A highway bridge engineering construction safety monitoring method

Abstract

The application discloses a kind of highway bridge engineering construction safety monitoring methods.The highway bridge engineering construction safety monitoring method includes: obtaining highway bridge construction historical accident data;Structured information is extracted from accident report text to determine accident label and the hazard factor associated with the accident label;Using preset correlation rule mining algorithm processes the co-occurrence data of hazard factor in accident case, and the hazard factor correlation rule set representing the potential causal relationship between hazard factors is generated from co-occurrence data;Based on the causal semantics of hazard factor correlation rule set, the probability graph model of hazard factor is constructed, and the probability graph model represents the dependency between hazard factors;Based on the probability graph model, at least part of the sensitivity value of hazard factor is determined using the preset importance measure algorithm;According to the sensitivity value of hazard factor and the preset sensitivity threshold, determine the key hazard factor;According to the key hazard factor, the safety monitoring index of highway bridge engineering construction is determined.

Description

Technical Field

[0001] This invention relates to the field of bridge engineering, specifically to a method for monitoring the construction safety of highway bridges. Background Technology

[0002] Bridge construction is a high-risk operation, and bridge construction accidents can have serious consequences for personnel safety, project progress, and economic benefits.

[0003] In the field of highway bridge construction safety management, numerous researchers have proposed safety monitoring schemes from different perspectives. However, existing monitoring indicators mainly rely on empirical methods, which suffers from a high degree of subjectivity. The monitoring indicators proposed by different researchers based on their respective backgrounds differ, leading to insufficient universality. Existing research often focuses only on specific aspects of monitoring, lacking comprehensiveness and systematicity.

[0004] Therefore, existing technologies have limitations in monitoring the safety of highway bridge construction and urgently need improvement.

[0005] The background description is provided for the purpose of understanding the relevant technologies in this field and is not intended as an admission of prior art. Summary of the Invention

[0006] Therefore, the objective of this invention is to provide a method for monitoring the construction safety of highway bridge projects, so as to overcome the shortcomings of existing technologies in which the determination of monitoring indicators relies on subjective experience, lacks objective data support, and has insufficient universality.

[0007] In a first aspect, embodiments of the present invention provide a method for monitoring the construction safety of highway bridge projects, which may include:

[0008] Obtain historical accident data for highway bridge construction, including unstructured accident report text.

[0009] Structured information is extracted from the accident report text to determine the accident tag and the risk factors associated with the accident type tag;

[0010] The co-occurrence data of the risk factors in accident cases are processed using a preset association rule mining algorithm, and a set of risk factor association rules representing the potential causal relationship between risk factors is generated from the co-occurrence data.

[0011] Based on the causal semantics of the association rule set of risk factors, a probabilistic graphical model of risk factors is constructed, which represents the dependency relationship between risk factors.

[0012] Based on the probabilistic graphical model, a preset importance measurement algorithm is used to determine the sensitivity values ​​of at least some of the risk factors, wherein the sensitivity values ​​represent the contribution of the corresponding risk factor to the occurrence of the accident.

[0013] Based on the sensitivity values ​​of the risk factors and preset sensitivity thresholds, key risk factors are determined; and

[0014] Safety monitoring indicators for highway bridge construction are determined based on the key risk factors mentioned above.

[0015] In some embodiments, the step of processing the co-occurrence data of the risk-causing factors in accident cases using a preset association rule mining algorithm, and generating a set of risk-causing factor association rules characterizing the potential causal relationship between risk-causing factors from the co-occurrence data, includes:

[0016] Based on the accident labels, the risk factors are organized into a transaction dataset according to accident cases. Each accident case corresponds to a transaction, and each transaction contains all the risk factors involved in the corresponding accident case.

[0017] Perform frequent itemset mining on the transaction dataset to obtain frequent itemsets that satisfy the preset support constraints;

[0018] Candidate association rules are generated based on the frequent itemsets;

[0019] Determine the confidence and lift of each candidate association rule, and retain the candidate association rules that simultaneously satisfy the preset confidence and preset lift constraints for the risk factor association rule set.

[0020] In some embodiments, determining the sensitivity values ​​of at least some of the risk factors based on the probabilistic graphical model using a preset importance measurement algorithm includes:

[0021] The probability of an accident occurring under a given state is determined based on the probabilistic graphical model as the first probability value;

[0022] For each risk factor to be evaluated, the probability of an accident occurring when the probability of occurrence of the corresponding risk factor is set to a preset value is determined based on the probability graphical model and used as the second probability value.

[0023] Based on the first probability value and the second probability value, the sensitivity value of the corresponding risk factor is determined.

[0024] In some embodiments, extracting structured information from the accident report text further includes: determining process tags associated with the accident tags, the process tags representing the construction process of highway bridge engineering.

[0025] In these embodiments, the step of using a preset association rule mining algorithm to process the co-occurrence data of the risk-causing factors in accident cases, and generating a set of risk-causing factor association rules characterizing the potential causal relationship between risk-causing factors from the co-occurrence data, includes: generating association rules for risk-causing factors within the same process and association rules for risk-causing factors across processes based on the process labels, wherein the set of risk-causing factor association rules includes the association rules for risk-causing factors within the same process and the association rules for risk-causing factors across processes.

[0026] In these embodiments, determining the safety monitoring indicators for highway bridge construction based on the key risk factors includes: determining different sets of safety monitoring indicators corresponding to different processes and / or combinations of processes based on the construction process configuration of the highway bridge project.

[0027] In some embodiments, constructing a probabilistic graphical model of the risk-causing factors based on the causal semantics of the risk-causing factor association rule set includes:

[0028] Set the risk factors involved in the association rule set as nodes in the probabilistic graph network;

[0029] Determine the dependencies between risk-causing factors based on association rules, and establish directed edges between the corresponding risk-causing factor nodes;

[0030] The conditional probability table of the probabilistic graphical model is determined based on the statistical frequency and dependency of the risk factors.

[0031] In some embodiments, extracting structured information from the accident report text further includes determining time information associated with the accident tag.

[0032] In these embodiments, the step of using a preset association rule mining algorithm to process the co-occurrence data of the risk-causing factors in accident cases, and generating a set of association rules for risk-causing factors that characterize the potential causal relationship between risk-causing factors from the co-occurrence data, further includes: identifying combinations of risk-causing factors with time dependence, and generating corresponding association rules that include time constraints.

[0033] In these embodiments, determining the sensitivity values ​​of at least some of the risk-causing factors using a preset importance measurement algorithm based on the probabilistic graphical model further includes: adjusting the sensitivity values ​​of the risk-causing factors based on the time constraint according to the construction time of the highway bridge project.

[0034] In some embodiments, extracting structured information from the accident report text further includes: determining a bridge-type label associated with the accident label.

[0035] In these embodiments, the step of using a preset association rule mining algorithm to process the co-occurrence data of the risk-causing factors in accident cases, and generating a set of risk-causing factor association rules representing the potential causal relationship between risk-causing factors from the co-occurrence data, further includes: generating corresponding association rules containing bridge constraints based on the bridge type label.

[0036] In these embodiments, determining the sensitivity values ​​of at least some of the risk-causing factors using a preset importance measurement algorithm based on the probabilistic graphical model further includes: adjusting the sensitivity values ​​of the corresponding risk-causing factors based on the bridge type constraints of the highway bridge.

[0037] In some embodiments, determining the safety monitoring indicators for highway bridge construction based on the key risk factors includes:

[0038] Based on the distribution characteristics of the sensitivity values ​​of each key risk factor, the monitoring level of the corresponding safety monitoring indicator is determined by a hierarchical mapping rule.

[0039] The monitoring frequency and / or monitoring threshold range of the corresponding safety monitoring indicators are determined based on the monitoring level.

[0040] In some embodiments, the method for monitoring the construction safety of highway bridge projects further includes: dynamically adjusting the safety monitoring indicators based on accident data acquired in real time during the construction of highway bridge projects.

[0041] In a second aspect, embodiments of the present invention provide a highway bridge construction safety monitoring system, which may include:

[0042] The acquisition unit is configured to acquire historical accident data of highway bridge construction, wherein the accident data includes unstructured accident report text;

[0043] The extraction unit is configured to extract structured information from the accident report text to determine the accident tag and the risk factors associated with the accident type tag;

[0044] The rule set generation unit is configured to use a preset association rule mining algorithm to process the co-occurrence data of the risk factors in accident cases, and generate a risk factor association rule set representing the potential causal relationship between risk factors from the co-occurrence data;

[0045] The construction unit is configured to construct a probabilistic graphical model of the risk factors based on the causal semantics of the risk factor association rule set, wherein the probabilistic graphical model represents the dependency relationship between risk factors;

[0046] The sensitivity determination unit is configured to determine the sensitivity values ​​of at least some of the risk factors based on the probabilistic graphical model using a preset importance measurement algorithm, wherein the sensitivity values ​​characterize the contribution of the corresponding risk factor to the occurrence of the accident.

[0047] The key factor determination unit is configured to determine the key risk factor based on the sensitivity value of the risk factor and a preset sensitivity threshold.

[0048] The indicator determination unit is configured to determine safety monitoring indicators for highway bridge construction based on the key risk factors.

[0049] In the highway bridge construction safety monitoring method of this invention, historical accident data of highway bridge construction is acquired, structured information is extracted from unstructured accident report texts to determine accident tags and risk factors, a preset association rule mining algorithm is used to process the co-occurrence data of risk factors in accident cases to generate a risk factor association rule set, a probabilistic graphical model is constructed based on the causal semantics of the association rule set, a preset importance measurement algorithm is used to determine the sensitivity value of the risk factors, and key risk factors are determined according to the sensitivity value and a preset threshold, thereby determining the safety monitoring indicators for highway bridge construction. This solves the technical problems of existing methods for determining highway bridge construction safety monitoring indicators, such as strong subjectivity, lack of data support, and inability to quantify the importance of risk factors. Therefore, it enables the objectification and quantification of monitoring indicators, automatically identifies the intrinsic relationships between risk factors, accurately locates risk factors with key impacts on accidents, and obtains a scientific, systematic, and universally applicable highway bridge construction safety monitoring indicator system.

[0050] In a further embodiment of the present invention, by introducing construction process dimension analysis, risk patterns within the same construction process and risk propagation relationships between different construction processes can be identified. In particular, by distinguishing between association rules of risk-causing factors within the same process and association rules of risk-causing factors across processes, a precise grasp of the construction process context is achieved. This process context-based analysis can not only track the risk impact of preceding construction processes on subsequent processes, but also identify mutual interference between parallel construction processes, the risk accumulation effect of repeated construction procedures, and the risk evolution law in the construction process cycle. It comprehensively grasps the risk transmission path, cross-influence, and dynamic change characteristics, dynamically adjusts monitoring priorities according to the current construction progress, fully considers the continuity and correlation of the construction process, and significantly improves the pertinence and timeliness of risk prevention.

[0051] This invention further enhances the adaptability and flexibility of the method by introducing multi-dimensional analysis, including time and bridge type. Time-based analysis can identify seasonal risk combinations and time-sensitive risks, enabling precise monitoring of time-related risks. Bridge type-based analysis can identify unique risk patterns for different bridge types, providing customized monitoring solutions for each type.

[0052] This invention further enhances the adaptability of the method by introducing a dynamic update mechanism. This mechanism enables the system to continuously optimize the model based on newly acquired accident data during construction, thereby constantly improving the accuracy of monitoring indicators and the reliability of early warnings.

[0053] It will be understood that various embodiments of the present invention may achieve only some of the technical effects described above, without requiring all of them, and that additional optional features and technical effects may also be present in the various embodiments of the present invention. Other optional features and technical effects of the embodiments of the present invention are described in part below, and in part will become apparent from reading this document. Attached Figure Description

[0054] Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. The elements shown are not limited to the scale shown in the drawings, and the same or similar reference numerals in the drawings denote the same or similar elements, wherein:

[0055] Figure 1 An exemplary flowchart of a method for monitoring the construction safety of highway bridges according to an embodiment of the present invention is shown;

[0056] Figure 2 An exemplary flowchart of a method for monitoring the construction safety of highway bridges according to an embodiment of the present invention is shown;

[0057] Figure 3 An exemplary flowchart of a method for monitoring the construction safety of highway bridges according to an embodiment of the present invention is shown;

[0058] Figure 4 An exemplary flowchart of a method for monitoring the construction safety of highway bridges according to an embodiment of the present invention is shown;

[0059] Figure 5 An exemplary block diagram of a highway bridge construction safety monitoring system according to an embodiment of the present invention is shown; and

[0060] Figure 6 An exemplary structural diagram of an electronic device capable of implementing the method according to an embodiment of the present invention is shown. Detailed Implementation

[0061] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments and accompanying drawings. Here, the illustrative embodiments and descriptions of this invention are used to explain the invention, but are not intended to limit the invention.

[0062] The term "comprising" and its variations as used herein signify open inclusion, i.e., "including but not limited to". Unless otherwise stated, the term "or" means "and / or". The term "based on" means "at least partially based on". The terms "one example embodiment" and "one embodiment" mean "at least one example embodiment". The term "another embodiment" means "at least one additional embodiment". The terms "first", "second", etc., may refer to different or the same objects. Other explicit and implicit definitions may also be included below.

[0063] The present invention will be further described below with reference to embodiments and accompanying drawings. Specific embodiments are only used to further illustrate the present invention and do not limit the scope of protection of the claims.

[0064] As mentioned above, the existing methods for determining safety monitoring indicators for highway bridge construction have technical problems such as strong subjectivity, lack of data support, and inability to quantify the importance of risk factors.

[0065] To address the aforementioned technical problems, this invention provides a data-driven method for monitoring the construction safety of highway bridges. This method mines deep information from historical accident data, utilizes association rule mining and probabilistic graphical modeling techniques to automatically identify the intrinsic relationships between risk-causing factors, and quantifies the contribution of each risk-causing factor to the occurrence of the accident, thereby objectively and scientifically determining key safety monitoring indicators.

[0066] like Figure 1 The diagram shows a flowchart of a method for monitoring the construction safety of highway bridges according to an embodiment of the present invention. The method may include steps S110 to S150.

[0067] S110: Obtain historical accident data for highway bridge construction.

[0068] In this embodiment of the invention, the accident data includes unstructured accident report text.

[0069] In this embodiment of the invention, in step S110, historical accident data that occurred during the construction of highway bridges can be collected through multiple data sources.

[0070] In some embodiments, accident data may include accident investigation reports issued by government safety regulatory departments, internal accident records of construction companies, accident notices from industry associations, and public media reports. Accident information from these data sources is typically in unstructured text format, including detailed descriptions such as the time and location of the accident, the construction procedures involved, the type of accident, casualties, and an analysis of the cause of the accident. In some examples, during data acquisition, accident reports from publicly available channels can be automatically collected using web crawling technology in a legal, compliant, and privacy-protected manner. In some examples, manual import of internal company accident records is supported. In these embodiments, the time span of the accident data can be determined as needed. In some examples, to ensure data integrity and representativeness, the data collection time span typically covers 5-10 years, and the geographical scope covers construction projects with different climatic conditions and geological environments.

[0071] In some embodiments, data quality control and / or preprocessing mechanisms may also be established. In some examples, the collected raw data may first undergo deduplication to identify and remove duplicate incident records; then, data integrity checks may be performed to mark records lacking critical information; and finally, data consistency verification may be performed.

[0072] In this embodiment of the invention, the acquired accident report text typically contains a large number of technical terms and complex language expressions, such as "the bridge beam overturned due to the instability of the bridge erecting machine's outriggers during the T-beam erection process" and "the formwork burst during pier construction due to improper formwork reinforcement." These unstructured text data can be structured in the subsequent step S120 to extract information that can be used for quantitative analysis.

[0073] S120: Extract structured information from accident report text to determine accident tags and risk factors associated with those accident type tags.

[0074] In this embodiment of the invention, in step S120, the unstructured accident report text obtained in step S110 can be subjected to in-depth analysis and information extraction to convert the natural language description into a structured data form, providing standardized input for subsequent quantitative analysis.

[0075] In this embodiment of the invention, structured information extraction mainly includes: generating accident tags to describe the basic attributes of the accident; and identifying risk factors associated with the accident to analyze the causal mechanism of the accident.

[0076] In some embodiments, accident tags may include information across multiple dimensions, such as, but not limited to, accident type and optional accident level and / or casualty data. Accident type may include, for example, collapse accidents, falls from heights, object strikes, and mechanical injuries. Accident level may be classified, for example, according to the number of deaths and missing persons, as general accidents, major accidents, serious accidents, and extremely serious accidents. Specific casualty data may include, for example, the number of deaths and missing persons and the number of injured persons.

[0077] In some embodiments, risk factors may include a variety of different types, including but not limited to unsafe human behavior factors, unsafe conditions of objects, management deficiencies, and adverse environmental factors. In some examples, unsafe human behavior factors include violations of operating procedures, misoperations, and lack of concentration; unsafe conditions of objects include equipment malfunctions, lack of protection, and material defects; management deficiencies include insufficient safety training, inadequate supervision, and lack of emergency plans; and adverse environmental factors include severe weather and complex geological conditions.

[0078] In some embodiments, one or more optional additional information may also be extracted, including but not limited to the time of the accident, the type of bridge involved (such as beam bridge, arch bridge, cable-stayed bridge, etc.), the construction process involved (such as sub-projects (such as foundation engineering, substructure, superstructure, ancillary works, etc.), sub-items (such as pile foundation, abutment, pier, cap beam, precast beam, beam erection, etc.), construction procedures (such as drilling, steel cage installation, concrete pouring, etc.), etc., as further described below.

[0079] In some embodiments, suitable natural language processing algorithms can be used to extract structured information from accident report text. In some specific examples, the algorithms employed include one or more of text classification algorithms, dependency parsing algorithms, named entity recognition (NER) algorithms, rule matching algorithms, and keyword extraction algorithms. In some examples, text classification algorithms include Support Vector Machines (SVM) or Naive Bayes algorithms, for example, for automatically classifying accident types; in some examples, named entity recognition algorithms include Conditional Random Fields (CRF) or BiLSTM-CRF models, for example, for identifying entities such as time, location, and project names in accident reports; in some examples, dependency parsing algorithms include graph-based MST algorithms or transition-based Arc-Standard algorithms, for example, for parsing sentence structure and identifying subject-verb-object relationships and causal relationships; in some examples, keyword extraction algorithms include TF-IDF algorithms or TextRank algorithms, for example, for extracting risk factors; in some examples, rule matching algorithms can, for example, use regular expressions for regular expression matching.

[0080] In a specific example, the structured information extracted from the processed report text of a bridge pier construction formwork rupture accident includes: accident tags: {Accident type: collapse accident, accident level: general accident, number of deaths: 1, number of injuries: 3}; risk factors: {unsafe human behavior: excessive pouring speed, improper vibration, unsafe conditions of the structure: insufficient formwork strength, unstable support system, management deficiencies: insufficient technical briefing, lack of on-site supervision, adverse environment: high temperature}; time information: {time: August 2022}; bridge type tag: {bridge type: continuous beam bridge}; construction process tags: {sub-project: substructure, sub-item project: pier construction, work procedure: pier pouring}.

[0081] In this embodiment of the invention, the extracted structured information can be used for subsequent association rule mining. Here, by converting unstructured text into structured data, the standardization and computability of accident information are achieved.

[0082] S130: Use a pre-defined association rule mining algorithm to process the co-occurrence data of risk factors in accident cases, and generate a set of association rules for risk factors that characterize the potential causal relationship between risk factors from the co-occurrence data.

[0083] In this embodiment of the invention, in step S130, the extracted structured information can be subjected to in-depth data mining analysis. Through mining algorithms, potential correlations between different risk factors can be discovered, and risk factor correlation rules can be generated.

[0084] In optional embodiments, any suitable preprocessing can be performed. In some embodiments, construction accidents and their hazard factors can be encoded, converting the textual accident types and hazard factors into standardized coded forms to facilitate subsequent data mining processing. For example, "collapse accident" can be encoded as "ACC001", "violation of regulations" as "HF001", and "equipment failure" as "EF001", etc. This not only improves the efficiency of data processing but also ensures consistent representation of the same type of accident and hazard factor in different cases.

[0085] In this embodiment of the invention, the association rule mining process described in step S130 can be implemented based on the Apriori algorithm. By progressively adjusting the minimum support and minimum confidence, association rules of risk factors that meet the conditions can be effectively discovered.

[0086] like Figure 2 The diagram illustrates a second flowchart of a highway bridge construction safety monitoring method according to an embodiment of the present invention, demonstrating a specific implementation of association rule mining based on the Apriori algorithm. Specifically, as shown... Figure 2 As shown, step S130 may include steps S131 to S134:

[0087] S131: Organize the risk factors into a transaction dataset based on accident cases according to the accident labels.

[0088] In this embodiment of the invention, in step S131, the structured accident data is converted into a transaction data set format suitable for association rule mining. Each accident case corresponds to a transaction, and each transaction contains all the risk factors involved in the corresponding accident case, forming transaction itemsets.

[0089] In an optional embodiment, a unique identifier can be assigned to each accident case, and then all the risk factors identified in that case can be used as the itemset of that transaction. For example, the risk factor set for a bridge erecting machine collapse accident case (transaction ID: T001) is {violation of operating procedures, aging equipment, excessive wind speed, insufficient safety training}; the risk factor set for a support collapse accident case (transaction ID: T002) is {insufficient support strength, overloaded construction, inadequate technical briefing, lack of supervision}.

[0090] In some embodiments, transactions can also be categorized according to incident type, forming subsets of transactions for different incident types. This helps to discover unique risk-causing patterns for specific incident types.

[0091] S132: Perform frequent itemset mining on the transaction dataset to obtain frequent itemsets that satisfy the preset support constraints.

[0092] In this embodiment of the invention, in step S132, the Apriori algorithm can be used to discover frequent itemsets iteratively layer by layer. In a specific instance, support constraints can be used to discover frequent itemsets. Support is an indicator that measures the frequency of an itemset occurrence, and can be defined as the proportion of transactions containing that itemset to the total number of transactions.

[0093] In some embodiments, the mining process can start with a single risk factor and gradually expand to combinations containing multiple risk factors. In one specific embodiment, the entire transaction dataset can be scanned, the support of each risk factor can be calculated, and items with support not lower than a preset minimum support threshold can be retained as frequent 1-itemsets; then, candidate 2-itemsets can be generated based on the frequent 1-itemsets, the dataset can be scanned again to calculate the support, and frequent 2-itemsets can be retained; and so on, until no new frequent itemsets can be generated.

[0094] In some embodiments, pruning can be applied during the generation of candidate itemsets to improve algorithm efficiency. In one specific embodiment, when generating candidate k-itemsets, only those itemsets where all (k-1)-subsets are frequent can be considered. For example, before generating the candidate 3-itemset {violation, equipment failure, lack of oversight}, it is confirmed that all its 2-subsets {violation, equipment failure}, {violation, lack of oversight}, and {equipment failure, lack of oversight} are frequent itemsets. This pruning strategy significantly reduces the number of candidate itemsets to be examined.

[0095] S133: Generate candidate association rules based on frequent itemsets.

[0096] In this embodiment of the invention, in step S133, all possible association rules can be generated based on frequent itemsets.

[0097] In some embodiments, the generated candidate association rules can be in the form of "precedence (condition) → consequence (result)". In some embodiments, for frequent itemsets, all possible combinations of precedence and a single consequence can be generated. For example, for frequent k-itemsets (k≥2), all possible combinations of (k-1) precedence and 1 consequence can be generated. For example, for frequent 3-itemsets {violation, equipment failure, lack of supervision}, the following candidate rules can be generated: {violation, equipment failure}→{lack of supervision}, {violation, lack of supervision}→{equipment failure}, {equipment failure, lack of supervision}→{violation}, etc. In optional embodiments, the semantic rationality of the rules is also considered during the candidate association rule generation process, and the generated combination rules are semantically filtered to obtain candidate association rules.

[0098] S134: Determine the confidence and lift of each candidate association rule, and retain the candidate association rules that simultaneously meet the pre-set confidence and pre-set lift constraints for the risk factor association rule set.

[0099] In this embodiment of the invention, in step S134, the candidate association rules can be evaluated and screened by combining confidence and lift, so as to ensure that the rules that are finally retained have practical application value.

[0100] In this paper, confidence level measures the reliability of a rule and is defined as the proportion of transactions that contain both the antecedent and consequent to the total number of transactions that contain only the antecedent.

[0101] In this paper, lift measures the relevance of a rule and is defined as the ratio of the rule's confidence to the probability of the subsequent term occurring independently. A lift greater than 1 indicates a positive correlation between the antecedent and the consequent; the larger the value, the stronger the correlation.

[0102] In some embodiments, the aforementioned preset confidence and lift constraints can be achieved by setting appropriate minimum confidence and minimum lift thresholds, thereby filtering out association rules that are both reliable and significantly relevant. In some embodiments, candidate rules that are greater than or equal to the preset minimum confidence and minimum lift thresholds can be identified as strongly associated rules and will be included in the risk factor association rule set. In a specific example, the rule {working at height + lack of protection + strong winds} → {fall from height accident} has a confidence of 0.82 and a lift of 3.1, which far exceeds the set thresholds, and is therefore identified as a strongly associated rule and included in the final risk factor association rule set.

[0103] The risk factor association rule set output in step S130 can be used to subsequently build a probabilistic graphical model and identify key risk factors.

[0104] S140: Construct a probabilistic graphical model of risk factors based on the causal semantics of the association rule set of risk factors.

[0105] In this embodiment of the invention, the probabilistic graphical model can characterize the dependencies between risk factors.

[0106] In this embodiment of the invention, in step S140, a probabilistic graphical model capable of characterizing the dependencies between risk-causing factors can be constructed based on the set of association rules for risk-causing factors obtained in step S130. This probabilistic graphical model can quantitatively describe the causal relationships and influence intensity between different risk-causing factors through a combination of graph structure and probability distribution.

[0107] In some embodiments, the probabilistic graphical model is a Bayesian network structure, having a network topology for representing risk factors and their dependencies, and an optional conditional probability table for quantifying the strength of dependencies. In some embodiments, the probabilistic graphical model, particularly one based on a Bayesian network structure, may include risk factor nodes and directed edges representing dependencies between risk factors. Specifically, the Bayesian network represents the causal relationships between risk factors through a directed acyclic graph, where nodes represent risk factors and directed edges represent the direction of causal influence. Optionally, the conditional probability table quantitatively describes the degree of influence of a parent node on its child nodes; alternatively, directed edges may carry corresponding conditional probabilities. This structure can not only represent direct causal relationships but also capture indirect influences between risk factors through a probability propagation mechanism.

[0108] More specifically, such as Figure 3 As shown, a third flowchart of the highway bridge construction safety monitoring method according to an embodiment of the present invention is illustrated, detailing the construction of the probabilistic graphical model. Figure 3 As shown, step S140 may include steps S141 to S143:

[0109] S141: Set the risk factors involved in the association rule set as nodes in the probabilistic graphical network.

[0110] In this sub-step, all relevant risk factors are extracted from the risk factor association rule set, and each risk factor is mapped to a node in a probabilistic graphical network. The node settings follow the principles of completeness and independence, ensuring that all risk factors appearing in the association rules have a corresponding node representation.

[0111] Node types include root nodes and intermediate nodes. Root nodes represent basic risk factors that do not depend on other risk factors, such as "violation of operating procedures," "equipment aging," and "severe weather." Intermediate nodes represent derived risk factors or accident outcomes that are affected by other risk factors, such as "support instability" and "collapse accident." Each node has a discrete state space, typically employing binary states (occurring / not occurring) or multi-valued states (low / medium / high risk).

[0112] In a specific example, based on association rule analysis of a bridge construction accident database, 15 major risk-causing factor nodes were identified, including 4 personnel factor nodes (violation of regulations, fatigue, insufficient skills, and weak safety awareness), 4 equipment factor nodes (equipment failure, lack of protection, material defects, and improper maintenance), 4 management factor nodes (inadequate supervision, insufficient training, lack of systems, and insufficient emergency preparedness), and 3 environmental factor nodes (severe weather, poor geological conditions, and complex construction environment).

[0113] S142: Determine the dependencies between risk-causing factors based on association rules, and establish directed edges between the corresponding risk-causing factor nodes.

[0114] In this embodiment of the invention, in step S142, the dependency relationship between risk-causing factor nodes can be determined and directed edges can be established based on the association relationship of association rules. A directed edge points from a cause node to a result node, representing the directionality of causal influence. As mentioned above, the establishment of edges follows the confidence and lift indices of association rules. Directed edges can be established according to the association rules of the risk-causing factor association rule set, that is, directed edges can be established between corresponding risk-causing factor nodes based on strong association rules that satisfy a preset threshold. For example, the rule {equipment aging, improper maintenance} → {equipment failure} indicates that "equipment aging" and "improper maintenance" are the causes of "equipment failure," therefore, a directed edge is established from the first two nodes to the latter. Here, for multiple factors, a node is allowed to have multiple parent nodes, indicating that the risk-causing factor is affected by multiple factors.

[0115] In some embodiments, during the construction process, methods are also employed to ensure that the generated graph structure is a directed acyclic graph (DAG) to avoid circular dependencies. For example, potential cycles can be detected and eliminated using topological sorting algorithms. In some embodiments, when a cycle is detected, edges with stronger and / or logically reasonable edges can be retained and edges with weaker and / or logically unreasonable edges can be deleted based on the strength of the association rules and domain knowledge. Here, the strength can be determined, for example, by the aforementioned support, confidence, and / or lift.

[0116] S143: Determine the conditional probability table of the probabilistic graphical model based on the statistical frequency and dependency of the risk factors.

[0117] In this embodiment of the invention, in step S143, probability distribution parameters for each node, such as prior probability and conditional probability, can be calculated based on statistical analysis of historical accident data. In some embodiments, the conditional probability table corresponding to a probabilistic graphical model, such as a Bayesian network, can be determined based on the statistical frequency and dependency of the risk-causing factors.

[0118] In specific embodiments, the prior probability of each risk factor node can be calculated based on the frequency of occurrence of the risk factor. Further, the conditional probability of each risk factor can be calculated based on dependencies (directed edges) in a probabilistic graphical model such as a Bayesian network. For non-root nodes with parent nodes, their conditional probabilities (tables) can be determined through the conditional associations between the parent and child nodes. More specifically, the conditional probability can be calculated using the maximum likelihood estimation method, by statistically analyzing the frequency of occurrence of specific state combinations of child nodes and their parent nodes, and dividing by the frequency of occurrence of specific states of the parent node to obtain the corresponding conditional probability value. In some embodiments, when the sample size for certain conditional combinations is insufficient, smoothing techniques or expert knowledge supplementation are used to estimate the conditional probabilities. In some embodiments, the root node in the conditional probability table can be represented as a special form where the conditional part is an empty set, and its probability value is the prior probability without conditional constraints.

[0119] This step allows us to establish a complete conditional probability table for each risk factor node in the probabilistic graphical model, which can be used for subsequent sensitivity determination based on probabilistic reasoning.

[0120] S150: Based on a probabilistic graphical model, use a pre-defined importance measurement algorithm to determine the sensitivity values ​​of at least some risk factors.

[0121] In this embodiment of the invention, in step S150, the influence of each risk factor on the occurrence of an accident can be quantitatively evaluated based on the probabilistic graphical model constructed in step S140 using an importance measurement algorithm. The sensitivity value represents the contribution of the corresponding risk factor to the occurrence of the accident.

[0122] In this embodiment of the invention, the importance measurement algorithm can assess the importance of a risk factor by calculating the impact of changes in the state of the risk factor on the probability of an accident. In a further embodiment, the sensitivity value of each risk factor is quantified by comparing the probability of accidents under different assumptions.

[0123] In some embodiments, the importance measurement algorithm employs the Fussel-Vesely method. Specifically, sensitivity calculation based on the Fussel-Vesely method is achieved by evaluating the relative reduction in accident probability after eliminating a specific risk factor.

[0124] In other embodiments, the Risk Increase Factor (RIF) method or the Risk Reduction Factor (RRF) method may be used.

[0125] In other embodiments, methods such as Birnbaum importance measure or critical importance measure may also be used.

[0126] like Figure 4 The diagram illustrates the fourth flowchart of a highway bridge construction safety monitoring method according to an embodiment of the present invention, detailing the determination of sensitivity. Figure 4 As shown, step S140 may include steps S151 to S153:

[0127] S151: Determine the probability of an accident occurring under a given state based on a probabilistic graphical model as the first probability value.

[0128] In this embodiment of the invention, in step S151, the probability of a specific type of accident occurring under a given risk factor state can be determined as a first probability value based on a probabilistic graphical model. In some embodiments, this first probability value is used as a benchmark probability value for subsequent sensitivity calculations.

[0129] In embodiments of the present invention, the given state can be set in various ways. In some embodiments, the given state refers to the state in which each risk factor takes a value according to the probability distribution obtained from historical data analysis; that is, each risk factor node takes the distribution corresponding to its prior probability or conditional probability, reflecting the typical risk level based on historical accident data. In other embodiments, the given state can be a combination of risk factor states set for a specific assessment scenario, such as assessing the risk level under specific construction conditions.

[0130] More specifically, in this step S151, the probabilistic reasoning function of Bayesian networks can be used to calculate the probability of a specific type of accident occurring under the actual state of a given risk factor.

[0131] S152: For each risk factor to be evaluated, the probability of an accident occurring when the probability of occurrence of the corresponding risk factor is set to a preset value is determined based on the probabilistic graphical model and used as the second probability value.

[0132] In this embodiment of the invention, in step S152, a hypothetical analysis can be performed on each risk factor to be evaluated. While keeping the states of other risk factors unchanged, the state of the risk factor can be changed, for example, its influence can be eliminated or minimized, to determine the probability of an accident occurring. In some embodiments, the second probability value represents the probability of an accident occurring after the influence of the risk factor is changed or preferably eliminated; it can also be called an updated probability value or an influence elimination probability value.

[0133] In further embodiments, the preset value can be set in several ways. In some embodiments, the preset value is set to 0, indicating that the risk factor will not occur at all. In other embodiments, the preset value can be set to the minimum possible probability of the risk factor occurring. In alternative embodiments, the preset value can also be set to 1 to assess the increased risk when the risk factor is certain to occur. In other embodiments, the preset value can be set to the maximum possible probability of the risk factor occurring. In different embodiments, depending on the importance calculation method, the probability of the risk factor occurring can be selected as a minimum (or 0) or a maximum (or 1), which falls within the scope of this invention.

[0134] In this embodiment of the invention, in step S152, the states of other risk factors remain unchanged, and only the probability value of the target risk factor is changed to a preset value. The calculation process can be similar to S151, and a new accident probability is obtained by re-performing probability reasoning.

[0135] S153: Determine the sensitivity value of the corresponding risk factor based on the first probability value and the second probability value.

[0136] In this embodiment of the invention, in step S153, the sensitivity value of each risk factor can be calculated based on the first probability value and the second probability value, wherein the sensitivity value represents the contribution of the corresponding risk factor to the occurrence of the accident.

[0137] In this embodiment of the invention, different sensitivity value algorithms can be adopted according to different preset values ​​in S152. In some embodiments, when the preset value is 0 or the minimum value, the Fussel-Vesely algorithm can be used. More specifically, the sensitivity value can be obtained by dividing the difference between the baseline probability value and the updated accident occurrence probability by the baseline probability value, i.e., (first probability value - second probability value) / first probability value, which reflects the relative reduction in the accident probability after eliminating (or approximately eliminating) the risk-causing factor.

[0138] In other embodiments, when the preset value is 1 or the maximum value, the risk increase factor (RIF) can be calculated, which is defined as the ratio of the second probability value to the first probability value, i.e., the second probability value / the first probability value, which reflects the increase in accident risk when the risk factor is bound to occur (or approximately bound to occur).

[0139] In an optional embodiment, the two sensitivity indices described above can be calculated simultaneously, and the sensitivity value can be determined based on both, which falls within the scope of the present invention.

[0140] S160: Determine the key risk factors based on the sensitivity values ​​of the risk factors and the preset sensitivity thresholds.

[0141] In this embodiment of the invention, in step S160, the key risk factors that significantly contribute to the occurrence of the accident can be determined based on the sensitivity values ​​of each risk factor obtained in step S150, for example by comparing and / or sorting with a preset threshold.

[0142] In some embodiments, the sensitivity threshold can be determined using an expert scoring method. Specifically, experts in the field of bridge construction safety are organized to conduct a comprehensive evaluation based on the distribution characteristics of sensitivity values ​​and engineering practice experience.

[0143] In some embodiments, risk factors with sensitivity values ​​higher than a preset threshold are identified as critical risk factors. In other embodiments, all risk factors are ranked according to their sensitivity values, and the top-ranked risk factors are selected as critical risk factors.

[0144] S170: Determine safety monitoring indicators for highway bridge construction based on key risk factors.

[0145] In this embodiment of the invention, in step S170, safety monitoring indicators for highway bridge construction can be determined based on the key risk factors identified in step S160.

[0146] In this embodiment of the invention, it is possible to analyze which of these factors can be monitored by technical means and which technical means can be used for monitoring. Specifically, suitable detection devices, such as image acquisition units like cameras, sensors, detectors, or other means, can be used to monitor the determined safety monitoring indicators. For example, image acquisition units, such as cameras, can be used to monitor personnel-related key risk factors corresponding to personnel behavior and status monitoring indicators; any suitable device sensor or detector can be used to detect equipment-related key risk factors corresponding to equipment operating status and structural safety monitoring indicators; any suitable environmental sensor can be used to monitor environment-related key risk factors corresponding to environmental condition monitoring indicators; and any suitable means can be used to monitor management-related key risk factors corresponding to management execution monitoring indicators.

[0147] The method of this invention solves the technical problems of existing methods for determining safety monitoring indicators for highway bridge construction, such as strong subjectivity, lack of data support, and inability to quantify the importance of risk factors.

[0148] Based on the above embodiments, this invention takes into account that highway bridge construction is a complex system engineering project involving multiple construction processes. Different construction processes have their own risk characteristics, and there are temporal relationships and risk transmission effects between construction processes. Another embodiment of this invention further introduces control of the construction process dimension, that is, introduces "construction process context" to more accurately identify and control construction risks.

[0149] In this embodiment of the invention, the construction process context includes the position of the construction process in the overall construction sequence, the temporal relationship, hierarchical inclusion relationship, and mutual influence relationship between construction processes. Specifically, the temporal relationship refers to the sequential order and parallel relationship of construction processes; the hierarchical inclusion relationship refers to the superior-subordinate relationship between sub-projects, sub-items, and procedures; and the mutual influence relationship refers to the influence of preceding processes on subsequent processes, the interaction between parallel processes, and the coupling effect generated by process combinations.

[0150] In this embodiment, as described above, in step S120, in addition to determining the accident label and the hazard factor, a process label associated with the accident label is also determined. The process label characterizes the construction process of the highway bridge engineering project, including, for example, sub-projects (such as foundation engineering, substructure, superstructure, and ancillary works), sub-items (such as pile foundations, abutments, piers, cap beams, precast beams, and beam erection), and specific procedures (such as drilling, rebar cage installation, concrete pouring, formwork erection, and concrete vibration). In some embodiments, the process label may also reflect the context of the construction process (if any), such as, but not limited to: the higher-level project stage to which the process belongs, the position of the process in the construction sequence, parallel construction activities carried out simultaneously with the process, and completed preceding processes.

[0151] In this embodiment, optionally, in step S140, when constructing the probabilistic graphical model of the risk-causing factors based on the causal semantics of the risk-causing factor association rule set, process nodes can also be introduced into the probabilistic graphical model to establish conditional dependencies between processes and risk-causing factors and / or conditional dependencies between processes. Correspondingly, the probabilistic graphical model may also include process nodes and directed edges representing the conditional dependencies between processes and risk-causing factors and / or directed edges representing the conditional dependencies between processes. Specifically, directed edges are established between risk-causing factor nodes within the same process based on same-process association rules, and cross-process directed edges are established between risk-causing factor nodes in different processes based on cross-process association rules; conditional dependency edges are established from process nodes to their contained risk-causing factor nodes, and dependency edges between process nodes based on construction sequence are also established.

[0152] In this embodiment, optionally, in step S150, when determining the sensitivity values ​​of at least some risk factors using a preset importance measurement algorithm based on the probabilistic graphical model, the conditional probabilities of the risk factors in different processes and / or process combinations can be adjusted according to the dependencies between processes and / or the conditional dependencies between processes and risk factors. For example, the sensitivity values ​​of each risk factor under specific process conditions can be calculated by propagation, considering the dependency constraints of process nodes.

[0153] In this embodiment, in step S170, different sets of safety monitoring indicators corresponding to different processes and / or combinations of processes can be determined according to the construction process configuration of the highway bridge project.

[0154] In a further embodiment, the identified monitoring indicators can be organized into monitoring indicator sets corresponding to different construction processes based on process tags. During actual construction, the monitoring indicator set corresponding to the specific construction process configuration (including the sub-projects, sub-items, and procedures to be executed) is selected and deployed to monitor the corresponding indicators.

[0155] In another embodiment, different sets of safety monitoring indicators corresponding to different processes and / or combinations of processes can take into account the construction context. For example, for construction processes with the same procedures but different construction sequences, different number of repetitions of procedures, or different implementation methods of procedure combinations, different sets of monitoring indicators may be applicable based on the intra-process association rules and cross-process association rules identified in the previous steps and / or the probabilistic graphical model with process nodes. Through this monitoring configuration method based on the construction process context, the same construction process can be configured with different sets of monitoring indicators under different construction sequence positions, different preceding states, and different combinations, achieving accurate matching between the monitoring scheme and the actual construction risk state.

[0156] In this embodiment, by adding consideration of the construction process context in key steps such as information extraction, association rule mining, and monitoring indicator determination, refined risk analysis based on the construction process context is achieved.

[0157] The inventors considered that in actual bridge construction safety management, accident risks are not only related to the construction process, but also significantly affected by time factors. Therefore, another embodiment of the present invention introduces time-dimensional control.

[0158] In this embodiment, in step S120, in addition to extracting accident tags and risk factors, time information associated with accident tags can also be determined, including season, month, process duration, etc.

[0159] In this embodiment, in step S130, when processing the co-occurrence data of risk-causing factors in accident cases using a preset association rule mining algorithm, it is also possible to identify risk-causing factor combinations with time dependence and generate corresponding association rules containing time constraints. For example, risk-causing factor combinations that show a strong correlation within a specific season or time period can be identified.

[0160] In this embodiment, in step S150, when determining the sensitivity values ​​of at least some risk factors using a preset importance measurement algorithm based on a probabilistic graphical model, the sensitivity values ​​of risk factors can also be adjusted based on time constraints according to the construction time of the highway bridge project. Specifically, the sensitivity values ​​of risk factors under specific time conditions are adjusted based on the time distribution patterns of historical accidents.

[0161] Considering that different bridge types have different structural characteristics and construction processes, and their construction risk modes also differ significantly, another embodiment of the present invention introduces control of the bridge type dimension.

[0162] In this embodiment, in step S120, when extracting structured information from the accident report text, the bridge type label associated with the accident label can also be determined, such as beam bridge, arch bridge, cable-stayed bridge, suspension bridge, etc.

[0163] In this embodiment, in step S130, when processing the co-occurrence data of risk-causing factors in accident cases using a preset association rule mining algorithm, corresponding association rules containing bridge constraints can also be generated based on bridge-type labels. Specifically, association patterns of risk-causing factors containing specific bridge-type labels can be identified.

[0164] In this embodiment, in step S150, when determining the sensitivity values ​​of at least some risk factors using a preset importance measurement algorithm based on a probabilistic graphical model, the sensitivity values ​​of the corresponding risk factors can also be adjusted based on the bridge type and the associated pattern of the bridge type.

[0165] Another embodiment of the present invention provides a dynamic update mechanism based on real-time data.

[0166] In this embodiment, the method also includes dynamically adjusting safety monitoring indicators based on accident data acquired in real time during highway bridge construction. Specifically, this includes: continuously collecting newly occurring accident data during construction; performing structured information extraction and association rule mining on the newly collected accident data to obtain new risk factors and association rules; fusing and updating the newly identified association rules with existing risk factor association rule sets, for example, through incremental updates, i.e., adding new nodes and edges to the existing network structure and updating the conditional probability tables of existing nodes; reconstructing the probabilistic graphical model based on the fused association rule set; recalculating the sensitivity values ​​of each risk factor based on the updated probabilistic graphical model; and dynamically adjusting key risk factors and corresponding safety monitoring indicators according to the new sensitivity values.

[0167] The accident data here includes not only past safety incidents but also data generated during current construction operations, such as safety records and anomaly monitoring data. Through this dynamic update mechanism, the safety monitoring system can continuously improve and optimize itself as engineering practice progresses, achieving a transformation from static pre-setting to dynamic adaptation.

[0168] refer to Figure 5 This invention also provides a highway bridge construction safety monitoring system 500, which may include an acquisition unit 510, an extraction unit 520, a rule set generation unit 530, a construction unit 540, a sensitivity determination unit 550, a key factor determination unit 560, and an index determination unit 570.

[0169] The acquisition unit 510 is configured to acquire historical accident data of highway bridge construction, the accident data including unstructured accident report text;

[0170] The extraction unit 520 is configured to extract structured information from the accident report text to determine the accident tag and the risk factors associated with the accident type tag;

[0171] The rule set generation unit 530 is configured to use a preset association rule mining algorithm to process the co-occurrence data of the risk factors in accident cases, and generate a risk factor association rule set representing the potential causal relationship between risk factors from the co-occurrence data;

[0172] The construction unit 540 is configured to construct a probabilistic graphical model of risk factors based on the causal semantics of the risk factor association rule set, wherein the probabilistic graphical model represents the dependency relationship between risk factors.

[0173] The sensitivity determination unit 550 is configured to determine the sensitivity values ​​of at least some of the risk factors based on the probability graphical model and using a preset importance measurement algorithm, wherein the sensitivity values ​​characterize the contribution of the corresponding risk factor to the occurrence of the accident.

[0174] The key factor determination unit 560 is configured to determine key risk factors based on the sensitivity value of the risk factor and a preset sensitivity threshold.

[0175] The indicator determination unit 570 is configured to determine safety monitoring indicators for highway bridge construction based on the key risk factors.

[0176] In this embodiment of the invention, the system embodiment can refer to the method embodiment and implement the steps described in the method embodiment, and the features of the system embodiment can be incorporated into the method embodiment in a non-contradictory manner, and vice versa.

[0177] In some embodiments of the present invention, an electronic device is provided, including a processor and a memory storing a computer program, the processor being configured to implement the method of any embodiment of the present invention when running the computer program.

[0178] Figure 6 A schematic diagram is shown of a method or electronic device 600 that can be used to implement embodiments of the present invention. In some embodiments, the number of electronic devices may be more or less than the number shown. In some embodiments, it can be implemented using a single or multiple electronic devices. It can also be implemented using cloud or distributed electronic devices in some embodiments.

[0179] like Figure 6 As shown, the electronic device 600 includes a processor 601 and a memory 602. The processor executes programs stored in the memory, which, when executed by a computer, can implement the methods, steps, or functions described in the above embodiments. The processor 601 may include various types of processors, such as a central processing unit (CPU), a graphics processing unit (GPU), a neural network processor (NPU), a digital signal processor (DSP), etc. The processor 601 and the memory 602 are interconnected via a bus 603. Input / output (I / O) interfaces may also be connected to the bus 603.

[0180] The systems, apparatuses, modules, or units described in the above embodiments can be implemented by a computer or its associated components. The computer may be, for example, a mobile terminal, smartphone, personal computer, laptop computer, in-vehicle human-machine interface device, personal digital assistant, industrial computer, server, or a combination thereof.

[0181] Although not shown, in this embodiment of the invention, a program product is provided, including a computer program that, when executed by a processor, implements the method of any embodiment of the invention.

[0182] Although not shown, in an embodiment of the invention, a storage medium is provided storing a computer program configured to be run to implement the method of any embodiment of the invention.

[0183] Storage media in embodiments of the present invention include articles that are permanent or non-permanent, removable or non-removable, and can store information by any method or technology. Examples of storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information that can be accessed by a computing device.

[0184] The methods, programs, systems, apparatuses, etc., in embodiments of the present invention can be executed or implemented in one or more networked computers, or practiced in a distributed computing environment. In the embodiments of this specification, in these distributed computing environments, tasks can be performed by remote processing devices connected via a communication network.

[0185] Those skilled in the art will understand that the embodiments described in this specification can be provided as methods, systems, or computer program products. Therefore, those skilled in the art will realize that the functional modules / units or controllers and related method steps described in the above embodiments can be implemented in software, hardware, or a combination of both.

[0186] Unless explicitly stated otherwise, the actions or steps of the methods and procedures described in the embodiments of the present invention do not necessarily have to be performed in a specific order and can still achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0187] Although the present invention has been described in detail above with general descriptions and specific embodiments, modifications or improvements can be made to it, which will be obvious to those skilled in the art. Therefore, all such modifications or improvements made without departing from the spirit of the present invention fall within the scope of protection claimed by the present invention. In this document, several embodiments of the present invention have been described, but for the sake of brevity, the descriptions of the embodiments are not exhaustive, and identical or similar features or parts between the embodiments may be omitted. In this document, "one embodiment," "some embodiments," "example," "specific example," or "some examples" refers to embodiments applicable to at least one embodiment or example according to the present invention, but not all embodiments. The above terms do not necessarily refer to the same embodiment or example. Without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification and the features of the different embodiments or examples.

[0188] The exemplary systems and methods of the present invention have been specifically shown and described with reference to the above embodiments, which are merely examples of the best mode for implementing the systems and methods. Those skilled in the art will understand that various changes can be made to the embodiments of the systems and methods described herein without departing from the spirit and scope of the invention as defined in the appended claims when implementing the systems and / or methods.

Claims

1. A method for monitoring the construction safety of highway bridge projects, characterized in that, include: Obtain historical accident data for highway bridge construction, including unstructured accident report text. Structured information is extracted from the accident report text to determine the accident tag and the risk factors associated with the accident tag, including: determining the process tag associated with the accident tag, the process tag representing the construction process of the highway bridge engineering construction; The co-occurrence data of the risk factors in accident cases are processed using a preset association rule mining algorithm. The co-occurrence data is used to generate a set of risk factor association rules that characterize the potential causal relationship between risk factors. The process includes generating association rules for risk factors in the same process and association rules for risk factors across processes based on the process labels. The set of risk factor association rules includes the association rules for risk factors in the same process and the association rules for risk factors across processes. Based on the causal semantics of the association rule set of risk factors, a probabilistic graphical model of risk factors is constructed. The probabilistic graphical model represents the dependency relationship between risk factors. Each risk factor corresponds to an independent node in the probabilistic graphical model, and the directed edges between nodes represent the conditional dependency relationship between risk factors. The step of constructing a probabilistic graphical model of risk-causing factors based on the causal semantics of the risk-causing factor association rule set includes: setting the risk-causing factors involved in the association rule set as nodes of the probabilistic graphical network; determining the dependencies between risk-causing factors according to the association rules, and establishing directed edges between the corresponding risk-causing factor nodes; and determining the conditional probability table of the probabilistic graphical model according to the statistical frequency and dependencies of the risk-causing factors. The probabilistic graphical model further includes process nodes, directed edges representing the conditional dependencies between processes and risk factors, and directed edges representing the conditional dependencies between processes. Specifically, directed edges are established between risk factor nodes within the same process based on the same-process risk factor association rules, and cross-process directed edges are established between risk factor nodes in different processes based on the cross-process risk factor association rules. Furthermore, conditional dependencies are established between process nodes and their contained risk factor nodes, as well as dependency edges between process nodes based on construction sequence. Based on the probabilistic graphical model, a preset importance measurement algorithm is used to determine the sensitivity values ​​of at least some of the risk factors, wherein the sensitivity values ​​characterize the contribution of the corresponding risk factor to the occurrence of the accident, including: determining the probability of accident occurrence under a given state as a first probability value based on the probabilistic graphical model; for each risk factor to be evaluated, determining the probability of accident occurrence when the probability of occurrence of the corresponding risk factor is set to a preset value as a second probability value based on the probabilistic graphical model; and determining the sensitivity value of the corresponding risk factor based on the first probability value and the second probability value. The step of determining the sensitivity values ​​of at least some of the risk factors based on the probabilistic graphical model using a preset importance measurement algorithm further includes: adjusting the conditional probabilities of risk factors in different processes and / or process combinations according to the dependencies between processes and / or the conditional dependencies between processes and risk factors, and propagating the calculation of the sensitivity values ​​of each risk factor under specific process conditions under the constraint of the dependencies between process nodes. Based on the sensitivity values ​​of the risk factors and preset sensitivity thresholds, key risk factors are determined; and Safety monitoring indicators for highway bridge construction are determined based on the key risk factors mentioned above.

2. The method for monitoring the construction safety of highway bridges according to claim 1, characterized in that, The process of using a preset association rule mining algorithm to process the co-occurrence data of the risk-causing factors in accident cases, and generating a set of association rules representing the potential causal relationships between risk-causing factors from the co-occurrence data, includes: Based on the accident labels, the risk factors are organized into a transaction dataset according to accident cases. Each accident case corresponds to a transaction, and each transaction contains all the risk factors involved in the corresponding accident case. Perform frequent itemset mining on the transaction dataset to obtain frequent itemsets that satisfy the preset support constraints; Candidate association rules are generated based on the frequent itemsets; Determine the confidence and lift of each candidate association rule, and retain the candidate association rules that simultaneously satisfy the preset confidence and preset lift constraints for the risk factor association rule set.

3. The method for monitoring the construction safety of highway bridges according to claim 1, characterized in that, The step of determining the safety monitoring indicators for highway bridge construction based on the key risk factors includes: determining different sets of safety monitoring indicators corresponding to different processes and / or combinations of processes based on the configuration of the construction process of the highway bridge project.

4. The method for monitoring the construction safety of highway bridges according to claim 1, characterized in that, The step of extracting structured information from the accident report text also includes: determining the time information associated with the accident tag; The step of using a preset association rule mining algorithm to process the co-occurrence data of the risk-causing factors in accident cases, and generating a set of association rules for risk-causing factors that characterize the potential causal relationship between risk-causing factors from the co-occurrence data, further includes: identifying combinations of risk-causing factors with time dependence, and generating corresponding association rules containing time constraints. The step of determining the sensitivity values ​​of at least some of the risk-causing factors using a preset importance measurement algorithm based on the probabilistic graphical model further includes: adjusting the sensitivity values ​​of the risk-causing factors based on the time constraint according to the construction time of the highway bridge project.

5. The method for monitoring the construction safety of highway bridges according to claim 1, characterized in that, The step of extracting structured information from the accident report text further includes: determining the bridge-type tag associated with the accident tag; The step of using a preset association rule mining algorithm to process the co-occurrence data of the risk-causing factors in accident cases, and generating a set of association rules for risk-causing factors that characterize the potential causal relationship between risk-causing factors from the co-occurrence data, further includes: generating corresponding association rules containing bridge constraints based on the bridge type label; The step of determining the sensitivity values ​​of at least some of the risk-causing factors using a preset importance measurement algorithm based on the probabilistic graphical model further includes: adjusting the sensitivity values ​​of the corresponding risk-causing factors based on the bridge type constraints of the highway bridge.

6. The method for monitoring the construction safety of highway bridges according to claim 1, characterized in that, The determination of safety monitoring indicators for highway bridge construction based on the key risk factors includes: Based on the distribution characteristics of the sensitivity values ​​of each key risk factor, the monitoring level of the corresponding safety monitoring indicator is determined by a hierarchical mapping rule. The monitoring frequency and / or monitoring threshold range of the corresponding safety monitoring indicators are determined based on the monitoring level.

7. The method for monitoring the construction safety of highway bridges according to claim 1, characterized in that, Also includes: The safety monitoring indicators are dynamically adjusted based on accident data acquired in real time during the construction of highway bridge projects.

8. A safety monitoring system for highway bridge construction, characterized in that, include: The acquisition unit is configured to acquire historical accident data of highway bridge construction, wherein the accident data includes unstructured accident report text; An extraction unit is configured to extract structured information from the accident report text to determine an accident tag and the risk factors associated with the accident tag; wherein, the extraction unit is further configured to: determine a process tag associated with the accident tag, the process tag representing the construction process of the highway bridge engineering construction; The rule set generation unit is configured to process the co-occurrence data of the risk-causing factors in accident cases using a preset association rule mining algorithm, and generate a set of risk-causing factor association rules representing the potential causal relationship between risk-causing factors from the co-occurrence data; wherein, the rule set generation unit is further configured to: generate association rules for risk-causing factors within the same process and association rules for risk-causing factors across processes based on the process labels, wherein the set of risk-causing factor association rules includes the association rules for risk-causing factors within the same process and the association rules for risk-causing factors across processes; The construction unit is configured to construct a probabilistic graphical model of risk-causing factors based on the causal semantics of the risk-causing factor association rule set. The probabilistic graphical model represents the dependencies between risk-causing factors, wherein each risk-causing factor corresponds to an independent node in the probabilistic graphical model, and directed edges between nodes represent conditional dependencies between risk-causing factors. The construction unit is further configured to: set the risk-causing factors involved in the association rule set as nodes in the probabilistic graphical network; determine the dependencies between risk-causing factors according to the association rules, and establish directed edges between the corresponding risk-causing factor nodes; and based on the risk-causing factors… The statistical frequency and dependencies determine the conditional probability table of the probabilistic graphical model; wherein, the probabilistic graphical model also includes process nodes, directed edges representing the conditional dependencies between processes and risk factors, and directed edges representing the conditional dependencies between processes; wherein, based on the association rules of risk factors within the same process, directed edges are established between risk factor nodes within the same process, and based on the association rules of risk factors across processes, directed edges are established between risk factor nodes in different processes, and conditional dependencies from process nodes to their contained risk factor nodes and dependencies between process nodes based on construction sequence are established; A sensitivity determination unit is configured to determine the sensitivity values ​​of at least some of the risk-causing factors based on the probabilistic graphical model using a preset importance measurement algorithm, wherein the sensitivity values ​​characterize the contribution of the corresponding risk-causing factor to the occurrence of an accident; wherein the sensitivity determination unit is further configured to: determine the probability of accident occurrence under a given state as a first probability value based on the probabilistic graphical model; for each risk-causing factor to be evaluated, determine the probability of accident occurrence when the probability of occurrence of the corresponding risk-causing factor is set to a preset value as a second probability value based on the probabilistic graphical model; determine the sensitivity value of the corresponding risk-causing factor according to the first probability value and the second probability value; wherein the sensitivity determination unit is further configured to: adjust the conditional probability of the risk-causing factor in different processes and / or process combinations according to the inter-process dependency relationship and / or the conditional dependency relationship between the process and the risk-causing factor, and, under the constraint of the dependency relationship of process nodes, propagate and calculate the sensitivity value of each risk-causing factor under a specific process condition; The key factor determination unit is configured to determine the key risk factor based on the sensitivity value of the risk factor and a preset sensitivity threshold. The indicator determination unit is configured to determine safety monitoring indicators for highway bridge construction based on the key risk factors.

Images