Concrete structure damage identification method based on multi-source data fusion
By using Dempster's rule for decision-level data fusion, the problems of time-consuming, labor-intensive, and inaccurate traditional concrete structure damage detection are solved. This enables efficient and accurate identification of multi-source data, improving the efficiency and accuracy of concrete structure damage detection.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- ZHEJIANG UNIV
- Filing Date
- 2024-11-29
- Publication Date
- 2026-06-02
Smart Images
Figure CN119670007B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of building structure condition detection technology, and in particular to a method for identifying damage in concrete structures based on multi-source data fusion. Background Technology
[0002] Due to material aging and load-bearing effects, reinforced concrete structures may experience various types of damage and material performance degradation during service, leading to a decrease in structural load-bearing capacity and functionality. In some cases, this can result in exposed and corroded internal steel reinforcement, and in severe cases, even complete structural failure or collapse. Therefore, it is necessary to conduct damage detection on concrete structures to promptly ascertain their performance status.
[0003] Traditional concrete structure damage detection mainly relies on visual inspection (observing directly visible damage such as concrete cracks, efflorescence, spalling, and exposed rebar), structural displacement measurement, concrete strength testing based on concrete rebound hammers, reinforcement detection based on rebar location instruments, structural tilt detection, and remodeling and recalculating bearing capacity. However, these methods are time-consuming and labor-intensive, and their accuracy is questionable. In recent years, the application of deep learning-based computer vision technology and image processing algorithms has attracted increasing attention from scholars. Their ability to extract, measure, and quantify the size of surface cracks in concrete structures helps determine the damage state of buildings. However, this approach mostly relies on a single source of damage identification data, making it difficult to guarantee the accuracy and reliability of detection and identification. If multiple different damage identification data sources are considered for fusion analysis, problems of processing uncertainty and conflict arise. Summary of the Invention
[0004] The purpose of this invention is to overcome the shortcomings of the existing technology by providing a method for identifying damage in concrete structures based on multi-source data fusion, which can solve the conflict problem between different data sources and improve the efficiency and accuracy of damage identification.
[0005] The objective of this invention can be achieved through the following technical solution: a method for identifying damage in concrete structures based on multi-source data fusion, comprising the following steps:
[0006] S1. Obtain the damage prediction results corresponding to different data sources;
[0007] S2. Based on Dempster's rule, different damage prediction results are fused at the decision level to obtain damage identification results.
[0008] Furthermore, the different data sources in step S1 include image information, sound information, and acceleration information;
[0009] Furthermore, the damage prediction results in step S1 include no damage, minor damage, moderate damage, or severe damage.
[0010] Further, step S2 includes the following steps:
[0011] S21. Predetermine the basic reliability allocation values for damage prediction corresponding to different data sources;
[0012] S22. Use Dempster's synthesis rule to calculate the basic reliability assignment value after fusion, which yields the damage identification result.
[0013] Furthermore, in step S21, if the data source is image information, then the basic reliability allocation value corresponding to the no-damage prediction result, the basic reliability allocation value corresponding to the minor damage prediction result, the basic reliability allocation value corresponding to the moderate damage prediction result, and the basic reliability allocation value corresponding to the severe damage prediction result are defined respectively.
[0014] Furthermore, in step S21, if the data source is sound information, then the basic reliability allocation value corresponding to the prediction result of no damage, the basic reliability allocation value corresponding to the prediction result of minor damage, the basic reliability allocation value corresponding to the prediction result of moderate damage, and the basic reliability allocation value corresponding to the prediction result of severe damage are defined respectively.
[0015] Furthermore, in step S21, if the data source is acceleration information, then the basic reliability allocation value corresponding to the no-damage prediction result, the basic reliability allocation value corresponding to the minor damage prediction result, the basic reliability allocation value corresponding to the moderate damage prediction result, and the basic reliability allocation value corresponding to the severe damage prediction result are defined respectively.
[0016] Further, step S22 specifically includes:
[0017] S221. Calculate the normalization coefficient;
[0018] S222. Use the normalization coefficient to solve for the basic confidence assignment value corresponding to each damage prediction result after fusion. The damage prediction result with the highest basic confidence assignment value after fusion is the damage identification result.
[0019] Furthermore, in step S221, since no two damage states can occur simultaneously, the intersection between any two is always an empty set. The normalization coefficient is then calculated as follows:
[0020]
[0021] Wherein, K3 is the normalization coefficient, A1, A2, and A3 represent different damage states, I represents the prediction result of no damage, II represents the prediction result of minor damage, III represents the prediction result of moderate damage, IV represents the prediction result of severe damage, m1 represents the basic reliability assignment value when the data source is image information, m2 represents the basic reliability assignment value when the data source is sound information, and m3 represents the basic reliability assignment value when the data source is acceleration information.
[0022] Furthermore, the basic confidence assignment values corresponding to each fused damage prediction result in step S222 are as follows:
[0023]
[0024] Where, m 123 (I) represents the basic reliability assignment value corresponding to the fused non-destructive prediction result, m 123 (II) represents the basic reliability assignment value for the fused minor injury prediction results, m 123 (III) represents the basic reliability assignment value corresponding to the fused moderate injury prediction result, m 123 (IV) represents the basic reliability assignment value corresponding to the fused severe damage prediction results.
[0025] Compared with the prior art, the present invention has the following advantages:
[0026] This invention addresses the conflicting issues among various damage identification data sources. Based on Dempster's rule, it merges evidence from different sources using certain rules, achieving the goals of processing multi-source information, handling uncertainty and conflict, and ensuring flexibility and theoretical completeness. It then performs decision-level data fusion on damage prediction results from different data sources, processing uncertainties and potentially conflicting parts of these data, effectively resolving conflicts between different data sources, and significantly improving the efficiency and accuracy of damage identification.
[0027] When applying a decision-level data fusion method based on DS evidence theory to different data sources, this invention predetermines the basic confidence assignment values corresponding to the four possible damage prediction results for a certain node from three different identification data sources. Then, by calculating the normalization coefficient, the basic confidence assignment value after fusion is calculated. This can improve the accuracy of damage judgment and accurately and efficiently identify and detect the damage state of any node in a concrete structure. Attached Figure Description
[0028] Figure 1 This is a schematic diagram of the method flow of the present invention;
[0029] Figure 2 This is a schematic diagram of the application process in an embodiment. Detailed Implementation
[0030] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments.
[0031] Example
[0032] To facilitate understanding of this plan, the following explanation will first address the conceptual basis of this plan:
[0033] To further improve the accuracy, efficiency, and reliability of concrete structure damage identification, this proposal aims to extract more comprehensive damage information from image-based damage information and damage information based on generalized acoustic features (including sound and acceleration features) using data fusion methods. Data fusion typically involves three levels of fusion operations: data-level fusion, feature-level fusion, and decision-level fusion. Since it is difficult to achieve data-level and feature-level fusion of image information and acoustic features at the first two levels, this proposal adopts the third-level—decision-level fusion—for research and discussion.
[0034] The primary goal of decision-level fusion is to combine information from multiple sources (in this case, acceleration, sound, and image information) to make more accurate and effective decisions. This level of fusion methods includes rule-based systems, Dempster-Shafer (DS) evidence theory, Bayesian networks, and other advanced statistical and artificial intelligence methods such as convolutional neural networks. These methods can handle large amounts of information from diverse data sources and make fast and accurate decisions.
[0035] This approach uses the DS evidence theory to perform decision-level data fusion on information from three different sources, processing the uncertain and potentially conflicting parts of this information to achieve a better and more stable damage identification effect.
[0036] Dempster-Shafer (DS) evidence theory, also known as the belief function theory, was proposed by Arthur Dempster and Glenn Shafer in the 1970s. It is a mathematical framework for understanding and processing uncertain information. DS evidence theory mainly consists of two parts: the belief function and evidence fusion.
[0037] In the DS evidence theory, after data fusion, all possible outcomes or events (in this scheme, the degree of node impairment, still categorized into four classes from 1 to 4) form a set, which is called the identification framework:
[0038] Θ = {θ1, θ2, ..., θ n},
[0039] Where n is the total number of hypotheses (n = 4 in this scheme). For each subset (each damage state), a confidence value can be assigned, ranging from 0 to 1, representing the degree of confidence in that subset. These confidence values collectively constitute a confidence function. The confidence function of proposition A (e.g., a node damage state of level 2—minor damage) under the identification framework Θ is defined as follows: For any subset A in the identification framework Θ, the confidence function Bel(A) is equal to the sum of the basic confidence assignments m(B) of all subsets B (B is contained in A), which can be written mathematically as:
[0040]
[0041] Where, mapping m: 2 Θ →[0,1] satisfies
[0042]
[0043] m(A) represents the degree of support for proposition A from the currently available evidence, also known as the basic confidence assignment. If the basic confidence assignment m(A) > 0 for a proposition, it is called the focal element of the confidence function Bel(A). The confidence function Bel expresses the degree of confidence in that proposition (considering cause and effect).
[0044] Besides the confidence function Bel, there is another function that measures the correctness of a proposition, called the plausibility function (Pl). Let there be a mapping Pl:2 Θ →[0,1] satisfies
[0045]
[0046] Then Pl is called the plausibility function. The plausibility of A, Pl(A), represents the degree to which A is not doubted. It includes the basic reliability assignment of all propositions compatible with A.
[0047] When there are two or more confidence functions (in this case, multiple sources of data for identifying impairments), the Dempster rule can be used to fuse them to obtain a new confidence function, thus achieving decision-level data fusion. The following is a brief introduction to the specific application and computational route of the Dempster rule when there are two or more confidence functions.
[0048] I. Dempster's Rule for Fusion of Two Belief Functions
[0049] Suppose that in a recognition framework Θ there exist two recognition functions Bel1 and Bel2, which have corresponding basic confidence assignments m1 and m2, respectively, and their focal elements are A1, A2, ..., Ai and B1, B2, ..., B j ,but
[0050]
[0051] This is Dempster's rule for combining two confidence functions, where K represents the degree of conflict between different pieces of evidence. If the evidence behind the propositions corresponding to m1 and m2 completely conflict, then K is 0, and the combination function does not exist; if the evidence behind the propositions corresponding to m1 and m2 completely support each other, then K is 1. These are the upper and lower bounds of K, and in practical applications, K will be between 0 and 1. Similarly, for propositions with three or more confidence functions, the Dempster rule for their combination can be generalized in a similar manner.
[0052] II. Dempster's Rule for Fusion of Multiple Belief Functions
[0053] Suppose that there are n confidence functions Bel1, Bel2, ..., Bel in a recognition framework Θ. n Each of them has a corresponding basic reliability assignment m1, m2, ..., m n ,like It exists and its basic credibility is assigned to m, and for All Then there is
[0054]
[0055] in
[0056] It can be seen that the synthesis formula of the multi-confidence function is obtained by repeatedly nesting and repeating the synthesis formula of the two-confidence function, and the synthesis operation... It has the commutative and associative laws, that is... This new confidence function will reflect the common information of all the original confidence functions, thus providing more comprehensive, reliable and robust results.
[0057] A key advantage of the DS evidence theory lies in its ability to handle incomplete and conflicting information. Compared to other methods for dealing with uncertainty, such as probability theory or fuzzy set theory, DS evidence theory provides a richer and more refined characterization of uncertainty. Therefore, it has been widely applied in many fields, including artificial intelligence, machine learning, information fusion, and decision analysis. This paper will apply DS evidence theory to the fusion of multi-source data on concrete structure damage, exploring the results of concrete structure damage identification based on multi-source data fusion.
[0058] This proposal suggests a method for identifying damage in concrete structures based on multi-source data fusion, such as... Figure 1 As shown, it includes the following steps:
[0059] S1. Obtain the damage prediction results corresponding to different data sources;
[0060] S2. Based on Dempster's rule, different damage prediction results are fused at the decision level to obtain damage identification results.
[0061] This embodiment applies the above-described solution, such as Figure 2 As shown, based on Dempster's rule, different confidence functions (based on image, sound, and acceleration recognition methods) are synthesized into a new confidence function using corresponding evidence information, thereby improving recognition accuracy and robustness. Specific content includes:
[0062] 1. Define the recognition framework
[0063] When applying a decision-level data fusion method based on DS evidence theory to different data sources, it is necessary to predetermine the basic confidence assignment values m corresponding to the four possible damage scenarios of a certain node for the three different identification methods. i In this embodiment, the average error of semantic segmentation in identifying cracks is 10.45%. According to the definition of damage level, no-damage (I) nodes have no visible cracks, minor-damage (II) nodes have cracks with a width less than 1 mm, moderate-damage (III) nodes have cracks with a width greater than 1 mm, and severe-damage (IV) nodes have obvious spalling or exposed rebar, which can be defined as cracks with a width greater than 5 mm. If the average crack width of a node (data source 1) falls into damage level A, then the basic confidence assignment value corresponding to level A can be set to 0.8955, and the basic confidence assignment values for the other three levels are (1-0.8955) / 3 = 0.035. For example, if the semantic segmentation of a crack at a node and the calculated average crack width are 1.2 mm, then it belongs to moderate damage, with m1(II) = 0.8955, m1(I) = m1(III) = m1(IV) = 0.035. For damage classification results based on sound information (data source 2), according to the corresponding confusion matrix and the prediction accuracy obtained from the test, its corresponding basic reliability assignment value can be set as m2(III) = 0.9805, m2(I) = m2(II) = m2(IV) = 0.0065. Similarly, for damage classification results based on acceleration information (data source 3), its corresponding basic reliability assignment value is m3(III) = 0.9472, m3(I) = m3(II) = m3(IV) = 0.0176.
[0064] In summary, for a damaged node S, if the predicted damage results from image, sound, and acceleration data are slight damage, moderate damage, and severe damage, respectively, then the following basic confidence assignment table can be constructed based on the above assumptions:
[0065] Table 1. Basic Reliability Allocation Table for Damage to Node S
[0066] <![CDATA[m1()]]> <![CDATA[m2()]]> <![CDATA[m3()]]> No damage (I) 0.035 0.0065 0.0176 Minor injury (II) 0.8955 0.0065 0.0176 Moderate injury (III) 0.035 0.9805 0.9472 Severe injury (IV) 0.035 0.0065 0.0176
[0067] The recognition framework is then Θ = {I,II,III,IV}; where subscript 1 represents data from the image, subscript 2 represents sound data, and subscript 3 represents acceleration data. Based on Dempster's rule theory and corresponding calculation formula for multi-belief function fusion, the basic confidence assignment value m for the damage state after fusing the three types of data can be calculated. 123 .
[0068] 2. Use Dempster's rule of composition to find the confidence function of the fusion.
[0069] First, the normalization coefficient K3 needs to be determined. Since no two damage states in the recognition framework can occur simultaneously, the intersection between any two damage states is always an empty set. K3 can be calculated as follows:
[0070]
[0071] The basic reliability assignment after fusion can be obtained using K3. Since no two elements in the recognition frame intersect, the only case where all three elements are A is the intersection of the three elements. Therefore:
[0072]
[0073] Finally, the above calculation process yielded the fused basic reliability assignment value m. 123 (), meaning the prediction results for this node show that damage state III (moderate damage) has the highest basic confidence score of 0.8454. Therefore, for this node, the decision-level data fusion prediction result is damage state III. Combining the above data with Table 1, we can obtain the following table:
[0074] Table 2. Basic Reliability Allocation of Damage to Node S After Data Fusion
[0075] <![CDATA[m1()]]> <![CDATA[m2()]]> <![CDATA[m3()]]> <![CDATA[m 123 ()]]> No damage (I) 0.035 0.0065 0.0176 0.005604 Minor injury (II) 0.8955 0.0065 0.0176 0.1433 Moderate injury (III) 0.035 0.9805 0.9472 0.8454 Severe injury (IV) 0.035 0.0065 0.0176 0.005604
[0076] Using this scheme, the damage status of any node can be achieved through decision-level data fusion via the above path, thereby improving the accuracy of damage assessment. When there is a conflict among the three classification methods, the data fusion method can make the optimal decision and avoid incorrect classification results as much as possible.
[0077] The above content uses a single node as an example to explore the entire process of decision-level data fusion based on Dempster's rule and demonstrates the final fusion result. To verify the effectiveness of this solution, this embodiment performs data fusion operations on more nodes and compares the final recognition accuracy to objectively present the effect of data fusion.
[0078] To compare the recognition accuracy, data fusion based on DS evidence theory and three prediction data sources was performed on seven nodes of the pre-reinforcement concrete structure to determine their corresponding damage states. The results are shown in Table 3. The table only includes nodes with slightly different results for different recognition methods, as decision-level data fusion is meaningless for nodes with identical results from all three methods. It should be noted that since the acoustic feature assessment of node damage is based on the signal generated by a single impact, the prediction results based on acoustic features for the same node may differ (due to unstable impact force or external noise), resulting in the same node appearing multiple times in the table.
[0079] Table 3 Comparison of Damage State Prediction Results Based on DS Evidence Theory
[0080]
[0081] As can be seen from Table 3, the accuracy of data fusion reached 100% at these "ambiguous" nodes, proving the effectiveness of decision-level data fusion and resolving the problem of conflicting evidence when different data sources are available.
[0082] In summary, this approach addresses the conflicting issues among various damage identification data sources by researching a decision-level data fusion method based on Dempster-Shafer evidence theory. This method merges evidence from different sources using certain rules, achieving the goals of handling multi-source information, addressing uncertainty and conflict, and providing flexibility and theoretical completeness. These characteristics enable the fusion method based on the DS evidence rule to achieve high accuracy while resolving conflicts between evidence, thus improving the efficiency and accuracy of damage identification in concrete structures.
Claims
1. A method for identifying damage to concrete structures based on multi-source data fusion, characterized in that, Includes the following steps: S1. Obtain the damage prediction results corresponding to different data sources, including image information, sound information and acceleration information. The damage prediction results include no damage, slight damage, moderate damage or severe damage. S2. Based on Dempster's rule, decision-level data fusion is performed on different damage prediction results to obtain damage identification results; Step S2 includes the following steps: S21. Predetermine the basic reliability allocation values for damage prediction corresponding to different data sources; S22. Use Dempster's synthesis rule to calculate the basic confidence assignment value after fusion, which yields the damage identification result. Step S22 specifically includes: S221. Calculate the normalization coefficient; S222. Use the normalization coefficient to solve for the basic confidence assignment value corresponding to each damage prediction result after fusion. The damage prediction result with the highest basic confidence assignment value after fusion is the damage identification result.
2. The method for identifying concrete structure damage based on multi-source data fusion according to claim 1, characterized in that, In step S21, if the data source is image information, then the basic reliability allocation value corresponding to the prediction result of no damage, the basic reliability allocation value corresponding to the prediction result of minor damage, the basic reliability allocation value corresponding to the prediction result of moderate damage, and the basic reliability allocation value corresponding to the prediction result of severe damage are defined respectively.
3. The method for identifying concrete structure damage based on multi-source data fusion according to claim 1, characterized in that, In step S21, if the data source is sound information, then the basic reliability allocation value corresponding to the prediction result of no damage, the basic reliability allocation value corresponding to the prediction result of minor damage, the basic reliability allocation value corresponding to the prediction result of moderate damage, and the basic reliability allocation value corresponding to the prediction result of severe damage are defined respectively.
4. The method for identifying concrete structure damage based on multi-source data fusion according to claim 1, characterized in that, In step S21, if the data source is acceleration information, then the basic reliability allocation value corresponding to the no-damage prediction result, the basic reliability allocation value corresponding to the minor damage prediction result, the basic reliability allocation value corresponding to the moderate damage prediction result, and the basic reliability allocation value corresponding to the severe damage prediction result are defined respectively.
5. The method for identifying concrete structure damage based on multi-source data fusion according to claim 1, characterized in that, In step S221, since no two damage states can occur simultaneously, the intersection between any two is always an empty set. Therefore, the normalization coefficient is calculated as follows: ， ， ， in, The normalization coefficient is... , , Different damage states, This indicates the result of the no-damage prediction. This indicates the prediction result for minor injury. This indicates the prediction result of moderate injury. This indicates the prediction results of severe injury. This represents the basic confidence assignment value when the data source is image information. This represents the basic reliability assignment value when the data source is audio information. This represents the basic reliability assignment value when the data source is acceleration information.
6. The method for identifying concrete structure damage based on multi-source data fusion according to claim 5, characterized in that, The basic confidence assignment values for each fused damage prediction result in step S222 are as follows: ， in, Assign a base reliability value to the fused non-destructive prediction result. Assigning basic reliability values to the fused minor injury prediction results. The base reliability assignment value is the result of the fused moderate injury prediction. The basic reliability assignment value is given to the fused severe damage prediction results.