.

[0090] In this embodiment, the basic information of railway facilities is the basic data for the construction and operation of railway facilities. The topographic data describes the natural topographic features of the area through which the railway line passes, such as plains, mountainous areas, and river distribution; the track design parameters cover the track type (such as ballastless track, ballasted track), track gauge, curve radius, etc.; the bridge structure data includes the bridge type (such as beam bridge, arch bridge, cable-stayed bridge), span, load-bearing design, etc.; the signal equipment layout diagram shows the installation position and connection relationship of equipment such as signal lights, switch controllers, and communication base stations.

[0091] Data analysis tools are software or algorithms used to process and analyze data, such as Excel, Python's Pandas library, SPSS and other statistical analysis software, which can perform data cleaning, statistical calculations, visualization and other operations.

[0092] Key monitoring points, such as the switch area, rail joints, and curved sections of the track; key stress points of bridge piers, towers, and main beams; and the core control box of signal equipment.

[0093] Data requirements refer to the type of data and accuracy requirements that need to be collected at key monitoring points. For example, key rail monitoring points need to collect vibration, temperature, and geometric dimension data, and the vibration sensor accuracy must reach ±0.1m / s. 2 ;Key monitoring points on the bridge need to monitor stress, strain and displacement data, and the accuracy of the stress and strain sensor must reach ±1με.

[0094] Sensor types are devices used to sense different physical quantities, such as vibration sensors, temperature sensors, stress and strain sensors, displacement sensors, current and voltage sensors, etc.

[0095] The electronic railway map presents the railway line direction, station distribution, topography and other information in digital form. It is the result of the application of GIS technology in the railway field.

[0096] The beneficial effect of the above technical solution is: through the analysis of basic information, key monitoring points such as track curves and bridge main spans are accurately located, and the vibration, stress, strain and other data monitoring requirements corresponding to each point are determined.

[0097] The present invention proposes a railway facility status prediction method based on working condition identification and multi-task learning, step 2, including:

[0098] Construct a CNN network to extract local features of the preprocessed data to obtain preliminary feature representation;

[0099] The preliminary feature representation is used as the input of the long short-term memory network and learned to mine the dynamic change patterns of the preprocessed data in the time dimension, and to extract the temporal features and contextual information from the preprocessed data;

[0100] The features output by the LSTM network are fused with some local features extracted by the CNN network in the middle layer to form a complete working condition feature vector containing the data's spatiotemporal information and dynamic change rules, and the working condition feature vector is output as the final extraction result.

[0101] In this embodiment, some representative local features are selected from the middle layer of the CNN network. These features can reflect the local details and structural information of the data. The temporal features and contextual information output by the LSTM network are fused with the selected local features of the middle layer of the CNN network by splicing, weighted summation, etc. For example, the feature vector output by the LSTM network and the feature map of the middle layer of the CNN network are expanded and spliced ​​to form a new feature vector; or weights are assigned to different features according to their importance, and the fused feature vector is obtained by weighted summation. The fused feature vector is normalized and other processing is performed to form the final working condition feature vector, which is used as the output result of the entire feature extraction process.

[0102] The beneficial effect of the above technical solution is that the fused feature vector obtained by using the features extracted by the CNN network and the LSTM network can more comprehensively describe the operating conditions of railway facilities.

[0103] The present invention proposes a railway facility status prediction method based on working condition identification and multi-task learning, step 3, including:

[0104] Use association rule mining algorithms to analyze working condition characteristics, calculate the support and confidence between different working condition characteristics, and find feature combinations with strong correlations;

[0105] Counting the number of combinations of the feature combinations, and determining the number of task branches and the state prediction target of each task branch in combination with the actual needs of railway facility state prediction;

[0106] Extract common features from all working condition features based on a shared underlying network;

[0107] Connecting the task branch to the shared underlying network;

[0108] The branch network of each task branch is screened according to the state prediction target, the general features are processed, the specific features related to the corresponding task branch are extracted, and the state prediction task is completed.

[0109] In this embodiment, the operating condition characteristics reflect the key characteristics of the operating conditions of railway facilities, such as the vibration frequency and temperature changes of the track, the stress and strain values ​​and displacement of the bridge, and the characteristics reflected by the train's operating speed and load.

[0110] Support: This indicates the proportion of all data sets that contain a certain feature combination, measuring how frequently a feature combination occurs. For example, if 200 out of 1,000 sets of railway operating data contain both "high-speed train operation" and "high track vibration frequency," the support of this feature combination is 20%.

[0111] Confidence: The probability of a feature appearing when another feature is present, reflecting the reliability of the association between features. For example, if 500 data sets containing "high-speed train" also contain "high track vibration frequency," the confidence level of the association rule "high-speed train → high track vibration frequency" is 80%.

[0112] Feature combination: A collection of multiple operating condition features, such as "high-speed train operation + heavy load + high vibration frequency of track" is a feature combination. The railway facility operating condition feature data obtained through the operating condition feature extraction model is organized into a format suitable for algorithm processing, such as a transaction data set. The operating condition feature data is analyzed using the Apriori algorithm or the FP-Growth algorithm, and the support threshold (such as 15%) and the confidence threshold (such as 70%) are set. The algorithm first finds all single features (frequent 1-item sets) that meet the minimum support, and then gradually generates higher-order frequent item sets, and calculates the support and confidence of each feature combination. Feature combinations with support and confidence higher than the threshold are screened out. These are feature combinations with strong correlation.

[0113] The state prediction target is the specific prediction content that each task branch needs to achieve, such as predicting the degree of track wear, the probability of damage to the bridge structure, the failure time of signal equipment, etc.

[0114] Common features are features that are valuable to multiple task branches and reflect the commonality of data, such as time information in railway operating data and basic parameters of train operation (speed, load, etc.). These features have a certain reference value for the status prediction of different facilities such as tracks, bridges, and signal equipment. Specific features: Features that are closely related to the status prediction goals of each task branch are obtained by further extraction and transformation of the branch network based on the common features and can be directly used for status prediction. For example, in the track wear prediction task branch, the features related to track material properties and train wheel-rail interaction forces obtained through branch network processing are specific features.

[0115] The beneficial effects of the above technical solution are: by counting the number of feature combinations, it is convenient to reasonably determine the number of task branches and clarify the state prediction target, and based on the shared underlying network, representative common features are successfully extracted from the working condition feature data. The connection between the task branches and the shared underlying network constructs a complete multi-task learning neural network model architecture. The branch network of each task branch can effectively extract specific features and complete the state prediction task.

[0116] The present invention proposes a railway facility status prediction method based on working condition identification and multi-task learning, which determines the number of task branches, including:

[0117] Count the total number of single working condition features that have strong correlation;

[0118] Counting the occurrence frequency and strong correlation coefficient of each operating condition feature with a strong correlation relationship, and setting a first coefficient for the corresponding operating condition feature;

[0119] Assign a representation target to the corresponding feature combination according to the feature representation of each feature combination, and perform cluster analysis on all representation targets to obtain the number of clusters;

[0120] Based on all first coefficients and a single total quantity, a comparative quantity is calculated;

[0121]

[0122] Among them, Nb represents the number of comparisons; N max Indicates the maximum frequency of occurrence; N max-1 Indicates the second most frequent occurrence; Indicates the rounding symbol;

[0123] If the comparison number is less than or equal to the cluster number, setting the cluster number as the task branch number;

[0124] If the comparison number is greater than the cluster number, the accuracy of the cluster analysis is refined, and cluster analysis is performed again on all representation targets, and the newly obtained number of clusters is set as the number of task branches.

[0125] Preferably, the calculation formula for setting the first coefficient to the corresponding operating condition characteristic is as follows:

[0126]

[0127] Where X represents the first coefficient of the corresponding working condition feature; Nc represents the frequency of occurrence of the corresponding working condition feature; Nd represents the total number of single i1 Indicates the strong correlation coefficient of the feature combination corresponding to the i1th occurrence of the working condition feature.

[0128] In this embodiment, the maximum occurrence frequency N max and the second most frequent occurrence N max-1 It reflects the frequency of occurrence of the main operating condition characteristics in the strong correlation relationship, and the single total number of Nd reflects the overall scale of the operating condition characteristics with strong correlation relationships.

[0129] In this embodiment, setting a reasonable coefficient for each operating condition feature is crucial for accurately measuring the importance of the operating condition feature in a strong correlation relationship. Dividing the two can get the relative frequency of the feature in the whole. The strong correlation coefficient of the feature combination each time the working condition feature appears is comprehensively considered to reflect the closeness of the correlation relationship when it appears.

[0130] The beneficial effects of the above technical solution are: it comprehensively considers multiple dimensions, avoiding the one-sidedness caused by determining the number of task branches based solely on a single factor, making the determination of the number of task branches more scientific and reasonable. The calculated comparison number Nb serves as a reference value, which is compared with the number of clusters and used as a basis for adjusting the number of task branches. The calculated first coefficient X can be used to assign a quantitative importance index to each working condition feature. When determining the number of task branches and allocating resources, different working condition features can be treated differently based on this coefficient, highlighting the role of important features and making the model focus more on key working condition features and their relationships.

[0131] This paper proposes a railway facility status prediction method based on working condition identification and multi-task learning. According to key monitoring points and data requirements, and combined with the operating environment characteristics of each key monitoring point, matching sensor types are screened, including:

[0132] Analyze the type of data and accuracy requirements that need to be monitored at each key monitoring point;

[0133] Conduct field surveys and data collection on the operating environment of each key monitoring point, analyze the environmental characteristics of the corresponding key monitoring point, and collect historical environmental data of each key monitoring point to analyze the patterns of environmental changes;

[0134] The fuzzy comprehensive evaluation method is introduced to deal with the uncertainty factors in the evaluation indicators. According to the data requirements, environmental characteristics and environmental change laws of each key monitoring point, the sensor types that meet the basic measurement requirements are screened from the sensor product library.

[0135] In this embodiment, the accuracy requirement is a quantitative standard for the accuracy of the collected data. For example, the accuracy of the track vibration sensor is required to reach ±0.1m / s 2 , ensuring that subtle changes in track vibration can be accurately captured; the accuracy of bridge stress and strain sensors must reach ±1με in order to accurately monitor changes in bridge structure stress.

[0136] In this embodiment, historical environmental data refers to environmental parameter records of key monitoring points over a period of time in the past, such as daily temperature and humidity data for the past year, and maximum wind speed and rainfall data for each of the past five years.

[0137] A fuzzy evaluation set of sensor performance indicators is established, and each indicator is divided into different fuzzy levels (such as "excellent", "good", "medium", and "poor"). The weight of each evaluation indicator is determined based on the data requirements, environmental characteristics, and change patterns of each key monitoring point. For example, in a strong electromagnetic interference environment, the anti-interference ability indicator has a higher weight. Performance parameter information of various sensors is collected from the sensor product library, and the fuzzy comprehensive evaluation method is used to evaluate candidate sensors for each key monitoring point. By calculating the membership degree of each sensor under different indicators and combining the indicator weights, a comprehensive score for each sensor is obtained. Sensor types that meet basic measurement requirements are screened according to the scores. Uncertainty factors include accuracy, range, and environmental adaptability.

[0138] The beneficial effects of the above technical solution are: clarifying the data type and accuracy requirement list of each key monitoring point provides a clear technical indicator basis for subsequent sensor selection. Collecting the operating environment of key monitoring points helps to gain an in-depth understanding of the environmental conditions of key monitoring points, which facilitates the subsequent screening of reasonable sensor types.

[0139] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A railway facility status prediction method based on working condition identification and multi-task learning, characterized in that: include: Step 1: Use sensor networks and monitoring equipment to collect multi-source data during railway facility operation and perform pre-processing; Step 2: Use a combination of convolutional neural network and long short-term memory network to build a working condition feature extraction model, and extract features from the preprocessed data to obtain working condition features; Step 3: Perform association mining on working condition features, determine the number of task branches and the state prediction target for each task branch, and build a multi-task learning neural network model; Step 4: Input the real-time collected railway facility operation data and current working condition information into the trained multi-task learning prediction model, and output the status prediction results of the railway facilities in different directions; Step 5: Adjust the structure and parameters of the working condition feature extraction model and multi-task learning model based on actual working condition changes and prediction effect feedback.

2. The railway facility status prediction method based on working condition identification and multi-task learning according to claim 1 is characterized in that: Utilize sensor networks and monitoring equipment to collect and pre-process multi-source data during railway facility operation, including: Develop a sensor network deployment plan based on the distribution and operational characteristics of railway facilities; The sensor network is activated, and each sensor collects railway facility operation data in real time at a set sampling frequency. Onboard sensors continuously collect train speed, load, and route data during train operation, and transmit this data to a ground-based data aggregation node via a dedicated railway communication network. The data aggregation node then performs preliminary integration and encoding of the received multi-source data. The data collected by the data aggregation node is transmitted to the data processing center, and the collected data is cleaned using the data cleaning algorithm; For problems where different types of data have different dimensions and numerical ranges, a normalization algorithm is used to process the cleaned data, and a one-hot encoding method is used to convert categorical data; Through statistical analysis of data, missing values ​​in the data are identified and filled.

3. The railway facility status prediction method based on working condition identification and multi-task learning according to claim 2 is characterized in that: Develop a sensor network deployment plan based on the distribution and operational characteristics of railway facilities, including: Collect basic information on railway facilities and use data analysis tools to conduct in-depth analysis of the basic information to determine the key monitoring points and data requirements of each part of the railway facilities; Screen matching sensor types based on key monitoring points and data requirements, and in combination with the operating environment characteristics of each key monitoring point; Based on the distribution and monitoring needs of railway facilities, geographic information system technology is used to conduct preliminary planning of the sensor network on the electronic map of the railway line. According to the key parts of the track and the characteristics of operation data, the deployment locations of sensors matching the key monitoring points involved in the key parts are determined to form a sensor network deployment plan.

4. The railway facility status prediction method based on working condition identification and multi-task learning according to claim 1 is characterized in that: Step 2 includes: Construct a CNN network to extract local features of the preprocessed data to obtain preliminary feature representation; The preliminary feature representation is used as the input of the long short-term memory network and learned to mine the dynamic change patterns of the preprocessed data in the time dimension, and to extract the temporal features and contextual information from the preprocessed data; The features output by the LSTM network are fused with some local features extracted by the CNN network in the middle layer to form a complete working condition feature vector containing the data's spatiotemporal information and dynamic change rules, and the working condition feature vector is output as the final extraction result.

5. The railway facility status prediction method based on working condition identification and multi-task learning according to claim 1 is characterized in that: Step 3 includes: Use association rule mining algorithms to analyze working condition characteristics, calculate the support and confidence between different working condition characteristics, and find feature combinations with strong correlations; Counting the number of combinations of the feature combinations, and determining the number of task branches and the state prediction target of each task branch in combination with the actual needs of railway facility state prediction; Extract common features from all working condition features based on a shared underlying network; Connecting the task branch to the shared underlying network; The branch network of each task branch is screened according to the state prediction target, the general features are processed, the specific features related to the corresponding task branch are extracted, and the state prediction task is completed.

6. The railway facility status prediction method based on working condition identification and multi-task learning according to claim 5 is characterized in that: Determine the number of task branches, including: Count the total number of single working condition features that have strong correlation; Counting the occurrence frequency and strong correlation coefficient of each operating condition feature with a strong correlation relationship, and setting a first coefficient for the corresponding operating condition feature; Assign a representation target to the corresponding feature combination according to the feature representation of each feature combination, and perform cluster analysis on all representation targets to obtain the number of clusters; Based on all first coefficients and a single total quantity, a comparative quantity is calculated; Among them, Nb represents the number of comparisons; N max Indicates the maximum frequency of occurrence; N max-1 Indicates the second most frequent occurrence; Indicates the rounding symbol; If the comparison number is less than or equal to the cluster number, the cluster number is set as the task branch number; If the comparison number is greater than the cluster number, the accuracy of the cluster analysis is refined, and cluster analysis is performed again on all representation targets, and the newly obtained number of clusters is set as the number of task branches.

7. The railway facility status prediction method based on working condition identification and multi-task learning according to claim 6 is characterized in that: The calculation formula for setting the first coefficient to the corresponding operating condition characteristics is as follows: Where X represents the first coefficient of the corresponding working condition feature; Nc represents the frequency of occurrence of the corresponding working condition feature; Nd represents the total number of single i1 Indicates the strong correlation coefficient of the feature combination corresponding to the i1th occurrence of the working condition feature.

8. The railway facility status prediction method based on working condition identification and multi-task learning according to claim 1 is characterized in that: The multi-source data includes: vibration data, temperature data and geometric dimension data of railway tracks, stress and strain data and displacement data of bridges, current and voltage data and operating frequency data of signal equipment, train running speed, load and travel route.

9. The railway facility status prediction method based on working condition identification and multi-task learning according to claim 3 is characterized in that: The basic information includes: topographic data of the railway line, track design parameters, bridge structure data and signal equipment layout diagram.

10. The railway facility status prediction method based on working condition identification and multi-task learning according to claim 1, characterized in that: Based on key monitoring points and data requirements, and combined with the operating environment characteristics of each key monitoring point, matching sensor types are screened, including: Analyze the type of data and accuracy requirements that need to be monitored at each key monitoring point; Conduct field surveys and data collection on the operating environment of each key monitoring point, analyze the environmental characteristics of the corresponding key monitoring point, and collect historical environmental data of each key monitoring point to analyze the patterns of environmental changes; The fuzzy comprehensive evaluation method is introduced to deal with the uncertainty factors in the evaluation indicators. According to the data requirements, environmental characteristics and environmental change laws of each key monitoring point, the sensor types that meet the basic measurement requirements are screened from the sensor product library.

Images