CN-120954574-B - Deep learning-based cement-based material infiltration corrosion degradation analysis method and system

CN120954574BCN 120954574 BCN120954574 BCN 120954574BCN-120954574-B

Abstract

The invention discloses a cement-based material infiltration corrosion degradation analysis method and a system based on deep learning, which relate to the technical field of data processing and comprise the steps of calling operation and maintenance information, and carrying out microstructure characterization acquisition to obtain K multi-scale characterizations of K operation and maintenance areas; the method comprises the steps of collecting K operation and maintenance areas to obtain M groups of operation and maintenance areas, dividing K multi-scale characterizations, carrying out penetration corrosion degradation characteristic networking matching to obtain M sample time sequence characterizations and M sample time sequence degradation characteristics, carrying out degradation conduction prediction of the K operation and maintenance areas to construct a degradation prediction graph neural network, inputting K time coding vectors of the K operation and maintenance areas into the degradation prediction graph neural network, carrying out penetration corrosion degradation prediction of cement-based materials in corrosion high-incidence areas, and outputting real-time operation and maintenance tasks. The method solves the technical problem of low prediction precision of the osmotic corrosion degradation of the cement-based material in the prior art, and achieves the technical effect of improving the prediction precision of the degradation of the cement-based material.

Inventors

ZHANG BOTAO
LIU DETAN
GE KAI
ZHANG TINGYU
Huang Mofei
SU JUN
HUANG ZONGGUI

Assignees

广西桂冠开投电力有限责任公司乐滩水电厂
大唐水电科学技术研究院有限公司

Dates

Publication Date: 20260512
Application Date: 20250620

Claims (6)

1. A method for analysis of penetration erosion degradation of cement-based materials based on deep learning, the method comprising: After operation and maintenance information of a dam corrosion high-rise area of a hydropower plant is called locally, microstructure representation acquisition is carried out according to the operation and maintenance information, K multi-scale representations of K operation and maintenance areas are obtained, wherein K sample metadata are marked by the K multi-scale representations; Aggregating the K operation and maintenance areas according to the K sample metadata to obtain M groups of operation and maintenance areas; Dividing the K multi-scale characterizations according to the M groups of operation and maintenance areas, and performing penetration corrosion degradation characteristic networking matching to obtain M sample time sequence characterizations and M sample time sequence degradation characteristics; Based on the M sample time sequence characterizations and the M sample time sequence degradation features, performing degradation conduction prediction of the K operation and maintenance areas to construct a degradation prediction graph neural network; Inputting K time coding vectors of the K operation and maintenance areas into the degradation prediction graph neural network, performing cement-based material infiltration corrosion degradation prediction of the corrosion high-incidence area, and outputting real-time operation and maintenance tasks; Wherein the degradation conduction prediction of the K operational areas is performed based on the M sample timing characterizations and the M sample timing degradation features to construct a degradation prediction graph neural network, the method comprising: Performing degradation trend fitting based on the M sample time sequence characterizations and the M sample time sequence degradation features, and outputting M material degradation curves; according to the M groups of operation and maintenance areas, mapping and distributing the M material degradation curves to the K operation and maintenance areas to generate K area degradation curves; Carrying out degradation conduction prediction of the K operation and maintenance areas according to the K area degradation curves by taking the hydraulic gradient direction as constraint, and constructing the degradation prediction graph neural network according to a prediction result; decomposing the K regional degradation curves to obtain K sample training data sets, and performing parameter adjustment optimization on K material degradation prediction models of K graph nodes in the degradation prediction graph neural network; the microstructure characterization acquisition is carried out according to the operation and maintenance information to obtain K multi-scale characterizations of K operation and maintenance areas, and the method comprises the following steps: Dividing the K operation and maintenance areas at equal intervals in the corrosion high-incidence area according to the operation and maintenance information; drilling core samples in the K operation and maintenance areas to obtain K infiltration corrosion core samples; Measuring K chloride ion penetration depths of the K penetration corrosion core samples by adopting a chromogenic method, and taking the K penetration depths as K macro-scale characterization; obtaining K communicated porosities as K millimeter scale characterizations by X-ray tomography of the K operation and maintenance areas; according to the ultrasonic longitudinal wave propagation characteristics of the K operation and maintenance areas, calculating and outputting K relative attenuation rates as K micron scale features; Performing operation and maintenance area binding storage on the K macro-scale characterizations, the K millimeter-scale characterizations and the K micrometer-scale characterizations to obtain K multi-scale characterizations; The K operation and maintenance areas are aggregated according to the K sample metadata to obtain M groups of operation and maintenance areas, and the method comprises the following steps: According to the K operation and maintenance areas, K sample metadata are called from the operation and maintenance information, wherein the sample metadata comprise repairing operation and maintenance time nodes, repairing material proportions, environment pH values and environment hydraulic gradients; Carrying out consistency judgment on the K sample metadata based on the patching material proportion, the environment pH value and the environment hydraulic gradient so as to aggregate the K sample metadata to obtain M groups of sample metadata; the K operation and maintenance areas are organized into M groups of operation and maintenance areas according to the M groups of sample metadata; After dividing the K multi-scale characterizations according to the M groups of operation and maintenance regions, performing penetration corrosion degradation characteristic networking matching to obtain M sample time sequence characterizations and M sample time sequence degradation characteristics, wherein the method comprises the following steps: Dividing the K multi-scale characterizations according to the M groups of operation and maintenance areas to obtain M groups of multi-scale characterizations; According to the mapping relation between the M groups of multi-scale characterizations and the M groups of sample metadata, invoking M groups of repairing operation and maintenance time nodes; performing degradation rate calculation on the M groups of multi-scale characterizations according to the M groups of repair operation and maintenance time nodes to obtain M degradation rate time-varying sequences; and performing penetration corrosion degradation characteristic networking matching by adopting the M degradation rate time-varying sequences to obtain M sample time sequence characterization and M sample time sequence degradation characteristics.
2. The deep learning-based cement-based material infiltration erosion degradation analysis method of claim 1, wherein the cement-based material infiltration erosion degradation prediction of the erosion high-rise region is performed by inputting K time-coded vectors of the K operation and maintenance regions into the degradation prediction graph neural network, outputting real-time operation and maintenance tasks, the method comprising: Updating K time coding vectors of the K operation and maintenance areas based on a preset operation and maintenance interval; mapping and inputting the K time coding vectors into K material degradation prediction models in the degradation prediction graph neural network to obtain K real-time degradation trend characteristics; And judging operation and maintenance requirements according to the K real-time degradation trend characteristics and the region boundaries of the K operation and maintenance regions, and outputting the real-time operation and maintenance tasks.
3. The deep learning based cement-based material infiltration erosion degradation analysis method of claim 1, wherein before locally invoking the operation and maintenance information of the hydropower plant dam erosion high-rise area, the method comprises: Extracting distributed time sequence seepage flow velocity through a hydropower plant dam embedded distributed optical fiber sensor network; taking the seepage flow speed threshold value and the continuous superthreshold time as double constraints, traversing the distributed time sequence seepage flow speed, and positioning a high seepage region spatial distribution map; Locally calling dam operation and maintenance records to analyze the space density of the defects and generating a defect density thermodynamic diagram; and after the spatial distribution map and the defect density thermodynamic diagram of the hypertonic flow region are spatially overlapped, performing multi-criterion decision fusion analysis, and positioning the corrosion high-incidence region.
4. The deep learning-based cement-based material infiltration erosion degradation analysis method of claim 1, wherein the M sets of multi-scale characterizations are subjected to degradation rate computation according to the M sets of repair operation and maintenance time nodes to obtain M degradation rate time-varying sequences, the method comprising: Arranging a first group of repairing operation time nodes in a time ascending order to obtain a first repairing operation time sequence; serializing a first group of multi-scale characterization to obtain a first multi-scale characterization sequence according to the first repair operation and maintenance time sequence; And according to the repair operation interval of the first repair operation time sequence, performing degradation rate calculation on the first multi-scale characterization sequence map to obtain a first degradation rate time-varying sequence, wherein each degradation rate time-varying characteristic in the degradation rate time-varying sequence consists of a penetration depth growth rate, a communication pore growth rate and an ultrasonic attenuation growth rate.
5. The deep learning-based cement-based material infiltration erosion degradation analysis method according to claim 1, wherein degradation conduction prediction of the K operation and maintenance areas is performed according to the K area degradation curves with a hydraulic gradient direction as a constraint, and the degradation prediction graph neural network is constructed according to a prediction result, the method comprising: Extracting K degradation extension distance extremum from the K regional degradation curves; Taking the regional centers of the K operation and maintenance regions as degradation starting points, taking the hydraulic gradient direction as degradation directions, and carrying out degradation linkage fitting on the K operation and maintenance regions in the corrosion high-incidence region based on the K degradation extension distance extremum to obtain K degradation influence region information; and constructing the K graph nodes according to the K operation and maintenance areas, and topologically connecting the K graph nodes according to the K degradation influence areas to finish the initialization of the degradation prediction graph neural network.
6. A deep learning-based cement-based material infiltration erosion degradation analysis system, the system comprising: The microstructure characterization acquisition module is used for carrying out microstructure characterization acquisition according to the operation and maintenance information after the operation and maintenance information of the high-incidence area of the dam corrosion of the hydropower station is locally called, so as to obtain K multi-scale characterizations of K operation and maintenance areas, wherein K multi-scale characterizations identify K sample metadata; The operation and maintenance area determining module is used for aggregating the K operation and maintenance areas according to the K sample metadata to obtain M groups of operation and maintenance areas; The matching module is used for carrying out the networking matching of the infiltration corrosion degradation characteristics after dividing the K multi-scale characterizations according to the M groups of operation and maintenance areas to obtain M sample time sequence characterizations and M sample time sequence degradation characteristics; the network construction module is used for carrying out degradation conduction prediction on the K operation and maintenance areas based on the M sample time sequence characterizations and the M sample time sequence degradation characteristics so as to construct a degradation prediction graph neural network; the degradation prediction module is used for performing cement-based material infiltration corrosion degradation prediction of the corrosion high-incidence area by inputting K time coding vectors of the K operation and maintenance areas into the degradation prediction graph neural network and outputting real-time operation and maintenance tasks; Wherein the network construction module is further configured to perform: Performing degradation trend fitting based on the M sample time sequence characterizations and the M sample time sequence degradation features, and outputting M material degradation curves; according to the M groups of operation and maintenance areas, mapping and distributing the M material degradation curves to the K operation and maintenance areas to generate K area degradation curves; Carrying out degradation conduction prediction of the K operation and maintenance areas according to the K area degradation curves by taking the hydraulic gradient direction as constraint, and constructing the degradation prediction graph neural network according to a prediction result; decomposing the K regional degradation curves to obtain K sample training data sets, and performing parameter adjustment optimization on K material degradation prediction models of K graph nodes in the degradation prediction graph neural network; wherein the microstructure characterization acquisition module is further configured to perform: Dividing the K operation and maintenance areas at equal intervals in the corrosion high-incidence area according to the operation and maintenance information; drilling core samples in the K operation and maintenance areas to obtain K infiltration corrosion core samples; Measuring K chloride ion penetration depths of the K penetration corrosion core samples by adopting a chromogenic method, and taking the K penetration depths as K macro-scale characterization; obtaining K communicated porosities as K millimeter scale characterizations by X-ray tomography of the K operation and maintenance areas; according to the ultrasonic longitudinal wave propagation characteristics of the K operation and maintenance areas, calculating and outputting K relative attenuation rates as K micron scale features; Performing operation and maintenance area binding storage on the K macro-scale characterizations, the K millimeter-scale characterizations and the K micrometer-scale characterizations to obtain K multi-scale characterizations; wherein the operation and maintenance area determining module is further configured to perform: According to the K operation and maintenance areas, K sample metadata are called from the operation and maintenance information, wherein the sample metadata comprise repairing operation and maintenance time nodes, repairing material proportions, environment pH values and environment hydraulic gradients; Carrying out consistency judgment on the K sample metadata based on the patching material proportion, the environment pH value and the environment hydraulic gradient so as to aggregate the K sample metadata to obtain M groups of sample metadata; the K operation and maintenance areas are organized into M groups of operation and maintenance areas according to the M groups of sample metadata; wherein the matching module is further configured to perform: Dividing the K multi-scale characterizations according to the M groups of operation and maintenance areas to obtain M groups of multi-scale characterizations; According to the mapping relation between the M groups of multi-scale characterizations and the M groups of sample metadata, invoking M groups of repairing operation and maintenance time nodes; performing degradation rate calculation on the M groups of multi-scale characterizations according to the M groups of repair operation and maintenance time nodes to obtain M degradation rate time-varying sequences; and performing penetration corrosion degradation characteristic networking matching by adopting the M degradation rate time-varying sequences to obtain M sample time sequence characterization and M sample time sequence degradation characteristics.

Description

Deep learning-based cement-based material infiltration corrosion degradation analysis method and system Technical Field The invention relates to the technical field of data processing, in particular to a cement-based material infiltration corrosion degradation analysis method and system based on deep learning. Background At present, cement-based materials are widely applied to various infrastructure constructions, in particular to important projects such as hydroelectric power plant dams and the like. However, cement-based materials are susceptible to factors such as osmotic erosion during long-term use, resulting in deterioration of their performance. Such degradation may not only affect the durability of the material, but may also threaten engineering safety. The traditional cement-based material degradation prediction method mainly depends on manual monitoring and an empirical model, cannot realize accurate degradation trend prediction, and has low prediction accuracy for the cement-based material infiltration corrosion degradation. Disclosure of Invention The application provides a deep learning-based cement-based material infiltration corrosion degradation analysis method and a deep learning-based cement-based material infiltration corrosion degradation analysis system, which are used for solving the technical problem that the prediction precision of cement-based material infiltration corrosion degradation is low in the prior art. In view of the above, the present application provides a deep learning-based cement-based material infiltration erosion degradation analysis method and system. In a first aspect of the present application, there is provided a deep learning-based cement-based material infiltration erosion degradation analysis method comprising: The method comprises the steps of carrying out microstructure representation acquisition according to operation and maintenance information after the operation and maintenance information of a dam corrosion high-rise area of a hydropower plant is called locally, obtaining K multi-scale representations of K operation and maintenance areas, wherein K sample metadata are marked on the K multi-scale representations, M groups of operation and maintenance areas are obtained by aggregating the K operation and maintenance areas according to the K sample metadata, carrying out infiltration corrosion degradation characteristic networking matching after dividing the K multi-scale representations according to the M groups of operation and maintenance areas, obtaining M sample time sequence representations and M sample time sequence degradation characteristics, carrying out conduction degradation prediction of the K operation and maintenance areas according to the M sample time sequence representations and the M sample time sequence degradation characteristics, so as to construct a degradation prediction graph neural network, and carrying out cement-based material infiltration corrosion degradation prediction of the corrosion high-rise area by inputting K time coding vectors of the K operation and maintenance areas into the degradation prediction graph neural network, and outputting real-time operation and maintenance tasks. In a second aspect of the present application, there is provided a deep learning based cementitious material infiltration erosion degradation analysis system, the system comprising: The system comprises a microstructure representation acquisition module, an operation and maintenance area determination module, a matching module, a network construction module and a degradation prediction module, wherein the microstructure representation acquisition module is used for carrying out microstructure representation acquisition according to operation and maintenance information of a dam corrosion high-incidence area of a hydropower station after the operation and maintenance information is locally called, K multiscale representations of K operation and maintenance areas are obtained, K sample metadata are marked on the K multiscale representations, the operation and maintenance area determination module is used for aggregating the K operation and maintenance areas according to the K sample metadata to obtain M groups of operation and maintenance areas, the matching module is used for carrying out infiltration corrosion degradation characteristic networking matching after the K multiscale representations are classified according to the M groups of operation and maintenance areas to obtain M sample time sequence representations and M sample time sequence degradation characteristics, the network construction module is used for carrying out degradation conduction prediction of the K operation and maintenance areas on the basis of the M sample time sequence representations and the M sample time sequence degradation characteristics, and the degradation prediction module is used for carrying out degradation prediction of the K operation and maintenance image ne