Method for real-time monitoring and intelligent regulation of ancient building

Abstract

The application relates to the technical field of ancient building protection and repair, and discloses a method for real-time monitoring and intelligent regulation and control of ancient building ramming and straightening, which comprises the following steps: setting an AE warning baseline and an AE danger threshold of an ancient building structure; updating a dynamic structure admittance matrix through a bimodal identification step based on AE event rate gating; online calibration of the brittleness coefficient and the AE baseline of the ancient building structure, establishment of damage stiffness correlation; construction of a predictive feedforward control under dynamic constraints, taking the AE danger threshold as a constraint to solve an optimal force increment for regulation and control of the ancient building structure; and finally, while the optimal force increment is executed, the measured AE event rate is monitored in parallel, and a highest priority veto instruction is sent when the limit is exceeded. The application realizes double safety guarantee through predictive avoidance and real-time veto, realizes dynamic balance between identification accuracy and operation safety through bimodal identification and online calibration, and improves the safety and accuracy of regulation and control.

Description

Technical Field

[0001] This invention relates to the field of ancient building protection and restoration technology, specifically a method for real-time monitoring and intelligent control of ancient buildings to correct their orientation. Background Technology

[0002] Ancient buildings are irreplaceable cultural heritage. Their wooden, brick, or rammed earth structures often suffer from tilting and deformation due to uneven foundation settlement, material aging, or natural disasters during long-term service. Correction and jacking (i.e., lifting and adjusting) is a key engineering technique for repairing these defects and restoring the structural integrity of ancient buildings. However, the structural materials of ancient buildings (especially the mortise and tenon joints of load-bearing wooden components) are usually quite fragile and prone to damage under stress, and their internal damage is difficult to accurately assess. During the adjustment process of applying huge lifting or translation forces, unpredictable and irreversible secondary damage to the structure is highly likely.

[0003] Existing projects for correcting deviations in ancient buildings often suffer from a disconnect between monitoring and control. Although current projects employ 3D laser scanning to verify the amount of deviation and utilize modern monitoring technologies such as high-precision inclinometers, laser rangefinders, and crack sensors, their core control methods still rely on traditional manual command and operation (such as manual hoist pulling). This model of modern monitoring combined with traditional control has the following significant drawbacks: First, existing methods for correcting the alignment of ancient buildings often rely on the commands of the supervisor, the feel of the puller, and the visual observation of the observer. For example, when pre-tightening ropes, workers rely on experience to "pull until they exert all their strength." Because the pulling speed is too fast, structural problems are very likely to occur. Relying on "stopping immediately" for remedial action results in a serious delay in response.

[0004] Secondly, the inclinometers, laser rangefinders, mortar spots, and digital scales used in the correction project all belong to macroscopic deformation monitoring. They can only read data after the structure has undergone significant displacement or cracks (i.e., damage has already occurred), and cannot capture early signs of internal damage such as the initiation and propagation of microcracks in mortise and tenon joints.

[0005] Furthermore, the safety measures in the corrective engineering project (such as observers and record sheets) are all reactive. When dangerous situations such as bending of the support steel pipes or lifting of anchor bolts occur, the project can only be stopped immediately and remedial measures discussed. This indicates that the existing corrective engineering project lacks the ability to predict and avoid disasters.

[0006] Finally, the correction engineering lacks the ability to identify the real-time characteristics of the structure. The entire process relies on preset fixed thresholds and cannot dynamically adjust the loading strategy based on the real-time changes in the mechanical properties of the structure during the stress process (such as the dynamic structural admittance matrix).

[0007] In summary, there is an urgent need for an intelligent control method that can incorporate real-time damage signals (such as acoustic emission AE) into a closed loop to achieve predictive avoidance and real-time rejection, in order to resolve the contradiction between identification accuracy and operational safety. Summary of the Invention

[0008] This invention provides a method for real-time monitoring and intelligent control of ancient buildings, aiming to solve the problem of fragmented monitoring and control in existing technologies, and the difficulty in achieving precise attitude control while ensuring safety (especially brittle damage).

[0009] The technical solution provided by this invention includes the following steps: S1. Initialization and parameter setting: Set the target posture of the ancient building structure, AE warning baseline and AE danger threshold; The target posture can be derived from existing engineering practices, such as by using a 3D laser scanner (e.g., Focus SPlus 150) to collect comprehensive data on the ancient building structure, obtain deformation data, and calculate it in conjunction with the ideal state.

[0010] S2. Dual-modal D-SAM identification based on acoustic emission rate gating: The real-time AE event rate of the ancient building structure is collected and compared with the AE warning baseline to switch between active identification mode and passive identification mode, thereby updating and outputting a dynamic structural admittance matrix. S3. Online calibration of damage stiffness correlation: Calculate the stiffness degradation rate based on the dynamic structural admittance matrix output in S2, and combine it with the AE event rate to calibrate and output the brittleness coefficient and AE baseline of the ancient building structure online; S4. Predictive feedforward control under dynamic constraints: Combining the target attitude, AE hazard threshold, dynamic structural admittance matrix, and the brittleness coefficient and AE baseline of the ancient building structure, a constrained optimization problem with the AE hazard threshold as the constraint is constructed and solved to output an optimal force increment for the regulation of the ancient building structure. S5. Continuous closed-loop iteration and parallel safety veto monitoring: Execute the optimal force increment to the ancient building structure and return to step S2 in a loop; and in parallel, use the AE danger threshold to monitor the measured AE event rate of the ancient building structure so as to issue the highest priority veto instruction when the limit is exceeded.

[0011] In one specific embodiment, the active identification mode in S2 specifically includes: When the real-time AE event rate is lower than the AE warning baseline, the system determines that the structure is in a safe state. At this time, a preset micro-perturbation force signal is applied to the ancient building structure. The system collects the micro-perturbation force signal and the corresponding micro-vibration response, and uses the recursive least squares (RLS) algorithm to recursively update the dynamic structure admittance matrix online based on the above signal and response data.

[0012] In one specific embodiment, the passive identification mode in S2 specifically includes: when the real-time AE event rate is greater than or equal to the AE warning baseline, the system determines that the structure has entered a warning state, and at this time, the application of the micro-perturbation force signal is immediately stopped to avoid additional disturbances; The data input source for the identification algorithm is switched to the main loading force and macroscopic displacement of the ancient building structure collected; and the parameter estimates from the previous moment are retained, and the recursive estimation algorithm is used to update the dynamic structural admittance matrix to ensure the continuity of the identification process.

[0013] Preferably, the step of calculating the stiffness degradation rate in S3 specifically includes: calculating the difference matrix between the dynamic structural admittance matrix at the current moment and the dynamic structural admittance matrix at the previous moment; calculating the Frobenius norm of the difference matrix; and dividing the Frobenius norm value by the time interval to obtain the stiffness degradation rate.

[0014] Preferably, the core principle of the online calibration of the brittleness coefficient and AE baseline in step S3 lies in establishing a damage signal (AE event rate). ) and structural physical degradation (stiffness degradation rate) The dynamic relationship between them. This step specifically includes: Establish a linear adaptive model to represent and Relationships, for example ,in The brittleness coefficient is mentioned. The AE baseline; Calculate the estimation error between the measured AE event rate and the estimated value of the linear adaptive model; The Least Mean Square (LMS) algorithm is employed, utilizing the estimation error and the stiffness degradation rate. Adaptively iteratively update the brittleness coefficient and the AE baseline .

[0015] In one specific embodiment, before solving the constrained optimization problem, step S4 further includes constructing a multi-step state prediction chain, which is used to predict future states caused by candidate force increments. This prediction chain specifically includes: Displacement prediction: Based on the dynamic structural admittance matrix and the current macroscopic displacement of the ancient building structure, predict the future displacement of the ancient building structure; Degradation prediction: Using an online identified stiffness degradation model, the future stiffness degradation rate caused by the optimal force increment and the current macroscopic displacement of the ancient building structure is predicted. ; Damage prediction: based on the brittleness coefficient calibrated in step S3. The AE baseline and the predicted future stiffness degradation rate Through the linear adaptive model (i.e. Predict the future AE event rate of the ancient building structure. .

[0016] Preferably, in the above prediction chain, the online identified stiffness degradation model is a nonparametric regression model. This implementation includes: constructing a historical database to store historical state-action pairs (i.e., historical displacements and historical force increments) and their corresponding measured stiffness degradation rates; When a prediction query (i.e., the current displacement and candidate force increment) is received, the distance between the query point and the data points in the historical database is calculated. A local weighted regression or kernel regression algorithm is then used to calculate weights based on these distances, and a weighted average of the historical stiffness degradation rates is applied to obtain the predicted future stiffness degradation rate. .

[0017] Preferably, the constrained optimization problem in S4 aims to minimize the error between the predicted future displacement and the target attitude. Its constraints, besides being based on the predicted future AE event rate... Not exceeding the AE hazard threshold In addition to core security constraints, it also includes: Strain constraints: require that the strain of the ancient building structure corresponding to the predicted future displacement of the ancient building structure does not exceed a preset strain safety threshold; and actuator physical constraints: including constraints on the magnitude of the optimal force increment (force increment constraint) and constraints on the applied total force (total force constraint).

[0018] In one specific embodiment, the highest priority rejection instruction in S5 is a parallel safety mechanism independent of the predictive control loop. This instruction includes at least one of the following: Hold command: Commands all actuators to immediately stop operating and lock in the current state; or Unload command: Commands all actuators to synchronously reduce output force at a preset rate.

[0019] Preferably, the settings of each parameter in S1 are based on clear engineering principles: The AE warning baseline is determined based on environmental noise monitoring or materials science experiments to distinguish structural response from background noise; the AE hazard threshold is determined based on loading failure test data or historical engineering data to represent unacceptable damage levels to the structure.

[0020] This invention provides a method for real-time monitoring and intelligent control of ancient building alignment. It has the following beneficial effects: 1. This invention proactively avoids structural damage such as steel pipe bending in existing construction projects by predicting whether an action will cause the Action AE (Advanced Effect) to exceed the limit before execution. Through parallel safety veto monitoring, once the measured Action AE exceeds the limit, the system will issue an automatic veto command in milliseconds. Its speed and reliability exceed those of manual command in existing technologies, solving the safety risks caused by delayed manual response and operational errors. 2. This invention uses dual-modal D-SAM identification based on AE gating to identify the current dynamic structural admittance matrix of the structure in real time and proactively, while ensuring safety (low AE event rate). This means that the central control unit accurately knows the current mechanical characteristics of the structure at every moment of decision-making, thereby calculating the truly scientific optimal force increment, completely replacing the static threshold based on experience or guesswork in the existing technology, and realizing high-precision adaptive loading.

[0021] 3. The present invention, through the online calibration module for damage stiffness correlation, can update the brittleness coefficient and AE baseline characterizing the structural properties in real time and dynamically during the correction process. This ensures that the prediction model and identification model of the present invention can closely fit the current actual working conditions of the structure (including damage accumulation and creep effect), so that the prediction and control decisions can maintain high accuracy and effectiveness throughout the construction process that lasts for several days. Attached Figure Description

[0022] Figure 1 This is a system framework diagram of the present invention; Figure 2 This is a schematic diagram of the method flow of the present invention.

[0023] Among them, 100 is the perception and monitoring subsystem; 200 is the execution and actuation subsystem; 300 is the central control and processing unit; 310 is the D-SAM identification module; 320 is the mode switching module; 330 is the correlation calibration module; 340 is the predictive control module; and 350 is the security veto module. Detailed Implementation The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] See attached document Figure 1 This invention provides a real-time monitoring and intelligent control system for the lifting and straightening of ancient buildings. The system aims to provide predictive safety control for multi-point jacking operations during the lifting and straightening process of ancient building structures. The system may include: a sensing and monitoring subsystem 100, an execution and excitation subsystem 200, and a central control and processing unit 300.

[0025] A sensing and monitoring subsystem 100 is used to acquire multi-source physical signals of the ancient building structure in real time during the process of mortaring and alignment. This sensing and monitoring subsystem 100 may include: Multiple displacement sensors (such as fiber Bragg gratings (FBGs) or differential transformer low-voltage transformers (LVDTs) are deployed at key structural nodes. In this embodiment, draw-wire displacement sensors (range 0-500mm, accuracy 0.05mm) are specifically selected and installed at the column base and column head of the load-bearing column, respectively, to provide real-time, high-frequency closed-loop control feedback and output displacement vectors. ; Multiple acoustic emission (AE) sensors are deployed in areas of structural stress concentration or historical damage. Acoustic emission refers to the physical phenomenon where materials (especially timber and brick components of ancient buildings) release strain energy in the form of transient elastic waves when subjected to external or internal forces, resulting in deformation or fracture. In this embodiment, a wideband piezoelectric sensor (frequency response range 20-100kHz, preamplifier gain 40dB) is specifically selected to capture stress wave signals released during the initiation or propagation of microcracks in the material and output the AE event rate, which characterizes the precursory damage. ; Multiple force sensors, specifically spoke-type pressure sensors (range 500kN), are integrated on the top of the jack in the actuation and excitation subsystem 200 to measure the applied force vector. ; And multiple accelerometers, specifically low-frequency, high-sensitivity accelerometers, used to acquire dynamic response signals of the structure to perturbation excitations under specific identification modes. ; An execution and excitation subsystem 200 is used to apply jacking control forces to the ancient building structure and identify the required perturbation excitations. This execution and excitation subsystem 200 may include: Multiple hydraulic jacks or servo electric cylinders serve as lifting actuators. In this embodiment, four synchronous hydraulic jacks are used, with their stroke and thrust controlled by the central control and processing unit 300. Multiple high-frequency servo valves or actuators, specifically electro-hydraulic proportional servo valves (response frequency > 20Hz), are linked with the actuators to control the force vector. High-frequency, low-amplitude micro-perturbation signals are superimposed on the basis. .

[0026] In this embodiment, the central control and processing unit 300 employs an industrial control computer (IPC) with a real-time operating system and a high-speed data acquisition card (sampling rate set to 1kHz). It is electrically connected to the sensing and monitoring subsystem 100 via a data acquisition interface and to the execution and excitation subsystem 200 via a control command interface. The central control and processing unit 300 internally contains multiple functional modules for executing the method of this invention.

[0027] In one embodiment, the central control and processing unit 300 may include: D-SAM identification module 310 is used for real-time estimation of dynamic structure admittance matrix. This module receives excitation signals from the sensing and monitoring subsystem 100. or ) and response signal ( or It is configured to perform recursive estimation algorithms, such as recursive least squares (RLS), and a forgetting factor is set. , to be continuously updated The matrix coefficients.

[0028] The mode switching module 320 is used to dynamically adjust the working mode of the D-SAM identification module 310 according to the safety status of the structure. The mode switching module 320 receives the AE event rate in real time. and compare it with the preset warning baseline. Comparison: In this embodiment, the preset warning baseline Set to 20 times / second (based on ambient background noise calibration).

[0029] when In the (static) state, the mode switching module 320 instructs the D-SAM identification module 310 to operate in "active identification mode" and instructs the execution and excitation subsystem 200 to apply a perturbation force. (For example, white noise signals of 5-20Hz); when In the (alert state) state, the mode switching module 320 instructs the D-SAM identification module 310 to switch to "passive identification mode" and immediately stop. The application of.

[0030] The correlation calibration module 330 is used for online learning and calibration of the current fragility of ancient building structures. The correlation calibration module 330 receives the output from the D-SAM identification module 310. The output of the sequence and sensing and monitoring subsystem 100 And it is configured as: (a) Calculation The rate of change over time (e.g., by calculating the Frobenius norm of the matrix difference) yields the rate of stiffness degradation. ; (b) Based on the following model, an adaptive algorithm (such as LMS with a learning rate of 0.01) is used to estimate the fragility coefficient online. and AE baseline : ; in This is an estimate of the AE event rate.

[0031] (c) The correlation calibration module 330 can also be configured to identify a stiffness degradation model. Used to describe stiffness variation With load and displacement The relationship between them.

[0032] Predictive control module 340 is used to calculate the next optimal and safe force vector increment. The predictive control module 340 receives... Identified , calibrated , and model And based on the preset target posture and safety constraints : (a) Construct a state prediction chain to predict the applied state. After , as well as ; (b) Construct and solve a constrained optimization problem whose objective is to minimize and The error between them, and the constraints include (For example, set it to 150 times / second).

[0033] The safety veto module 350 provides the highest priority safety protection independent of the predictive control module 340. This safety veto module 350 receives measured data directly from the sensing and monitoring subsystem 100. And compare it with the "danger threshold". Comparison. When When this occurs, the safety veto module 350 immediately bypasses the predictive control module 340 and issues a "hold" or "unload" veto command to the execution and excitation subsystem 200 (such as power-off locking of the servo valve).

[0034] During system operation, the sensing and monitoring subsystem 100 continuously collects data and sends it to the central control and processing unit 300. The D-SAM identification module 310, mode switching module 320, and correlation calibration module 330 within the central control and processing unit 300 work together to provide real-time model parameters to the predictive control module 340. The predictive control module 340 calculates... The instruction is sent to the execution and activation subsystem 200 for execution. The safety veto module 350 monitors the entire process in parallel to ensure that the system does not exceed the danger limit of damage under any circumstances.

[0035] See attached document Figure 2 This figure is a flowchart of a method according to an embodiment of the present invention. The method for real-time monitoring and intelligent control of ancient buildings provided by the present invention is executed by a central control and processing unit 300. The execution process includes multiple steps, which are described in detail below. S1. Initialization and parameter setting steps.

[0036] In this step, the central control and processing unit 300 establishes a time-varying linear system model to describe the dynamic relationship between the force vectors applied by multiple lifting points (corresponding to the execution and excitation subsystem 200) and the displacement vectors measured by M key monitoring points (corresponding to the sensing and monitoring subsystem 100). This model is expressed as: ; in It is the N-dimensional input force vector measured by the force sensor (in this embodiment, a spoke-type pressure sensor with a range of 500kN is selected). The output displacement vector is M-dimensional, measured by a displacement sensor (such as a fiber optic grating (FBG) or a differential transformer (LVDT). In this embodiment, an MPS series draw-wire displacement sensor with an accuracy of 0.05 mm is specifically selected.

[0037] It is the M×N dimensional dynamic structure admittance matrix (D-SAM) to be identified. The physical meaning lies in its matrix elements Characterized in At that moment, the The force input at the apex point affects the first... The real-time influence coefficient of the displacement output of each monitoring point, i.e. the flexibility of the structure; this matrix is ​​time-varying, and its coefficients will change with the redistribution of stress, damage evolution or misalignment of tenons and mortises within the structure.

[0038] It is an M-dimensional vector of system noise and unmodeled dynamics (such as higher-order modes or nonlinearities).

[0039] During the initialization and parameter setting steps, the central control and processing unit 300 also receives or sets a series of control targets and safety constraint parameters, which are stored and used for calculations in subsequent modules.

[0040] These parameters include an M-dimensional target attitude vector. M-dimensional target attitude vector This represents the final displacement state that the ancient building structure is expected to achieve after the straightening and adjustment work is completed. The target attitude vector can be pre-input into the central control and processing unit 300 by the operator according to the engineering design plan. For example, the straightening target of monitoring point Z1 is set to move 120mm in the northwest direction along the horizontal axis.

[0041] These parameters include a strain safety threshold. The strain safety threshold is a scalar or vector quantity. It defines the maximum allowable strain at critical monitoring points of the structure. In this embodiment, based on the longitudinal compressive strength of ancient pine wood, the strain safety threshold is... Set to 0.002 (i.e., 2000 microstrain), strain safety threshold. With displacement vector Related, for example, through a known displacement-strain transformation matrix ,Right now .

[0042] When the strain is predicted in subsequent steps Exceeding the strain safety threshold When this occurs, control constraints will be triggered; these parameters also include critical acoustic emission (AE) rate thresholds for grading damage risk, including an acoustic emission (AE) warning baseline. and an acoustic emission (AE) hazard threshold .

[0043] It is a low AE event rate threshold used to distinguish the "silent state" of a structure. ) and "alert state" In its specific implementation, this The value can be determined in one or more of the following ways: (a) Before the alignment operation begins, environmental noise monitoring is performed on the structure to obtain the background noise level of the AE signal, and a safety margin is added on this basis; (b) Based on materials science experiments, determine the typical AE (Advanced Effects) event rate of this type of material (such as aged wood or masonry) in the early stages of plastic deformation or microcrack initiation. In this embodiment, the background noise measured by on-site silent testing is approximately 5-8 times / second. Considering system redundancy, [the following is omitted as the original text is incomplete and requires further context]. The specific setting is 20 times per second.

[0044] Acoustic emission (AE) hazard threshold It is a relatively high AE event rate threshold, and its value is always greater than ,Should This indicates that structural damage is accelerating or is about to enter a critical state of instability. In practical implementation, this... The value can be determined in one or more of the following ways: (a1) Based on the loading failure test data of similar materials, the AE event rate corresponding to the yield point or the inflection point of rapid damage propagation is determined. (b1) The lower limit of the event rate of AEs (Advanced Effects) leading to visible damage or failure of the structure, obtained from historical engineering data. In this embodiment, destructive experimental data of similar decaying wood are referenced. The specific setting is 150 times per second, which corresponds to the acoustic emission characteristic frequency when wood fibers begin to undergo macroscopic fracture.

[0045] and Once set, it is stored in the central control and processing unit 300 and used to trigger the mode switching module 320 and the security veto module 350, respectively.

[0046] S2: Dual-modal D-SAM identification based on acoustic emissivity gating Following the initialization and parameter setting steps, this method performs dual-modal D-SAM identification based on acoustic emission rate gating, namely, active identification mode (corresponding to the silent state) and passive identification mode (corresponding to the alert state). The initiation and mode selection of this step are implemented by the mode switching module 320 deployed in the central control and processing unit 300. In this embodiment, this process is executed periodically at a frequency of 10Hz (i.e., once every 0.1 seconds) in the real-time calculation loop of the central controller.

[0047] The mode switching module 320 is used to dynamically and automatically adjust the operating mode of the D-SAM identification module 310 according to the safety status of the structure. The mode switching module 320 is configured to perform the following sub-steps: S2211, Data Reception: The mode switching module 320 obtains the latest measured AE event rate from the sensing and monitoring subsystem 100 in real time via the internal data bus of the central control and processing unit 300. .

[0048] S2212, Status Comparison: The mode switching module 320 compares the AE event rate received in real time in S2211. Compared with the AE warning baseline set in the initialization step (In this embodiment, the number of comparisons is 20 times per second) to perform logical comparisons.

[0049] S2213. Generation and distribution of switching instructions: Based on the comparison results of S2212, the mode switching module 320 generates a set of mode switching instructions. The switching instructions take effect inside the central control and processing unit 300 and are simultaneously sent to the execution and excitation subsystem 200.

[0050] When the comparison result of S2212 is When the mode switching module 320 determines that the structure is in a "silent state", the mode switching module 320 executes the following: (a) Send an active identification mode command to the D-SAM identification module 310.

[0051] (b) Send an incentive enable command to the execution and incentive subsystem 200.

[0052] When the comparison result of S2212 is At this time, the mode switching module 320 determines that the structure has entered the "alert state". At this point, the mode switching module 320 immediately executes: (a) Send a passive identification mode command to the D-SAM identification module 310.

[0053] (b) Send an incentive prohibition command to the execution and incentive subsystem 200.

[0054] S2221, Perturbation Excitation Application: The central control and processing unit 300 sends an excitation enable command to the execution and excitation subsystem 200, and the execution and excitation subsystem 200 applies the force vector... Based on this, a preset, low-amplitude N-dimensional micro-perturbation force signal is superimposed. In this embodiment, the micro-perturbation force signal The signal is designed as Gaussian white noise, with a frequency bandwidth set at 5Hz-30Hz and an amplitude set at 2% of the main lifting force (approximately 1-2kN). This signal is converted into an analog voltage signal via a D / A module to drive the high-frequency port of the electro-hydraulic servo valve. S2222, Micro-vibration response acquisition: The sensing and monitoring subsystem 100 synchronously acquires data from micro-perturbation forces. M-dimensional micro-vibration response caused Micro-vibration response After low-pass filtering (cutoff frequency 50Hz), and... Sample values They were sent together to the D-SAM identification module 310.

[0055] S2223, Model Discretization: The D-SAM identification module 310 uses a discrete-time model to process the acquired data. The system model under active identification mode can be represented as: ; in For discrete time steps, The N-dimensional perturbation force signal applied in S2221 The sampled values, M-dimensional micro-vibration response collected in S2222 The sampled values, For measuring noise.

[0056] S2224, Parameter Recursive Estimation: The D-SAM identification module 310 identifies each... (from 1 to Execute the following RLS iterative algorithm: 1. Define the regression vector and the vector of parameters to be estimated : ; ; 2. Calculation Prediction error at time : ; in yes Always The estimated value.

[0057] 3. Calculate the RLS gain vector : ; in yes The N×N covariance matrix at time t, It is the forgetting factor (a scalar close to 1, such as 0.995), used for tracking. The time-varying characteristics; For regression vectors, Indicates to The matrix transpose.

[0058] 4. Update Parameter vector estimate at time step : ; 5. Update Covariance matrix at time step : ; in It is an N×N dimensional identity matrix.

[0059] S2225, Matrix Reconstruction: D-SAM Identification Module 310 in At time S2224, the M N×1 parameter vectors computed in parallel will be... By combining the data, a complete estimate of the M×N dimensional dynamic structural admittance matrix can be reconstructed. .

[0060] When the mode switching module 320 determines in step S2212 that the structure has entered the "alert state" (i.e. When this occurs, the mode switching module 320 immediately issues a passive identification mode command and an excitation disable command in step S2213. The central control and processing unit 300 will then execute the following steps: S2231, Excitation Stop: The execution and excitation subsystem 200 receives an excitation disable command. It immediately stops the high-frequency servo valve or exciter from generating micro-perturbation signals. That is, forced The specific operation is as follows: the controller clamps the analog output voltage of the corresponding servo valve perturbation channel to 0V.

[0061] S2232, Data Source Switching: Simultaneously, the D-SAM identification module 310 receives a passive identification mode command. This command triggers the D-SAM identification module 310 to immediately switch the data input source for its algorithm.

[0062] In one specific implementation, the switching is seamless, meaning that the internal state of the recursive estimation algorithm (such as RLS described in S2224) running in the D-SAM identification module 310, including the parameter estimates from the previous time step, is seamless. or Covariance Matrix All are retained and will not be reinitialized.

[0063] S2233, Algorithm Input Redirection: The D-SAM identification module 310 redirects its data input port from the sampled value of the perturbation signal. and the sampled values ​​of the micro-vibration response Switch to the main load signal for correction. Specifically: (a) The input vector of the algorithm, from Switch to . It is the N-dimensional main loading force vector measured by the force sensor (spoke pressure sensor) in the sensing and monitoring subsystem 100.

[0064] (b) The algorithm's output vector, from Switch to . It is an M-dimensional macroscopic displacement vector measured by displacement sensors (such as FBG or LVDT) in the sensing and monitoring subsystem 100.

[0065] S2234: Recursive estimation of passive mode parameters.

[0066] by Taking the nth parallel RLS estimator as an example, for the nth... Output (Macroscopic displacement vector) The (each component), the D-SAM identification module 310 performs the following iterations: 1. Define a new regression vector and the vector of parameters to be estimated ( (The definition remains unchanged) ; .

[0067] 2. Calculation Prediction error at time ,at this time These are macroscopic displacement measurements: ; in It is an estimate inherited from the previous moment (which could be the last moment of active mode or the moment before passive mode).

[0068] 3. Calculate the RLS gain vector ,at this time It is the main loading force vector : ; 4. Update Parameter vector estimate at time step : ; 5. Update Covariance matrix at time step : ; S2235: Matrix Reconstruction and Output. The D-SAM identification module 310 converts the M parameter vectors calculated in S2234 into... Reconstructed into complete 3D matrix The results are then output to the correlation calibration module 330 and the prediction control module 340.

[0069] S3. Online calibration of damage stiffness correlation: In the central control and processing unit 300, the correlation calibration module 330 is connected to the D-SAM identification module 310.

[0070] Correlation calibration module 330 in each control cycle Real-time reception of the latest estimated dynamic structure admittance matrix output by the D-SAM identification module 310 (This matrix is ​​M×N dimensional, and in this embodiment it is 8×4 dimensional). The first step of the correlation calibration module 330 is to perform a quantitative calculation of the stiffness degradation rate.

[0071] The purpose of this step is to The time-varying properties of the matrix sequence are transformed into a one-dimensional scalar signal. This scalar signal It is used to quantitatively characterize the degree of drastic change in the overall flexibility (or its reciprocal, i.e., stiffness) of a structure over time.

[0072] In one specific implementation, the correlation calibration module 330 performs the following sub-steps: S2311, Data Reception and Storage: In At any given time, the correlation calibration module 330 receives the estimated value of the current dynamic structure admittance matrix output by the D-SAM identification module 310. and retrieve the previous time step from its internal memory. Stored dynamic structure admittance matrix estimate .

[0073] S2312, Matrix Difference: Calculate the difference matrix between the current matrix and the matrix from the previous time step. : ; Should Matrix Each element reflects a control cycle The change in each component of the structural flexibility matrix.

[0074] S2313, Scalarization of Change Amplitude: To obtain a representation The scalar value of the overall change magnitude is used by the correlation calibration module 330 to calculate the difference matrix. The Frobenius norm is denoted as , is defined as the square root of the sum of the squares of all elements in a matrix, and is calculated as follows: ; in and They are and In the line, number The elements of the column.

[0075] S2314, Rate Calculation: The correlation calibration module 330 calculates the rate obtained in S2313. The value, divided by Time's up Time interval (i.e., control period or sampling time interval) to obtain Stiffness degradation rate at time : ; In this embodiment, the control cycle Set to 0.1 seconds.

[0076] S2315, Data Output: The correlation calibration module 330 will output the calculated stiffness degradation rate. AE event rate synchronized with the sensing and monitoring subsystem 100 Together, these serve as input data for performing subsequent brittleness coefficient calculations. Adaptive calibration steps.

[0077] The stiffness degradation rate is calculated in step S2314. Subsequently, the correlation calibration module 330 in the central control and processing unit 300 continues to perform the adaptive calibration step of the "brittleness coefficient".

[0078] The purpose of this step is to learn online and quantify the current brittleness of the structure, i.e., the strength of the correlation between stiffness degradation (increased flexibility) and damage signals (AE events).

[0079] S2321. Correlation Model Establishment: The correlation calibration module 330 establishes a linear adaptive model to describe... (as model input) and The dynamic relationship between (as model output). Linear adaptive models are used to... Time, based on Parameters at time, estimation Model estimates at time 1 : ; in It is the stiffness degradation rate. yes The brittleness coefficient estimated at time is a scalar whose physical meaning is to characterize the "AE event rate increment" corresponding to the "unit stiffness degradation rate". The larger the value of the coefficient, the more "brittle" the structure is (i.e., a small change in stiffness is accompanied by a large number of AE events). yes The "AE baseline noise" estimated at time step is a scalar whose physical meaning lies in characterizing the situation when the stiffness remains unchanged (i.e., ...). During system initialization, the structure's own AE background event rate. The initial value is set to 0. The initial value is set to the ambient background noise level (e.g., 5). S2322, Estimation Error Calculation: Correlation Calibration Module 330 Calculation Real-time measured AE event rate Compared with the model estimates obtained in S2321 estimation error between : ; Substituting the linear adaptive model of S2322, we get: .

[0080] S2323, Brittleness coefficient Update: The update and iteration formula is as follows: ; in It is a preset, used for The learning rate (or step size) is used. In this embodiment, to ensure the stability of parameter convergence and avoid drastic fluctuations in coefficients due to occasional noise, It is a small positive number whose value determines Track and adapt The speed. The meaning of this update formula is to utilize the estimation error. Multiply by the input corresponding to the error , to adjust parameters Make corrections.

[0081] S2324, AE baseline Update: The update and iteration formula is as follows: ; in It is a preset, used for The learning rate. In this embodiment, Set to 0.05, slightly greater than In order to adapt more quickly to changes in ambient background noise. It is also a small positive number. The update formula (corresponding to...) The constant term in the model, whose input can be considered as "1") utilizes the estimation error Directly to parameters Make corrections.

[0082] S2325, Parameter Output: The correlation calibration module 330 will output the calculated parameters. Latest fragility coefficient and AE baseline The output is stored in the memory of the central control and processing unit 300 for the predictive control module 340 to call when constructing the state prediction chain in subsequent steps.

[0083] To support subsequent predictive control steps, the correlation calibration module 330 in the central control and processing unit 300 can also be configured to identify a stiffness degradation model online while executing steps S2311 to S2325. .

[0084] The stiffness degradation model Its function is to predict the "stiffness degradation rate" that will result from a "candidate" future control action and the current structural state.

[0085] In one specific implementation, this stiffness degradation model It is implemented as a non-linear, instance-based regression model (e.g., locally weighted regression (LWR) or kernel regression), which is achieved through steps S2331 to S2335: S2331, Model Input and Output Definition: This model... A method was established from the input vector To output scalar The mapping relationship.

[0086] (a) Input vector of the model Defined as A "state-action" pair at any given moment. In one embodiment, this input vector... The dimension is (N+M)×1 (i.e., 4+8=12 dimensions), by Increment of N-dimensional force vector applied at time t and M-dimensional displacement vector at time t It is pieced together: ; Model output scalar Defined as by Caused by, in The rate of stiffness degradation occurring at any given moment, i.e.: ; The stiffness degradation model The identification objective is to learn this mapping relationship. .

[0087] S2332, Online Database Construction: Correlation calibration module 330 in each control cycle of the alignment operation Continuously pair its (input, output) data with Store in an internal historical database .

[0088] Specifically, in At that moment, the correlation calibration module 330 retrieves from memory Moment ,Right now and And obtain the result calculated in step S2314. Stiffness degradation rate at time ,Right now Then As a data point Add to database middle.

[0089] S2333, Receiving Predictive Query: In the subsequent predictive control step (S2411), the predictive control module 340 evaluates a "candidate force vector increment". At that time, a prediction query will be sent to the relevance labeling module 330. This query vector The increment of the candidate force vector and the current measured displacement vector Its composition, expressed as follows: ; S2334. Calculate local weights: The correlation calibration module 330 receives the query. Then, iterate through its internal database. All historical data points The relevance calibration module 330 calculates the query vector. With each historical input Distance between (For example, Euclidean distance) ).

[0090] Subsequently, the correlation calibration module 330 uses a kernel function to measure the distance. Convert to a weight .

[0091] In one implementation, a Gaussian kernel function is used: ; in It is a preset scalar, called kernel bandwidth, used to control the query vector. The influence range of nearby data points; in this embodiment, the core bandwidth Setting it to 1.0 (the normalized value) means that only historical data that is very close to the current state in the state space will have a significant impact on the prediction. Representing the base of the natural logarithm An exponential function with base 0.

[0092] S2335. Calculate the weighted predicted value: The correlation calibration module 330 uses the weights calculated in S2334. For the database All historical outputs (i.e., historical) Perform a weighted average to obtain the results. Predicted value of the stiffness degradation rate at time step : ; S4. Predictive feedforward control under dynamic constraints: S2411, Displacement State Prediction: Based on the principle of linear superposition, calculate the incremental candidate force vector. After that, Predicted displacement vector at time 1 The calculation formula is as follows: ; in It is the predicted M-dimensional displacement vector. It has been identified. Dimensional flexibility matrix It is the N-dimensional candidate force increment vector to be evaluated.

[0093] S2412, Stiffness Degradation Rate Prediction: This involves predicting the current "state-action" pair, i.e., the current measured displacement vector. With candidate force vector increment This is combined into an N+M dimensional query vector. .

[0094] Subsequently, the predictive control module 340... As input, query the stiffness degradation model constructed in step S3. In order to obtain the Predicted value of the stiffness degradation rate at time step : ; S2413, Damage Rate Prediction: Utilizing the predicted stiffness degradation rate from step S2412. And the latest brittleness parameters, calculated in Predicted AE event rate at any time Its calculation formula is based on the correlation model established in S2322: ; For example, if the current moment is identified , Furthermore, the model predicts that a certain candidate force increment will lead to a stiffness degradation rate. The system predicts that this action will result in an AE rate of 50 × 0.8 + 10 = 50 times / second.

[0095] After executing the multi-step state prediction chain from S2411 to S2413, the predictive control module 340 continues with a constraint optimization step to solve for the current control cycle. Optimal force vector increment .

[0096] The constrained optimization step is achieved by constructing and solving a mathematically constrained optimization problem. In this embodiment, the problem is constructed as a quadratic programming problem.

[0097] S2421. Optimization Objective Function Construction: Finding a Force Vector Increment This causes the predicted displacement vector resulting from the force vector increment. With the final target posture The error between them is minimal.

[0098] Predictive control module 340 constructs a system based on The objective function with independent variable In one specific implementation, the objective function is defined as the square of the Euclidean norm of the difference between the predicted displacement and the target displacement (i.e., the least squares error): ; Substituting the displacement prediction formula in S2411, the objective function can be explicitly expressed as about Functions: ; in This is the currently measured displacement vector. This is an estimate of the dynamic structure admittance matrix. The preset target attitude vector; Indicate the optimization objective; Let be the objective function, which is a function about The quadratic form function.

[0099] S2422, Constraint Set Construction: To ensure the safety and feasibility of control actions, the predictive control module 340 constructs a constraint set containing multiple inequality constraints. All candidate constraints... All conditions in this set must be met.

[0100] This set of constraints includes: Dynamic safety constraints, their mathematical expression is: ; Substituting the prediction formulas for S2412 and S2413, the constraint can be explicitly expressed as about Functions: ; In this embodiment, the constraint is specified as follows: the predicted AE event rate must be less than 150 times / second.

[0101] Strain constraints are expressed as: ; Substituting into the displacement prediction formula, we obtain about Linear inequality constraints: ; in The displacement-strain transformation matrix is ​​determined by the sensor's geometric position. The preset safety strain threshold is set to 0.002 in this embodiment.

[0102] Actuator physical constraints. This set of constraints ensures that control commands are physically executable, including: Force increment constraint, expressed as: ; in It is the first of the force increment vectors One portion, It is the first The maximum single-step increment allowed for each actuator. To prevent shocks, this embodiment will... The setting is 5kN (meaning that the force applied in a single step does not exceed 500 kg).

[0103] Total force constraint: Ensures that the total applied force is within the actuator's range.

[0104] ; in It is the total force from the previous step. and It is the first The minimum and maximum output forces of each actuator. In this embodiment, the jack range... , .

[0105] S2423, Optimal Command Output: Solve the objective function in S2421 under the constraints in S2422 to obtain the optimal force increment vector. The solution is obtained iteratively using the interior point method or the effective set method algorithm library, and the solution time is usually less than 10ms.

[0106] Predictive control module 340 will As the final control command of the current control cycle, it is output to the execution and excitation subsystem 200 for execution.

[0107] S5. Through an adaptive closed-loop iterative process, the ancient building structure is continuously corrected.

[0108] S2511: In At that moment, the execution and incentive subsystem 200 executed the force increment. After that, the system entered the... One control cycle.

[0109] S2512: In At the start of the cycle, the sensing and monitoring subsystem 100 collects and uploads new structural state data, including new displacement vectors. New force vector And the new AE event rate .

[0110] S2513: The central control and processing unit 300 received... After receiving the new data at a given time, the aforementioned identification, calibration, and predictive control steps are automatically and cyclically re-executed to calculate the next optimal force increment. To ensure real-time performance, the entire computation process is strictly constrained to be completed within a single control cycle (100ms in this embodiment). S2514: This It is sent to the execution and stimulus subsystem 200 for execution. The system then enters the next step. Each control cycle repeats steps S2512 to S2514. This iterative process continues until the error between the displacement vector of the monitoring point and the target attitude vector is less than a preset engineering tolerance (e.g., ±2mm). At this point, the system determines that the correction is complete, automatically exits the closed-loop control, and locks the current state. Simultaneously with the entire closed-loop iterative process, a highest-priority safety veto mechanism monitors the system in parallel and independently. This mechanism is implemented by the safety veto module 350 in the central control and processing unit 300.

[0111] S2521, Independent Monitoring: The safety veto module 350 does not rely on any prediction results from the predictive control module 340, but instead receives the measured AE event rate directly and continuously (e.g., at a frequency much faster than the control cycle) from the sensing and monitoring subsystem 100. .

[0112] S2522, Critical Comparison: The safety veto module 350 will perform real-time... Compared with the AE hazard threshold set in the initialization step, which represents a dangerous state. Conduct continuous comparisons.

[0113] S2523, Rejection Instruction Generation: Once the security rejection module 350 detects... If this situation occurs, it will immediately generate a "safety veto instruction" with the highest system priority. This instruction is triggered directly through the hardware interrupt line, bypassing the operating system's scheduling queue, with a response time of less than 1ms.

[0114] S2524, Instruction Overriding and Execution: The security denial instruction is directly sent to the execution and activation subsystem 200. In one specific implementation, the security denial instruction can be one of the following: (a) Holding Command: All actuators in the command execution and excitation subsystem 200 immediately cease all movement and lock at their current position or force output state. Specifically, the servo valve power supply is instantly cut off, and the cylinder chamber is sealed using a hydraulic lock. This instantly freezes the structural state, preventing further damage.

[0115] (b) Unloading Command: The command execution and excitation subsystem 200 begins to synchronously reduce the output force of all actuators according to a preset, slow, and safe rate program to actively eliminate stress on the structure and bring it into a safe state. In this embodiment, the safe unloading rate is set to reduce the current total load by 5% per second (i.e., linear decay) to prevent structural rebound impact caused by sudden unloading.

Claims

1. A method for real-time monitoring and intelligent control of ancient buildings to correct and adjust their structure, characterized in that: Includes the following steps: S1. Initialization and parameter setting: Set the target posture of the ancient building structure, AE warning baseline and AE danger threshold; S2. Dual-modal D-SAM identification based on acoustic emission rate gating: The real-time AE event rate of the ancient building structure is collected and compared with the AE warning baseline to switch between active identification mode and passive identification mode, thereby updating and outputting a dynamic structural admittance matrix. S3. Online calibration of damage stiffness correlation: Calculate the stiffness degradation rate based on the dynamic structural admittance matrix output in S2, and combine it with the AE event rate to calibrate and output the brittleness coefficient and AE baseline of the ancient building structure online; S4. Predictive feedforward control under dynamic constraints: Combining the target attitude, AE hazard threshold, dynamic structural admittance matrix, and the brittleness coefficient and AE baseline of the ancient building structure, construct and solve a constrained optimization problem with the AE hazard threshold as the constraint, and output an optimal force increment for the regulation of the ancient building structure. S5. Continuous closed-loop iteration and parallel safety veto monitoring: Execute the optimal force increment to the ancient building structure and return to step S2 in a loop; In parallel, the measured AE event rate of the ancient building structure is monitored using the AE danger threshold, so as to issue the highest priority veto command when the limit is exceeded.

2. The method for real-time monitoring and intelligent control of ancient buildings according to claim 1, characterized in that, The active identification mode specifically includes: applying a micro-perturbation force signal and using the recursive least squares method to recursively estimate the dynamic structural admittance matrix based on the micro-perturbation force signal and the collected micro-vibration response of the ancient building structure.

3. The method for real-time monitoring and intelligent control of ancient buildings for adjusting and correcting structures according to claim 1, characterized in that, The passive identification mode specifically includes: stopping the application of micro-perturbation force signals, switching the data input source of the identification algorithm to the main loading force and macroscopic displacement of the ancient building structure, and retaining the parameter estimates from the previous moment to continue updating the dynamic structure admittance matrix.

4. The method for real-time monitoring and intelligent control of ancient buildings for adjusting and correcting structures according to claim 1, characterized in that, The calculation of stiffness degradation rate specifically includes: Calculate the difference matrix between the dynamic structure admittance matrix at the current time and the dynamic structure admittance matrix at the previous time. Calculate the Frobenius norm of the difference matrix; The stiffness degradation rate is obtained by dividing the Frobenius norm value by the time interval.

5. The method for real-time monitoring and intelligent control of ancient buildings for adjusting and correcting structures according to claim 1, characterized in that, The steps for online calibration of the structural brittleness coefficient and AE baseline of the ancient building specifically include: Establish a linear adaptive model; Calculate the estimation error between the AE event rate and the estimated value of the linear adaptive model; The Least Mean Square (LMS) algorithm is used to adaptively update the brittleness coefficient and the AE baseline using the estimation error and the stiffness degradation rate.

6. The method for real-time monitoring and intelligent control of ancient buildings according to claim 1, characterized in that, In the predictive feedforward control under dynamic constraints, before solving the constrained optimization problem, a multi-step state prediction chain is constructed, specifically including: Based on the dynamic structural admittance matrix and the current macroscopic displacement of the ancient building structure, predict the future displacement of the ancient building structure. Using an online identified stiffness degradation model, the future stiffness degradation rate caused by the optimal force increment and the current macroscopic displacement of the ancient building structure is predicted; Based on the brittleness coefficient, the AE baseline, the future predicted displacement, and the future stiffness degradation rate, the future predicted AE event rate of the ancient building structure is predicted.

7. The method for real-time monitoring and intelligent control of ancient buildings for adjusting and correcting structures according to claim 6, characterized in that, The steps of the online identification of the stiffness degradation model include: Construct a historical database to store historical state-action pairs and historical stiffness degradation rates; When a predictive query is received, the distance between the query point and the data point in the historical database is calculated; The system employs local weighted regression or kernel regression, calculates weights based on the distance between the query point and data points in the historical database, and performs a weighted average of the historical stiffness degradation rates to obtain the predicted future stiffness degradation rate.

8. The method for real-time monitoring and intelligent control of ancient buildings for adjusting and correcting structures according to claim 6, characterized in that, The constraints of the constrained optimization problem also include: The strain constraint requires that the strain of the ancient building structure corresponding to the predicted future displacement does not exceed a preset strain safety threshold. And actuator physical constraints, including force increment constraints and total force constraints.

9. The method for real-time monitoring and intelligent control of ancient buildings for adjusting and correcting structures according to claim 1, characterized in that, The highest priority veto instructions include: The hold command instructs all actuators to immediately stop moving and lock in their current state. Alternatively, an unload command can be issued, instructing all actuators to synchronously reduce their output force at a preset rate.

10. The method for real-time monitoring and intelligent control of ancient buildings for adjusting and correcting structures according to claim 1, characterized in that, In the initialization and parameter setting steps, the AE warning baseline is determined based on environmental noise monitoring or materials science experiments; The AE hazard threshold is determined based on load damage test data or historical engineering data.

Images