Sulfuric acid erosion resistant self-repairing geopolymer system and self-repairing method thereof

Abstract

The invention discloses a sulfuric acid erosion resistant self-repairing geopolymer system and a self-repairing method thereof, and relates to the technical field of composite materials and artificial intelligence. At least one self-repairing unit optimized for sulfuric acid erosion embedded within the geopolymer matrix; the in-situ multipoint pH sensing network is integrated in the geopolymer matrix and is used for monitoring spatio-temporal evolution data of a pH field in a sulfuric acid erosion process in real time; the edge calculation module is used for identifying a preliminary erosion indication according to the received spatio-temporal evolution data of the pH field; and the central AI decision module is used for evaluating an erosion state and predicting a development trend by utilizing an AI model optimized aiming at sulfuric acid erosion according to the identified preliminary erosion indication and the pH field spatio-temporal evolution data, and generating an optimal repair strategy to control the self-repair unit to perform self-repair. According to the invention, intelligent, active and refined protection and life management of the geopolymer in the sulfuric acid erosion process can be realized.

Description

Technical Field

[0001] This invention relates to the fields of composite materials and artificial intelligence technology, and in particular to a self-healing geopolymer system resistant to sulfuric acid corrosion and its self-healing method. Background Technology

[0002] Geopolymers, as inorganic non-metallic cementitious materials formed through complex condensation reactions of silica-alumina precursors (such as industrial solid wastes or natural minerals like fly ash, slag, and metakaolin) under the action of strongly alkaline activators (such as sodium hydroxide and sodium silicate), have been widely considered a key candidate with great development potential in the field of sustainable civil engineering materials due to their significant advantages over traditional silicate cement, such as low carbon emissions and low energy consumption. Furthermore, their application prospects have gradually expanded from initially being special cementitious materials to multiple fields such as precast components, road materials, refractory materials, nuclear waste solidification, and marine engineering. However, in many application scenarios, such as structures for industrial wastewater treatment facilities, production workshop floors and storage facilities in chemical enterprises, and certain types of contaminated soil and groundwater environments, geopolymer structures inevitably suffer from sulfuric acid (…). Long-term and complex corrosion by sulfate media, or sulfuric acid corrosion, is recognized as one of the most critical and destructive factors affecting the long-term durability of geopolymers. Its destructive mechanism is particularly complex and has significant harmful effects.

[0003] Currently, research on improving the sulfuric acid erosion resistance of geopolymers mainly focuses on material composition optimization and macroscopic modification, surface protection and physical isolation technologies, and preliminary exploration and passive response of self-healing technologies. However, existing research has certain limitations. There are still some problems in the intelligent, proactive, and precise monitoring and control of the sulfuric acid erosion process of geopolymers. (1) There is a lack of real-time, in-situ, high-precision, and long-term stable monitoring methods for core chemical indicators of sulfuric acid erosion (especially the dynamic evolution of the internal pH field). Traditional erosion assessment methods, such as periodic sampling for chemical analysis, mechanical property testing, pore structure characterization, or micromorphology and phase analysis, are all offline, lagging, or even destructive detection methods. They cannot capture in real time the dynamic and complete process information of the formation of internal pH gradient, the advancement of erosion front, the occurrence area and intensity of key chemical reactions, and the evolution of microcrack network in geopolymers under sulfuric acid erosion. Existing pH sensor technology faces multiple severe challenges in achieving long-term (e.g., months to years) stable and accurate operation in the extreme and dynamically changing chemical environment of geopolymers, which are initially strongly alkaline (pH>11-13) and may subsequently transform into strongly acidic (pH<3-4). These challenges include corrosion of sensitive materials, failure of reference electrodes, signal drift, and aging of encapsulation materials.

[0004] (2) Limited ability to deeply understand, accurately assess, and reliably predict complex sulfuric acid erosion processes. Sulfuric acid-eroded geopolymers are an extremely complex multi-physics coupled process involving multiple chemical reactions (including acid-base neutralization, gel depolymerization, ion exchange, precipitation dissolution, redox, etc.), multiphase ion diffusion and transport, pore structure evolution, and the development of internal stress and physical damage caused by chemical reaction products. Existing empirical models, simplified phenomenological theoretical models, or predictive models based on extrapolation of short-term accelerated test results are often unable to accurately describe and quantitatively predict their long-term development trends under different sulfuric acid concentrations, temperatures, humidity, material compositions, and structural dimensions. In particular, it is difficult to accurately assess the complex nonlinear relationship between pH field evolution and the degree of material damage (such as the coupling of chemical and physical damage), the decline of key properties (such as permeability and mechanical strength), and the remaining service life of the structure.

[0005] (3) Existing self-healing technologies have low levels of intelligence and lack the ability to actively intervene and precisely control. Currently, the self-healing technologies explored in geopolymers are basically at the passive response level. The timing of repair, the location of repair, the type and dosage of repair agents, as well as the efficiency and effectiveness of the repair process are difficult to control and optimize precisely, and they cannot be effectively actively controlled according to the actual situation. For example, physical filling of cracks alone may not be able to effectively prevent subsequent chemical erosion, nor can it actively inhibit the continuous generation and destructive effects of sulfate expansion products. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention proposes a self-healing geopolymer system and its self-healing method resistant to sulfuric acid corrosion. This system utilizes an embedded multi-point pH sensing network within the geopolymer matrix under sulfuric acid corrosion conditions, combined with intelligent AI-assisted regulation. Specifically, it employs artificial intelligence (AI) based on pH spatiotemporal evolution data and integrating physicochemical mechanisms to perform in-depth analysis of the corrosion state and accurate prediction of development trends. This drives a targeted, multi-strategy, closed-loop feedback self-healing unit regulation mechanism tailored to the characteristics of sulfuric acid corrosion. This enables high-precision in-situ monitoring, intelligent damage assessment and prediction, and proactive self-healing of the internal pH dynamics of geopolymer materials under sulfuric acid corrosion conditions. Ultimately, it achieves intelligent, proactive, and refined protection and lifespan management of geopolymers during sulfuric acid corrosion.

[0007] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions: In a first aspect, the present invention provides a self-healing geopolymer system resistant to sulfuric acid erosion.

[0008] A self-healing geopolymer system resistant to sulfuric acid erosion, comprising: The geopolymer matrix exists in a highly alkaline environment when not subjected to sulfuric acid erosion. At least one self-healing unit optimized for sulfuric acid erosion is embedded in a geopolymer matrix. The self-healing unit contains a repair agent for neutralizing acidity, sealing pores, or inhibiting sulfate reactions. An in-situ multi-point pH sensing network integrated within a geopolymer matrix is ​​used to monitor the spatiotemporal evolution of the pH field during sulfuric acid erosion in real time. The edge computing module is used to identify preliminary erosion indicators based on the received spatiotemporal evolution data of the pH field; The central AI decision-making module is used to assess the erosion status and predict the development trend based on the identified preliminary erosion indications and pH field spatiotemporal evolution data, using an AI model optimized for sulfuric acid erosion, and to generate the optimal remediation strategy to control the self-repair unit to perform self-repair.

[0009] A further technical solution is that the in-situ multi-point pH sensing network includes several miniature pH sensors uniformly distributed at multiple points inside the geopolymer matrix, and each miniature pH sensor is used to collect the pH value of each monitoring point inside the geopolymer matrix in real time. The miniature pH sensor employs a solid-state potentiometric sensor based on an iridium oxide thin-film electrode or a pH-sensitive functionalized polymer, and features a dual-resistance encapsulation structure designed to withstand both the initial high alkalinity of the polymer and the subsequent strong sulfuric acid erosion environment.

[0010] A further technical solution is that the edge computing module is used to preprocess the collected pH field spatiotemporal evolution data, including: Based on the real-time collected pH values ​​and coordinated temperature data of each monitoring point inside the geopolymer matrix, the pH value is compensated and corrected by a nonlinear temperature compensation algorithm based on polynomial fitting or lookup table. For the pH value after temperature compensation and correction, a second correction is performed using a consistency verification algorithm for neighboring sensor data to obtain the corrected pH value, and the pH field feature vector, including pH change rate and pH spatial gradient, is extracted. Preliminary erosion indicators are identified based on the corrected pH value.

[0011] A further technical solution is that the central AI decision-making module is equipped with a dynamic tracking and quantification model of the sulfuric acid erosion front, a prediction model based on physical information, and an AI decision-making model; Using a dynamic tracking and quantification model of the sulfuric acid erosion front, we analyze the spatial distribution characteristics of the current three-dimensional pH gradient field and the average normal advance rate of the erosion front. Using a prediction model based on physical information, we can predict the evolution trend of the pH field, the development of erosion depth, and the risk level of expansion and cracking within a specific time window in the future. Using an AI decision-making model, based on the repair knowledge base and the assessment and prediction results of the current erosion state, a reinforcement learning algorithm or a genetic algorithm based on multi-objective optimization is used to generate the optimal repair strategy for the current erosion state.

[0012] Further technical solutions, targeting the dynamic tracking and quantification model of the sulfuric acid erosion front: inputting the spatiotemporal evolution data of the pH field collected by the in-situ multi-point pH sensing network at continuous time points into the model, and setting the spatial resolution to track the three-dimensional position and morphological evolution of a specific pH isosurface in the geopolymer matrix in real time, and calculating the average normal advance rate of the sulfuric acid erosion front and the spatial distribution characteristics of the pH gradient field. For the prediction model based on physical information: the average normal advance rate of the sulfuric acid erosion front and the spatial distribution characteristics of the pH gradient field are input into the model to predict the evolution trend of the pH field, the development of erosion depth, and the accumulation or distribution area of ​​key erosion products within a set time window in the future, and to clarify the risk level of expansion and cracking. Specifically, for the prediction model based on physical information, the rate equation describing the key chemical reaction between sulfuric acid and the main hydration products in the geopolymer matrix, as well as the diffusion transport equation describing the diffusion coefficient of key ions in the porous structure of the geopolymer matrix, are used as partial differential equation constraints and incorporated into the loss function of the neural network model, thereby continuously iterating and training the prediction model.

[0013] In a further technical solution, the AI ​​decision-making model generates the optimal repair strategy based on a preset knowledge base for repairing sulfuric acid corrosion and the current corrosion status assessment and prediction results, using a reinforcement learning algorithm or a genetic algorithm based on multi-objective optimization. The optimal repair strategy is generated using a reinforcement learning algorithm. This involves using a deep Q-network, setting the state space to include the current pH field feature vector, propulsion rate, and predicted expansion and cracking risk level, and setting the action space to include the selection of the type and combination of repair agents, the spatial range of the activation location, the activation intensity, and the delivery dosage of the repair agent. The optimal repair strategy is obtained through iterative optimization of the action-state relationship. The optimal repair strategy is generated by a genetic algorithm based on multi-objective optimization. The strategy is as follows: with the repair effect, repair cost and repair time as objectives, the type and combination of repair agents, the spatial range of activation sites, activation intensity and delivery dosage of repair agents are randomly selected as a group of individuals. Through fitness evaluation and crossover and mutation operations, the individuals are continuously updated iteratively until the multi-objective optimality is reached, and the optimal repair strategy is obtained.

[0014] A further technical solution is that the repair knowledge base includes: performance parameters of different pH-responsive repair agents, curing time and sealing efficiency of pore plugging agents under different pH and humidity conditions, and effective concentration range and potential side effects of sulfate reaction inhibitors.

[0015] A further technical solution, wherein the optimal repair strategy includes: the type and combination of repair agents, the range of three-dimensional spatial regions activated by the repair agents, the release dosage of the repair agents, and repair regulation parameters.

[0016] In a further technical solution, the self-healing unit includes one or more of pH-responsive microcapsules and microvascular network systems; Among them, for pH-responsive microcapsules, when the pH value drops to a preset threshold, the shell material undergoes structural damage, controllable swelling, or a sharp increase in permeability, releasing the internally encapsulated alkaline neutralizer or pore-blocking agent precursor at a controlled rate. For the microvascular network system, the microvessels are made of acid-resistant materials, and the inner diameter of the microvascular channels is three-dimensionally distributed according to the set vascular length density. The microvascular network is connected to one or more replaceable repair agent reservoirs, and the repair agent is delivered to specific damaged areas on demand, in a quantitative and timed manner through a micro piezoelectric pump and a micro solenoid valve array controlled by an AI decision module.

[0017] Secondly, the present invention provides a self-healing method for a sulfuric acid-resistant self-healing geopolymer system.

[0018] A self-healing method for a sulfuric acid-resistant self-healing geopolymer system as proposed in the first aspect, comprising: Real-time monitoring of the spatiotemporal evolution of pH field in geopolymer matrices during sulfuric acid erosion; Based on the received spatiotemporal evolution data of pH field, preliminary erosion indicators are identified. Then, using the onboard dynamic tracking and quantification model of sulfuric acid erosion front and the prediction model based on physical information, the erosion status is assessed and the development trend is predicted. Based on a pre-defined knowledge base for repairing sulfuric acid erosion and the current erosion status assessment and prediction results, an optimal repair strategy is generated using a reinforcement learning algorithm or a genetic algorithm based on multi-objective optimization to control the self-repairing unit to perform self-repair. During the self-repair process, the repair parameters in the optimal repair strategy are dynamically fine-tuned based on the real-time feedback of pH recovery rate and spatial range to achieve closed-loop control.

[0019] The above one or more technical solutions have the following beneficial effects: 1. This invention proposes a self-healing geopolymer system and its self-healing method resistant to sulfuric acid erosion. Through innovative deep integration and synergistic optimization of multiple technologies, it achieves technical effects far exceeding the simple superposition of existing technologies. It solves the problems of outdated monitoring methods, limited damage assessment and prediction capabilities, passive and inefficient repair mechanisms, and lack of intelligent active regulation in existing technologies for geopolymer resistance to sulfuric acid erosion. Specifically, this invention establishes for the first time a four-in-one intelligent closed-loop system: "pH field spatiotemporal evolution data - sulfuric acid erosion mechanism model - AI prediction decision - targeted self-healing." This system uses an embedded multi-point pH sensing network within the geopolymer matrix under sulfuric acid erosion conditions to perform high-precision in-situ monitoring of the internal pH field of the geopolymer during the sulfuric acid erosion process, thereby acquiring high spatiotemporal resolution data. Based on this, it combines intelligent AI-assisted regulation, namely, artificial intelligence (AI) based on pH spatiotemporal evolution data and integrating physicochemical mechanisms. By combining the specific chemical reaction kinetic equations of sulfuric acid erosion embedded in the Physical Information Neural Network (PINN), it regulates the erosion process... This invention performs in-depth analysis, precise modeling, and accurate prediction of the erosion state and development trend. Based on the prediction results, it proactively and precisely regulates the self-healing units, thereby driving a targeted, multi-strategy, closed-loop feedback self-healing unit regulation mechanism targeting the characteristics of sulfuric acid erosion. This enables high-precision in-situ monitoring, intelligent damage assessment and prediction, and proactive self-healing regulation of the internal pH dynamic evolution of geopolymer materials under sulfuric acid erosion. This effectively resists sulfuric acid erosion, extends the material's service life, and realistically reproduces and intelligently responds to the complex physicochemical evolution process inside geopolymers under sulfuric acid erosion. It achieves intelligent, proactive, and refined protection and lifespan management of geopolymers during sulfuric acid erosion. Compared to existing technologies, this invention can extend the service life of geopolymers in sulfuric acid environments, reduce the erosion rate, and improve the utilization efficiency of repair agents.

[0020] 2. Targeting the specificity, high precision, and in-situ dynamic intelligent sensing of sulfuric acid erosion, this invention, through a specially optimized in-situ multi-point pH sensing network and AI-assisted signal conditioning and health management, can capture the fine dynamic evolution of the internal pH field of geopolymers in complex sulfuric acid erosion environments in real time, accurately, and over a long period of time. This overcomes the lag, destructiveness, and instability of traditional monitoring methods in extreme chemical environments, providing an unprecedented high-quality data foundation for accurately assessing the state of sulfuric acid erosion.

[0021] 3. In this invention, by innovatively introducing advanced AI models such as Physical Information Neural Network (PINN), the complex chemical reaction kinetics and ion transport mechanism of sulfuric acid-eroded geopolymers are deeply integrated with data-driven learning. This significantly improves the accuracy and reliability of predicting the risk of erosion depth, rate, long-term evolution trend of pH field, and formation of key erosion products (such as expansive sulfate). This provides a more scientific and forward-looking basis for the durability assessment of structures and preventive maintenance decisions.

[0022] 4. This invention, through the design of an AI decision engine, can intelligently generate and execute optimal, multi-stage or synergistic repair strategies based on real-time monitoring and deep prediction results, combined with a repair knowledge base targeting the characteristics of sulfuric acid corrosion, such as pH reduction, ion dissolution, and swelling products, and advanced optimization algorithms, such as reinforcement learning. This enables intelligent, proactive, precise, and multi-strategy repair regulation by controlling the timing, location, type of repair agent, release dosage, and other factors. Furthermore, this approach can achieve precise parameter control, significantly improving repair efficiency and effectiveness, delaying the sulfuric acid corrosion process to the maximum extent, and potentially actively inhibiting harmful swelling reactions.

[0023] 5. In this invention, continuous monitoring and quantitative evaluation of the repair effect are carried out, and a complete empirical data chain is used for the iterative optimization of the AI ​​model. Through this closed-loop feedback and system adaptive evolution, the system performance is improved, enabling the system to have a strong adaptive learning and continuous improvement capability. It can continuously improve its response accuracy, prediction ability and repair efficiency in the complex and ever-changing sulfuric acid erosion environment, and achieve true "intelligent evolution".

[0024] 6. Through the above-mentioned intelligent, proactive, and precise monitoring, prediction, and repair closed-loop control, the present invention can effectively inhibit the destructive effects of sulfuric acid corrosion, protect the polymer matrix structure, thereby significantly extending the service life of the material in the sulfuric acid corrosion environment, improving the safety and reliability of the structure, and significantly reducing the inspection, maintenance, and repair costs throughout the entire life cycle.

[0025] 7. The integrated intelligent system and working method proposed in this invention not only provide an advanced experimental platform and powerful data analysis tools for in-depth research on the complex degradation mechanism of geopolymers under sulfuric acid erosion, but also open up a new technical path for developing a new generation of ultra-durable geopolymer materials with active protection and intelligent repair capabilities and their engineering applications in harsh sulfuric acid erosion environments (such as chemical, mining and metallurgical, sewage treatment, marine acidification and other fields). Attached Figure Description

[0026] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0027] Figure 1 This is a schematic diagram of the structure of the sulfuric acid-resistant self-healing geopolymer system in an embodiment of the present invention; Figure 2 This is a schematic diagram of the three-dimensional layout of the in-situ multi-point pH sensing network in the geopolymer specimen in an embodiment of the present invention; Figure 3 This is a flowchart of the self-healing method of the sulfuric acid erosion-resistant self-healing geopolymer system in an embodiment of the present invention.

[0028] Among them, 101 is the geopolymer matrix; 102 is the self-healing unit; 103 is the in-situ multi-point pH sensing network; 301 is the pH sensor; 104 is the edge computing module; 105 is the central AI decision-making module; and 106 is other types of sensor arrays. Detailed Implementation

[0029] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0030] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0031] To address the shortcomings of existing technologies in the resistance of geopolymers to sulfuric acid erosion, such as outdated monitoring methods, limited damage assessment and prediction capabilities, passive and inefficient repair mechanisms, and a lack of intelligent active regulation, this invention proposes a self-healing geopolymer system and its self-healing method. This intelligent geopolymer system highly integrates in-situ pH sensing technology optimized for sulfuric acid erosion environments, an AI-powered intelligent analysis and prediction engine based on pH spatiotemporal evolution, and an AI-driven targeted, multi-strategy self-healing regulation mechanism. The system can intelligently analyze key erosion state information, such as real-time monitoring of internal pH field evolution, and the AI's accurate prediction of deterioration trends. The root causes of damage and dynamic optimization of repair strategies, such as intervention in the early stages of corrosion, selection of the most suitable combination of repair agents according to the corrosion stage, precise control of the release location and dosage of repair agents, and proactive adjustment of the repair environment to improve repair efficiency, are all addressed. Furthermore, a closed-loop proactive control of the repair process is implemented to form an integrated intelligent solution that combines "sensing, diagnosis, prediction, decision-making, and repair" for the characteristics of sulfuric acid corrosion. This solution can fundamentally improve the durability, reliability, and service life of geopolymer materials in sulfuric acid corrosion environments, reduce their total life-cycle maintenance costs, and effectively expand their engineering applications in key fields such as chemical engineering, mining and metallurgy, wastewater treatment, and marine acidification.

[0032] Example 1 This embodiment proposes a self-healing geopolymer system resistant to sulfuric acid erosion, such as... Figure 1 As shown, it includes: Geopolymer matrix 101 exists in a highly alkaline environment when not subjected to sulfuric acid erosion; At least one self-healing unit 102 optimized for sulfuric acid erosion is embedded in a geopolymer matrix. The self-healing unit contains a repair agent for neutralizing acidity, sealing pores, or inhibiting sulfate reactions. An in-situ multi-point pH sensing network 103 integrated within the geopolymer matrix is ​​used to monitor the spatiotemporal evolution of the pH field during sulfuric acid erosion in real time. The edge computing module 104 is communicatively connected to the in-situ multi-point pH sensing network 103 and is used to identify preliminary erosion indicators based on the received pH field spatiotemporal evolution data. The central AI decision-making module 105 is communicatively connected to the edge computing module 104 and the self-repairing unit 102. It is used to assess the erosion status and predict the development trend based on the identified preliminary erosion indications and pH field spatiotemporal evolution data, using an AI model optimized for sulfuric acid erosion, and to generate the optimal repair strategy to control the self-repairing unit to perform self-repair.

[0033] Specifically, the geopolymer matrix 101 serves as the functional carrier of the entire intelligent system; it is itself a composite material structure specially designed to integrate various functional units. In a preferred embodiment, the geopolymer matrix 101 is prepared as follows: (1) Low-calcium fly ash (e.g., CaO content less than 5%) and granulated blast furnace slag mixed in a specific ratio (e.g., mass ratio range of 60:40 to 80:20) are used as the main silica-alumina precursors.

[0034] (2) The alkaline activator is a sodium hydroxide (NaOH) solution of a specific concentration (e.g., 6M to 12M) and a solution of a specific modulus (e.g., ...). A water glass (i.e., sodium silicate) solution with a molar ratio of 1.0 to 2.0 is mixed in a certain proportion (e.g., the mass ratio of NaOH solution to water glass solution is 1:1 to 1:3). The liquid-solid ratio (the ratio of the total mass of the alkaline activator solution to the total mass of the aluminosilicate precursor) is controlled within a set range, such as 0.35 to 0.50.

[0035] (3) After mixing the silicon-aluminum precursor with the alkaline activator solution to form a uniform slurry, before or during casting, the sensor components of the pre-prepared or selected self-healing unit 102 and the in-situ multi-point pH sensing network 103 are precisely embedded or integrated into the geopolymer slurry according to the preset spatial distribution scheme.

[0036] (4) Subsequently, the geopolymer specimens or components are cured and molded under specific curing conditions (e.g., cured at 50°C to 80°C for 12 to 48 hours, and then transferred to standard curing conditions, i.e., 20±2°C, relative humidity RH≥95%, for a predetermined age, such as 28 days), forming an intelligent geopolymer matrix integrating sensing, computing, decision-making, and repair potential. The initial pore fluid pH value of the geopolymer matrix 101 when not subjected to sulfuric acid erosion is typically higher than 11.5, and can even reach above 13.0.

[0037] like Figure 2 The diagram shows a three-dimensional layout of the in-situ multi-point pH sensing network 103 in the geopolymer specimen. The core function of the in-situ multi-point pH sensing network 103 is to monitor the spatiotemporal dynamic evolution of pH value inside the geopolymer matrix 101 during sulfuric acid erosion in real time, in situ, and in a distributed manner. The in-situ multi-point pH sensing network includes several miniature pH sensors 301 uniformly distributed at multiple points inside the geopolymer matrix. Each miniature pH sensor is used to collect the pH value at each monitoring point inside the geopolymer matrix in real time.

[0038] In one implementation, the network consists of an array of multiple (e.g., from eight to hundreds, depending on the size of the monitoring area and accuracy requirements) miniature pH sensors 301. The design of this sensor network includes: (1) Selection and characteristics of pH sensor 301: The miniature pH sensor adopts an iridium oxide (IrO)-based design. x Solid-state potentiometric sensors using thin-film electrodes or specific pH-sensitive functionalized polymers (such as conductive polymers doped with proton acceptors, such as polyaniline derivatives) have an effective operating pH range covering, for example, pH 2.0 to pH 13.5, and exhibit near-Nernst linearity in the pH 4.0 to pH 12.0 range. For example, at 25°C, the absolute value of the potential response slope is between 50 mV / pH unit and 60 mV / pH unit, and the response time (i.e., the time to reach 90% stable reading) is less than, for example, 60 seconds.

[0039] In this embodiment, it is preferable to use iridium oxide-based ( A solid-state potentiometric pH sensor with a thin-film electrode, which controls... Thin film fabrication processes (such as oxygen partial pressure and annealing conditions in reactive sputtering) can achieve a wide pH response range (e.g., pH 1 to pH 14) and good linearity (e.g., at 25°C, the absolute value of its potential response slope is close to the theoretical Nernst value, approximately 59 mV / pH unit), and the sensor's response time... Designed to last less than 30 seconds.

[0040] (2) Encapsulation Structure: Each pH sensor 301 has a dual-resistance encapsulation structure for both the initial high-alkalinity environment of the polymer and the subsequent strong acid corrosion environment of sulfuric acid. Its inner layer is a dense alumina (300 μm to 25 μm thick) prepared by chemical vapor deposition (CVD) or atomic layer deposition (ALD) technology. ) or silicon nitride ( The passivation layer, with the outer layer being a chemically resistant polymer coating such as polytetrafluoroethylene (PTFE), polyvinylidene fluoride (PVDF), or Parylene-C with a thickness of 10µm to 100µm, ensures that the sensor has a continuous operating life of more than, for example, 1 to 3 years in the complex chemical environment.

[0041] In this embodiment, the sensor's sensitive core is first deposited using atomic layer deposition (ALD) technology, forming an ultrathin, dense, and chemically inert layer of aluminum oxide with a thickness of 10 nm to 50 nm. ) or hafnium dioxide ( The inner passivation and isolation layer is then used as the outer layer. Then, a chemically resistant and flexible polymer material, such as polytetrafluoroethylene (PTFE) or Parylene-C, with a thickness of 20µm to 100µm is used for encapsulation through precision coating or vapor deposition processes, leaving only sensitive areas in contact with the environment. Finally, the lead wires are also made of corrosion-resistant materials (such as precious metal wires insulated with PTFE) and are sealed.

[0042] (3) Spatial Layout: The miniature pH sensor 301 is embedded in the polymer matrix 101 in a pre-defined three-dimensional grid, such as... Figure 2 As shown, in a 100mm × 100mm × 100mm cubic specimen, a uniformly arranged 5×5×5 sensor array with a 20mm spacing can be formed, or the sensors can be embedded in the geopolymer matrix 101 with a gradient density (higher sensor density near the surface and lower density in the interior) according to the expected sulfuric acid erosion direction (e.g., from the exposed surface to the interior). Before casting, the sensors can be fixed in the mold using precision positioning clamps. For key monitoring areas (e.g., the exposed surface depth range of 0-20mm), the sensor spacing is 10-15mm; for the interior area (depth above 20mm), the sensor spacing can be increased to 20-30mm. A typical 100mm cubic specimen is recommended to use a 3×3×3 array arrangement of 27 sensors.

[0043] (4) Collaborative temperature monitoring: Integrate a miniature temperature sensor (such as an NTC thermistor or a PT1000 thin-film platinum resistance thermometer with a temperature measurement accuracy of ±0.2°C) near the pH sensor array or on some of the pH sensors to perform real-time temperature compensation for pH readings.

[0044] Furthermore, the edge computing module 104 is connected to the pH sensor network and the central AI decision-making module. As the front end of the distributed intelligence, it is responsible for collecting, preprocessing, performing preliminary analysis, and managing the health of the raw data from the pH sensor network. Specifically, the edge computing module is equipped with a low-power microcontroller (MCU), such as an ARM Cortex-M4 or Cortex-M7 series processor with a floating-point unit (FPU), or a small field-programmable gate array (FPGA). Its real-time data processing capability can support at least, for example, 8 to 32 pH sensor channels to perform data acquisition and preliminary analysis at a sampling frequency of 0.1 to 10 times per second. It also has built-in AI-assisted sensor signal conditioning algorithms, including nonlinear temperature compensation algorithms based on polynomial fitting or lookup table (LUT) (such as using co-integrated miniature temperature sensors, such as NTC thermistors or digital temperature sensors, with a temperature measurement accuracy of, for example, ±0.1°C to 0.5°C), and sensor drift online monitoring and preliminary correction algorithms or fault alarm mechanisms based on neighboring sensor data consistency verification (such as comparing the difference between adjacent sensor readings with a preset threshold) and historical data trend analysis (such as detecting sudden changes in readings or long-term drift exceeding the allowable range, such as ±0.2 pH units / month).

[0045] In this embodiment, the edge computing module employs a microcontroller (MCU) 401 based on the ARM Cortex-M7 core. This MCU features high processing speed (e.g., operating frequency of 200MHz to 600MHz), ample RAM (e.g., 512KB to 2MB), and Flash storage (e.g., 1MB to 4MB), and integrates a high-precision multi-channel ADC (e.g., 16-bit or 24-bit resolution). Specifically, this microcontroller includes the following functions: (1) Data acquisition and synchronization: The edge computing module 104 acquires the potential signals of all pH sensors 301 and the temperature signals of temperature sensors synchronously via ADC at a configurable sampling frequency (e.g., from once per minute to once per hour depending on the rate of erosion).

[0046] (2) Signal Conditioning and AI-Assisted Calibration: The module incorporates digital filtering algorithms (such as median filtering and Kalman filtering) to reduce noise in the original signal. The module runs a lightweight AI model (e.g., a small feedforward neural network or a Gaussian process regression-based model) that can be trained using data obtained through multi-point calibration (performed at different known pH buffers and temperatures). Through a nonlinear temperature compensation algorithm based on polynomial fitting or lookup tables, it performs nonlinear, personalized temperature compensation and pH conversion on the potential signal of each pH sensor based on real-time temperature readings.

[0047] In addition, this module can perform secondary correction through a consistency verification algorithm for neighboring sensor data. That is, it uses redundant information in the sensor array (such as comparing whether the readings of multiple spatially adjacent sensors are within a reasonable consistency range, such as a difference of less than 0.3 pH units) and historical data trends of individual sensors (such as detecting whether there are sudden, physically irregular jumps or continuous, abnormal drift rates) to conduct online assessment of the health status of the sensors, and marks or attempts to perform preliminary deviation correction on sensors that may fail or drift significantly.

[0048] (3) Preliminary erosion indication identification: The module has built-in rule-based logical judgment or simple pattern recognition algorithm (such as pH change rate threshold judgment based on sliding window, for example, when the average pH change rate of a certain area continues to exceed 0.05 pH units / hour for more than 6 hours, it is determined to be a preliminary erosion indication) to quickly identify areas inside the geopolymer that may be undergoing significant sulfuric acid erosion.

[0049] (4) Data communication: The edge computing module 104 uploads the processed pH data (e.g., pH value, temperature value, and health status indicator after each sensor is calibrated) and preliminary erosion indication information to the central AI decision module 105 at preset time intervals (e.g., once per hour to once per day, or immediately when there is an erosion indication) through a low power wide area network (LPWAN) communication module (such as LoRaWAN or NB-IoT module) or a short-range wireless communication module (such as BLE5.0).

[0050] Furthermore, the central AI (Artificial Intelligence) decision-making module 105 is the intelligent core of the entire system, responsible for in-depth analysis and accurate prediction of sulfuric acid erosion status, and generating and executing the optimal repair strategy. This module can be deployed on a local embedded AI platform (such as the NVIDIA Jetson series), edge server, or cloud platform with strong computing power. Its core is a multi-model, multi-layered AI engine. The AI ​​models optimized for sulfuric acid erosion that this AI decision-making module runs include: a dynamic tracking and quantification model of the sulfuric acid erosion front, a prediction model based on physical information, and an AI decision-making model. Among them, the dynamic tracking and quantification model of the sulfuric acid erosion front analyzes the spatial distribution characteristics of the current three-dimensional pH gradient field and the average normal advance rate of the erosion front; the prediction model based on physical information predicts the pH field evolution trend, erosion depth development, and expansion and cracking risk level within a specific time window in the future; and the AI ​​decision-making model, based on the repair knowledge base and the current erosion status assessment and prediction results, uses reinforcement learning algorithms or genetic algorithms based on multi-objective optimization to generate the optimal repair strategy for the current erosion status.

[0051] Specifically, for the dynamic tracking and quantification model of sulfuric acid erosion front: a model based on a three-dimensional convolutional neural network (3DCNN) or a convolutional long short-term memory network (ConvLSTM) is used. This model takes pH field data collected by a pH sensor array at continuous time points (e.g., time intervals of 1 hour to 24 hours) as input (e.g., forming a four-dimensional tensor of N×M×K×T, where N, M, and K are the number of sensors in the spatial dimensions, and T is the length of the time series). It can track the three-dimensional position and morphological evolution of specific pH isosurfaces (e.g., key erosion indicator thresholds of pH=10.5, pH=9.0, and pH=7.5) in geopolymer in real time with a spatial resolution of 0.5 mm to 2.5 mm, and calculate the average normal advance rate of the erosion front (e.g., ranging from 0.001 mm / day to 1.0 mm / day, depending on the sulfuric acid concentration and geopolymer type) and the spatial distribution characteristics of the pH gradient field (e.g., maximum gradient value, gradient direction, etc.).

[0052] In this embodiment, the dynamic tracking and quantification model of sulfuric acid erosion front receives pH spatiotemporal data sequences from the edge computing module 104. As one implementation, it employs a deep neural network architecture (such as a combination of 3DU-Net and ConvLSTM) integrating three-dimensional convolutional layers (for extracting spatial features of the pH field) and recurrent layers (such as LSTM or GRU, for capturing the dynamic features of pH field evolution over time). The model is trained with a large amount of simulation data (such as pH field evolution data under different conditions obtained through finite element simulation of sulfuric acid erosion of geopolymers using software such as COMSOL) and accelerated erosion experimental data from the laboratory (including real readings of the pH sensor array and corresponding offline characterization results of erosion depth and mineral phase changes). 1) Reconstruct a three-dimensional pH field distribution cloud map inside the geopolymer in real time with a spatial resolution of, for example, 1 mm; (2) Dynamically track the movement trajectory and morphological changes of the preset pH isosurface inside the geopolymer, where the pH isosurface refers to, for example, pH=11.0 as the boundary of the initial alkaline environment, pH=9.0 as the neutralization front, and pH=7.0 as the beginning of the acidic environment; (3) Calculate the average normal advance rate of the erosion front (e.g., by comparing the positional changes of the isosurface at consecutive time points) and the volumetric erosion rate; (4) Identify areas with abnormally strong pH gradients, which can indicate local erosion acceleration or crack propagation.

[0053] For Physical Information Neural Networks (PINN) or Mechanism-Enhanced AI Models (i.e., prediction models based on physical information): This model will describe sulfuric acid ( ) and geopolymer main hydration products (e.g., NASH gel, C-(A)-SH gel, and possibly free ions) Key chemical reactions (e.g., neutralization reactions) between alkali metal hydroxides or other metal hydroxides + → Acidolysis of the silicon-aluminum framework; + +2 → (Gypsum); and the related reactions for the formation of ettringite) rate equations (such as reaction rate constants based on the law of mass action and considering pH dependence, with an order of magnitude, for example, to mol· · )as well as , , , The effective diffusion coefficient of key ions in the geopolymer porous structure (e.g., range) to The Fickian or non-Fickian diffusion transport equations (considering the dynamic changes in porosity and tortuosity with erosion) are used as partial differential equation (PDE) constraints and are explicitly incorporated into the loss function of a deep neural network (e.g., containing multiple hidden layers, each with 32 to 256 neurons, using activation functions such as ReLU or tanh) for training. The model is used to predict the detailed evolution of the pH field, the development of erosion depth (defined as the boundary of the region where pH is below a certain failure threshold, such as pH=X), and the accumulation or distribution area of ​​key erosion products (such as gypsum) within a specific future time window (e.g., the next 7 days to 1 year). The prediction accuracy is, for example, a relative error of less than 10%-20% compared to experimental measurements. Specifically, 100mm × 100mm × 100mm geopolymer specimens equipped with 25 pH sensors are tested for 180 days under cyclic erosion conditions with 3% sulfuric acid solution. Complete pH field data are recorded every 24 hours and compared with offline pH test results obtained through destructive sampling. The results showed that the PINN model had a small average relative error between the predicted and measured values ​​of the pH field after 7 days and a small prediction error after 30 days, both within the range of 10%-20%.

[0054] In this embodiment, to improve the accuracy of prediction and extrapolation ability, a PINN model is adopted. The neural network part of this model (e.g., a multilayer perceptron containing 5-10 hidden layers, each with 64-512 neurons, using swish or tanh activation functions) is time (t) and spatial coordinates (t). As input, it outputs the pH value at the corresponding location and time, as well as key ions (such as...). , , , The concentration of ) and its loss function not only include the fitting error term with the sensor measurement data (pH value), but also require the network output to satisfy the residuals of a set of core partial differential equations (PDEs) describing the sulfuric acid erosion geopolymerization process as small as possible. These PDEs include: (1) and The unsteady-state diffusion equation for plasma, which considers concentration gradient-driven diffusion and electromigration, with the diffusion coefficient being a function of porosity and pH; (2) The main alkaline components of geopolymers (such as soluble alkali in NASH gel, )and The kinetic equation for the neutralization reaction; (3) In geopolymers The precipitation reaction kinetics equation for gypsum formation (from slag hydration products or gel decomposition); (4) If the geopolymer has a high aluminum content and a calcium source, then further consideration is needed. , With activity The reaction kinetic equation for the formation of ettringite.

[0055] By solving the highly nonlinear optimization problem described above, the trained PINN model can predict the evolution of the pH field in the future (e.g., from one month to several years), the development of erosion depth (defined as the boundary of a region where the pH is below a certain critical value, such as pH=X), and the generation and spatial distribution of key erosion products such as gypsum and ettringite. It may also be able to assess the level of internal stress or the risk of cracking caused by expansion products.

[0056] For the central AI decision-making module, based on a pre-set knowledge base for sulfuric acid erosion repair and the current erosion status assessment and prediction results, it generates the optimal repair strategy using either a reinforcement learning (RL) algorithm or a genetic algorithm based on multi-objective optimization. The RL algorithm employs a deep Q-network (DQN), setting the state space to include the current pH field feature vector, propagation rate, and predicted expansion cracking risk level, while the action space includes selecting the type and combination of repair agents, determining the spatial range of activation locations, activation intensity, and the dosage of repair agents. The optimal repair strategy is obtained through iterative optimization of the action-state relationship. The genetic algorithm based on multi-objective optimization uses the following criteria: repair effect, repair cost, and repair time as objectives. It randomly selects the type and combination of repair agents, the spatial range of activation locations, activation intensity, and the dosage of repair agents as a group of individuals. Through fitness evaluation and crossover / mutation operations, the individuals are iteratively updated until the multi-objective optimum is reached, yielding the optimal repair strategy.

[0057] In addition, the remediation knowledge base includes: (a) performance parameters of different pH-responsive remediation agents, such as encapsulated nanoparticles. (a) Average release rate curves of microcapsules (particle size, e.g., 50 nm-200 nm) at pH below 9.0 (e.g., 60%-80% release of active ingredient within 24 hours at pH=8); (b) Curing time (e.g., 1 hour to 48 hours) and sealing efficiency (e.g., 30%-70% reduction in porosity) of pore-blocking agents (e.g., water glass-based or silane-based remediation agents) under different pH (e.g., pH 4-9) and humidity (e.g., RH 60%-95%) conditions; (c) Sulfate reaction inhibitors (e.g., ... The effective concentration range (e.g., 0.01M to 0.5M) and potential side effects (such as the introduction of other harmful ions) of lithium salt solutions.

[0058] In this embodiment, the repair strategy optimization and decision-making model (i.e., the AI ​​decision-making model) is built on a reinforcement learning (RL) framework (such as using an Actor-Critic algorithm). Its "State" is jointly defined by the erosion state assessed by the current AI (such as pH field characteristics, erosion rate, predicted expansion risk level, and remaining lifetime prediction) and the available state of the repair unit (such as the remaining amount of repair agent and the patency of the microvascular network). The "Action" space includes whether to initiate repair, which repair agent to select (if the system integrates multiple agents), determining the target three-dimensional region for repair agent activation / delivery, and specific activation parameters (such as heating temperature curve, flow rate and total amount delivered to microvessels, etc.). The "Reward" function is designed to maximize the expected repair effect (such as the degree of pH recovery, the reduction in erosion rate, and the extension of remaining lifetime) and minimize the repair cost (such as the amount of repair agent consumed and energy consumption). Through training in a large number of simulated erosion and repair scenarios, the RL agent can learn which repair actions to take under different erosion states to obtain the optimal long-term cumulative reward, thereby generating a highly intelligent repair strategy. In addition, a pre-defined repair knowledge base assists in RL decision-making, including various repair agents (such as nano-repair agents). Sodium silicate solution Performance data of the solution (such as neutralization capacity, pore-sealing efficiency, inhibition of sulfate reaction, release / reaction kinetics at different pH, compatibility with geopolymer matrix, cost, etc.) as well as the applicable conditions and potential risks of different remediation mechanisms.

[0059] Furthermore, the self-healing unit 102 optimized for sulfuric acid erosion is an actuator for actively repairing damage. This self-healing unit optimized for sulfuric acid erosion includes at least one of the following: (1) pH-responsive microcapsules: Their shell material (such as a chitosan / sodium alginate composite nanocoating prepared by layer-by-layer self-assembly technology) undergoes structural destruction, controlled swelling, or a sharp increase in permeability (e.g., a 10-fold to 1000-fold increase in permeability) when the pH value drops to a preset threshold (e.g., pH < 9.0 or pH < 7.0, depending on the type of repair agent and the target pH), thereby releasing the internally encapsulated alkaline neutralizing agent (e.g., nanoscale) at a controlled rate. , Or buffering organic amine compounds such as triethanolamine) or pore-blocking precursors such as low-viscosity (e.g., <100 mPa·s) tetraethoxysilane TEOS with acidic or basic properties. Initiator system, such as hydroxyethyl methacrylate (HEMA) and benzoyl peroxide (BPO).

[0060] In this embodiment, the core of the microcapsule is encapsulated with an alkaline neutralizing agent (such as nano-sized...). or Particle size 50-300 nm (to improve reactivity and dispersibility) or pore-blocking agent precursors (such as low viscosity (<50 mPa·s) ethyl silicate (TEOS) with acidic... The microcapsules are made from a mixture of oxidizing agents (such as dilute hydrochloric acid) or methyl methacrylate (MMA) monomers and benzoyl peroxide (BPO) initiators. The shell of the microcapsules is made from a pH-sensitive polymer (such as poly(dimethylaminoethyl methacrylate) (PDMAEMA) with a specific grafting ratio, which swells under protonation under acidic conditions, or a composite shell formed by the layer-by-layer self-assembly of calcium alginate and chitosan, which increases permeability at low pH due to the dissociation of calcium alginate). The microcapsules are uniformly dispersed or selectively enriched in areas expected to be susceptible to corrosion at a dosage of, for example, 1%-5% of the total polymer volume. When sulfuric acid corrosion causes the local pH value to drop to the response threshold of the microcapsule shell material (e.g., pH < 9.0 or pH < 7.5), the shell material structure changes, resulting in the controlled release of the repair agent into cracks or pores to neutralize the acid or form a filler.

[0061] (2) AI-controlled microvascular network system: The microvessels are made of acid-resistant flexible (e.g., PTFE or silicone rubber) or rigid (e.g., borosilicate glass microtubules) materials, with channel inner diameters of, for example, 100µm to 500µm, and are three-dimensionally distributed at a vessel length density of, for example, 0.5cm to 5cm per cubic centimeter of polymer. This network is connected to one or more replaceable remedial reservoirs (each reservoir volume, for example, 1mL to 10mL, containing, for example, 0.1M to a saturated concentration of...). Solution, 10%-30% sodium silicate solution, or 0.05M to 0.2M solution. The solution is delivered to specific damaged areas on demand, quantitatively (e.g., single delivery volume of 0.01 mL to 0.5 mL), and at timed intervals (e.g., delivery duration of 1 minute to 30 minutes) by a micro piezoelectric pump (with flow control accuracy of, for example, 1 nL / min to 10 µL / min) and a micro solenoid valve array (with a switching response time of less than, for example, 100 ms) through a micro piezoelectric pump (with a flow control accuracy of, for example, 1 nL / min to 10 µL / min) and a micro solenoid valve array (with a switching response time of less than, for example, 100 ms).

[0062] In one implementation, the self-healing unit 102 is a three-dimensional interconnected microvascular network, formed by pre-embedding a sacrificial template (such as a 3D-printed PLA or PVA fiber network, 150-800 µm in diameter) before casting the geopolymer and removing the template after the geopolymer has hardened. The network's inlets are connected via micro-connectors to externally or internally integrated replaceable repair agent reservoirs (e.g., reservoir A containing 0.5 M...). The suspension was prepared in two containers: reservoir B contained a 20% sodium silicate solution, and reservoir C contained a 0.1M solution. The solutions are interconnected, and each reservoir's outlet or key node in the network branch is equipped with a miniature solenoid valve (e.g., normally closed type, response time <50ms) and a miniature piezoelectric ceramic pump (e.g., capable of precise flow control from 10nL / min to 50µL / min), precisely controlled by a central AI decision module 105 via digital control signals. Based on the generated optimal repair strategy, the AI ​​can independently or collaboratively control the pump and valve systems of different reservoirs, delivering specific types and dosages of repair agents through the microvascular network to the target damage area determined by the AI.

[0063] Optionally, the proposed system may also include other types of sensor arrays 106 for assisting in the assessment of the degree of sulfuric acid erosion. For these other types of sensor arrays 106, an electrical impedance spectroscopy (EIS) electrode array (e.g., a four-electrode configuration, operating frequency range of 1Hz to 1MHz, excitation voltage of, for example, 10mV to 100mV) or an acoustic emission (AE) sensor (e.g., resonant frequency of 50kHz to 300kHz, sensitivity better than -70dB) can be integrated. The data collected includes charge transfer resistance, double-layer capacitance, energy, amplitude, and duration of AE events in the EIS spectrum. This data, combined with the pH sensing data, can be fused and analyzed by the central AI decision module using a multimodal deep learning model or a Bayesian inference network to more comprehensively assess changes in pore structure (e.g., percentage increase in porosity, changes in pore size distribution), alterations in ion migration characteristics, and the initiation and propagation of microcracks (e.g., crack density, length) caused by sulfuric acid erosion. This assists the AI ​​in more accurately determining the erosion stage, assessing the degree of damage, and verifying the repair effect.

[0064] In this embodiment, the system integrates an electrical impedance spectroscopy (EIS) electrode array or an acoustic emission (AE) sensor (collectively referred to as other types of sensor array 106). The EIS electrode array (e.g., using stainless steel or graphite electrodes, arranged in a four-electrode configuration such as Wenner or Schlumberger) acquires impedance spectroscopy data reflecting the polymer's pore structure, ionic conductivity, and interface state by applying a small AC voltage (e.g., 10-50 mVRMS) in a frequency range of, for example, 10 mHz to 1 MHz and measuring the current response. The AE sensor (e.g., with a resonant frequency in the range of 100 kHz to 500 kHz) is used to capture elastic wave signals released by the initiation and propagation of microcracks caused by sulfate crystallization expansion or skeletal dissolution. The aforementioned auxiliary sensing data, along with pH sensing data, is fused in a multimodal manner by constructing a joint feature space or employing an attention-based fusion network. This allows for a more comprehensive and accurate assessment of the physical damage (such as increased porosity and crack density) and chemical degradation state of sulfuric acid corrosion, thereby further improving the accuracy of AI diagnostic predictions and the rationality of repair decisions.

[0065] Furthermore, the system proposed in this embodiment also includes an energy management module. As a key supporting component of the sulfuric acid-resistant self-healing geopolymer system proposed in this embodiment, the energy management module's main function is to provide a reliable and continuous power supply for the entire intelligent monitoring and self-healing system, ensuring stable operation of the system throughout the entire lifecycle of the geopolymer structure. Specifically, the core functions of the energy management module include: First, the energy management module is electrically connected to various modules or components in the system, providing a suitable power supply for each component, including: the signal acquisition and conditioning circuit of the in-situ multi-point pH sensor network (operating voltage 3.3V to 5V, power consumption approximately 10μW to 100μW per sensor), the microcontroller of the edge computing module (operating voltage 1.8V to 3.3V, average power consumption 10mW to 100mW), the processing unit of the central AI decision module (operating voltage 5V to 12V, peak power consumption up to several watts), and actuators in the self-healing unit such as miniature piezoelectric pumps (drive voltage 12V to 24V, instantaneous power 1W to 5W) and miniature solenoid valve arrays (operating voltage 5V to 12V, single valve power consumption 0.5W to 2W). Preferably, if other types of sensor arrays are provided in the system, the energy management module also supplies power to those sensor arrays.

[0066] Secondly, the energy management module integrates an intelligent power consumption management strategy, which optimizes energy utilization efficiency by dynamically adjusting the system's operating status. During normal monitoring, the system operates in a low-power mode, with the pH sensor sampling at a lower frequency (e.g., once per hour). When the edge computing module detects an erosion indication, the energy management module immediately switches the relevant components to a high-performance mode, increasing the sampling frequency and computing power to support real-time monitoring and rapid response.

[0067] Furthermore, considering that geopolymer structures are often in environments where frequent maintenance is difficult, the energy management module is equipped with a high-energy-density lithium thionyl chloride battery. It can be powered by a lithium-ion battery pack or other lithium-ion battery pack, with a capacity designed to support continuous system operation for 3 to 5 years. Additionally, the module can be equipped with an energy harvesting unit, such as a piezoelectric vibration energy harvester or a thermoelectric generator, to obtain supplementary energy from the environment, further extending the system's autonomous operating time.

[0068] In addition, the energy management module has comprehensive power monitoring and protection functions, monitoring the remaining battery power, output voltage and current status in real time, and ensuring system safety through overvoltage, overcurrent, and short-circuit protection circuits. When it detects that the battery is about to run out, the module will send a low battery warning to the central AI decision-making module in advance, enabling the system to take appropriate energy-saving measures or remind maintenance personnel to replace the battery.

[0069] By integrating the above-mentioned energy management module, the sulfuric acid erosion-resistant self-healing geopolymer system of this embodiment can achieve truly autonomous and long-term operation without frequent manual intervention and maintenance, which significantly improves the practicality and reliability of the system and is particularly suitable for remote areas, underground facilities or other inaccessible application scenarios.

[0070] As another implementation, the specific chemical composition of the geopolymer matrix 101 in this embodiment can be pure metakaolin-based or geopolymer containing volcanic ash, to adapt to different sulfuric acid concentrations or temperature environments; the specific type of sensors in the in-situ multi-point pH sensing network 103 can be fiber optic pH sensors to resist electromagnetic interference, and their quantity and spatial layout can be optimized for components of specific shapes, such as pipes or plates; the specific hardware platform used by the edge computing module 104 and the central AI decision-making module 105 can be a low-power dedicated AI chip for edge computing or a distributed cloud computing resource for central decision-making, and the specific implementation details of the AI ​​algorithm can be imputed using a generative adversarial network (GAN) with a specific structure, or a transfer learning method can be used. The method allows for the rapid adaptation of models trained in other material systems to geopolymers. The specific chemical formulation of the repair agent in the self-healing unit 102 can utilize novel, efficient, long-lasting repair agents highly compatible with geopolymers. Its encapsulation / delivery technology can employ magnetically controlled or acoustically controlled microcapsule release technology, or utilize microfluidic chips to precisely mix and deliver multi-component repair agents. The specific configuration of the energy management module can integrate higher-efficiency energy harvesting devices or larger-capacity energy storage units. All of the above settings can be flexibly selected, combined, and optimized based on actual application scenarios (such as long-term monitoring and maintenance of large-scale infrastructure, structural safety assurance in high-risk chemical environments, or as a tool for the research and evaluation of novel sulfuric acid-resistant materials), cost budgets, and the technological development level in related fields. For example, for critical structures requiring ultra-high reliability and extremely long service life, a synergistic repair scheme with higher-density redundant sensor designs and multiple repair mechanisms (such as pH-responsive microcapsules and AI-controlled microvascular networks) can be adopted, supplemented by a more powerful AI model for refined management. For cost-sensitive large-scale applications, the focus can be on optimizing the cost-effectiveness of sensors, the energy efficiency ratio of edge computing, and the economics of the repair agent.

[0071] Preferably, the system was compared with traditional geopolymer materials in a laboratory accelerated sulfuric acid erosion test. The test conditions were immersion in 5% sulfuric acid solution at 25°C. The results showed that the traditional geopolymer specimen, as a control group, experienced a 68% loss in compressive strength after 90 days, while the geopolymer specimen integrated with the system only experienced a 15% loss. Through the prediction and active repair of the AI ​​system, the system successfully controlled the acid erosion area with pH < 7 to within 2 mm of the surface layer, while the erosion depth of the control group reached 12 mm. This further verifies the superiority of the proposed solution in this embodiment.

[0072] Example 2 This embodiment proposes a self-healing method for a sulfuric acid-resistant self-healing geopolymer system, applicable to the sulfuric acid-resistant self-healing geopolymer system proposed in Embodiment 1 above, such as... Figure 3As shown, the specific steps include: Step S0, Preliminary Preparation: Start the system and perform a self-test to confirm that all sensors are functioning properly. In normal monitoring mode, pH sampling is performed every 4 hours. When any sensor detects pH < 10.5, the system automatically switches to intensive monitoring mode, increasing the sampling frequency to once every 30 minutes. When pH < 9.0 and the rate of decrease is > 0.05 pH / h, the AI ​​deep analysis process is triggered. The AI ​​system completes erosion assessment and prediction within 10 minutes of receiving the data and generates a repair plan within 5 minutes after confirming the need for repair. During the repair process, the pH recovery is continuously monitored, and the repair effect is evaluated every 5 minutes. Repair parameters are dynamically adjusted as needed.

[0073] Step S1: Real-time monitoring of the spatiotemporal evolution of the pH field in the geopolymer matrix during sulfuric acid erosion. Specifically, initialization and in-situ dynamic pH field monitoring: After system startup, the edge computing module 104 initializes all pH sensors 301 and temperature sensors, and begins collecting data at a preset baseline sampling frequency (e.g., once every 30 minutes). For example, the in-situ multi-point pH sensing network collects pH values ​​and collaboratively measured temperature data at various monitoring points inside the geopolymer matrix at a frequency of 5 minutes to 4 hours (this frequency can be dynamically adjusted by the central AI decision module according to the current erosion rate, with increased sampling density during the rapid pH drop phase). The raw electrical signals (e.g., at the mV level) are then transmitted to the edge computing module for edge intelligent preprocessing. Preferably, the edge computing module performs real-time temperature compensation (e.g., correction based on the Nernst equation or a pre-stored temperature coefficient), AI-based dynamic calibration (e.g., updating the calibration curve using periodic calibration data from a reference electrode or a known pH buffer), and validity verification (e.g., checking whether the detection signal is within a reasonable range or whether the noise level is too high) on the original signal, converting it into an accurate pH value, and extracting preliminary feature vectors such as the pH change rate (e.g., calculating the slope by performing linear regression on the most recent N sampling points) and the pH spatial gradient (e.g., calculating the ratio of the pH difference between adjacent sensors to their distance).

[0074] Step S2: Based on the received spatiotemporal evolution data of the pH field, preliminary erosion indicators are identified. Then, using the onboard dynamic tracking and quantification model of the sulfuric acid erosion front and the prediction model based on physical information, the erosion status is assessed and the development trend is predicted. That is, AI-driven in-depth assessment and trend prediction of sulfuric acid erosion status: First, when the edge computing module identifies a preliminary erosion indication (e.g., the pH value at any monitoring point drops by more than 0.8 pH units continuously within 24 hours, or the pH values at multiple adjacent monitoring points are simultaneously lower than a preset early warning threshold such as pH = 10.5), or at a preset evaluation period (e.g., every 12 hours to 72 hours), it uploads the processed pH feature data or the original pH spatio-temporal data sequence to the central AI decision-making module.

[0075] In this embodiment, the edge computing module uploads the processed pH data sequence (including the pH value, change rate, and sensor health status identifier of each sensor in the past 24 hours) regularly (e.g., once every 6 hours) or when a preliminary erosion indication is detected (the average pH value in a certain area is lower than 11.0 and the continuous decline rate exceeds 0.1 pH unit / day) to the central AI decision-making module via wireless communication.

[0076] Second, the central AI decision-making module uses the sulfuric acid erosion front dynamic tracking and quantification model to analyze the distribution characteristics of the current three-dimensional pH field (such as the shape of the isopH surface, the area with the maximum gradient), the precise position, geometric morphology, and average normal advancement rate of the erosion front, and uses the PINN or mechanism-enhanced AI model to predict the development of the erosion depth within a specific future time window (e.g., the next 15 days to 6 months) (e.g., predicting that the area with pH < X will expand by Y millimeters), the evolution trend of the pH field (e.g., predicting the change of the entire pH profile over time), and the potential risk level of expansion cracking caused by sulfate reactions (e.g., outputting a low / medium / high risk level based on the predicted amount and distribution of gypsum or ettringite and the anti-expansion ability of the material).

[0077] In this embodiment, the central AI decision-making module uses its built-in sulfuric acid erosion front dynamic tracking and quantification model to analyze the three-dimensional pH field distribution of the entire monitoring area, accurately identify the position, shape, and advancement rate of the erosion front, quantify the pH gradient characteristics, and use the PINN or mechanism-enhanced AI model to combine the current pH field state, known sulfuric acid exposure conditions (such as external sulfuric acid concentration, temperature), and the initial characteristic parameters of the geopolymer (such as porosity, main chemical composition content) to predict the development of the erosion depth, the evolution trend of the pH field, and the potential risk level and occurrence area of expansion cracking caused by sulfate reactions (such as the formation of gypsum or ettringite) within a specific future time window (e.g., the next 1 month, 3 months, 6 months). Additionally, if other types of sensors are integrated, their data are also fused and analyzed at this stage.

[0078] Step S3: Based on the preset repair knowledge base for sulfuric acid corrosion and the current corrosion status assessment and prediction results, an optimal repair strategy is generated using a reinforcement learning algorithm or a genetic algorithm based on multi-objective optimization to control the self-repair unit to perform self-repair. That is, intelligent repair strategy generation and optimization: If the central AI decision module assesses that the current corrosion status has reached or is predicted to reach the preset repair initiation conditions in the short term (e.g., within the next 3 to 14 days) (e.g., the corrosion depth has exceeded the design allowable value by 20%-50%, or the predicted pH drop rate will cause the pH to fall below a certain key failure threshold such as pH=7.0 or 6.5 within 7 days, or the predicted expansion stress has reached 30%-60% of the material's tensile strength), then based on the repair knowledge base and reinforcement learning (RL) algorithm or multi-objective genetic algorithm, an optimal repair strategy is generated for the current corrosion characteristics (e.g., the lowest pH point cloud, the maximum gradient path, the predicted high-risk expansion core area, or the key bottleneck node of the corrosion channel).

[0079] The optimal remediation strategy specifies in detail: (a) the type or combination of remediation agents selected (e.g., for areas with rapidly decreasing pH but low risk of expansion, highly efficient alkaline neutralizers are preferred; for areas with microcracks and risk of expansion, a combination of alkaline neutralizers and pore-sealing agents, supplemented with expansion inhibitors, may be selected); (b) the precise three-dimensional spatial range activated by the remediation agent (e.g., defining a target remediation area centered on the erosion front with a radius of R and a depth of D); and (c) the release dosage of the remediation agent (e.g., based on AI predictions). (d) Calculate the required molar amount of alkaline substance or the volume of pore-blocking agent based on the total acid consumption and pore volume in the region; (e.g., for pH-responsive microcapsules requiring external heat triggering, set the target heating temperature curve to 55-70°C, the heating rate to 1-5°C / min, and the holding time to 15-120 minutes; for microvascular network delivery, set the flow rate of repair agent A to 0.01-0.1 mL / min and the total delivery volume to 0.2-2 mL, then switch to repair agent B and set different parameters).

[0080] In this embodiment, when the evaluation and prediction results of the central AI decision-making module indicate that the current erosion state has reached or is predicted to reach the preset repair initiation conditions in the short term (e.g., within the next 7 days) (e.g., the erosion depth has exceeded X% of the design allowable value (e.g., 30%), or the predicted pH drop rate will cause the pH to fall below a certain critical failure threshold Z (e.g., pH=8.0) within Y days (e.g., 14 days), or the predicted expansion stress has reached A% of the material's tensile strength (e.g., 40%)), the repair decision-making process is initiated. Then, based on the repair knowledge base and reinforcement learning (RL) algorithms or multi-objective genetic algorithms, an optimal repair strategy is generated for the current erosion characteristics (e.g., the lowest pH core area, the maximum gradient penetration path, the predicted high-risk expansion area, or the key connecting nodes of the erosion channel). This strategy specifies in detail the target three-dimensional region for repair, the type or combination of repair agents selected (e.g., for areas with rapidly decreasing pH but low expansion risk, highly efficient alkaline neutralizers such as nano-alkaline neutralizers are preferentially selected). Microcapsules; for areas with existing microcracks and a high risk of expansion, a combination of alkaline neutralizers and pore-sealing agents (such as TEOS-based repair agents) may be chosen, and small amounts can be delivered via microvascular networks. This includes precise dosage of the solution as a swelling inhibitor, accurate release of the repair agent (e.g., calculating the required molar amount of alkaline substance or the volume of pore-blocking agent based on the total acid consumption and pore volume to be filled predicted by AI, with an accuracy of, for example, ±10% of the required amount), and detailed control parameters (e.g., for pH-responsive microcapsules requiring external heat triggering, setting a target heating temperature profile (e.g., increasing from room temperature to 65°C at a rate of 2°C / min and holding for 30 minutes) and specifying the microheater to be activated; for microvascular network delivery, setting the delivery flow rate (e.g., 5 µL / min) and total delivery volume (e.g., 0.5 mL) of repair agent A, and then switching to repair agent B with different parameters). The dosage of the alkaline neutralizer is also important. Where k is an empirical coefficient (0.1-0.3), and A is the area of ​​the eroded region ( ), where d is the erosion depth (mm). The target pH recovery value.

[0081] Preferably, during the self-repair process described above, the repair parameters in the optimal repair strategy are dynamically fine-tuned based on the real-time feedback of the pH recovery rate and spatial range, thereby achieving closed-loop control. In other words, targeted repair implementation and closed-loop process control: The central AI decision module sends precise control command sequences to the corresponding self-repair units (e.g., auxiliary activation devices for microcapsules in specific areas, or pump-valve drive systems for microvascular networks) through digital or analog control interfaces (e.g., sending instructions to the microcontroller via SPI or I2C bus, which in turn drives the power devices) to release / deliver the repair agent in stages or in a coordinated manner within the target area specified by the AI ​​and under preset parameters. During the release / delivery of the repair agent (e.g., every 1 to 15 minutes), the central AI decision module can dynamically fine-tune the repair parameters (e.g., adjusting the delivery flow rate of the microvascular network or the power output of the micro-heating device in real time) based on the real-time recovery rate and spatial range of the pH value fed back by the pH sensing network (e.g., monitoring whether the pH value of the target area rises at the expected rate, or whether the target pH value is reached, such as pH>9.0) using PID (proportional-integral-derivative) control algorithms or fuzzy logic control algorithms, in order to achieve closed-loop fine control of the repair process, ensure the repair effect, and avoid waste of the repair agent or negative impact on the surrounding area.

[0082] In this embodiment, the central AI decision-making module sends precise control command sequences to the corresponding self-repairing units (e.g., the power controller of the auxiliary activation device for microcapsules in a specific region, or the pump-valve drive board of the microvascular network) via a digital or analog control interface. During the release or delivery of the repair agent (e.g., for microvascular delivery, this can last from several minutes to several hours), the central AI decision-making module obtains pH feedback of the target repair area and its surroundings from the pH sensing network at a higher frequency (e.g., every 1 to 10 minutes), and uses a PID (proportional-integral-derivative) control algorithm or a fuzzy logic control algorithm to monitor the pH recovery rate and spatial range in real time, while comparing it with the expected pH change curve in the repair strategy. If there is a significant deviation (e.g., pH recovery is too slow or too fast, or the repair range does not reach the expected level), the AI ​​dynamically fine-tunes the repair parameters (e.g., adjusting the delivery flow rate or pressure of the microvascular network in real time, or extending / shortening the heating time of the micro-heating device or changing the heating power) to achieve closed-loop fine control of the repair process, ensuring the repair effect and avoiding excessive consumption of the repair agent or potential negative impacts on the surrounding healthy areas.

[0083] Preferably, the process also includes step S4, long-term evaluation of the repair effect, and adaptive evolution of the AI ​​model. Specifically, after the repair action is completed, during an evaluation period (e.g., 7 days to several months, or even longer, depending on the severity of the erosion environment and the long-lasting effect of the repair agent), the central AI decision-making module continuously monitors the pH field stability of the repaired area and surrounding areas, the quiescence or reactivation status of the erosion front (e.g., comparing the erosion rate before and after repair) using a pH sensing network (and optional other sensors, such as EIS or AE), quantifies the repair effect (e.g., the percentage of area where the pH returns to a safe range (e.g., pH > 9.5) and maintains this state, the duration of a significant reduction in erosion rate (e.g., a reduction of more than 80%), and the predicted increase in the remaining useful life of the material (e.g., the percentage of area where the pH returns to a safe range (e.g., pH > 9.5) and maintains this state), and the amount of increase in the predicted remaining useful life of the material (e.g., the percentage of area where the pH returns to a safe range (e.g., pH > 9.5) and maintains this state). (Increase by 20%-200%), and then use this complete data chain of "sulfuric acid attack event characteristics - multi-source sensor data sequence - AI diagnosis and prediction results - repair decision parameters - repair execution process data - long-term repair effect evaluation" as a high-quality experience sample. This data will be used offline or online to update and optimize the sulfuric acid attack prediction model (especially the physicochemical parameters in PINN or the weights and structure of the AI ​​model) and the repair strategy optimization model (e.g., the Q-value table of the RL agent, policy network, or reward function) in its central AI decision module. This will improve the system's intelligence level and adaptability in dealing with future sulfuric acid attack events or different polymer formulations and environmental conditions. During the initial operation phase (first 6 months), the model will be updated after each repair cycle. During the stable operation phase (after 6 months), the model will be optimized quarterly. An emergency update mechanism will be triggered when the prediction error exceeds 25%.

[0084] In this embodiment, after the repair operation is completed, the system enters a long-term evaluation period (e.g., 7 days to 6 months, or even longer, the specific duration of which can be dynamically set by the AI ​​according to the repair type and the severity of the erosion environment). During this period, the central AI decision-making module continuously monitors the pH field stability (e.g., whether the pH value can be maintained within a safe range, such as pH > 9.0) and the quiescence or reactivation status of the erosion front (e.g., whether the erosion rate before and after repair is significantly reduced and remains at a low level) of the repaired area and surrounding area using a pH sensing network (and optional other sensors, such as EIS or AE, whose measurement frequency can be appropriately reduced). By comparing the pH field data, erosion parameters (depth, rate), and optional EIS / AE data before and after repair, the repair effect is quantified (e.g., the percentage of area where the pH returns to a safe range and the length of time it remains in this state, the percentage of effective reduction in erosion rate (e.g., a reduction of more than 90%) and its duration, and the predicted increase in the remaining service life of the material (e.g., an increase of 50%-300%). Simultaneously, the complete data chain of this "sulfuric acid erosion event characteristics (initial pH field, sulfuric acid concentration, etc.) - multi-source sensor data sequence (pH, temperature, EIS, AE) - AI diagnosis and prediction results (erosion depth, rate, risk level) - remediation decision parameters (remediation agent, location, dosage, control process) - remediation execution process data (actual release / delivery, pH recovery curve) - long-term remediation effect evaluation results" is stored as a high-quality, clearly labeled experience sample in the training database of the AI ​​model. Through regular offline retraining or incremental online learning, the sulfuric acid erosion prediction model (especially the physicochemical parameters in PINN or the weights and network structure of the AI ​​model) and the remediation strategy optimization model (e.g., Q-value table of RL agent, policy network or reward function) in its central AI decision module are continuously updated and optimized to improve the system's intelligence level, prediction accuracy and adaptability in dealing with more complex or unseen sulfuric acid erosion events in the future.

[0085] Furthermore, in the aforementioned process, the energy management module in the system, as a key supporting component of the sulfuric acid-resistant self-healing geopolymer system, provides a reliable and continuous power supply to the entire intelligent monitoring and self-healing system, ensuring stable operation of the system throughout the entire lifecycle of the geopolymer structure. Preferably, this energy management module is equipped with a high-energy-density lithium thionyl chloride battery (…). The system can be powered primarily by a lithium-ion battery pack or other energy harvesting unit (such as a piezoelectric vibration energy harvester or a thermoelectric generator module), ensuring reliable power supply. The module integrates intelligent power management strategies to optimize energy efficiency by dynamically adjusting the system's operating state. It also features power monitoring and protection functions, real-time monitoring of remaining battery charge, output voltage, and current status, and overvoltage, overcurrent, and short-circuit protection circuits to ensure system safety. Through this energy management module, the self-healing process of the sulfuric acid-resistant self-healing geopolymer system can achieve truly autonomous and long-term operation, eliminating the need for frequent manual intervention and maintenance, significantly improving the system's practicality and reliability.

[0086] Through the aforementioned system design and highly intelligent collaborative self-healing method, this embodiment endows geopolymer materials with unprecedented autonomous perception, deep understanding, accurate prediction, active repair, and continuous evolution capabilities in response to the specific and severe challenge of sulfuric acid corrosion. It can not only promptly and effectively repair internal damage caused by sulfuric acid corrosion, significantly extending the service life of materials and structures in sulfuric acid environments and improving their reliability and safety; but also, through continuous learning and optimization by AI, it can continuously adapt to changing service conditions and material aging states, achieving continuous improvement in repair performance and continuous evolution of intelligence. This opens up a new technical path for developing a new generation of ultra-durable, self-healing, and intelligent sulfuric acid-resistant geopolymer materials and their application in various harsh engineering environments, possessing significant technological advancements and broad industrialization prospects. It also provides advanced scientific tools for in-depth research into the mechanisms of material degradation under complex environmental factors.

[0087] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0088] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A self-healing geopolymer system resistant to sulfuric acid erosion, characterized in that, include: The geopolymer matrix exists in a highly alkaline environment when not subjected to sulfuric acid erosion. At least one self-healing unit optimized for sulfuric acid erosion is embedded in a geopolymer matrix. The self-healing unit contains a repair agent for neutralizing acidity, sealing pores, or inhibiting sulfate reactions. An in-situ multi-point pH sensing network integrated within a geopolymer matrix is ​​used to monitor the spatiotemporal evolution of the pH field during sulfuric acid erosion in real time. The edge computing module is used to identify preliminary erosion indicators based on the received spatiotemporal evolution data of the pH field; The central AI decision-making module is used to assess the erosion status and predict the development trend based on the identified preliminary erosion indications and pH field spatiotemporal evolution data, using an AI model optimized for sulfuric acid erosion, and to generate the optimal remediation strategy to control the self-repair unit to perform self-repair.

2. The sulfuric acid-resistant self-healing geopolymer system as described in claim 1, characterized in that, The in-situ multi-point pH sensing network includes several miniature pH sensors uniformly distributed at multiple points inside the geopolymer matrix. Each miniature pH sensor is used to collect the pH value of each monitoring point inside the geopolymer matrix in real time. The miniature pH sensor employs a solid-state potentiometric sensor based on an iridium oxide thin-film electrode or a pH-sensitive functionalized polymer, and features a dual-resistance encapsulation structure designed to withstand both the initial high alkalinity of the polymer and the subsequent strong sulfuric acid erosion environment.

3. The sulfuric acid-resistant self-healing geopolymer system as described in claim 1, characterized in that, The edge computing module is used to preprocess the collected spatiotemporal evolution data of the pH field, including: Based on the real-time collected pH values ​​and coordinated temperature data of each monitoring point inside the geopolymer matrix, the pH value is compensated and corrected by a nonlinear temperature compensation algorithm based on polynomial fitting or lookup table. For the pH value after temperature compensation and correction, a second correction is performed using a consistency verification algorithm for neighboring sensor data to obtain the corrected pH value, and the pH field feature vector, including pH change rate and pH spatial gradient, is extracted. Preliminary erosion indicators are identified based on the corrected pH value.

4. The sulfuric acid-resistant self-healing geopolymer system as described in claim 1, characterized in that, The central AI decision-making module is equipped with a dynamic tracking and quantification model of the sulfuric acid erosion front, a prediction model based on physical information, and an AI decision-making model. Using a dynamic tracking and quantification model of the sulfuric acid erosion front, we analyze the spatial distribution characteristics of the current three-dimensional pH gradient field and the average normal advance rate of the erosion front. Using a prediction model based on physical information, we can predict the evolution trend of the pH field, the development of erosion depth, and the risk level of expansion and cracking within a specific time window in the future. Using an AI decision-making model, based on the repair knowledge base and the assessment and prediction results of the current erosion state, a reinforcement learning algorithm or a genetic algorithm based on multi-objective optimization is used to generate the optimal repair strategy for the current erosion state.

5. The sulfuric acid-resistant self-healing geopolymer system as described in claim 4, characterized in that, A dynamic tracking and quantification model for the sulfuric acid erosion front: The spatiotemporal evolution data of pH field collected by an in-situ multi-point pH sensing network at continuous time points are input into the model. The spatial resolution is set to track the three-dimensional position and morphological evolution of a specific pH isosurface in the geopolymer matrix in real time, and calculate the average normal advance rate of the sulfuric acid erosion front and the spatial distribution characteristics of the pH gradient field. For the prediction model based on physical information: the average normal advance rate of the sulfuric acid erosion front and the spatial distribution characteristics of the pH gradient field are input into the model to predict the evolution trend of the pH field, the development of erosion depth, and the accumulation or distribution area of ​​key erosion products within a set time window in the future, and to clarify the risk level of expansion and cracking. Specifically, for the prediction model based on physical information, the rate equation describing the key chemical reaction between sulfuric acid and the main hydration products in the geopolymer matrix, as well as the diffusion transport equation describing the diffusion coefficient of key ions in the porous structure of the geopolymer matrix, are used as partial differential equation constraints and incorporated into the loss function of the neural network model, thereby continuously iterating and training the prediction model.

6. The sulfuric acid-resistant self-healing geopolymer system as described in claim 4, characterized in that, In the AI ​​decision-making model, based on a pre-set knowledge base for the repair of sulfuric acid corrosion and the current corrosion status assessment and prediction results, the optimal repair strategy is generated through reinforcement learning algorithm or genetic algorithm based on multi-objective optimization. The optimal repair strategy is generated using a reinforcement learning algorithm. This involves using a deep Q-network, setting the state space to include the current pH field feature vector, propulsion rate, and predicted expansion and cracking risk level, and setting the action space to include the selection of the type and combination of repair agents, the spatial range of the activation location, the activation intensity, and the delivery dosage of the repair agent. The optimal repair strategy is obtained through iterative optimization of the action-state relationship. The optimal repair strategy is generated by a genetic algorithm based on multi-objective optimization. The strategy is as follows: with the repair effect, repair cost and repair time as objectives, the type and combination of repair agents, the spatial range of activation sites, activation intensity and delivery dosage of repair agents are randomly selected as a group of individuals. Through fitness evaluation and crossover and mutation operations, the individuals are continuously updated iteratively until the multi-objective optimality is reached, and the optimal repair strategy is obtained.

7. The sulfuric acid-resistant self-healing geopolymer system as described in claim 6, characterized in that, The remediation knowledge base includes: performance parameters of different pH-responsive remediation agents, curing time and sealing efficiency of pore plugging agents under different pH and humidity conditions, effective concentration range and potential side effects of sulfate reaction inhibitors.

8. The sulfuric acid-resistant self-healing geopolymer system as described in claim 6, characterized in that, The optimal repair strategy includes: the type and combination of repair agents, the range of three-dimensional spatial regions activated by the repair agents, the release dosage of the repair agents, and repair regulation parameters.

9. The sulfuric acid-resistant self-healing geopolymer system as described in claim 1, characterized in that, The self-healing unit includes one or more of pH-responsive microcapsules and microvascular network systems; Among them, for pH-responsive microcapsules, when the pH value drops to a preset threshold, the shell material undergoes structural damage, controllable swelling, or a sharp increase in permeability, releasing the internally encapsulated alkaline neutralizer or pore-blocking agent precursor at a controlled rate. For the microvascular network system, the microvessels are made of acid-resistant materials, and the inner diameter of the microvascular channels is three-dimensionally distributed according to the set vascular length density. The microvascular network is connected to one or more replaceable repair agent reservoirs, and the repair agent is delivered to specific damaged areas on demand, in a quantitative and timed manner through a micro piezoelectric pump and a micro solenoid valve array controlled by an AI decision module.

10. A self-healing method for a sulfuric acid-resistant self-healing geopolymer system as described in any one of claims 1-9, characterized in that, include: Real-time monitoring of the spatiotemporal evolution of pH field in geopolymer matrices during sulfuric acid erosion; Based on the received spatiotemporal evolution data of pH field, preliminary erosion indicators are identified. Then, using the onboard dynamic tracking and quantification model of sulfuric acid erosion front and the prediction model based on physical information, the erosion status is assessed and the development trend is predicted. Based on a pre-defined knowledge base for repairing sulfuric acid erosion and the current erosion status assessment and prediction results, an optimal repair strategy is generated using a reinforcement learning algorithm or a genetic algorithm based on multi-objective optimization to control the self-repairing unit to perform self-repair. During the self-repair process, the repair parameters in the optimal repair strategy are dynamically fine-tuned based on the real-time feedback of pH recovery rate and spatial range to achieve closed-loop control.

Images