Concrete carbon sequestration production process optimization method

Abstract

The invention provides a concrete carbon sequestration production process optimization method, and belongs to the technical field of concrete carbon sequestration production. After residual air in pores is discharged through vacuum pretreatment, pressurized carbon dioxide is injected, and ultrasonic waves are applied to assist gas dissolution and diffusion; a distributed temperature sensor array is used for collecting temperature field data in real time and inputting the temperature field data into a carbonization temperature control optimization model to calculate the optimal maintenance temperature set value of each area, and the flow of circulating cooling water and the spray humidification system are regulated and controlled according to the set values to accurately maintain the maintenance temperature and humidity. Meanwhile, a near infrared spectrum detection device is adopted for scanning every 2-4 hours, the current carbon sequestration amount is analyzed through a carbon sequestration amount rapid evaluation model, and the carbonization maintenance time or the carbon dioxide injection pressure is dynamically adjusted according to the deviation degree; the technical problem that it is difficult to achieve temperature field accurate control and carbon sequestration efficiency collaborative optimization in the concrete carbonization curing process is solved.

Description

Technical Field

[0001] This invention belongs to the field of concrete carbon sequestration production technology, and more specifically, relates to an optimization method for concrete carbon sequestration production process. Background Technology

[0002] Concrete carbon sequestration technology achieves carbon capture and storage by promoting the carbonization reaction between carbon dioxide and cement hydration products, making it an important pathway for carbon neutrality in the building materials industry. Traditional concrete carbon sequestration processes employ atmospheric pressure natural carbonization, placing molded concrete components in an air environment containing carbon dioxide for curing. Carbon dioxide naturally diffuses into the concrete pores and reacts with cement hydration products. Curing temperature and humidity are only roughly controlled, and carbon sequestration is measured by observing carbonization depth with phenolphthalein solution spraying or by offline testing of samples using a thermogravimetric analyzer. Due to the significant temperature gradient and uneven carbonization within concrete, current technologies lack the ability to perceive the temperature field distribution in real time, making it impossible to dynamically adjust curing parameters according to the carbonization process. This leads to localized excessively high temperatures causing rapid moisture evaporation or excessively low temperatures causing a decrease in the reaction rate. In other words, existing technologies face the technical challenge of achieving precise temperature field control and synergistic optimization of carbon sequestration efficiency during concrete carbonization curing. Summary of the Invention

[0003] In view of this, the present invention provides an optimization method for concrete carbonation production process, which can solve the technical problem in the prior art that it is difficult to achieve precise temperature field control and synergistic optimization of carbonation efficiency during concrete carbonation curing.

[0004] This invention is implemented as follows: An optimized method for concrete carbon sequestration production process is provided, comprising mixing mineral admixtures with cement to prepare a basic concrete slurry, adding nano-calcium carbonate seeds and a pore conditioner during the mixing process, injecting the basic concrete slurry into a mold, sending it to a curing chamber, and using a vacuum pretreatment device to negatively pressure-extract residual air from the concrete pores, immediately switching to pressurization after the vacuum pretreatment is completed. Injection mode will Gas is injected into the curing chamber, and an ultrasonic generator is simultaneously activated to apply ultrasonic-assisted gas dissolution and diffusion. A distributed temperature sensor array collects real-time temperature field data within the concrete and inputs this data into a carbonation temperature control optimization model to obtain the optimal curing temperature setpoints for each zone. Based on these optimal setpoints, the circulating cooling water flow rate and the operating parameters of the spray humidification system are adjusted to precisely maintain the curing temperature within the set range. During the carbonation curing process, a near-infrared spectroscopy device is used at regular intervals to scan the concrete surface and borehole sampling points to acquire spectral data, which is then input into a rapid carbon fixation assessment model for analysis. The current carbon fixation value calculated by the rapid carbon fixation assessment model is used to determine the degree of deviation between the current value and the target carbon fixation value, and the carbonation curing time is adjusted or increased accordingly. Inject pressure.

[0005] The mineral admixture is mixed with cement in a mass ratio of 30 to 45% fly ash, 25 to 40% slag powder, and 5 to 10% silica fume to prepare the basic concrete slurry.

[0006] The nano-calcium carbonate seed crystals have a particle size of 30 to 80 nm. Particles are used to provide nucleation sites for calcium carbonate crystal growth in the pore liquid phase of concrete.

[0007] The pore regulator is a compound of polycarboxylate superplasticizer and expansion agent, used to regulate the pore size distribution and connectivity during the setting and hardening process of concrete.

[0008] The vacuum pretreatment device is controlled at a suction pressure of -0.08 to -0.12 MPa for a duration of 15 to 25 minutes.

[0009] Wherein, the pressurization In injection mode The gas purity is not less than 99.5%, and the injection pressure is 0.3 to 0.6 MPa.

[0010] The ultrasonic generator applies ultrasonic waves at a frequency of 20 to 40 kHz to assist in the dissolution and diffusion of the gas.

[0011] The distributed temperature sensor array consists of fiber optic temperature sensors arranged on the surface, middle layer and core area of ​​the concrete component, with a sensor spacing of no more than 100mm.

[0012] The temperature field data includes real-time temperature values, temperature change rates, and temperature distribution uniformity indices measured by each fiber optic temperature sensor in the distributed temperature sensor array.

[0013] The carbonization temperature control optimization model is a temperature field prediction and control model based on neural networks, which is based on the input temperature field data, The output parameters of injection pressure, curing time, and mineral admixture proportions are used to determine the optimal curing temperature settings for each zone that maximize the carbonization reaction rate and ensure uniform temperature distribution.

[0014] The carbonization temperature control optimization model utilizes a sparse connection learning framework based on dynamic topology reconstruction. After each training batch, it calculates the average activation intensity and activation frequency of each neuron, calculates the contribution of the connection weights between adjacent neurons to the gradient of the loss function, prunes and deletes connections with a contribution lower than the dynamic threshold, and randomly adds new connections between highly activated neurons.

[0015] The circulating cooling water flow rate is adjustable from 5 to 20 L / min, the atomized particle diameter of the spray humidification system is controlled from 10 to 50 μm, the spray pressure is from 0.2 to 0.5 MPa, the curing temperature is precisely maintained in the range of 40 to 55 °C, and the pore relative humidity is maintained at 75 to 90%.

[0016] The near-infrared spectroscopy detection device scans the concrete surface and the borehole sampling points every 2 to 4 hours to acquire spectral data in the 1200 to 2200 nm band, with a spectral resolution of not less than 5 nm and a single scan time of not more than 30 seconds.

[0017] The rapid carbon fixation assessment model is a spectral data processing model based on a convolutional neural network. It takes spectral data in the 1200 to 2200 nm band collected by a near-infrared spectral detection device as input and outputs the corresponding current carbon fixation value.

[0018] The current carbon sequestration value is calculated as the carbon content per unit mass of concrete. Fixed form The quality and deviation are calculated as a percentage by dividing the absolute value of the difference between the current carbon fixation value and the target carbon fixation value by the target carbon fixation value and then multiplying by 100%.

[0019] Specifically, when the deviation exceeds 8%, the carbonization curing time should be extended or the intensity increased. Injection pressure; when the deviation is less than 3%, carbonization curing is terminated and the post-treatment stage begins. The post-treatment stage includes stopping... Supply, slowly reduce pressure to ambient pressure, and maintain wet curing for 48 to 72 hours.

[0020] This invention utilizes a distributed temperature sensor array to collect real-time three-dimensional temperature field data within concrete. This data is then processed by a neural network-based carbonation temperature control optimization model, which outputs optimal curing temperature setpoints for each region. Based on these setpoints, the flow rate of circulating cooling water and the parameters of the spray humidification system are precisely adjusted, achieving accurate maintenance of the curing temperature within the range of 40 to 55 degrees Celsius. The carbonation temperature control optimization model employs a sparse connection learning framework based on dynamic topology reconstruction. By continuously monitoring the statistical characteristics of neuron activation and the influence of connection weights on the objective function, it automatically identifies and removes redundant connections while exploring new connection paths. This significantly reduces the number of parameters and computational complexity while maintaining the accuracy of temperature field prediction, resulting in a significant improvement in real-time response speed. The dynamically adjusted network topology better adapts to the differences in temperature field evolution under different mineral admixture ratios and curing conditions, avoiding overfitting and enhancing the model's generalization ability to new conditions. In summary, this invention solves the technical problem mentioned in the background art of achieving precise temperature field control and synergistic optimization of carbon sequestration efficiency during concrete carbonation curing. Attached Figure Description

[0021] Figure 1 This is a flowchart of the method of the present invention.

[0022] Figure 2 This is a diagram showing the structure of the vacuum pretreatment device and the change curve of air pressure in the concrete pores.

[0023] Figure 3 To pressurize Schematic diagram of injection and ultrasonic-assisted diffusion system Concentration distribution map.

[0024] Figure 4 This diagram illustrates the arrangement of a distributed temperature sensor array and real-time temperature field monitoring data. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below.

[0026] like Figure 1 The diagram shown is a flowchart of an optimized method for concrete carbon sequestration production provided by this invention. This method includes the following steps:

[0027] S01. Prepare a basic concrete slurry by mixing mineral admixtures with cement in a mass ratio of 30-45% fly ash, 25-40% slag powder, and 5-10% silica fume, and add nano-calcium carbonate seeds and pore conditioners during the mixing process.

[0028] S02. After the foundation concrete slurry is injected into the mold, it is sent to the curing room. A vacuum pretreatment device is used to perform negative pressure suction on the concrete pores. The suction pressure is controlled at -0.08 to -0.12 MPa for 15 to 25 minutes to remove residual air inside the pores.

[0029] S03. Immediately after vacuum pretreatment, switch to pressurized carbon dioxide injection mode and inject carbon dioxide gas with a purity of not less than 99.5% into the curing chamber at a pressure of 0.3 to 0.6 MPa. Simultaneously, start the ultrasonic generator to apply ultrasonic waves with a frequency of 20 to 40 kHz to assist in gas dissolution and diffusion.

[0030] S04. Real-time acquisition of internal temperature field data of concrete through distributed temperature sensor array, input of temperature field data into carbonation temperature control optimization model for processing, and obtaining the optimal curing temperature setpoint for each area.

[0031] S05. Based on the optimal curing temperature setpoints for each region output by the carbonization temperature control optimization model, adjust the circulating cooling water flow rate and the operating parameters of the spray humidification system to precisely maintain the curing temperature within the range of 40–55℃ and the relative humidity of the pores within the range of 75–90%.

[0032] S06. During the carbonization curing process, the concrete surface and borehole sampling points are scanned every 2 to 4 hours using a near-infrared spectroscopy detection device to obtain spectral data in the 1200 to 2200 nm band and input into the carbon fixation rapid assessment model for analysis and processing.

[0033] S07. Based on the current carbon fixation value calculated by the rapid carbon fixation assessment model, determine the degree of deviation between the current carbon fixation value and the target carbon fixation value. If the deviation exceeds 8%, extend the carbonization curing time or increase the carbon dioxide injection pressure. If the deviation is less than 3%, terminate the carbonization curing and enter the post-treatment stage.

[0034] Among them, the traditional concrete carbonation production process uses atmospheric pressure natural carbonation, placing the molded concrete components in a solution containing... Maintenance is carried out in the air environment, relying on Carbonation occurs naturally in concrete pores and reacts with cement hydration products. Curing temperature and humidity are only roughly controlled. Carbonation detection relies on phenolphthalein solution spraying to observe carbonation depth or using a thermogravimetric analyzer to test samples offline. This approach has problems such as limited carbonation depth, low carbonation efficiency, uneven curing conditions, long testing cycles, and no real-time feedback.

[0035] Among them, the nano-calcium carbonate seed crystals have a particle size of 30–80 nm. The particles are used to provide nucleation sites for calcium carbonate crystal growth in the pore liquid phase of concrete, accelerate the carbonation process, and promote the uniform distribution of calcium carbonate crystals.

[0036] The pore regulator is a compound of polycarboxylate superplasticizer and expansion agent, used to regulate the pore size distribution and connectivity during the setting and hardening process of concrete, concentrating the pore size in the range of 50–500 nm to facilitate… Gas diffusion and permeation.

[0037] The vacuum pretreatment device includes a vacuum pump and a sealed curing chamber. It uses negative pressure suction to remove air and water vapor from the concrete's pore network, preparing it for subsequent treatment. Injection creates low-resistance diffusion channels, avoiding mass transfer limitations caused by air blockage.

[0038] The mechanism of ultrasonic-assisted gas dissolution and diffusion is that the cavitation and acoustic flow effects generated by ultrasound in the porous liquid phase disrupt the mass transfer resistance at the gas-liquid interface, thus... The diffusion coefficient of molecules in the liquid phase is improved.

[0039] The distributed temperature sensor array consists of fiber optic temperature sensors arranged on the surface, middle layer and core area of ​​the concrete component, with a sensor spacing of no more than 100mm, used to construct the three-dimensional temperature field distribution state inside the concrete.

[0040] The temperature field data includes real-time temperature values, temperature change rates, and temperature distribution uniformity indices measured by each fiber optic temperature sensor in the distributed temperature sensor array.

[0041] Among them, the carbonization temperature control optimization model is a temperature field prediction and control model based on neural networks. It outputs the optimal curing temperature setting value for each region to maximize the carbonization reaction rate and achieve uniform temperature distribution based on the input temperature field data, carbon dioxide injection pressure, curing time, and mineral admixture proportion parameters.

[0042] The circulating cooling water flow rate is adjustable from 5 to 20 L / min, the atomized particle diameter of the spray humidification system is controlled from 10 to 50 μm, and the spray pressure is 0.2 to 0.5 MPa. The two work together to achieve precise control of temperature and humidity in the maintenance environment.

[0043] Among them, the spectral resolution of the near-infrared spectroscopy detection device is not less than 5nm, and the single scan time is not more than 30 seconds. The carbon fixation amount is evaluated by collecting the intensity changes of characteristic absorption peaks of calcium carbonate in the 1400-1500nm and 2300-2350nm bands in concrete.

[0044] Among them, the rapid carbon sequestration assessment model is a spectral data processing model based on convolutional neural networks. It takes spectral data in the 1200-2200nm band collected by a near-infrared spectral detection device as input and outputs the corresponding current carbon sequestration value.

[0045] The current carbon sequestration value is calculated as the carbon content per unit mass of concrete. Fixed form Mass, in kg / t.

[0046] The target carbon sequestration amount is a pre-set target value for carbon sequestration based on the design requirements of concrete components and the theoretical carbonization potential of mineral admixtures, and the unit is kg / t.

[0047] The deviation is calculated as a percentage by dividing the absolute value of the difference between the current carbon sequestration value and the target carbon sequestration value by the target carbon sequestration value and then multiplying by 100%.

[0048] The post-treatment stage includes steps such as stopping the carbon dioxide supply, slowly reducing the pressure to atmospheric pressure, and maintaining wet curing for 48 to 72 hours, which are used to stabilize carbonation products and prevent shrinkage cracks from forming on the concrete surface due to rapid water loss.

[0049] The carbonization temperature control optimization model is a feedforward network consisting of an input layer, three hidden layers, and an output layer. The input layer contains 28 neurons representing the temperature values ​​at each measurement point in the temperature field data, carbon dioxide injection pressure, curing time, and mineral admixture ratio parameters. The first hidden layer has 64 neurons using a modified linear unit activation function (MRU), the second hidden layer has 128 neurons using an exponential linear unit activation function (EXU), and the third hidden layer has 64 neurons using a hyperbolic tangent activation function. The output layer contains 12 neurons representing the optimal curing temperature setpoints for each region. The carbonization temperature control optimization model utilizes a sparse connection learning framework based on dynamic topology reconstruction. After each training batch, the average activation intensity and frequency of each neuron in that batch are calculated. The contribution of the connection weights between adjacent neurons to the gradient of the loss function is calculated. Connections with a contribution below a dynamic threshold are pruned and deleted with a probability of 0.05–0.15. Simultaneously, new connections are randomly added between highly activated neurons with a probability of 0.02–0.08. The retained connection weights are grouped according to the output variance of their respective neurons and then applied. Norm regularization constraints are applied to re-evaluate connection importance and update the network topology every 5 training epochs.

[0050] The implementation of the sparse connection learning framework based on dynamic topology reconstruction is as follows: During the forward propagation phase of each batch in the training process, the activation values ​​of all neurons are recorded. The average activation intensity of each neuron in the current batch is calculated as the arithmetic mean of the activation values ​​of all samples of that neuron. The activation frequency is the proportion of samples whose activation values ​​exceed a threshold of 0.1 out of the total number of samples in the batch. During the backpropagation phase, the gradient of the loss function with respect to each connection weight is calculated. The absolute value of the gradient is multiplied by the absolute value of the corresponding connection weight to obtain the contribution of that connection. A dynamic threshold is set to the 20th percentile of the current contribution of all connections. For connections with a contribution value lower than the dynamic threshold, a uniformly distributed random number between 0 and 1 is generated. When the random number is less than 0.05 to 0.15, the connection is deleted. Neurons with an average activation intensity exceeding 0.5 and an activation frequency exceeding 0.7 are identified as highly activated neurons. A uniformly distributed random number between 0 and 1 is generated between highly activated neurons in different layers. When the random number is less than 0.02 to 0.08 and there is no connection between the two neurons, a new connection is established and the weights are initialized to random values ​​following a standard normal distribution. All connections are grouped according to their starting neurons, and a weight vector is applied to the weight vector of each group of connections. Norm regularization, the The norm is calculated as the weight vector of each group. The sum of norms and the regularization coefficient are set to 0.001. After every 5 training cycles, the above contribution calculation and connection addition / removal process are repeated to continuously optimize the network topology until the training ends.

[0051] The sparse connection learning framework based on dynamic topology reconstruction continuously monitors the activation statistics of neurons and the influence of connection weights on the objective function during training. It automatically identifies and removes redundant connections that contribute little to the predictive ability of the carbonization temperature control optimization model. Simultaneously, it explores and establishes new connection paths during training to compensate for information loss caused by pruning. This framework significantly reduces the number of parameters and computational complexity of the carbonization temperature control optimization model while maintaining temperature field prediction accuracy. It also significantly improves the real-time response speed of the carbonization temperature control optimization model in industrial production environments. The dynamically adjusted network topology better adapts to the differences in temperature field evolution under different mineral admixture ratios and curing conditions, avoiding overfitting and enhancing carbonization temperature control. The optimization model's generalization ability to new working conditions is improved. The grouping regularization constraint of the connection ensures that the network maintains the integrity of the key feature extraction path during the sparsification process. This enables the carbonation temperature control optimization model to accurately capture the nonlinear mapping relationship between the temperature field and carbon fixation efficiency in the concrete carbonation reaction. This provides a reliable decision basis for achieving precise temperature control in the carbonation process. Compared with traditional neural networks with fixed topology, the framework shortens the model inference time under the condition of similar temperature prediction error. This allows the system to quickly respond to changes in the temperature field during the curing process and adjust cooling and humidification parameters in a timely manner, avoiding local overheating or moisture loss due to control lag. This ensures uniform carbonation reaction across the entire concrete cross section and improves the stability of total carbon fixation and product quality.

[0052] The steps for establishing the training dataset for the carbonation temperature control optimization model include: preparing 200 sets of concrete specimens with different mineral admixture proportions under laboratory conditions, wherein the mineral admixture proportions cover a range of 20-50% fly ash by mass, 15-45% slag powder by mass, and 3-12% silica fume by mass; during the curing process of each set of concrete specimens, a distributed temperature sensor array is deployed and temperature field data is recorded at 1-minute intervals, while the corresponding carbon dioxide injection pressure and curing time parameters are also recorded; after curing, the carbonation depth and carbon fixation amount of each set of concrete specimens are measured, and the optimal curing temperature setpoints for each region to achieve the optimal carbonation effect are calculated in reverse based on the carbonation depth and carbon fixation amount as label samples; the temperature field data, carbon dioxide injection pressure, curing time, and mineral admixture proportion parameters are used as input samples, and the optimal curing temperature setpoints for each region are used as label samples; after normalization of the input samples and label samples, they are divided into training set, validation set, and test set in a ratio of 8:1:1.

[0053] The training steps of the carbonization temperature control optimization model include: using an adaptive moment estimation optimization algorithm for gradient descent, setting the initial learning rate to 0.001 and dynamically adjusting it according to the cosine annealing strategy, setting the batch size to 32, and the total number of training cycles to 300; defining the loss function as the weighted sum of mean squared error loss and connection sparsity penalty term, wherein the mean squared error loss is the sum of the squares of the differences between the predicted optimal maintenance temperature setpoints for each region and the actual optimal maintenance temperature setpoints for each region divided by the number of samples, and the connection sparsity penalty term is the number of non-zero connections in the network divided by the total number of possible connections and then multiplied by the weight coefficient 0.01; monitoring the change of the loss function value on the validation set, and terminating the training early when the value of the loss function on the validation set does not decrease within 20 consecutive training cycles, and saving the model weight parameters that perform best on the validation set as the final carbonization temperature control optimization model.

[0054] The structure of the rapid carbon fixation assessment model is a spectral data processing network consisting of convolutional layers, pooling layers, and fully connected layers. The input layer receives spectral reflectance data at 1024 wavelength points. The first convolutional layer uses 32 convolutional kernels of size 7 to extract local features of the spectral curve. After dimensionality reduction by a max pooling layer, the data is fed into the second convolutional layer. The second convolutional layer uses 64 convolutional kernels of size 5 to further extract abstract features. After flattening, the data is connected to a fully connected layer containing 128 neurons. The final output layer is a single neuron used for regression prediction of the current carbon fixation value. The rapid carbon fixation assessment model also utilizes a sparse connection learning framework based on dynamic topology reconstruction to optimize the connection structure of the fully connected layer. During training, the connection topology is dynamically adjusted according to the correlation strength between spectral features and the current carbon fixation value, pruning connections that have little impact on the prediction results and retaining key feature transmission paths.

[0055] The steps for establishing the training dataset for the rapid carbon sequestration assessment model include: collecting 1500 concrete samples with different carbonization degrees; simultaneously performing near-infrared spectroscopy scanning and thermogravimetric analysis on each concrete sample; using the 1200–2200 nm spectral data obtained from near-infrared spectroscopy scanning as input samples; and using the accurate carbon sequestration amount measured by thermogravimetric analysis as label samples; adding Gaussian noise of different intensities and baseline drift to the spectral data using data augmentation techniques to improve the robustness of the rapid carbon sequestration assessment model, wherein the standard deviation of the Gaussian noise ranges from 0.001 to 0.005, and the baseline drift ranges from -0.02 to 0.02; and dividing the dataset into training, validation, and test sets in a ratio of 7:2:1.

[0056] The training steps of the rapid carbon sequestration assessment model include: updating parameters using the root mean square propagation optimization algorithm, setting the learning rate to 0.0005, the batch size to 16, and training for 200 epochs; using mean square error as the loss function, where the mean square error is the sum of the squares of the differences between the predicted current carbon sequestration value and the actual current carbon sequestration value divided by the number of samples; evaluating the prediction accuracy of the rapid carbon sequestration assessment model on the validation set and selecting the model with the smallest validation error as the deployment version; and ensuring the stability and reliability of the rapid carbon sequestration assessment model through 10-fold cross-validation.

[0057] The specific implementation methods of the above steps are described in detail below.

[0058] The specific implementation of step S01 is as follows: First, weigh 30-45% fly ash, 25-40% slag powder, 5-10% silica fume, and cement according to the mass percentage. Add the above mineral admixtures and cement to a forced mixer for dry mixing. The dry mixing time is controlled at 3-5 minutes to ensure uniform dispersion of each component. Then, add mixing water according to a water-cement ratio of 0.35-0.45 and start the mixer for wet mixing. During wet mixing, add nano-sized particles (30-80 nm) to the slurry. Seed crystals are added at a rate of 0.5–1.5% of the total mass of the cementitious material. Simultaneously, a pore conditioner, a mixture of polycarboxylate superplasticizer and expanding agent in a 3:1 mass ratio, is added at a rate of 2–4% of the total mass of the cementitious material. Stirring continues for 8–12 minutes until the slurry achieves a slump of 180–220 mm. The purpose of this step is to prepare a basic concrete slurry with optimized pore structure and nucleation sites. The addition of seed crystals, based on the heterogeneous nucleation theory, can significantly reduce the energy barrier for calcium carbonate crystal formation. Pore regulators, by controlling the growth of hydration products and the micro-expansion effect, can adjust the pore size to the range of 50–500 nm, facilitating subsequent... Diffusion provides an efficient mass transfer channel.

[0059] The specific implementation of step S02 involves injecting the foundation concrete slurry into a pre-prepared steel mold and vibrating it to remove air. The vibration time is controlled at 30-60 seconds to avoid segregation. Then, the mold containing the foundation concrete slurry is transferred to a sealed curing chamber, and the chamber door is closed. The vacuum pump system connected to the curing chamber is started, and the pressure inside the curing chamber is gradually reduced to a negative pressure state of -0.08 to -0.12 MPa through vacuum pump suction. The suction process adopts a staged pressure reduction method. First, the pressure is reduced to -0.04 MPa within 5 minutes, and then further reduced to the target negative pressure value within 10 minutes, maintaining the target negative pressure state for 15-25 minutes. During the negative pressure maintenance stage, the air and water vapor in the concrete pores are forcibly extracted under the pressure difference. The purpose of this step is to remove gas phase obstacles in the pore network. Injection creates a low-resistance diffusion path, and vacuum pretreatment is based on the principle of pressure difference mass transfer in gas dynamics. The negative pressure environment reduces the partial pressure of gas in the pores, allowing it to flow outward along the pressure gradient. Staged depressurization can avoid damage to the pore structure caused by sudden depressurization.

[0060] The specific implementation of step S03 is to immediately turn off the vacuum pump and turn on the vacuum pump after the vacuum pretreatment is completed. The gas source valve supplies liquefied gas with a purity of not less than 99.5%. After being converted into a gaseous state by the vaporizer, the gas is injected into the curing chamber through the inlet pipe. The initial injection pressure is set to 0.3 MPa and gradually increased to a working pressure of 0.4–0.6 MPa within 5–8 minutes. Simultaneously, the ultrasonic generator installed on the wall of the curing chamber is started. The output frequency of the ultrasonic generator is set to the range of 20–40 kHz, and the sound intensity is controlled at 1–3 kHz. When ultrasound propagates in the pore liquid phase of concrete, the cavitation bubbles generated form microjets and shock waves during the periodic compression and expansion process. These microjets and shock waves disrupt... The mass transfer boundary layer at the gas-liquid interface enhances molecular perturbation, making... The dissolution rate and diffusion coefficient in porous solutions are significantly improved; the purpose of this step is to establish high concentrations. The environment, through physical enhancement methods, promotes the deep penetration of gas into the interior of concrete, and pressurized injection increases the penetration of gas based on Henry's Law. Solubility in the liquid phase, ultrasound-assisted mass transfer enhancement mechanism based on acoustic cavitation and acoustic flow effects.

[0061] The specific implementation of step S04 involves real-time acquisition of temperature data using fiber optic temperature sensors pre-embedded at different depths and locations within the concrete component. These sensors operate based on the fiber Bragg grating principle, and their sensitivity to temperature variations in reflected wavelength is 10 pm / ℃. The data acquisition system reads the temperature values ​​from each sensor once per second and simultaneously records the current temperature reading. The injection pressure reading and the curing time from the start of curing, along with the aforementioned temperature field data, The injection pressure, curing time, and pre-stored mineral admixture proportion parameters form a 28-dimensional input vector. After normalization, the input vector is input to the input layer of the carbonation temperature control optimization model. After forward propagation calculation, the model obtains 12 values ​​at the output layer, corresponding to the optimal curing temperature settings for four regions on the surface, four regions in the middle layer, and four regions in the core layer of the concrete component. The purpose of this step is to use the nonlinear mapping capability of the neural network to predict the temperature control target that maximizes the carbonation reaction rate and achieves uniform temperature distribution based on real-time monitored multi-source data. The carbonation temperature control optimization model obtains the implicit relationship between the temperature field evolution law and carbon fixation efficiency based on a large amount of historical curing data.

[0062] The specific implementation of step S05 involves transmitting the optimal curing temperature setpoints for each region output by the carbonization temperature control optimization model to the temperature control execution system. The temperature control execution system first calculates the deviation between the current temperature and the optimal curing temperature setpoint for each region. For overheated regions with positive temperature deviations, the cooling water flow is increased by adjusting the opening of the electromagnetic proportional valve in the corresponding region's circulating cooling pipe. The cooling water flow adjustment uses a proportional-integral-derivative (PID) control algorithm, with the proportional coefficient set to 0.8, the integral time constant set to 120 seconds, and the derivative time constant set to 30 seconds. For underheated regions with negative temperature deviations, the cooling water flow is reduced or the cooling water supply is shut off. To achieve heat preservation, the spray humidification system is adjusted based on humidity values ​​fed back from the pore relative humidity sensor. When the humidity is below 75%, the high-pressure spray pump is activated to spray water mist with a particle diameter of 10–50 μm into the curing chamber, with the spray pressure controlled within the range of 0.2–0.5 MPa. When the humidity exceeds 90%, the spraying is stopped and micro-ventilation is initiated. The purpose of these steps is to precisely maintain the curing temperature at 40–55℃ and the humidity at 75–90% within the optimal carbonization reaction range through closed-loop feedback control. This coordinated temperature and humidity control is based on the principles of carbonization reaction kinetics; suitable temperature accelerates ion diffusion and chemical reactions, while sufficient humidity ensures… The liquid phase environment required for dissolution and ion migration.

[0063] The specific implementation of step S06 is to pause every 2 to 4 hours after the carbonization curing begins. The curing chamber door is opened, and the operator uses a handheld near-infrared spectroscopy device to perform non-contact scanning of marked measuring points on the concrete surface. Each measuring point scan is completed within 30 seconds. Simultaneously, a handheld core drill is used to drill a 10mm diameter, 50mm deep core sample at a preset location, and the core sample end face is spectrally scanned. The near-infrared spectroscopy device emits near-infrared light in the 1200–2200nm wavelength band and collects the reflectance spectrum. After baseline correction and smoothing filtering preprocessing, the spectral data forms a spectral reflectance vector of 1024 wavelength points. This spectral reflectance vector is input into a rapid carbon fixation assessment model, which extracts the carbon fixation content through a convolutional layer. The intensity and peak shape parameters of the characteristic absorption peaks at 1400–1500 nm and 2300–2350 nm are calculated by the fully connected layer, and the current carbon fixation amount is output. The purpose of this step is to achieve rapid, non-destructive, or minimal-destructive detection of the carbon fixation process, providing real-time feedback for process adjustment. Near-infrared spectroscopy detection is based on the principle of molecular vibrational spectroscopy. The overtone and combination vibrations of the carbon-oxygen bond produce characteristic absorption in the near-infrared region, with absorption peak intensity similar to... The content showed a positive correlation.

[0064] The specific implementation of step S07 involves comparing the current carbon sequestration value output by the rapid carbon sequestration assessment model with the target carbon sequestration value pre-set according to the concrete design requirements. The absolute value of the difference is calculated, divided by the target carbon sequestration value, and then multiplied by 100% to obtain the percentage deviation. When the deviation exceeds 8%, it is determined that the carbon sequestration progress is lagging and reinforcement measures need to be taken. These reinforcement measures include extending the carbonation curing time by 2-4 hours each time, or... Increase the injection pressure by 0.05–0.1 MPa. If the deviation is between 3% and 8%, maintain the current process parameters and continue curing. If the deviation is less than 3%, it is determined that the carbon fixation target has been reached, and the termination procedure is immediately executed. The termination procedure includes shutting down... The air source valve is activated, and the slow pressure relief valve is started to reduce the pressure of the curing chamber to atmospheric pressure at a rate of 0.05 MPa per minute. The curing chamber is kept sealed and spray humidification continues for 48 to 72 hours. The purpose of these steps is to dynamically optimize process parameters based on real-time carbon fixation feedback to ensure that the carbon fixation target is achieved and to avoid excessive carbonization. The threshold setting is based on a large number of experimental statistical results. The upper threshold of 8% corresponds to the critical state where the carbon fixation efficiency is obviously insufficient and intervention is required. The lower threshold of 3% corresponds to the state where the carbon fixation reaction is basically completed and the benefits of continued curing are decreasing.

[0065] It should be noted that one of the key technical concepts of this invention is vacuum pretreatment and pressurization. The alternating cycle injection process, under traditional atmospheric pressure carbonization method Entering concrete pores requires overcoming the resistance of the existing air within the pores. The diffusion path is tortuous and the mass transfer resistance is high, resulting in gas only being able to penetrate to a depth of a few millimeters on the surface. This invention uses vacuum pretreatment to forcibly expel air and water vapor from the pores, creating a negative pressure environment, followed by immediate injection of high pressure. Using pressure differential as a driving force, gas is rapidly filled into the entire pore network under pressurized conditions. Solubility in porous solutions is significantly increased according to Henry's law, and ultrasonic cavitation further disrupts the mass transfer resistance at the gas-liquid interface. The synergistic effect of these three mechanisms leads to… It can penetrate deep into the core area of ​​large-sized components, achieving uniform carbonization across the entire cross section rather than just surface carbonization, thus significantly improving the overall carbon fixation capacity and efficiency of concrete.

[0066] The second key technical idea of ​​this invention is to achieve precise control of the temperature field using a carbonization temperature control optimization model based on a neural network. The carbonization reaction is exothermic, and different regions exhibit varying degrees of temperature control due to... Differences in concentration and degree of hydration lead to uneven reaction rates. Traditional, crude temperature control methods cannot cope with localized temperature fluctuations. Localized overheating accelerates water evaporation, hindering pore drying. Dissolution forms carbonation dead zones. This invention utilizes a distributed temperature sensor array to construct a three-dimensional temperature field and inputs it into a carbonation temperature control optimization model. By learning the correlation between temperature distribution and carbon fixation effect from a large amount of historical data, the model can predict the optimal temperature setpoints for each region that maximize the carbonation reaction rate and ensure uniform temperature distribution. The temperature control execution system dynamically adjusts the circulating cooling water flow rate and spray humidification parameters based on the model output, precisely maintaining the temperature within the optimal reaction window of 40–55°C. This prevents local overheating that leads to moisture loss and pore drying, ensuring the uniformity and continuity of the carbonation reaction across the entire concrete cross-section.

[0067] The third key technical approach of this invention is to achieve real-time quality monitoring through a rapid carbon fixation assessment model based on near-infrared spectroscopy and machine learning. Traditional thermogravimetric analysis requires sampling, crushing, grinding, and high-temperature heating for testing, a process that takes several hours and is destructive. The phenolphthalein indicator method can only semi-quantitatively determine the carbonization depth through color changes and cannot accurately measure the carbon fixation amount. This invention utilizes near-infrared spectroscopy detection technology to... The quantitative relationship between the absorption peak intensity of molecular vibrations in characteristic bands and their content is established by automatically extracting spectral features through convolutional neural networks and establishing a nonlinear mapping model with carbon fixation amount. The detection process is completed within 30 seconds and is non-destructive to concrete components or requires only a small amount of core sampling. High-frequency detection every 2 to 4 hours provides timely feedback for dynamic adjustment of process parameters. When the carbon fixation progress is lagging, strengthening measures such as extending the time or increasing the pressure can be taken immediately. When the target carbon fixation amount is reached, the curing is terminated in time to avoid energy waste and excessive carbonization, thus realizing closed-loop quality control of the carbon fixation production process.

[0068] The synergistic effect of the three key technological approaches mentioned above lies in constructing a system from... A comprehensive optimization system encompassing efficient mass transfer, precise temperature and humidity control during carbonization, and real-time feedback on carbon fixation quality has been established. This system utilizes a vacuum pressurized alternating cycle process to solve [the following issues]. The fundamental problem of limited mass transfer prevents gases from penetrating deeply. A neural network temperature control model ensures deep penetration by maintaining optimal reaction temperature and sufficient humidity. The continuous reaction and transformation, and the real-time monitoring of the rapid spectral evaluation model enable the process parameters to be dynamically adjusted according to the progress of carbon fixation. The three elements form a positive cycle of mass transfer enhancement to promote reaction, temperature control optimization to ensure uniformity, and real-time monitoring to guide adjustment. Compared with the traditional process that relies solely on the open-loop control method of fixed parameters, this invention significantly improves the stability of carbon fixation depth, total carbon fixation, and product quality through multi-dimensional synergistic optimization, laying a solid foundation for the industrial application of concrete carbon fixation technology.

[0069] It should be noted that this invention also solves the following technical problems: Traditional concrete carbon fixation production processes rely on phenolphthalein solution spraying to observe carbonation depth or offline testing of samples using a thermogravimetric analyzer. These methods suffer from long testing cycles and lack of real-time feedback, making it impossible to adjust curing parameters promptly according to the carbonation process, thus affecting carbon fixation efficiency and product quality stability. This invention, by employing a near-infrared spectroscopy detection device to scan the concrete surface and drilled sampling points every 2 to 4 hours, acquires spectral data in the 1200 to 2200 nanometer band and inputs it into a rapid carbon fixation assessment model based on a convolutional neural network for analysis. This achieves rapid, non-destructive detection of carbon fixation, with a single scan taking no more than 30 seconds to obtain the current carbon fixation value. This model learns the mapping relationship between near-infrared spectral data of 1500 concrete samples with different carbonation levels and accurate carbon fixation measured by thermogravimetric analysis. It can accurately assess the carbon fixation based on the intensity changes of the characteristic absorption peaks of calcium carbonate in a certain wavelength band. Combined with a sparse connection learning framework based on dynamic topology reconstruction, it optimizes the connection structure of the fully connected layer, improving computational efficiency while ensuring prediction accuracy. This allows the system to adjust the carbonation curing time or carbon dioxide injection pressure in real time according to the deviation between the current carbon fixation value and the target carbon fixation value, thus realizing closed-loop feedback control of the carbonation process.

[0070] Specifically, the principle of this invention is as follows: The fundamental reason why this invention can solve this technical problem lies in establishing a closed-loop control system from temperature field sensing to curing parameter regulation. The distributed temperature sensor array overcomes the limitations of traditional single-point temperature measurement. By arranging fiber optic temperature sensors with a spacing of no more than 100 mm in the surface, middle, and core areas of concrete components, a complete representation of the three-dimensional temperature field distribution within the concrete is constructed, providing a data foundation for precise control. The carbonation temperature control optimization model, based on a neural network architecture, establishes a nonlinear mapping relationship between temperature field data and the optimal curing temperature setpoint. By learning from the curing process data of 200 sets of concrete test blocks with different mix proportions, the intrinsic law of carbonation reaction rate and temperature distribution is mastered, enabling the output of control strategies that maximize the carbonation reaction rate and ensure uniform temperature distribution for specific working conditions. The coordinated regulation mechanism of the circulating cooling water flow and the spray humidification system transforms the abstract temperature setpoint output by the model into specific physical operations, achieving precise control of the temperature and humidity of the curing environment, ensuring uniform carbonation reaction across the entire concrete cross-section, thereby improving the total carbon sequestration and product quality stability.

[0071] The following provides a specific embodiment 1 of the present invention. The specific implementation of steps S01-S03 in this embodiment 1 is the same as that described above, and will not be repeated in detail here. The specific implementation of other steps is described in detail below.

[0072] The specific implementation of step S04 is as follows: The distributed temperature sensor array consists of several fiber optic temperature sensors, numbered as follows: ,in The total number of sensors is [number], with a spacing of no more than 100mm between them. They are arranged in the surface, middle, and core areas of the concrete component. Each sensor records the temperature value at a sampling interval of 1 minute. ,in Maintenance time, in minutes. The unit is ℃, and the temperature field data matrix is ​​represented as follows: The rate of temperature change is calculated as follows: ,in This is the time interval, with a default value of 1 minute, and the unit is min. The unit is ℃ / min, and the temperature distribution uniformity index is calculated as follows: ,in The average temperature is expressed in °C. The unit is ℃, and the temperature field data matrix is... Temperature change rate vector Temperature distribution uniformity index Carbon dioxide injection pressure Maintenance time Mineral admixture proportion parameter vector The carbonization temperature control optimization model is fed in as input, where The unit is MPa. The percentage of fly ash by mass is expressed in %. This represents the percentage of slag powder by mass, in %. The output of the carbonization temperature control optimization model is a vector of optimal curing temperature setpoints for each region, representing the percentage of silica fume by mass. ,in This specifies the number of temperature control zones, typically set to 12. For area code, The unit is ℃.

[0073] The specific implementation of step S05 is as follows: based on the optimal curing temperature setpoints for each region output by the carbonization temperature control optimization model. The circulating cooling water flow rate in the corresponding area is adjusted by a proportional-integral-derivative controller. The operating parameters of the spray humidification system are as follows: the circulating cooling water flow rate is adjustable from 5 to 20 L / min, and the relationship between flow rate and temperature deviation is expressed as follows: ,in The baseline flow rate is empirically estimated at 10 L / min. For the first The deviation between the actual temperature of the area and the set value, in °C. For the first The actual temperature of the area, in °C. To control the time window, the unit is minutes, and the default is 10 minutes. This is the proportionality constant, with a default value of 0.8, and its dimension is L / (min·℃). This is the integral coefficient, with a default value of 0.15 and dimensions of L / (°C·m·K). ), This is the differential coefficient, with a default value of 0.05 and dimensions in L / ℃. The unit is L / min. The atomized particle diameter of the spray humidification system is controlled between 10 and 50 μm, the spray pressure is 0.2 to 0.5 MPa, and the formula for adjusting the pore relative humidity is: ,in For the first Area spray flow rate, in L / h. The baseline spray flow rate is empirically estimated at 5 L / h. For the first The deviation between the actual relative humidity and the target relative humidity in the region, expressed in %. For the first Current relative humidity in the area, in %. The target relative humidity ranges from 75% to 90%, and the unit is %. The humidity adjustment coefficient is set to 0.3 by default, with dimensions L / (h·%), to precisely maintain the curing temperature in the range of 40–55℃ and the relative humidity of the pores in the range of 75–90%.

[0074] The specific implementation of step S06 is as follows: During the carbonization curing process, the near-infrared spectroscopy detection device scans the concrete surface and borehole sampling points every 2-4 hours, with a spectral resolution of not less than 5 nm and a single scan time not exceeding 30 seconds, acquiring spectral reflectance data in the wavelength range of 1200-2200 nm. The number of sampling wavelength points is 1024, and the wavelength point index is... , No. The reflectance corresponding to each wavelength point is... Spectral data vector representation is , spectral data vector The input is processed by the rapid carbon sequestration assessment model, which then outputs the current carbon sequestration value. The unit is kg / t, and the intensity of the characteristic absorption peaks of calcium carbonate in the 1400-1500nm and 2300-2350nm bands is positively correlated with the carbon fixation amount.

[0075] The specific implementation of step S07 is as follows: the current carbon sequestration value is calculated based on the rapid carbon sequestration assessment model. Compared with the preset target carbon sequestration amount The formula for calculating the degree of deviation is as follows: ,in The unit is kg / t. For dimensionless percentages, when If the current carbon sequestration level is deemed insufficient, extend the carbonization curing time or increase the carbon dioxide injection pressure by 0.05–0.1 MPa, and extend the time by 10–20% of the original plan. Once the current carbon sequestration level is determined to be within the acceptable range, carbonization curing is terminated and the post-treatment stage begins. The post-treatment stage includes stopping the supply of carbon dioxide, slowly reducing the pressure to atmospheric pressure at a rate of 0.01–0.02 MPa per minute, and maintaining wet curing for 48–72 hours.

[0076] It should be explained that the input layer of the carbonization temperature control optimization model contains 28 neurons, and the input vector is represented as follows: The first hidden layer contains 64 neurons and uses a modified linear unit activation function. ,in The input values ​​for the neurons are used; the second hidden layer contains 128 neurons and employs the exponential linear unit activation function. ,in This is a hyperparameter, with a default value of 1.0. To control the slope of the negative regions, the third hidden layer contains 64 neurons and uses the hyperbolic tangent activation function. The output layer contains 12 neurons, and the output vector is .

[0077] In the sparse connection learning framework based on dynamic topology reconstruction, neurons In the The average activation intensity of the batch was calculated as follows: ,in Number the neurons. For batch number, For the first The number of samples in a batch. For sample number, For neurons In the The activation value of each sample, and the activation frequency are calculated as follows: ,in This is an indicator function; it takes a value of 1 when the condition inside the parentheses is true and a value of 0 when the condition is false. The connection weights are also specified. The contribution is calculated as follows ,in For loss function, The gradient of the loss function with respect to the connection weights. Indicates from neurons To neurons The connection weights, The target neuron is numbered, and the dynamic threshold is set to the 20th percentile of the contribution of all current connections. Uniformly distributed random numbers are generated for connections whose contribution is lower than the dynamic threshold. ,in Representing an interval Uniform distribution on The random number used for pruning decisions, when Delete the connection at that time. The pruning probability ranges from 0.05 to 0.15. For highly activated neurons with an average activation intensity exceeding 0.5 and an activation frequency exceeding 0.7, uniformly distributed random numbers are generated among highly activated neurons in adjacent layers. ,in For adding a random number to the connection check, when Furthermore, new connections are established when no existing connections exist between the two neurons. To add a probability, the value ranges from 0.02 to 0.08, and the new connection weights are initialized to... ,in This represents a standard normal distribution with a mean of 0 and a standard deviation of 1, categorized by the initial neuron. After grouping, the weight vector Apply Norm regularization, where For neurons The number of output connections, the regularization term is calculated as follows: ,in The total number of neurons is 0.001, and the regularization coefficient is 0.001. The importance of connections is re-evaluated and the network topology is updated every 5 training cycles.

[0078] The steps for establishing the training dataset for the carbonation temperature control optimization model include: preparing 200 groups of concrete specimens with different mineral admixture proportions under laboratory conditions. The mineral admixture proportions cover a range of 20-50% fly ash by mass, 15-45% slag powder by mass, and 3-12% silica fume by mass. During the curing process of each group of concrete specimens, a distributed temperature sensor array is deployed, recording temperature field data at 1-minute intervals. Simultaneously, the corresponding carbon dioxide injection pressure and curing time parameters are recorded. After curing, the carbonation depth and carbon fixation amount of each group of concrete specimens are measured, and the inverse calculation formula is as follows: ,in For the first Optimal maintenance temperature label for the region, in °C. For the first Measured regional curing temperature, in °C. This represents the maximum carbon fixation amount measured in the experimental group, expressed in kg / t. The measured carbon fixation amount of the current experimental group is expressed in kg / t. The optimal curing temperature setpoints for each region to achieve the best carbonization effect are calculated in reverse based on the carbonization depth and carbon fixation amount and used as label samples. Temperature field data, carbon dioxide injection pressure, curing time, and mineral admixture mix proportion parameters are used as input samples. The optimal curing temperature setpoints for each region are used as label samples. After normalization of the input samples and label samples, they are divided into training set, validation set and test set in a ratio of 8:1:1.

[0079] The training steps for the carbonization temperature control optimization model include: using an adaptive moment estimation optimization algorithm for gradient descent, setting the initial learning rate to 0.001 and dynamically adjusting it according to a cosine annealing strategy, setting the batch size to 32, and the total number of training epochs to 300. The loss function expression is as follows: ,in For the sample size, For the first The first sample The optimal maintenance temperature setpoint predicted for the region, in °C. For the first The first sample The actual optimal maintenance temperature setting for the region, in °C. This is the temperature normalization reference value, with an empirical value of 50℃. The number of non-zero connections in the network. To determine the total number of possible connections, the loss function is defined as the weighted sum of the mean squared error loss and the connection sparsity penalty term. The change in the value of the loss function is monitored on the validation set. Training is terminated early when the value of the loss function on the validation set does not decrease within 20 consecutive training cycles. The model weight parameters that perform best on the validation set are saved as the final carbonization temperature control optimization model.

[0080] The input layer of the rapid carbon sequestration assessment model receives spectral reflectance data at 1024 wavelength points. The first convolutional layer uses 32 kernels of size 7 for convolution operations, and the second convolutional layer uses 64 kernels of size 5. After being flattened, they are connected to a fully connected layer containing 128 neurons. The output layer outputs the current carbon fixation value for a single neuron. .

[0081] The steps for establishing the training dataset for the rapid carbon sequestration assessment model include: collecting 1500 concrete samples with different carbonization degrees; simultaneously performing near-infrared spectroscopy scanning and thermogravimetric analysis on each concrete sample; using the 1200–2200 nm spectral data obtained from near-infrared spectroscopy scanning as input samples; using the accurate carbon sequestration amount measured by thermogravimetric analysis as label samples; and adding Gaussian noise of different intensities and baseline drift to the spectral data using data augmentation techniques to improve the robustness of the rapid carbon sequestration assessment model. The standard deviation of the Gaussian noise ranges from 0.001 to 0.005, and the baseline drift ranges from -0.02 to 0.02. The dataset is divided into training, validation, and test sets in a ratio of 7:2:1.

[0082] The training steps for the rapid carbon sequestration assessment model include: updating parameters using the root mean square propagation optimization algorithm, setting the learning rate to 0.0005, the batch size to 16, training for 200 epochs, and using the loss function expression as follows: ,in For the sample size, For the first The predicted current carbon sequestration value for each sample, in kg / t. For the first The actual current carbon sequestration value for each sample, in kg / t. The normalized baseline value for carbon sequestration is 20 kg / t, with the mean squared error used as the loss function. The prediction accuracy of the rapid carbon sequestration assessment model is evaluated on the validation set, and the model with the smallest validation error is selected as the deployment version. Ten-fold cross-validation is used to ensure the stability and reliability of the rapid carbon sequestration assessment model.

[0083] To better understand and implement this invention, a specific application scenario, Example 2, is provided below: The technical team first prepared a basic concrete slurry by mixing fly ash (38%), slag powder (32%), and silica fume (7%) with 525 ordinary Portland cement. During the mixing process, 8 kg of nano-calcium carbonate seed crystals with a particle size of approximately 55 nm were added to each cubic meter of concrete slurry, along with 6.5 kg of a pore conditioner composed of polycarboxylate superplasticizer and expansion agent in a 3:1 mass ratio. The mixing time was controlled at 180 seconds, and the rotation speed was maintained at 45 rpm to ensure that all components were fully and uniformly dispersed. The prepared basic concrete slurry was injected into the steel mold of the precast component, and the inner surface of the mold was pre-coated with a release agent for subsequent demolding operations.

[0084] like Figure 2As shown, the mold containing concrete slurry was hoisted into the sealed curing chamber, and the vacuum pretreatment device was activated to perform negative pressure suction treatment on the internal pores of the concrete. The technical team set the vacuum pump suction pressure to -0.10 MPa and continued suction for 20 minutes. During the suction process, the pressure gauge reading in the curing chamber gradually decreased from 101.3 kPa to 1.3 kPa, indicating that a large amount of air and water vapor in the concrete pore network was expelled, and the porosity was adjusted from the initial 12.3% to 15.8%, significantly improving pore connectivity. After the vacuum pretreatment was completed, the technical team immediately shut down the vacuum pump and switched to pressurization. The injection mode prevents outside air from re-entering the pores and causing blockage.

[0085] Industrial purity of 99.7% Gas is precisely metered from the storage tank via a pressure reducing valve and flow meter, and then injected into the curing chamber through multi-point distributed nozzles. The technical team will... The injection pressure is set to 0.45 MPa, and the flow rate is controlled at 280. .like Figure 3 As shown, in Simultaneously with the injection, an array of ultrasonic generators mounted on the walls of the curing chamber begins operation, outputting ultrasonic waves at a frequency of 30kHz and a power of 150W. The ultrasonic waves generate and collapse cavitation bubbles in the pore liquid phase of the concrete, disrupting the mass transfer resistance layer at the gas-liquid interface. The effective diffusion coefficient of molecules in the liquid phase increases from that under normal conditions. Upgraded to Significantly accelerated The process of penetrating deep into the concrete.

[0086] The technical team pre-embedded a distributed fiber optic temperature sensor array inside the concrete component. The sensors were positioned in the surface area (20mm from the component surface), the middle area (80mm from the surface), and the core area at the geometric center of the component. Four temperature measurement points were set in each area, totaling 12 temperature monitoring nodes, with an 85mm spacing between adjacent sensors. Figure 4 As shown, the sensor collects temperature data from each measuring point in real time at 1-minute intervals and transmits it to the carbonation temperature control optimization model for processing via the data acquisition module. In the early stages of carbonation curing, the internal structure of the concrete... The exothermic reaction of dissolution and carbonization causes the temperature to rise rapidly. The surface temperature rises from 23°C to 49°C within 15 minutes, the middle temperature rises to 46°C, and the core temperature reaches 52°C due to heat accumulation.

[0087] After receiving temperature field data, the carbonization temperature control optimization model combines it with the current... Information such as injection pressure (0.45 MPa), curing time (15 minutes), and mineral admixture proportions is used for inference calculations via a feedforward neural network. The 28 neurons in the model input layer correspond to 12 temperature measurement points, 12 temperature change rates, 1 temperature distribution uniformity index, and 1... The model incorporates a pressure value, a curing time value, and a blending ratio as comprehensive parameters. After a nonlinear transformation through three hidden layers, the output layer generates optimal curing temperature setpoints for 12 regions. Based on the model output, the technical team adjusts the flow distribution valves of the circulating cooling water system, reducing the cooling water flow rate in the surface region from the initial 8... Upgraded to 14 The middle layer area remains at 10 The core area has been adjusted to 12. Simultaneously, the spray humidification system was activated, with the atomized particle diameter controlled at 25μm and the spray pressure set at 0.35MPa, stabilizing the relative humidity inside the curing chamber at 82%. Through precise temperature and humidity control, the temperature of each area of ​​the concrete gradually adjusted to the set range within 30 minutes, with the surface area stabilizing at 47℃, the middle area at 49℃, and the core area at 51℃. The temperature distribution uniformity index decreased from the initial 6.8℃ to 4℃.

[0088] During the carbonation curing process, the technical team used a near-infrared spectroscopy device to scan the concrete surface and borehole sampling points every 3 hours. The fiber optic probe of the device acquired reflectance spectral data in the 1200–2200 nm band with a spectral resolution of 5 nm, and each scan took 25 seconds. As shown in Table 1, the spectral data obtained from the first scan after 6 hours of curing showed that the intensities of the characteristic absorption peaks of calcium carbonate in the 1450 nm and 2330 nm bands were 0.342 and 0.287, respectively. Based on these spectral characteristics, the rapid carbon sequestration assessment model calculated the current carbon sequestration to be 9.8. A second scan after 12 hours of curing showed that the characteristic peak intensities increased to 0.461 and 0.385, respectively, corresponding to a carbon fixation content of 16.5%. After 18 hours of curing, the characteristic peak intensities of the third scan reached 0.538 and 0.452, indicating a carbon fixation content of 21.2%. After 24 hours of curing, the characteristic peak intensities of the fourth scan were 0.592 and 0.498, and the carbon fixation content was 24.1. .

[0089] Table 1. Changes in the intensity of near-infrared spectral characteristic peaks and carbon fixation during carbonization curing.

[0090]

[0091] The technical team calculated the current carbon sequestration to be 24.1 after 24 hours of maintenance. With target carbon sequestration of 25 The deviation was 3.6%, indicating a deviation exceeding 3% but not reaching 8%, thus the carbonization curing time was extended. After an additional 3 hours of curing, reaching a total curing time of 27 hours, a fifth spectral scan was performed, yielding characteristic peak intensities of 0.621 and 0.523, respectively, resulting in a calculated carbon sequestration amount of 25.6%. The deviation from the target value decreased to 2.4%, meeting the termination condition. The technical team immediately stopped... The supply is achieved by slowly opening the exhaust port of the curing chamber via an electromagnetic control valve, allowing the pressure inside the chamber to drop evenly to atmospheric pressure within 20 minutes. This prevents rapid pressure drop from causing a pressure gradient inside the concrete that could trigger microcracks. Subsequently, the spray humidification system is kept running continuously to maintain the relative humidity inside the curing chamber at 85%, while the temperature is gradually reduced to room temperature. Wet curing continues for 60 hours to stabilize the calcium carbonate crystal structure and prevent surface water loss and shrinkage.

[0092] After post-treatment, samples were taken from the concrete components for testing. Core drilling revealed an average carbonation depth of 38 mm, far exceeding the 8–12 mm carbonation depth of traditional atmospheric pressure natural carbonation processes. Scanning electron microscopy showed that calcium carbonate crystals were uniformly distributed in the concrete pores, with grain sizes concentrated in the 200–500 nm range, effectively filling the capillary pores of the cement paste. X-ray diffraction analysis confirmed that the calcium carbonate was predominantly calcite phase, with minor amounts of aragonite and spherulite phases, and a total carbon fixation content of 25.6%. The results meet design requirements. Compressive strength tests show that the 28-day compressive strength is 56.3 MPa, 12.7% higher than the control group without carbonation treatment. This indicates that the calcium carbonate generated during carbonation fills the pores, enhancing the concrete's density and improving its mechanical properties. Chloride ion penetration tests measured an electrical flux of 1280 coulombs, and the durability rating is high, indicating that carbonation treatment significantly reduced the concrete's permeability.

[0093] The technological advancements brought about by this invention compared to traditional concrete carbon sequestration processes are mainly reflected in the following aspects. Traditional processes rely on atmospheric pressure environments. The natural diffusion is driven solely by the concentration gradient, and the residual air in the concrete pores creates mass transfer resistance, leading to... Due to the difficulty in penetrating deep into the component, the carbonization reaction is limited to a thin surface layer. This invention removes air from the pores and creates low-resistance channels through vacuum pretreatment, followed by the application of pressure. Providing forced convection driving force, combined with the ultrasonic cavitation effect to destroy the gas-liquid interfacial resistance, This invention enables rapid penetration into the deep layers of concrete, significantly improving carbonation depth and total carbon sequestration. Traditional curing processes rely on crude temperature and humidity control, maintaining environmental parameters within a wide range of fluctuations. They cannot finely adjust for spatial differences in the internal temperature field of the concrete, leading to uneven carbonation reaction rates between the surface and core areas. This can easily cause localized overheating or moisture loss, affecting the uniformity of carbonation products and final performance. This invention uses a distributed temperature sensor array to monitor the internal temperature field in real time. A neural network model intelligently calculates the optimal temperature setpoint for each area based on temperature distribution characteristics. A circulating cooling water and spray humidification system achieve precise temperature control in different zones, ensuring that the carbonation reaction across the entire component cross-section occurs synchronously under optimal temperature and humidity conditions, eliminating the non-uniformity caused by temperature gradients. Traditional carbon sequestration detection relies on phenolphthalein solution spray colorimetric method or offline thermogravimetric analysis. The former can only qualitatively determine the boundaries of carbonation areas and cannot quantify the amount of carbon sequestration, while the latter requires destructive sampling and has a testing cycle of several hours, failing to provide real-time feedback on carbonation progress during curing. This results in curing time being estimated based on experience, leading to the risk of insufficient or excessive curing. This invention uses near-infrared spectroscopy detection technology combined with a convolutional neural network model to achieve rapid and non-destructive assessment of carbon sequestration. A single detection takes only tens of seconds and can be repeated multiple times during the maintenance process. The maintenance strategy can be dynamically adjusted based on the real-time carbon sequestration data, which avoids the failure to achieve the carbon sequestration target due to insufficient maintenance, and also prevents energy waste and efficiency reduction caused by over-maintenance, thus realizing closed-loop precise control of the carbonization process.

[0094] It should be noted that the variables involved in this invention are explained in detail in Tables 2 and 3.

[0095] Table 2. Variable Explanation Table (Part 1)

[0096]

[0097] Table 3. Variable Explanation Table (Part Two)

[0098]

[0099] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. An optimized method for concrete carbon sequestration production process, characterized in that, This process involves mixing mineral admixtures with cement to prepare a base concrete slurry, adding nano-calcium carbonate seed crystals and pore conditioners during the mixing process, injecting the base concrete slurry into molds, sending it to a curing chamber, and using a vacuum pretreatment device to negatively pressure-extract residual air from the concrete pores. Immediately after vacuum pretreatment, the process is switched to pressurization. Injection mode will Gas is injected into the curing chamber, and an ultrasonic generator is simultaneously activated to apply ultrasonic-assisted gas dissolution and diffusion. A distributed temperature sensor array collects real-time temperature field data within the concrete and inputs this data into a carbonation temperature control optimization model to obtain the optimal curing temperature setpoints for each zone. Based on these optimal setpoints, the circulating cooling water flow rate and the operating parameters of the spray humidification system are adjusted to precisely maintain the curing temperature within the set range. During the carbonation curing process, a near-infrared spectroscopy device is used at regular intervals to scan the concrete surface and borehole sampling points to acquire spectral data, which is then input into a rapid carbon fixation assessment model for analysis. The current carbon fixation value calculated by the rapid carbon fixation assessment model is used to determine the degree of deviation between the current value and the target carbon fixation value, and the carbonation curing time is adjusted or increased accordingly. Inject pressure.

2. The method for optimizing the concrete carbon sequestration production process according to claim 1, characterized in that, The mineral admixtures are mixed with cement in a mass ratio of 30 to 45% fly ash, 25 to 40% slag powder, and 5 to 10% silica fume to prepare the basic concrete slurry.

3. The method for optimizing the concrete carbon sequestration production process according to claim 2, characterized in that, The nano-calcium carbonate seed crystals have a particle size of 30 to 80 nm. Particles are used to provide nucleation sites for calcium carbonate crystal growth in the pore liquid phase of concrete.

4. The method for optimizing the concrete carbon sequestration production process according to claim 3, characterized in that, The pore regulator is a compound of polycarboxylate superplasticizer and expansion agent, used to regulate the pore size distribution and connectivity during the setting and hardening process of concrete.

5. The method for optimizing the concrete carbon sequestration production process according to claim 4, characterized in that, The suction pressure of the vacuum pretreatment device is controlled at -0.08 to -0.12 MPa for a duration of 15 to 25 minutes.

6. The method for optimizing the concrete carbon sequestration production process according to claim 5, characterized in that, The pressurization In injection mode The gas purity is not less than 99.5%, and the injection pressure is 0.3 to 0.6 MPa.

7. The method for optimizing the concrete carbon sequestration production process according to claim 6, characterized in that, The ultrasonic generator applies ultrasonic waves at a frequency of 20 to 40 kHz to assist in the dissolution and diffusion of the gas.

8. The method for optimizing the concrete carbon sequestration production process according to claim 7, characterized in that, The distributed temperature sensor array consists of fiber optic temperature sensors arranged on the surface, middle layer and core area of ​​the concrete component, with a sensor spacing of no more than 100mm.

9. The method for optimizing the concrete carbon sequestration production process according to claim 8, characterized in that, The temperature field data includes real-time temperature values, temperature change rates, and temperature distribution uniformity indices measured by each fiber optic temperature sensor in the distributed temperature sensor array.

10. The method for optimizing the concrete carbon sequestration production process according to claim 9, characterized in that, The carbonization temperature control optimization model is a temperature field prediction and control model based on neural networks, which is based on the input temperature field data, The output parameters of injection pressure, curing time, and mineral admixture proportions are used to determine the optimal curing temperature settings for each zone that maximize the carbonization reaction rate and ensure uniform temperature distribution.

Images