An indoor concrete quality objectification intelligent acceptance evaluation method

Abstract

The application discloses an indoor concrete quality objectification intelligent acceptance evaluation method and relates to the technical field of acceptance management. The whole scheme focuses on the management scene of engineering quality acceptance control and responsibility distribution, improves the rigorousness of concrete entity quality acceptance, and further improves the management logic of on-site acceptance. Relying on a hierarchical classification attribution judgment mode, management disputes caused by premature locking of a single responsibility object are avoided. Relying on the matching of quality judgment standards of corrected acceptance measured data, the acceptance qualification evaluation is adapted to the actual on-site management and control demand. The corresponding division of responsibility teams can be completed automatically relying on the final attribution conclusion, and the management problems such as repeated rectification and responsibility shirking caused by attribution misjudgment and inaccurate data are reduced.

Description

Technical Field

[0001] This invention relates to the field of acceptance management technology, specifically to an intelligent acceptance evaluation method for indoor concrete quality. Background Technology

[0002] In actual engineering projects, after the construction of concrete walls, beams, columns, and floors is completed, the acceptance process needs to be recorded, analyzed, and the attribution determined. The acceptance objects are then mapped to specific components to support subsequent acceptance evaluation and responsibility allocation management.

[0003] In the intelligent acceptance process of indoor concrete, existing methods, when mapping the acceptance object to the component, usually prioritize the principle of consistent attachment to complete the component attribution in order to speed up the acceptance process. However, if there are multiple component attachment conditions, there is often a lack of allocation and dynamic correction of the rationality of attribution, resulting in erroneous acceptance as the acceptance process progresses. The above problems are essentially because the traditional acceptance model focuses more on process efficiency and ignores the complexity of concrete distribution and the non-ideal nature of on-site testing operations, resulting in chaotic division of responsibility teams, misassignment of maintenance tasks, repeated rectification or no one in charge, which further reduces the reliability of subsequent operation and maintenance management. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides an intelligent acceptance and evaluation method for indoor concrete quality, which solves the problems in the background technology.

[0005] To achieve the above objectives, the present invention provides the following technical solution: A method for intelligent acceptance evaluation of indoor concrete quality includes the following steps: Obtain the original acceptance record set and management attribute data of indoor concrete, and combine them with the component type status to obtain the acceptance group for attribution analysis; Based on the risk level of the acceptance group analysis, a risk index for the status stratification of the acceptance group is obtained. The acceptance group is hierarchically classified, and host components are allocated and candidate sets are screened according to the different status acceptance groups. Combined with the acceptance operation data during the acceptance process, several candidate paths are determined. Based on several candidate paths, the attribution label is identified and updated for the acceptance group to obtain the attribution results used for responsibility allocation, and the acceptance qualification is determined by the attribution results.

[0006] Preferably, the original acceptance record set and management attribute data of indoor concrete are obtained, and combined with the component type status, acceptance groups for attribution analysis are obtained, including: Perform integrity verification on the original acceptance record set of indoor concrete to obtain a standardized dataset; the original acceptance record set should at least contain the defect type, location description, size information, collection time, construction batch and acceptance operation data of each text record; Based on a standardized dataset, the type status of components in each text record is analyzed, potential group records are marked, and the acceptance records of indoor concrete are classified and managed according to the potential group records to obtain acceptance groups.

[0007] Preferably, the process of acquiring the original acceptance record set and management attribute data of indoor concrete, and combining this with the component type and status to obtain the acceptance group for attribution analysis, further includes: Read the management attribute data of concrete components, match each acceptance group with the management attribute data to obtain the matching degree, and filter out the components with the matching degree higher than the preset matching threshold as acceptance management components, thus obtaining the set of management components for each acceptance group.

[0008] Preferably, based on the acceptance group, the risk level of the assigned acceptance is analyzed to obtain a risk index for the status stratification of the acceptance group, including: Based on the set of management components, the matching degree between each text record in the acceptance group and the acceptance management component is converted into an attribution contribution value. The attribution contribution values ​​of all text records are summed and averaged to obtain the attribution confidence vector of each acceptance group to each acceptance management component. Extract the maximum and second-largest values ​​of the attribution confidence within the attribution confidence vector, analyze the intensity of competition among multiple acceptance management components for the acceptance group, obtain the attribution difference, and obtain the cross-component responsibility overlap ratio by statistically analyzing the proportion of text records that simultaneously meet several acceptance management components within the acceptance group. The risk index for attribution acceptance is constructed by calculating the ratio between the cross-component responsibility overlap ratio and the attribution difference. If the risk index exceeds the preset affixing threshold, it is pre-marked as unstable attribution; otherwise, it is pre-marked as stable attribution.

[0009] Preferably, the acceptance group status is hierarchically layered, and host component allocation and candidate set screening are performed according to different status acceptance groups. Combined with acceptance operation data during the acceptance process, several candidate paths are determined, including: Based on the attribution confidence vector, the acceptance group with stable responsibility attribution is assigned the acceptance management component with the highest attribution confidence as the host; For acceptance groups with unstable responsibility attribution, acceptance management components with attribution confidence exceeding the average attribution confidence are selected and included in the candidate host set. Each acceptance management component in the candidate host set is combined with its corresponding acceptance group to construct several candidate paths; Preferably, the acceptance group status is hierarchically layered, and host component allocation and candidate set screening are performed according to different status acceptance groups. Combined with acceptance operation data during the acceptance process, several candidate paths are determined. This also includes: Based on the acceptance operation data within all candidate paths, analyze the local density of acceptance operations, and combine the changes in operation result values ​​to determine the hardening layer and correct the acceptance operation data within it. Based on the standardized dataset of the acceptance group, the host corresponding to the hardening layer and several candidate paths are recorded.

[0010] Preferably, based on several candidate paths, the attribution label is identified and updated for the acceptance group to obtain the attribution result for responsibility allocation, and the acceptance qualification is determined based on the attribution result, including: Based on the cross-component responsibility overlap ratio, cross-boundary correlation analysis is performed on the attribution confidence of each candidate path to construct a post-evaluation index for acceptance judgment, and the index is grouped and aggregated according to the acceptance group to obtain the component evaluation value. Select the component with the highest component evaluation value that exceeds the preset evaluation threshold as the preliminary acceptance attribution label; otherwise, mark it as belonging to the undecided group. Record the acceptance group corresponding to the preliminary acceptance attribution label as the attribution group. Traverse the pending attribution group and the attribution group, analyze the impact of attribution propagation on the acceptance group, and update the labels of the pending attribution group and the attribution group, as well as the acceptance management components under the labels.

[0011] Preferably, based on several candidate paths, the attribution label identification and update are performed on the acceptance group to obtain the attribution result for responsibility allocation, and the acceptance qualification is determined by the attribution result. This also includes: Based on the updated pending attribution group and the attribution group's labels and the acceptance management components under the labels, high-conflict acceptance groups are identified, and each acceptance management component in the high-conflict acceptance group is used as an acceptance attribution label. Based on the acceptance groups corresponding to several candidate paths within the hardened layer, determine the corresponding acceptance attribution labels and record the attribution results. Then, automatically assign the corresponding responsible work teams based on the attribution results. Read the operation results from the corrected acceptance operation data, analyze the acceptance evaluation results of the assigned results. If the operation results exceed the preset acceptance threshold, the corresponding assigned results are determined to be acceptable; otherwise, the acceptance is unacceptable.

[0012] The above-described solution of the present invention has at least the following beneficial effects: This solution focuses on the management scenarios of engineering quality acceptance control and responsibility allocation. While improving the rigor of concrete entity quality acceptance, it further refines the management logic of on-site acceptance. Relying on a hierarchical and categorized attribution judgment model, it avoids management disputes caused by prematurely identifying a single responsible party. By matching the revised acceptance measurement data with quality judgment standards, it ensures that the acceptance qualification evaluation aligns with the actual on-site control needs. It can automatically complete the assignment of responsible teams based on the final attribution conclusion, reducing management problems such as repeated rectification and shirking of responsibility caused by misjudgment of attribution and inaccurate data.

[0013] By rationally dividing acceptance groups based on acceptance information and the actual correlation characteristics of components, and quantifying the risk of unstable attribution based on the degree of competition among component associations and cross-regional situations, and establishing a graded handling and differentiated attribution judgment path according to risk differences, it is possible to identify potential responsibility delineation hazards for defects in boundary areas in advance from a management perspective. At the same time, for operational problems that occur during on-site acceptance operations, abnormal operation areas are accurately identified and acceptance measurement results are corrected, avoiding the problem of quality evaluation distortion caused by acceptance operation errors. By using attribution weight constraints, multi-dimensional indicator fusion, and neighborhood correlation verification, ambiguous attribution problems are dynamically corrected, effectively improving the management drawbacks of rigid attribution judgment and one-sided responsibility division in traditional acceptance. Attached Figure Description

[0014] Figure 1 This is a flowchart of an intelligent acceptance and evaluation method for indoor concrete quality according to the present invention. Detailed Implementation

[0015] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0016] like Figure 1 As shown, an embodiment of the present invention provides an intelligent acceptance assessment method for indoor concrete quality, comprising the following steps: S100: Obtain the original acceptance record set and management attribute data of indoor concrete from the engineering quality management system, mobile data acquisition terminal or historical inspection batch form, and combine them with the component type status to obtain the acceptance group for attribution analysis. S200: Based on the risk level of the acceptance group analysis, a risk index is obtained for the status stratification of the acceptance group; S300: Execute the acceptance group status stratification, and perform host component allocation and candidate set screening according to the different status acceptance groups, and determine several candidate paths by combining the acceptance operation data during the acceptance process; S400: Based on several candidate paths, perform attribution label identification and update for the acceptance group to obtain the attribution results for responsibility allocation, and determine the acceptance qualification through the attribution results.

[0017] Acceptance group status stratification refers to classifying each acceptance group into different status categories based on the degree of uncertainty of its classification, so as to carry out differentiated processing in the future and avoid premature solidification of classification. In this embodiment of the invention, it is not merely about adding layers, paths, and update steps to the process, but rather about reorganizing the acceptance data in both time and structural dimensions, so that the attribution determination is transformed from a one-time static decision into a dynamic process with evolutionary capabilities. First, the acceptance group formed in S100 transforms discrete defect records into data carriers with internal correlation structures, so that all subsequent calculations revolve around this structure. On this basis, S200 introduces a risk index to quantify the uncertainty of attribution, which is equivalent to establishing an evolvable threshold boundary for each acceptance group. This boundary not only determines whether it enters the hierarchical processing, but also implicitly constrains its search range in subsequent path expansion. Furthermore, in S300, a differentiated candidate path generation mechanism is triggered by state layering, which expands the original single attribution relationship into multiple parallel paths. This path expansion essentially constructs an implicit multi-hypothesis reasoning space. The introduction of acceptance operation data is equivalent to superimposing behavioral constraints in this space, so that the path is no longer determined solely by the static matching relationship, but is modulated by the distribution of detected behavior, thereby introducing behavioral consistency screening at the path level. Finally, in S400, the multi-path competition results are compressed and output by identifying and updating the path-level attribution labels. However, this output no longer depends on the initial attachment relationship, but comes from the game and selection results between paths.

[0018] Specifically, this method forms a delayed convergence attribution determination mechanism at the data level, that is, it actively suppresses the certainty of the attribution result in the early stage, allowing information to spread fully in the path space, and then gradually converges through later constraints, thereby avoiding the problem of error locking caused by premature convergence in traditional methods. Meanwhile, since there is a coupling relationship between candidate paths and acceptance operation data, when a local detection behavior is abnormal, its impact will be naturally diluted in the path weight distribution, and will not directly affect the final assignment result. This indirect modulation mechanism makes the system exhibit stronger structural stability when facing abnormal inputs. Furthermore, since the risk index-driven state hierarchical structure essentially prioritizes the data for decision-making, the entire evaluation process exhibits adaptive characteristics in the allocation of computational resources, enabling the method to maintain stable processing efficiency and judgment consistency even in complex engineering scenarios.

[0019] In a preferred embodiment of the present invention, S100: Perform integrity verification on the original acceptance record set of indoor concrete and establish the association between defect records and component types to obtain a standardized dataset; this dataset is the defect records of the original records after field unification, format standardization, and enumeration value mapping to form a unified format data, so as to ensure that the correlation calculation is performed under a unified semantic; the original acceptance record set includes at least the defect type, location description, size information, collection time, construction batch, and acceptance operation data of each text record; Concrete structures are not limited to walls, columns, beams, floors, and repair areas; Defect types include at least cracks, honeycombing, pitting, and holes; location descriptions include at least axis position, elevation, and component number; and dimensional information includes at least length, width, and depth. Each record undergoes a field-by-field check, including checking the existence of required fields, whether the data format conforms to specifications, whether numerical fields are within a reasonable range, and whether there are completely duplicate records. If any check fails, the record is marked as abnormal and moved to the review queue. Data format conformity includes ensuring the defect type is within a preset enumeration value and that the location description conforms to axis numbering specifications. Based on a standardized dataset, the type status of components in each text record is analyzed, potential group records are marked, and the acceptance records of indoor concrete are classified and managed according to the potential group records to obtain acceptance groups.

[0020] In practical implementation, since defects may appear within a single component or at the boundary of components, the latter involves competition among multiple components when determining attribution. Therefore, cross-component associations must be considered during the unit division stage. Otherwise, cross-component defects that should belong to the same defect group will be forcibly split into multiple independent groups during division. Therefore, for each defect text record, the association set of other defect records related to each defect record within the coding threshold and the adjacent coding threshold is extracted; and the number of defect records within the same component and adjacent components is extracted. The conditions for obtaining this record set include association within the same component and association within adjacent components. Association within the same component is when the component ID is the same and the spatial location coding distance on the same component is less than the coding threshold; association between adjacent components is when the component IDs are different, but the two components have a topological adjacency relationship in the BIM model and the defect type is the same, and the spatial location coding distance on adjacent components is less than the adjacent coding threshold. For each record, the attribute similarity between it and each record in the associated set is calculated using the cosine similarity algorithm. Records whose attribute similarity variance exceeds a preset similarity threshold are marked as potential defect group boundary points, i.e., potential group records, which are used to delineate defect boundaries. At the same time, the local association density and its gradient magnitude of each record are calculated using the finite difference method to identify the abrupt change locations between the defect concentration area and the discrete distribution area. Records whose gradient magnitude exceeds a preset gradient threshold are marked as potential defect group boundary points. Traverse the records in the potential group, using non-potential group records as seeds, and employ a region growing algorithm to group interconnected records that are not separated by breakpoints into the same acceptance group. During partitioning, if the component type associated with a record is inconsistent with its neighboring records, the unit is marked as a cross-component boundary unit. Each group is accompanied by a list of record indices, feature vectors, and cross-component labels. After partitioning, ensure that records within the acceptance unit are strongly correlated and that records in different units do not overlap, while retaining the original concrete component information to which each record belongs.

[0021] Potential defect cluster boundary points refer to records where the attribute similarity changes drastically or the association density gradient changes abruptly. These records are located in the transition region between two or more defect clusters and are used to prevent different defect clusters from being mistakenly merged. The dimensions of the feature vector include the spatial location encoding, defect type encoding, and size. The management attribute data of concrete components is read, and each acceptance group is matched with the management attribute data to obtain the matching degree. Components with a matching degree higher than the preset matching threshold are selected as acceptance management components, resulting in the set of management components for each acceptance group. Through this step, a list of components that each acceptance group may belong to is generated, providing a complete management relationship basis for subsequent attribution analysis, while ensuring that the output data has continuity, integrity and traceability.

[0022] The management attribute data of concrete components is a structured data table extracted from the BIM model and oriented towards the needs of engineering quality acceptance management. It is used to associate the final acceptance conclusion with the responsible work team and support management decisions. It includes component index, type, construction time, responsible work team and acceptance standards. A weighted hybrid similarity algorithm is adopted to integrate matching degree components from multiple dimensions to calculate the matching degree used to screen acceptance management components and quantify their belonging priority. The matching degree components include component ID consistency component, component type matching degree component, spatial location matching degree component, and acceptance standard matching degree component. The weights of each item in this method can be obtained with reference to the analytic hierarchy process. The component ID consistency component is the ratio of the number of records with the same component ID to the total number of records in the acceptance group. It is used to measure the degree of consistency between the original component ID recorded in the acceptance group and the ID of the management attribute data. Both the component type matching degree component and the acceptance standard matching degree component are determined by the indicator function. When the component types are consistent or the defect type belongs to the corresponding component's defect type, the output is 1; otherwise, the output is 0. The spatial location matching degree component is based on Gaussian attenuation of the distance between the center of the acceptance group and the feature line of the component. First, the spatial location codes of all records in the acceptance group are averaged and used as the coordinates of the center point of the acceptance group. Then, the feature line of the component is obtained according to the BIM model to calculate the shortest distance from the center point of the acceptance group to the feature line of the component. This shortest distance is then determined by Gaussian attenuation mapping. The spatial attenuation parameter in the Gaussian attenuation mapping is used to control the distance attenuation rate. Specifically, it is calibrated by statistical analysis of historical data: the actual distance distribution from the correctly matched acceptance group to the feature line of the component in the completed acceptance projects is collected, and the standard deviation of this distribution is calculated as the optimized value of the spatial attenuation parameter. The spatial location matching degree component is used to measure the spatial proximity between the acceptance group and the management attribute data. The closer the distance, the higher the matching degree. If the shortest distance is 0, the spatial location matching degree component is 1. Characteristic lines are one-dimensional linear features extracted from the BIM model to represent key spatial positioning geometric elements of components; It should be noted that different matching thresholds are used for different component types. By collecting the project data of completed acceptance projects, for each component type, the matching degree distribution between the known correctly matched acceptance units and the component is statistically analyzed, and the mean-standard deviation method is used to determine the matching thresholds corresponding to different component types; the management component set includes multiple acceptance management components. In this embodiment of the invention, by reconstructing the defect records in terms of spatial relationships, attribute distribution and component topology, the originally discrete and unrelated acceptance records are gradually transformed into analysis units with structural continuity. In the initial field-level verification and standardization process, all records are unified into the same semantic space, so that the subsequent similarity calculation is no longer affected by format differences. Subsequently, the association set is extracted through dual association conditions of the same component and adjacent components, so that each record not only retains its own information, but also implicitly carries the structural relationship of its local neighborhood. On this basis, the variance discrimination of cosine similarity and the density gradient calculation of finite difference are introduced to identify similar but unstable and dense but abrupt records as potential boundary points, thereby forming a kind of structural fracture perception capability at the data level, so that subsequent regional growth no longer expands blindly, but forms an acceptance group with internal consistency but external separability under boundary constraints. Furthermore, the marking of cross-component boundary units during the region growth process gives each acceptance group a structural attribute of whether it crosses the boundary. This attribute does not directly participate in the attribution determination, but indirectly changes the ranking of the matching results by affecting the weight distribution of the component ID consistency component and the spatial location component in the subsequent matching stage. In the matching degree calculation stage, through the multi-component coupling of component ID consistency, type matching, spatial Gaussian decay, and acceptance criterion matching, each acceptance group is no longer forcibly mapped to a single component, but forms a set of components ordered by probability. This set actually retains the uncertainty of the attribution in advance. In particular, the introduction of an adaptive decay parameter based on the historical distribution standard deviation in the spatial location component makes the distance influence no longer a fixed threshold judgment, but dynamically adjusted with engineering experience, thereby automatically adapting to spatial scale differences in different projects.

[0023] Based on the above process, this method completes an implicit denoising and structural rearrangement in the data organization stage. That is, through boundary recognition and region growing, defects that may have been incorrectly split or merged are reorganized into units that are more in line with the real physical distribution, so that any subsequent calculations based on these units are built on a more stable structure. Meanwhile, since each acceptance group corresponds to a candidate set of components rather than a single result, decisions can be delayed in subsequent processing, and uncertainties can be reserved for processing at a higher level, thus avoiding the problem of early errors being amplified in traditional methods. Furthermore, this processing order of first constructing the structure and then calculating the matching makes the matching results depend not only on the features of a single record, but also on its position and relationship in the group. Thus, when facing cross-component defects or areas with ambiguous boundaries, it can naturally form a multi-component competitive relationship, rather than being forcibly assigned to a certain component. This competitive relationship becomes an important information carrier for subsequent decisions, ultimately transforming the entire acceptance evaluation process from point-to-point matching to structure-driven matching, exhibiting higher consistency and stability in complex scenarios.

[0024] In a preferred embodiment of the present invention, S200 includes: based on the set of management components, converting the matching degree between each text record in the acceptance group and the acceptance management component into an attribution contribution value through an exponential decay function, and averaging the attribution contribution values ​​of all text records to obtain an attribution confidence vector of each acceptance group for each acceptance management component; this vector is a multi-dimensional array containing the attribution confidence of the acceptance group for all associated components; used to characterize the attribution ambiguity of the unit itself, providing a preliminary basis for candidate path generation; By enhancing the decision weight of high-match records through nonlinear mapping, while suppressing the interference of low-match records, the attribution confidence more accurately reflects the overall attribution relationship between the acceptance group and the component.

[0025] The attribution contribution value is the voting strength of a single record on the attribution of a component, while the attribution confidence is the degree of confidence of the acceptance team as a whole in the attribution of the component. Extract the maximum and second-largest values ​​of the attribution confidence within the attribution confidence vector, analyze the intensity of competition among multiple acceptance management components for the acceptance group, obtain the attribution difference, and obtain the cross-component responsibility overlap ratio by statistically analyzing the proportion of text records that simultaneously meet several acceptance management components within the acceptance group. The attribution difference is the difference between the maximum and second-largest attribution confidence values. It is used to quickly determine whether the competition for attribution among multiple components is intense or whether the attribution is ambiguous. It serves as a preliminary judgment basis for risk index and status stratification. The smaller the difference, the more ambiguous the attribution. Specifically, for each acceptance group, its internal records are traversed, and the number of records whose matching degree with multiple acceptance management components is greater than the preset record matching threshold is counted. The ratio of this number to the total number of records in the acceptance group is calculated to obtain the cross-component responsibility overlap ratio. This ratio is used to identify the proportion of records in the acceptance group that are simultaneously attached to multiple acceptance management components, and to determine whether the acceptance group has crossed the component management boundary. The larger the ratio, the higher the degree of cross-component, and the more ambiguous the attribution. The risk index for attribution acceptance is constructed by calculating the ratio between the cross-component responsibility overlap ratio and the attribution difference. If the risk index exceeds the preset affixing threshold, it is pre-marked as unstable attribution; otherwise, it is pre-marked as stable attribution.

[0026] The risk index is used to measure the degree of ambiguity in the attribution of the acceptance team and the risk of responsibility division, thereby enabling rapid determination and stratification of the stability of the responsibility attribution of the acceptance team; The embodiments of the present invention construct a comparable competitive structure before the attribution decision, so that the local matching relationships originally scattered in each record are reorganized into a group decision expression with overall significance. First, the matching degree is compressed into the attribution contribution value through the exponential decay function, which essentially weakens the disturbance of low-confidence records on the overall judgment. At the same time, the influence of high-matching records is amplified by nonlinear mapping, so that the information distribution within the acceptance group changes from linear superposition to a weighted structure in which the strong dominate and the weak are restricted. Subsequently, by summing and averaging the contribution values ​​of all records, an attribution confidence vector is formed. This step is not a simple aggregation, but rather a transformation of single-point judgment into distributed judgment, so that each acceptance group maintains a relationship of different strengths with multiple components at the same time, thereby preserving the uncertainty of attribution. Based on this, by extracting the maximum and second-largest values ​​to construct the attribution difference, this distribution is further compressed into the intensity of competition. The cross-component responsibility overlap ratio characterizes the scope of competition from another dimension. The former reflects who has the advantage, and the latter reflects the scale of records participating in the competition. The two are combined by ratio to form a risk index, which is equivalent to coupling intensity and scope into a unified measure, so that fuzzy attribution no longer depends on empirical judgment, but is transformed into a calculable structural feature.

[0027] The resulting benefits are as follows: This method introduces a competitive visualization compression mechanism before attribution determination, compressing the multidimensional confidence distribution into a directional risk signal. This allows the method to identify which acceptance groups are in highly competitive and highly coupled unstable regions before making a final attribution decision, thus providing a basis for subsequent path development or decision delay. At the same time, since the risk index comes from the internal relationships of the group rather than a single record, the impact of local abnormal records is diluted by the overall distribution, which helps to reduce the phenomenon of abrupt changes in attribution determination. This group anti-disturbance characteristic makes the system more stable when facing noisy data or scenarios with ambiguous boundaries. Furthermore, since the cross-component responsibility overlap ratio characterizes the degree of spatial cross-boundary, while the attribution difference characterizes the degree of concentration of competitive focus, the coupling of the two actually constructs a structural uncertainty measure, enabling different types of uncertainty (such as multi-component attachment and equilibrium competition) to be uniformly expressed and ranked. This forms a rhythm control mechanism in the overall process of first identifying uncertainty and then processing uncertainty. This mechanism not only avoids the rigid problem of one-size-fits-all attribution in traditional methods, but also enables subsequent computing resources to be prioritized for truly complex acceptance groups, indirectly improving the computational efficiency and result consistency of the overall evaluation process.

[0028] In a preferred embodiment of the present invention, S300 includes: assigning the acceptance management component with the highest attribution confidence as the host to the acceptance group with stable responsibility attribution based on the attribution confidence vector; and automatically assigning the corresponding responsibility team according to the host. For acceptance groups with unstable responsibility attribution, acceptance management components with attribution confidence exceeding the average attribution confidence are selected and included in the candidate host set. Each acceptance management component in the candidate host set is combined with its corresponding acceptance group to construct several candidate paths; Candidate paths are obtained by expanding the candidate host set of the acceptance group. Specifically, each acceptance group and each acceptance management component in its candidate host set are combined into an independent evaluation path for the acceptance group and acceptance management component. Each path is bound to the complete attribution information of the acceptance group, including confidence level, matching degree statistics, risk index, etc., and a unique path identifier is assigned according to the defect ID, acceptance group ID and candidate sequence number format. Based on the acceptance operation data within all candidate paths, analyze the local density of acceptance operations, and combine the changes in operation result values ​​to determine the hardening layer and correct the acceptance operation data within it. A hardened layer refers to a dense surface layer formed on concrete due to repeated impacts and excessive compression in the same area using a rebound hammer. This layer has a higher density than the surrounding normal concrete, leading to inflated rebound values ​​and consequently, calculated strength values ​​that are higher than the actual strength. It should be noted that the hardened layer is not an original defect type but rather a result of errors during the acceptance assessment process. The acceptance operation data should include at least the operating position, number of operations, and operation results of the quality inspector using the rebound hammer, namely the impact point, the number of impact points, and the rebound value. Define a local area, for example, draw a circle with a radius of 25mm. Based on the three-dimensional spatial coordinates and point number of the operation position, count the number of operations in the local area and calculate the local density. Compare the local density with the preset density threshold. If it exceeds the density threshold, it is recorded as an over-dense bullet point. DBSCAN clustering is used, with a neighborhood radius and a minimum number of points set. Spatially adjacent overly dense impact points are grouped into the same overly dense cluster. Each cluster records the point index, convex hull boundary, centroid coordinates, average density, and maximum density. If the cluster area exceeds a preset area threshold, such as the area of ​​a circle with a diameter of 50mm, it is marked as an overly dense region. Read the timestamp interval between adjacent operations. If the interval is greater than the preset interval threshold, insert an interrupt flag, divide the operation sequence into several continuous subsequences, and record the start and end point numbers of each subsequence. For each continuous subsequence, calculate the first-order difference of the rebound value, with each difference value corresponding to a pair of impact points. Scan the difference sequence to find all continuous positive subsequences, i.e., multiple consecutive difference values ​​> 0. Treat all continuous positive subsequences as positive drift segments and record the length, start point number, and end point number of each subsequence. For each positive drift segment, extract the midpoint of the line connecting its first and last points, and use it as the point of a local area to draw a circle to obtain the drift region. Spatially overlay and match the drift region with the over-dense region to determine whether there are over-dense impact points falling into the region or the boundary of the over-dense cluster intersecting with the region. If so, it is confirmed that the drift segment is caused by over-dense impact, and all impact points in the segment are marked as hardening layer causative points. If there are multiple over-dense point matches, take the one with the closest midpoint of the line connecting the first and last points as the main causative point, and record its coordinates and the length of the drift segment in which it is located. A local region is generated centered on each hardening layer causation point. The union of all local regions is taken, and the convex hull is calculated to obtain the boundary of the hardened layer region. The total number of impact points, the mean of the original rebound value, and the mean of the density within the boundary are statistically analyzed.

[0029] If no match is found, it is marked as no impact hardening layer.

[0030] Based on the multiple sets of operation results of the hardened layer, the extreme values ​​of the operation results are removed, and the mean value is calculated on the removed data to obtain the average operation value. If the test is conducted in a non-horizontal direction, the corresponding correction value is determined by using the angle correction table in the instructions to correct the average operation value. The carbonization depth was determined by using a chemical colorimetric method with phenolphthalein indicator, combined with a method of detecting carbonization depth by drilling sampling and averaging multiple measurements. The carbonation depth and the corrected average operating value are substituted into the preset strength test curve to obtain the estimated strength, which is then used as the updated operating result to update the acceptance operation data. This strength value reflects the compressive strength of the concrete in the component and is used to determine the qualification of the concrete compressive strength. The strength curve is obtained as follows: First, a batch of concrete test blocks of different strength grades are made. The operation results are measured simultaneously on each test block using a rebound hammer. The actual strength is measured by crushing the test block with a press, and the carbonation depth is measured. Then, the operation results and actual strength of a large number of test blocks are used to construct a strength curve in the form of a power function. The least squares method is used to perform nonlinear regression analysis to calculate the specific values ​​of each regression coefficient, so as to minimize the error between the actual strength and the calculated estimated strength.

[0031] Carbonation depth is the depth at which carbon dioxide diffuses from the concrete surface inward and reacts with calcium hydroxide, a cement hydration product, to form calcium carbonate. It is the vertical distance from the concrete surface to the boundary between the purplish-red and colorless areas. The greater the carbonation depth, the more serious the artificially inflated result. Based on the standardized dataset of the acceptance group, the host corresponding to the hardening layer and several candidate paths are recorded.

[0032] This invention introduces a behavior-driven adaptive correction structure at the path level, enabling the attribution determination process to have reverse constraint capabilities on detection behavior while the structure is unfolded. First, when the unstable acceptance group is unfolded into multiple candidate paths, each path is bound with complete attribution information. This step indicates that multiple possible attribution interpretations are retained within the same acceptance group. Next, acceptance operation data is introduced into the path, and through local density calculation, clustering, time series segmentation, and first-order difference analysis, the originally uninterpretable operation behavior is transformed into a feature expression with spatial and temporal structure, such as overly dense regions, positive drift segments, and their spatial overlapping relationships. In effect, a layer of behavior consistency constraint is constructed within the path. Subsequently, by identifying excessively dense and drift coupling as the causes of hardening layer, and performing regional-level convex hull reconstruction and data removal correction, the operation results that each path depends on are no longer the original collected values, but effective values ​​after behavioral filtering and physical correction. Combined with carbonization depth detection and strength curve regression, the strength evaluation in the path is transformed from single-point measurement to an inferred result under multi-source constraints.

[0033] This method first reconstructs the behavior consistency of the data within each path, so that the differences between paths not only come from the attribution relationship itself, but also from the quality of the detection behavior they carry, forming a path quality stratification. That is, some paths are automatically weakened by the system due to operational anomalies in their corresponding areas, while other paths retain more stable data support due to the uniform distribution of operations. Furthermore, since the hardening layer identification is completed by coupling spatial density and temporal trend, it is essentially extracting traces of human intervention from the detection data. Therefore, the system can reverse identify and correct operational deviations in the detection process without relying on manual annotation, making the final intensity calculation closer to the real physical state. In addition, by binding and recording the hardening layer region with the candidate path, the subsequent attribution decision can perceive whether the path is based on abnormal operations, thus naturally forming an implicit penalty mechanism in the path competition stage. In a preferred embodiment of the present invention, S400 includes: performing cross-boundary correlation analysis on the attribution confidence of each candidate path based on the cross-component responsibility overlap ratio, constructing a post-evaluation index for acceptance judgment, and grouping and aggregating by acceptance group to obtain component evaluation value; Select the acceptance management component whose highest component evaluation value exceeds the preset evaluation threshold as the preliminary acceptance attribution label; otherwise, mark it as belonging to the undecided group. If there are multiple acceptance management components whose highest component evaluation value exceeds the preset evaluation threshold, select the acceptance management component with the highest component evaluation value as the preliminary acceptance attribution label. Record the acceptance group corresponding to the preliminary acceptance attribution label as the attribution group. The component evaluation value corresponding to the preliminary acceptance attribution label is recorded as the deterministic score; The classification of undecided groups needs further revision to ensure the accuracy of acceptance classification. In this invention, all parameters are dimensionless by using dimensionless processing technology to remove their dimensions; and all thresholds can be obtained by the mean-standard deviation method. In the acceptance evaluation of concrete components, if a path crosses multiple component boundaries, it indicates that its reference value for the attribution determination is low and it is easy to cause acceptance ambiguity. Therefore, it is necessary to impose a cross-boundary penalty on its confidence level to weaken its influence in the voting and ensure the reliability of the acceptance attribution evaluation. Thus, the original attribution confidence level is first used to reflect the credibility of the path itself to the component attribution, and then it is multiplied by the cross-boundary penalty coefficient to reduce the weight of the cross-component path. Finally, an unnormalized weight that reflects the attribution credibility and suppresses cross-boundary interference is obtained for subsequent acceptance attribution voting. Subsequently, normalization is performed on all candidate components in the same acceptance group, that is, the unnormalized weights are normalized to obtain the posterior evaluation index, which represents the normalized number of votes cast by the acceptance group when participating in the acceptance attribution vote. This is used to eliminate the voting bias caused by the uneven total confidence of different acceptance groups and to suppress the ambiguity of cross-component paths on the attribution determination.

[0034] Based on this, voting groups are constructed according to the acceptance group, and the posterior evaluation indicators of all paths are accumulated according to the component dimension to obtain the component evaluation value. The component evaluation value is then normalized to quantify the final belonging competitiveness of each component to the acceptance group, which serves as the core judgment basis for determining the preliminary belonging label. The cross-boundary penalty coefficient is the proportion of overlapping responsibilities across components. The larger the proportion of overlapping responsibilities across components, the more serious the cross-boundary violation. The smaller the cross-boundary penalty coefficient, the stronger the penalty and the lower the path weight.

[0035] The attribution difference is used to analyze the intensity of competition for attribution among multiple components in the acceptance group from the perspective of the original confidence level; the normalized component evaluation value is obtained by normalizing the voting accumulation result, and is used to quantify the final attribution competitiveness of each component and support the acceptance attribution determination. The attribution competitiveness of each component to the acceptance group is a comprehensive result obtained by dynamically weighting and fusing multiple candidate paths, and is used for the final attribution decision; Traverse the pending attribution group and the attribution group, analyze the impact of attribution propagation on the acceptance group, and update the labels of the pending attribution group and the attribution group, as well as the acceptance management components under the labels. In practice, neighborhood consistency propagation is used to correct the attribution of undecided units and optimize the accuracy of acceptance judgment. For each undecided group, its neighborhood set is extracted. Based on the attribution labels and deterministic scores of neighboring nodes, weighted neighborhood consistency is calculated to characterize the influence of neighborhood on the attribution of undecided units. Specifically, first, all adjacent acceptance groups of each undecided group are identified. Based on the attribution labels and deterministic scores of each adjacent acceptance group, the adjacent acceptance groups with confirmed attribution are counted. The deterministic scores of adjacent acceptance groups belonging to the same component are accumulated and used as the total support of the component. Then, the certainty scores of all adjacent acceptance groups are summed up as the total weight. The total support is divided by the total weight to obtain the corresponding weighted neighborhood consistency for the current unresolved group. This is used to determine the reasonable attribution of the unresolved group and to identify whether the attribution group conflicts with the neighborhood spatial pattern, so as to ensure the spatial consistency of acceptance attribution and the reliability of the assessment.

[0036] When the weighted neighborhood consistency exceeds the preset consistency threshold, the corresponding pending group will be updated to the corresponding component; for the assigned group, if the weighted neighborhood consistency does not exceed the preset consistency threshold, it will be recorded as a high-conflict acceptance group. Based on the updated pending group and the labels of the group and the acceptance management components under the labels, the high-conflict acceptance group is determined, and each acceptance management component in the high-conflict acceptance group is used as the acceptance attribution label; that is, the acceptance management component corresponding to each candidate path in the high-conflict acceptance group is used as the acceptance attribution label. Based on the acceptance groups corresponding to several candidate paths within the hardened layer, determine the corresponding acceptance attribution label at the hardened layer and record it as the attribution result. Then, automatically assign the corresponding responsible work group based on the attribution result. Read the operation results in the corrected acceptance operation data, analyze the acceptance evaluation results of the assigned results. If the operation results exceed the preset acceptance threshold, it means that the compressive strength of the concrete of the component meets the design requirements, and the corresponding assigned result is judged to be qualified for acceptance; otherwise, the acceptance is unqualified.

[0037] The attribution results reflect the component types where the hardened layer exists; the acceptance evaluation results include acceptance qualified and unqualified. In this embodiment of the invention, a multi-dimensional, full-process attribution determination and optimization mechanism is used. The cross-boundary correlation analysis is performed on the attribution confidence of each candidate path by combining the cross-component responsibility overlap ratio. A cross-boundary penalty coefficient is introduced to weaken the weight of cross-component paths. Then, a posterior evaluation index is constructed through normalization. The component evaluation value is obtained by grouping and aggregating according to the acceptance group. Based on this, the preliminary acceptance attribution label is accurately screened, the attribution group and the attribution undecided group are divided, and the component evaluation value is used as a deterministic score to support subsequent corrections. Simultaneously, through a neighborhood consistency propagation mechanism, the neighborhood set of unresolved groups is extracted, and weighted neighborhood consistency is calculated to achieve unresolved group attribution correction and attribution group conflict identification. Attribution groups with conflicting neighborhood spatial patterns are marked as high-conflict acceptance groups, and all acceptance management components corresponding to candidate paths within high-conflict groups are used as attribution labels. The attribution results are determined by combining the acceptance groups corresponding to candidate paths within the hardened layer, and responsibility teams are automatically assigned. Finally, acceptance qualification is judged based on the corrected acceptance operation data. Compared to traditional acceptance methods that rely solely on static matching relationships to determine attribution and are susceptible to cross-component ambiguity, this method... By using cross-boundary penalties and normalization, the interference weight of cross-component paths is automatically reduced, avoiding misjudgments of attribution caused by cross-boundary paths. At the same time, by utilizing neighborhood consistency propagation, the attribution correction of unresolved groups is made more in line with spatial distribution patterns, effectively solving the attribution problem of component boundary areas and complex defect areas. For example, for high-conflict acceptance groups at the junction of walls and columns, traditional methods often forcibly designate a single component as the attribution subject, which easily leads to the responsibility team shirking responsibility and inadequate maintenance and rectification. However, this invention uses all candidate components in the group as attribution labels, which not only clarifies the collaborative rectification responsibility of each responsible team, but also avoids the omission of responsibility caused by a single attribution.

[0038] Furthermore, this invention deeply integrates hardened layer detection data with attribution determination. It not only ensures the authenticity of acceptance evaluation results by correcting acceptance operation data, but also accurately matches the hardened layer attribution results with the responsible work teams, achieving a closed-loop linkage of defect attribution, responsibility allocation, and acceptance evaluation. This improves acceptance efficiency and the targeted nature of rectification. Simultaneously, the dual verification of deterministic scores and weighted neighborhood consistency ensures that the attribution results are both data-supported and spatially reasonable, effectively preventing the gradual amplification of initial attribution deviations. In addition, the differentiated handling of high-conflict acceptance groups preserves the integrity of multi-component attribution while providing a comprehensive basis for subsequent quality traceability and responsibility definition. This breaks the traditional either-or attribution limitations of acceptance, making the acceptance process more flexible and practical. It not only improves the accuracy and scientific nature of indoor concrete quality acceptance but also reduces engineering quality hazards and management costs caused by attribution errors and distorted evaluations.

[0039] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for indoor concrete quality objectification intelligent acceptance evaluation, characterized in that, The method includes: Obtain the original acceptance record set and management attribute data of indoor concrete, and combine them with the component type status to obtain the acceptance group for attribution analysis. The management attribute data includes component index, type, construction time, responsible team and acceptance standard. Based on the risk level of acceptance group analysis, a risk index for acceptance group status stratification is obtained, including: Based on the set of management components, the matching degree between each text record in the acceptance group and the acceptance management component is converted into an attribution contribution value. The attribution contribution values ​​of all text records are summed and averaged to obtain the attribution confidence vector of each acceptance group to each acceptance management component. Extract the maximum and second-largest values ​​of the attribution confidence within the attribution confidence vector, analyze the intensity of competition among multiple acceptance management components for the acceptance group, obtain the attribution difference, and obtain the cross-component responsibility overlap ratio by statistically analyzing the proportion of text records that simultaneously meet several acceptance management components within the acceptance group. The risk index for attribution acceptance is constructed by calculating the ratio between the cross-component responsibility overlap ratio and the attribution difference. If the risk index exceeds the preset affixing threshold, it is pre-marked as unstable attribution of responsibility; otherwise, it is pre-marked as stable attribution of responsibility. The acceptance group is hierarchically classified, and host components are allocated and candidate sets are screened according to the different status acceptance groups. Combined with the acceptance operation data during the acceptance process, several candidate paths are determined. Based on several candidate paths, the attribution label is identified and updated for the acceptance group to obtain the attribution results used for responsibility allocation, and the acceptance qualification is determined by the attribution results.

2. The method of claim 1, wherein, Obtain the original acceptance record set and management attribute data of indoor concrete, and combine them with the component type status to obtain the acceptance groups used for attribution analysis, including: Perform integrity verification on the original acceptance record set of indoor concrete to obtain a standardized dataset; the original acceptance record set should at least contain the defect type, location description, size information, collection time, construction batch and acceptance operation data of each text record; Based on a standardized dataset, the type status between components in each text record is analyzed to obtain an association set. For each text record, the attribute similarity between it and each text record in the association set is calculated using a cosine similarity algorithm. Text records whose attribute similarity variance exceeds a preset similarity threshold are identified as potential group records. The association set refers to the set of the same component and adjacent components corresponding to each text record. The acceptance records of indoor concrete are classified and managed according to the potential group records, resulting in acceptance groups.

3. The method of claim 2, wherein the method further comprises: Obtain the original acceptance record set and management attribute data of indoor concrete, and combine them with the component type status to obtain the acceptance group used for attribution analysis, which also includes: Read the management attribute data of concrete components, match each acceptance group with the management attribute data to obtain the matching degree, and filter out the components with the matching degree higher than the preset matching threshold as acceptance management components, thus obtaining the set of management components for each acceptance group.

4. The method of claim 3, wherein, The acceptance group is stratified by status, and host component allocation and candidate set screening are performed according to different status acceptance groups. Based on the acceptance operation data during the acceptance process, several candidate paths are determined, including: Based on the attribution confidence vector, the acceptance group with stable responsibility attribution is assigned the acceptance management component with the highest attribution confidence as the host; For acceptance groups with unstable responsibility attribution, acceptance management components with attribution confidence exceeding the average attribution confidence are selected and included in the candidate host set. Each acceptance management component in the candidate host set is combined with its corresponding acceptance group to construct several candidate paths.

5. The method of claim 4, wherein, The system performs hierarchical status stratification for acceptance groups, allocates host components and filters candidate sets based on different status acceptance groups, and determines several candidate paths by combining acceptance operation data during the acceptance process. This also includes: Based on the acceptance operation data within all candidate paths, analyze the local density of acceptance operations, and combine the changes in operation result values ​​to determine the hardening layer and correct the acceptance operation data within it. Based on the standardized dataset of the acceptance group, the host corresponding to the hardening layer and several candidate paths are recorded.

6. The method for intelligent acceptance and evaluation of indoor concrete quality according to claim 5, characterized in that, Based on several candidate paths, the attribution label is identified and updated for the acceptance group to obtain the attribution results used for responsibility allocation. The acceptance qualification is then determined based on the attribution results, including: Based on the cross-component responsibility overlap ratio, cross-boundary correlation analysis is performed on the attribution confidence of each candidate path to construct a post-evaluation index for acceptance judgment, and the index is grouped and aggregated according to the acceptance group to obtain the component evaluation value. Select the component with the highest component evaluation value that exceeds the preset evaluation threshold as the preliminary acceptance attribution label; otherwise, mark it as belonging to the undecided group. Record the acceptance group corresponding to the preliminary acceptance attribution label as the attribution group. Traverse the pending attribution group and the attribution group, analyze the impact of attribution propagation on the acceptance group, and update the labels of the pending attribution group and the attribution group, as well as the acceptance management components under the labels.

7. The method for intelligent acceptance and evaluation of indoor concrete quality according to claim 6, characterized in that, Based on several candidate paths, the attribution label is identified and updated for the acceptance group to obtain the attribution results used for responsibility allocation. The acceptance qualification is determined based on the attribution results. This also includes: Based on the updated pending group and the labels of the group and the acceptance management components under the labels, the high-conflict acceptance group is identified, and each acceptance management component in the high-conflict acceptance group is used as the acceptance attribution label. Based on the acceptance groups corresponding to several candidate paths within the hardened layer, determine the corresponding acceptance attribution labels and record the attribution results. Then, automatically assign the corresponding responsible work teams based on the attribution results. Read the operation results from the corrected acceptance operation data, analyze the acceptance evaluation results of the assigned results. If the operation results exceed the preset acceptance threshold, the corresponding assigned results are determined to be acceptable; otherwise, the acceptance is unacceptable.

Images