A train operation risk assessment method and system based on a dynamic Bayesian network

Abstract

This invention provides a method and system for train operation risk assessment based on dynamic Bayesian networks. The method includes: acquiring multi-source heterogeneous data involved in train operation; modeling the multi-source heterogeneous data using a dynamic Bayesian network to construct a causal relationship graph containing time slices, generating an initial risk assessment model; updating the node variables in the initial risk assessment model in real time to obtain a target risk assessment model; calculating the probability and severity of each risk factor using a conditional probability inference algorithm based on the target risk assessment model; generating risk response strategies using a decision tree algorithm based on the probability and severity of each risk factor; and introducing a distributed monitoring mechanism to continuously track the train operation status and record feedback data for optimizing the risk assessment model during the execution of the risk response strategies. This invention significantly improves the accuracy, timeliness, and adaptability of train risk assessment.

Description

Technical Field

[0001] This invention relates to the field of train operation risk assessment technology, and in particular to a train operation risk assessment method and system based on dynamic Bayesian networks. Background Technology

[0002] With the continuous development of railway transportation systems, the risk factors faced by trains during operation are becoming increasingly complex. Effectively assessing and addressing these risks has become a crucial issue in ensuring railway traffic safety. Existing train operation risk assessment methods mostly rely on static models or single-dimensional data analysis, which have limitations in comprehensively reflecting risk changes in a dynamic environment. Therefore, developing a dynamic risk assessment method that can combine multi-source data and update it in real time is of great significance.

[0003] One existing method for dynamic risk analysis of high-speed railway overhead contact lines based on Bayesian networks involves establishing a risk propagation chain model for flashover of insulators in the overhead contact lines and using a combination of dynamic probability and passenger flow density index to dynamically assess the probability and severity of risk occurrence. The drawbacks of this method include: it focuses on risk analysis in specific scenarios of high-speed railway overhead contact lines and fails to extend to broader train operation risk scenarios. Furthermore, its dynamic probability modeling process considers only a limited number of input variables, which may lead to insufficient adaptability and accuracy in complex and changing real-world environments.

[0004] One existing method for risk assessment of train operation under earthquake conditions involves constructing a risk assessment model to guide emergency response for train operations by acquiring earthquake epicenter location information, train location information, and historical disaster data. The drawbacks of this method include: while it demonstrates high applicability in specific disaster scenarios such as earthquakes, its assessment scope primarily focuses on earthquake-related risks, with limited ability to cover other potential risk factors (such as abnormal weather and equipment failure). Furthermore, this method lacks "closed-loop feedback" after the implementation of risk response strategies. That is, after issuing instructions, it fails to effectively track the actual execution effects of the instructions and feed them back into the model, resulting in the model's inability to self-evolve and calibrate its parameters over time. Summary of the Invention

[0005] This invention provides a train operation risk assessment method and system based on dynamic Bayesian networks to effectively improve train operation safety and efficiency.

[0006] To achieve the above objectives, the present invention adopts the following technical solution.

[0007] According to one aspect of the present invention, a method for assessing train operation risk in a dynamic Bayesian network is provided, comprising:

[0008] Acquire multi-source heterogeneous data involved in train operation, including environmental data, equipment status data, and historical operation data;

[0009] The multi-source heterogeneous data is modeled using a dynamic Bayesian network to construct a causal relationship graph containing time slices and generate an initial risk assessment model.

[0010] Based on real-time collected train operation status data, the node variables in the initial risk assessment model are updated in real time to obtain the target risk assessment model.

[0011] Based on the target risk assessment model, the probability and severity of each risk factor are calculated using a conditional probability inference algorithm;

[0012] Based on the probability and severity of each risk factor, a risk response strategy is generated using a decision tree algorithm.

[0013] In the process of implementing the aforementioned risk response strategy, a distributed monitoring mechanism is introduced to continuously track the train's operating status and record feedback data to optimize the risk assessment model.

[0014] Preferably, the acquisition of multi-source heterogeneous data involved in train operation includes environmental data, equipment status data, and historical operation data, including:

[0015] The system acquires multi-source heterogeneous data involved in train operation, including environmental data, equipment status data, and historical operation data. The environmental data comes from external sensors or third-party meteorological service interfaces. The equipment status data is collected in real time by various sensors on the train. The historical operation data is extracted from the train operation record database to reflect the operation trend over a period of time.

[0016] After denoising and normalization, the multi-source heterogeneous data is used to form an observation vector.

[0017] Preferably, the step of using a dynamic Bayesian network to model the multi-source heterogeneous data, constructing a causal relationship graph containing time slices, and generating an initial risk assessment model includes:

[0018] The multi-source heterogeneous data is modeled using a dynamic Bayesian network, and the node set of the DBN is defined to include environment nodes, device nodes, and running nodes.

[0019] A causal relationship graph containing time slices is constructed to describe the dependencies between nodes. This causal relationship graph is used to describe the direct dependencies between nodes. Based on the causal relationship graph, the prior probability distribution of each node is initialized using the maximum likelihood estimation method. and conditional probability distribution ;

[0020] Prior probability distribution of nodes Represented as:

[0021]

[0022] Among the symbols Indicates multiplication. =1 to , Represents a node The set of parent nodes;

[0023] For dynamic processes, consider time slices. Conditional probability distribution of nodes for:

[0024]

[0025] Among the symbols Indicates to arrive Perform a series of multiplications. express All state variables at any given time;

[0026] An initial risk assessment model is formed based on the prior probability distribution of the nodes and the conditional probability distribution of the system using the maximum likelihood estimation method.

[0027] Preferably, the method of updating the node variables in the initial risk assessment model in real time based on real-time collected train operation status data to obtain the target risk assessment model includes:

[0028] Real-time data streams during train operation are collected, including sensor data, communication data, and manually input data. The real-time data streams are preprocessed to remove noise and outliers, resulting in cleaned data. The cleaned data is then used as observation evidence E and mapped to the corresponding nodes of a dynamic Bayesian network.

[0029] The posterior probability of a node is updated using a dynamic Bayesian network, and the update formula is as follows:

[0030] in, This indicates the hypothetical risk event to be assessed; Indicates at time The collected observational evidence; This indicates that, without incorporating current observational evidence, risk events... The prior probability; Indicates in risk events Under the conditions of occurrence, observational evidence The conditional probability of occurrence; Indicates observational evidence The marginal probability of occurrence; Indicating in observational evidence Under these conditions, risk events The posterior probability of occurrence;

[0031] By adjusting the conditional probability distribution among nodes based on the posterior probability distribution, the initial risk assessment model is updated, resulting in a target risk assessment model that reflects the true current status of the nodes in the initial risk assessment model.

[0032] Preferably, the step of calculating the probability and severity of each risk factor based on the target risk assessment model using a conditional probability reasoning algorithm includes:

[0033] Based on the target risk assessment model and current train operation status data, the probability of occurrence of each risk factor is calculated, and the severity of the risk factor is calculated by combining the loss function.

[0034] Based on the target risk assessment model and current train operation data, the probability of a specific risk event is calculated using a joint tree algorithm or variable elimination method. ;

[0035] Based on the probability of occurrence of the specific risk event The severity of risk factors is obtained by quantifying them using a loss function.

[0036] Set risk threshold ,when At that time, the decision-making process is triggered.

[0037] Preferably, the step of generating risk response strategies using a decision tree algorithm based on the probability and severity of each risk factor includes:

[0038] Define the feature set of the decision tree, which includes the probability of occurrence, severity, current position of the train, and available resources for each risk factor;

[0039] Information gain is used as the basis for node splitting. The information gain of each feature is calculated. Let the sample set be... ,feature The information gain is:

[0040] in In order to reach a settlement, arrive ;in: Indicates the current training sample set; Represents the sample set The number of samples in the sample; This represents the feature attributes used to partition the sample set; Indicates in features Take the first The sample subset corresponding to each value; Represents a subset of samples The number of samples in the sample; Representation of features The number of all possible values, ; Represents the sample set Information entropy is used to measure the uncertainty of a sample set; Represents a subset of samples Information entropy. Among them, The calculation formula is:

[0041] in: Represents the sample set The Middle The proportion of samples of each class; Indicates the sample category index; when the sample set Includes When there are different categories, The value range is 1 to Information entropy Entropy is used to measure the degree of uncertainty in the distribution of categories in a sample set. The larger the entropy value, the more dispersed the category distribution of the sample set.

[0042] Based on the information gain of each feature, the optimal splitting attribute is selected to construct a decision tree. Each leaf node of the decision tree corresponds to a risk response strategy. The decision tree is optimized by pre-pruning or post-pruning. Based on the optimized decision tree, the risk level is mapped to a specific control instruction to generate a risk response strategy.

[0043] Preferably, during the execution of the risk response strategy, a distributed monitoring mechanism is introduced to continuously track the train's operating status and record feedback data for optimizing the risk assessment model, including:

[0044] Multiple distributed monitoring modules are deployed in the train system. During the execution of the risk response strategy, each monitoring module collects the operating status data of a local area of ​​the train. The monitoring module uploads the operating status data to the central processing unit. The central processing unit summarizes and analyzes the operating status data reported by each monitoring module and generates a global operating status report.

[0045] Define the deviation in execution effect .

[0046] in: This indicates the expected operating state predicted by the system after the risk response strategy is implemented; This indicates the actual train operating status collected by the distributed monitoring module; Used to measure the degree of deviation between the actual results of risk response strategies and the expected results;

[0047] If deviation If the value exceeds the preset range, it indicates a discrepancy between the prior probability distribution and the conditional probability distribution of the nodes in the original dynamic Bayesian network. The system uses this event and the discrepancy as feedback data and employs the Bayesian learning mechanism to correct the node parameters in the dynamic Bayesian network.

[0048]

[0049] in The learning rate; Represents a node The conditional probability distribution before the update; Represents a node In the updated conditional probability distribution; The learning rate is used to control the step size of parameter updates; Indicates deviation based on execution effect The calculated parameter correction amount;

[0050] Adjust the prior probability distribution and conditional probability distribution of nodes to optimize the risk assessment model.

[0051] According to another aspect of the present invention, a train operation risk assessment system based on dynamic Bayesian networks is provided, comprising:

[0052] The data acquisition module acquires multi-source heterogeneous data involved in the train operation process, including environmental data, equipment status data, and historical operation data.

[0053] The modeling module is used to model the multi-source heterogeneous data using a dynamic Bayesian network, construct a causal relationship graph containing time slices, and generate an initial risk assessment model.

[0054] The update module is used to update the node variables in the initial risk assessment model in real time based on the real-time collected train operation status data, so as to obtain the target risk assessment model.

[0055] The assessment module is used to calculate the probability and severity of each risk factor based on the target risk assessment model using a conditional probability reasoning algorithm.

[0056] The decision-making module is used to generate risk response strategies based on the probability and severity of occurrence of each risk factor using a decision tree algorithm.

[0057] The monitoring module is used to introduce a distributed monitoring mechanism to continuously track the train's operating status and record feedback data to optimize the risk assessment model during the execution of the risk response strategy.

[0058] A computing device includes a processing component and a storage component. The storage component stores one or more computer instructions that are invoked and executed by the processing component to implement the train operation risk assessment method based on a dynamic Bayesian network. The processing component performs data acquisition, modeling, updating, evaluation, decision-making, and monitoring operations. The storage component stores intermediate results and final output.

[0059] As can be seen from the technical solutions provided by the embodiments of the present invention above, the present invention can effectively solve the problems of static models and lack of real-time feedback loops in existing technologies. Through mathematical dynamic updating and decision-making mechanisms, it significantly improves the accuracy, timeliness and adaptive capability of train risk assessment.

[0060] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and will become apparent from the description or may be learned by practice of the invention. Attached Figure Description

[0061] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0062] Figure 1 A structural diagram of a train operation risk assessment system based on a dynamic Bayesian network is provided for an embodiment of the present invention;

[0063] Figure 2 This is a flowchart illustrating a train operation risk assessment method based on dynamic Bayesian networks, provided as an embodiment of the present invention.

[0064] Figure 3 This is a schematic diagram of the structure of a dynamic Bayesian network node set and its causal relationship graph provided in an embodiment of the present invention.

[0065] Figure 4 This is a schematic diagram of a process for obtaining a target risk assessment model by real-time updating of an initial risk assessment model, as provided in an embodiment of the present invention.

[0066] Figure 5This is a schematic diagram illustrating a process for calculating the probability and severity of each risk factor based on a conditional probability reasoning algorithm, as provided in an embodiment of the present invention.

[0067] Figure 6 This is a schematic diagram illustrating a process for generating risk response strategies using a decision tree algorithm, as provided in an embodiment of the present invention.

[0068] Figure 7 This is a schematic diagram illustrating a distributed monitoring mechanism for continuously tracking train operation status, as provided in an embodiment of the present invention.

[0069] Figure 8 This is a schematic diagram of the hardware structure of a computing device provided in an embodiment of the present invention.

[0070] The attached figures are labeled as follows:

[0071] 1. Data acquisition module; 2. Modeling module; 3. Update module; 4. Evaluation module; 5. Decision-making module; 6. Monitoring module; 7. Processing component; 8. Storage component. Detailed Implementation

[0072] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0073] Those skilled in the art will understand that, unless specifically stated otherwise, the singular forms “a,” “an,” “the,” and “the” used herein may also include the plural forms. It should be further understood that the term “comprising” as used in this specification means the presence of the stated features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof. It should be understood that when we say an element is “connected” or “coupled” to another element, it can be directly connected or coupled to the other element, or there may be intermediate elements. Furthermore, “connected” or “coupled” as used herein can include wireless connections or couplings. The term “and / or” as used herein includes any and all combinations of one or more of the associated listed items.

[0074] It will be understood by those skilled in the art that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. It should also be understood that terms such as those defined in general dictionaries should be understood to have the same meaning as in the context of the prior art, and should not be interpreted in an idealized or overly formal sense unless defined as herein.

[0075] To facilitate understanding of the embodiments of the present invention, the following will provide further explanation and description with reference to the accompanying drawings and several specific embodiments. These embodiments do not constitute a limitation on the embodiments of the present invention.

[0076] Example 1

[0077] This invention provides a structure for a train operation risk assessment system based on dynamic Bayesian networks, as shown in the following embodiment. Figure 1 As shown, it includes the following modules:

[0078] The data acquisition module is used to acquire multi-source heterogeneous data involved in train operation, including environmental data, equipment status data, and historical operation data.

[0079] The modeling module is used to model multi-source heterogeneous data using dynamic Bayesian networks and generate an initial risk assessment model.

[0080] The update module is used to update the node variables in the initial risk assessment model in real time to obtain the target risk assessment model;

[0081] The assessment module is used to calculate the probability and severity of each risk factor based on the target risk assessment model and the current train operation status data, using a conditional probability reasoning algorithm.

[0082] The decision-making module is used to generate risk response strategies based on the probability and severity of occurrence of each risk factor using a decision tree algorithm.

[0083] The monitoring module is used to introduce a distributed monitoring mechanism to continuously track the train's operating status during the execution of risk response strategies, and to record feedback data for optimizing the risk assessment model.

[0084] First, data acquisition module 1 is responsible for acquiring multi-source heterogeneous data involved in train operation, including environmental data, equipment status data, and historical operational data. Data acquisition module 1 transmits this multi-source heterogeneous data to modeling module 2 via wired or wireless communication. Internally, data acquisition module 1 employs a distributed data acquisition architecture to ensure the diversity and real-time nature of data sources. To guarantee data quality, data acquisition module 1 also incorporates preliminary data cleaning functions, such as removing obvious outliers and duplicate data, thus providing a reliable foundation for subsequent modeling.

[0085] Modeling module 2 uses a dynamic Bayesian network to model multi-source heterogeneous data, generating an initial risk assessment model. Update module 3 updates the node variables in the initial risk assessment model in real time, obtaining the target risk assessment model. Evaluation module 4, based on the target risk assessment model and combined with current train operation status data, uses a conditional probability inference algorithm to calculate the probability and severity of each risk factor.

[0086] Decision module 5 generates risk response strategies based on the probability and severity of each risk factor using a decision tree algorithm. Finally, based on the optimized decision tree and current train operation status data, decision module 5 selects the leaf node corresponding to the optimal path to determine the risk response strategy. The risk response strategy includes deceleration commands, stopping commands, or emergency braking commands, which are executed through the train operation control system.

[0087] Monitoring module 6 introduces a distributed monitoring mechanism to continuously track the train's operating status during the execution of risk response strategies and records feedback data to optimize the risk assessment model.

[0088] The specific process of performing train operation risk assessment based on dynamic Bayesian networks using the system of this invention is similar to the method embodiments described below, and will not be repeated here.

[0089] Example 2

[0090] The processing flow of a train operation risk assessment method based on dynamic Bayesian networks provided in this embodiment of the invention is as follows: Figure 2 As shown, the processing steps include the following:

[0091] Step S10: The data acquisition module acquires multi-source heterogeneous data involved in the train operation process, including environmental data, equipment status data and historical operation data.

[0092] Step S20: Use dynamic Bayesian networks to model multi-source heterogeneous data, construct a causal relationship graph containing time slices, and generate an initial risk assessment model;

[0093] Step S30: Update the node variables in the initial risk assessment model in real time to obtain the target risk assessment model;

[0094] Step S40: Based on the target risk assessment model and combined with the current train operation status data, use the conditional probability reasoning algorithm to calculate the probability and severity of each risk factor.

[0095] Step S50: Generate risk response strategies using a decision tree algorithm based on the probability and severity of each risk factor.

[0096] Step S60: During the implementation of risk response strategies, a distributed monitoring mechanism is introduced to continuously track the train operation status and record feedback data to optimize the risk assessment model.

[0097] Optionally, step S10 above includes:

[0098] Environmental data comes from external sensors or third-party meteorological service interfaces, covering information such as weather conditions and geological activity, including rainfall. Wind speed ,earthquake intensity wait.

[0099] Equipment status data is collected in real time by various sensors on the train, such as braking system pressure values ​​and signal system status; including: braking pressure. Motor temperature Signal delay wait.

[0100] Historical operational data is extracted from the train operation record database to reflect operational trends over a past period. This includes historical fault logs and corresponding handling records.

[0101] After denoising and normalization, the aforementioned multi-source heterogeneous data forms an observation vector. .

[0102] Optionally, step S20 specifically includes: Figure 3 This is a schematic diagram of the structure of a dynamic Bayesian network node set and its causal relationship graph provided in an embodiment of the present invention.

[0103] Dynamic Bayesian Network (DBN) is used to model multi-source heterogeneous data. The node set of DBN is defined, which includes environment nodes, equipment nodes, and operation nodes. Environment nodes represent the external environment information during train operation, such as weather conditions and geological activities. Equipment nodes represent the status information of key equipment for train operation, such as braking system and signaling system. Operation nodes represent the dynamic status information of train operation, such as speed and position.

[0104] A causal relationship graph containing time slices is constructed to describe the dependencies between nodes. This graph describes the direct dependencies between nodes. Based on the causal relationship graph, the prior probability distribution of each node is initialized using maximum likelihood estimation. and conditional probability distribution .

[0105] Prior probability distribution of nodes Represented as:

[0106] Among the symbols Indicates multiplication. arrive , Represents a node The set of parent nodes;

[0107] For dynamic processes, consider time slices. The conditional probability distribution of the system for:

[0108] Among the symbols Indicates to arrive Perform a series of multiplications. express All state variables at any given time;

[0109] The modeling module uses historical data and the maximum likelihood estimation (MLE) or EM algorithm to form an initial risk assessment model based on the prior probability distribution of the above nodes and the conditional probability distribution of the system.

[0110] Optionally, step S30 specifically includes: the update module updating the node variables in the initial risk assessment model in real time.

[0111] Figure 4This is a flowchart illustrating a process for obtaining a target risk assessment model through real-time updating of an initial risk assessment model, as provided in an embodiment of the present invention. Real-time data streams during train operation are collected, including sensor data, communication data, and manually input data. The real-time data streams are preprocessed to remove noise and outliers, resulting in cleaned data. This cleaned data is then used as observation evidence E and mapped to the corresponding nodes in a dynamic Bayesian network. The update module uses Bayes' theorem to update the posterior probability of the nodes and update the observation values ​​of the nodes.

[0112] When the update module receives the observation evidence E at time t t When a sensor detects abnormal braking pressure, Bayesian inference is used to update the posterior probability of the node. The update formula is as follows:

[0113] in, This indicates the hypothetical risk event to be assessed; Indicates at time The collected observational evidence; This indicates that, without incorporating current observational evidence, risk events... The prior probability; Indicates in risk events Under the conditions of occurrence, observational evidence The conditional probability of occurrence; Indicates observational evidence The marginal probability of occurrence; Indicating in observational evidence Under these conditions, risk events The posterior probability of occurrence;

[0114] The time transition probability matrix describes the same node in different time slices. and The state evolution relationship between them. Introducing a time decay factor. The weighting of historical evidence is adjusted to ensure the model focuses more on the recent state:

[0115]

[0116] in: Represents a node At the present moment The probability distribution; This represents the node calculated based on current observational evidence. The probability of observation; Represents a node In the previous moment The probability distribution; Represents a node State transition probability matrix between adjacent time slices; This is the time decay factor, with a value ranging from 0 to 1, used to adjust the weight of the influence of current observational evidence and historical state on the current probability distribution;

[0117] By adjusting the conditional probability distribution among nodes based on the posterior probability distribution, the initial risk assessment model is updated, resulting in a target risk assessment model that reflects the true current status of the nodes in the initial risk assessment model.

[0118] Optionally, step S40 specifically includes: Figure 5 This is a schematic diagram illustrating a process for calculating the probability and severity of each risk factor based on a conditional probability reasoning algorithm, as provided in an embodiment of the present invention.

[0119] Based on the target risk assessment model and current train operation status data, the probability of occurrence of each risk factor is calculated, and the severity of the risk factor is calculated by combining the loss function.

[0120] The assessment module performs inference based on the target risk assessment model and current train operation data. It employs either the junction tree algorithm or variable elimination for precise inference to calculate specific risk events. The probability of occurrence (such as derailment, rear-end collision) .

[0121] Specific risk events Severity Through the loss function Quantification, taking into account both casualties, property damage, and operational disruption time:

[0122] Set risk threshold ,when At that time, the decision-making process is triggered.

[0123] The overall risk value of each risk factor is obtained by combining the probability of occurrence and the severity of each risk factor. The overall risk value is used to quantify the degree of impact of the risk and provide a basis for subsequent decision-making.

[0124] Optionally, step S50 specifically includes:

[0125] The decision-making module generates risk response strategies based on the probability and severity of occurrence of each risk factor using a decision tree algorithm.

[0126] 1. Feature Definition: Define the feature set of the decision tree. This feature set includes the probability of occurrence, severity, current position of the train, and available resources for each risk factor. = {Risk probability, severity, current speed, position, remaining braking distance}.

[0127] 2. Splitting Criterion: Information gain is used as the basis for node splitting; the information gain of each feature is calculated. Let the sample set be... ,feature The information gain is:

[0128] in In order to reach a settlement, arrive ;

[0129] in, Information entropy is calculated using the following formula:

[0130] in: Represents the sample set The Middle The proportion of samples of each class; Indicates the sample category index; when the sample set Includes When there are different categories, The value range is 1 to Information entropy Entropy is used to measure the degree of uncertainty in the distribution of categories in a sample set. The larger the entropy value, the more dispersed the category distribution of the sample set.

[0131] 3. Pruning: Select the optimal splitting attribute to construct a decision tree, where each leaf node corresponds to a risk response strategy. Perform post-pruning on the generated decision tree to prevent overfitting. Optimize the decision tree using pre-pruning or post-pruning (such as pruning based on pessimistic error rate) to avoid overfitting.

[0132] 4. Strategy Generation: Map risk levels to specific control commands (such as speed limit, stop), and generate risk response strategies. Traverse the decision tree based on the current state, and when a leaf node is reached, output the strategy: Strategy ∈ {Emergency Braking, Degraded Operation, Lazy Operation, Normal Operation}.

[0133] Optionally, step S60 specifically includes: Figure 7 This is a schematic diagram illustrating a distributed monitoring mechanism for continuously tracking train operation status, as provided in an embodiment of the present invention. Multiple distributed monitoring modules are deployed in the system, distributed across different carriages and key equipment of the train. After the system executes a risk response strategy (such as "downgrading to 80km / h"), each monitoring module is responsible for collecting operation status data within its local area. The monitoring modules continuously track the train's status (such as actual deceleration and stability indicators).

[0134] The monitoring module uploads operational status data to the central processing unit. The central processing unit aggregates and analyzes the operational status data reported by each monitoring module, generating a global operational status report. Then, based on the global operational status report, it assesses the effectiveness of risk response strategies and feeds the effectiveness data back to the dynamic Bayesian network to adjust the prior probability distribution and conditional probability distribution of nodes, thereby optimizing the risk assessment model.

[0135] Define the deviation in execution effect .

[0136] in: This indicates the expected operating state predicted by the system after the risk response strategy is implemented; This indicates the actual train operating status collected by the distributed monitoring module; Used to measure the degree of deviation between the actual results of risk response strategies and the expected results;

[0137] If deviation If the value exceeds the preset range, it indicates a discrepancy between the prior probability distribution and the conditional probability distribution of the nodes in the original dynamic Bayesian network. The system uses this event and the discrepancy as feedback data and employs a Bayesian learning mechanism to correct the node parameters in the dynamic Bayesian network.

[0138] in: The learning rate; Represents a node The conditional probability distribution before the update; Represents a node In the updated conditional probability distribution; The learning rate is used to control the step size of parameter updates; Indicates deviation based on execution effect The calculated parameter corrections. This feedback mechanism enables the adaptive evolution of the model;

[0139] Each distributed monitoring module is responsible for collecting operational status data within a local area and uploading the data to the central processing unit. For example, a monitoring module in one carriage might collect vibration and temperature data for that carriage, while a monitoring module in another carriage might collect braking system status data. The central processing unit aggregates and analyzes the received data to generate a global operational status report. Based on this report, the central processing unit assesses the effectiveness of risk response strategies. For instance, if the speed reduction does not reach the expected level after a train executes a deceleration command, it indicates a potential problem with the current strategy. Monitoring module 6 feeds back the performance data to a dynamic Bayesian network to adjust the prior and conditional probability distributions of the nodes, thereby optimizing the risk assessment model.

[0140] Figure 8 This is a schematic diagram of the hardware structure of a computing device provided in an embodiment of the present invention, such as... Figure 8 As shown, this embodiment of the invention also provides a computing device, including a processing component 7 and a storage component 8. The storage component 8 stores one or more computer instructions, which are invoked and executed by the processing component 7 to implement the train operation risk assessment method based on a dynamic Bayesian network as described above. The processing component 7 is responsible for performing operations such as data acquisition, modeling, updating, evaluation, decision-making, and monitoring, while the storage component 8 is used to store intermediate results and final output. Through the collaborative work of hardware and software, this invention can efficiently complete the train operation risk assessment task.

[0141] In summary, the embodiments of the present invention improve the mathematical rigor and accuracy of the evaluation by introducing specific probability update formulas and decision tree splitting algorithms; in particular, the introduction of distributed monitoring and parameter correction mechanisms solves the problems of existing technologies lacking closed-loop feedback and being unable to cope with sudden unknown environments, and significantly improves the active safety protection capability of train operation.

[0142] This invention solves the problems of static models and lack of real-time feedback loops in existing technologies. Through a mathematical dynamic update and decision-making mechanism, it significantly improves the accuracy, timeliness, and adaptive capability of train risk assessment.

[0143] Those skilled in the art will understand that the accompanying drawings are merely schematic diagrams of one embodiment, and the modules or processes shown in the drawings are not necessarily essential for implementing the present invention.

[0144] As can be seen from the above description of the embodiments, those skilled in the art can clearly understand that the present invention can be implemented by means of software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in various embodiments or some parts of the embodiments of the present invention.

[0145] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, for apparatus or system embodiments, since they are basically similar to method embodiments, the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments. The apparatus and system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0146] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for assessing train operation risk using dynamic Bayesian networks, characterized in that, include: Acquire multi-source heterogeneous data involved in train operation, including environmental data, equipment status data, and historical operation data; The multi-source heterogeneous data is modeled using a dynamic Bayesian network to construct a causal relationship graph containing time slices and generate an initial risk assessment model. Based on real-time collected train operation status data, the node variables in the initial risk assessment model are updated in real time to obtain the target risk assessment model. Based on the target risk assessment model, the probability and severity of each risk factor are calculated using a conditional probability inference algorithm; Based on the probability and severity of each risk factor, a risk response strategy is generated using a decision tree algorithm. In the process of implementing the aforementioned risk response strategy, a distributed monitoring mechanism is introduced to continuously track the train's operating status and record feedback data to optimize the risk assessment model.

2. The method according to claim 1, characterized in that, The acquisition of multi-source heterogeneous data involved in train operation includes environmental data, equipment status data, and historical operation data, including: The system acquires multi-source heterogeneous data involved in train operation, including environmental data, equipment status data, and historical operation data. The environmental data comes from external sensors or third-party meteorological service interfaces. The equipment status data is collected in real time by various sensors on the train. The historical operation data is extracted from the train operation record database to reflect the operation trend over a period of time. After denoising and normalization, the multi-source heterogeneous data is used to form an observation vector.

3. The method according to claim 1, characterized in that, The method of using dynamic Bayesian networks to model the multi-source heterogeneous data, constructing a causal relationship graph containing time slices, and generating an initial risk assessment model includes: The multi-source heterogeneous data is modeled using a dynamic Bayesian network, and the node set of the DBN is defined to include environment nodes, device nodes, and running nodes. A causal relationship graph containing time slices is constructed to describe the dependencies between nodes. This causal relationship graph is used to describe the direct dependencies between nodes. Based on the causal relationship graph, the prior probability distribution of each node is initialized using the maximum likelihood estimation method. and conditional probability distribution ; Prior probability distribution of nodes ) is represented as: Among the symbols Indicates multiplication. =1 to , Represents a node The set of parent nodes; For dynamic processes, consider time slices. Conditional probability distribution of nodes for: Among the symbols Indicates to arrive Perform a series of multiplications. express All state variables at any given time; An initial risk assessment model is formed based on the prior probability distribution of the nodes and the conditional probability distribution of the system using the maximum likelihood estimation method.

4. The method according to claim 3, characterized in that, The method of updating the node variables in the initial risk assessment model in real time based on the real-time collected train operation status data to obtain the target risk assessment model includes: Real-time data streams during train operation are collected, including sensor data, communication data, and manually input data. The real-time data streams are preprocessed to remove noise and outliers, resulting in cleaned data. The cleaned data is then used as observation evidence E and mapped to the corresponding nodes of a dynamic Bayesian network. The posterior probability of a node is updated using a dynamic Bayesian network, and the update formula is as follows: in, This indicates the hypothetical risk event to be assessed; Indicates at time The collected observational evidence; This indicates that, before the introduction of current observational evidence, risk events... The prior probability; Indicates in risk events Under the conditions of occurrence, observational evidence The conditional probability of occurrence; Indicates observational evidence The marginal probability of occurrence; This represents the posterior probability of the occurrence of risk event H given the observed evidence Et. By adjusting the conditional probability distribution among nodes based on the posterior probability distribution, the initial risk assessment model is updated, resulting in a target risk assessment model that reflects the true current status of the nodes in the initial risk assessment model.

5. The method according to claim 4, characterized in that, The aforementioned calculation of the probability and severity of each risk factor based on the target risk assessment model and using a conditional probability inference algorithm includes: Based on the target risk assessment model and current train operation status data, the probability of occurrence of each risk factor is calculated, and the severity of the risk factor is calculated by combining the loss function. Based on the target risk assessment model and current train operation data, the probability of a specific risk event is calculated using a joint tree algorithm or variable elimination method. ; Based on the probability of occurrence of the specific risk event The severity of risk factors is obtained by quantifying them using a loss function. ; Set risk threshold ,when At that time, the decision-making process is triggered.

6. The method according to claim 5, characterized in that, The method of generating risk response strategies based on the probability and severity of each risk factor using a decision tree algorithm includes: Define the feature set of the decision tree, which includes the probability of occurrence, severity, current position of the train, and available resources for each risk factor; Information gain is used as the basis for node splitting. The information gain of each feature is calculated. Let the sample set be... ,feature The information gain is: in In order to reach a settlement, arrive ;in: Indicates the current training sample set; Represents the sample set The number of samples in the sample; This represents the feature attributes used to partition the sample set; Indicates in features Take the first The sample subset corresponding to each value; Represents a subset of samples The number of samples in the sample; Representation of features The number of all possible values, ; Represents the sample set Information entropy is used to measure the uncertainty of a sample set; Represents a subset of samples Information entropy, where The calculation formula is: in: Represents the sample set The Middle The proportion of samples of each class; Indicates the sample category index; when the sample set Includes When there are different categories, The value range is 1 to Information entropy Entropy is used to measure the degree of uncertainty in the distribution of categories in a sample set. The larger the entropy value, the more dispersed the category distribution of the sample set. Based on the information gain of each feature, the optimal splitting attribute is selected to construct a decision tree. Each leaf node of the decision tree corresponds to a risk response strategy. The decision tree is optimized by pre-pruning or post-pruning. Based on the optimized decision tree, the risk level is mapped to a specific control instruction to generate a risk response strategy.

7. The method according to claim 6, characterized in that, The aforementioned risk response strategy involves introducing a distributed monitoring mechanism to continuously track train operation status and record feedback data for optimizing the risk assessment model, including: Multiple distributed monitoring modules are deployed in the train system. During the execution of the risk response strategy, each monitoring module collects the operating status data of a local area of ​​the train. The monitoring module uploads the operating status data to the central processing unit. The central processing unit summarizes and analyzes the operating status data reported by each monitoring module and generates a global operating status report. Define the deviation in execution effect ; in: This indicates the expected operating state predicted by the system after the risk response strategy is implemented; This indicates the actual train operating status collected by the distributed monitoring module; Used to measure the degree of deviation between the actual results of risk response strategies and the expected results; If the deviation Err exceeds the preset range, it indicates that there is a deviation between the prior probability distribution and the conditional probability distribution of the nodes in the original dynamic Bayesian network. The system uses this event and the deviation as feedback data and uses the Bayesian learning mechanism to correct the node parameters in the dynamic Bayesian network. in The learning rate; Represents a node The conditional probability distribution before the update; Represents a node In the updated conditional probability distribution; The learning rate is used to control the step size of parameter updates; Indicates deviation based on execution effect The calculated parameter correction amount; Adjust the prior probability distribution and conditional probability distribution of nodes to optimize the risk assessment model.

8. A train operation risk assessment system based on dynamic Bayesian networks, characterized in that, include: The data acquisition module acquires multi-source heterogeneous data involved in the train operation process, including environmental data, equipment status data, and historical operation data. The modeling module is used to model the multi-source heterogeneous data using a dynamic Bayesian network, construct a causal relationship graph containing time slices, and generate an initial risk assessment model. The update module is used to update the node variables in the initial risk assessment model in real time based on the real-time collected train operation status data, so as to obtain the target risk assessment model. The assessment module is used to calculate the probability and severity of each risk factor based on the target risk assessment model using a conditional probability reasoning algorithm. The decision-making module is used to generate risk response strategies based on the probability and severity of occurrence of each risk factor using a decision tree algorithm. The monitoring module is used to introduce a distributed monitoring mechanism to continuously track the train's operating status and record feedback data to optimize the risk assessment model during the execution of the risk response strategy.

9. A computing device comprising a processing component and a storage component, the storage component storing one or more computer instructions for execution by the processing component to implement the train operation risk assessment method based on a dynamic Bayesian network, the processing component performing data acquisition, modeling, updating, assessment, decision-making, and monitoring operations, and the storage component storing intermediate results and final output.

Images