Deep valley long-span bridge cantilever casting construction safety risk dynamic assessment method

Abstract

The invention relates to a deep-valley large-span bridge suspension pouring construction safety risk dynamic assessment method. The method comprises the following steps: S1, constructing a suspension pouring construction stage subjective and objective fusion analysis database; s2, preliminarily establishing a Bayesian network model, a network structure and network parameters, and carrying out adaptability optimization during target bridge risk assessment; s3, constructing a Bayesian network model adaptability module; s4, constructing a dynamic assessment guidance-time module, and establishing a risk dynamic assessment model without evidence guidance; s5, constructing a dynamic assessment guidance-evidence module, and establishing a risk dynamic assessment model with evidence guidance; according to the method, comprehensive safety evaluation of comprehensively considering the risk occurrence probability and the risk loss is realized, and the objectivity of safety evaluation is ensured; according to the method, the dynamic assessment of the risk is realized, so that the security risk assessment process is closely combined with the actual site construction, the reliability and accuracy of the assessment result are ensured, and finally a management basis is provided for a decision maker through a highly visual result.

Description

Technical Field

[0001] The invention relates to a dynamic assessment method for safety risks in suspended pouring construction of a deep valley large-span bridge, and belongs to the technical field of bridge construction safety risk assessment. Background Art

[0002] In recent years, my country's road network construction has developed rapidly. As a link connecting various geographical locations in my country, bridges have great significance for safety evaluation during the construction phase. Especially for large-span bridges under complex construction conditions, such as the deep valley large-span bridges in the mountainous areas of southwest my country, the construction technology is complex and the construction conditions are harsh. It is extremely important to strictly control the safety issues during the construction phase to ensure the safety and efficiency of the entire construction process. The existing safety evaluation methods for bridge construction safety consider a small safety evaluation system, rarely involving large-span bridges under complex construction conditions, and cannot reflect the time-varying characteristics of safety risks during the construction period.

[0003] At present, the existing bridge construction safety evaluation methods mainly evaluate bridges under general construction conditions. For deep valley long-span bridges under complex construction conditions, due to the many factors affecting the construction process and the complex influencing paths, the relevant bridge construction safety evaluation results are mostly based on the probability of risk occurrence or risk loss, and the evaluation results are relatively single and one-sided; the evaluation basis is mainly obtained by expert scoring, which is highly subjective. It is important that the probability of risk factors occurring during bridge construction is in a time-varying state, which is the difficulty of risk assessment work and is also a point that is easily overlooked in the research process. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide a method for dynamically assessing the safety risks of suspended cast-in-place construction of a deep valley large-span bridge, which can realize dynamic risk assessment of the entire bridge construction stage and ensure the safety of on-site construction.

[0005] The present invention adopts the following technical solution:

[0006] The dynamic assessment method for safety risk of suspended pouring construction of deep valley large-span bridges of the present invention adopts the following steps:

[0007] S1. Construct a subjective and objective fusion analysis data module, establish a target bridge safety risk assessment index system, collect the frequency and disaster degree of risk factors that have occurred in the target bridge, conduct special expert surveys, count the subjective and objective data of risk factors, and construct a subjective and objective fusion analysis database for the suspended pouring construction stage; S2. Use the Bayesian network as the basic evaluation model framework, preliminarily establish the Bayesian network model, network structure and network parameters, and optimize the adaptability when conducting the target bridge risk assessment; S3. Construct the Bayesian network model adaptability module; S4. Construct a dynamic assessment guidance-time module to establish a risk dynamic assessment model without evidence guidance; S5. Construct a dynamic assessment guidance-evidence module to establish a risk dynamic assessment model with evidence guidance; S6. Construct an integrated risk loss and risk value result analysis visualization model.

[0008] Furthermore, S1 specifically adopts the following steps:

[0009] S11. Obtain the construction design plan, construction organization design plan, safety management manual and hydrogeological survey data of the target bridge, and preliminarily complete the identification of risk sources of the target bridge; S12. Count the frequency of past risk sources and the degree of risk loss of the target bridge, and establish a database of safety factors for the suspended pouring construction of deep valley large-span bridges. 1 ; S13. Establish an expert evaluation scoring table for identifying risk sources, collect expert scoring data, and build an expert scoring database for safety factors of suspended cast-in-place construction of deep valley large-span bridges A 2 ; S14, perform database A 1 , A 2 Data fusion uses weighted fusion method, and the fusion formula is: In the formula, i takes values ​​of 1, 2, ω i Indicates the weight of the i-th database, and the weights are ω 1 ,ω 2 The weights are obtained based on the evaluation of experts in the bridge field. The data samples of each data source are x 1 , x 2 , x 3 , …, x n ; S15, database A 1 , A 2 Together they form a statistical database A for quantifying the risk sources of the target bridge, and pre-process the database A by filling in missing data values ​​and removing abnormal data values; S16, establishing a safety risk evaluation index system, specifically: constructing a three-level index system of target layer, criterion layer, and indicator layer, the target layer corresponds to whether the safety risk of the target bridge occurs, the criterion layer corresponds to whether various risk sources occur, and the indicator layer corresponds to whether a single risk source occurs.

[0010] Furthermore, S2 specifically adopts the following steps:

[0011] S21. Based on the safety risk assessment index system and the causal relationship between factors, safety risk factors are selected as nodes to establish the network structure topology of the target bridge risk assessment model;

[0012] S22. Based on the statistical database A, the probability of occurrence of the risk source is calculated using the following formula:

[0013] P 风险源x =n 风险源x / N,

[0014] In the formula, P 风险源x represents the calculated probability of occurrence of risk source x, n 风险源x Represents the number of risk sources x existing in database A, and N represents the statistical number of all risk sources in database A; the probability parameters of all nodes in the network structure topology of the risk assessment model are calculated according to the formula, and the node probability part of the network parameters of the target bridge assessment model is constructed; S23, based on the Bayesian network visualization UI platform, an initial basic assessment model is constructed, referring to the network structure topology, corresponding nodes are added to the Bayesian network visualization UI platform, and the nodes are connected using arrows representing the definition of causal relationships to complete the work of initializing the network structure; continue to define the nodes, introduce the calculated node probabilities, and initialize the network parameters.

[0015] Furthermore, in S3, the network parameter conditional probability table in the evaluation model is replaced by 2 n The term is transformed into 2n, and Leaky nodes are added to the basic assessment model to introduce leakage probability, so as to introduce the influence of secondary risk factors on the suspended pouring construction of deep valley large-span bridges into the basic assessment model; the specific steps are as follows: S31, the risk sources in the database are classified according to primary and secondary, and the risk sources as nodes in the network structure are set as primary risk sources, and the remaining risk sources are classified as secondary risk sources according to the risk source category; S32, the secondary risk sources are set as Leaky nodes, and the leakage probability is assigned. The leakage nodes and the corresponding nodes of the main risk sources belong to the same level; the influence of secondary risk sources on the safety of target bridge construction is introduced into the initial network structure to improve the adaptability of the assessment model to the assessment object and enhance the accuracy of the assessment results; S33, the Leaky Noisy-or theory is used to construct the conditional probability table of the complex assessment model. The Leaky Noisy-or theoretical calculation formula is as follows:

[0016]

[0017] P xi =(P|X 1 ,X 2 ,…,X i ,…,X n )

[0018] P L =(P|X 1 ,X 2 ,…,X n ,…,X L )

[0019] In the formula, P a (Y) represents the set of parent nodes of the Y node, Pxi represents the joint probability when one parent node of the Y node appears and the other parent nodes do not appear, P L Represents the joint probability of leaky nodes.

[0020] Furthermore, in S4, on top of the initial Bayesian network model, a temporal node, i.e., a transfer node, is created, and the time series and transfer probability are introduced to associate the initial network model on the adjacent time series. Specifically, the following steps are adopted: S41. The top-level node, deep valley large-span bridge suspended pouring construction safety risk, in the Bayesian network structure is selected as the transfer node, and its time series is assigned, and the transfer probability is introduced for it. The transfer node and the transfer probability together constitute the Bayesian network evaluation model temporal module, thereby constructing the temporal relationship of the initial Bayesian network of two adjacent time segments; S42. The number of construction days is the concept of time series time, i.e., each construction day corresponds to an initial evaluation model network model; S43. On the basis of the time series, the correlation between the two time segment evaluation models is added, i.e., the transfer probability is added to the time series. After the transfer probability is added, the construction of the initial evaluation model time module is realized, and the transfer probability is calculated using the following formula:

[0021]

[0022] In the formula, represents the value of the i-th state at time t, for The parent node of

[0023] S44. Run the target bridge suspended cast construction safety risk dynamic assessment model that introduces the time module to conduct a dynamic assessment of the safety risk of suspended cast construction of deep valley large-span bridges without evidence guidance.

[0024] Furthermore, in S5, several new nodes are created, relationships are established with the transfer nodes, and attributes of the new nodes are defined. Specifically, the following steps are adopted: S51. Several new nodes are connected to the transfer nodes of the time module to define evidence nodes. The causal arrow points to the transfer node, indicating that the node needs to be set as a risk influencing factor when it is defined. The arrow points to the newly added node, indicating that the node needs to be set as a risk characterization factor when it is defined; risk influencing factors include wind force during construction, rainfall during construction, and physical signs of construction workers; risk characterization factors include the vertical line shape of the construction beam, the stress of the construction beam, and the stress of the main components of the construction hanging basket; risk characterization factors mainly play an evidence guiding role, and the role of risk influencing factors is reflected in improving the risk assessment model and improving the adaptability of the model; S52. The monitoring data-risk probability level conversion method is adopted to convert The collected monitoring data are converted into the parameters required by the dynamic assessment model of the safety risk of the target bridge after preprocessing; the monitoring data-risk probability level conversion method specifically adopts the following steps: S521. Collect previous monitoring data from the construction site, and use the monitoring data-risk probability level conversion method to complete the conversion, and the results are used for model training and tuning; S522. Write a special construction monitoring plan, conduct on-site monitoring of the required monitoring data, and collect, summarize and preprocess the monitoring results; S53. Use the monitoring data-risk probability level conversion method again to convert the on-site monitoring results, introduce the conversion results into the corresponding nodes, realize the model evidence import, run the evidence module to build the target bridge risk dynamic assessment model, and realize the dynamic assessment of the safety risk of suspended cast-in-place construction of deep valley large-span bridges under the guidance of evidence.

[0025] Furthermore, S6 specifically adopts the following steps:

[0026] S61. Collect the risk assessment results of the target bridge as basic data, including the operation results of the dynamic Bayesian network model without or with evidence, that is, the dynamic probability of safety risk P, and collect and count the analysis results data; S62. Based on the statistical database A, use fuzzy rules to calculate the loss membership Q and risk factor weight W in the risk loss assessment; According to the statistical data of the risk loss degree in database A, obtain the membership vector Q of each risk factor by weighted average; Use the 1-9 scoring method to obtain the weight judgment matrix between risk factors, and use the SPSS data analysis tool to calculate the risk factor weight W; S63. Based on the Python language, build an integrated risk loss and risk value result analysis visualization model; Among them: According to the definition of risk loss C, C = Q × W, first complete the risk loss calculation and visualization; Further according to the definition of risk value R, R = C × P, complete the calculation and visualization of risk value R.

[0027] The positive effects of the present invention are as follows: the present invention is based on the Bayesian network (BN / DBN) framework, combined with the subjective and objective data fusion database, to achieve a comprehensive assessment of the risk probability of the suspended casting construction stage of the deep valley large-span bridge from a qualitative and quantitative perspective; at the same time, the real-time monitoring data of the on-site construction bridge status is introduced, including the linearity, stress and construction influencing factors of the construction beam body, to achieve a dynamic assessment of the safety of bridge construction, and the assessment results can be used by the target bridge construction managers to make accurate decisions.

[0028] The monitoring data-risk probability level conversion method cleverly links the real-time construction monitoring data with the safety risk, providing reliable evidence for the temporalization of the dynamic assessment model. The fuzzy comprehensive evaluation method is used to conduct risk loss assessment. Through fuzzy rules, combined with the historical risk data of the target bridge and expert survey data, the individual subjectivity of experts is fully reduced.

[0029] The present invention realizes a comprehensive safety evaluation that takes into account the probability of risk occurrence and risk loss; secondly, it can appropriately weaken the subjective influence of experts to ensure the objectivity of safety evaluation; at the same time, it can realize dynamic risk evaluation, so that the safety risk evaluation process is closely integrated with the actual on-site construction, ensuring the reliability and accuracy of the evaluation results, and finally providing a management basis for decision makers through highly visualized results. The present invention completes the dynamic evaluation of safety risks in complex working conditions during the suspended pouring construction of deep valley large-span bridges through improvement and adaptability optimization. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 It is a flowchart of the present invention;

[0031] Figure 2 This is a flow chart of the monitoring data-risk probability level conversion method of the present invention;

[0032] Figure 3 This is a schematic diagram of the target bridge risk assessment index system of the present invention;

[0033] Figure 4 A schematic diagram of the network structure topology of the target bridge safety risk assessment model of the present invention;

[0034] Figure 5 Constructing a flow chart for the evaluation result visualization model of the present invention;

[0035] Figure 6 This is a schematic diagram of the structure of the visualization model for the evaluation results of the present invention. DETAILED DESCRIPTION

[0036] The present invention will be further described below in conjunction with the accompanying drawings and specific examples.

[0037] like Figure 1As shown, the present invention provides a dynamic assessment method for safety risk of suspended pouring construction of deep valley large-span bridges, which adopts the following steps:

[0038] Step S1, constructing a subjective and objective fusion analysis data module: identifying the risk factors affecting the safety of cantilevered large-span bridge construction in deep valleys, establishing a safety evaluation index system for target bridges, collecting the frequency and degree of disaster-causing risk factors that have occurred in target bridges, conducting special expert surveys, and statistically analyzing the subjective and objective data of risk factors, and constructing a subjective and objective fusion analysis database for the cantilevered construction stage; the specific steps are as follows:

[0039] Step S11: Conduct on-site inspections of deep valley large-span bridges under construction, obtain construction design plans, construction organization design plans, safety management manuals, and hydrogeological survey data of target bridges, and preliminarily complete risk source identification of target bridges;

[0040] Step S12: Count the frequency and degree of the risk sources of the target bridge in the past, and establish a database A of safety factors for the suspended pouring construction of deep valley large-span bridges 1 ;

[0041] Step S13: Establish an expert evaluation scoring table for identifying risk sources, collect expert scoring data, and build an expert scoring database A for safety factors of suspended cast-in-place construction of deep valley large-span bridges 2 ;

[0042] Step S14: Database A 1 , A 2 Data fusion uses weighted fusion method, and the fusion formula is:

[0043]

[0044] In the formula, i takes values ​​of 1, 2, ω i Indicates the weight of the i-th database, and the weights are ω 1 ,ω 2 The weights are obtained based on the evaluation of experts in the bridge field. The data samples of each data source are x 1 , x 2 , x 3 , …, x n ;

[0045] Step S15: Database A 1 , A 2 Together they form a statistical database A for quantifying the risk sources of the target bridge, and preprocess the database A by filling in missing data values ​​and removing abnormal data values;

[0046] S16. Establish a safety risk assessment index system, specifically: construct a three-level index system of target layer, criterion layer, and index layer. The target layer corresponds to whether the safety risk of the target bridge occurs, the criterion layer corresponds to whether various risk sources occur, and the index layer corresponds to whether a single risk source occurs. Furthermore, for the risk source assessment scoring table of the target bridge, refer to the provisions on risk probability and risk loss in the "Technical Specifications for Risk Management of Railway Construction Projects" (QCR9006-2004), and attach the scoring standards to the scoring table for experts to refer to for scoring.

[0047] In order to deal with the weight problem of different experts in the bridge field, an expert weighting table is defined, and the expert weighting table is shown in Table 1:

[0048] Table 1 Expert empowerment table:

[0049]

[0050] Get the expert weight coefficient, the weight coefficient is calculated by the formula:

[0051]

[0052] In the formula, D t represents the total score of the t-th expert, P is the total number of experts, and λ t represents the weight of the t-th expert, where λ t ∈[0,1];

[0053] The questionnaire was distributed to experts in related fields such as universities, construction units, and research institutes. A total of 100 questionnaires were distributed and used as data standards;

[0054] Collect expert feedback data and calculate the expert evaluation results of risk factors. The results constitute the expert scoring database of safety factors of suspended cast-in-place construction of deep valley large-span bridges. 2 , calculation formula:

[0055]

[0056] In the formula, λ ijt It means that when the state of risk factor j is i, the tth is the weight coefficient of the expert, It represents the probability when the risk factor j is in state i.

[0057] Step S2: Using the Bayesian network as the basic assessment model framework, preliminarily establish the Bayesian network model, network structure and network parameters, and optimize the adaptability when conducting the target bridge risk assessment; the specific steps are as follows:

[0058] S21. Based on the safety risk assessment index system and the causal relationship between factors, safety risk factors are selected as nodes to establish the network structure topology of the target bridge risk assessment model;

[0059] Step S2: Using the Bayesian network as the basic assessment model framework, preliminarily establish the Bayesian network model, network structure and network parameters, and optimize the adaptability when conducting the target bridge risk assessment;

[0060] S22. Based on the statistical database A, the probability of occurrence of the risk source is calculated using the following formula: P 风险源x =n 风险源x / N,

[0061] In the formula, P 风险源x represents the calculated probability of occurrence of risk source x, n 风险源x represents the number of risk sources x in database A, and N represents the statistical number of all risk sources in database A; the probability parameters of all nodes in the network structure topology of the risk assessment model are calculated according to the formula, and the node probability part of the network parameters of the target bridge assessment model is constructed;

[0062] S23. Build an initial basic evaluation model based on the Bayesian network visualization UI platform. Refer to the network structure topology, add corresponding nodes in the Bayesian network visualization UI platform, connect the nodes using arrows representing causal relationships, and complete the initialization of the network structure. Continue to define the nodes, introduce the calculated node probabilities, initialize the network parameters, and then improve and optimize the adaptability.

[0063] Step S3, construct the Bayesian network model adaptability module: transform the network parameter conditional probability part table in the evaluation model from 2 n The term is transformed into 2n, which simplifies the calculation. At the same time, a "Leaky" node is added to the initial model, and the "leakage probability" is introduced to introduce the influence of secondary risk factors on the suspended pouring construction of deep valley large-span bridges into the initial Bayesian network evaluation model, so as to improve the adaptability of the model to the safety risk evaluation index system of suspended pouring construction of deep valley large-span bridges; specifically:

[0064] Step S31, classify the risk sources in the database according to primary and secondary categories, set the risk sources that are nodes in the network structure as primary risk sources, and classify the remaining risk sources as secondary risk sources according to the risk source categories;

[0065] Step S32: Set the "secondary risk source" as the "leakage node" and assign a leakage probability. The leakage node and the node corresponding to the primary risk source belong to the same level. The impact of the secondary risk source on the construction safety of the target bridge is introduced into the initial network structure to improve the adaptability of the assessment model to the assessment object and enhance the accuracy of the assessment results.

[0066] Step S33: Use the Leaky Noisy-or theory to construct a conditional probability table for a complex evaluation model, simplify the amount of calculation, and adapt to the risk assessment process of multiple complex construction conditions of deep valley large-span bridge suspended casting construction. The Leaky Noisy-or theoretical calculation formula is:

[0067]

[0068] P xi =(P|X 1 ,X 2 ,…,X i ,…,X n )

[0069] P L =(P|X 1 ,X 2 ,…,X n ,…,X L )

[0070] In the formula: P a (Y) represents the set of parent nodes of the Y node, Pxi represents the joint probability when one parent node of the Y node appears and the other parent nodes do not appear, P L represents the joint probability of “Leaky” nodes;

[0071] Step S4, constructing a dynamic assessment guidance module - time module: on the initial Bayesian network model, create a time-series node, namely a "transfer node", introduce time series and transfer probability, associate the initial network model on the adjacent time series, and construct a dynamic assessment model for the risk of suspended pouring construction of deep valley large-span bridges without evidence guidance; specifically:

[0072] Step S41, select the top-level node "safety risk of suspended pouring construction of deep valley large-span bridge" in the network structure as the "transfer node" and assign it a time series to build a time series relationship of the initial Bayesian network of two adjacent time segments;

[0073] Step S42, setting "construction days" as the concept of "time" in the time series, that is, each construction day corresponds to an initial evaluation model network model;

[0074] Step S43: On the basis of the time series, add the correlation between the two time segment evaluation models, that is, add "transition probability" to the time series. After the "transition probability" is added, the construction of the initial evaluation model "time module" is realized. The following is the calculation formula for the transition probability:

[0075]

[0076] In the formula, represents the value of the i-th state at time t, for Step S44, run the target bridge suspended pouring construction safety risk dynamic assessment model introduced by the "time module", and conduct a dynamic assessment of the risk of suspended pouring construction of deep valley large-span bridges without evidence guidance;

[0077] Step S5, constructing a dynamic assessment guidance module - evidence module: creating several new nodes, establishing relationships with transfer nodes, defining new node attributes, and constructing a dynamic assessment model for the risk of suspended pouring construction of deep valley large-span bridges with evidence guidance; the details are as follows:

[0078] Step S51: Connect several new nodes to the transfer node of the time module to define evidence nodes. The causal relationship arrow points to the transfer node, indicating that the node needs to be set as a risk influencing factor when it is defined. The arrow points to the newly added node, indicating that the node needs to be set as a risk characterization factor when it is defined.

[0079] Risk influencing factors include wind force during construction, rainfall during construction and physical signs of construction workers; risk characterization factors include vertical alignment of construction beams, stress of construction beams and stress of main components of construction hanging baskets; risk characterization factors mainly play an evidence-guiding role, and the role of risk influencing factors is reflected in improving the risk assessment model and enhancing the adaptability of the model;

[0080] S52. The monitoring data-risk probability level conversion method is used to convert the collected monitoring data into the parameters required by the target bridge safety risk dynamic assessment model after preprocessing. The monitoring data-risk probability level conversion method specifically adopts the following steps:

[0081] S521. Collect previous monitoring data from the construction site, use the monitoring data-risk probability level conversion method to complete the conversion, and use the results for model training and optimization;

[0082] S522. Write a special construction monitoring plan, conduct on-site monitoring of the required monitoring data, and collect, summarize and pre-process the monitoring results;

[0083] S53. Use the monitoring data-risk probability level conversion method again to convert the field monitoring results, introduce the conversion results into the corresponding nodes, realize the model evidence import, run the target bridge risk dynamic assessment model built by the evidence module, and realize the dynamic assessment of the safety risk of suspended cast-in-place construction of deep valley large-span bridges under the guidance of evidence.

[0084] like Figure 2 As shown in the figure, the monitoring data-risk probability level conversion method converts the monitoring data into the content required by the assessment model; the details are as follows:

[0085] The monitoring data-risk probability level conversion method consists of two key parts: the first is the acquisition of the theoretical value and the measured value of the monitoring data; the second is the determination of the threshold value of the difference between the theoretical value and the actual value;

[0086] The measured values ​​of the monitoring data, for risk characterization factors, are obtained by arranging measurement points at the construction site to obtain the stress and deformation of the beam body during hanging pouring. For risk influencing factors, such as the influence of wind, they are obtained through on-site wind speed detectors. For the theoretical values ​​of the monitoring data, the theoretical value data is obtained by obtaining bridge construction data and establishing a finite element analysis model.

[0087] After obtaining the theoretical value and measured value of bridge construction monitoring data, the difference between them is calculated. By consulting specifications, literature, and construction design plans, the specification limit between the theoretical value and the actual value of monitoring data is found. Then, after consulting risk management literature, specifications, and combined with expert discussion and review, the difference threshold is determined, thereby establishing a connection between monitoring data and the probability of risk occurrence.

[0088] Step S6, constructing an integrated risk loss and risk value result analysis visualization model: Based on fuzzy rules, the safety risk loss assessment of the cantilever cast-in-place construction of a deep valley large-span bridge is completed, the assessment results of the risk dynamic probability P are integrated, and the risk value result analysis and visualization are completed, as follows:

[0089] S61. Collect the risk assessment results of the target bridge as basic data, including the operation results of the dynamic Bayesian network model without or with evidence, that is, the risk dynamic probability P, and collect and count the analysis results data;

[0090] S62. Based on the statistical database A, fuzzy rules are used to calculate the loss membership Q and risk factor weight W in risk loss assessment; according to the statistical data of risk loss degree in database A, the membership vector Q of each risk factor is obtained by weighted average; the weight judgment matrix between risk factors is obtained by 1 to 9 scoring method, and the risk factor weight W is calculated using SPSS data analysis tools.

[0091] S63. Based on Python language, construct an integrated risk loss and risk value result analysis visualization model; wherein: based on the definition of risk loss C, C = Q × W, first complete the risk loss calculation and visualization; further based on the definition of risk value R, R = C × P, complete the calculation and visualization of risk value R.

[0092] Example:

[0093] A deep valley long-span bridge is located in the mountainous area of ​​southwest my country, spanning over the canyon. The main span of the bridge is a (65+120+65)m continuous rigid frame. Both ends of the bridge are laid along the ridge slope, and the terrain is undulating. Natural disasters such as landslides and mudslides occur frequently during the rainy season. The hydrogeological conditions are harsh and the construction is extremely difficult. The main beam of the bridge is a concrete box beam, which is constructed by cantilever casting with a hanging basket. The construction process is relatively complicated. In order to accurately control the safety status of the bridge under this special construction condition in real time, it is necessary to conduct a dynamic safety risk assessment of the construction process.

[0094] Step 1: Construct a subjective and objective fusion analysis data module: Identify the risk factors that affect the safety of cantilevered large-span bridge construction in deep valleys, establish a safety evaluation index system for target bridges, collect the frequency and degree of disaster-causing risk factors that have occurred in target bridges, conduct special expert surveys, collect subjective and objective data on risk factors, and construct a subjective and objective fusion analysis database for the cantilevered construction stage;

[0095] For the target bridge, analyze its cantilever pouring construction process, identify risks according to the types of risks it has, and identify risk sources including equipment risk sources A: A 1 , A 2 , A 3 , …, A n ; Material risk source B: B 1 , B 2 , B 3 , …, B n ; Risk source of construction method C: C 1 , C 2 , C 3 ,…,C n ; Environmental risk source D: D 1 , D 2 , D 3 ,…,D n ; Behavioral risk source E: E 1 、E 2 、E 3 ,…,E n ; Protection risk source F: F 1 、F 2 、F 3 ,…,F n ; Monitoring risk source G: G 1 , G 2 , G 3 , …, G n ; A total of 41 risk sources were identified;

[0096] like Figure 3As shown in the figure, based on the risk source identification results, a target bridge risk assessment index system is established. The risk assessment index system is divided into three layers: target layer, criterion layer and indicator layer. The target layer corresponds to the "safety risk of suspended pouring construction of deep valley large-span bridges"; the criterion layer corresponds to the risk source types [A, B, C, D, E, F, G]; the indicator layer corresponds to the risk source [A 1 ,A 2 ,A 3 ,…,G n ].

[0097] Based on the risk source identification results, the risk factors of the target bridge were first counted, including the frequency of risk source occurrence and the degree of disaster caused by the risk source. The statistical results were used to construct a database of safety factors for the suspended casting of deep valley large-span bridges. 1 ; Secondly, a special risk source expert scoring survey was conducted, and the scoring content included the risk source occurrence probability rating and risk loss rating. The statistical results were used to construct an expert scoring database of safety factors for cantilever cast-in-place construction of long-span bridges in the valley. 2 ; The scoring table for the safety risk investigation of suspended cast-in-place construction of deep valley large-span bridges is shown in Table 2.

[0098] Table 2 Scoring table for safety risk survey on suspended cast-in-place construction of deep valley large-span bridges

[0099]

[0100]

[0101] The expert scoring results need to be further processed to weaken the subjective influence of the experts. Taking 5 experts as an example, the weights are assigned as shown in Table 3:

[0102] Table 3 Expert weight assignment

[0103]

[0104] After obtaining the expert weight coefficient, the probability of the investigated content is calculated. Taking the "environmental risk" risk as an example, the calculation results are shown in Table 4:

[0105] Table 4 Summary of risk source scoring results Table 4 Summary of risk source scoring results

[0106]

[0107]

[0108] Step 2: Construct an initial assessment method model: Use the Bayesian network as the basic assessment model framework, preliminarily establish the Bayesian network model network structure and network parameters, and then optimize the adaptability of the target bridge risk assessment work;

[0109] like Figure 4As shown in the figure, according to the target bridge safety risk evaluation index system and based on the causal relationship between risk identification results, the network structure of the target bridge risk assessment model is established; specifically: the bottom layer is the basic parent node, which refers to the risk source identified in the safety risk evaluation index system; the middle-layer child nodes are the categories to which the risk source belongs; the final child node refers to whether the safety risk of the target bridge occurs or not, and the connecting arrows between different levels represent the causal relationship between nodes.

[0110] Construct the network parameters of the Bayesian network risk assessment model and collect database A 1 With database A 2 The data on the probability of risk occurrence is fused, and the fusion result is used as the basic node probability in the network parameters of the Bayesian network risk assessment model;

[0111] Importing this part of network parameters into the initial Bayesian network model, waiting for further optimization of the Bayesian network risk assessment model;

[0112] Step 3: Construct the model adaptability module: transform the network parameter conditional probability part of the evaluation model into 2 n The term is transformed into 2n to simplify the calculation. At the same time, a "Leaky" node is added to the initial model, and the "leakage probability" is introduced to introduce the influence of secondary risk factors on the suspended pouring construction of deep valley large-span bridges into the model, thereby improving the adaptability of the model to the safety risk evaluation index system of suspended pouring construction of deep valley large-span bridges.

[0113] The impact of the secondary risk sources identified in the database on the construction safety of the target bridge is introduced into the initial network structure to further improve the adaptability of the assessment model to the assessment object and enhance the accuracy of the assessment results;

[0114] Calculate the probability parameters of the leaky nodes, and use the Leaky Noisy-or theory to construct a conditional probability table for the complex evaluation model, connect the network parameters in step 2, and complete the overall network parameter construction of the evaluation model;

[0115] Step 4: Construct a dynamic assessment guidance module - time module: On top of the initial Bayesian network model, create a time-series node, namely the "transfer node", introduce time series and transfer probability, associate the initial network model on adjacent time series, and construct a dynamic assessment model for the risk of suspended pouring construction of deep valley large-span bridges without evidence guidance;

[0116] The final sub-node "safety risk of suspended pouring construction of deep valley large-span bridge" is selected as the basic node of the dynamic Bayesian network time module, that is, the transfer node. The transfer time segment selects 1 construction day as a segment, and the time series selects 15 days as the period. The transfer probability is introduced for it. The calculation results of the transfer probability are shown in Table 5:

[0117] Table 5 Transition probability calculation results:

[0118]

[0119]

[0120] Step 5: Construct a dynamic assessment guidance module - evidence module: Create several new nodes, establish relationships with transfer nodes, define new node attributes, and construct a dynamic assessment model for the risk of suspended pouring construction of deep valley large-span bridges guided by evidence;

[0121] The final sub-node "Safety Risk of Cast-in-place Construction of Deep Valley Large-span Bridge" was selected as the basic node of the dynamic Bayesian network evidence module. For the dynamic risk assessment of the bridge, the vertical linear deviation and beam stress of the construction bridge beam were selected as risk characterization factors, and its monitoring data were obtained. The monitoring data-risk probability level conversion method was introduced to complete the introduction of evidence; the measured value of the beam monitoring data was obtained by on-site real-time monitoring. Measuring points were arranged on the hanging basket construction beam, and three measuring points were arranged on the upper edge of the box beam, namely T1, T2, and T3, of which T1 and T3 were located on the center line of the box beam web, 100 mm away from the top surface of the box beam; T2 was located on the central axis of the box beam, 100 mm away from the top surface of the box beam; three measuring points were arranged on the lower edge of the box beam, namely B1, B2, and B3, of which B1 and B3 were located on the center line of the box beam web, 100 mm away from the bottom surface of the box beam; T2 was located on the central axis of the box beam, 100 mm away from the bottom surface of the box beam, and the monitoring period was 15 days;

[0122] The theoretical value of the beam monitoring data is obtained based on the finite element software modeling and analysis. The measured value of the monitoring data is compared with the theoretical value to obtain the difference between the two, which will provide evidence for the evidence module later.

[0123] With reference to the relevant contents of the "Technical Code for Safety in Construction of Railway Bridges and Culverts" (TB 10303-2020), "Quality Acceptance Standards for Construction of High-speed Railway Bridges and Culverts" (TB10752-2018), and "Technical Standards for Bridge Cantilever Casting Construction" (CJJ / T281-2018), the specifications stipulate that the deviation between the measured value and the theoretical value of the vertical line shape of each beam segment shall not exceed ±20mm; the stress during the construction stage shall not exceed ±10% of the design calculation result. With reference to engineering risk management literature and specifications, combined with expert discussion and review, the thresholds for converting the two monitoring data of the vertical line shape deviation of the main beam and the axial stress deviation of the beam body into the risk probability level are finally determined, as shown in Table 6:

[0124] Table 6 Monitoring data - risk probability level conversion threshold

[0125]

[0126] According to the threshold, the difference of monitoring data is quantified, and the dynamic Bayesian network evidence module is introduced to complete the upgrade and optimization of the initial Bayesian network to the dynamic Bayesian network;

[0127] Run the dynamic assessment model of the target bridge's cast-in-place construction risk to obtain the target bridge's construction phase risk change trend under both the introduction of evidence and the non-introduction of evidence, for further analysis;

[0128] Step S6, constructing an integrated risk loss and risk value result analysis visualization model: Based on fuzzy rules, the safety risk loss assessment of the cantilever cast-in-place construction of deep valley large-span bridges is completed, the assessment results of the risk dynamic probability P are integrated, and the risk value results are analyzed and visualized, so that the safety managers of the cantilever cast-in-place construction of deep valley large-span bridges can make accurate decisions;

[0129] According to the 1-9 scale scoring method, the importance of risk factors in the risk assessment index system is scaled and a judgment matrix is ​​constructed. The 1-9 scale scoring method is shown in Table 7:

[0130] Table 7: Scoring method on a scale of 1 to 9

[0131] Scale Explanation of the importance of two factors 1 Factor a is as important as factor b. 3 Factor a is slightly more important than factor b. 5 Factor a is significantly more important than factor b. 7 Factor a is much more important than factor b. 9 Factor a is extremely important compared to factor b. 2、4、6、8 The importance of the two factors is between the above scale values. reciprocal The opposite of the above importance: j is more important than i

[0132] According to the scaling standard, the “environmental risk” judgment matrix is ​​obtained, as shown in Table 8.

[0133] Table 8 Judgment Matrix for “Environmental Risks”

[0134] Environmental risks Risk Source 1 Risk Source 2 Risk Source 3 Risk Source 4 Risk Source 5 Risk Source 1 1 2 1 / 3 3 5 Risk Source 2 1 / 2 1 1 / 3 5 6 Risk Source 3 3 3 1 5 7 Risk Source 4 1 / 3 1 / 5 1 / 5 1 2 Risk Source 5 1 / 5 1 / 6 1 / 7 1 / 2 1

[0135] SPSS data analysis software is used to calculate the judgment matrix of the feedback, requiring the consistency test to pass, that is, CR<1, and finally the weight vector of each indicator level is obtained as follows:

[0136] Equipment risk: W A = [0.076, 0.344, 0.063, 0.259, 0.144, 0.039, 0.028, 0.031, 0.017];

[0137] Material risk: W B = [0.381, 0.410, 0.147, 0.061];

[0138] Construction method risk: W C = [0.215, 0.192, 0.084, 0.048, 0.044, 0.169, 0.072, 0.127, 0.027, 0.023];

[0139] Environmental risks: W D=[0.229, 0.200, 0.457, 0.070, 0.043];

[0140] Behavioral risk: W E = [0.086, 0.054, 0.253, 0.108, 0.319, 0.179];

[0141] Protection risk: W F = [0.129, 0.514, 0.075, 0.282];

[0142] Monitoring risks: W G =[0.653, 0.096, 0.251].

[0143] The criterion layer weight vector is:

[0144] W=[0.163, 0.051, 0.070, 0.165, 0.312, 0.210, 0.029].

[0145] The membership vector Q is obtained by sorting out the survey results on risk losses in the safety risk survey scoring table of cantilever cast-in-place construction of deep valley large-span bridges, clarifying the fuzzy relationship between each risk factor and the risk loss level in the safety risk evaluation index system. According to statistics, the membership vector of "environmental risk" is:

[0146]

[0147] The risk assessment results of the target bridge are collected as basic data, including the operation results of the dynamic Bayesian network model with / without evidence, that is, the dynamic probability P of safety risk, for subsequent application.

[0148] Construct an integrated risk loss and risk value result analysis visualization model, such as Figure 5 As shown in the figure, the main functions of the visualization model are: first, according to the definition of risk loss C (C = Q × W), the calculation and visualization of risk loss are completed; secondly, according to the definition of risk value R (R = C × P), the combined analysis of the dynamic probability P of safety risk and risk loss C is completed, and the risk value R is visualized and the dangerous points of the risk value change trend are marked; the marked content mainly highlights the rising section of the risk value and the extreme point of the risk value. The marked content indirectly reflects the dangerous section in the construction process, which can provide management personnel with more intuitive construction site safety status information and play a key role in hazard warning.

[0149] like Figure 6As shown in the figure, the specific results of the visualization model operation are as follows: first, the operation results of the risk loss C are displayed. Decision makers can view the risk loss status and level of the target bridge during the construction period, and provide a basis for decision-making based on the degree of loss; secondly, the results of the dynamic probability of safety risk P are displayed. This content is the operation result of the Bayesian network evaluation model, which is displayed together with the risk loss C to facilitate decision makers to manage safety risks; further, the risk value R is visualized, which shows the changing trend of the risk value during the construction stage in the form of a point-line graph, and marks the dangerous state of the risk value, providing the most intuitive on-site safety status for decision makers to guide decision-making.

[0150] The operation result is the change trend of the target bridge risk value. The change of risk value can reflect the dynamic risk status of the construction site, which is consistent with the on-site early warning situation and can provide certain help to the construction site manager in making decisions. The above application examples are used to explain the present invention, not to limit the present invention. Any modifications and changes made to the present invention within the spirit and protection scope of the claims of the present invention shall fall within the protection scope of the present invention.

Claims

1. A dynamic assessment method for safety risk of suspended pouring construction of deep valley large-span bridges, characterized in that: Use the following steps: S1. Construct a subjective and objective fusion analysis data module, establish a target bridge safety risk assessment index system, collect the frequency and degree of disaster-causing risk factors that have occurred on the target bridge, conduct special expert surveys, collect subjective and objective data on risk factors, and construct a subjective and objective fusion analysis database for the suspended pouring construction stage; S2. Using the Bayesian network as the basic assessment model framework, the Bayesian network model, network structure and network parameters are initially established, and the adaptability is optimized when the target bridge risk assessment is performed; S3, constructing the adaptability module of the Bayesian network model; S4. Construct a dynamic assessment guidance-time module and establish a risk dynamic assessment model without evidence guidance; S5. Construct a dynamic assessment guidance-evidence module and establish a risk dynamic assessment model guided by evidence; S6. Construct a visualization model for integrated risk loss and risk value result analysis.

2. According to claim 1, a dynamic assessment method for safety risk of suspended casting construction of deep valley large-span bridges is characterized in that: S1 specifically adopts the following steps: S11. Obtain the construction design plan, construction organization design plan, safety management manual and hydrogeological survey data of the target bridge, and preliminarily complete the identification of risk sources of the target bridge; S12. Count the frequency of past risk sources and the degree of risk loss of the target bridge, and establish a database A1 of safety factors for cast-in-place construction of deep-ravine large-span bridges; S13. Establish an expert evaluation scoring table for identifying risk sources, collect expert scoring data, and build an expert scoring database A2 for safety factors of suspended cast-in-place construction of deep valley large-span bridges; S14, perform data fusion of database A1 and A2, using weighted fusion method, fusion formula: In the formula, i takes values ​​of 1, 2, ω i represents the weight of the i-th database, and the weights are ω1, ω2 respectively. The weights are obtained based on the evaluation of experts in the bridge field. The data samples of each data source are x1, x2, x3, …, x n ; S15, databases A1 and A2 together constitute a statistical database A for quantifying the risk sources of the target bridge, and preprocessing is performed on database A to fill in missing data values ​​and remove abnormal data values; S16. Establish a safety risk assessment index system, specifically: construct a three-level index system consisting of a target layer, a criterion layer, and an indicator layer. The target layer corresponds to whether the safety risk of the target bridge occurs, the criterion layer corresponds to whether various risk sources occur, and the indicator layer corresponds to whether a single risk source occurs.

3. According to claim 1, a dynamic assessment method for safety risk of suspended casting construction of deep valley large-span bridges is characterized in that: S2 specifically adopts the following steps: S21. Based on the safety risk assessment index system and the causal relationship between factors, safety risk factors are selected as nodes to establish the network structure topology of the target bridge risk assessment model; S22. Based on the statistical database A, the probability of occurrence of the risk source is calculated using the following formula: P 风险源x =n 风险源x / N, In the formula, P 风险源x represents the calculated probability of occurrence of risk source x, n 风险源x represents the number of risk sources x in database A, and N represents the statistical number of all risk sources in database A; the probability parameters of all nodes in the network structure topology of the risk assessment model are calculated according to the formula to complete the construction of the node probability part of the network parameters of the target bridge assessment model; S23. Construct an initial basic evaluation model based on the Bayesian network visualization UI platform. Referring to the network structure topology, add corresponding nodes in the Bayesian network visualization UI platform, connect the nodes using arrows representing causal relationships, and complete the initialization of the network structure. Continue to define the nodes, introduce the calculated node probabilities, and initialize the network parameters.

4. According to claim 3, a dynamic assessment method for safety risk of suspended casting construction of deep valley large-span bridges is characterized in that: In S3, the network parameter conditional probability part of the evaluation model is represented by 2 n The term is transformed into 2n, and Leaky nodes are added to the basic assessment model to introduce leakage probability, so as to introduce the influence of secondary risk factors on the cast-in-place construction of deep valley large-span bridges into the basic assessment model; the specific steps are as follows: S31. Classify the risk sources in the database according to primary and secondary categories, set the risk sources that are nodes in the network structure as primary risk sources, and classify the remaining risk sources as secondary risk sources according to the risk source categories; S32. Set the secondary risk source as a leaky node and assign a leakage probability. The leaky node and the node corresponding to the primary risk source belong to the same level. The impact of the secondary risk source on the construction safety of the target bridge is introduced into the initial network structure to improve the adaptability of the assessment model to the assessment object and enhance the accuracy of the assessment results. S33. Use the Leaky Noisy-or theory to construct a conditional probability table for a complex evaluation model. The calculation formula for the Leaky Noisy-or theory is as follows: P xi (P|X1,X2,…,X i ,…,X n ) P L (P|X1,X2,…,X n ,…,X L ) In the formula, P a (Y) represents the set of parent nodes of the Y node, Pxi represents the joint probability when one parent node of the Y node appears and the other parent nodes do not appear, P L Represents the joint probability of leaky nodes.

5. A dynamic assessment method for safety risk of suspended casting construction of deep valley large-span bridges according to claim 4, characterized in that: In S4, on top of the initial Bayesian network model, a temporal node, i.e., a transfer node, is created, time series and transfer probability are introduced, and the initial network model on the adjacent time series is associated. Specifically, the following steps are adopted: S41. Select the top node of the Bayesian network structure, the safety risk of the suspended pouring construction of the deep valley large-span bridge, as the transfer node, assign it a time series, and introduce a transfer probability for it. The transfer node and the transfer probability together constitute the time series module of the Bayesian network evaluation model, thereby constructing the time series relationship of the initial Bayesian network of two adjacent time segments; S42, the number of construction days is a concept of time series time, that is, each construction day corresponds to an initial assessment model network model; S43. On the basis of the time series, the correlation between the two time segments evaluation models is added, that is, the transition probability is added to the time series. After the transition probability is added, the construction of the time module of the initial evaluation model is realized. The transition probability is calculated using the following formula: In the formula, represents the value of the i-th state at time t, for The parent node of S44. Run the target bridge suspended cast construction safety risk dynamic assessment model that introduces the time module to conduct a dynamic assessment of the safety risk of suspended cast construction of deep valley large-span bridges without evidence guidance.

6. A dynamic assessment method for safety risk of suspended casting construction of deep valley large-span bridges according to claim 4, characterized in that: In S5, several new nodes are created, relationships are established with the transfer nodes, and attributes of the new nodes are defined. Specifically, the following steps are adopted: S51. Connect several new nodes to the transfer node of the time module to define evidence nodes. The causal relationship arrow points to the transfer node, indicating that the node needs to be set as a risk influencing factor when it is defined. The arrow points to the newly added node, indicating that the node needs to be set as a risk representation factor when it is defined. Risk influencing factors include wind force during construction, rainfall during construction and physical signs of construction workers; risk characterization factors include vertical alignment of construction beams, stress of construction beams and stress of main components of construction hanging baskets; risk characterization factors mainly play an evidence-guiding role, and the role of risk influencing factors is reflected in improving the risk assessment model and enhancing the adaptability of the model; S52. The monitoring data-risk probability level conversion method is used to convert the collected monitoring data into the parameters required by the target bridge safety risk dynamic assessment model after preprocessing. The monitoring data-risk probability level conversion method specifically adopts the following steps: S521. Collect previous monitoring data from the construction site, use the monitoring data-risk probability level conversion method to complete the conversion, and use the results for model training and optimization; S522. Write a special construction monitoring plan, conduct on-site monitoring of the required monitoring data, and collect, summarize and pre-process the monitoring results; S53. Use the monitoring data-risk probability level conversion method again to convert the field monitoring results, introduce the conversion results into the corresponding nodes, realize the model evidence import, run the target bridge risk dynamic assessment model built by the evidence module, and realize the dynamic assessment of the safety risk of suspended cast-in-place construction of deep valley large-span bridges under the guidance of evidence.

7. The method for dynamic assessment of safety risks of suspended casting construction of deep valley large-span bridges according to claim 1 is characterized in that: The S6 specifically adopts the following steps: S61. Collect the risk assessment results of the target bridge as basic data, including the operation results of the dynamic Bayesian network model without or with evidence, that is, the dynamic probability of safety risk P, and collect and count the analysis results data; S62. Based on the statistical database A, the fuzzy rules are used to calculate the loss membership Q and the risk factor weight W in the risk loss assessment; According to the statistical data of risk loss degree in database A, the membership vector Q of each risk factor is obtained by weighted average; the weight judgment matrix between risk factors is obtained by using the 1-9 scoring method, and the risk factor weight W is calculated using the SPSS data analysis tool; S63. Based on Python language, construct an integrated risk loss and risk value result analysis visualization model; wherein: based on the definition of risk loss C, C = Q × W, first complete the risk loss calculation and visualization; further based on the definition of risk value R, R = C × P, complete the calculation and visualization of risk value R.

Images