CN112070399B - Safety risk assessment method and system for large-scale engineering structure - Google Patents
Safety risk assessment method and system for large-scale engineering structure Download PDFInfo
- Publication number
- CN112070399B CN112070399B CN202010938563.0A CN202010938563A CN112070399B CN 112070399 B CN112070399 B CN 112070399B CN 202010938563 A CN202010938563 A CN 202010938563A CN 112070399 B CN112070399 B CN 112070399B
- Authority
- CN
- China
- Prior art keywords
- brb
- optimized
- processing block
- rule base
- confidence rule
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 46
- 238000012502 risk assessment Methods 0.000 title claims abstract description 39
- 238000012545 processing Methods 0.000 claims abstract description 77
- 238000000605 extraction Methods 0.000 claims abstract description 65
- 238000012544 monitoring process Methods 0.000 claims abstract description 29
- 230000003044 adaptive effect Effects 0.000 claims description 12
- 230000006870 function Effects 0.000 claims description 11
- 238000005457 optimization Methods 0.000 claims description 11
- 230000008569 process Effects 0.000 claims description 11
- 238000010276 construction Methods 0.000 claims description 10
- 239000011159 matrix material Substances 0.000 claims description 10
- 239000010410 layer Substances 0.000 description 35
- 238000010586 diagram Methods 0.000 description 11
- 238000011156 evaluation Methods 0.000 description 5
- 238000012423 maintenance Methods 0.000 description 4
- 230000000875 corresponding effect Effects 0.000 description 3
- 239000003949 liquefied natural gas Substances 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 238000004062 sedimentation Methods 0.000 description 3
- 238000001994 activation Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- 229910000831 Steel Inorganic materials 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000013210 evaluation model Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 239000002356 single layer Substances 0.000 description 1
- 239000010959 steel Substances 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0635—Risk analysis of enterprise or organisation activities
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Economics (AREA)
- Entrepreneurship & Innovation (AREA)
- General Physics & Mathematics (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- Artificial Intelligence (AREA)
- Operations Research (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Game Theory and Decision Science (AREA)
- Educational Administration (AREA)
- Quality & Reliability (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Development Economics (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- General Engineering & Computer Science (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to a security risk assessment method and system for a large-scale engineering structure. The method comprises the following steps: constructing an extraction block according to the correlation between the historical monitoring index and the historical risk index; combining expert knowledge to construct a processing block; optimizing parameters of the extraction block by adopting a layer-by-layer learning strategy to obtain an optimized extraction block; optimizing the structure and parameters of the processing block by adopting a layer self-adaptive growth strategy to obtain an optimized processing block, wherein the optimized extraction block and the optimized processing block form an optimized depth confidence rule base together; and taking the monitoring index and the risk index as inputs, and inputting the inputs into the optimized deep confidence rule base to obtain a safety risk assessment result. The method and the system can deeply mine rule information and realize accurate assessment of safety risks.
Description
Technical Field
The invention relates to the technical field of risk assessment of large-scale engineering structures, in particular to a safety risk assessment method and a system for a large-scale engineering structure.
Background
In recent years, the economic development of China is rapid, and the construction and development of large-scale engineering structures represented by large bridges, extra-high voltage power towers, large storage tanks and the like are effectively promoted. Large engineering structures occupy an important position in the infrastructure of China. In order to prevent potential safety hazards and ensure safe and healthy operation, more and more researches treat security risk assessment as core content of maintenance and management of large-scale engineering structures. The safety risk assessment is beneficial to early finding potential safety hazards, and maintenance measures are convenient to take in time. At present, aiming at the problem of risk assessment of large-scale engineering structures, students at home and abroad develop extensive researches and obtain certain achievements. For example, university of homography Pan Liming et al first propose an index system for cable-stayed bridge safety, and based on analytic hierarchy process and fuzzy comprehensive judgment principle, the bridge safety risk is evaluated. The risk of the storage tank is analyzed by the university of North He Guo Bing and the like from the aspects of failure possibility and failure result, and a quantitative risk analysis model of the storage tank is established, so that the security risk level of the storage tank is determined.
There are several problems with existing research: firstly, the large-scale engineering structure has complex composition and huge scale, and the key parts to be monitored are numerous; secondly, the working environment of the large-scale engineering structure is bad, the running state of the large-scale engineering structure is interfered by various internal and external factors, the large-scale engineering structure has stronger uncertainty, and an expert can hardly judge the state of the large-scale engineering structure only by experience; third, large-scale engineering structure anomaly data is limited, and it is difficult to obtain high-quality data samples. In view of this, the above-mentioned problems must be fully considered, and limited data samples and expert knowledge must be fully combined to realize high-precision evaluation of large-scale engineering structure risk, so as to provide support for subsequent maintenance decisions.
Disclosure of Invention
The invention aims to provide a safety risk assessment method and a system for a large engineering structure, which can deeply mine rule information and realize accurate assessment of safety risk.
In order to achieve the above object, the present invention provides the following solutions:
a security risk assessment method for a large-scale engineering structure comprises the following steps:
constructing an extraction block according to the correlation between the historical monitoring index and the historical risk index;
combining expert knowledge to construct a processing block;
optimizing parameters of the extraction block by adopting a layer-by-layer learning strategy to obtain an optimized extraction block;
optimizing the structure and parameters of the processing block by adopting a layer self-adaptive growth strategy to obtain an optimized processing block, wherein the optimized extraction block and the optimized processing block form an optimized depth confidence rule base together;
and taking the monitoring index and the risk index as inputs, and inputting the inputs into the optimized deep confidence rule base to obtain a safety risk assessment result.
Optionally, the constructing and extracting block according to the correlation between the historical monitoring index and the historical risk index specifically includes:
calculating mutual information between the historical monitoring index and the historical risk index;
determining relative mutual information according to the mutual information;
dividing all historical monitoring indexes into even-numbered combinations and odd-numbered combinations according to the relative mutual information;
inputting the even combination or the odd combination into a hierarchical confidence rule base BRB to obtain an output u of the hierarchical confidence rule base BRB of the previous level H-1 And u H ；
Output u of the higher-level confidence rule base BRB H-1 And u H Input to the confidence rule base BRB-0 to obtain the output y of the confidence rule base BRB-0 0 The method comprises the steps of carrying out a first treatment on the surface of the The output u of the higher-level confidence rule base BRB H-1 And u H And the output y of the confidence rule base BRB-0 0 Together as the output of the extraction block, also the input of the processing block.
Optionally, the combined expert knowledge building processing block specifically includes:
output y of the confidence rule base BRB-0 0 Input as a first input to the confidence rule base BRB-1, output y of BRB-1 1 Input BRB-2, and so on to BRB-M, while outputting the output u of the hierarchical confidence rule base BRB in the extraction block H-1 And u H Respectively inputting the data into the confidence rule base BRB-1, BRB-2, BRB-M as a second input, and finally obtaining the processing block output y M 。
Optionally, the optimizing the parameters of the extraction block by adopting a layer-by-layer learning strategy to obtain an optimized extraction block specifically includes:
constructing an objective function of the extraction block;
and recursively optimizing parameters of each sub BRB in the hierarchical confidence rule base BRB by using a projection covariance matrix self-adaptive evolution strategy until all sub BRB models are optimized, and determining an optimized extraction block.
Optionally, the optimizing the structure and parameters of the processing block by adopting a layer adaptive growth strategy to obtain an optimized processing block specifically includes:
constructing an objective function of the processing block;
optimizing a confidence rule base BRB-1 in the initial processing block by using a projection covariance matrix self-adaptive evolution strategy;
generating a confidence rule base BRB-2 in the processing block by taking the optimized BRB-1 parameter as an initial parameter, wherein the output u of the BRB of the confidence rule base of the upper level H-1 、u H And the optimized BRB-1 output is used as the input of the processing block confidence rule base BRB-2;
optimizing the processing block confidence rule base BRB-2 by utilizing a projection covariance matrix self-adaptive evolution strategy until the optimization of the processing block is stopped when the end condition is met; the end condition is g q ≥G max or q≥Q max or E q ＜e；
Wherein g q Is the iteration number of the BRB-q optimization process, G max Is the maximum iteration number, Q max Is the maximum number of layers and e is the expected modeling error.
A large-scale engineering structure security risk assessment system, comprising:
the extraction block construction module is used for constructing an extraction block according to the correlation between the historical monitoring index and the historical risk index;
the processing block construction module is used for constructing a processing block by combining expert knowledge;
the extraction block optimization module is used for optimizing parameters of the extraction blocks by adopting a layer-by-layer learning strategy to obtain optimized extraction blocks;
the processing block optimizing module is used for optimizing the structure and parameters of the processing block by adopting a layer self-adaptive growth strategy to obtain an optimized processing block, and the optimized extraction block and the optimized processing block jointly form an optimized depth confidence rule base;
and the security risk assessment result determining module is used for taking the monitoring index and the risk index as inputs, and inputting the inputs into the optimized depth confidence rule base to obtain a security risk assessment result.
According to the specific embodiment provided by the invention, the invention discloses the following technical effects:
the invention provides a security risk assessment method and system for a large-scale engineering structure. The method comprises the following steps: constructing an extraction block according to the correlation between the historical monitoring index and the historical risk index; combining expert knowledge to construct a processing block; optimizing parameters of the extraction block by adopting a layer-by-layer learning strategy to obtain an optimized extraction block; optimizing the structure and parameters of the processing block by adopting a layer self-adaptive growth strategy to obtain an optimized processing block, wherein the optimized extraction block and the optimized processing block form an optimized depth confidence rule base together; and taking the monitoring index and the risk index as inputs, and inputting the inputs into the optimized deep confidence rule base to obtain a safety risk assessment result. The method and the system can deeply mine rule information and realize accurate assessment of safety risks.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings that are needed in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart of a security risk assessment method for a large-scale engineering structure;
FIG. 2 is a block diagram of a security risk assessment system for a large-scale engineering structure according to the present invention;
FIG. 3 is a schematic diagram of a deep belief rule base model structure;
FIG. 4 is a schematic diagram of the structure of an extraction block;
FIG. 5 is a schematic diagram of a processing block;
FIG. 6 is a schematic diagram of a layer-by-layer learning strategy;
FIG. 7 is a schematic diagram of an adaptive growth strategy;
FIG. 8 is a schematic diagram of the structure of an initial assessment model;
FIG. 9 is a diagram showing comparison of evaluation results;
fig. 10 is a schematic diagram of the traceability of the risk status.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
The invention aims to provide a safety risk assessment method and a system for a large engineering structure, which can deeply mine rule information and realize accurate assessment of safety risk.
In order that the above-recited objects, features and advantages of the present invention will become more readily apparent, a more particular description of the invention will be rendered by reference to the appended drawings and appended detailed description.
FIG. 1 is a flow chart of a security risk assessment method for a large-scale engineering structure. As shown in fig. 1, a security risk assessment method for a large engineering structure includes:
step 101: the construction and extraction block is constructed according to the correlation between the historical monitoring index and the historical risk index, and specifically comprises the following steps:
calculating mutual information between the historical monitoring index and the historical risk index;
determining relative mutual information according to the mutual information;
dividing all historical monitoring indexes into even-numbered combinations and odd-numbered combinations according to the relative mutual information;
inputting the even combination or the odd combination into a hierarchical confidence rule base BRB to obtain an output u of the hierarchical confidence rule base BRB of the previous level H-1 And u H ；
Output u of the higher-level confidence rule base BRB H-1 And u H Input to the confidence rule base BRB-0 to obtain the output y of the confidence rule base BRB-0 0 The method comprises the steps of carrying out a first treatment on the surface of the The output u of the higher-level confidence rule base BRB H-1 And u H And the output y of the confidence rule base BRB-0 0 Together as the output of the extraction block, also the input of the processing block.
Step 102: the method combines expert knowledge to construct a processing block, and specifically comprises the following steps:
output y of the confidence rule base BRB-0 0 Input as a first input to the confidence rule base BRB-1, output y of BRB-1 1 Input BRB-2, and so on to BRB-M, while outputting the output u of the hierarchical confidence rule base BRB in the extraction block H-1 And u H Respectively inputting the data into the confidence rule base BRB-1, BRB-2, BRB-M as a second input, and finally obtaining the processing block output y M 。
Step 103: optimizing parameters of the extraction block by adopting a layer-by-layer learning strategy to obtain an optimized extraction block, wherein the method specifically comprises the following steps:
constructing an objective function of the extraction block;
and recursively optimizing parameters of each sub BRB in the hierarchical confidence rule base BRB by using a projection covariance matrix self-adaptive evolution strategy until all sub BRB models are optimized, and determining an optimized extraction block.
Step 104: and optimizing the structure and parameters of the processing block by adopting a layer self-adaptive growth strategy to obtain an optimized processing block, wherein the optimized extraction block and the optimized processing block jointly form an optimized depth confidence rule base. FIG. 3 is a schematic diagram of a deep belief rule base model structure.
Step 104 specifically includes:
constructing an objective function of the processing block;
optimizing a confidence rule base BRB-1 in the initial processing block by using a projection covariance matrix self-adaptive evolution strategy;
generating a confidence rule base BRB-2 in the processing block by taking the optimized BRB-1 parameter as an initial parameter, wherein the confidence rule of the upper level is providedOutput u of library BRB H-1 、u H And the optimized BRB-1 output is used as the input of the processing block confidence rule base BRB-2;
optimizing the processing block confidence rule base BRB-2 by utilizing a projection covariance matrix self-adaptive evolution strategy until the optimization of the processing block is stopped when the end condition is met; the end condition is g q ≥G max or q≥Q max or E q ＜e；
Wherein g q Is the iteration number of the BRB-q optimization process, G max Is the maximum iteration number, Q max Is the maximum number of layers and e is the expected modeling error.
Step 105: and taking the monitoring index and the risk index as inputs, and inputting the inputs into the optimized deep confidence rule base to obtain a safety risk assessment result.
Regarding step 101, the main steps of the method for constructing the extraction block are as follows:
step 1: because of the lack of system integrity mechanism information, mutual information is needed to quantify the correlation between variables, enabling the grouping of features. The mutual information between the feature X and the security risk factor Y can be calculated as:
I(X,Y)＝∑ i,j p(x i ,y j )log(p(x i ,y j )/(p(x i )p(y j ))) (1)
where p (-) is the probability distribution of the variables. I (X, Y) represents mutual information between X and Y. Thus, the Relative Mutual Information Ratio (RMIR) can be defined as:
φ Y (X)＝I(X,Y)/(I(Y,Y)),φ Y (X)∈[0,1] (2)
wherein phi is Y (X) represents RMIR between X and Y. If completely correlated, phi Y (X) is equal to 1. If phi Y (X) is 0, there is no correlation between them.
Step 2: RMIR between the feature and risk factors is calculated using the equation above. It is assumed that it satisfies the following expression:
wherein N is X Is the number of features. Y represents a risk factor.
Step 3: all features are divided into two groups according to the RMIR between the features and the risk factors. Both even and odd cases are shown below.
Step 4: let the number of actual features be 4h. The structure of the extraction block is shown in fig. 4 and consists of two parts. Part 1 is the hierarchy BRB and part 2 is BRB-0. In section 1, two adjacent features within each group are input into the sub-BRBs, and the k-th rule of the h-th sub-BRB is described as follows:
wherein the method comprises the steps ofIs a reference value. h θ k Is a rule weight. h δ i (i=1, 2) represents an attribute weight. u (u) h The output of the h-th sub-BRB, which is taken as the sub-BRB input of the next stage together with the adjacent BRB output. As shown in FIG. 2,u H-1 And u H The output of section 1, indicated, is used as input to BRB-0 in section 2.
In section 2, the kth rule for BRB-0 is described as follows:
wherein d= { D 1 ,...,D N And the reference value of the input and result. u (u) H-1 ，u H And output y of BRB-0 0 Will be considered the output of the whole extraction block, also the input of the processing block.
Regarding step 102, the method for constructing the processing block mainly includes the following steps:
as shown in fig. 5, the adaptive BRB in the processing block is constructed by processing the information in the extraction block layer by layer. y is 0 ，u H-1 And u H To extract the input of the first layer of the block, this layer is typically built using expert knowledge, while the growth layers (layer 2 to M) are generated using a layer adaptive growth strategy. The growth layer has the same form as the first layer, taking BRB-m as an example, and the kth rule is as follows:
according to the DLBRB-based large-scale engineering structure security risk assessment method, a evidence reasoning method (ER) is adopted as a reasoning engine by the DLBRB model, and results are generated layer by layer. Taking the reasoning step of the h sub-BRB as an example:
step 1: will input x i The conversion is to a confidence distribution as follows:
h S(x i )＝{( h A i,j , h a i,j ),i＝1,…, h M；j＝1,…, h J i } (8)
wherein x is i Representing the input of each BRB model, e.g., x for the h sub-BRB i X represents 2h-1 And X is 2h Specific values of (2). h A i,j The j-th reference value representing the i-th input. h a i,j Representation of h A i,j Is a matching degree of (a).
Step 2: calculating rule activation weights:
wherein, h θ k ∈[0,1]the weight of the kth rule is represented.
Step 3: converting the confidence in the rule to a basic probability quality (BPA):
wherein, h m j,k representation and D j Related BPA. m is m D,k Representation pair d= { D 1 ,...,D N Support degree of }.
Step 4: the aggregate confidence is calculated by iterating through a combination of the preceding rules:
h m j,I(k+1) ＝K I(k+1) ( h m j,I(k) h m j,k+1 + h m j,I(k) h m D,k+1 + h m D,I(k) h m j,k+1 ) (11-a)
wherein,indicating the remaining confidence.
Step 5: the final distributed result is expressed as:
y h ＝ h S( h A * )＝{(D j , h β j )；j＝1,..., h N} (12)
wherein, h A * representing the actual input vector.
Regarding step 103, the layer-by-layer learning strategy of the extraction layer is shown in fig. 6, and the main steps are as follows:
step 1: an objective function is constructed. Taking the h sub-BRB as an example, the objective function is as follows:
wherein, psi is% h θ, h δ, h β, h A) Representing the error between the output and the actual value. The ID is the interpretability distribution of the confidence in the kth rule.And->Representing the lower and upper bounds, respectively, of the t-th reference value in the k-th rule. V represents the threshold for human understanding of the activation process, which is subjective in engineering practice, and is related to the decision maker and the expected accuracy of the result.
Step 2: the parameters of sub-BRB 1 are optimized using a projection covariance matrix adaptive evolution strategy (P-CMA-ES).
Step 3: the parameters of sub-BRB 2 are optimized in a similar way as in step 2. This process is recursively performed until all sub-BRB models (including BRB-0) are optimized.
Regarding step 104, the adaptive growth strategy of the process layer is shown in fig. 7, which mainly includes the steps of:
step 1: an objective function is constructed. Taking BRB-m as an example, the objective function is expressed as follows:
min{ψ(θ m ,δ m ,β m ,A m ,q)} (14)
where q is the number of layers of the processing block. The constraint is similar to equation (13).
Step 2: the initial BRB-1 is optimized using P-CMA-ES. The error between the BRB-1 output and the actual value is denoted as E 1 The calculation formula is
E 1 ＝ψ(θ 1 ,δ 1 ,β 1 ,A 1 ,q) (15)
Step 3: and generating BRB-2 by taking the optimized BRB-1 parameter as an initial parameter. u (u) H-1 ，u H And the optimized BRB-1 output is used as the input of BRB-2. BRB-2 is optimized using P-CMA-ES. The stop condition of the optimization process is E 2 ＜ηE 1 Wherein E is 2 Representing the error between the BRB-2 output and the actual value. η is the learning rate. After BRB-2 is optimized, the next BRB model should be generated and optimized in the same manner.
Step 4: the number of layers in the processing block will increase until either of the following conditions is met.
g q ≥G max or q≥Q max or E q ＜e (16)
Wherein g q Is the number of iterations of the BRB-q optimization process. G max Is the maximum number of iterations. Q (Q) max Is the maximum number of layers. e is the expected modeling error.
Corresponding to the security risk assessment method of the large engineering structure, the invention also provides a security risk assessment system of the large engineering structure, and fig. 2 is a structural diagram of the security risk assessment system of the large engineering structure, as shown in fig. 2, the system comprises:
an extraction block construction module 201, configured to construct an extraction block according to a correlation between a history monitoring index and a history risk index;
a processing block construction module 202 for constructing a processing block in combination with expert knowledge;
the extraction block optimization module 203 is configured to optimize parameters of the extraction block by adopting a layer-by-layer learning strategy, so as to obtain an optimized extraction block;
the processing block optimizing module 204 is configured to optimize the structure and parameters of the processing block by using a layer adaptive growth strategy to obtain an optimized processing block, where the optimized extraction block and the optimized processing block together form an optimized deep confidence rule base;
the security risk assessment result determining module 205 is configured to input the monitoring index and the risk index as input to the optimized deep confidence rule base, and obtain a security risk assessment result.
Example 1:
in this embodiment, the safety risk assessment is performed on a large lng tank located in hainan province, china. 12 monitoring points are arranged on the support column to monitor sedimentation. However, disturbance factors such as tidal-induced foundation pressure changes may introduce more uncertainty into the safe risk assessment of the storage tanks. Methods such as artificial neural networks and support vector machines cannot describe these uncertainties. The single-layer BRB can handle uncertainty well, but data from 12 monitoring points can lead to rule explosion. Thus, DLBRB can be a good choice for accurately estimating the safe status of LNG storage tanks. In the experiment, the obtained monitoring data are manually measured from 2012 to 2019, and the corresponding safety state can be determined by an expert group according to the construction deformation measurement Specification (JGJ 8-2007) and the petrochemical steel storage tank foundation construction Specification (SH 3528-2005). The obtained tank sedimentation mechanism information is assumed to be limited.
The specific implementation steps of the safety risk assessment of the large-scale liquefied natural gas storage tank are as follows:
step 1: an initial assessment model is constructed.
From the RMIR between the feature and the security risk, the two feature groups are represented as { E } 12 ,E 6 ,E 2 ,E 1 ,E 7 ,E 4 Sum { E } 8 ,E 3 ,E 10 ,E 11 ,E 5 ,E 9 And determines the structure of the initial DLBRB as shown in fig. 8. The extraction block consists of 10 sub-BRB models and one BRB-0 model. As a local judgment of the safety risk of the storage tank. Extracting blocksThe output is used as the input of a processing block, an initial BRB-1 model is established by using expert knowledge, and the rest models are generated by adopting a layer self-adaptive growth strategy.
According to expert knowledge, the semantic values describing sedimentation are "high" and "low" (denoted "H" and "L"), the reference values of which are given in table 1. According to JGJ8-2007 and SH3528-2005, the safety state of a tank can be evaluated with risk factors from 1 to 10. When the risk factor is greater than 10, further maintenance should be performed. The security risk states are described by "security", "slight risk" and "high risk". Thus, the reference values are shown in Table 2. The initial attribute weight and rule weight are equal to 1. The initial extraction block and the initial BRB-1 are given in tables 3 and 4, and after the initial DLBRB determination, the hierarchical learning strategy and the layer adaptive growth strategy are adopted to perform adaptive learning, so as to obtain a final model.
Table 1 monitoring point reference value
TABLE 2 reference value for tank safety risk
TABLE 3 initial extraction Block
TABLE 4 initial BRB-1
Step 2: and (5) evaluating the adaptive learning of the model.
Assuming that the reference values in tables 1 and 2 are accurate, the tolerance threshold v is equal to 0.003. The objective function is constructed as follows:
from the analysis, the ID can be expressed as:
the stopping condition of the layer self-adaptive growth strategy is set to be that G is more than or equal to G max . The hierarchical adaptive growth strategy and the initial parameters of P-CMA-ES-I are given: expert knowledge confidence kappa expert A population size lambda of 23, a number of generation of extraction blocks G of 50, and a maximum number of iterations of processing blocks G of 0.85 max 400. The learning rate η is 1. In the self-adaptive learning process, 80 groups of data are randomly selected as training data, and all 105 groups of data are used as test data.
Step 3: security risk assessment based on DLBRB model.
As shown in fig. 9, the evaluation results of the initial DLBRB and the training DLBRB are compared. The result shows that the evaluation error of the initial DLBRB is 0.5297, the safety risk state of the storage tank can only be roughly obtained, the trained DLBRB has good performance in evaluating the safety risk of the storage tank, the evaluation error is only 0.0607, and the accuracy is improved by 88.54%. In order to further combine the DLBRB evaluation model to trace the safety risk of the storage tank, the hidden danger is eliminated. The first five data points with corresponding risk factors equal to (9, 8.8, 9, 8.9), respectively, were chosen, and the local output of the tank safety risk is shown in fig. 10, and it can be seen that the confidence level of "S" and "M" decreases and the confidence level of "H" increases, which indicates that the local output gets closer to the actual output (global output) during the reasoning process. In engineering practice, the clear local output change process is helpful for judging and positioning the reason of the unreliable result, so that the related parameters of the model can be manually adjusted or monitoring points of abnormal change can be further found out.
In the present specification, each embodiment is described in a progressive manner, and each embodiment is mainly described in a different point from other embodiments, and identical and similar parts between the embodiments are all enough to refer to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant points refer to the description of the method section.
The principles and embodiments of the present invention have been described herein with reference to specific examples, the description of which is intended only to assist in understanding the methods of the present invention and the core ideas thereof; also, it is within the scope of the present invention to be modified by those of ordinary skill in the art in light of the present teachings. In view of the foregoing, this description should not be construed as limiting the invention.
Claims (6)
1. The safety risk assessment method for the large-scale engineering structure is characterized by comprising the following steps of:
constructing an extraction block according to the correlation between the historical monitoring index and the historical risk index;
combining expert knowledge to construct a processing block;
optimizing parameters of the extraction block by adopting a layer-by-layer learning strategy to obtain an optimized extraction block;
optimizing the structure and parameters of the processing block by adopting a layer self-adaptive growth strategy to obtain an optimized processing block, wherein the optimized extraction block and the optimized processing block form an optimized depth confidence rule base together;
and taking the monitoring index and the risk index as inputs, and inputting the inputs into the optimized deep confidence rule base to obtain a safety risk assessment result.
2. The method for evaluating the safety risk of a large-scale engineering structure according to claim 1, wherein the constructing the extraction block according to the correlation between the historical monitoring index and the historical risk index specifically comprises:
calculating mutual information between the historical monitoring index and the historical risk index;
determining relative mutual information according to the mutual information;
dividing all historical monitoring indexes into even-numbered combinations and odd-numbered combinations according to the relative mutual information;
inputting the even combination or the odd combination into a hierarchical confidence rule base BRB to obtain an output u of the hierarchical confidence rule base BRB of the previous level H-1 And u H ；
Output u of the higher-level confidence rule base BRB H-1 And u H Input to the confidence rule base BRB-0 to obtain the output y of the confidence rule base BRB-0 0 The method comprises the steps of carrying out a first treatment on the surface of the The output u of the higher-level confidence rule base BRB H-1 And u H And the output y of the confidence rule base BRB-0 0 Together as the output of the extraction block, also the input of the processing block.
3. The method for evaluating the security risk of a large engineering structure according to claim 2, wherein the building processing block combining expert knowledge specifically comprises:
output y of the confidence rule base BRB-0 0 Input as a first input to the confidence rule base BRB-1, output y of BRB-1 1 Input BRB-2, and so on to BRB-M, while outputting the output u of the hierarchical confidence rule base BRB in the extraction block H-1 And u H Respectively inputting the data into the confidence rule base BRB-1, BRB-2, BRB-M as a second input, and finally obtaining the processing block output y M 。
4. The method for evaluating the security risk of a large engineering structure according to claim 1, wherein the optimizing the parameters of the extraction block by adopting a layer-by-layer learning strategy to obtain an optimized extraction block specifically comprises:
constructing an objective function of the extraction block;
and recursively optimizing parameters of each sub BRB in the hierarchical confidence rule base BRB by using a projection covariance matrix self-adaptive evolution strategy until all sub BRB models are optimized, and determining an optimized extraction block.
5. The method for evaluating the security risk of a large-scale engineering structure according to claim 1, wherein the optimizing the structure and parameters of the processing block by adopting a layer adaptive growth strategy, to obtain an optimized processing block, specifically comprises:
constructing an objective function of the processing block;
optimizing a confidence rule base BRB-1 in the initial processing block by using a projection covariance matrix self-adaptive evolution strategy;
generating a confidence rule base BRB-2 in the processing block by taking the optimized BRB-1 parameter as an initial parameter, wherein the output u of the BRB of the confidence rule base of the upper level H-1 、u H And the optimized BRB-1 output is used as the input of the processing block confidence rule base BRB-2;
optimizing the processing block confidence rule base BRB-2 by utilizing a projection covariance matrix self-adaptive evolution strategy until the optimization of the processing block is stopped when the end condition is met; the end condition is g q ≥G max or q≥Q max or E q ＜e；
Wherein g q Is the iteration number of the BRB-q optimization process, G max Is the maximum iteration number, Q max Is the maximum number of layers and e is the expected modeling error.
6. A large-scale engineering structure security risk assessment system, comprising:
the extraction block construction module is used for constructing an extraction block according to the correlation between the historical monitoring index and the historical risk index;
the processing block construction module is used for constructing a processing block by combining expert knowledge;
the extraction block optimization module is used for optimizing parameters of the extraction blocks by adopting a layer-by-layer learning strategy to obtain optimized extraction blocks;
the processing block optimizing module is used for optimizing the structure and parameters of the processing block by adopting a layer self-adaptive growth strategy to obtain an optimized processing block, and the optimized extraction block and the optimized processing block jointly form an optimized depth confidence rule base;
and the security risk assessment result determining module is used for taking the monitoring index and the risk index as inputs, and inputting the inputs into the optimized depth confidence rule base to obtain a security risk assessment result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010938563.0A CN112070399B (en) | 2020-09-09 | 2020-09-09 | Safety risk assessment method and system for large-scale engineering structure |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010938563.0A CN112070399B (en) | 2020-09-09 | 2020-09-09 | Safety risk assessment method and system for large-scale engineering structure |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112070399A CN112070399A (en) | 2020-12-11 |
| CN112070399B true CN112070399B (en) | 2024-02-13 |
Family
ID=73662936
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010938563.0A Active CN112070399B (en) | 2020-09-09 | 2020-09-09 | Safety risk assessment method and system for large-scale engineering structure |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112070399B (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2007219769A (en) * | 2006-02-15 | 2007-08-30 | Yamaguchi Univ | Hazard evaluation system |
| CA2795828A1 (en) * | 2010-05-07 | 2011-11-10 | Exxonmobil Research And Engineering Company | Integrated expert system for identifying abnormal events in an industrial plant |
| CN109685283A (en) * | 2019-01-07 | 2019-04-26 | 东华大学 | A kind of mineral hot furnace operating condition prediction technique based on confidence rule base reasoning |
| CN110084518A (en) * | 2019-04-29 | 2019-08-02 | 杭州电子科技大学 | A kind of slope stability evaluation method based on Mixed design confidence rule-based reasoning model |
| CN110348752A (en) * | 2019-07-16 | 2019-10-18 | 中国人民解放军火箭军工程大学 | A kind of large scale industry system structure security assessment method considering environmental disturbances |
| CN110824018A (en) * | 2019-11-12 | 2020-02-21 | 宁波市劳动安全技术服务公司 | Risk assessment and safety detection evaluation method for atmospheric storage tank |
-
2020
- 2020-09-09 CN CN202010938563.0A patent/CN112070399B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2007219769A (en) * | 2006-02-15 | 2007-08-30 | Yamaguchi Univ | Hazard evaluation system |
| CA2795828A1 (en) * | 2010-05-07 | 2011-11-10 | Exxonmobil Research And Engineering Company | Integrated expert system for identifying abnormal events in an industrial plant |
| CN109685283A (en) * | 2019-01-07 | 2019-04-26 | 东华大学 | A kind of mineral hot furnace operating condition prediction technique based on confidence rule base reasoning |
| CN110084518A (en) * | 2019-04-29 | 2019-08-02 | 杭州电子科技大学 | A kind of slope stability evaluation method based on Mixed design confidence rule-based reasoning model |
| CN110348752A (en) * | 2019-07-16 | 2019-10-18 | 中国人民解放军火箭军工程大学 | A kind of large scale industry system structure security assessment method considering environmental disturbances |
| CN110824018A (en) * | 2019-11-12 | 2020-02-21 | 宁波市劳动安全技术服务公司 | Risk assessment and safety detection evaluation method for atmospheric storage tank |
Non-Patent Citations (2)
| Title |
|---|
| Self-tuning of fuzzy belief rule bases for engineering system safety analysis;Liu, Jun;《ANNALS OF OPERATIONS RESEARCH》;全文 * |
| 基于FEMA P-58 的RC 框架结构抗震及减隔震性能评估;杜 轲;《工程力学》;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112070399A (en) | 2020-12-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110147450B (en) | Knowledge complementing method and device for knowledge graph | |
| CN109885907A (en) | A kind of Satellite Attitude Control System health state evaluation and prediction technique based on cloud model | |
| CN109961186A (en) | Desulphurization system operating parameter prediction technique based on decision tree and BP neural network | |
| CN111723527B (en) | Method for predicting residual life of gear based on cocktail long-short-term memory neural network | |
| CN110689171A (en) | Turbine health state prediction method based on E-LSTM | |
| CN107133400A (en) | A kind of aircraft structure fatigue reliability Bayes's combination forecasting method | |
| CN107545375A (en) | A kind of high-risk equipment fault quantifying risk evaluation method of petroleum and petrochemical industry | |
| CN105740984A (en) | Product concept performance evaluation method based on performance prediction | |
| Okoro et al. | Multi-criteria risk assessment approach for components risk ranking–The case study of an offshore Wave Energy Converter | |
| CN107608658A (en) | Fault tree synthesis Similarity Match Method based on hidden Markov | |
| CN114898121A (en) | Concrete dam defect image description automatic generation method based on graph attention network | |
| CN115982141A (en) | Characteristic optimization method for time series data prediction | |
| CN112529053A (en) | Short-term prediction method and system for time sequence data in server | |
| CN116611556A (en) | Compact gas well single well yield prediction method based on hybrid neural network | |
| CN115238573A (en) | Hydroelectric generating set performance degradation trend prediction method and system considering working condition parameters | |
| CN114564881A (en) | Drilling riser fracture failure risk evaluation method based on fuzzy Bayesian network | |
| CN112070399B (en) | Safety risk assessment method and system for large-scale engineering structure | |
| Yan et al. | Reliability prediction of CNC machine tool spindle based on optimized cascade feedforward neural network | |
| Yang | Risk management of prefabricated building construction based on fuzzy neural network | |
| CN115618751B (en) | Steel plate mechanical property prediction method | |
| Dui et al. | Reliability Evaluation and Prediction Method with Small Samples. | |
| CN115564247A (en) | Intelligent identification and decision-making method and system for nuclear power plant accidents | |
| CN115153549A (en) | BP neural network-based man-machine interaction interface cognitive load prediction method | |
| Sharma et al. | Stochastic behaviour and performance analysis of an industrial system using GABLT technique | |
| CN115510752A (en) | Data-driven lateral drilling well position optimization method and device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |