Space-time-corrosion dynamic environment effect grading method for coastal concrete structure

Abstract

The invention discloses a dynamic environmental effect grading method for a coastal concrete structure. The method comprises the following steps: firstly, constructing a coupling space-time database of the surface chloride ion concentration, the chloride ion diffusion coefficient and the steel bar corrosion rate; then, an improved space-time Kriging interpolation model (MSTK) is adopted to reconstruct space chloride ion distribution, and the weight coefficient of the space-time distance is optimized through a particle swarm; and meanwhile, a Hilbert dynamic probability interpolation model (HDPI) is combined with Bayesian optimization LSTM to realize time dimension continuous prediction, and the chloride ion concentration is decomposed into a deterministic trend, low-frequency environment fluctuation and high-frequency material fluctuation. Based on the interpolation result and the actually measured corrosion rate, high, medium and low corrosion risk grades are automatically divided through hierarchical clustering; and finally, quantizing the contribution degree of the detection points based on the diagonal elements of the prediction matrix, and optimizing the layout of the detection points. The method is high in precision (predicting R2gt; and the cost is low (detection points are reduced by 37.5%), and differentiated protection design of coastal electric power facilities is supported.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of concrete structure durability evaluation and marine engineering corrosion monitoring, and more particularly to a coastal concrete structure dynamic environmental action grade division method based on space-time interpolation and corrosion response coupling. BACKGROUND

[0002] The current national standard "Concrete Structure Durability Design Standard" (GB / T50476) divides marine environment into qualitative grades such as III-A, III-B and III-C, lacking quantitative, dynamic and spatial differentiation description of key parameters such as chloride ion concentration and diffusion coefficient. In actual engineering, the corrosion rates of coastal power facilities such as substations and transmission towers are significantly different at different elevations such as splash zone, tidal zone and atmospheric zone. The traditional "one-size-fits-all" protection design is easy to lead to insufficient protection or over-protection, causing safety hazards or economic losses.

[0003] The existing environmental action evaluation method mainly relies on static Kriging spatial interpolation or single time series prediction, which is difficult to simultaneously depict the coupled evolution law of chloride ion concentration under spatial non-uniformity (such as elevation, wind direction) and time non-stationarity (such as dry-wet cycle, seasonal change), and has no correlation with the actual corrosion response of steel bars, which cannot support precise differentiated durability design. SUMMARY

[0004] The technical problem to be solved by the present application is to overcome the deficiencies in the prior art and provide a coastal concrete structure dynamic environmental action grade division method based on space-time interpolation and corrosion response coupling.

[0005] The coastal concrete structure dynamic environmental action grade division method based on space-time interpolation and corrosion response coupling of the present application is realized through the following technical scheme, and a four-step closed-loop method of "database construction-space-time interpolation-grade division-detection optimization" is proposed, including the following steps:

[0006] 1. Constructing a "chloride ion concentration-diffusion coefficient-corrosion response" coupling database

[0007] Through field exposure test and indoor accelerated test, the time series data of surface chloride ion concentration C s (t) and chloride ion diffusion coefficient D(t) of concrete specimens at different elevations (atmospheric zone AZ, splash zone SP, tidal zone TZ) are collected, and the real corrosion rate p of steel bars is measured synchronously, and a space-time-corrosion response coupling database is established.

[0008] 2. Proposing an MSTK-HDPI mixed space-time interpolation model

[0009] MSTK (Modified Spatio-Temporal Kriging) model: Particle Swarm Optimization (PSO) is introduced to dynamically generate non-uniform spatio-temporal distance field, accurately reconstructing the spatial distribution of missing points of chloride ions.

[0010] The spatio-temporal distance is defined as:

[0011]

[0012] where d s is the spatial distance (Euclidean distance is adopted), d t is the time interval, w s and w t are the weight coefficients optimized by PSO.

[0013] The fitness function of PSO is the cross-validation mean square error (MSE):

[0014]

[0015] HDPI (Hilbert Dynamic Probability Interpolation) model: Bayesian optimization LSTM (BO-LSTM) is combined to realize the continuous and fluctuation feature extraction of chloride ion parameter time series.

[0016] The HDPI model decomposes the chloride ion concentration into three parts: deterministic trend, low-frequency environmental fluctuation, and high-frequency material fluctuation, and predicts the fluctuation components through BO-LSTM.

[0017] The chloride ion concentration time-varying model is decomposed as:

[0018] C s (t) = f det (t) + f low (t) + f high (t) + ∈ (t)

[0019] where f det is the deterministic trend (such as the CECS model), f low (t) and f high (t) are the low-frequency (environmental) and high-frequency (material) fluctuation components, respectively.

[0020] The BO-LSTM hyperparameters (learning rate, number of layers, etc.) are optimized by a Gaussian process proxy model, and the objective function is:

[0021]

[0022] 3. Establish dynamic environmental action level division criteria

[0023] C s(x, y, z, t) and D(x, y, z, t) as input, combined with the measured corrosion rate of steel bar ρ, the high, medium and low corrosion risk areas are automatically divided by hierarchical clustering (Hierarchical Clustering):

[0024] RiskLevel=Cluster([C s ,D,ρ])

[0025] The output environment effect level mapping diagram guides the differentiated protection design (such as high-risk area adopts sacrificial anode + epoxy coating composite protection).

[0026] The corrosion risk level division result is used to guide the differentiated protection design of the concrete structure of the coastal power engineering, including the selection of protection materials, the thickness of the protection layer and the development of maintenance strategies.

[0027] 4. Contribute to the optimization method of detection points

[0028] The prediction matrix H (HatMatrix) based on MSTK interpolation quantifies the contribution of each detection point to the global information:

[0029] Contributionx=h ii ,i=1,2,...,n

[0030] Where h ii is the diagonal element of the prediction matrix. Through hierarchical clustering, the contribution is divided into high (> 0.75), medium (0.5-0.75) and low (< 0.5) three grades, and the low contribution points are removed, so that the cost-benefit ratio of detection is improved by 37.5%.

[0031] Compared with the prior art, the beneficial effects of the present application are:

[0032] 1. Precision improvement: the MSTK interpolation error is reduced by 42% compared with the traditional Kriging, and the prediction R 2 >0.96 of the chloride ion concentration.

[0033] 2. Cost optimization: the number of detection points is reduced by 37.5%, and the cost-benefit ratio of detection is improved from 62.5% to 100%.

[0034] 3. Engineering value: it is the first time to realize the leap from "static qualitative partitioning" to "dynamic quantitative partitioning", supporting the differentiation, precision and intelligent durability design of coastal power facilities. BRIEF DESCRIPTION OF DRAWINGS

[0035] Figure 1 is a space-time interpolation fusion flowchart;

[0036] Figure 2 is a detection point optimization and level division application flowchart. Detailed Implementation

[0037] To better understand the above-mentioned objects, features, and advantages of the present invention, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. Many specific details are set forth in the following description to provide a thorough understanding of the present invention. The described embodiments are merely some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this invention pertains. The terminology used herein is for the purpose of describing specific embodiments only and is not intended to limit the present invention.

[0038] Combination Figure 1 and Figure 2 This invention employs a four-step closed-loop method: "database construction—spatiotemporal interpolation—level classification—detection optimization," which includes the following steps:

[0039] 1. Construct a coupled database of "chloride ion concentration – diffusion coefficient – ​​corrosion response".

[0040] The surface chloride ion concentration (C0) of concrete specimens at different elevations (atmospheric zone AZ, splash zone SP, and tidal zone TZ) was collected through field exposure tests and indoor accelerated tests. s A spatiotemporal-corrosion response coupled database was established by using time-series data of chloride ion diffusion coefficient D(t) and the actual corrosion rate ρ of steel bars to be measured simultaneously.

[0041] 2. A hybrid spatiotemporal interpolation model of MSTK–HDPI is proposed.

[0042] The MSTK (Modified Spatio-Temporal Kriging) model introduces particle swarm optimization (PSO) to dynamically generate a non-uniform spatiotemporal distance field, accurately reconstructing the chloride ion distribution at spatially missing points.

[0043] Spatiotemporal distance is defined as:

[0044]

[0045] Where d s d represents the spatial distance (using Euclidean distance). t w is the time interval s With w t Weighting coefficients optimized for PSO.

[0046] The PSO fitness function is the cross-validation mean square error (MSE):

[0047]

[0048] HDPI (Hilbert Dynamic Probability Interpolation) model: combined with Bayesian optimization LSTM (BO-LSTM), to realize the continuous and volatility feature extraction of chloride ion parameter time series.

[0049] The HDPI model decomposes the chloride ion concentration into deterministic trend, low-frequency environmental fluctuations, and high-frequency material fluctuations, and predicts the fluctuation components through BO-LSTM.

[0050] The chloride ion concentration time-varying model is decomposed into:

[0051] C s (t)=f det (t)+f low (t)+f high (t)+∈(t)

[0052] Where f det is the deterministic trend (such as the CECS model), f low (t) and f high (t) are low-frequency (environmental) and high-frequency (material) fluctuation components, respectively.

[0053] BO-LSTM hyperparameters (learning rate, number of layers, etc.) are optimized through a Gaussian process proxy model, with the objective function being:

[0054]

[0055] 3. Establish dynamic environmental action level division criteria

[0056] The interpolated C s (x,y,z,t) and D(x,y,z,t) are taken as inputs, combined with the measured corrosion rate of steel bars ρ, and through hierarchical clustering (Hierarchical Clustering), high, medium, and low corrosion risk areas are automatically divided:

[0057] RiskLevel=Cluster([C s D,ρ])

[0058] Output the environmental action level mapping diagram to guide differentiated protection design (such as using sacrificial anode + epoxy coating composite protection in high-risk areas).

[0059] The corrosion risk grading result is used for guiding the differentiated protection design of the coastal electric power engineering concrete structure, including the protection material selection, the protection layer thickness and the maintenance strategy formulation.

[0060] 4. A contribution-driven detection point optimization method is proposed

[0061] A prediction matrix H (HatMatrix) based on MSTK interpolation is used to quantify the contribution of each detection point to the global information:

[0062] Contribution i =h ii , i = 1, 2,..., n

[0063] Where h ii is the diagonal element of the prediction matrix. By hierarchical clustering, the contribution is divided into high (> 0.75), medium (0.5-0.75), and low (< 0.5) three grades, and the low contribution points are removed to achieve a 37.5% improvement in the cost-benefit ratio of detection.

[0064] Embodiment:

[0065] Taking a 220kV substation concrete pillar in Tianjin Binhai as an example:

[0066] 1. Data collection: 24 detection points are arranged at 0-5m elevation of the pillar, and C s and D are collected every 3 months, and p is obtained by synchronous power-on accelerated corrosion.

[0067] 2. Spatio-temporal interpolation:

[0068] The PSO is used to optimize the MSTK weight, and w s = 3.2 and w t = 1.8;

[0069] The BO-LSTM is used to predict the missing time series, and the RMSE is 0.0174.

[0070] 3. Grade division: the cluster output splash zone (1.5-3.0m) is classified as high-risk area (grade III-C+), and the atmospheric zone (> 3.0m) is classified as medium-risk area (grade III-B).

[0071] 4. Detection optimization: 8 low contribution points h ii <0.5 are removed, and 16 high / medium contribution points are retained, and the detection cost is reduced by 37.5%.

[0072] The above only describes the preferred embodiments of the present application, and it should be pointed out that for ordinary skilled persons in the technical field, several improvements and refinements can be made without departing from the principles of the present application, and these improvements and refinements should also be regarded as the protection scope of the present application.

Claims

1. A method for dividing a time-space-corrosion dynamic environmental action grade of a coastal concrete structure, characterized in that, Comprising the following steps: S1. Constructing the "surface chloride concentration C s - chloride diffusion coefficient D - steel reinforcement corrosion rate p" coupled spatio-temporal database; S2. Reconstructing the spatial dimension chloride distribution by using the improved spatiotemporal Kriging interpolation model MSTK, wherein the spatiotemporal distance is defined as Where d s d represents the spatial distance (using Euclidean distance). t w is the time interval s With w t Weighting coefficients optimized for PSO; weight coefficient w s and w t Solved by particle swarm optimization algorithm with the goal of minimizing the cross-validation mean square error. S3. Realizing the time dimension continuous prediction by using the Hilbert dynamic probability interpolation model HDPI combined with the Bayesian optimization LSTM; The HDPI model: combined with the Bayesian optimization BO-LSTM, realizes the continuous and fluctuation feature extraction of the chloride ion parameter time series; The chloride ion concentration time-varying model is decomposed into: C s (t) = f det (t) + f low (t) + f high (t) + e(t) where f det is a deterministic trend, f low (t) is a low frequency ambient component, and f high (t) is a high frequency material fluctuation component; The BO-LSTM hyperparameters are optimized by the Gaussian process proxy model, and the objective function is: S4. Obtain C based on interpolation s (x, y, z, t) and D(x, y, z, t) and the measured p, automatically divide high, medium and low corrosion risk levels by hierarchical clustering; S5. Diagonal elements h of the MSTK prediction matrix ii Quantify the contribution of the detection points and optimize the layout of the detection points.

2. The method for dividing the time-space-corrosion dynamic environmental action grade of coastal concrete structures according to claim 1, characterized in that, In the MSTK model, the fitness function of the particle swarm optimization is 3. The method for dividing the time-space-corrosion dynamic environmental action grade of the coastal concrete structure according to claim 1, characterized in that, The HDPI model decomposes the chloride ion concentration into three parts: deterministic trend, low-frequency environmental fluctuation and high-frequency material fluctuation, and predicts the fluctuation components by BO-LSTM.

4. The method for dividing the time-space-corrosion dynamic environmental action grade of the coastal concrete structure according to claim 1, characterized in that, The detection point contribution threshold is set as: high contribution > 0.75, medium contribution 0.5-0.75, low contribution < 0.5, and low contribution points are removed to optimize the detection cost.

5. The method for dividing the time-space-corrosion dynamic environmental action grade of the coastal concrete structure according to claim 1, characterized in that, The corrosion risk grade division result is used to guide the differentiated protection design of coastal power engineering concrete structure, including protection material selection, protection layer thickness and maintenance strategy formulation.

Images