Damage identification method and system based on data fusion and adaptive sparse regularization

Abstract

The present invention discloses a damage identification method and system based on data fusion and adaptive sparse regularization. Leveraging multidisciplinary Internet of Things and structural dynamics analysis technology, the system collects and analyzes a variety of structural dynamic response monitoring data from key parts of civil structures in real time. This method is based on a set of damage identification indicators fused from a small amount of multivariate heterogeneous data and adaptive sparse regularization solution technology for structural health monitoring and damage identification. This method addresses the issues of insufficient mining and fusion of multi-source heterogeneous sensory data in civil structure health monitoring systems, as well as the issues of insufficient sensitivity of damage identification indicators or their susceptibility to interference from environmental noise, leading to misjudgment. The technical solution provided by the present invention can be deployed in a structural health monitoring system to diagnose and assess the safety of engineering structures, achieving early warning, early detection, and early disposal of potential safety risks to engineering structures, thereby ensuring the safe operation of urban civil engineering structures and efficient response to emergencies.

Description

Technical Field

[0001] The present invention relates to the technical field of civil engineering structure safety monitoring and structural damage identification, and in particular to a damage identification method and system based on multivariate heterogeneous data fusion and adaptive sparse regularization. Background Art

[0002] The health of major civil engineering structures such as long-span bridges, super-high-rise buildings, long-span space structures, reservoirs and dams, offshore platforms, and high-voltage transmission towers is directly related to the safety of people's lives and property, and the normal operation of civil engineering infrastructure is related to the normal operation of the national economy. However, during the decades or even hundreds of years of service of civil engineering structures, the coupling of disaster factors such as environmental erosion, material aging and load effects, man-made or natural mutation effects will inevitably lead to the accumulation of structural damage and attenuation of resistance, thereby reducing the ability to resist natural disasters, normal loads and environmental effects, and triggering catastrophic accidents. There are countless tragedies caused by the failure to detect structural damage in time. For example, the collapse of the Silver Bridge in West Virginia, USA in 1967 killed 46 people and injured 9 people; the collapse of the Sungsoo Bridge in Seoul, South Korea in 1994 The collapse of the Grand Bridge killed 32 people and injured 17 people; in 2007, the Jiujiang Bridge in Guangdong was hit by a sand-carrying ship, causing the bridge deck to collapse about 200 meters and 9 people to die; on June 27, 2009, a 13-story residential building under construction in "Lotus Riverside Garden" in Minhang District, Shanghai collapsed uprooted; on July 27, 2018, the Tonglu Corridor Bridge in Zhejiang collapsed, causing a total of 8 deaths and 3 injuries; in 2019, the collapse of the National Highway 312 overpass in Wuxi caused a total of 3 deaths and 2 injuries; on July 8, 2019, the Shenzhen Sports Center accidentally collapsed during the demolition process, and some construction workers were buried; the Shenzhen building collapse accident on August 28, 2019 and the strong shaking incident of the Shenzhen super-high-rise building SEG Tower on May 18, 2021 did not cause any casualties, but caused large economic losses. Numerous accidents have demonstrated that when damage to critical structural components accumulates to a certain level, if not promptly detected and addressed, it can rapidly expand, leading to the destruction of the entire structure. Furthermore, the overall destruction or widespread failure of civil structures often originates from localized damage, such as tiny fatigue cracks or corrosion damage to functional components. To promptly detect potential safety hazards and ensure structural safety, research in structural health monitoring (SHM) and structural damage identification technologies is of great significance and faces significant application demand.

[0003] In the field of structural health monitoring and damage identification in civil engineering, dynamic damage identification is one of the most widely used techniques, having developed rapidly both domestically and internationally over the past few decades. The core concept of this technique is that a structure's vibration characteristics (such as frequency and mode shapes) and dynamic responses (such as acceleration, displacement, velocity, strain, and stress) are functions of its physical parameters (such as mass and stiffness). Structural damage implies changes in its physical parameters, which inevitably lead to changes in its vibration characteristics and dynamic responses. Dynamic damage identification methods can be categorized into frequency-domain methods (using frequency, mode shapes, and their derivatives as damage identification indicators) and time-domain methods (using dynamic responses and their derivatives as damage identification indicators). Although numerous vibration-based structural damage methods have been proposed, many challenges remain in their application, such as insufficient sensitivity of damage identification indicators (structural frequency or mode shapes), susceptibility to interference from environmental noise, and misjudgment. Furthermore, the limited number of sensors and the large number of structural parameters to be identified represent significant obstacles to the practical application of existing dynamic damage identification techniques, requiring continued research and continuous advancement.

[0004] Furthermore, current SHM systems utilize a wide variety of intelligent sensors, including acceleration, displacement, and strain sensors. However, most existing damage identification metrics and methods utilize data from only one sensor type for damage identification, failing to fully utilize the data from various sensors and failing to integrate the respective advantages of different data types for high-precision structural damage identification. The data measured by various sensors possess distinct characteristics. For example, acceleration responses can easily yield high signal-to-noise ratio measurements and contain higher kinetic energy in higher-order vibration modes. In contrast, displacement responses contain more kinetic energy in lower-order vibration modes. Strain or stress responses are highly sensitive to local damage changes near the sensor but less sensitive to local damage changes farther away. Given the varying strengths and limitations of these sensors, the combined use of multiple sensor types can complement each other's strengths, improving data quality and damage identification effectiveness. However, the data from various sensors vary in physical meaning, dimensions, and characteristics, leading to a lack of research on methods for integrating multi-type sensor data for damage identification.

[0005] In summary, most existing structural health monitoring systems can only issue simple alarms based on alarm thresholds and cannot accurately and effectively locate and quantitatively analyze structural damage. Furthermore, many existing structural damage identification methods use data from only a single type of sensor for structural health assessment, failing to fully leverage the strengths of multiple heterogeneous data to mitigate their weaknesses. Furthermore, damage identification indicators are not sensitive enough or are easily affected by environmental noise, leading to misjudgments. Furthermore, the convergence rate of the damage identification process is slow. Therefore, developing highly sensitive damage identification indicators and corresponding damage identification methods based on multiple heterogeneous data is an engineering bottleneck that urgently needs to be addressed in the engineering field. Summary of the Invention

[0006] (1) Technical issues to be resolved

[0007] Most structural health monitoring systems can only issue simple alarms based on alarm thresholds, failing to accurately and effectively locate and quantitatively analyze structural damage. Furthermore, most structural damage identification methods rely solely on data from a single sensor type, failing to fully utilize multi-dimensional, heterogeneous data. Furthermore, damage identification indicators are insufficiently sensitive or easily interfered with by environmental noise, leading to misjudgments and slow convergence in the damage identification process. Therefore, developing highly sensitive damage identification indicators and corresponding damage identification methods based on multi-dimensional, heterogeneous data is an urgent engineering bottleneck that needs to be addressed in the engineering field.

[0008] (2) Technical solution

[0009] In order to overcome the shortcomings of the existing technology, the present invention uses multidisciplinary Internet of Things and structural dynamics analysis technology to collect, transmit, store, analyze and apply structural dynamic response monitoring data such as acceleration, displacement, strain, etc. of key parts of civil structures in real time, and provides a set of damage identification indicators and adaptive sparse regularization solution technology that can realize the fusion of multi-source heterogeneous data such as acceleration, displacement, strain, etc. to perform structural health monitoring and damage identification methods. The main purpose is to solve the problem that multi-source heterogeneous perception data in civil structure health monitoring systems cannot be fully mined and integrated, and the problem that damage identification indicators are not sensitive enough or damage indicators are easily interfered with by environmental noise and cause misjudgment. An adaptive sparse regularization method is proposed to solve the problem that the pathological inverse problem of damage identification is prone to misjudgment and the iterative solution has a slow convergence speed.

[0010] The damage identification method based on data fusion and adaptive sparse regularization is characterized by performing multivariate perception data fusion based on the covariance of data measured by a small number of optimized sensors and solving it using an adaptive sparse regularization method. This method can effectively overcome the ill-posed problems of the inverse problem of damage identification and obtain sparse and accurate damage identification results. The method specifically includes the following steps:

[0011] S1. By presetting a finite element model of the civil structure, the optimal layout calculation and analysis of the measurement points of the structural health monitoring sensors are performed to obtain the optimal layout plan of multiple structural health monitoring sensors;

[0012] S2. Real-time collection of multi-sensory structural dynamic response monitoring data of the target civil structure, and standardization and dimensionless processing of the structural dynamic response monitoring data to obtain a standardized structural response vector;

[0013] S3, using the covariance function to perform multivariate heterogeneous data fusion on the standardized structural response vector, and constructing a damage identification index V based on the covariance-based multivariate perception fusion according to the result of the multivariate heterogeneous data fusion. pq ,The damage identification index based on covariance multivariate perception fusion is sensitive to local damage but insensitive to measurement noise;

[0014] S4, according to the damage identification index of the multivariate perception fusion based on covariance and the actual measured multivariate perception structural dynamic response monitoring data, calculate the multivariate perception data fusion damage index vector of the target civil structure containing damage information

[0015] S5, according to the damage identification index of the multivariate perception fusion based on covariance and the finite element model of the preset civil structure in a healthy benchmark state, a simulation calculation is performed to obtain a multivariate perception data fusion damage index vector based on the finite element model

[0016] S6. According to and A damage identification equation is constructed, and the damage identification equation is solved and the model is updated in combination with the adaptive sparse regularization technology. Finally, the damage identification result is obtained by iterative solution, so that the location and extent of the damage can be determined.

[0017] S7. A damage identification method based on data fusion and adaptive sparse regularization, characterized in that it also includes automatic early warning of the damage identification results through wireless data communication.

[0018] A damage identification system based on data fusion and adaptive sparse regularization is used to execute the above-mentioned damage identification method based on data fusion and adaptive sparse regularization, and comprises:

[0019] (1) Data real-time acquisition and transmission module, which is used for 24-hour unattended continuous acquisition of structural dynamic response data such as structural vibration acceleration, displacement, and strain, and transmits the monitoring data back to the management cloud platform in real time via 4G / 5G or a dedicated network, so as to remotely view and set the status of the sensor and related parameters;

[0020] (2) Data storage and management module, which is used to store and manage massive multi-source heterogeneous monitoring data generated by different types of sensors in multiple structural arrays, and includes: building a high-performance database with dynamically scalable storage capacity and dynamic hierarchical management of data based on cloud storage technology;

[0021] (3) Data analysis and structural safety assessment module, which is used to provide basic data analysis such as data cleaning, data integration, data conversion, data reduction, data integration, spectrum analysis, and statistical value analysis for structural safety assessment. It also includes: a built-in damage identification and analysis algorithm based on covariance-based multivariate data fusion, which realizes automatic analysis of massive data and damage diagnosis based on multivariate perception data fusion. The process does not require human intervention, achieving automated and efficient structural status assessment;

[0022] (4) a structural safety warning and warning information sending module, which is used to establish a structural safety multi-level warning threshold indicator system based on the standard limits and structural damage identification and assessment results, and use the structural safety multi-level warning threshold indicator system as the basis for structural safety warning;

[0023] (5) System visualization module, which is used to provide a user-friendly system visualization interface based on B / S architecture.

[0024] (3) Beneficial effects

[0025] The present invention proposes a structural damage identification index and damage identification method based on multivariate sensing data fusion of response covariance that is sensitive to local structural damage but insensitive to measurement noise. The method comprehensively considers the overall response and local response of the structure, obtains the state information of the structure more comprehensively, and can quantitatively identify the location and degree of structural damage based on the measurement data of a small number of optimized sensors, thereby significantly improving the accuracy of structural damage identification. The purpose of the present invention is to provide an innovative civil structure health monitoring and damage identification method based on multivariate sensing data fusion and adaptive regularization technology to address the shortcomings of the existing structural damage identification methods mentioned in the above background technology, and to form a systematic comprehensive system architecture for civil structure health diagnosis based on intelligent sensors, Internet of Things technology, multivariate sensing data fusion damage identification indicators and adaptive sparse regularization damage identification solution algorithms, which integrates software, hardware and analysis algorithms.

[0026] To address the problem of underutilization of multi-source heterogeneous sensing data in civil structure health monitoring systems, the present invention proposes a damage identification indicator based on response covariance multivariate sensing data fusion. This indicator can fuse multi-source heterogeneous monitoring data and jointly extract components sensitive to structural damage from various structural response data such as measured acceleration, displacement, and strain to construct a structural dynamic indicator for structural damage identification. This combines the advantages of various structural dynamic response monitoring data, leveraging their strengths and minimizing their weaknesses. The damage identification indicator based on response covariance multivariate sensing obtained after data fusion is more sensitive to local structural damage and insensitive to environmental measurement noise. Finally, combined with an adaptive sparse regularized damage solution technology, more accurate damage identification can be achieved. This effectively overcomes the problem of prone to misjudgment in the inverse problem of pathological damage identification and efficiently obtains sparse and accurate damage identification results. The damage identification indicator and damage identification method can be deployed in a structural health monitoring system to diagnose and evaluate the safety of engineering structures, achieve early warning, early detection, and early disposal of engineering structure safety risks, and ensure the safe operation of urban civil engineering structures and efficient response to emergencies. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 Schematic diagram of an application scenario of a damage identification method and system based on data fusion and adaptive sparse regularization in an embodiment of the present invention;

[0028] Figure 2 Schematic diagram of the technical route of the damage identification method and system based on data fusion and adaptive sparse regularization in an embodiment of the present invention;

[0029] Figure 3 Schematic diagram of an iterative solution process for damage identification based on multi-sensory data fusion and adaptive sparse regularization in an embodiment of the present invention;

[0030] Figure 4 Schematic diagram of the iterative solution results of the damage identification method based on multi-sensory data fusion and adaptive sparse regularization in an embodiment of the present invention;

[0031] Figure 5 This is the technical roadmap for the civil structure safety monitoring and early warning software system with a B / S architecture deployed in the cloud in an embodiment of the present invention. DETAILED DESCRIPTION

[0032] In order to make the technical problems, technical solutions and beneficial effects to be solved by the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0033] A damage identification method based on data fusion and adaptive sparse regularization and a system for executing the method are applied to Figure 1 In the scenario shown, accelerometers, displacement sensors, and strain sensors are deployed within the civil engineering structure, but the sensors that can be deployed are not limited to these. The sensors communicate with a cloud server via a 4G / 5G network. The cloud server uses an algorithm library to identify damage to the civil engineering structure and sends warning information to personnel using a visual software front-end platform based on a B / S architecture.

[0034] The structural damage identification method based on covariance multivariate perception data fusion index in the algorithm library includes four main sub-algorithms: data normalization and dimensionless, multivariate perception data fusion calculation, structural damage location and quantitative analysis based on adaptive sparse regularization technology, structural health assessment and decision support. Its functions and implementation can be found in Figure 2 , mainly includes the following steps:

[0035] (1) Combined with the finite element model of the civil engineering structure, the optimal layout calculation and analysis of the measurement points of various structural health monitoring sensors such as acceleration, displacement, and strain are performed. That is, the optimized layout plan includes obtaining the optimal layout position and number of various sensors;

[0036] (2) Standardize and dimensionlessly process different types of measurement data, so that data with different dimensions and variation characteristics measured by various sensors can be fused. The various data that need to be standardized and dimensionlessly processed mainly include structural dynamic response monitoring data recorded by intelligent sensors such as acceleration, displacement, and strain sensors. These data are recorded together in the observation vector and standardized and dimensionlessly processed by dividing the raw data collected by various sensors by their respective standard deviations.

[0037] (3) Using the covariance function to fuse multivariate heterogeneous data and construct a damage identification index based on multivariate perception data fusion that is sensitive to local damage. First, by calculating the cross-covariance function between any two normalized and dimensionless structural response data, the information fusion of multivariate heterogeneous data can be performed. Then, the covariance matrix obtained by the above calculation is converted into a one-dimensional data vector to obtain the covariance-based multivariate perception fusion damage identification index vector.

[0038] (4) According to the structural dynamic response data such as acceleration response, displacement response and strain response obtained from the damaged structure and the corresponding finite element model, the damage identification index vector based on the fusion of response covariance multivariate perception data is calculated respectively, and the damage identification equation is expressed as a first-order Taylor expansion. The time domain iteration method is used to further rewrite the damage identification equation into an iterative linear damage identification equation. On this basis, the adaptive sparse regularization technology proposed in this patent can be used to perform refined damage positioning and quantitative analysis.

[0039] (5) Based on the above damage identification equation, damage identification is performed using a damage identification index based on response covariance multivariate perception data fusion and a sensitivity analysis method.

[0040] Therefore, by collecting and feeding back multi-dimensional sensory data on the service status of structures in real time and combining it with the finite element model of the actual engineering structure, the location and extent of damage can be effectively determined, the safety of the structure can be timely and effectively assessed, changes in structural performance can be predicted, and early warnings can be issued for emergencies. This allows for a more comprehensive understanding of the stress and damage evolution laws throughout the entire process of structural construction and service, ensuring the service safety of large-scale engineering structures.

[0041] Specifically, in step S1, the finite element model of the civil engineering structure is pre-set, which has two functions: first, it is used for optimizing the arrangement of sensors, screening the standardized structural response vector in step S2, and obtaining the data measured by sensors based on a small number of optimized arrangements; second, it is used for simulation to obtain the parameters of the healthy civil engineering structure, that is, the damage index vector based on the multivariate perception data fusion of the finite element model.

[0042] In step S2, sensors deployed in the target civil structure collect data in real time. Suppose the structural dynamic response monitoring data is y(t), which includes acceleration, displacement, strain, etc., and is multivariate heterogeneous measurement data, which can be expressed by formula (1):

[0043] y(t)=C c x(t)+D c f(t) (1)

[0044] in,

[0045] in, is the observation vector, ε(t) is the strain response time history, z(t) is the displacement response time history, is the acceleration response time history; is the state vector, satisfying the dynamic equation:

[0046]

[0047] in,

[0048] Where ψ = BGL d Φ is the strain modal matrix; L d is the selection matrix of the node displacement for matching strain calculation; the matrix G is the coordinate transformation matrix from global coordinates to local coordinates; the vector B defines the local strain-displacement relationship; Φ is the modal matrix; ω and ξ are the diagonal matrices of the structural natural frequency and damping ratio, respectively; q and are the displacement and velocity in modal coordinates respectively; f(t) is the external excitation vector.

[0049] That is, by adopting the method of multi-sensory data fusion, structural dynamic indicators that are sensitive to structural damage are jointly extracted from various structural dynamic response data such as acceleration, displacement, and strain. The different characteristics of various types of data can be used to effectively improve the quality of structural dynamic indicator data, thereby achieving sensitive and high-precision damage location and quantitative analysis.

[0050] The structural dynamic response monitoring data y(t) are standardized and dimensionless, that is, the original structural response data are divided by their respective standard deviations to obtain the standardized structural response vector According to formula (5):

[0051]

[0052] Among them, y p (t) is the original structural dynamic response monitoring data, For y p (t) The corresponding standard deviation, i.e., the raw structural response data are divided by their respective standard deviations.

[0053] In step S3, a covariance-based multivariate perception data fusion damage identification index is adopted. This index can effectively fuse multivariate heterogeneous data, improve the sensitivity of the damage identification index to structural damage, and reduce the impact of measurement noise on damage identification.

[0054] The covariance function is used to fuse multivariate heterogeneous data and reduce measurement noise, and the cross-covariance function formula (6) of any two standardized structural responses is calculated:

[0055]

[0056] Among them, y p is the structural response recorded by sensor p, is the normalized structural response, represents the standard deviation of the structural dynamic response recorded by sensor p on the intact structure; E represents the expectation, the variable τ is the time interval, and the subscripts p and q represent the calculated values ​​from the responses measured by sensors p and q;

[0057] After the cross-covariance function of the structural response is obtained according to formula (6), it can be further assembled into a damage identification index vector V based on covariance-based multivariate perception data fusion: pq , the calculation method is as follows:

[0058]

[0059] Among them, p i ∈[p1,p s ],q j ∈[q1,q s ]The subscript s indicates the total number of selected sensors; nt is the total number of time intervals selected for damage identification.

[0060] In steps S4 and S5, the V obtained in step S3 is respectively pq , calculate the damage index vector of the target civil structure using multi-sensor data fusion and damage index vector based on multivariate perception data fusion of finite element model

[0061] Among them, in S5, based on the damage identification of finite element model and sensitivity analysis, the finite element model of the preset civil structure is simulated and calculated. Its goal is to identify the location and degree of structural stiffness reduction caused by local structural damage. In order to perform quantitative structural damage identification, the stiffness matrix of the damaged structure can be expressed by the mathematical formula:

[0062]

[0063] Where 0≤α i ≤1, -1≤Δα i ≤0,α i ∈α is the coefficient of the stiffness matrix corresponding to the i-th unit; Δα i ∈Δα is the local stiffness change of the i-th unit; ne is the total number of units; α is the vector of stiffness matrix coefficients; Δα is the vector of local stiffness change of the damaged unit.

[0064] At the same time, in S4, The damage identification equation is expressed as follows using the first-order Taylor expansion:

[0065]

[0066] The time domain iteration method is used to transform the formula (9) into a linear damage identification equation integrated with the iterative Gauss-Newton algorithm:

[0067]

[0068] Where k = 0, 1, 2, 3, ...,

[0069]

[0070] is the damage identification index vector calculated by data fusion of the response of the finite element model; vector ΔV pq for and The difference between them; Δα represents the vector of local change of stiffness of damaged element, and Δα i ∈Δα(-1≤Δα i ≤0) is the local stiffness change of the i-th unit; S is the sensitivity matrix of the damage identification index vector of data fusion to the local stiffness change vector, calculated using the finite difference method; the superscript k is the number of iterations.

[0071] Since only a small number of sensors can be installed on the structure for structural health monitoring in engineering practice, the number of units that need to be damaged and quantitatively analyzed is far greater than the number of damage identification equations. Therefore, solving the damage identification equation (10) is often a problem of solving an ill-conditioned equation. Generally, the most widely used method for solving ill-conditioned loss identification problems is the Tikhonov regularization method. However, since local damage often only occurs in a local area of ​​the structure and the damage spatial distribution is sparse, the Tikhonov regularization method will produce some misjudgments on some non-damaged units.

[0072] To this end, the present invention proposes an innovative adaptive sparse regularization method to solve the iterative damage identification equation (10) (e.g. Figure 3 As shown): In S6, let the damage identification parameter α be expressed by formula (13), and Δα=∑Δα k+1 Obtain updated damage parameters, which are used to update the finite element model of the engineering structure, thereby identifying the corresponding updated structural damage location and severity;

[0073]

[0074] st-1≤Δα i k+1 +∑Δα i k ≤0;λ *k+1 ≥0

[0075] Among them, λ *k+1 is the adaptive sparse regularization coefficient, ∑Δα k+1 It is the cumulative damage identification amount representing the damage location and extent of structural damage; represents data fidelity; Represents the sparse constraint of damage identification solution.

[0076] Formula (13) innovatively considers the prior condition that the damage of actual engineering structures often only occurs in a small number of spatial locations of the structure, and thus uses the L1 norm to construct the sparse regularization constraint term Replaces the L2 regularization term of the traditional Tikhonov regularization method Therefore, the damage identification solution based on adaptive sparse regularization proposed by the present invention can automatically perform sparsity constraints, making the solution converge faster and greatly reducing false positives on non-damage identification units. Therefore, in the process of iteratively solving equations (10) and (13), a converged solution (such as Figure 4 As shown in the figure, it can provide accurate structural damage location and damage extent, realize accurate and reliable diagnosis and assessment of engineering structure safety, achieve early warning, early detection and early disposal of engineering structure safety risk hazards, and ensure the safe operation of urban civil engineering structures and efficient response to emergencies.

[0077] Furthermore, in the process of damage identification based on the adaptive sparse regularization method, its adaptive ability comes from the adaptive regularization coefficient λ in formula (13): *k+1 Since formula (13) is an L1 regularized optimization equation, it often does not have an analytical solution, so the current research literature does not give λ *k+1 An expression with a mathematical formula is mainly a subjective regularization coefficient given by experience or a large number of trial calculations. Therefore, λ is proposed to address the above problems. *k+1 A new expression formula is proposed, which assumes that the data fidelity and represents the sparse constraint of damage identification solution It is of equal importance in the damage identification process. A new calculation method is proposed, including formula (14) and formula (15):

[0078]

[0079] in, is the reference solution for damage identification in step k. The specific calculation formula is as follows:

[0080]

[0081] That is, it is proposed to use a semidefinite programming optimization algorithm to solve the problem. This method can efficiently search for the optimal solution in the feasible domain in each step of the solution process.

[0082] After the damage and identification results are obtained, step S7 is also included, in which an early warning of the damage and identification results is issued through wireless data communication methods such as text messages, emails, etc.

[0083] A damage identification system based on data fusion and adaptive sparse regularization is proposed to implement the above-mentioned damage identification method based on data fusion and adaptive sparse regularization. Specifically, Figure 5 Shown, including:

[0084] (1) Real-time data acquisition and transmission module: This module continuously collects structural dynamic response data such as structural vibration acceleration, displacement, and strain 24 hours a day without human intervention. The monitoring data is transmitted back to the management in real time via 4G / 5G or a dedicated network. The status of the sensor and related parameters can be viewed and set remotely. When the transmission is interrupted, the module provides a data retransmission function and alerts the on-duty personnel. When an event such as an earthquake, typhoon, impact, or explosion occurs, the system automatically triggers the recording of the event data.

[0085] (2) Data storage and management module: In response to the massive multi-source heterogeneous monitoring data generated by different types of sensors in multiple structural arrays, a high-performance database with dynamically expandable storage capacity and dynamically hierarchical data management is established based on cloud storage technology to solve the problems of limited storage space and low reading efficiency of traditional monitoring data.

[0086] (3) Data analysis and structural safety assessment module: It provides basic data analysis such as data cleaning, data integration, data conversion, data reduction, data integration, spectrum analysis and statistical value analysis (including maximum value, average value, peak-to-peak value and effective value); for structural safety assessment, it embeds a damage identification and analysis algorithm based on covariance-based multivariate data fusion to realize automatic analysis of massive data and damage diagnosis based on multivariate perception data fusion. The process does not require human intervention, thus achieving automated and efficient structural status assessment.

[0087] (4) Structural safety warning and warning information transmission module: Based on the standard limits and structural damage identification and assessment results, a scientific and reasonable multi-level warning threshold indicator system for structural safety is established as the basis for structural safety warning. Warning information can be automatically pushed to management personnel through various forms such as email, WeChat, and SMS as needed.

[0088] (5) System visualization module: It provides a user-friendly system visualization interface based on B / S architecture, realizing comprehensive functions such as structural information management, sensor management, dynamic visualization of real-time data and spectrum, automatic background analysis and result storage of real-time data, historical data call and analysis, structural safety status assessment and early warning information management and release.

[0089] That is, based on the data transmission mode, model application mode and system integration technology of the cloud platform, data and analysis algorithms based on different platforms and different interfaces are integrated to form a unified data layer, model layer, evaluation algorithm layer and auxiliary decision layer, and finally form a set of B / S architecture software system deployed in the cloud to realize the integrated and efficient management of intelligent monitoring, evaluation and early warning, and auxiliary decision-making of civil structure disaster prevention and safety.

[0090] In summary, the present invention utilizes multidisciplinary Internet of Things and structural dynamics analysis technologies to collect, transmit, store, analyze, and apply real-time structural dynamic response monitoring data, such as acceleration, displacement, and strain, at key parts of civil structures. This invention provides a set of damage identification indicators and adaptive sparse regularization solution techniques that can integrate heterogeneous data such as acceleration, displacement, and strain to perform structural health monitoring and damage identification methods and systems. These methods enable health diagnosis, assessment, and early warning of the safety status of civil structures, with early warning information automatically delivered to relevant personnel, providing a scientific basis for the safe operation of civil structures in disaster prevention. To address the issues of multi-level data transmission and the independence of monitoring systems and decision support in traditional health monitoring systems, this invention proposes a cloud-based data transmission model, model application model, and system integration technology. This integrates data and analysis algorithms based on different platforms and interfaces to form a unified data layer, model layer, assessment algorithm layer, and decision support layer. Finally, this invention proposes an integrated cloud-based software platform for efficient and integrated management of intelligent monitoring, assessment, early warning, and decision support for civil structure disaster prevention safety.

[0091] The above is an explanation of the sparse regularized damage identification method and device based on data fusion and adaptation of the present invention, which is used to help understand the present invention; however, the implementation methods of the present invention are not limited to the above embodiments, and any changes, modifications, substitutions, combinations, and simplifications made without departing from the principles of the present invention should be equivalent replacement methods and are included in the scope of protection of the present invention.

Claims

1. Damage identification method based on data fusion and adaptive sparse regularization, characterized by The data measured by a small number of optimized sensors are fused through covariance analysis and solved using an adaptive sparse regularization method. The specific steps are as follows: S1. By presetting a finite element model of the civil structure, the optimal layout calculation and analysis of the measurement points of the structural health monitoring sensors are performed to obtain the optimal layout plan of multiple structural health monitoring sensors; S2. Real-time collection of multi-sensory structural dynamic response monitoring data of the target civil structure, and standardization and dimensionless processing of the structural dynamic response monitoring data to obtain a standardized structural response vector; S3. Using the covariance function to perform multivariate heterogeneous data fusion on the standardized structural response vector, and constructing a damage identification index based on the covariance-based multivariate perception fusion according to the result of the multivariate heterogeneous data fusion. ,The damage identification index based on covariance multivariate perception fusion is sensitive to local damage but insensitive to measurement noise; S4, according to the damage identification index of the multivariate perception fusion based on covariance and the actual measured multivariate perception structural dynamic response monitoring data, calculate the multivariate perception data fusion damage index vector of the target civil structure containing damage information ; S5, according to the damage identification index of the multivariate perception fusion based on covariance and the finite element model of the preset civil structure in a healthy benchmark state, a simulation calculation is performed to obtain a multivariate perception data fusion damage index vector based on the finite element model ; S6. According to and A damage identification equation is constructed and solved using adaptive sparse regularization technology. The model is updated and the damage identification result is obtained through iterative solution, thereby locating the location and extent of the damage. The structural dynamic response monitoring data includes structural response variables: acceleration, displacement, and strain, and is expressed by the state space equation as follows: （1） in, , (2) is the observation vector, is the strain response time, is the displacement response time history, is the acceleration response time history; is the state vector, satisfying the dynamic equation: （3） in, , (4) is the strain modal matrix; is the selection matrix of the nodal displacements for matching strain calculations; the matrix G is the coordinate transformation matrix from global coordinates to local coordinates; the vector B defines the local strain-displacement relationship; is the modal matrix; and are the diagonal matrices consisting of the structural natural frequency and damping ratio respectively; and are the displacement and velocity in modal coordinates respectively; is the external excitation vector; In S2, the structural dynamic response monitoring data is standardized and dimensionless to obtain a standardized structural response vector, which is calculated using the following formula: （5） in, is the original structural dynamic response monitoring data, For The corresponding standard deviation; In S3, the covariance function is used to perform multivariate heterogeneous data fusion on the standardized structural response vector, including calculating the cross-covariance function of any two standardized structural responses. The formula is as follows: （6） in, is the structural response recorded by sensor p, is the normalized structural response, represents the standard deviation of the structural dynamic response recorded by sensor p on the intact structure; E represents the expectation, and the variable is the time interval, and the subscripts p and q indicate that they are calculated from the responses measured by sensors p and q; Based on the above formula (6), the damage identification index based on covariance multivariate perception fusion is constructed by the following formula: : （7） in, The subscript s denotes the total number of selected sensors; nt is the total number of time intervals selected for damage identification.

2. The damage identification method based on data fusion and adaptive sparse regularization according to claim 1, characterized in that: After obtaining the standardized structural response vector in S2, the method further includes: selecting an optimized structural dynamic response data vector that is sensitive to local structural damage from the standardized structural response vector according to the optimized layout plan of the structural health monitoring sensor.

3. The damage identification method based on data fusion and adaptive sparse regularization according to claim 1, characterized in that: In S5, simulation calculation is performed on the finite element model of the preset civil engineering structure, including: Assume that the stiffness matrix of the damaged structure can be expressed by the mathematical formula: （8） in, , , is the coefficient of the stiffness matrix corresponding to the i-th element; is the local stiffness change of the i-th unit; ne is the total number of units; is the vector of stiffness matrix coefficients; is the vector of local stiffness change of the damaged element.

4. The damage identification method based on data fusion and adaptive sparse regularization according to claim 3, characterized in that: The S4 includes: The damage identification equation is expressed as follows using the first-order Taylor expansion: （9） The time domain iteration method is used to transform the formula (9) into a linear damage identification equation integrated with the iterative Gauss-Newton algorithm: （10） Where k=0,1,2,3,..., （11） （12） is the damage identification index vector obtained by data fusion calculated from the response of the finite element model; vector for and The difference between The vector representing the local change in stiffness of the damaged element, and is the local stiffness change of the i-th unit; S is the sensitivity matrix of the damage identification index vector of data fusion to the local stiffness change vector, calculated using the finite difference method; the superscript k is the number of iterations; The S6 includes: setting the damage identification parameter It is expressed by formula (13) and Obtain updated damage parameters, which are used to update the finite element model of the engineering structure, thereby identifying the corresponding updated structural damage location and severity; （13） in, is the adaptive sparse regularization coefficient, It is the cumulative damage identification amount representing the damage location and extent of structural damage; represents data fidelity; Represents the sparse constraint of damage identification solution.

5. The damage identification method based on data fusion and adaptive sparse regularization according to claim 4, characterized in that: set up and Of equal importance in the damage identification process, the Expressed as: （14） in, is the reference solution for damage identification in step k. The specific calculation formula is as follows: （15） 6. The damage identification method based on data fusion and adaptive sparse regularization according to claim 1, characterized in that: Also includes the steps: S7. Automatically issue an early warning on the damage identification result through wireless data communication.

7. Damage identification system based on data fusion and adaptive sparse regularization, characterized by: The system is used to execute the damage identification method based on data fusion and adaptive sparse regularization according to any one of claims 1 to 6, and includes: (1) Real-time data acquisition and transmission module, used for 24-hour unattended continuous acquisition of structural vibration acceleration, displacement, and strain data, and real-time transmission of monitoring data to the management cloud platform via 4G / 5G or a dedicated network, so as to remotely view and set the status of the sensor and related parameters; (2) Data storage and management module, which is used to store and manage massive multi-source heterogeneous monitoring data generated by different types of sensors in multiple structural arrays, and includes: building a high-performance database with dynamically scalable storage capacity and dynamic hierarchical management of data based on cloud storage technology; (3) Data analysis and structural safety assessment module, which is used to provide data cleaning, data integration, data conversion, data reduction, data integration, spectrum analysis, and statistical value analysis for structural safety assessment. It also includes: an embedded damage identification and analysis algorithm based on covariance-based multivariate data fusion, which realizes automatic analysis of massive data and damage diagnosis based on multivariate perception data fusion. The process does not require human intervention, achieving automated and efficient structural status assessment; (4) Structural safety warning and warning information sending module, which is used to establish a structural safety multi-level warning threshold indicator system based on the standard limits and structural damage identification and assessment results, and use the structural safety multi-level warning threshold indicator system as the basis for structural safety warning; (5) System visualization module, which is used to provide a user-friendly system visualization interface based on B / S architecture.

Images