Building structure reinforcing method and monitoring system based on data identification

Abstract

The invention relates to the technical field of building engineering, in particular to a building structure reinforcing method and monitoring system based on data identification, and the method comprises the steps: obtaining the full life cycle data of a building structure; establishing a theoretical reference model; establishing a real-time simulation model; performing deviation analysis; reasoning through a pre-constructed causal reasoning model; the building structure is decomposed into a plurality of independent components; establishing a reinforcement measure library; generating an optimal reinforcement scheme; according to the full-life-cycle modeling method fusing original data, historical data and real-time data, the cumulative influence of each load event and maintenance event on the structure is accurately simulated according to the time sequence, and therefore a high-fidelity reference model capable of reflecting the theoretical state of the structure from completion to the current moment is constructed; according to the method, the accuracy and representativeness of the reference model are greatly improved, and a reliable comparison reference is provided for subsequent state recognition and diagnosis.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of building engineering, and in particular to a building structure reinforcement method and monitoring system based on data recognition. BACKGROUND

[0002] During its long service life, a building structure will inevitably be affected by material aging, environmental erosion, load changes, and even accidental events, and its safety performance will gradually degrade. Traditional structural reinforcement decisions are highly dependent on engineers' on-site investigation experience and qualitative theoretical analysis, and this process urgently needs to introduce more objective, accurate, and data-driven methods to improve its scientificity and efficiency.

[0003] Existing structure evaluation and reinforcement technologies have many limitations. First, the evaluation of the structure state often relies on discrete detection data and static calculation models, which are usually based on the ideal state of the structure at completion and cannot accurately reflect the cumulative damage and performance evolution of the structure after multiple load events and maintenance interventions during its long-term use. Second, defect diagnosis is heavily dependent on expert experience, and there is a lack of effective prediction and reasoning ability for defects hidden inside the structure that cannot be directly observed and the causal relationship between defects. Finally, when developing reinforcement schemes, it is usually based on "place mode" specification requirements or local experience, and it is difficult to comprehensively optimize from multiple dimensions such as overall structural safety, durability, and life cycle cost, which may result in insufficient or excessive reinforcement measures, causing economic waste. These shortcomings make the existing reinforcement methods face great challenges in terms of accuracy, predictability, and economy.

[0004] The present application aims to fundamentally solve the above problems and proposes a building structure reinforcement method and monitoring system based on data recognition. SUMMARY

[0005] In order to overcome the problems presented in the background art, the present application proposes a building structure reinforcement method and monitoring system based on data recognition.

[0006] The technical solution of the present application is: a building structure reinforcement method based on data recognition, comprising the following steps: S11: obtaining full life cycle data of the building structure, including original data, historical data, and real-time monitoring data; S12: based on the original data and the historical data, establishing a theoretical benchmark model reflecting the theoretical state of the building structure under the influence of historical loads and maintenance; S13: based on the theoretical benchmark model and the real-time monitoring data, establishing a real-time simulation model reflecting the real state of the building structure through model updating technology; S14: Comparing the real-time simulation model with the theoretical benchmark model, performing deviation analysis, and based on the deviation analysis result, performing root cause analysis to identify the explicit defects and hidden defects of the building structure; S15: Based on the identified defects, reasoning through the pre-constructed causal reasoning model to predict potential weak links and invisible defects in the building structure; S16: Decomposing the building structure into multiple independent components, and defining a performance attribute vector for each component, including stiffness, strength, toughness, and support; S17: Establishing a reinforcement measure library, each reinforcement measure in the library is quantified as an impact on the performance attribute vector of at least one component; S18: Taking the performance attribute vector of all components reaching the preset reinforcement target as the constraint condition, and taking the minimum total reinforcement cost as the optimization objective, combining and optimizing the measures in the reinforcement measure library to generate the optimal reinforcement scheme.

[0007] As a preferred, when establishing a theoretical benchmark model reflecting the theoretical state of the building structure under historical load and maintenance influence based on original data and historical data, specifically includes: S21: Based on the construction drawing information in the obtained original data, extracting the geometric information, material constitutive information and boundary conditions of the structure, establishing an initial finite element model, and the initial state of the initial finite element model corresponds to the theoretical state of the structure at completion; S22: Reading the load change record in the historical data, the load change record records the change events of permanent load, variable load and accidental load borne by the structure in time sequence; S23: Each load event in the load change record is applied to the initial finite element model as an independent load case according to its time sequence and duration, and the response of the structure under each load case is calculated, and the cumulative effect of the key components is recorded; S24: Reading the historical maintenance record in the historical data, the maintenance record describes the reinforcement and repair measures for specific parts of the structure; S25: Quantifying each maintenance event as a modification operation on the material properties, geometric properties and boundary conditions of the corresponding component in the model, and applying the modification operation to the processed finite element model in time sequence; S26: After sequentially applying and calculating all historical load events and maintenance events, the final theoretical benchmark model is obtained.

[0008] As a preferred, when establishing a real-time simulation model reflecting the real state of the building structure based on the theoretical benchmark model and real-time monitoring data through model updating technology, the following steps are included: S31: define a model parameter vector to be updated, the parameter vector including parameters representing the physical state of the structure, including the elastic modulus, cross-sectional area, moment of inertia and boundary condition coefficient of the key components; S32: define a structure response vector corresponding to the parameter vector, the structure response vector being composed of the calculated values of the theoretical reference model under the parameter vector, including the displacement, strain, acceleration frequency and mode shape of the key measuring points; S33: obtain a measured structure response vector corresponding to the structure response vector from real-time monitoring data; S34: construct an objective function to quantify the difference between the calculated response vector and the measured response vector; S35: minimize the objective function as the optimization goal, and repeatedly adjust the parameter vector through the iterative algorithm to make the calculated response continuously approach the measured response; S36: when one of the two conditions that the value of the objective function is less than the preset tolerance and the number of iterations reaches the upper limit occurs, stop iteration, and at this time, update the theoretical reference model with the finally determined parameter vector to obtain the real-time simulation model.

[0009] As preferred, when comparing the real-time simulation model with the theoretical reference model, performing deviation analysis, and based on the deviation analysis result, identifying the explicit defects and hidden defects of the building structure, the following steps are included: S41: define a set of key performance indicators for model comparison, the key performance indicators including global indicators and local indicators; wherein the global indicators include the first n order natural frequencies and main modes of the structure, and the local indicators include the displacement, stress, strain of the key nodes and the internal force of the main components; S42: extract the calculated values of the key performance indicators from the real-time simulation model and the theoretical reference model respectively to obtain the measured response vector and the theoretical response vector; S43: calculate the deviation degree of each key performance indicator to form a deviation vector; S44: compare the deviation vector with the preset multi-level threshold to identify significant deviations that exceed the allowed range; S45: based on the identified significant deviations, combined with the mechanical properties of the structure and the construction process historical data, perform root cause analysis to establish a mapping relationship from the deviation pattern to the defect type, and diagnose the type, location and severity of the defects; S46: output a defect diagnosis report, the report including a defect list, a defect location distribution map and corresponding root cause inference.

[0010] As preferred, when based on the identified defects, reasoning through the pre-constructed causal reasoning model to predict potential weak links and invisible defects in the building structure, the following steps are included: S51: Construct a causal knowledge graph of the building structure, and the causal knowledge graph takes the components as nodes and the mechanical interaction relationship between the components as directed edges; S52: Map the identified explicit defects and hidden defects to defect evidence of corresponding nodes in the causal knowledge graph, and assign an initial confidence to each defect evidence; S53: Define a causal rule of defect propagation based on the directed edge relationship in the causal knowledge graph, and the causal rule describes the influence probability of a node on the state of downstream nodes when a specific type of defect exists in the node; S54: Starting from the defect evidence, reasoning and calculation are performed along the directed edges of the causal knowledge graph, and the defect risk probability of the downstream nodes is updated step by step according to the causal rule; S55: Screen out the nodes whose defect risk probability exceeds a preset threshold, diagnose the corresponding components of the nodes as potential weak links and invisible defects, and evaluate the risk level thereof; S56: Output a potential defect prediction report, including the predicted weak link position, defect type, risk probability and reasoning path chain.

[0011] As preferred, when the building structure is decomposed into a plurality of independent components, and a performance attribute vector including stiffness, strength, toughness and support is defined for each component, it specifically includes: S61: Based on the BIM model of the building structure, the overall structure topology is decomposed into a plurality of independent component units, each component unit is assigned a unique identifier, and the component unit includes beams, plates, columns, walls and nodes; S62: Define a set of core performance attributes for each component unit, including stiffness attribute, strength attribute, toughness attribute and support attribute; S63: Based on the calculation results of the real-time simulation model, the initial values of the performance attributes of each component unit are quantified, and a performance attribute vector is constructed; S64: Based on the structure design specification and safety requirements, set the corresponding performance attribute target vector for each component unit.

[0012] As preferred, when the reinforcement measure library is established, and each reinforcement measure in the library is quantified as an influence amount on the performance attribute vector of at least one component, it specifically includes: S71: Establish a reinforcement measure knowledge base, and the knowledge base stores a plurality of standard reinforcement measures, each measure including a measure name, an applicable component type, a construction process and a cost parameter; S72: For each reinforcement measure in the knowledge base, define its influence domain, and the influence domain includes direct action components and indirect influence components; S73: Establish a performance influence quantification model for each reinforcement measure, and map the implementation amount of the measure to the change amount of the performance attribute vector of the components in the influence domain; S74: Store the quantified impact in matrix form, build a reinforcement measure-attribute impact matrix library for subsequent optimization calculation.

[0013] As a preferred, the performance impact quantification model is represented by the following formula: ; Wherein, is the change of performance attribute vector, is the reinforcement measure type coefficient matrix, is the measure application amount matrix, is the component-measure coupling coefficient matrix; wherein, the measure application amount matrix is calculated by the following formula: For the reinforcement measures of the pasting type: ; For the reinforcement measures of the section increase type: ; Wherein, is the elastic modulus of the reinforcement material, is the cross-sectional area of the reinforcement material, is the length or coverage of the reinforcement material, is the cross-sectional area of the reinforced component, is the cross-sectional area of the original component.

[0014] As a preferred, in generating the optimal reinforcement scheme, specifically includes: S81: Define the decision variable, which is a binary variable, indicating whether to adopt the jth measure in the reinforcement measure library; S82: Build an optimization objective function, aiming to minimize the total cost of all adopted reinforcement measures; S83: Build the constraint condition, requiring the final performance attribute vector of each component unit after reinforcement to be no lower than its preset target vector; S84: Formulate the reinforcement scheme optimization problem into a binary integer programming problem with constraints; S85: Use optimization algorithms to solve the programming problem to obtain the optimal combination of reinforcement measures; S86: Convert the solution into an executable reinforcement scheme, including the measure list, construction sequence and cost budget.

[0015] The monitoring system for building structure reinforcement based on data recognition includes: A data acquisition module for acquiring whole life cycle data of building structures; A model management module for building and managing theoretical benchmark models and real-time simulation models; An intelligent diagnosis module for performing model comparison, deviation analysis and root cause analysis; An optimization solving module is used for performing component attribute quantification, measure influence quantification and combined optimization solving; A scheme output module is used for generating and visualizing the optimal reinforcement scheme.

[0016] The present application has the following advantages: 1. Compared with the prior art which usually only relies on real-time monitoring data or a simple initial design model to evaluate the structure state, this method cannot reflect the performance evolution of the structure caused by historical loads and maintenance during long-term use, and the evaluation result is biased. The present application innovatively proposes a whole life cycle modeling method that fuses original data, historical data and real-time data, accurately simulates the cumulative influence of each load event and maintenance event on the structure in time sequence, and thus builds a high-fidelity benchmark model that can reflect the theoretical state of the structure from completion to the current time. This method greatly improves the accuracy and representativeness of the benchmark model, and provides a reliable comparison benchmark for subsequent state identification and diagnosis. 2. Compared with the prior art in which the structure analysis model is usually static and unchangeable, it is difficult to truly reflect the damage and degradation of the structure over time, resulting in a simulation result that is out of touch with the actual situation. The present application uses a dynamic model updating technology based on sensitivity analysis and optimization algorithm, iteratively matches the real-time monitoring data with the theoretical benchmark model, automatically corrects the physical parameters in the model, and enables the simulation model to continuously approach the real state of the structure. This scheme realizes online self-calibration of the structure digital twin model, ensures the consistency of the simulation model and the physical entity, and provides a dynamic and real virtual experimental environment for accurate diagnosis. 3. Compared with the prior art which relies on artificial experience to compare limited data points to make defect judgments, the present application defines a multi-level key performance index and threshold system, automatically compares the output of the real-time simulation model and the theoretical benchmark model, quantitatively calculates the deviation of each index, and intelligently associates specific deviation patterns with potential defect types based on the Bayesian inference probability framework. This scheme realizes the leap from "phenomenon perception" to "root cause diagnosis", can systematically identify both explicit defects and hidden defects, and gives a quantitative diagnosis conclusion, significantly improving the objectivity and depth of diagnosis. 4. Compared with the prior art which mainly focuses on discovered defects and lacks the ability to predict defect chain reaction and potential risks, the present application cannot prevent problems from happening. The present application builds a causal knowledge graph with components as nodes and mechanical relationships as edges, and designs a probability reasoning algorithm for defect propagation along the network, which can simulate and deduce the influence of identified defects on associated components. This method realizes the risk probability prediction of potential weak links and invisible defects in the structure that have not been observed, advances the reinforcement decision from "after-the-fact remedy" to "pre-incident warning", and enhances the predictability of structure safety management. 5. Compared to existing technologies that often only assess structural performance at the overall level or with qualitative descriptions, making it difficult to support precise reinforcement design, this invention decomposes the overall structure into independent components and defines a multi-dimensional performance attribute vector for each component, including stiffness, strength, and toughness. This transforms the complex structural performance state into a quantifiable mathematical vector. This method provides precise, component-level state descriptions and target constraints for subsequent reinforcement optimization, enabling targeted reinforcement design and moving from extensive reinforcement to precise enhancement. 6. Compared to existing technologies where the selection of reinforcement measures relies heavily on engineers' experience and qualitative judgment, making it difficult to accurately assess the improvement effect of a single or combined measure on the overall structural performance, this invention establishes a standardized knowledge base of reinforcement measures and quantifies the implementation effect of each measure into an influence matrix on the performance attribute vector of a specific component. Simultaneously, it considers the indirect effects caused by the measures through finite element analysis. This approach transforms fuzzy engineering experience into precise, calculable influence quantities, laying a solid foundation for the scientific comparison and optimization of various reinforcement measures. 7. Compared to existing technologies that often employ trial calculations or empirical methods when formulating reinforcement schemes, making it difficult to achieve an optimal balance between safety and economic cost, this invention formalizes the formulation of reinforcement schemes into a mathematical programming problem with the goal of minimizing total cost and the constraint that all components meet performance standards. It then applies an efficient integer programming algorithm to solve the problem. This method can automatically and quickly search for the most cost-effective reinforcement scheme from a vast array of measures, achieving a decision-making upgrade from "meeting safety" to "meeting safety optimally and economically," significantly improving resource utilization efficiency. Attached Figure Description

[0017] Figure 1 The diagram shows a flowchart of the data recognition-based building structure reinforcement method of the present invention. Figure 2 The diagram shown is a schematic representation of the construction of the monitoring system for building structure reinforcement based on data recognition according to the present invention. Detailed Implementation

[0018] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0019] Please see Figure 1 The present invention provides an embodiment of a building structure reinforcement method based on data identification, comprising the following steps: Step 1: Acquire full lifecycle data of the building structure, including raw data, historical data, and real-time monitoring data. The full lifecycle data includes: The raw data includes construction drawings, BIM models, and construction process records; The historical data includes historical maintenance records and load change records. The real-time monitoring data includes sensor monitoring data and manual detection data.

[0020] Step two: Based on the original data and the historical data, a theoretical benchmark model reflecting the theoretical state of the building structure under the influence of historical loads and maintenance is established, specifically including: S21: Based on the construction drawing information in the obtained original data, the geometric information, material constitutive information and boundary conditions of the structure are extracted, and an initial finite element model is established, and the initial state of the initial finite element model corresponds to the theoretical state of the structure at completion; S22: Read the load change record in the historical data, and the load change record records the change events of the permanent load, variable load and accidental load borne by the structure in time sequence; S23: Each load event in the load change record is applied to the initial finite element model as an independent load case according to its time sequence and duration, and the response of the structure under each load case is calculated, and the cumulative effect of the key component is recorded; S24: Read the historical maintenance record in the historical data, and the maintenance record describes the reinforcement and repair measures for specific parts of the structure; S25: Each maintenance event is quantified as a modification operation on the material properties, geometric properties and boundary conditions of the corresponding component in the model, and the modification operation is applied to the processed finite element model in time sequence; S26: After the sequential application and calculation of all historical load events and maintenance events, the final theoretical benchmark model is obtained.

[0021] In this embodiment, when calculating the cumulative effect of the key component, the linear accumulation of the load effect is performed by the following formula: ; Wherein, represents the cumulative effect of a component, represents the size of the i-th load event, represents the duration of the i-th load event, represents the repetition frequency factor of the i-th load event.

[0022] In this embodiment, when quantifying the maintenance event as a modification operation, specifically including: When the maintenance record is an increase in the cross section of the component, the modification operation corresponds to updating the cross-sectional geometric properties of the component in the model; When the repair record is a pasted steel plate or carbon fiber cloth, the modification operation corresponds to updating the material stiffness matrix of the component in the model, and the equivalent stiffness after repair is calculated by the following key formula: ; wherein, is the equivalent stiffness of the component after repair, is the stiffness of the original component before repair, is the elastic modulus of the reinforcing material, is the cross-sectional area of the reinforcing material, is the geometric size of the original component.

[0023] Specifically, after each important load event or repair event is applied, the state of the current finite element model and its calculation results are saved to form a model state sequence, which is used to trace the historical path of the evolution of the structure state.

[0024] In the embodiment, the application first establishes an initial finite element model reflecting the theoretical state of the structure at completion according to original data such as construction drawings; then, historical load records are read in chronological order, each load event is applied to the model as an independent working condition for static and dynamic analysis, and the cumulative effect of key components is calculated using a linear cumulative formula; at the same time, historical repair records are read, each repair event is quantified as an accurate modification to the properties of the model components (such as the stiffness after reinforcement calculated by the equivalent stiffness formula), and the model is updated synchronously according to the order of occurrence; finally, after all historical loads and repair interventions are sequentially reproduced, a high-precision reference model reflecting the theoretical state of the structure under historical influence is obtained. The beneficial effects of this scheme are that it breaks through the limitations of traditional static models that cannot reflect the time-varying evolution of structure performance, and through sequential simulation of the load history and repair intervention, the reference model more truly reflects the actual "age" and "experience" of the structure, providing a reliable reference for subsequent accurate comparison with the real-time state, and the model state sequence formed provides a complete data chain for structure state evolution tracing and defect diagnosis.

[0025] Step three: Based on the theoretical reference model and real-time monitoring data, a real-time simulation model reflecting the actual state of the building structure is established through model updating technology, including the following steps: S31: define the model parameter vector to be updated, the parameter vector including parameters representing the physical state of the structure, including the elastic modulus, cross-sectional area, moment of inertia and boundary condition coefficient of the key component; S32: define the structure response vector corresponding to the parameter vector, the structure response vector being composed of the calculated values of the theoretical reference model under the parameter vector, including the displacement, strain, acceleration frequency and mode shape of the key measuring point; S33: obtaining a measured structural response vector corresponding to the structural response vector from the real-time monitoring data; S34: constructing a target function to quantify the difference between the calculated response vector and the measured response vector; S35: taking the minimization of the target function as the optimization objective, repeatedly adjusting the parameter vector through an iterative algorithm to make the calculated response continuously approach the measured response; S36: when one of the following two conditions occurs, stopping iteration, and updating the theoretical reference model with the finally determined parameter vector to obtain a real-time simulation model.

[0026] In the embodiment, the constructed target function adopts a weighted least squares form, and the key formula is: ; wherein, is the target function, is a diagonal weight matrix, the diagonal elements of which are valued according to the reliability and importance of the sensor data of each measuring point, is the measured response vector, is the calculated response vector, represents a matrix transposition operation.

[0027] In the embodiment, the adopted iterative algorithm is an optimization algorithm based on sensitivity analysis, and the iterative formula is: ; ; wherein, is the model parameter vector at the k+1th iteration, is the model parameter vector at the kth iteration, is the sensitivity matrix at the kth iteration, is a regularization factor, is a unit matrix, represents a matrix transposition operation.

[0028] Specifically, when defining the model parameter vector to be updated, the definition of the model parameter vector to be updated specifically includes two ways of parameter type model updating and matrix type model updating, specifically: Parameter type model updating: directly associating the parameter vector with the physical properties in the finite element model, and the parameters have clear physical meaning; Matrix type model updating: associating the parameter vector with the correction amount of the overall stiffness matrix or mass matrix, and directly correcting the system matrix through the following formula: ; wherein, is the updated global stiffness matrix, is the original global stiffness matrix, is the pre-selected base matrix, is the correction coefficient to be updated.

[0029] Specifically, in the iteration process, a parameter constraint step is further included: The parameter vector updated in each iteration is subjected to physical constraints to ensure that its value is within a reasonable physical range, i.e., satisfies: ; wherein, is a parameter lower limit vector determined according to material properties and design requirements, is a parameter upper limit vector determined according to material properties and design requirements.

[0030] In the present embodiment, the present application establishes a real-time simulation model through a dynamic model updating technique based on parameter optimization. Specifically, first, a set of updatable parameters representing the physical state of the structure and the corresponding structural responses are defined, then a target function is constructed in the form of weighted least squares of the difference between the calculated response of the theoretical benchmark model and the measured response of the real-time monitoring data, and subsequently, an iterative algorithm based on sensitivity analysis is used to automatically and repeatedly adjust the model parameters under the premise of considering the physical constraints of the parameters, until the calculated response infinitely approximates the measured data, and finally a real-time simulation model highly consistent with the true state of the structure is obtained. The beneficial effects of this scheme lie in that it successfully integrates the static theoretical model with the dynamic measured data, continuously tracks the real performance degradation of the physical structure through mathematical optimization means, and solves the core problem of the disconnection between the traditional model and the actual structure state, providing the most real and reliable digital twin body basis for subsequent precise deviation analysis, defect diagnosis and reinforcement decision.

[0031] Step four: The real-time simulation model is compared with the theoretical benchmark model to perform deviation analysis, and based on the deviation analysis results, root cause analysis is performed to identify the explicit defects and hidden defects of the building structure, including the following steps: S41: defining a set of key performance indicators for model comparison, the key performance indicators including global indicators and local indicators; wherein the global indicators include the first n order natural frequencies and main modes of vibration of the structure, and the local indicators include the displacements, stresses, strains of key nodes and internal forces of main members; S42: extracting the calculated values of the key performance indicators from the real-time simulation model and the theoretical benchmark model respectively to obtain a measured response vector and a theoretical response vector; S43: calculating the deviation degree of each key performance indicator to form a deviation vector; S44: Compare the deviation vector with the preset multi-level threshold value, and identify the significant deviation exceeding the allowed range; S45: Based on the identified significant deviation, combined with the mechanical properties of the structure and the construction process historical data, the root cause analysis is carried out, the mapping relationship from the deviation mode to the defect type is established, and the type, position and severity of the defect are diagnosed; S46: Output the defect diagnosis report, which includes the defect list, defect position distribution map and corresponding root cause inference.

[0032] In this embodiment, the deviation degree is calculated by using the following key formula: ; Or: ; Wherein, represents the percentage deviation degree of the i th key performance indicator, is the i th indicator value extracted from the real-time simulation model, is the i th indicator value extracted from the theoretical benchmark model, is the mode confidence factor, is the i th mode vector extracted from the real-time simulation model, is the i th mode vector extracted from the theoretical benchmark model, represents the matrix transposition operation.

[0033] In this embodiment, the preset multi-level threshold value includes attention threshold value, warning threshold value and alarm threshold value, and the judgment logic is: If < attention threshold value, the index deviation is considered as normal fluctuation and is not marked; If attention threshold value≤ ≤ warning threshold value, the index is marked as slight deviation and needs to be observed continuously; If warning threshold value≤ ≤ alarm threshold value, the index is marked as significant deviation and needs to be analyzed for root cause; If ≥ alarm threshold value, the index is marked as serious deviation and needs to start detailed evaluation and reinforcement process immediately.

[0034] Specifically, the root cause analysis is realized by deviation mode recognition and knowledge base matching, including: Stiffness defect diagnosis: if the overall natural frequency of the structure is generally reduced and the mid-span displacement is increased, the root cause is inferred as stiffness degradation of the overall structure or main load-bearing member, which may be caused by material aging and micro-crack development; Damage location diagnosis: if the MAC value of a specific mode is significantly reduced, the local damage near the node of the mode can be located according to the node displacement pattern of the mode; Process defect diagnosis: if the angle or relative displacement deviation of a certain connection node is abnormally large, and the construction history record shows that the node is a construction joint or post-cast strip, the root cause is inferred to be a connection process defect; Hidden defect prediction: if the stress level of a certain component is much lower than the theoretical value, it is inferred that there may be a change in the load transfer path, and the adjacent component may have hidden stiffness enhancement or weakening.

[0035] Specifically, the root cause analysis adopts a probabilistic diagnosis framework based on Bayesian inference, and the key formula is: ; Wherein, The posterior probability of the defect type exists under the condition of observing the deviation mode Δ, The likelihood probability of the deviation mode Δ appearing when the defect exists, which is obtained from the historical case database or finite element parameter analysis, The prior probability of the defect exists, which can be determined according to the structure type, age and construction record.

[0036] In this embodiment, the present application identifies structural defects through systematic model comparison and intelligent diagnosis mechanism. This method first defines global and local key performance indicators, respectively extracts data from real-time simulation models and theoretical benchmark models to quantitatively calculate the percentage deviation or mode confidence factor of each indicator, and then compares the deviation vector with the preset multi-level threshold to automatically identify significant deviations; On this basis, combined with mechanical properties and historical data, a probabilistic diagnosis framework based on Bayesian inference is used to map specific deviation patterns to potential defect types, thereby systematically diagnosing explicit and hidden defects including stiffness degradation, local damage, process defects, etc. and generating a detailed diagnosis report. The beneficial effects of this scheme are that it realizes the intelligent conversion from numerical difference to engineering significance, improves the fuzzy experience judgment to objective diagnosis based on probability, not only can accurately locate the defects, but also can reveal the root cause, and can speculate the hidden risks, greatly improving the depth, accuracy and forward-looking of the structure state evaluation, providing a scientific basis for formulating targeted reinforcement scheme.

[0037] Step five: Based on the identified defects, the potential weak links and invisible defects in the building structure are predicted through the pre-constructed causal reasoning model, including the following steps: S51: Construct a causal knowledge graph of the building structure, the causal knowledge graph taking a component as a node and a mechanical interaction relationship between components as a directed edge; S52: Map the identified explicit defects and hidden defects to defect evidence of corresponding nodes in the causal knowledge graph, and give each defect evidence an initial confidence degree; S53: Define a causal rule of defect propagation based on the directed edge relationship in the causal knowledge graph, the causal rule describing the influence probability of a node on a downstream node state when the node has a specific type of defect; S54: Take the defect evidence as a starting point, perform reasoning calculation along the directed edges of the causal knowledge graph, and update the defect risk probability of the downstream nodes according to the causal rule; S55: Screen out nodes whose defect risk probability exceeds a preset threshold, diagnose the corresponding components of the nodes as potential weak links and invisible defects, and evaluate the risk levels of the nodes; S56: Output a potential defect prediction report, including predicted weak link positions, defect types, risk probabilities and reasoning path chains.

[0038] In the embodiment, the construction of the causal knowledge graph specifically includes: a node set, wherein each node represents a structural component, and node attributes include component type, material, geometric attribute and element group thereof in a finite element model; a directed edge set, wherein a directed edge is from a node to a node , indicating that the mechanical state change of component has an influence on component ; Each directed edge is associated with an influence weight matrix, which quantifies the influence degree of various defect types of node on various state attributes of node .

[0039] In the embodiment, the reasoning calculation adopts a causal reasoning algorithm based on a probability soft logic, and a specific formula is as follows: ; wherein, is the risk probability of node having a potential defect, is a set of all parent nodes of node , is the probability of the parent node having been observed to have a defect, is an influence intensity factor from node to node .

[0040] Specifically, the influence intensity factor is calculated by the following formula: ; wherein, is the internal force redistribution amount of the node due to the stiffness damage of the node , which is obtained by finite element analysis under a unit damage working condition, is the nominal internal force of the node under the normal use state, is the modal confidence factor of the node related part in the change of the overall structure mode before and after the damage, and and are weight coefficients.

[0041] Specifically, the risk level evaluation is comprehensively determined according to the risk probability and the importance coefficient of the component , and the specific formula is: ; wherein, is the risk level, and the importance coefficient of the component is determined according to the key degree of the force transmission path, the redundancy and the failure consequence severity of the component in the structure system.

[0042] In the embodiment, the present application predicts potential defects by constructing a causal knowledge graph and implementing probabilistic reasoning. The method first abstracts the building structure into a causal knowledge graph with components as nodes and mechanical relationships as directed edges, and maps the identified defects to the corresponding nodes and gives confidence; then, based on the pre-defined defect propagation causal rules, the probabilistic soft logic algorithm is used to reason and calculate along the directed edges of the graph, quantifying the influence of upstream node defects on downstream nodes, thereby updating the defect risk probability of all nodes in the network step by step; finally, the components with high importance and risk probability exceeding the threshold are diagnosed as potential weak links. The beneficial effects of this scheme lie in that it upgrades the structure analysis from the identification of isolated defects to the systematic prediction of defect transmission chains, revealing secondary defects and hidden risks that may be triggered by mechanical interaction, realizing the leap from "seeing the present" to "seeing the future", providing key decision-making basis for formulating forward-looking and preventive reinforcement strategies, and significantly improving the initiative and globality of structure safety management.

[0043] Step six: The building structure is decomposed into a plurality of independent components, and a performance attribute vector including stiffness, strength, toughness and support is defined for each component, specifically including: S61: Based on the BIM model of the building structure, the overall structure topology is decomposed into a plurality of independent component units, each component unit is assigned a unique identifier, and the component unit includes beams, plates, columns, walls and nodes; S62: A set of core performance attributes is defined for each component unit, including stiffness attribute, strength attribute, toughness attribute and support attribute; S63: Based on the calculation results of the real-time simulation model, the performance attribute initial value of each component unit is quantified, and a performance attribute vector is constructed; S64: Based on the structure design specification and safety requirements, a corresponding performance attribute target vector is set for each component unit.

[0044] Specifically, the toughness attribute value The following key formula is used for calculation: ; Wherein, is the energy dissipation value of the component under cyclic load, is the elastic strain energy of the component, is the maximum allowable deformation angle of the component, is the yield deformation angle of the component.

[0045] In this embodiment, the performance state of the building structure is accurately evaluated by quantitatively describing the component level. This method first systematically decomposes the overall structure into independent component units such as beams, plates, columns, walls and nodes based on the BIM model, and defines a set of core performance attributes including stiffness, strength, toughness and support for each component; Then, by using the calculation results of the real-time simulation model, the performance attribute initial value of each component in the current state is quantified, especially the toughness attribute is calculated by combining the energy dissipation and deformation capacity formula, so as to construct a quantifiable performance attribute vector; Finally, according to the design specification, the performance attribute target vector that each component needs to reach is set. The beneficial effects of this scheme are that it converts the complex and overall structure performance evaluation into accurate, component-level, multi-dimensional vector description, provides clear and measurable targets and mathematical basis for subsequent reinforcement optimization, realizes the leap from fuzzy qualitative judgment to precise quantitative management, and significantly improves the scientificity and pertinence of reinforcement design.

[0046] Step seven: A reinforcement measure library is established, and each reinforcement measure in the library is quantified as an influence amount on the performance attribute vector of at least one component, specifically including: S71: Establish a reinforcement measure knowledge base, and the knowledge base stores a plurality of standard reinforcement measures, each measure including measure name, applicable component type, construction process and cost parameter; S72: define the influence domain of each reinforcement measure in the knowledge base, the influence domain including direct-acting components and indirectly-affected components; S73: establish a performance influence quantification model for each reinforcement measure, mapping the implementation amount of the measure to the change amount of the performance attribute vector of the components in the influence domain; S74: store the quantified influence amount in matrix form, and build a reinforcement measure-attribute influence matrix library for subsequent optimization calculation.

[0047] In this embodiment, the performance influence quantification model is represented by the following formula: ; wherein, is the change amount of the performance attribute vector, is a reinforcement measure type coefficient matrix, is a measure application amount matrix, is a component-measure coupling coefficient matrix; wherein the measure application amount matrix is calculated by the following formula: For the reinforcement measure of the adhesive type: ; For the reinforcement measure of the section increase type: ; wherein, is the elastic modulus of the reinforcement material, is the cross-sectional area of the reinforcement material, is the length or coverage of the reinforcement material, is the cross-sectional area of the reinforced component, is the cross-sectional area of the original component.

[0048] Specifically, the determination method of the indirectly-affected components includes: Based on the real-time simulation model, after virtually applying a certain reinforcement measure, the internal force redistribution of the structure is calculated through finite element analysis, and the components whose internal force changes exceed a preset threshold are determined as indirectly-affected components.

[0049] In the embodiment, the application systematically manages the effects of various reinforcement schemes by establishing a quantified reinforcement measure knowledge base. The method first constructs a knowledge base containing various standard reinforcement measures and their process and cost parameters, and defines the direct influence and indirect influence component range of each measure through internal force redistribution analysis; then establishes a performance influence quantification model for each measure, converts the implementation amount of the measure into a specific change in the performance attribute vector of the component through a mathematical model, and stores all the influence quantities in matrix form to form a complete measure-attribute influence matrix library. The beneficial effect of the scheme is that it changes the traditional qualitative reinforcement selection process that relies on experience into a quantitative decision support system that can accurately calculate and predict the overall effect, taking into account not only the direct influence of the measure but also the indirect effect on the overall structure, thereby providing a reliable data basis for subsequent global optimal reinforcement scheme combination optimization, significantly improving the scientificity and economy of reinforcement design.

[0050] Step eight: With all the performance attribute vectors of the components reaching the preset reinforcement target as the constraint condition and the total cost of reinforcement being the lowest as the optimization target, the measures in the reinforcement measure library are combined and optimized to generate an optimal reinforcement scheme, which specifically includes: S81: Define the decision variable, which is a binary variable, representing whether to adopt the jth measure in the reinforcement measure library; S82: Build the optimization objective function, with the goal of minimizing the total cost of all adopted reinforcement measures; S83: Build the constraint condition, requiring the final performance attribute vector of each component unit after reinforcement to be no lower than its preset target vector; S84: Formalize the reinforcement scheme optimization problem into a binary integer programming problem with constraints; S85: Use an optimization algorithm to solve the programming problem and obtain the optimal reinforcement measure combination; S86: Convert the solution into an executable reinforcement scheme, including the measure list, construction sequence, and cost budget.

[0051] In the embodiment, the optimization objective function is represented by the following formula: ; Wherein, is the total cost objective function, is the unit cost of the jth reinforcement measure, is the decision variable of the jth reinforcement measure, is the total number of measures in the reinforcement measure library.

[0052] Specifically, the constraint condition is represented by the following formula: ; ; wherein, is the initial performance attribute vector of the i-th component unit, is the impact of the j-th reinforcement measure on the performance attribute vector of the i-th component unit, is the performance attribute target vector of the i-th component unit, is the total number of component units.

[0053] In this embodiment, when the optimization algorithm is used to solve the planning problem and obtain the optimal combination of reinforcement measures, the optimization algorithm used is the branch and bound algorithm, which is solved by the following steps: Step a: relax the integer constraint, convert the original problem into a linear programming problem for solving, and obtain the lower bound; Step b: if the relaxed solution satisfies the integer constraint, the optimal solution is obtained; otherwise, select a fractional variable for branching; Step c: solve the linear programming relaxation at the branching node, and eliminate the branches that are not possible to produce better solutions through pruning strategies; Step d: repeat the branching, bounding and pruning process until the optimal solution or the approximate optimal solution that satisfies the integer constraint is found.

[0054] Specifically, in the branch and bound algorithm, the cutting plane method is used to strengthen the relaxation, and the following effective inequalities are added to tighten the relaxation gap: ; wherein, and are subsets of variables determined according to the characteristics of the problem and the known solution structure.

[0055] In this embodiment, the present application generates the optimal reinforcement scheme through the mathematical programming method, formulates the scheme selection problem into a binary integer programming problem with the minimum total cost as the optimization objective and all component performance meeting the standard as the constraint condition, represents whether the measure is adopted by defining binary decision variables, constructs a mathematical model including the cost function and performance constraints, and then efficiently solves the combination optimization problem by using the branch and bound algorithm combined with the cutting plane method. Finally, the solution result is converted into an executable scheme including a specific measure list, construction sequence and cost budget. The beneficial effects of the scheme are that it realizes the leap from the qualitative scheme selection depending on the experience of engineers to the quantitative automatic generation based on mathematical optimization, can quickly and scientifically screen out the globally optimal solution with the highest economy under the premise of strictly meeting all safety performance requirements from the vast number of possible measure combinations, and thus significantly improves the scientificity, economy and decision-making efficiency of the reinforcement scheme development.

[0056] As Figure 2As shown, the embodiment also provides a monitoring system for building structure reinforcement based on data recognition, comprising: a data acquisition module for acquiring full life cycle data of the building structure; a model management module for constructing and managing theoretical benchmark models and real-time simulation models; an intelligent diagnosis module for performing model comparison, deviation analysis and root cause analysis; an optimization solving module for performing component attribute quantification, measure impact quantification and combined optimization solving; a scheme output module for generating and visualizing the optimal reinforcement scheme.

[0057] Embodiment one: Structure reinforcement of a large public building (stadium) A stadium has been in use for twenty years. Due to changes in its use function (such as hosting larger events, adding large suspended display screens) and long-term bearing of crowd loads and environmental effects, the management unit hopes to conduct a comprehensive safety assessment and targeted reinforcement of the main structure.

[0058] Data acquisition and establishment of theoretical benchmark model The technical personnel first collected the full set of construction drawings, original BIM model and construction records of the stadium at completion as original data. At the same time, historical data was reviewed, including maintenance records (such as reinforcement of some steel members of the roof, repair of the stand board) and important load change records (such as installation of large equipment, snow load records, etc.) saved in the archives. Subsequently, a sensor network was laid out at key parts of the structure (such as main truss supports, key rod members, main column feet, etc.) to obtain real-time monitoring data, including strain, displacement and vibration frequency.

[0059] Based on the construction drawings, an initial finite element model reflecting the ideal state of the stadium at completion was established. Then, in chronological order, each major load event (such as a rare snowstorm) and each maintenance event in the historical records were sequentially applied to the model. For example, for a maintenance record of pasting carbon fiber cloth, the stiffness properties of the corresponding components in the model were improved. This process ultimately generates a theoretical benchmark model that can reflect what "state should the structure be in" after experiencing all historical events.

[0060] Real-time simulation model updating and deviation analysis The calculation results of the theoretical benchmark model (such as the displacement of key points, the natural frequency of the structure) are compared with the real-time monitoring data returned by the sensors. Through an iterative optimization algorithm, the physical parameters (such as the elastic modulus of some components) in the theoretical benchmark model are automatically adjusted so that the output of the model and the measured data are as consistent as possible, thereby obtaining a real-time simulation model that accurately reflects the true current state of the structure.

[0061] The real-time simulation model is systematically compared with the theoretical benchmark model. Through the calculation, it is found that the displacement values of the main truss in the roof part exceed the early warning threshold of the theoretical values, and the main vibration frequency of the structure is reduced. Through deviation analysis and root cause analysis (combined with historical maintenance records and mechanical principles), the system diagnoses that the main load-bearing truss of the roof has a dominant defect of overall stiffness degradation, which may be caused by the development of microscopic fatigue cracks in the steel node area.

[0062] Causal reasoning and potential defect prediction The system pre-builds a causal knowledge graph of the stadium structure, where the nodes are beams, columns, trusses, etc. components, and the edges represent the mechanical support relationship between them. The identified defect of "main truss stiffness degradation" is input into the graph as evidence.

[0063] The system reasons according to the pre-set rules: the insufficient stiffness of the main truss will cause it to transfer more load to the supporting columns and surrounding components connected to it. Through calculation, the system predicts that several adjacent secondary subordinate components have a high risk of potential damage due to internal force redistribution, although the current monitoring data does not directly show abnormalities in these components.

[0064] Component attribute quantification and reinforcement measure optimization The system decomposes the entire stadium structure into thousands of independent components and defines a performance attribute vector for each component, including stiffness, strength, toughness, etc. It also calculates the current attribute values of each component based on the real-time simulation model. At the same time, according to the new use load and safety standards, it sets the target performance attribute vector that each component needs to achieve.

[0065] The system's built-in reinforcement measure library includes measures such as "pasting carbon fiber cloth", "external steel reinforcement", "increasing section", etc. Each measure has been quantified as the impact on the performance attribute vector of a specific component (for example, "pasting carbon fiber cloth" will significantly improve the stiffness attribute of the component).

[0066] Finally, the system takes the minimum total reinforcement cost as the optimization goal, and takes the performance attributes of all components reaching the target value as the constraint condition, searches for all feasible combinations in the measure library to solve. Finally, it generates an optimal reinforcement scheme: it suggests using a specific external steel reinforcement method for the main truss, and performing preventive carbon fiber cloth pasting on several predicted high-risk potential weak links, and provides detailed construction sequence and budget. The scheme saves about 15% of the cost compared to the traditional experience scheme under the premise of ensuring the safety of the structure.

[0067] Example Two: Structural Assessment and Preventive Reinforcement of a Large-span Bridge in Service A prestressed concrete continuous girder bridge with heavy traffic was found to have slight changes in the bridge deck alignment and cracks in some areas during regular inspection. The management department hopes to conduct accurate assessment and preventive reinforcement without interrupting traffic.

[0068] Data acquisition and theoretical benchmark model establishment Collect original data: design drawings, material strength reports. Historical data: traffic flow records over the years (reflecting load changes), overload vehicle records, and previous crack sealing and repair records. Real-time monitoring data: real-time acquisition of bridge responses under vehicle load and temperature changes through installed GPS displacement monitoring points, strain gauges, acceleration sensors, etc.

[0069] After establishing the initial finite element model, apply historical traffic load spectrum and temperature load in chronological order, and simulate the impact of each repair (such as grouting for a crack) on the model to build a theoretical benchmark model considering historical operating conditions.

[0070] Real-time simulation model updating and deviation analysis Use the monitoring data of the current season (such as mid-span displacement, first-order frequency) to update the theoretical benchmark model to obtain a real-time simulation model. Comparison shows that the deflection of the mid-span has exceeded the theoretical value, and the vibration mode of the pier has a slight but critical difference from the theoretical model.

[0071] Deviation analysis indicates that the mid-span deflection exceeding the limit is a significant deviation. Root cause analysis combined with causal reasoning diagnoses that this is not only due to the loss of prestress of the prestressed steel tendon (explicit defect), but also infers that there may be hidden defects such as uneven settlement of the pier foundation, which changes the support conditions of the structure, thereby exacerbating the internal force redistribution of the main girder.

[0072] Causal reasoning and potential defect prediction In the causal knowledge graph of the bridge, "uneven settlement of the pier" as the parent node defect will affect the "moment distribution of the main girder" and "bridge deck alignment" and other child nodes. The system predicts that if the settlement continues to develop, the moment of some areas of the main girder will significantly increase, posing a potential risk of new cracks.

[0073] Component attribute quantification and reinforcement measure optimization The bridge is decomposed into components such as main girder segments, piers, bearings, etc., and their current performance attributes are quantified. The target attributes are set according to the current specifications and safety operation requirements.

[0074] The reinforcement measure library contains measures such as "external prestress reinforcement", "pier foundation grouting reinforcement", and "additional bearings" tailored to the characteristics of the bridge. The improvement amount of each measure on the component performance attributes is quantified.

[0075] The system performs a combined optimization. The optimization goal is to minimize cost and construction impact on traffic (which is also a cost), while ensuring that the strength, stiffness of the main girder are up to code and foundation settlement is suppressed. The final solution can recommend: first, to perform preventive grouting reinforcement on the pier foundation to stabilize the foundation (to address hidden defects), and then to apply external prestress to the main girder to compensate for the loss and improve the bearing capacity (to address the explicit defects), rather than to perform a more expensive large-scale bridge deck overhaul. This solution precisely targets the root cause, is economical and effective.

[0076] The embodiments of the present application are described in detail above with reference to the accompanying drawings, but the present application is not limited to the above-described embodiments, and various changes can be made within the knowledge of those skilled in the art without departing from the spirit of the present application.

Claims

1. A building structure reinforcement method based on data recognition, characterized in that: Includes the following steps: S11: Obtain full lifecycle data of building structures, including raw data, historical data, and real-time monitoring data; S12: Based on raw and historical data, establish a theoretical benchmark model that reflects the theoretical state of the building structure under the influence of historical loads and maintenance. S13: Based on the theoretical benchmark model and real-time monitoring data, a real-time simulation model reflecting the true state of the building structure is established through model update technology; S14: Compare the real-time simulation model with the theoretical benchmark model, perform deviation analysis, and conduct root cause analysis based on the deviation analysis results to identify the explicit and implicit defects of the building structure. S15: Based on the identified defects, inference is performed using a pre-built causal reasoning model to predict potential weak points and invisible defects in the building structure; S16: Decompose the building structure into multiple independent components and define a performance attribute vector for each component, including stiffness, strength, toughness, and support. S17: Establish a reinforcement measure library, where each reinforcement measure in the library is quantified as the amount of influence on the performance attribute vector of at least one component; S18: With the constraint that the performance attribute vectors of all components reach the preset reinforcement target, and with the optimization objective of minimizing the total reinforcement cost, the measures in the reinforcement measure library are combined and optimized to generate the optimal reinforcement scheme.

2. The building structure reinforcement method based on data recognition according to claim 1, characterized in that: When establishing a theoretical benchmark model based on raw and historical data to reflect the theoretical state of a building structure under historical loads and maintenance influences, the specific steps include: S21: Based on the construction drawing information in the acquired raw data, extract the geometric information, material constitutive information and boundary conditions of the structure, and establish an initial finite element model. The initial state of the initial finite element model corresponds to the theoretical state of the structure when it is completed. S22: Read the load change records in the historical data. The load change records record the changes in permanent loads, variable loads and accidental loads borne by the structure in time series. S23: Each load event in the load change record is treated as an independent load case according to its occurrence time and duration. It is then applied to the initial finite element model for static and dynamic analysis. The response of the structure under each load case is calculated, and the cumulative effect of key components is recorded. S24: Read historical maintenance records from historical data. These records describe reinforcement and repair measures for specific parts of the structure. S25: Quantify each maintenance event into a modification operation on the material properties, geometric properties and boundary conditions of the corresponding components in the model, and apply the modification operation to the processed finite element model in chronological order; S26: After sequentially applying and calculating all historical load events and maintenance events, the final theoretical baseline model is obtained.

3. The building structure reinforcement method based on data recognition according to claim 2, characterized in that: When establishing a real-time simulation model reflecting the true state of a building structure based on a theoretical benchmark model and real-time monitoring data, using model update technology, the following steps are included: S31: Define the model parameter vector to be updated. The parameter vector includes parameters that characterize the physical state of the structure, including the elastic modulus, cross-sectional area, moment of inertia, and boundary condition coefficients of key components. S32: Define the structural response vector corresponding to the parameter vector. The structural response vector consists of the calculated values ​​of the theoretical reference model under the parameter vector, including the displacement, strain, acceleration frequency and mode shape of key measuring points. S33: Obtain the measured structural response vector corresponding to the structural response vector from the real-time monitoring data; S34: Construct the objective function and quantify the difference between the response vector and the measured response vector; S35: With minimizing the objective function as the optimization objective, the parameter vector is repeatedly adjusted through an iterative algorithm so that the calculated response continuously approaches the measured response; S36: When the value of the objective function is less than the preset tolerance or the number of iterations reaches the upper limit, the iteration stops. At this time, the theoretical benchmark model is updated with the finally determined parameter vector to obtain the real-time simulation model.

4. The building structure reinforcement method based on data recognition according to claim 3, characterized in that: When comparing the real-time simulation model with the theoretical benchmark model, performing deviation analysis, and conducting root cause analysis based on the deviation analysis results to identify explicit and implicit defects in the building structure, the following steps are included: S41: Define a set of key performance indicators for model comparison. Key performance indicators include global indicators and local indicators. Among them, the global indicators include the first n natural frequencies and main mode shapes of the structure, and the local indicators include the displacement, stress, strain of key nodes and the internal forces of main components. S42: Extract the calculated values ​​of key performance indicators from the real-time simulation model and the theoretical benchmark model respectively to obtain the measured response vector and the theoretical response vector; S43: Calculate the deviation of each key performance indicator and form a deviation vector; S44: Compare the deviation vector with preset multi-level thresholds to identify significant deviations that exceed the allowable range; S45: Based on the identified significant deviations, combined with the mechanical properties of the structure and historical construction technology data, root cause analysis is conducted to establish a mapping relationship from deviation patterns to defect types, and to diagnose the type, location, and severity of defects. S46: Output a defect diagnosis report, which includes a defect list, a defect location distribution map, and corresponding root cause inferences.

5. The building structure reinforcement method based on data recognition according to claim 4, characterized in that: When predicting potential weak points and invisible defects in a building structure based on identified defects using a pre-built causal reasoning model, the following steps are included: S51: Construct a causal knowledge graph of building structures, with components as nodes and the mechanical interaction relationships between components as directed edges; S52: Map the identified explicit and implicit defects to defect evidence of the corresponding nodes in the causal knowledge graph, and assign an initial confidence level to each defect evidence. S53: Based on the directed edge relationships in the causal knowledge graph, define the causal rules for defect propagation. The causal rules describe the probability of the influence of a certain type of defect on the state of downstream nodes. S54: Starting with defective evidence, reasoning and calculation are performed along the directed edges of the causal knowledge graph, and the defect risk probability of downstream nodes is updated level by level according to the causal rules. S55: Filter out nodes whose defect risk probability exceeds a preset threshold, diagnose the corresponding components as potential weak links and invisible defects, and assess their risk level. S56: Output a potential defect prediction report, including the predicted location of the weak link, defect type, risk probability, and reasoning path chain.

6. The building structure reinforcement method based on data recognition according to claim 5, characterized in that: When decomposing a building structure into multiple independent components and defining a performance attribute vector for each component, including stiffness, strength, toughness, and support, the specific steps include: S61: Based on the BIM model of the building structure, the overall structure is topologically decomposed into multiple independent component units, each of which is assigned a unique identifier. Component units include beams, slabs, columns, walls and nodes. S62: Define a set of core performance properties for each component element, including stiffness properties, strength properties, toughness properties, and support properties; S63: Based on the calculation results of the real-time simulation model, quantify the initial values ​​of the performance attributes of each component unit and construct a performance attribute vector; S64: Based on structural design specifications and safety requirements, set a corresponding performance attribute target vector for each component unit.

7. The building structure reinforcement method based on data recognition according to claim 6, characterized in that: When establishing a reinforcement measure library, where each reinforcement measure in the library is quantified as an impact on the performance attribute vector of at least one component, the specific steps include: S71: Establish a knowledge base for reinforcement measures. The knowledge base stores a variety of standard reinforcement measures. Each measure includes the measure name, applicable component type, construction process and cost parameters. S72: For each reinforcement measure in the knowledge base, define its domain of influence, which includes directly affected components and indirectly affected components; S73: Establish a performance impact quantification model for each reinforcement measure, mapping the implementation amount of the measure to the change in the performance attribute vector of the component within the impact domain; S74: Store the quantified impact in matrix form to construct a reinforcement measure-attribute impact matrix library for subsequent optimization calculations.

8. The building structure reinforcement method based on data recognition according to claim 7, characterized in that: The performance impact quantification model is expressed by the following formula: ； in, This represents the change in the performance attribute vector. For the reinforcement measure type coefficient matrix, For the measure application quantity matrix, Let be the component-measure coupling coefficient matrix; where the measure application amount matrix is ​​calculated using the following formula: For adhesive reinforcement measures: ; For reinforcement measures involving increased cross-section: ; in, To strengthen the elastic modulus of the material, To reinforce the cross-sectional area of ​​the material, To reinforce the length or coverage area of ​​the material, This represents the cross-sectional area of ​​the reinforced component. This represents the cross-sectional area of ​​the original component.

9. The building structure reinforcement method based on data recognition according to claim 8, characterized in that: The specific steps involved in generating the optimal reinforcement scheme include: S81: Define a decision variable, which is a binary variable representing whether to adopt the j-th measure in the reinforcement measure library; S82: Construct an optimization objective function with the goal of minimizing the total cost of all reinforcement measures adopted; S83: Construct constraints that require the final performance attribute vector of each component unit after reinforcement to be no less than its preset target vector; S84: The reinforcement scheme optimization problem is formalized as a constrained binary integer programming problem; S85: The planning problem is solved using an optimization algorithm to obtain the optimal combination of reinforcement measures; S86: Convert the solution results into an actionable reinforcement plan, including a list of measures, construction sequence, and cost budget.

10. A monitoring system for building structure reinforcement based on data recognition, used to implement the building structure reinforcement method based on data recognition as described in any one of claims 1-9, characterized in that: include: The data acquisition module is used to acquire data on the entire lifecycle of the building structure; The model management module is used to build and manage theoretical benchmark models and real-time simulation models; The intelligent diagnostic module is used to perform model comparison, deviation analysis, and root cause analysis. The optimization solution module is used to perform component attribute quantification, measure impact quantification, and combinatorial optimization solution. The solution output module is used to generate and visualize the optimal reinforcement solution.

Images