A method and system for simulating concrete deterioration under sustained load-chloride ion coupling

Abstract

The application relates to a concrete deterioration simulation method and system under the coupling of sustained load and chloride ions, and solves the problems that the overall mechanical performance deterioration of concrete structures under the coupling of load and chloride ions cannot be systematically studied, and the deterioration of concrete structures in actual service environments cannot be comprehensively reflected. The method comprises the following steps: a sustained load is applied to a test block, a solution is configured according to a stress type to determine a chloride ion diffusion direction, an electric field is applied to accelerate migration and maintain test stability, the penetration depth is measured after the experiment, and the diffusion coefficient is calculated; in the process, multidimensional data such as cracks and acoustic emission are collected, dimensionality reduction is performed through PCA (principal component analysis), and redundancy is removed, then a deep learning model is established, two-stage optimization and physical constraint inversion are combined to invert damage factors, and finally, gradientized parameters conforming to actual deterioration are formed by fusing damage and chloride ion parameters. The application has the following effects: the concrete deterioration process can be accurately simulated, the durability evaluation and service life prediction can be provided, and the engineering safety and economy can be improved.

Description

Technical Field

[0001] This invention relates to the technical field of building material durability assessment, and in particular to a method and system for simulating concrete deterioration under continuous load-chloride ion coupling. Background Technology

[0002] Coastal service environments present concrete structures with high concentrations of corrosive ions and accelerated erosion under continuous loads, seriously threatening concrete durability and safe application. Therefore, research and accurate evaluation of concrete's resistance to chloride ion penetration under load conditions are crucial for concrete durability design and life prediction.

[0003] Currently, the main methods for testing and evaluating the chloride ion permeability of concrete include natural diffusion, wet-dry cycle acceleration, and applied electric field accelerated diffusion. Among these, the applied electric field accelerated diffusion method accelerates chloride ion migration by applying an external electric field, enabling the acquisition of experimental results in a relatively short time. Furthermore, some studies utilize finite element analysis software to establish structural mechanical models, assigning deteriorated material parameters to the models to analyze the overall mechanical performance degradation of the structure. These methods provide technical support for concrete durability research to a certain extent.

[0004] Existing technologies have the following shortcomings: First, existing methods are mostly based on single factors (such as load or chloride ion corrosion), lacking a systematic study of the overall mechanical property degradation of concrete structures under the coupled effects of load and chloride ions, making it difficult to comprehensively reflect the degradation of concrete structures in actual service environments. Second, existing experimental devices lack accurate measurement of the chloride ion diffusion coefficient under continuous load, making it impossible to clarify the influence mechanism of load on chloride ion corrosion. Summary of the Invention

[0005] In order to accurately simulate the deterioration process of concrete, provide a scientific basis for durability assessment and life prediction, and improve engineering safety and economy, this application provides a method and system for simulating concrete deterioration under continuous load-chloride ion coupling.

[0006] Firstly, this application provides a method for simulating concrete deterioration under continuous load-chloride ion coupling, employing the following technical solution:

[0007] A method for simulating concrete deterioration under sustained load-chloride ion coupling includes:

[0008] Concrete test blocks were prepared using a pre-designed method. After standard curing and pretreatment with a saturated solution, the support points were adjusted and the test blocks were placed according to the three-point bending specification to determine the tension and compression zones of the test blocks. The test blocks were loaded using a pre-designed loading device, and the load was continuously increased until the load reached its peak value. This peak value was recorded as the ultimate load of the test block using a pre-designed sensor and load display panel. A specific ratio of continuous tension / compression load was applied and maintained based on the ultimate load, and real-time monitoring and feedback control were performed using the sensor and load display panel to ensure the stability of the load value. The tension load corresponds to the tension zone of the test block, and the compression load corresponds to the compression zone of the test block.

[0009] Assemble the test components based on the stress area of ​​the test block, adjust the solution according to the stress type to determine the direction of chloride ion diffusion, connect the electrodes and apply an external electric field to accelerate chloride ion migration; during this period, replenish the electrode solution at timed intervals according to the preset solution replenishment method to maintain the stability of the test environment, and the entire chloride ion migration stage continues until the preset experimental end requirements are met.

[0010] After the experiment, the power was disconnected, the test block was processed and sprayed with a preset colorant, the chloride ion penetration depth was measured, and the chloride ion diffusion coefficient under different loading degrees and different stress types was calculated based on the experimental parameters.

[0011] By combining diffusion coefficient data under different conditions, a correlation is established between loading degree, stress type and chloride ion diffusion coefficient.

[0012] By employing the above technical solution, this method simulates the coupling effect of load and chloride ions, precisely controls experimental conditions, accelerates chloride ion migration, accurately measures chloride ion penetration depth, calculates diffusion coefficient, and establishes relevant correlations. This provides reliable data for concrete durability research, helps optimize concrete structure design, extends its service life, and has significant engineering application value.

[0013] Secondly, this application provides a concrete deterioration simulation system under continuous load-chloride ion coupling, which adopts the following technical solution:

[0014] A concrete deterioration simulation system under continuous load-chloride ion coupling includes a memory, a processor, and a program stored in the memory and executable on the processor. When the program is loaded and executed by the processor, it implements the concrete deterioration simulation method under continuous load-chloride ion coupling as described in the first aspect. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the overall process of a method for simulating concrete deterioration under continuous load-chloride ion coupling according to an embodiment of this application.

[0016] Figure 2This is a schematic diagram of the loading device configured with a chloride ion testing component in an embodiment of this application.

[0017] Figure 3 This is a schematic diagram of the chloride ion testing component in an embodiment of this application.

[0018] Figure 4 This is a top view of the chloride ion testing component in an embodiment of this application.

[0019] Figure 5 This is a partial schematic diagram of the chloride ion testing component in an embodiment of this application, showing the assembly relationship between the sponge, titanium mesh, solution tank, and test block. Detailed Implementation

[0020] The present application will be further described in detail below with reference to the accompanying drawings.

[0021] Reference Figure 1 This application discloses a method for simulating concrete deterioration under sustained load-chloride ion coupling, comprising:

[0022] Step S100: Concrete test blocks are prepared using a pre-defined preparation method. After standard curing and saturated solution pretreatment, the support points are adjusted and the test blocks are placed according to the three-point bending specification to determine the tension and compression zones of the test blocks. A pre-defined loading device is used to load the test blocks, continuously increasing the load until it reaches its peak value. This peak value is recorded as the ultimate load of the test block using a pre-defined sensor and load display panel. A specific ratio of continuous tension / compression load is applied and maintained based on the ultimate load, and real-time monitoring and feedback control are performed using the sensor and load display panel to ensure the stability of the load value. The tension load corresponds to the tension zone of the test block, and the compression load corresponds to the compression zone. Specific details of the loading device can be found in [link to details]. Figure 2 .

[0023] The experiment includes the following components: Pre-determined preparation method: Based on the experimental design requirements, a pre-determined concrete mix ratio and mixing process are used to ensure the concrete specimens possess specific properties. Standard curing: The concrete specimens are placed in a standard curing chamber, with the temperature controlled at 20±2℃ and the relative humidity above 95%, and cured for a certain period to ensure the concrete reaches its design strength. Saturated solution pretreatment: The cured specimens are immersed in a saturated Ca(OH)2 solution to enhance their resistance to erosion. Three-point bending test: A mechanical experimental method that tests the bending performance of the specimen by applying a concentrated load. Loading device: Equipment used to apply loads to the concrete specimens, typically including loading wheels and bending clamps. Sensor: A device used to monitor the load on the specimens in real time, converting the load signal into an electrical signal output. Load display panel: A device connected to the sensor to display real-time load values ​​for easy observation and recording by experimental personnel. Continuous tensile / compressive load: During the experiment, a certain magnitude of tensile or compressive load is applied to and maintained on the specimens to simulate the stress state of concrete structures in actual engineering. Tension zone / compression zone: Under load, the upper and lower sides of the concrete specimen are subjected to tensile stress and compressive stress, respectively.

[0024] The overall process is as follows:

[0025] First, accurately weigh the cementitious materials, sand, and gravel according to the designed mix proportions. Place these materials in a mixer and premix at low speed for 20-30 seconds. Then add the predetermined proportions of water and admixtures, and continue mixing at high speed for 3-5 minutes to ensure the concrete is uniformly mixed. After mixing, quickly pour the fresh concrete into molds of the appropriate size and use vibration equipment to compact its structure, ensuring the internal quality of the concrete. After pouring, cover the test blocks with a film and remove the mold after 24 hours. Immediately after demolding, transfer the test blocks to a standard curing room and cure them at a temperature of 20±2℃ and a relative humidity of over 95% until the specified age (3-90 days) to ensure the concrete reaches its design strength. After the curing period, pretreat the test blocks by immersing them in a saturated Ca(OH)2 solution for 3 days to enhance their resistance to erosion. After pretreatment, the support point positions were adjusted according to the three-point bending test specifications. The soaked concrete beam was placed on the support points, the surface was wiped dry, and both sides were sealed with resin to determine the tension zone (lower side) and compression zone (upper side) of the concrete under continuous load. Finally, the loading wheel of the loading device was used to directly load the specimen in contact with the flexural clamp. The load change was monitored in real time using sensors, and the load value was displayed in real time on the load display panel. The above process can be referred to Figure 2During the process, the sensor monitors the load on the test block in real time and converts it into an electrical signal, which is simultaneously displayed on the load display panel. When the load increases to the "peak value" (i.e., the maximum load value at which the load no longer increases and the test block is about to suffer significant damage), the peak value is recorded. This peak value is the ultimate load of the test block. After obtaining the ultimate load of the test block, the required continuous tensile / compressive load is accurately controlled and monitored in real time to ensure the stability of the load during the experiment. Specifically, the load display panel compares the real-time monitored "current load value" with the preset "target load value". In comparison, if a deviation occurs (e.g., the current load is lower than the target value, possibly due to insufficient loading force caused by slight deformation of the test block; or higher than the target value, possibly due to overload caused by the inertia of the loading device), an "adjustment signal" is immediately sent to the control module of the loading device. After receiving the adjustment signal, the loading device drives the actuators (such as the loading wheel and hydraulic components) through the control module to make fine adjustments: if the load is too low, the loading force is increased (e.g., pushing the loading wheel to apply greater pressure to the test block); if the load is too high, the loading force is decreased (e.g., slowing down the speed of the loading wheel or slightly retracting it). Specific proportions need to be set in conjunction with the experimental objectives (e.g., simulating the actual service load level of concrete structures). These are adjustable parameters in the experimental design stage. The core principle is to use the "ultimate load of the test block" as a benchmark (e.g., taking 50%, 60%, 70% of the ultimate load, etc.), and the common value range is mostly 50%-80% of the ultimate load.

[0026] In step S200, the test component is assembled based on the stress area of ​​the test block, the solution is adjusted according to the stress type to determine the direction of chloride ion diffusion, the electrode is connected and an external electric field is applied to accelerate chloride ion migration; during this period, the electrode solution is replenished periodically according to the preset solution replenishment method to maintain the stability of the test environment, and the entire chloride ion migration stage continues until the preset experimental end requirements are met.

[0027] The test components are auxiliary devices used in the experiment, including upper and lower solution tanks, titanium mesh, and sponge tanks, used to simulate the actual service environment and accelerate chloride ion migration. Chloride ion diffusion direction: The migration path of chloride ions in the concrete specimen, determined by the arrangement of different solutions in the solution tank. External electric field: An electric field applied through electrodes to accelerate chloride ion migration in the concrete. Preset solution replenishment method: A method to periodically replenish electrode solution to maintain a stable test environment according to experimental requirements. Preset experimental completion requirements (all must be set in advance), the core of which are as follows: 1. Reaching the preset migration time (e.g., "7 days of migration" in the document example); 2. The chloride ion penetration depth is sufficient to reach the measurable threshold (avoiding excessive shallowness leading to large measurement errors); 3. A clear concentration gradient of chloride ions is formed within the specimen (satisfying subsequent diffusion coefficient calculations).

[0028] The overall process (the parts involving test components can be referred to) Figures 3 to 5 )as follows:

[0029] 1. Test Component Assembly: Two appropriately sized titanium meshes are laid on the upper side of the test block after the load is applied and connected by copper wire. The length of the copper wire is adjustable to avoid the bending clamps. Then, the prefabricated upper solution tank (made of acrylic sheet) is bonded to the upper side of the test block with glass glue. Finally, the test block is wrapped on three sides with sponge, which is filled with electrode solution and fixed using a prefabricated foldable sponge tank (made of acrylic sheet). The nuts on the sponge tank can be adjusted for tightness, and the sponge with the bottom opening is in direct contact with the solution in the lower solution tank. A titanium mesh is placed between the lower surface of the test block and the sponge. Before the experiment, the titanium meshes on the upper and lower surfaces are connected to the corresponding electrodes.

[0030] 2. Adjustment of chloride ion diffusion direction: Adjust the solutions in the upper and lower solution tanks according to the type of force to determine the direction of chloride ion diffusion. For tensile loads, the upper solution tank contains NaOH solution and the lower solution tank contains NaCl solution, with the chloride ion diffusion direction from the tensile side to the compressive side; for compressive loads, the upper solution tank contains NaCl solution and the lower solution tank contains NaOH solution, with the chloride ion diffusion direction from the compressive side to the tensile side.

[0031] 3. Accelerating chloride ion migration with an external electric field: After injecting the selected solution into the upper and lower solution tanks, connect the power supply electrode to the corresponding upper and lower titanium meshes, and adjust to a suitable voltage or current to accelerate chloride ion migration. The power supply electrode is... Figures 3 to 5 Electrodes A and B are shown. Electrodes A and B are made of titanium mesh (titanium mesh has good conductivity and is resistant to chloride ion corrosion). They are connected to an external power source via copper wire (avoiding the bending clamp to prevent short circuits). Electrode A corresponds to the tension zone of the concrete specimen (as shown in the solution tank below, filled with NaCl solution), and electrode B corresponds to the compression zone (as shown in the solution tank above, filled with NaOH solution). Both are arranged along the load direction (tension / compression) to form an electric field gradient. Electrode A is connected to the anode (positive potential) of the power source, and electrode B is connected to the cathode (negative potential), forming an electric field inside the specimen from the tension zone (A) to the compression zone (B).

[0032] 4. Solution replenishment: During the experiment, the electrode solution is replenished regularly according to the preset method to ensure the stability of the solution concentration and volume and maintain the stability of the test environment.

[0033] The following example illustrates the experiment: Assuming the specimen is under tension, a 0.1 mol / L NaOH solution is added to the upper solution tank, and a 0.1 mol / L NaCl solution is added to the lower solution tank. Titanium meshes are placed in both tanks and connected to power electrodes. A 10V DC voltage is applied to accelerate the migration of chloride ions towards the stressed side. During the experiment, the electrode solution is replenished every 2 hours to ensure stable solution concentration and volume. This method effectively simulates the migration process of chloride ions in concrete under actual service conditions and provides accurate data for subsequent diffusion coefficient calculations.

[0034] Step S300: After the experiment, disconnect the power supply, process the test block and spray the preset color developer, measure the chloride ion penetration depth, and calculate the chloride ion diffusion coefficient under different loading degrees and different stress types based on the experimental parameters.

[0035] Among them, chloride ion penetration depth: the depth to which chloride ions diffuse in the concrete specimen, measured after color development with a colorimetric agent. Chloride ion diffusion coefficient: a parameter describing the diffusion rate of chloride ions in concrete, calculated from experimental data. Preset colorimetric agent: a chemical reagent used to indicate the chloride ion diffusion depth, usually AgNO3 solution. Experimental parameters: parameters recorded during the experiment, such as voltage, temperature, and load, used to calculate the chloride ion diffusion coefficient.

[0036] The overall process is as follows:

[0037] 1. Post-Experiment Procedure: Immediately after the experiment, disconnect the power supply and remove the test block from the loading device. Rinse the test block thoroughly and dry it, ensuring no residual solution remains on the surface. Use a cutting machine to cut the test block perpendicular to the direction of ion diffusion, ensuring a smooth cut surface.

[0038] 2. Chloride ion penetration depth measurement: A pre-set colorimetric reagent (such as 0.1 mol / L AgNO3 solution) is sprayed onto the surface of the cut test block. Chloride ions react with AgNO3 to form insoluble AgCl precipitate, creating a white colored area. The chloride ion penetration depth is determined by measuring the boundary of the colored area.

[0039] 3. Calculation of chloride ion diffusion coefficient: a. Calculate the chloride ion diffusion coefficient based on experimental parameters (such as voltage, temperature, load, etc.) and chloride ion penetration depth, using Fick's second law. b. The formula for Fick's second law is: ,in, Let be the chloride ion concentration at a depth x from the concrete surface and a diffusion time t. Let be the initial concentration, D be the diffusion coefficient, t be time, and erfc be the complementary error function. c. Calculate the diffusion coefficient D by using the chloride ion penetration depth and time recorded in the experiment.

[0040] Example: Assume that in the experiment, a specimen is subjected to a sustained tensile load, and a 10V voltage is applied to accelerate chloride ion migration for 7 days. After the experiment, the specimen is cut open and sprayed with a 0.1mol / L AgNO3 solution, and the chloride ion penetration depth is measured to be 10mm. Based on the experimental parameters (voltage 10V, temperature 20℃, time 7 days) and Fick's second law, the chloride ion diffusion coefficient D is calculated to be 1.5×10−12m² / s. This method allows for accurate evaluation of the chloride ion diffusion performance of concrete under different loading levels and stress types.

[0041] Step S400: Combine diffusion coefficient data under different conditions to establish the correlation between loading degree, force type and chloride ion diffusion coefficient.

[0042] Among these, the loading degree refers to the magnitude of the load applied to the concrete specimen, usually expressed as stress or strain. The stress type refers to the type of load the specimen bears, such as tensile or compressive loads. The chloride ion diffusion coefficient is a parameter describing the diffusion rate of chloride ions in concrete, obtained through experimental calculations. The correlation model is a mathematical model used to describe the relationship between the loading degree, stress type, and chloride ion diffusion coefficient.

[0043] The overall process is as follows: 1. Data processing: Collect chloride ion diffusion coefficient data under different loading levels and force types, including the load magnitude, force type (tension or compression), and corresponding chloride ion diffusion coefficient recorded in the experiment. 2. Establishing a correlation model: Use regression analysis or machine learning algorithms to establish a relationship model between loading level, force type, and chloride ion diffusion coefficient. For example, a linear regression model or a nonlinear regression model can be used to fit the data. Assuming the relationship between the diffusion coefficient D and the loading level (represented by stress σ) and force type (tension or compression) can be expressed as: Where a, b, and c are model parameters, and type is an indicator variable representing the type of force (tension or compression). 3. Model Validation and Optimization: The model is trained using a portion of the experimental data, and then validated using the remaining data to ensure its accuracy and reliability. If the model's prediction error is large, the model parameters can be adjusted, or other more complex models, such as multinomial regression or neural network models, can be tried. 4. Application of Results: The established model is applied to actual engineering projects. Based on specific loading conditions and force types, the diffusion coefficient of chloride ions in concrete structures is predicted, thereby assessing the durability of the concrete structure.

[0044] The test components assembled based on the stress area of ​​the test block include:

[0045] Step S201 involves configuring a chloride ion testing component, including a solution tank, titanium mesh, conductive connectors, and a fixing device, and pre-treating the surfaces of the tensile and compressive zones of the test block. The chloride ion testing component can be referenced from [reference needed]. Figures 3 to 5 As shown.

[0046] The components include: a solution tank (usually made of insulating material like acrylic sheet) to hold electrolyte solutions (such as sodium chloride solution) and to simulate chloride ion corrosion conditions in actual service environments; a titanium mesh (used as an electrode to accelerate chloride ion migration in concrete); conductive connectors (such as copper wires) to connect the titanium mesh and power source, ensuring stable current flow; a fixing device to secure the solution tank to the concrete specimen, typically made of silicone sealant and nuts, ensuring stability during the experiment; and pre-treatment cleaning to remove impurities and dust from the tensile and compressive zones of the specimen, ensuring good contact between the titanium mesh and the specimen surface.

[0047] The overall process is as follows: 1. Configure chloride ion testing components: a. Solution tank: Prefabricate upper and lower solution tanks according to the size of the test block, usually made of acrylic sheet, ensuring that its size is compatible with the test block. b. Titanium mesh: Select a titanium mesh of appropriate size for the tension and compression zones of the test block. The size of the titanium mesh should be slightly smaller than the inner diameter of the solution tank to ensure that it can be completely immersed in the solution. c. Conductive connectors: Prepare conductive connectors such as copper wire to connect the titanium mesh and the power supply to ensure that the current can pass through stably. d. Fixing device: Prepare fixing devices such as glass glue and nuts to fix the solution tank to the test block to ensure that it is stable and does not shift during the experiment. 2. Surface cleaning and pretreatment of the test block: Use sandpaper or wire brush to gently polish the surface of the tension and compression zones of the test block to remove surface dust and impurities. Wipe the surface of the test block with a clean damp cloth to ensure that the surface is clean and dust-free. Dry the surface of the test block to ensure that there is no moisture residue on the surface to ensure good contact between the titanium mesh and the surface of the test block.

[0048] In step S202, based on the positions of the tension and compression zones of the pretreated test block, a titanium mesh is laid on the stress surface and a stable conductive connection is achieved through conductive connectors. Then, a fixing device is used to position and assemble the solution tank on the outside of the stress area of ​​the test block to form a test assembly adapted to the stress area.

[0049] The overall process is as follows: 1. Titanium mesh laying and conductive connection: a. Based on the location of the tension and compression zones of the test block, lay appropriately sized titanium meshes on the upper and lower surfaces of the test block respectively. b. Use conductive connectors (such as copper wires) to connect the titanium mesh to the power supply to ensure stable current flow. The copper wire length is adjustable to avoid bending clamps and prevent interference with the experimental setup. 2. Solution tank positioning and fixing: a. Adhere the prefabricated solution tank (made of acrylic sheet) to the outside of the stress area of ​​the test block using fixing devices (such as silicone sealant). b. Ensure that the solution tank is tightly fitted to the surface of the test block, without air bubbles or gaps, to ensure that the solution is evenly distributed and effectively contacts the surface of the test block. 3. Assembly of the overall test assembly: a. Check whether the connection between the titanium mesh and the solution tank is firm to ensure that it will not loosen during the experiment. b. Confirm that the conductive connectors are connected correctly, without short circuits or open circuits. c. Finally, form a test assembly adapted to the stress area, preparing for the subsequent chloride ion diffusion experiment.

[0050] Adjusting the solution according to the type of force applied to determine the direction of chloride ion diffusion includes:

[0051] Step S20A: Determine the type of sustained load that the test block will bear, and mark the specific spatial locations of the tension zone and compression zone of the test block. The sustained load type includes tensile load and compressive load.

[0052] Among them, sustained load type refers to the nature of the load that the specimen bears during the experiment, divided into tensile load (tension) and compressive load (compression). Tension zone: The area of ​​the specimen subjected to tensile stress under load. Compression zone: The area of ​​the specimen subjected to compressive stress under load. Marking: The specific locations of the tension and compression zones are marked on the surface of the specimen using physical or chemical methods for subsequent operations.

[0053] The overall process is as follows:

[0054] 1. Determine the load type: Based on the experimental design, determine the type of sustained load that the specimen will bear, i.e., tensile load or compressive load. For example, if the experiment simulates the load-ion coupling effect in the tension zone of the arch waist of a tunnel concrete segment structure, the specimen will bear a tensile load; if it simulates the compression zone on the outer side of the top and bottom of the tunnel, the specimen will bear a compressive load.

[0055] 2. Mark the stress areas: Mark the specific locations of the tension and compression zones on the surface of the test block. Physical marking can be done using a marker pen or sticker, or chemical marking can be done by applying a small amount of distinguishable chemical reagent to the surface of the test block. Ensure the markings are clear, accurate, and clearly differentiate between the tension and compression zones.

[0056] Step S20B: Based on the marked stress area, configure the solution according to the load type, as follows: Under tensile load, inject NaOH solution into the upper solution tank corresponding to the compression zone and NaCl solution into the lower solution tank corresponding to the tension zone; Under compressive load, inject NaCl solution into the upper solution tank corresponding to the compression zone and NaOH solution into the lower solution tank corresponding to the tension zone.

[0057] The solution tank is a container used to hold electrolyte solutions (such as NaCl or NaOH solutions), typically made of insulating materials (such as acrylic sheets). NaCl solution is a sodium chloride solution used to simulate chloride ion corrosion conditions in actual service environments, serving as the source of chloride ions. NaOH solution is a sodium hydroxide solution used to create an electrochemical gradient in the solution tank, promoting chloride ion migration. The direction of chloride ion diffusion is the migration path of chloride ions in the concrete specimen, determined by the arrangement of different solutions in the solution tank.

[0058] The overall process is as follows:

[0059] 1. Solution preparation according to load type: a) Tensile load: When the specimen is subjected to a tensile load, NaOH solution is injected into the upper solution tank corresponding to the compression zone, and NaCl solution is injected into the lower solution tank corresponding to the tension zone. This setting is to simulate the migration of chloride ions from the tension zone to the compression zone. b) Compressive load: When the specimen is subjected to a compressive load, NaCl solution is injected into the upper solution tank corresponding to the compression zone, and NaOH solution is injected into the lower solution tank corresponding to the tension zone. This setting is to simulate the migration of chloride ions from the compression zone to the tension zone.

[0060] 2. Solution Injection Procedure: a. Accurately measure the required concentrations of NaCl and NaOH solutions using a graduated cylinder or pipette. b. Slowly inject the solutions into the corresponding solution tanks, ensuring uniform distribution and no air bubbles. c. Check the sealing of the solution tanks to ensure no leakage occurs during the experiment.

[0061] Step S20C: Determine the chloride ion diffusion direction based on the solution configuration, as follows: Under tensile load, chloride ions migrate from the tensile region containing NaCl solution to the compressive region containing NaOH solution, with the diffusion direction from the tensile side to the compressive side; Under compressive load, chloride ions migrate from the compressive region containing NaCl solution to the tensile region containing NaOH solution, with the diffusion direction from the compressive side to the tensile side.

[0062] The overall process is as follows:

[0063] 1. Determine the diffusion direction based on the load type: a) Tensile load: Chloride ions migrate from the tension zone containing NaCl solution to the compression zone containing NaOH solution, with the diffusion direction from the tension side to the compression side. b) Compressive load: Chloride ions migrate from the compression zone containing NaCl solution to the tension zone containing NaOH solution, with the diffusion direction from the compression side to the tension side. 2. Record the diffusion direction: Record the determined chloride ion diffusion direction in the experimental log for subsequent analysis and verification.

[0064] Example: Assuming the specimen is subjected to a tensile load, according to the solution preparation in step S20B, NaCl solution is injected into the tension zone (lower side) and NaOH solution is injected into the compression zone (upper side). Therefore, chloride ions will migrate from the tension zone (lower side) to the compression zone (upper side), with the diffusion direction from the tension side to the compression side. This direction is recorded in the experimental log.

[0065] Step S20D: Associate and store the diffusion direction with the stress type of the test block and the solution configuration parameters.

[0066] Among these, the associated storage involves systematically recording and storing key experimental parameters (such as load type, solution configuration, diffusion direction, etc.) for subsequent analysis and verification. Parameter recording involves detailed recording of various experimental parameters, including load type, solution type, concentration, diffusion direction, etc.

[0067] The test block was treated and sprayed with a pre-set color developer. The chloride ion penetration depth was measured, including:

[0068] Step S301: Rinse the test block clean and dry it, and cut the test block along the chloride ion diffusion direction to obtain a cross section.

[0069] Among them, the chloride ion diffusion direction is the migration path of chloride ions in the concrete specimen, which is determined by the experimental setup. The cross-section is a plane perpendicular to the diffusion direction obtained after cutting the specimen along the chloride ion diffusion direction; it is used to observe the chloride ion penetration depth.

[0070] The overall process is as follows: Test block cleaning: After the experiment, remove the test block from the loading device and rinse its surface with clean water to remove any residual solution and impurities. Wipe the surface of the test block dry with a clean cloth or paper towel, ensuring that no moisture remains.

[0071] Step S302: Spray a pre-concentrated AgNO3 colorimetric agent onto the cross section and let it stand for a pre-concentrated time to allow chloride ions to react with silver ions to form a white silver chloride precipitate boundary.

[0072] The AgNO3 colorimetric reagent is a silver nitrate solution used to react with chloride ions to form an insoluble silver chloride (AgCl) precipitate, thus indicating the diffusion boundary of chloride ions. Preset concentration: The concentration of the AgNO3 solution is determined according to the experimental design, typically 0.1 mol / L. Standing time: After spraying the colorimetric reagent, the test block needs to stand for a period of time to ensure that chloride ions and silver ions react fully and form a clear precipitation boundary.

[0073] The overall process is as follows: 1. Spraying the color developer: Using a sprayer or brush, evenly spray the pre-concentrated (e.g., 0.1 mol / L) AgNO3 solution onto the cross-section of the test block. Ensure the spray covers the entire cross-section without any missed areas. 2. Static reaction: Place the sprayed test block in a well-ventilated environment and allow it to stand for a pre-concentrated time (e.g., 30 minutes) to ensure that chloride ions and silver ions react fully. During the static reaction, avoid external interference to the test block, such as vibration or airflow.

[0074] Step S303: Measure the vertical distance from the surface of the test block to the precipitation boundary at multiple points along the cross section, and take the average value as the chloride ion penetration depth.

[0075] Penetration depth: The maximum distance chloride ions diffuse in a concrete specimen, typically determined by measuring the AgCl precipitation boundary formed by the reaction of chloride ions with AgNO3. Multi-point measurement: Measurements are taken at multiple locations on the specimen cross-section to ensure data accuracy and reliability. Average value: The data from multiple measurements are averaged to reduce measurement errors and obtain a more accurate penetration depth.

[0076] The overall process is as follows: 1. Select measurement points: Select multiple measurement points on the cross-section of the test block, usually choosing evenly distributed points to ensure comprehensive measurement. For example, select measurement points at the center, four corners, and middle of the cross-section. 2. Measure the penetration depth: Use vernier calipers or digital calipers to measure vertically from the surface of the test block to the AgCl precipitation boundary. Record the penetration depth data for each measurement point. 3. Calculate the average value: Add the penetration depth data from all measurement points together, then divide by the number of measurement points to obtain the average penetration depth.

[0077] Step S304: Take samples in layers according to the permeation depth range, detect the chloride ion concentration in each layer, and establish a concentration gradient distribution.

[0078] The process includes: stratified sampling: dividing the cross-section of the test block into multiple layers based on the chloride ion penetration depth, and sampling from each layer. Chloride ion concentration detection: determining the concentration of chloride ions in the sample using chemical analysis methods, typically ion chromatography or potentiometric titration. Concentration gradient distribution: describing the concentration variation of chloride ions at different depths within the test block, usually represented by a curve of concentration versus depth.

[0079] The overall process is as follows: 1. Determine the stratification range: Divide the test block cross-section into multiple layers based on the chloride ion penetration depth range. For example, if the penetration depth is 10mm, the test block can be divided into layers such as 0-2mm, 2-4mm, 4-6mm, 6-8mm, and 8-10mm. 2. Layered sampling: Using drilling or cutting tools, take an appropriate amount of concrete sample from each layer. Ensure that the integrity of the sample is not damaged during the sampling process. Label and save the samples from each layer for subsequent analysis. 3. Chloride ion concentration detection: Determine the chloride ion concentration in the sample of each layer using ion chromatography or potentiometric titration. Record the chloride ion concentration data for each layer. 4. Establish a concentration gradient distribution: Based on the chloride ion concentration data of each layer, plot the concentration change curve with depth to establish a concentration gradient distribution.

[0080] Following the calculation of chloride ion diffusion coefficients under different loading levels and stress types using experimental parameters, the process also includes steps for obtaining degradation parameters and simulating structural performance, as detailed below:

[0081] Step SA00: Acquire images of crack distribution on the cross section of the test block and extract the geometric features of the cracks; export the acoustic emission signals recorded during the experiment and analyze the acoustic emission characteristic parameters; conduct a uniaxial stress-strain curve experiment on the test block and obtain the mechanical property parameters.

[0082] Among them, the crack distribution image is a visual representation of the location, length, width, and other information of cracks on the cross-section of the test block. Crack geometric features include geometric parameters such as crack length, width, and distribution density. Acoustic emission signals are the acoustic wave signals generated during the experiment due to the generation and propagation of microcracks.

[0083] Acoustic emission characteristic parameters: parameters such as the energy, amplitude, and duration of the acoustic emission signal. Uniaxial stress-strain curve: a curve describing the stress-strain relationship of a material under uniaxial loading conditions. Mechanical property parameters: material property parameters obtained from the uniaxial stress-strain curve, such as elastic modulus, yield strength, and ultimate strength.

[0084] The complete process is as follows: 1. Acquire crack distribution images: Use a high-resolution industrial camera (≥20 megapixels) to capture crack distribution images of the specimen cross-section. Extract the geometric features of the cracks, including crack length, width, and distribution density, using image processing software (such as MATLAB or OpenCV). 2. Export and analyze acoustic emission signals: Export acoustic emission signals from the experimental records, with a signal sampling frequency of 1MHz. Analyze the acoustic emission signals using acoustic emission analysis software (such as AEwin) to extract acoustic emission characteristic parameters, including energy count, amplitude, and duration. 3. Uniaxial stress-strain curve experiment: Perform uniaxial compression or tension experiments on the specimen and record the stress-strain curves. Obtain mechanical property parameters, such as elastic modulus, yield strength, and ultimate strength, from the stress-strain curves.

[0085] Step SB00: Substitute the mechanical performance parameters into the relevant formulas of concrete damage plastic constitutive model to calculate the tensile and compressive damage evolution parameters of the deteriorated part.

[0086] Mechanical property parameters: Material property parameters obtained from uniaxial stress-strain curves, such as elastic modulus, yield strength, and ultimate strength. Damage evolution parameters: Parameters describing the accumulation and development of damage in a material during stress, typically used in concrete damage plastic constitutive models. Concrete Damage Plastic Constitutive Model (CDP): A mathematical model describing the damage and plastic deformation of concrete during stress, usually including damage variables and plastic flow rules.

[0087] The complete process is as follows: 1. Obtain mechanical property parameters: Obtain the elastic modulus (E) and yield strength (E) from the uniaxial stress-strain curve experiment. Ultimate strength () ) and other parameters.

[0088] 2. Calculate damage evolution parameters: Using the relevant formulas of the concrete damage plastic constitutive model (CDP), substitute the mechanical property parameters to calculate the tensile and compressive damage evolution parameters. For example, the tensile damage evolution parameters ( ) and pressure damage evolution parameters ( It can be calculated using the following formula:

[0089] ;

[0090] ;

[0091] in, For the current stress, For ultimate strength, denoted as yield strength, and n as hardening exponent.

[0092] Record damage evolution parameters: Record the calculated damage evolution parameters in the experimental log for subsequent analysis.

[0093] Step SC00: Construct a multidimensional feature matrix containing crack features, acoustic emission parameters, residual strength and chloride ion diffusion coefficient. Establish an end-to-end mapping relationship between the feature matrix and CDP parameters through a deep learning network, and obtain the damage factor by inversion.

[0094] The multidimensional feature matrix is ​​a matrix containing multiple feature parameters used to describe the deterioration state of the specimen. These feature parameters include crack geometry, acoustic emission parameters, residual strength, and chloride ion diffusion coefficient. The deep learning network is a neural network-based machine learning model used to learn complex mapping relationships from large amounts of data. CDP parameters are parameters in the concrete damage-plastic constitutive model (CDP) used to describe the damage and plastic behavior of concrete.

[0095] The overall process can be referred to in steps SC10 to SC50, and will not be repeated here.

[0096] Step SD00 involves fusing the damage factors obtained from the inversion with the calculated damage evolution parameters, and combining them with the concentration gradient and penetration depth distribution corresponding to the chloride ion diffusion coefficient to form comprehensive material parameters for the deteriorated parts.

[0097] Among them, the damage factor is a parameter describing the degree of degradation of the specimen obtained through inversion using a deep learning network. The damage evolution parameters are parameters describing damage accumulation and development calculated based on mechanical property parameters. The comprehensive material parameters are a set of parameters describing the material properties of the deteriorated areas, formed by fusing the damage factor and damage evolution parameters, and are used for subsequent structural performance simulations.

[0098] The overall process can be referred to in steps SD10 to SD60, and will not be repeated here.

[0099] Step SE00: Establish a structural service mechanics model. Assign the integrated material parameters to the structural service mechanics model according to the actual tension and compression zones of the structure. Set boundary conditions and continuous load conditions consistent with the preset structural service scenarios.

[0100] Among them, the structural service mechanics model is a mathematical model used to simulate the mechanical behavior of a structure under actual service conditions, usually based on finite element analysis. Synthetic material parameters are parameters that integrate information such as damage factors, damage evolution parameters, and chloride ion diffusion coefficients, used to describe the material properties of deteriorated areas. Finite element analysis is a numerical analysis method used to solve complex mechanical problems by approximating the solution by dividing the structure into a finite number of elements.

[0101] The complete process is as follows: 1. Establish a structural service mechanics model: Use finite element analysis software (such as Abaqus or ANSYS) to establish the geometric model of the structure. Set boundary conditions and load conditions according to the actual stress conditions of the structure. 2. Assign synthetic material parameters: Assign the synthetic material parameters obtained in step SD00 to the corresponding deteriorated parts in the structural model. Ensure that the material parameters of different parts in the model are consistent with the actual service conditions. 3. Verify the convergence of the model: Check whether the mesh of the finite element model is reasonable and ensure the convergence of the model. Adjust the mesh density or element type to improve the accuracy and convergence of the model.

[0102] Example: Suppose we need to simulate the mechanical behavior of a tunnel structure under service conditions. First, use Abaqus software to create a geometric model of the tunnel structure, including its shape, dimensions, and support structure. Based on the actual stress conditions of the tunnel, set boundary conditions (such as in-situ stress and groundwater pressure) and load conditions (such as vehicle loads). Then, assign the comprehensive material parameters obtained in step SD00 to the corresponding deteriorated parts of the tunnel structure. Finally, check the mesh generation of the finite element model for rationality and adjust the mesh density to ensure model convergence.

[0103] Step SF00: After verifying the convergence of the structural service mechanics model, submit the model for mechanical performance calculation; based on the calculation results, analyze the structural bearing capacity degradation law, stiffness attenuation characteristics and stress distribution state, and compare the calculation results with the experimental measured data of the test block under the same conditions to verify the accuracy of the model. If the error is ≤ the preset threshold, the simulation of the mechanical performance degradation of the concrete structure is completed.

[0104] Among these, model convergence refers to whether the finite element model can stably reach the solution state during the calculation process, which is usually verified by checking the stability of mesh generation and calculation results. Load-bearing capacity degradation refers to the decrease in load-bearing capacity of the structure due to material deterioration during service. Stiffness reduction refers to the decrease in stiffness of the structure due to material deterioration during service. Stress distribution refers to the stress distribution of the structure under stress conditions, usually obtained through finite element analysis.

[0105] The complete process is as follows: 1. Verify model convergence: Check if the mesh of the finite element model is reasonable to ensure model convergence. Adjust the mesh density or element type to improve the model's accuracy and convergence. Run the model calculation and check if the results are stable to ensure model convergence. 2. Calculate mechanical performance indicators: Run the finite element model to calculate the structure's load-bearing capacity, stiffness, and stress distribution. Analyze the changes in load-bearing capacity degradation, stiffness attenuation, and stress distribution to assess the degree of structural deterioration. 3. Result analysis and verification: Compare the calculation results with experimental data or theoretical values ​​to verify the model's accuracy. If there is a large deviation between the calculation results and experimental data, adjust the model parameters or re-mesh and recalculate. The preset threshold can be 5%.

[0106] Assuming a finite element model of a tunnel structure has been established and composite material parameters have been assigned, the mesh generation is first checked to ensure its rationality and sufficient mesh density to capture structural details. The model is then run to calculate the tunnel structure's load-bearing capacity, stiffness, and stress distribution. Analysis reveals a 20% decrease in load-bearing capacity, a 30% reduction in stiffness, and the emergence of significant stress concentration areas. Comparison of these calculations with experimental data shows a strong consistency, validating the model's accuracy.

[0107] An end-to-end mapping relationship between the feature matrix and CDP parameters is established through a deep learning network, and the damage factors are obtained by inversion, including:

[0108] Step SC10 involves using principal component analysis to reduce the dimensionality of the multidimensional feature matrix and remove redundant information.

[0109] The multidimensional feature matrix, a matrix containing multiple feature parameters, describes the degradation state of the test block. These parameters include crack geometry, acoustic emission parameters, residual strength, and chloride ion diffusion coefficient. Principal Component Analysis (PCA) is a statistical method that projects data into a new coordinate system through linear transformation, maximizing the variance of the data in the new coordinate system, thereby removing redundant information and reducing data dimensionality. Redundant information refers to parts of the data that are repeated or do not provide additional information; these can typically be removed using dimensionality reduction methods.

[0110] The complete process is as follows: 1. Construct a multidimensional feature matrix: Collect data on crack geometric features (such as crack length, width, and distribution density), acoustic emission characteristic parameters (such as energy count, amplitude, and duration), residual intensity, and chloride ion diffusion coefficient. Organize this data into a multidimensional feature matrix, where each row represents a sample and each column represents a feature parameter. 2. Apply Principal Component Analysis (PCA): Use the PCA algorithm to reduce the dimensionality of the multidimensional feature matrix. Select the number of principal components to retain; typically, the number of retained principal components should be able to explain most of the variance of the data (e.g., more than 95%). Use the dimensionality-reduced data as input for subsequent deep learning networks.

[0111] Step SC20: Construct an end-to-end deep learning network. The input layer receives the dimensionality-reduced multidimensional feature matrix, the hidden layer adopts a transfer learning strategy, the initial weights of the network are pre-trained based on the mechanical-erosion dataset of standard concrete specimens, and the number of neurons in the hidden layer is dynamically adjusted to adapt to the feature dimension. The output layer is directly mapped to the damage factor.

[0112] The network consists of several key components: an end-to-end deep learning network (a deep learning model that directly translates input data into output, eliminating the need for manual feature extraction); a transfer learning strategy (utilizing the weights of a pre-trained model as initial weights to accelerate training and improve model performance); a mechanical-erosion dataset of standard concrete specimens (containing experimental data of standard concrete specimens under different mechanical and erosion conditions for pre-training); and the number of neurons in the hidden layers (the number of neurons in the hidden layers of the deep learning network, affecting the model's complexity and fitting ability).

[0113] The complete process is as follows: 1. Construct an end-to-end deep learning network: Design a deep learning network where the input layer receives the dimensionality-reduced multidimensional feature matrix, and the output layer directly maps to damage factors. The hidden layers employ a transfer learning strategy, with initial weights pre-trained based on a mechanical-erosion dataset of standard concrete specimens. The number of neurons in the hidden layers is dynamically adjusted to adapt to the feature dimension and improve model performance. 2. Pre-train the network: Pre-train the network using the mechanical-erosion dataset of standard concrete specimens. During pre-training, adjust the number of neurons in the hidden layers to ensure the model can effectively learn the features in the data. 3. Optimize the network structure: Based on the pre-training results, further optimize the network structure, such as adjusting the number of layers and neurons. Ensure the network structure can adapt to the dimensionality of the reduced feature matrix.

[0114] Example: Assuming the reduced feature matrix has a dimension of 3, design a simple end-to-end deep learning network containing one input layer, two hidden layers, and one output layer. The input layer receives 3 features, the first hidden layer has 10 neurons, the second hidden layer has 5 neurons, and the output layer outputs a damage factor. Pre-training is performed using a mechanical-erosion dataset of standard concrete specimens. Assume the dataset contains 1000 samples, each with 3 features and a corresponding damage factor. During pre-training, the number of neurons in the hidden layers is dynamically adjusted, ultimately determining that the first hidden layer has 12 neurons and the second hidden layer has 6 neurons.

[0115] Step SC30 involves introducing a pre-defined two-stage optimization mechanism to optimize the constructed deep learning network. See steps SC31 and SC32 for details, which will not be elaborated upon here.

[0116] Step SC40: During the training of the deep learning network, the uniaxial tension / compression constitutive equation of concrete is transformed into network constraints. A physical constraint penalty term is added to the loss function. When the predicted damage factor deviates from the mechanical law, the penalty weight is increased according to the degree of deviation.

[0117] The loss function includes: a physical constraint penalty term, an additional term added to ensure the model's output conforms to physical laws, such as the mechanical properties of concrete; a loss function, which evaluates the difference between the model's predictions and the true values, typically minimized during training; and a deviation factor, representing the degree of difference between the model's predictions and physical laws, used to adjust the weights of the penalty term.

[0118] The overall process is as follows: 1. Define physical constraints: Based on the mechanical properties of concrete, define physical constraints. For example, the damage factor of concrete should be between 0 and 1, and should increase with increasing stress. Transform these physical constraints into mathematical expressions, such as: 0 ≤ D ≤ 1. Where D is the damage factor and σ is the stress. 2. Construct the loss function: Add a physical constraint penalty term to the loss function to ensure that the model's output conforms to physical laws. The loss function can be expressed as: ;in, For data-driven loss terms, Here, λ represents the penalty term for physical constraints, and λ is the weight of the penalty term. Dynamically adjusting the penalty weight involves adjusting the weight of the penalty term based on the degree of deviation between the model's predicted value and the physical laws. If the predicted value deviates significantly from the physical laws, the penalty weight is increased; if the deviation is small, the penalty weight is decreased. 3. Training the model: The deep learning network is trained by minimizing the loss function using an optimization algorithm (such as the Adam optimizer). During training, the model's output is monitored to ensure it conforms to physical laws, guaranteeing that the model's predictions are reasonable.

[0119] Example: Suppose an end-to-end deep learning network has been built. The input layer receives 3 features, the two hidden layers have 12 and 6 neurons respectively, and the output layer outputs a damage factor. During training, the physical constraint is defined as the damage factor should be between 0 and 1, and it increases with increasing stress. Assuming the initial weight λ = 0.1, during training, if the model predicts a damage factor that exceeds the range of 0 to 1, the value of λ is increased; if the predicted value conforms to physical laws, λ remains unchanged.

[0120] Step SC50 utilizes a deep learning network model trained with two-stage optimization and physical constraints to infer from the input dimensionality-reduced feature matrix, outputting damage factors that meet accuracy requirements and conform to physical laws. The damage factors are then associated with the confidence scores of the corresponding feature matrices.

[0121] Among them, the confidence score is the degree of confidence that the model has in the prediction results, which is usually expressed by probability or uncertainty measure.

[0122] The complete process is as follows: 1. Prepare input data: Use the dimensionality-reduced feature matrix as input data, ensuring the input data format is consistent with that used during training. 2. Model inference: Use a deep learning network model trained with two-stage optimization (NSGA-II algorithm and Bayesian regularization algorithm) to infer the input data. The model outputs the predicted damage factors. 3. Calculate confidence scores: Assign a confidence score to each predicted damage factor using the model's uncertainty estimate or confidence interval calculation. The confidence score can be based on the model's output probability distribution or estimated using the Monte Carlo method. 4. Output results: Output the predicted damage factors and their corresponding confidence scores. Store or display the results for subsequent analysis and validation.

[0123] The two-stage optimization mechanism includes:

[0124] Step SC31: In the first stage, the NSGA-II algorithm is used for multi-objective optimization. The sum of squared errors between the predicted and measured values ​​of CDP parameters is used as the accuracy index, and the number of network iterations is used as the efficiency index. The Pareto optimal solution with accuracy and efficiency is output.

[0125] Among them, the NSGA-II algorithm is a multi-objective genetic algorithm based on non-dominated sorting, used to simultaneously optimize multiple objective functions and output Pareto optimal solutions. Pareto optimal solution: In multi-objective optimization, a solution that cannot improve a particular objective without worsening other objectives. Accuracy metric: The sum of squared errors between the predicted and measured values ​​of the CDP parameters, used to evaluate the model's prediction accuracy. Efficiency metric: The number of network iterations, used to evaluate the model's training efficiency.

[0126] The complete process is as follows: 1. Define the optimization objective: Accuracy index: the sum of squared errors between the predicted CDP parameter values ​​and the experimentally measured values, i.e. ;in, It is the i-th damage factor predicted by the model. It is the i-th damage factor measured in the experiment, and n is the number of samples.

[0127] Efficiency Metric: 1. Number of network iterations: The number of iterations required for the model to converge. 2. Initialize the population: Randomly generate an initial population, where each individual represents a set of network parameters (e.g., weights and biases). 3. Non-dominated sorting: Perform non-dominated sorting on the individuals in the population, dividing them into different non-dominated layers. Individuals within the same non-dominated layer are sorted according to crowding distance to maintain population diversity. 4. Genetic operations: Selection: Select individuals from the population for crossover and mutation operations. Crossover: Combine the parameters of two parent individuals to generate new offspring individuals. Mutation: Randomly perturb the parameters of individuals, introducing new genetic mutations. 5. Iterative optimization: Repeat the non-dominated sorting and genetic operations until the preset number of iterations is reached or the population converges. Output the Pareto optimal solution, which is the solution set that achieves the best balance between accuracy and efficiency.

[0128] Example: Suppose we have a dataset containing 100 samples, each with 3 features and a corresponding impairment factor. We use the NSGA-II algorithm for multi-objective optimization, aiming to find model parameters with high accuracy and few iterations. The initial population size is 50, the crossover probability is 0.9, and the mutation probability is 0.1. After 100 iterations, we obtain a set of Pareto optimal solutions, where the sum of squared errors of the optimal solutions is 0.05, and the number of iterations is 80.

[0129] Step SC32: Based on the optimal solution, proceed to the second stage, where the network weights are dynamically adjusted using the Bayesian regularization algorithm to reduce the risk of overfitting.

[0130] Among them, the Bayesian regularization algorithm is a regularization method based on Bayesian theory, used to dynamically adjust network weights and reduce the risk of overfitting. Overfitting is the phenomenon where a model performs well on the training set but poorly on the test set or real-world data. Weight adjustment optimizes model performance by adjusting the network's weights and biases.

[0131] The overall process is as follows: 1. Initialize network weights: Initialize network weights using the optimal solution obtained in the first stage (NSGA-II algorithm). 2. Bayesian regularization: During training, a Bayesian regularization term is introduced to dynamically adjust the network weights. The regularization term typically includes the prior distribution of the weights and the likelihood function, used to control the complexity of the weights. 3. Dynamically adjust weights: In each iteration, update the network weights according to the Bayesian regularization term. By adjusting the regularization parameter, the model's fitting ability and generalization ability are balanced. 4. Monitor the training process: Use a validation set to monitor the model's performance to ensure that the model does not overfit. If the loss function on the validation set starts to rise, training can be stopped early. 5. Output optimized weights: After training is complete, output the network weights optimized by Bayesian regularization.

[0132] Example: Suppose that an optimal set of network weights has been obtained in the first stage (NSGA-II algorithm). In the second stage, these weights are further optimized using a Bayesian regularization algorithm. Assuming the regularization parameter is λ=0.01, the formula for updating the weights in each iteration during training is: Where θ is the network weight, α is the learning rate, and L is the loss function. By dynamically adjusting the weights, we ensure that the model performs well on both the training and validation sets.

[0133] The damage factors obtained from the inversion are combined with the calculated damage evolution parameters, and then, by incorporating the concentration gradient and penetration depth distribution corresponding to the chloride ion diffusion coefficient, comprehensive material parameters for the deteriorated sites are formed, including:

[0134] Step SD10: Collect the damage factor obtained from the inversion, the tensile / compressive damage evolution parameters obtained from the calculation, and the concentration gradient and penetration depth distribution data corresponding to the chloride ion diffusion coefficient. Based on the inversion error, theoretical fit and measurement accuracy of each parameter, determine their respective fusion weights.

[0135] Among them, fusion weights are the weights assigned to each parameter based on its inversion error, theoretical fit, and measurement accuracy when fusing different parameters. Inversion error is the difference between the model's predicted values ​​and the experimentally measured values. Theoretical fit is the degree of matching between the model's predicted values ​​and the theoretical values.

[0136] Measurement accuracy: The accuracy and reliability of experimental measurements.

[0137] The complete process is as follows: 1. Data collection: Collect the damage factors obtained from the inversion, the calculated tensile / compressive damage evolution parameters, and the concentration gradient and permeation depth distribution data corresponding to the chloride ion diffusion coefficient. 2. Reliability assessment of each parameter: Calculate the inversion error of each parameter, for example: 3. Assess the theoretical fit of each parameter, for example, whether the damage factor conforms to physical laws. Assess the measurement accuracy of each parameter, for example, the measurement error range of the concentration gradient. 4. Determine the fusion weights: Assign fusion weights to each parameter based on inversion error, theoretical fit, and measurement accuracy. The weight allocation can be based on the following rules: parameters with small inversion errors, high theoretical fit, and high measurement accuracy should be assigned higher weights. Parameters with large inversion errors, low theoretical fit, and low measurement accuracy should be assigned lower weights.

[0138] For example, if the inversion error of the damage factor is 5%, the theoretical fit is 90%, and the measurement accuracy is 95%, a higher weight can be assigned; if the inversion error of the concentration gradient is 10%, the theoretical fit is 80%, and the measurement accuracy is 90%, a lower weight can be assigned.

[0139] Step SD20 involves parameter adaptation for the division of the structure into tension and compression zones. In the tension zone, the damage factor is correlated through a quantified relationship between crack density and penetration depth. In the compression zone, damage evolution parameters are matched through the correspondence between stiffness degradation rate and concentration gradient. Specifically, the tension zone is the area of ​​the structure subjected to tensile stress under load. The compression zone is the area of ​​the structure subjected to compressive stress under load. Crack density is the number of cracks per unit area, typically used to assess the degree of damage in the tension zone. Stiffness degradation rate is the degree of reduction in structural stiffness, typically used to assess the degree of damage in the compression zone. Quantified relationships are established by quantifying the relationships between crack density and damage factor, and between stiffness degradation rate and damage evolution parameters, to facilitate parameter adaptation.

[0140] The complete process is as follows: 1. Determine the tension and compression zones: Based on the structural stress analysis, determine the specific locations of the tension and compression zones. 2. Quantify the relationship between crack density and damage factor: In the tension zone, correlate the damage factor by quantifying the relationship between crack density and damage factor. For example, the following formula can be used: ;in, 3. The relationship between quantized stiffness degradation rate and damage evolution parameters: In the compression zone, the damage evolution parameters are matched by the correspondence between stiffness degradation rate and damage evolution parameters. For example, the following formula can be used: ;

[0141] in, 3. Parameter Adaptation: Based on the quantization relationship, the crack density is adapted to the damage factor, and the stiffness degradation rate is adapted to the damage evolution parameters to ensure the rationality and consistency of the parameters in the tension and compression zones.

[0142] Step SD30: Based on the partition adaptation relationship, the mechanical parameters and the corresponding chloride ion parameters are weighted according to the preset weights, and the erosion influence coefficient is introduced simultaneously to achieve quantitative coupling, so as to obtain the partition fusion parameters. Among them, the mechanical parameters include the tensile zone damage factor and the compressive zone damage evolution parameters, and the chloride ion parameters include the penetration depth and concentration gradient.

[0143] The process includes: Zonal adaptation: Based on the different characteristics of the tension and compression zones, corresponding mechanical and chloride ion parameters are adapted. Mechanical parameters include the damage factor of the tension zone and the damage evolution parameters of the compression zone. Chloride ion parameters include penetration depth and concentration gradient. Erosion influence coefficient: Used to quantify the degree of influence of chloride ion erosion on mechanical properties. Weighted average: Different parameters are weighted and summed according to preset weights to obtain comprehensive material parameters.

[0144] The complete process is as follows: 1. Determine the preset weights: Based on the fusion weights determined in step SD10, assign weights to the mechanical parameters and chloride ion parameters. For example, the weight of the damage factor in the tension zone is 0.3, the weight of the damage evolution parameter in the compression zone is 0.28, the weight of the penetration depth is 0.05, and the weight of the concentration gradient is 0.12. 2. Calculate the fusion parameters for each zone: In the tension zone, the damage factor and penetration depth are weighted according to their respective weights to obtain the fusion parameters for the tension zone. The specific formula is as follows: .

[0145] In the compression zone, the damage evolution parameters and concentration gradient are weighted according to their respective weights to obtain the compression zone fusion parameters, as shown in the following formula: .

[0146] in, and These are the fusion parameters for the tension region and the compression region, respectively. , For mechanical parameter weights, , For chloride ion parameter weights, These are the damage evolution parameters of the compression zone. denoted as the damage factor of the tensile zone, d as the penetration depth, and C as the concentration gradient.

[0147] 3. Introducing an Erosion Influence Coefficient: An erosion influence coefficient γ is introduced to quantify the impact of chloride ion erosion on mechanical properties. Fusion parameters are then corrected. ; ;in, and These are the erosion influence coefficients for the tension zone and the compression zone, respectively.

[0148] 4. Output partition fusion parameters: Output the corrected fusion parameters of the tension and compression regions for subsequent spatial mapping and construction of comprehensive material parameters.

[0149] Step SD40 involves spatially mapping the partition fusion parameters according to the penetration depth, and then interpolating to form a gradient parameter that continuously varies along the erosion depth to match the actual spatial distribution of degradation.

[0150] Among them, the partition fusion parameters are the combined parameters of the tension and compression zones after weighting and correction by the erosion influence coefficient. Spatial mapping distributes the parameter values ​​according to their spatial location in the structure, forming a continuous parameter field. Interpolation uses mathematical methods to estimate the values ​​of unknown data points among known data points, thus forming a continuous parameter distribution.

[0151] The overall process is as follows:

[0152] 1. Determine spatial location: Based on the structure's geometry and stress analysis, determine the spatial locations of the tension and compression zones. Associate the fusion parameters of each zone with their corresponding spatial coordinates.

[0153] 2. Spatial Mapping: The partitioned fusion parameters are spatially mapped according to the penetration depth to form a parameter distribution along the erosion depth. Interpolation methods (such as linear interpolation, spline interpolation, etc.) are used to estimate the values ​​of unknown data points between known data points to form a continuous parameter distribution.

[0154] 3. Generate gradient parameters: Generate gradient parameters that vary continuously along the erosion depth using interpolation methods to ensure that the parameter distribution matches the actual spatial distribution of degradation.

[0155] Step SD50 verifies the gradient parameters based on the uniaxial tension / compression constitutive equation of concrete, and corrects parameters that deviate from the physical laws according to constraints.

[0156] Among them, the gradient parameter is a parameter that changes continuously along the erosion depth, reflecting the spatial distribution of material properties. The uniaxial tension / compression constitutive equation is a mathematical model describing the stress-strain relationship of concrete under uniaxial tension or compression conditions. Physical laws are the fundamental physical principles that material properties should conform to, such as the damage factor being between 0 and 1 and increasing with increasing stress. Constraint correction adjusts parameters that deviate from physical laws to make them conform to them.

[0157] 1. Define the uniaxial tension / compression constitutive equations: Verify the gradient parameters using the uniaxial tension / compression constitutive equations for concrete. For example, the tension constitutive equation can be expressed as:

[0158] ;

[0159] The constitutive equation under compression can be expressed as: ;in, The tensile stress per unit area that concrete bears under uniaxial tension. Stiffness index of concrete in the tensile elastic stage (stress-strain proportional stage), i.e., the ratio of stress to strain in the elastic stage. The deformation of concrete under tension (dimensionless, or expressed as microstrain με) is the degree of elongation under tensile stress. : Dimensionless parameter that describes the degree of deterioration of concrete under tension. The compressive stress per unit area that concrete bears under uniaxial compression. Stiffness index of concrete in the compressive elastic stage (stress-strain proportional stage). : The deformation of concrete under compression (dimensionless), encompassing elastic strain (when no significant plastic deformation occurs) and plastic / damage strain. : Dimensionless parameter, quantifying the degree of weakening of mechanical properties of concrete during compression due to damage such as "microcrack propagation, aggregate-mortar interface slippage, and internal defect accumulation".

[0160] 2. Verify the gradient parameters: Substitute the gradient parameters into the uniaxial tension / compression constitutive equations and check whether they conform to physical laws. For example, check whether the damage factor is between 0 and 1 and increases with increasing stress.

[0161] 3. Constraint Correction: If the gradient parameters deviate from the physical laws, they are corrected according to the constraints of the constitutive equations. For example, if the damage factor exceeds 1, it is corrected to 1; if the damage factor is less than 0, it is corrected to 0.

[0162] 4. Output corrected gradient parameters: Output the gradient parameters after constraint correction to ensure that they conform to physical laws.

[0163] Step SD60 involves associating the corrected gradient parameters with the tensile / compression zone identifier and the penetration depth coordinates to form comprehensive material parameters that include mechanical degradation, chloride ion erosion, and spatial distribution characteristics.

[0164] Among them, comprehensive material parameters integrate parameters considering mechanical property degradation, chloride ion corrosion effects, and spatial distribution characteristics to comprehensively describe the performance changes of materials under service conditions. Mechanical degradation: The deterioration of material properties during stress, such as reduced strength and stiffness. Chloride ion corrosion: The chemical erosion of materials by chloride ions, leading to a decline in material properties. Spatial distribution characteristics: The spatial variation of material properties, typically related to the depth of corrosion.

[0165] The overall process is as follows: 1. Integrate the corrected gradient parameters: Associate the constrained and corrected gradient parameters with the tension / compression zone identifiers and permeation depth coordinates. For example, for each depth point in the tension and compression zones, record the corresponding gradient parameter value. 2. Construct comprehensive material parameters: Integrate mechanical degradation parameters (such as damage factor, stiffness degradation rate), chloride ion erosion parameters (such as permeation depth, concentration gradient), and spatial distribution characteristics (such as depth coordinates) into a comprehensive parameter system. 3. Output the integrated comprehensive material parameters as a table or data file for subsequent structural performance analysis and simulation.

[0166] In addition, after completing the mechanical property degradation simulation, a method for simulating concrete degradation under continuous load-chloride ion coupling also includes a simulation-prediction-verification closed-loop analysis step, as detailed below:

[0167] Step 1: Construct a structural digital twin, embed the CDP constitutive parameters (i.e., comprehensive material parameters) calibrated in the laboratory and the chloride ion diffusion model, deploy distributed fiber optic sensors, wireless chloride ion probes and 3D laser scanners, collect structural strain field, chloride ion concentration gradient and surface morphology data in real time, process them with spatiotemporal data fusion algorithm, update the twin load and environmental parameters at a preset frequency, and construct an initial model with real-time degradation conditions.

[0168] Among them, the structural digital twin is a virtual model corresponding to the actual structure, constructed using digital technology, used for real-time monitoring and prediction of structural performance changes. CDP constitutive parameters are parameters of the concrete damage-plastic constitutive model, used to describe the damage and plastic behavior of concrete during stress.

[0169] Chloride ion diffusion model: A mathematical model describing the diffusion process of chloride ions in concrete, used to predict chloride ion concentration distribution. Spatiotemporal data fusion algorithm: An algorithm that fuses data from different times and spaces to improve data accuracy and completeness.

[0170] The overall process is as follows: 1. Building a digital twin: Embedding laboratory-calibrated CDP constitutive parameters and a chloride ion diffusion model to construct a structural digital twin. Deploying distributed fiber optic sensors, wireless chloride ion probes, and a 3D laser scanner to collect real-time data on structural strain field, chloride ion concentration gradient, and surface morphology. 2. Data processing and model updating: Processing the collected data using a spatiotemporal data fusion algorithm to update the twin's load and environmental parameters at a preset frequency. Constructing an initial model with real-time degradation conditions to provide a foundation for subsequent prediction and verification. Example: Assume a concrete structure is exposed to a marine environment with severe chloride ion corrosion. Data is collected in real-time by deployed sensors, processed using a spatiotemporal data fusion algorithm, and the twin's load and environmental parameters are updated. The embedded chloride ion diffusion model predicts the chloride ion concentration distribution, and the CDP constitutive parameters describe the damage behavior of the concrete. Ultimately, the constructed initial model can reflect the structural degradation state in real time.

[0171] Step 2: Construct a hybrid attention mechanism LSTM prediction model. The input layer integrates strain, chloride ion concentration, temperature and humidity, and time-series crack image features. The multi-head attention mechanism is used to dynamically allocate parameter weights. The output is the bearing capacity degradation coefficient and crack propagation rate. After training until the validation set error is less than the preset threshold, the prediction uncertainty is quantified. When the predicted bearing capacity decreases by more than the preset ratio and the uncertainty index is greater than the preset value, an early warning is triggered.

[0172] Among them, the hybrid attention mechanism combines multiple attention mechanisms (such as self-attention and multi-head attention) to dynamically allocate weights for input features, increasing the model's attention to important features. LSTM (Long Short-Term Memory) network: a special type of recurrent neural network (RNN) capable of learning long-term dependencies, suitable for predicting time-series data. Load-bearing capacity degradation coefficient: a parameter describing the decrease in structural load-bearing capacity over time or due to damage. Crack propagation rate: the rate at which cracks propagate over time or under varying loads. Prediction uncertainty: the degree of uncertainty in the model's prediction results, typically represented by confidence intervals or probability distributions.

[0173] The complete process is as follows: 1. Construct a hybrid attention mechanism LSTM model: The input layer integrates strain, chloride ion concentration, temperature and humidity, and time-series crack image features. A multi-head attention mechanism is used to dynamically allocate parameter weights, improving the model's attention to important features. The output layer outputs the bearing capacity degradation coefficient and crack propagation rate. 2. Model training and validation: The hybrid attention mechanism LSTM model is trained using historical data until the validation set error is less than a preset threshold. Prediction uncertainty is quantified to evaluate the model's prediction accuracy and reliability. 3. Early warning mechanism: When the predicted bearing capacity decreases by more than a preset proportion and the uncertainty index is greater than a preset value, an early warning is triggered. The early warning mechanism can notify relevant personnel to take timely measures to prevent further structural deterioration. Example: Suppose that real-time monitoring data of a concrete structure shows that the crack width gradually increases over time. A hybrid attention mechanism LSTM model is used for prediction, with input features including strain, chloride ion concentration, temperature and humidity, and crack image features. After model training, the predicted bearing capacity degradation coefficient is 0.8, and the crack propagation rate is 0.05 mm / month. When the predicted bearing capacity decreases by more than 20% and the uncertainty index is greater than 0.1, an early warning is triggered, prompting relevant personnel to conduct inspections and maintenance.

[0174] Step 3: Periodically acquire data on internal defects and compositional evolution using ultrasonic tomography and X-ray fluorescence spectroscopy. Calculate the weighted error between measured and predicted values. If the error is greater than or equal to a preset threshold, initiate the particle swarm optimization-Bayesian joint optimization algorithm to correct the diffusion coefficient and CDP parameters (the parameter space is divided according to engineering experience and constrained boundaries are set; after global search by particle swarm optimization and local refinement by Bayesian optimization, verify whether the parameters conform to the physical laws of concrete deterioration), and update the twin material library. Internal defects refer to damage or defects within the structure, such as cracks and pores, affecting the integrity and durability of the structure. Compositional evolution refers to the process of material composition changing over time or in the environment, such as changes in chloride ion concentration. The particle swarm optimization-Bayesian joint optimization algorithm combines particle swarm optimization (PSO) and Bayesian optimization for parameter optimization. Parameter space partitioning divides the possible value range of the parameters into multiple sub-intervals to facilitate the search of the optimization algorithm. Physical law verification checks whether the optimized parameters conform to the physical properties and behavior laws of the material.

[0175] The complete process is as follows: 1. Acquire internal defect and composition evolution data: Regularly acquire data on internal defects and composition evolution of the structure using techniques such as ultrasonic tomography and X-ray fluorescence spectroscopy. This data is used to evaluate the accuracy of the model prediction and provide a basis for parameter correction. 2. Calculate weighted error: Calculate the weighted error between the measured and predicted values ​​to evaluate the accuracy of the model prediction. 3. Activate the optimization algorithm: If the weighted error is greater than a preset threshold, activate the particle swarm optimization-Bayesian joint optimization algorithm to correct the diffusion coefficient and CDP parameters. The parameter space is divided according to engineering experience and constrained boundaries are set. After global search by particle swarm optimization and local refinement by Bayesian optimization, verify whether the parameters conform to the physical laws of concrete deterioration.

[0176] 4. Update the twin material library: Update the corrected parameters to the material library of the structural digital twin to ensure that the twin can accurately reflect the current state of the structure.

[0177] Example: Suppose that during the service life of a concrete structure, ultrasonic tomography reveals internal cracks, and X-ray fluorescence spectroscopy shows changes in chloride ion concentration. The weighted error between the measured and predicted values ​​is calculated, and it is found to exceed a preset threshold. A particle swarm optimization-Bayesian joint optimization algorithm is then initiated to correct the diffusion coefficient and CDP parameters. After the corrected parameters are verified to conform to the physical laws of concrete deterioration, they are updated in the twin material library.

[0178] Step 4: If the accuracy is still not up to standard after a preset number of optimizations, the sensor configuration is automatically adjusted (e.g., the sampling frequency is increased), triggering the UAV to inspect and acquire panoramic images. The incremental data is used to train the meta-learning model until the prediction error stabilizes within the preset range, forming a closed-loop system.

[0179] The system includes: Sensor configuration adjustment: Automatically adjusting sensor parameters such as sampling frequency and location based on the accuracy and reliability of monitoring data. Drone inspection: Using drones to acquire panoramic images and obtain macroscopic state information of structures. Incremental data: New data added during the optimization process for further training and optimization of the model. Meta-learning model: A machine learning model capable of quickly adapting to new tasks and data, improving its generalization ability by learning from experience across multiple tasks. Closed-loop system: A system that automatically adjusts and optimizes its own parameters to ensure system stability and performance.

[0180] The complete process is as follows: 1. Evaluate and optimize accuracy: If the accuracy still fails to meet the standard after multiple optimizations, the system is triggered to enter the adjustment phase. The accuracy of the current sensor configuration and data acquisition is evaluated. 2. Automatically adjust sensor configuration: The sampling frequency of the sensors or the density of encrypted scan points is automatically adjusted to improve data accuracy and reliability. For example, the sampling frequency of the sensors is increased from once per hour to once every half hour. 3. Trigger UAV inspection: The UAV is triggered to acquire panoramic images to obtain macroscopic state information of the structure. The image data acquired by the UAV can be used to supplement the sensor data, providing more comprehensive structural state information. 4. Train a meta-learning model using incremental data: The incremental data acquired by the UAV is used to train a meta-learning model to further optimize the model's performance. The meta-learning model can quickly adapt to new tasks and new data, improving the model's generalization ability. 5. Form a closed-loop system: The adjusted sensor data and the image data acquired by the UAV are integrated into the system to form a closed-loop system. The system automatically adjusts and optimizes its own parameters to ensure system stability and performance.

[0181] The process of starting the particle swarm optimization-Bayes joint optimization algorithm and correcting the diffusion coefficient and CDP parameters is as follows:

[0182] Step 3.1, Parameter Space Partitioning: Based on engineering experience, key parameters such as diffusion coefficient and damage factor are divided into multiple subspaces. Each subspace has reasonable upper and lower bound constraints (e.g., diffusion coefficient ≤ 3 times the initial value), forming parameter constraint boundaries. Wherein, parameter space: the set of all possible parameter values, including key parameters such as diffusion coefficient and damage factor. Subspace: the partitioning of the parameter space into multiple smaller regions, facilitating the search by the optimization algorithm. Upper and lower bound constraints: setting reasonable value ranges for each parameter to ensure that the parameter is reasonable in both physical meaning and engineering experience. The complete process is as follows: 1. Determine key parameters: Determine the key parameters to be optimized, such as diffusion coefficient and damage factor. 2. Partition parameter space: Divide the parameter space into multiple subspaces based on engineering experience. Set reasonable upper and lower bound constraints for each subspace. For example, the upper bound of the diffusion coefficient can be set to 3 times the initial value. 3. Set constraint boundaries: Set upper and lower bound constraints for each subspace, forming parameter constraint boundaries. For example, the upper bound of the diffusion coefficient can be set to 3 times the initial value, and the lower bound can be set to 1 / 3 of the initial value.

[0183] Step 3.2: Global Particle Swarm Optimization (PSO) searches the particle swarm and iteratively calculates the fitness value (error between measured and predicted values) to quickly locate the optimal parameter region and select elite parameter combinations. Particle Swarm Optimization (PSO) is a swarm intelligence-based optimization algorithm that finds the optimal solution by simulating the foraging behavior of a flock of birds. Fitness value: A function used to evaluate the quality of the solution, typically the error between the measured and predicted values. Elite parameter combination: The parameter combination that performs best during the optimization process, usually with the lowest fitness value. The complete process is as follows: 1. Initialize the particle swarm: Randomly generate a group of particles, each representing a parameter combination, and initialize their position and velocity. The particle's position corresponds to a point in the parameter space, and the velocity determines the particle's direction and distance of movement in space. 2. Calculate the fitness value: For each particle's parameter combination, calculate the fitness value, i.e., the error between the measured and predicted values. The lower the fitness value, the closer the particle's position is to the optimal solution. 3. Update particle position and velocity: Update the position and velocity of each particle based on its individual and swarm experience. Individual experience refers to the optimal position found by a single particle, while swarm experience refers to the optimal position found by the entire particle swarm. 4. Selecting Elite Parameter Combinations: In each iteration, the position of the particle with the lowest fitness value is recorded; this is the elite parameter combination. Through multiple iterations, the particle swarm gradually approaches the optimal solution, ultimately selecting the elite parameter combinations. Example: Assume the particle swarm size in the particle swarm optimization algorithm is 50, and the maximum number of iterations is 100. Each particle represents a parameter combination, including the diffusion coefficient and damage factor. In each iteration, the fitness value of each particle is calculated, which is the error between the measured value and the predicted value. By updating the particle's position and velocity, the optimal solution is gradually found. Finally, the elite parameter combination with the lowest fitness value is selected.

[0184] Step 3.3, Bayesian Local Refinement: A local search space is constructed based on the elite solution. Gaussian process regression is used to model the objective function, and the optimal parameter combination is selected through a sampling function to achieve local fine-tuning. Specifically, Bayesian optimization is a global optimization algorithm based on Bayesian theory, which finds the optimal solution by constructing a prior distribution of the objective function and continuously updating the posterior distribution. Gaussian process regression is a non-parametric Bayesian method used to model and predict the value of the objective function, commonly used in Bayesian optimization. The sampling function is the function used to determine the next sampling point; common examples include expected improvement, probabilistic improvement, and upper confidence bounds. The complete process is as follows: 1. Constructing the Local Search Space: A local search space is constructed based on the elite parameter combinations selected by the particle swarm optimization algorithm. The range of the local search space can be set as the neighborhood of the elite parameter combinations; for example, the upper and lower limits can be set to ±10% of the elite parameters. 2. Gaussian Process Regression Modeling: Gaussian process regression is used to model the objective function, which is typically the error between the measured and predicted values. Gaussian process regression can provide mean and variance predictions for the objective function, used to assess the uncertainty of each parameter combination. 3. Selecting the optimal parameter combination: The next sampling point is selected using a sampling function (such as the expected improvement EI). The sampling function selects the parameter combination most likely to improve the current optimal solution based on the prediction results of the Gaussian process regression. 4. Local fine-tuning optimization: Within the local search space, the optimal solution is gradually approximated through multiple iterations. In each iteration, the Gaussian process regression model is updated, and the optimal parameter combination is reselected until the convergence condition is met.

[0185] Step 3.4, Physical Constraint Verification: Verify whether the optimized parameters conform to the physical laws of concrete deterioration (e.g., elastic modulus degradation rate ≤ 60%). If not, adjust the search strategy and iterate again until the parameters stabilize and converge. Physical constraint verification checks whether the optimized parameters conform to the physical properties and behavior of the material. Elastic modulus degradation rate: The percentage decrease in the material's elastic modulus over time or due to damage, typically used to assess the degree of material deterioration. Convergence condition: The condition under which the optimization algorithm stops iterating, usually based on parameter stability or error minimization. The overall process is as follows: 1. Define physical constraints: Define physical constraints based on the material's physical properties and behavior. For example, the elastic modulus degradation rate should not exceed 60%. Other possible constraints include a damage factor between 0 and 1 that increases with stress. 2. Verify optimized parameters: Check whether the optimized parameters meet the physical constraints. For example, calculate the optimized elastic modulus degradation rate to ensure it does not exceed 60%. If the parameters do not meet the physical constraints, adjust the search strategy and re-optimize. 3. Adjust the search strategy: If the parameters do not meet the physical constraints, adjust the search strategy, such as narrowing the search range or changing the parameters of the optimization algorithm. Optimize again until the parameters meet the physical constraints. 4. Verify parameter stability: Verify whether the optimized parameters converge stably. If the parameters remain unstable after multiple iterations, continue adjusting the search strategy. Ensure that the parameters converge stably after multiple iterations, satisfying the convergence conditions of the optimization algorithm.

[0186] Based on the same inventive concept, embodiments of the present invention provide a concrete deterioration simulation system under continuous load-chloride ion coupling, including a memory and a processor, wherein the memory stores information that can be run on the processor to implement, as described above. Figure 1 The procedure for the method shown.

[0187] The embodiments described in this specific implementation are preferred embodiments of this application and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.

Claims

1. A method for simulating concrete deterioration under continuous load-chloride ion coupling, characterized in that, include: Concrete specimens were prepared using a pre-designed method. After standard curing and pretreatment with a saturated solution, the support points were adjusted and the specimens were placed according to the three-point bending specification to determine the tension and compression zones of the specimens. The test block is loaded using a preset loading device, and the load is continuously increased until the load reaches its peak value. This peak value is recorded as the ultimate load of the test block using a preset sensor and load display panel. A specific ratio of continuous tensile / compressive load is applied and maintained based on the ultimate load, and real-time monitoring and feedback control are performed using sensors and load display panels to ensure the stability of the load value. The tensile load corresponds to the tensile zone of the test block, and the compressive load corresponds to the compressive zone of the test block. Assemble the test components based on the stress area of ​​the test block, adjust the solution according to the stress type to determine the direction of chloride ion diffusion, connect the electrodes and apply an external electric field to accelerate chloride ion migration; during this period, replenish the electrode solution at timed intervals according to the preset solution replenishment method to maintain the stability of the test environment, and the entire chloride ion migration stage continues until the preset experimental end requirements are met. After the experiment, the power was disconnected, the test block was processed and sprayed with a preset colorant, the chloride ion penetration depth was measured, and the chloride ion diffusion coefficient under different loading degrees and different stress types was calculated based on the experimental parameters. By combining diffusion coefficient data under different conditions, a correlation is established between loading degree, stress type and chloride ion diffusion coefficient; Adjusting the solution according to the type of force applied to determine the direction of chloride ion diffusion includes: Determine the type of sustained load that the test block will bear, and mark the specific spatial locations of the tension and compression zones of the test block. The sustained load types include tensile loads and compressive loads. Based on the marked stress areas, the solutions are configured according to the load type, as follows: Under tensile load, NaOH solution is injected into the upper solution tank corresponding to the compression zone, and NaCl solution is injected into the lower solution tank corresponding to the tension zone; Under compressive load, NaCl solution is injected into the upper solution tank corresponding to the compression zone, and NaOH solution is injected into the lower solution tank corresponding to the tension zone. The direction of chloride ion diffusion is determined based on the solution configuration, as follows: Under tensile load, chloride ions migrate from the tensile region containing NaCl solution to the compressive region containing NaOH solution, with the diffusion direction from the tensile side to the compressive side; under compressive load, chloride ions migrate from the compressive region containing NaCl solution to the tensile region containing NaOH solution, with the diffusion direction from the compressive side to the tensile side. The diffusion direction is associated with the stress type of the test block and the solution configuration parameters and stored.

2. The method for simulating concrete deterioration under continuous load-chloride ion coupling according to claim 1, characterized in that, The test components assembled based on the stress region of the test block include: The chloride ion testing component is equipped with a solution tank, titanium mesh, conductive connectors and fixing devices, and the surfaces of the tensile and compressive areas of the test block are cleaned and pretreated. Based on the positions of the tension and compression zones of the pretreated test block, titanium mesh is laid on the stress surface and a stable conductive connection is achieved through conductive connectors. Then, a fixing device is used to position and assemble the solution tank on the outside of the stress area of ​​the test block, forming a test assembly adapted to the stress area.

3. The method for simulating concrete deterioration under continuous load-chloride ion coupling according to claim 1, characterized in that, The test block was treated and sprayed with a pre-set color developer. The chloride ion penetration depth was measured, including: Rinse the test block clean and dry it, then cut the test block along the direction of chloride ion diffusion to obtain a cross section; Spray a pre-concentrated AgNO3 colorant onto the cross section and let it stand for a pre-concentrated time to allow chloride ions to react with silver ions to form a white silver chloride precipitate boundary. Measure the vertical distance from the surface of the test block to the precipitation boundary at multiple points along the cross section, and take the average value as the chloride ion penetration depth; Sampling was performed in layers according to the depth of penetration, and the chloride ion concentration in each layer was detected to establish a concentration gradient distribution.

4. The method for simulating concrete deterioration under continuous load-chloride ion coupling according to claim 3, characterized in that, Following the calculation of chloride ion diffusion coefficients under different loading levels and stress types using experimental parameters, the process also includes steps for obtaining degradation parameters and simulating structural performance, as detailed below: Images of crack distribution in the cross section of the specimen were acquired and the geometric features of the cracks were extracted. The acoustic emission signals recorded during the experiment were exported and the acoustic emission characteristic parameters were analyzed. Uniaxial stress-strain curve experiments were conducted on the specimen to obtain mechanical property parameters. Based on the mechanical performance parameters, the relevant formulas of concrete damage plastic constitutive model are substituted into the formulas to calculate the tensile and compressive damage evolution parameters of the deteriorated part. A multidimensional feature matrix containing crack characteristics, acoustic emission parameters, residual strength, and chloride ion diffusion coefficient is constructed. An end-to-end mapping relationship between the feature matrix and CDP parameters is established through a deep learning network, and the damage factor is obtained by inversion. The damage factors obtained by inversion are fused with the calculated damage evolution parameters, and combined with the concentration gradient and penetration depth distribution corresponding to the chloride ion diffusion coefficient, to form comprehensive material parameters of the deteriorated part. Establish a structural service mechanics model, assign the integrated material parameters to the structural service mechanics model according to the actual tension and compression zones of the structure, and set boundary conditions and continuous load conditions consistent with the preset structural service scenarios. After verifying the convergence of the structural service mechanics model, the model is submitted for mechanical performance calculation. Based on the calculation results, the structural bearing capacity degradation law, stiffness attenuation characteristics and stress distribution state are analyzed. The calculation results are compared with the experimental measured data of the test block under the same conditions to verify the accuracy of the model. If the error is ≤ the preset threshold, the simulation of the mechanical performance degradation of the concrete structure is completed.

5. The method for simulating concrete deterioration under continuous load-chloride ion coupling according to claim 4, characterized in that, An end-to-end mapping relationship between the feature matrix and CDP parameters is established through a deep learning network, and the damage factors are obtained by inversion, including: Principal component analysis is used to reduce the dimensionality of the multidimensional feature matrix and remove redundant information. An end-to-end deep learning network is constructed. The input layer receives the dimensionality-reduced multidimensional feature matrix, the hidden layer adopts the transfer learning strategy, the initial weights of the network are pre-trained based on the mechanical-erosion dataset of standard concrete specimens, and the number of neurons in the hidden layer is dynamically adjusted to adapt to the feature dimension. The output layer is directly mapped to the damage factor. A pre-defined two-stage optimization machine is introduced to optimize the constructed deep learning network; During the training of deep learning networks, the constitutive equation of uniaxial tension / compression of concrete is transformed into network constraints. A physical constraint penalty term is added to the loss function. When the predicted damage factor deviates from the mechanical law, the penalty weight is increased according to the degree of deviation. Using a deep learning network model trained with two-stage optimization and physical constraints, inference is performed on the input dimensionality-reduced feature matrix to output a damage factor that meets the accuracy requirements and conforms to physical laws.

6. The method for simulating concrete deterioration under continuous load-chloride ion coupling according to claim 5, characterized in that, The two-stage optimization mechanism includes: The first stage uses the NSGA-II algorithm for multi-objective optimization, with the sum of squared errors between the CDP parameter predictions and experimental measurements as the accuracy index and the number of network iterations as the efficiency index, outputting the accuracy-efficiency Pareto optimal solution. Based on the optimal solution, we proceed to the second stage, where the network weights are dynamically adjusted using the Bayesian regularization algorithm to reduce the risk of overfitting.

7. The method for simulating concrete deterioration under continuous load-chloride ion coupling according to claim 4, characterized in that, The damage factors obtained from the inversion are combined with the calculated damage evolution parameters, and then, by incorporating the concentration gradient and penetration depth distribution corresponding to the chloride ion diffusion coefficient, comprehensive material parameters for the deteriorated sites are formed, including: Collect the damage factors obtained from the inversion, the tensile / compressive damage evolution parameters obtained from the calculation, and the concentration gradient and penetration depth distribution data corresponding to the chloride ion diffusion coefficient. Based on the inversion error, theoretical fit and measurement accuracy of each parameter, determine their respective fusion weights. Parameters are adapted for the division of the tensile and compressive zones of the structure. In the tensile zone, the damage factor is associated with the quantitative relationship between crack density and penetration depth. In the compressive zone, the damage evolution parameters are matched with the correspondence between stiffness degradation rate and concentration gradient. Based on the zonal adaptation relationship, the mechanical parameters and corresponding chloride ion parameters are weighted according to preset weights, and the erosion influence coefficient is introduced simultaneously to achieve quantitative coupling, so as to obtain the zonal fusion parameters. Among them, the mechanical parameters include the damage factor of the tensile zone and the damage evolution parameters of the compressive zone, and the chloride ion parameters include the penetration depth and concentration gradient. The partition fusion parameters are spatially mapped according to the penetration depth, and a gradient parameter that changes continuously along the erosion depth is formed by interpolation, which fits the actual spatial distribution of degradation. The gradient parameters are verified based on the constitutive equation of uniaxial tension / compression of concrete, and the parameters that deviate from the physical laws are corrected according to the constraints. The corrected gradient parameters are correlated with the tensile / compression zone identifier and the penetration depth coordinates to form comprehensive material parameters that include mechanical degradation, chloride ion erosion, and spatial distribution characteristics.

8. A simulation system for concrete deterioration under continuous load-chloride ion coupling, characterized in that, It includes a memory, a processor, and a program stored in the memory and executable on the processor, which, when loaded and executed by the processor, implements a method for simulating concrete deterioration under continuous load-chloride ion coupling as described in any one of claims 1 to 7.

Images