Corrosion testing system and corrosion rate monitoring method for tunnel lining steel bars

Abstract

The invention provides a corrosion testing system and a corrosion rate monitoring method for tunnel lining steel bars, and relates to the technical field of tunnel engineering, and the method comprises the following steps: constructing a prediction model, determining the hidden layer weight and bias of the prediction model by an artificial ecosystem optimization algorithm, and obtaining an initial population of the artificial ecosystem optimization algorithm by cubic mapping; acquiring a training sample and actually measured strain data; the prediction model is trained through a training sample, a corrosion rate prediction model is obtained, and the training sample comprises lining strain data and steel bar corrosion rate data; and inputting the actually measured strain data into the corrosion rate prediction model to obtain the corrosion rate. According to the method, ICM, AEO and ELM are combined, a population in an AEO optimization algorithm is initialized through a random number generated by a chaotic system, the diversity of the population is improved, and falling into a local optimal solution is avoided. And carrying out hyper-parameter optimization on the ELM by using ICM-AEO to obtain optimal values of the weight and the offset of the hidden layer, thereby improving the prediction precision of the whole prediction model.

Description

Technical Field

[0001] The present invention relates to the technical field of tunnel engineering, and in particular to a corrosion testing system and a corrosion rate monitoring method for tunnel lining steel bars. Background Art

[0002] The geological conditions in the mountainous areas of western my country are complex. The tunnels to be built need to pass through complex and dangerous geological environments. There are many adverse engineering geological conditions, and the working environment of the tunnel structure is very complex. Under the long-term action of high-pressure water head and chemical erosion, the steel bar corrosion problem is often caused by excessive chloride ion concentration. After the steel bar is corroded, the volume expands, causing the surrounding concrete to crack under tension and the cracks gradually extend to the surface of the structure. The generation of cracks will further accelerate the corrosion of the steel bar. When the amount of steel bar corrosion reaches a certain extent, the bonding force between the steel bar and the concrete will begin to decrease, and the cross-sectional area participating in the force will decrease after corrosion, which seriously reduces the bearing capacity of the structure and has a significant impact on the safety and durability of the lining structure. However, since the steel bar is buried inside the concrete, the current non-destructive testing technology cannot directly measure the amount of steel bar corrosion in the structure. Therefore, it is urgent to establish a new type of non-destructive intelligent prediction method to realize the continuous monitoring of the corrosion rate of tunnel lining steel bars. Summary of the invention

[0003] The purpose of the present invention is to provide a corrosion testing system and corrosion rate monitoring method for tunnel lining steel bars to improve the above problems. In order to achieve the above purpose, the technical solution adopted by the present invention is as follows: In a first aspect, the present application provides a method for monitoring the corrosion rate of tunnel lining steel bars, comprising: Constructing a prediction model, wherein the hidden layer weights and biases of the prediction model are determined by an artificial ecosystem optimization algorithm, and the initial population of the artificial ecosystem optimization algorithm is obtained by cubic mapping; Acquire training samples and measured strain data; the training samples are obtained by a corrosion device test, the corrosion device includes a lining component in which a plurality of steel bars are buried, and a plurality of strain optical fibers are arranged on the surface of the lining component; The prediction model is trained by training samples to obtain a corrosion rate prediction model, wherein the training samples include lining strain data and steel bar corrosion rate data; The measured strain data are input into the corrosion rate prediction model to obtain the corrosion rate.

[0004] In a second aspect, the present application also provides a tunnel lining steel bar corrosion testing system, comprising: The first module is used to construct a prediction model, wherein the hidden layer weights and biases of the prediction model are determined by an artificial ecosystem optimization algorithm, and the initial population of the artificial ecosystem optimization algorithm is obtained by cubic mapping; The second module is used to obtain training samples and measured strain data; the training samples are obtained by a corrosion device test, the corrosion device includes a lining component with a plurality of steel bars buried therein, and a plurality of strain optical fibers are arranged on the surface of the lining component; The third module is used to train the prediction model through training samples to obtain a corrosion rate prediction model, wherein the training samples include lining strain data and steel corrosion rate data; The fourth module is used to input the measured strain data into the corrosion rate prediction model to obtain the corrosion rate.

[0005] The beneficial effects of the present invention are: The present invention constructs an ICM-AEO-ELM prediction model, combines ICM, AEO and ELM, and can initialize the population in the AEO optimization algorithm through random numbers generated by a chaotic system, thereby improving the diversity of the population and avoiding falling into a local optimal solution. ICM-AEO is used to optimize the hyperparameters of ELM to obtain the optimal values ​​of the hidden layer weights and biases in the extreme learning machine, thereby improving the prediction accuracy of the entire prediction model, and learning the training samples obtained from the experiment according to the prediction model, and finally forming an accurate corrosion rate prediction model, which provides a reference for on-site non-destructive detection of lining steel corrosion rate.

[0006] Other features and advantages of the present invention will be set forth in the following description, and in part will become apparent from the description, or may be understood by practicing embodiments of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0008] Figure 1 This is a flow chart of the method for monitoring the corrosion of tunnel lining steel bars in an embodiment of the present application; Figure 2 This is a schematic diagram of the structure of the corrosion device in the embodiment of the present application; Figure 3 This is a structural diagram of the tunnel lining steel bar corrosion testing system in an embodiment of the present application.

[0009] Markings in the figure: 100-lining member; 101-corbel; 200-optical fiber; 300-corrosion component; 400-reaction device; 401-reaction steel beam; 402-threaded steel bar; 500-jack; 600-adjustment device; 601-pad; 602-guide block; 710-first module; 711-first unit; 712-second unit; 713-first subunit; 714-second subunit; 715-third subunit; 720-second module; 730-third module; 740-fourth module; 750-fifth module; 760-sixth module; 770-seventh module; 780-eighth module. DETAILED DESCRIPTION

[0010] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0011] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0012] At present, my country's standards mainly use steel bar corrosion potential, concrete resistivity and structural surface disease conditions to qualitatively describe the corrosion condition of steel bars. However, these traditional lining steel bar corrosion rate measurement methods are greatly affected by environmental interference and the measurement results are inaccurate. Therefore, a new corrosion rate monitoring method is needed to improve the prediction accuracy of the corrosion rate and provide a reference for on-site detection of lining steel bar corrosion rate.

[0013] Embodiment 1: See also Figure 1 , this embodiment provides a method for monitoring the corrosion rate of tunnel lining steel bars, comprising: steps S100, S200, S300 and S400; S100, constructing a prediction model, wherein the hidden layer weights and biases of the prediction model are determined by an artificial ecosystem optimization algorithm, and the initial population of the artificial ecosystem optimization algorithm is obtained by cubic mapping; At least 20 populations are generated in sequence by using cubic mapping, and during the cubic mapping process of each population generation, information of the first two generated populations is fused according to a preset weight, and the last generated population is used as the target population; and the target population is used as the initial population; ICM mapping (Improved Cubic Map) helps the group-based intelligent optimization algorithm generate a diverse initial population by introducing the memory effect. The memory mechanism enables the optimization algorithm to better explore the solution space, reduce the initial redundant individuals, and avoid falling into the local optimal solution.

[0014] In this embodiment, 30 populations are generated by using cubic mapping (i.e., the number of iterations of cubic mapping is 29 times), and m represents the mth population; The expression of the cubic mapping with memory effect is: ; in, is the individual state in the m-1th population, is the individual state in the mth population, is the individual state in the m+1th population; , is the weight coefficient of memory effect, when When , the mapping system is a standard cubic mapping, when When the current mapping depends entirely on ; , is the control parameter of the cubic mapping, by changing As the value of , it can be observed that this chaotic mapping gradually transitions from periodic behavior to chaotic behavior.

[0015] Iterate and update the roles of individuals starting from the initial population, specifically: Producers, consumers and decomposers are updated according to the predator-prey relationship between roles and the energy level of roles. The Artificial Ecosystem Optimization Algorithm (AEO) is a new intelligent optimization algorithm proposed based on the energy flow in the ecosystem and the behaviors of producers, consumers and decomposers. It mainly includes the production stage, consumption stage and decomposition stage. Organisms represent feasible solutions to the problem, and decomposers represent the optimal solution to the problem and approach the theoretical optimal solution with iteration. The role update process is as follows: Producer Update: Producer (The worst individual) through the decomposer (best individual) and random individual The linear coupled iterative update of to guide the evolution of herbivores and omnivores is expressed as: ; in: is the population size, is the weighting coefficient, and are the current and maximum number of iterations, respectively. is a random number; is a random individual in the solution space [L,U] and For one dimensional random vector.

[0016] Consumer update: Consumers can be divided into herbivores, omnivores, and carnivores according to their feeding characteristics. In the model, they are randomly assigned role categories according to probability and iteratively updated in the following three modes.

[0017] Case 1: Herbivores Only prey on producers , the corresponding behavior model can be mathematically expressed as: ; in: , is the consumption factor based on the Levy flight improvement, and and are all random numbers that follow a standard normal distribution.

[0018] Case 2: Carnivores Can only randomly prey on a consumer with a higher energy level , the corresponding behavior model can be expressed as: ; in: , .

[0019] Case 3: Omnivore Prey producers and a random consumer with a higher energy level , the corresponding behavior model is: , ; in: , is a random number.

[0020] Decomposer Update: Every organism in an ecosystem Will perish and be decomposed Decomposition, the corresponding decomposition behavior model is: ; in: , is the factorization and It follows a standard normal distribution. and are weighted coefficients and is a random number.

[0021] The hidden layer weights and biases of the prediction model are modified according to the roles obtained in each round of updates, and the test set is input into the prediction model to obtain the classification accuracy corresponding to each role; The classification accuracy is used as the energy level (fitness value) of each role to update the population, and iterative calculation is performed. When the maximum number of iterations is reached or the convergence condition is met, the iteration is stopped, the optimal solution of the decomposer is output, and the optimal values ​​of the hidden layer weights and biases are obtained.

[0022] In each iteration, a chaotic sequence is generated by cubic mapping, a disturbance value is generated based on the chaotic sequence mapping, and the individual position is disturbed according to the disturbance value. When the maximum number of iterations is reached or the convergence condition is met, the iteration is stopped to obtain the optimal value of the hidden layer weight and bias.

[0023] In this application, the extreme learning machine (ELM) is used to build the model, and the AEO algorithm is used to optimize the optimal values ​​of the hidden layer weights and biases of the extreme learning machine. The hidden layer output is calculated by the activation function for the input data. , taking the minimization of the square difference between the network output and the sample output target matrix as the training error objective function, solving the output weight, and using the trained weights for prediction.

[0024] S200, obtaining training samples and measured strain data; the training samples are obtained by a corrosion device test, the corrosion device includes a lining component in which a plurality of steel bars are buried, and a plurality of strain optical fibers are arranged on the surface of the lining component; The training samples include lining strain data and steel corrosion rate data, and the method for obtaining the lining strain data and steel corrosion rate data includes: Obtaining a preset corrosion rate; the preset corrosion rate may be randomly given according to a known steel corrosion rate of tunnel lining in the art, which is used to determine experimental parameters to simulate a real environment; Calculating the corrosion duration based on the preset corrosion rate and test parameters; The steel bars are corroded according to the test parameters and corrosion duration, and the lining strain data generated during and after the corrosion process is measured by strain optical fiber; After the corrosion is complete, the corroded steel bars are taken out, and multiple sections of the corroded steel bars are cut at the same length as the spacing of the strain optical fiber layout, and the cross-sectional diameter of each section of the corroded steel bars is measured; Based on the cross-sectional diameter and length of the corroded steel bar, steel bar corrosion rate data is calculated; When obtaining measured strain data, strain optical fiber is laid on the surface of the tunnel lining to be measured in the same way as the corrosion device; the use of distributed optical fiber can simultaneously measure the strain data of multiple points on the tunnel lining to be measured.

[0025] S300, training a prediction model through training samples to obtain a corrosion rate prediction model, wherein the training samples include lining strain data and steel bar corrosion rate data; S400, inputting the measured strain data into a corrosion rate prediction model to obtain a corrosion rate; In the lining structure, generally, multiple steel bars are buried at the same time, and the cement cracking stress generated by the corrosion of each steel bar will affect the surrounding steel bars, that is, the strain distribution on the lining surface is jointly determined by multiple internal steel bars; in order to improve the prediction accuracy of the model under different conditions, this embodiment optimizes the corrosion rate prediction model based on the actual situation of the lining to be tested, as follows: Acquire strain data of multiple points on the surfaces of multiple lining components, construct multiple surface strain matrices, calculate the difference between adjacent elements to construct a distribution gradient matrix, and convert the distribution gradient matrix into a first heat map; The strain optical fibers are arranged equidistantly on the surface of the lining components and are perpendicular to the internal steel bars. The intersection of each strain optical fiber and the steel bar is the measurement point of the strain optical fiber. Therefore, the measurement points are actually distributed in an array form. The strain data of each point is filled into the matrix with the actual position to obtain the surface strain matrix. The difference between adjacent elements in the surface strain matrix can be calculated to construct a distribution gradient matrix, which can reflect the changes in the strain values ​​on the lining.

[0026] During the experiment, multiple lining components were made by using steel bars of different sizes and arranged at different intervals, and multiple first thermal maps were obtained respectively; The strain data of multiple points on the lining to be tested are obtained, and the measured gradient matrix of the lining to be tested is constructed and converted into a second thermal map. It should be noted that when the lining to be tested is tested, strain optical fibers are arranged on it in the same manner as the experimental device to obtain a measured gradient matrix of the same size as the distribution gradient matrix, and the second thermal map can be obtained based on the same operation; updating training weights for strain data corresponding to the first heat map based on a similarity between the second heat map and each of the first heat maps; In this step, the structural similarity index (SSIM) can be used to evaluate the similarity between heat maps or the images can be converted into hash values, and then the differences in hash values ​​can be compared to evaluate the similarity; a high similarity of the heat maps indicates that the strain distribution on the lining to be tested is similar to the strain distribution on the lining component of the experiment, so the strain data of the lining component has a higher reference value for the lining to be tested this time; The prediction model is trained using strain data with updated training weights to obtain an updated prediction model. After the update, test data with similar lining strain distribution has a greater impact on the prediction results, making the prediction results of the prediction model more accurate.

[0027] In actual measurement, the strain distribution gradient of the lining to be measured may be larger than the strain distribution gradient during the experiment due to differences in construction scale. In order to adjust the strain distribution of linings of different scales to the same scale for comparison; this step can normalize the data in the distribution gradient matrix and the measured gradient matrix respectively. After normalization, the data in the two matrices are between [0,1] and have the same scale.

[0028] As another optional implementation, the method for adjusting the data in the distribution gradient matrix and the measured gradient matrix to the same scale is: Calculating the mean of the strain data of the lining component to obtain a first mean; Calculate the mean of the strain data of the lining to be tested to obtain a second mean; Scaling the strain data of the lining to be measured according to the ratio of the first mean value to the second mean value to obtain processed data; A measured strain matrix of the lining to be measured is constructed based on the processed data.

[0029] In this step, after the corrosion rate prediction model is optimized through the above operations, the measured strain data is input into the optimized corrosion rate prediction model to obtain the corrosion rate.

[0030] Based on the above technical concept, see Figure 2 , the present invention provides a corrosion device, including a lining member 100, wherein the lining member 100 includes steel bars to be corroded buried in concrete; An optical fiber 200, wherein the optical fiber 200 is arranged on the surface of the lining member 100, and the optical fiber 200 and the steel bar to be corroded are perpendicular to each other; A corrosion component 300, wherein the corrosion component 300 is connected to the steel bar to be corroded and is used to corrode the steel bar to be corroded; A reaction device 400, wherein the reaction device 400 is disposed around the outside of the lining member 100, one end of the reaction device 400 is in contact with the lining member 100, and a gap is left between the other end of the reaction device 400 and the lining member 100; The jack 500 is disposed in a gap between the reaction device 400 and the lining member 100. The jack 500 is used to apply a load to the lining member 100, and the direction of the applied load is the same as the layout direction of the steel bars to be corroded.

[0031] The corrosion test system of the present invention is used to simulate a certain Cl- The corrosion of lining steel bars and the strain of lining under concentration conditions and bending moment axial force. During the corrosion process, concrete will produce corrosion cracks in the same direction as the steel bars, and tensile strains perpendicular to the direction of the steel bars will be generated on both sides of the cracks. Optical fibers can be laid at equal intervals on the concrete surface in the vertical direction of the steel bars to be corroded, and connected to the optical fiber demodulator to obtain and record the strain values ​​caused by steel bar corrosion cracking.

[0032] When designing lining components, it is necessary to obtain design data based on the tunnel site to be monitored and manufacture reinforced concrete lining components of corresponding sizes based on the design data.

[0033] As an optional embodiment, the upper and lower ends of the lining member 100 are provided with brackets 101, and the jack 500 applies load to the brackets 101. Lining brackets are a support structure provided in the tunnel lining structure, usually located in the tunnel side wall or arch, and used to support crane beams, rails or other equipment.

[0034] As an optional embodiment, the corrosion assembly 300 includes a corrosion tank, a stainless steel mesh, a corrosion liquid and a power supply; The stainless steel mesh is placed in a rusting tank, and the rusting liquid immerses the stainless steel mesh; The steel bar to be corroded is connected to the positive electrode of the power supply, and the stainless steel mesh is connected to the negative electrode of the power supply.

[0035] The power supply is a constant current DC power supply, and the rust solution is a NaCl solution, wherein Cl - The concentration is determined according to the chloride ion concentration of the tunnel site to be monitored. - The concentration can be measured using a chloride ion content meter with the aid of the ion selective electrode method (ISE method).

[0036] As an optional embodiment, the corrosion test system also includes an adjustment device 600. Since the actual force on the tunnel is constantly changing, the adjustment device is provided to adjust the position of the force applied by the jack to the lining. The adjustment device 600 includes a pad 601, and a lifting mechanism is connected and provided below the pad 601; the lifting mechanism includes two symmetrically arranged slide grooves, and a sliding bracket arranged in the slide groove, each sliding bracket is in an X-shaped structure and can rotate around a midpoint, the upper end of the sliding bracket is hinged to the pad, and the lower end of the sliding bracket slides in the slide groove and can be fixed, thereby realizing the lifting and lowering adjustment of the pad.

[0037] The jack is cylindrical in shape. In order to prevent the jack from shifting during the load application process, a guide block 602 is further provided on the pad 601. The jack 500 is arranged in the guide block 602. The jack just fits into the concave structure of the guide block and will not roll. The two ends of the jack 500 are respectively against the reaction device 400 and the lining member 100.

[0038] As an optional embodiment, the reaction device 400 includes two reaction steel beams 401 located at both ends, and also includes threaded steel bars 402 connecting the two reaction steel beams 401; each of the reaction steel beams 401 is provided with threaded holes at both ends; the two ends of the threaded steel bars 402 are respectively connected to the threaded holes of the two reaction steel beams 401, and the threaded steel bars 402 and the reaction steel beams 401 are perpendicular to each other to form a rectangular frame.

[0039] When the jack is working, its two ends respectively rest on the reaction steel beam and the lining component, applying load to the lining component. At the same time, the other end of the lining component is constrained by another reaction steel beam, which generates a reaction force on the lining component.

[0040] In addition, a pad is provided on the pad plate for raising the reaction steel beam. After raising the pad, the reaction steel beam is kept at the same height as the axis of the jack to ensure that there is no deviation in the force.

[0041] As an optional implementation, the lining component 100 is further provided with ordinary steel bars, the surface of which is coated with epoxy resin and wrapped with insulating tape. When making the lining component, multiple steel bars are embedded in the concrete in the horizontal, vertical and along the inclined surface of the corbel, among which only the horizontal steel bars need to be tested for corrosion, and the other steel bars need to be insulated.

[0042] Based on the corrosion device provided in this application, the lining strain data and steel corrosion rate data are obtained. The specific operations are as follows: First determine the parameters for applying the load: The stress data of the lining in the rusted area can be obtained by monitoring the tunnel on site. Specifically, vibrating wire concrete strain gauges can be arranged on the inner and outer sides of the lining to obtain the inner strain and lateral strain ; According to the inner strain and lateral strain Calculate the eccentricity of the jack : ; In the formula, is the moment of inertia of the lining member, is the cross-sectional thickness of the lining member, is the inner strain, is the outer strain, is the eccentric load magnitude, and Should be less than the cracking load To avoid concrete cracking affecting the corrosion rate; Cracking load The calculation formula is: ; In the formula, is the plastic coefficient of section resistance moment, is the tensile stress limiting coefficient of concrete, is the standard value of concrete axial tensile strength, To convert the cross-sectional area (equalizing the steel bar area to the concrete area), is the converted interface tensile edge resistance moment.

[0043] Before testing, an eccentric load can be estimated , then calculate the eccentricity , then judge Is it less than If yes, then you can use this parameter to test, if not, adjust the eccentric load Recalculate until the requirements are met.

[0044] Then determine the duration of corrosion: Get the preset corrosion rate; Calculating the corrosion duration based on the preset corrosion rate and test parameters, wherein the test parameters include the molar mass of the iron element, the surface current density of the steel bar to be corroded, the radius of the steel bar to be corroded, and the density of the steel bar to be corroded; The corroded steel bars are corroded according to the test parameters and the corrosion time, and the lining strain data generated during and after the corrosion process is measured through the optical fiber 200.

[0045] Specifically, first preset a suitable corrosion rate, and calculate the corrosion duration according to the preset corrosion rate. The calculation formula is as follows: ; In the formula, is the molar mass of iron, 56 g / mol, The current density on the surface of the steel bar to be corroded should be less than or equal to 500 , For the duration of rust, is the Faraday constant, 96500C / mol, is the radius of the steel bar to be corroded, The density of the steel bars to be corroded.

[0046] According to the parameters determined above, the corrosion test is carried out, the steel bar to be corroded is connected to the positive electrode of the constant current DC power supply, the stainless steel mesh is connected to the negative electrode of the constant current DC power supply, the power is turned on for corrosion, and the eccentric load is applied to the lining component at the same time; the strain value measured by the optical fiber during and after the corrosion process is recorded.

[0047] After the corrosion is complete, the corroded steel bars are taken out and cleaned with a rust removal liquid to remove the corrosion products on the surface of the corroded steel bars; The length is the same as the optical fiber 200 layout spacing. Cut off multiple sections of corroded steel bars and measure the cross-sectional diameter of each section of corroded steel bars. It should be noted that the setting position of each optical fiber corresponds to the buried position of a section of corroded steel bars. Therefore, the strain of the optical fiber is caused by the cracks caused by the corrosion of the section of steel bars. The data of each optical fiber corresponds to a section of corroded steel bars one by one, forming a strain-corrosion rate data set. The steel bar corrosion rate data is calculated based on the cross-sectional diameter and length of the corroded steel bar.

[0048] For each section of corroded steel bar, measure the and vertical cross rib diameter ,Pick and The average value is taken as the corrosion diameter of the remaining section of the corroded steel bar. , The calculation formula of steel bar corrosion rate is: ; In the formula, is the steel bar corrosion rate, is the length of the corroded steel bar, is the rust diameter.

[0049] Taking a tunnel as an example, the corrosion rate monitoring method of the tunnel lining steel bars of this application is described as follows: The length of a certain tunnel is 8624.5m (average length of the left and right tunnels), and the maximum burial depth of the tunnel is 923.2m. The gypsum-salt strata are rich in gypsum-salt minerals. The tunnel site is severely corroded by chlorates, resulting in a series of defects such as rust of lining steel bars, cracking and spalling of concrete, surface crystallization, and strength degradation. There is an urgent need for "diagnosis and treatment" projects for the tunnel.

[0050] Determine the size of the eccentric load applied in the test based on the actual situation on site , eccentricity , the concentration of NaCl solution is 3%. In order to avoid the difference between the test results and the steel corrosion under natural conditions caused by excessive current density, the current density of this test is , nominal diameter of corroded steel bars , the target corrosion zone length is , when the corrosion test is carried out for 674.4 h (i.e. 28.1 days), the steel bar corrosion rate can reach The current of each steel bar to be corroded during the corrosion period During the test, the four steel bars to be corroded in each component are connected in parallel as anodes and share a DC power supply with an output current of 1590mA. The optical fiber is laid in the same range as the corrosion area, with a spacing of After the corrosion is over, destroy the concrete, take out the steel bars, and Steel bars were intercepted to calculate the steel bar corrosion rate, and a total of 264 training data were obtained. Some training samples are shown in Table 1.

[0051] Table 1 Training samples

[0052] Based on the prediction model of the present application, the above-mentioned training samples are trained; in addition, some test components are prepared for testing, and the actual corrosion rate of the test components is obtained according to the piezoelectric ceramic technology. Then, the corrosion rate of the test components is predicted using the above-mentioned prediction model, and the prediction error rate is calculated based on the obtained partial actual data and predicted data. As shown in Table 2, the error rate between the prediction of some data and the actual data is roughly within 10%, and the average error rate is 5.935%. It can be seen that the above prediction model has good accuracy.

[0053] Table 2 Prediction samples and prediction error rates

[0054] Embodiment 2: See also Figure 3 This embodiment provides a tunnel lining steel bar corrosion testing system, comprising: The first module 710 is used to construct a prediction model, wherein the hidden layer weights and biases of the prediction model are determined by an artificial ecosystem optimization algorithm, and the initial population of the artificial ecosystem optimization algorithm is obtained by cubic mapping; The second module 720 is used to obtain training samples and measured strain data; the training samples are obtained by a corrosion device test, the corrosion device includes a lining component with a plurality of steel bars buried therein, and a plurality of strain optical fibers are arranged on the surface of the lining component; The third module 730 is used to train the prediction model through training samples to obtain a corrosion rate prediction model, wherein the training samples include lining strain data and steel corrosion rate data; The fourth module 740 is used to input the measured strain data into the corrosion rate prediction model to obtain the corrosion rate.

[0055] As an optional implementation, the first module 710 includes: The first unit 711 is used to sequentially generate at least 20 populations by using cubic mapping, and in the process of generating each population by cubic mapping, information of the first two generated populations is fused according to a preset weight, and the last generated population is used as the target population; and the target population is used as the initial population; The second unit 712 is used to iteratively update the role of the individual starting from the initial population; in each iteration, a cubic mapping is used to generate a chaotic sequence, a disturbance value is generated based on the chaotic sequence mapping, and the individual position is disturbed according to the disturbance value. When the maximum number of iterations is reached or the convergence condition is met, the iteration is stopped to obtain the optimal value of the hidden layer weight and bias.

[0056] As an optional implementation manner, the second unit 712 includes: The first subunit 713 is used to update producers, consumers and decomposers according to the predator-prey relationship between the roles and the energy level of the roles; The second subunit 714 is used to modify the hidden layer weights and biases of the prediction model according to the roles obtained in each round of updates, and input the test set into the prediction model to obtain the classification accuracy corresponding to each role; The third subunit 715 is used to update the classification accuracy as the energy level of each role to the population, perform iterative calculation, and stop iteration when the maximum number of iterations is reached or the convergence condition is met, output the optimal solution of the decomposer, and obtain the optimal values ​​of the hidden layer weights and biases.

[0057] As an optional implementation, the corrosion testing system further includes: The fifth module 750 is used to obtain strain data of multiple points on the surface of multiple lining components, construct multiple surface strain matrices, calculate the difference between adjacent elements to construct a distribution gradient matrix, and convert the distribution gradient matrix into a first heat map; The sixth module 760 is used to obtain strain data of multiple points on the lining to be tested, construct a measured gradient matrix of the lining to be tested and convert it into a second thermal map; A seventh module 770 is used to update the training weights for the strain data corresponding to the first heat map based on the similarity between the second heat map and each first heat map; The eighth module 780 is used to train the prediction model using the strain data with updated training weights to obtain an updated prediction model.

[0058] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A method for monitoring the corrosion rate of tunnel lining steel bars, characterized in that: include: Constructing a prediction model, wherein the hidden layer weights and biases of the prediction model are determined by an artificial ecosystem optimization algorithm, and the initial population of the artificial ecosystem optimization algorithm is obtained by cubic mapping; Acquire training samples and measured strain data; the training samples are obtained by a corrosion device test, the corrosion device includes a lining component in which a plurality of steel bars are buried, and a plurality of strain optical fibers are arranged on the surface of the lining component; The prediction model is trained by training samples to obtain a corrosion rate prediction model, wherein the training samples include lining strain data and steel bar corrosion rate data; The measured strain data are input into the corrosion rate prediction model to obtain the corrosion rate.

2. The method for monitoring the corrosion rate of tunnel lining steel bars according to claim 1, characterized in that: The constructing of the prediction model comprises: At least 20 populations are generated in sequence by using cubic mapping, and during the cubic mapping process of each population generation, information of the first two generated populations is fused according to a preset weight, and the last generated population is used as the target population; and the target population is used as the initial population; Iteratively update the roles of individuals starting from the initial population; in each iteration, a chaotic sequence is generated by using a cubic mapping, a disturbance value is generated based on the chaotic sequence mapping, and the individual position is disturbed according to the disturbance value. When the maximum number of iterations is reached or the convergence condition is met, the iteration is stopped to obtain the optimal values ​​of the hidden layer weights and biases.

3. The method for monitoring the corrosion rate of tunnel lining steel bars according to claim 2, characterized in that: Iteratively updating the roles of individuals starting from the initial population includes: Producers, consumers, and decomposers are updated based on the predator-prey relationships between roles and the energy levels of the roles; The hidden layer weights and biases of the prediction model are modified according to the roles obtained in each round of updates, and the test set is input into the prediction model to obtain the classification accuracy corresponding to each role; The classification accuracy is used as the energy level of each role to update the population, and iterative calculation is performed. When the maximum number of iterations is reached or the convergence condition is met, the iteration is stopped, the optimal solution of the decomposer is output, and the optimal values ​​of the hidden layer weights and biases are obtained.

4. The method for monitoring the corrosion rate of tunnel lining steel bars according to claim 1, characterized in that: The method comprises: Acquire strain data of multiple points on the surfaces of multiple lining components, construct multiple surface strain matrices, calculate the difference between adjacent elements to construct a distribution gradient matrix, and convert the distribution gradient matrix into a first heat map; Acquire strain data of multiple points on the lining to be tested, construct a measured gradient matrix of the lining to be tested and convert it into a second heat map; updating training weights for strain data corresponding to the first heat map based on a similarity between the second heat map and each of the first heat maps; The prediction model is trained using the strain data with updated training weights to obtain an updated prediction model.

5. The method for monitoring the corrosion rate of tunnel lining steel bars according to claim 4, characterized in that: Obtain strain data of multiple points on the lining to be tested and construct the measured strain matrix of the lining to be tested, including: Calculating the mean of the strain data of the lining component to obtain a first mean; Calculate the mean of the strain data of the lining to be tested to obtain a second mean; Scaling the strain data of the lining to be measured according to the ratio of the first mean value to the second mean value to obtain processed data; A measured strain matrix of the lining to be measured is constructed based on the processed data.

6. The method for monitoring the corrosion rate of tunnel lining steel bars according to claim 1, characterized in that: The method for obtaining the lining strain data and the steel bar corrosion rate data comprises: Get the preset corrosion rate; Calculating the corrosion duration based on the preset corrosion rate and test parameters; The steel bars are corroded according to the test parameters and corrosion duration, and the lining strain data generated during and after the corrosion process is measured by strain optical fiber; After the corrosion is complete, the corroded steel bars are taken out, and multiple sections of the corroded steel bars are cut at the same length as the spacing of the strain optical fiber layout, and the cross-sectional diameter of each section of the corroded steel bars is measured; The steel bar corrosion rate data is calculated based on the cross-sectional diameter and length of the corroded steel bar.

7. A tunnel lining steel bar corrosion testing system, characterized in that: include: The first module is used to construct a prediction model, wherein the hidden layer weights and biases of the prediction model are determined by an artificial ecosystem optimization algorithm, and the initial population of the artificial ecosystem optimization algorithm is obtained by cubic mapping; The second module is used to obtain training samples and measured strain data; The training samples are obtained from a corrosion device test, wherein the corrosion device includes a lining component in which a plurality of steel bars are buried, and a plurality of strain optical fibers are arranged on the surface of the lining component; The third module is used to train the prediction model through training samples to obtain a corrosion rate prediction model, wherein the training samples include lining strain data and steel corrosion rate data; The fourth module is used to input the measured strain data into the corrosion rate prediction model to obtain the corrosion rate.

8. A tunnel lining steel bar corrosion testing system according to claim 7, characterized in that: The first module includes: The first unit is used to sequentially generate at least 20 populations by using cubic mapping, and in the cubic mapping process of each population generation, information of the first two generated populations is fused according to a preset weight, and the last generated population is used as the target population; and the target population is used as the initial population; The second unit is used to iteratively update the role of the individual starting from the initial population; in each iteration, a cubic mapping is used to generate a chaotic sequence, a disturbance value is generated based on the chaotic sequence mapping, and the individual position is disturbed according to the disturbance value. When the maximum number of iterations is reached or the convergence condition is met, the iteration is stopped to obtain the optimal value of the hidden layer weight and bias.

9. A tunnel lining steel bar corrosion testing system according to claim 8, characterized in that: The second unit comprises: The first subunit is used to update producers, consumers and decomposers according to the predator-prey relationship between roles and the energy level of roles; The second subunit is used to modify the hidden layer weights and biases of the prediction model according to the roles obtained in each round of updates, and input the test set into the prediction model to obtain the classification accuracy corresponding to each role; The third subunit is used to update the classification accuracy as the energy level of each role to the population, perform iterative calculation, and stop the iteration when the maximum number of iterations is reached or the convergence condition is met, output the optimal solution of the decomposer, and obtain the optimal values ​​of the hidden layer weights and biases.

10. A tunnel lining steel bar corrosion testing system according to claim 7, characterized in that: Also includes: The fifth module is used to obtain strain data of multiple points on the surface of multiple lining components, construct multiple surface strain matrices, calculate the difference between adjacent elements to construct a distribution gradient matrix, and convert the distribution gradient matrix into a first heat map; The sixth module is used to obtain strain data of multiple points on the lining to be tested, construct a measured gradient matrix of the lining to be tested and convert it into a second thermal map; A seventh module, configured to update training weights for strain data corresponding to the first heat map based on a similarity between the second heat map and each of the first heat maps; The eighth module is used to train the prediction model using the strain data with updated training weights to obtain an updated prediction model.

Images