Intelligent pouring and curing method for large-volume concrete in pile-raft foundation construction

Abstract

This invention relates to the field of intelligent management technology for pile-raft foundation construction, and particularly to an intelligent pouring and curing method for large-volume concrete used in pile-raft foundation construction. The method includes constructing an integrated monitoring system, establishing a data-driven intelligent prediction model for the concrete temperature field, applying low-heat, high-crack-resistant concrete mix design technology, implementing a low-noise construction technology system, and constructing a multi-objective collaborative optimization model. This invention employs a material-monitoring-prediction triple control mechanism, resulting in a lower maximum temperature rise and smaller internal and external temperature difference during the curing process, ensuring virtually no cracks and fundamentally improving the crack resistance and durability of large-volume concrete. Intelligent scheduling through optimized spatiotemporal layout achieves precise, intelligent, and minimally disruptive noise control. Through a multi-objective collaborative optimization algorithm, it solves for the minimization of temperature quality objectives, noise impact objectives, and total cost, achieving a globally optimal solution with the lowest total project cost and shortest total construction period, thus improving the accuracy, reliability, and efficiency of decision-making.

Description

Technical Field

[0001] This invention relates to the field of intelligent management technology for pile-raft foundation construction, and in particular to an intelligent method for pouring and curing large-volume concrete for pile-raft foundation construction. Background Technology

[0002] In pile-raft foundation construction, especially during the pouring and curing of large-volume concrete, controlling the heat of hydration is a core challenge in ensuring the integrity and durability of the raft slab. Traditional curing methods mainly rely on manual experience for heat preservation and moisture retention, lacking real-time and precise control of the internal temperature field, which can easily lead to temperature cracks.

[0003] In recent years, with the introduction of the Internet of Things and intelligent sensing technologies, real-time temperature monitoring systems based on wireless transmission have been gradually applied to major projects. By dynamically tracking the internal temperature gradient of concrete, precise control and early warning of curing measures have been achieved. Meanwhile, non-traditional post-pouring strip technologies, such as material optimization, are also being promoted, effectively reducing shrinkage constraints and improving the crack resistance of ultra-long structures. However, for large-volume concrete, traditional construction methods still have the following technical shortcomings when dealing with ultra-thick raft slabs:

[0004] 1) Relying on empirical maintenance methods makes it difficult to accurately control the internal temperature stress caused by the heat of cement hydration, which can easily lead to harmful temperature cracks. This not only damages the integrity, waterproofness and durability of the structure, but also brings high costs or even permanent quality hazards to the later leakage treatment.

[0005] 2) At the same time, with the continuous expansion of building scale and the innovation of design concepts (such as the optimization of raft slab thickness brought about by variable stiffness leveling design), higher requirements are put forward for the crack resistance performance of concrete. It is urgent to achieve a fundamental transformation from passive treatment to active prevention and control through refined processes such as intelligent temperature monitoring system, high-performance material formula optimization and skip-pour method.

[0006] In addition, research and application in construction noise reduction mainly focus on process substitution and passive protection. Static pressure piling technology, as a representative low-noise and vibration-free process, has become the preferred choice in urban center projects, with its equipment tonnage and piling capacity continuously improving to adapt to more complex geological conditions. For processes where noise cannot be avoided, such as drilling, mobile sound barriers and local sound enclosures are commonly used, combined with construction time management, to reduce the impact on the acoustic environment. However, a comprehensive, multi-layered active noise reduction technology system and an intelligent noise monitoring and management platform are still not widespread, and a real-time acoustic environment data sharing mechanism linked with the community is lacking. Summary of the Invention

[0007] The purpose of this invention is to address the shortcomings of existing technologies by proposing an intelligent method for pouring and curing large-volume concrete for pile-raft foundation construction.

[0008] To achieve the above objectives, the present invention adopts the following technical solution:

[0009] The intelligent pouring and curing method for large-volume concrete used in pile-raft foundation construction includes the following steps:

[0010] S1: Construct an integrated monitoring system by arranging a wireless temperature sensor array inside the concrete structure and a noise monitoring instrument at the construction site boundary to achieve synchronous acquisition of temperature field and acoustic environment data.

[0011] S2: Establish a data-driven intelligent prediction model for concrete temperature field. Based on the time-series data collected in S1, predict the future development of the temperature field and generate early warning of excessive temperature difference and intelligent maintenance suggestions accordingly.

[0012] S3: Applying low-heat, high-crack-resistant concrete preparation technology, by compounding large-volume admixtures, high-performance additives and fibers, the heat of hydration is reduced and crack resistance is improved from the material source.

[0013] S4: Implement a low-noise construction technology system, including the application of static pressure piles under complex geological conditions, the deployment of modular mobile sound barriers, and the intelligent scheduling of high-noise operations based on noise monitoring data from S1.

[0014] S5: Construct a multi-objective collaborative optimization model with constraints such as concrete internal and external temperature difference ≤25℃, crack width <0.2mm, and field boundary noise ≤70dB(A) during the day and ≤55dB(A) at night. The objective functions are minimizing the total project cost and minimizing the total construction period. Solve for the optimal combination of construction parameters to achieve global optimization of quality, environmental protection, cost and construction period.

[0015] Preferably, the intelligent prediction model in S2 is a Long Short-Term Memory (LSTM) network model. Its inputs include historical temperature sequences, ambient temperature and humidity, and static characteristics of concrete mix proportions. The output is the predicted value of the internal temperature of the concrete in a specific future time window. When the predicted internal and external temperature difference exceeds 23°C, an early warning is triggered. When it exceeds 25°C, an active curing control command is triggered.

[0016] Furthermore, the input and output of the LSTM model are specific engineering data, but the internal structure of the model involves mathematical processing; the LSTM model comprises the following three-layer architecture:

[0017] 1) Engineering data interface (input and output)

[0018] Input vector x t :

[0019]

[0020] T N,tThe temperature measurement value (°C) of the Nth sensor at time step t; this is the most crucial input.

[0021] Ta t : Ambient temperature (°C) at time step t;

[0022] RH t : Ambient relative humidity (%) at time step t;

[0023] S c : Static feature vector, containing concrete mix proportions (cement dosage, fly ash content, etc.) and geometric parameters (sensor location, raft foundation thickness); these are input once at the start of the project: c: Cement dosage (kg / m³) 3 ); w / c: water-cement ratio; pfa: fly ash content (%); raft thickness (m); sensor burial depth (m);

[0024] 2) Internal computation of LSTM (mathematical abstraction process)

[0025] LSTM models include:

[0026] ForgetGate: Determines which historical information in the cell state needs to be "forgotten" (by outputting a weight of 0 to 1 using the sigmoid function, where 0 represents complete forgetting and 1 represents complete retention).

[0027] Input Gate: Determines which new information (such as features of the current word) at the current time step needs to be "written" into the cell state;

[0028] Output Gate: Determines which information from the cell state needs to be "output" to the hidden state of the LSTM (for subsequent computation or prediction).

[0029] The following formula describes the mathematical transformation process of data within the model. Its parameters need to be obtained through training with a large amount of data, rather than being calculated through physical formulas.

[0030] ;

[0031] f t The output of the forget gate is a vector with values ​​between (0,1), activated by the sigmoid function:

[0032] ;

[0033] The Sigmoid function is: This maps the input to the (0,1) interval, providing a "weighting coefficient" for the forget gate (controlling the degree of information retention / forgetting).

[0034] h t-1 The LSTM hidden state (of dimension d) at the previous time step t-1 contains the "short-term memory" processed in the previous time step; x t Given the input sequence elements (of dimension n) at the current time step t; concatenate the two elements column-wise (or row-wise) into a long vector of dimension d+n, this concatenated vector [h t-1 ,x t ] is used as the input for the forget gate; f t Determining the state of LSTM cells C t-1 The retention rate of "historical information" (element-wise, i.e., each dimension is controlled independently) is determined by the value. The closer the value is to 1, the more historical information is retained in the corresponding dimension; the closer it is to 0, the more historical information is forgotten in the corresponding dimension. t The decision is made from internal memory (cell state C) t-1 The model automatically learns to discard information and determine which long-term trends are no longer important. For example, when hydration enters a period of stable cooling, the model may choose to "forget" certain short-term fluctuation details from the early stage of rapid warming and focus more on the current cooling rate.

[0035] W f Here is the weight matrix for the forget gate, with dimension d: assuming the hidden layer dimension is d, and the input x... t If the dimension is n, then W f The dimension is d×(d+n), for input [h t-1 ,x t Perform a linear transformation to capture the "historical hidden state h". t-1 "and "Current input x t The combined effects of "forgetting decisions";

[0036] b f This is the bias vector for the forget gate, with the same dimension d as the hidden layer (length d); it provides an "offset" for the linear transformation, increasing the model's expressive power (similar to the role of bias in neural networks, making the output of the activation function more flexible).

[0037] C t-1 It is the cell state at the previous time step (historical memory).

[0038] i t The output value of the input gate is a vector with values ​​between (0,1), activated by the Sigmoid function:

[0039] ;

[0040] Among them W i Let b be the weight matrix of the forget gate. i This is the bias vector for the forget gate;

[0041] The candidate cell state is a vector with values ​​between (-1, 1), activated by the tanh function:

[0042] ;

[0043] Where tanh is the hyperbolic tangent activation function, the formula is: This maps the input to the (-1,1) interval, providing a "numerical range constraint" for the candidate cell state, while introducing nonlinearity to enhance the model's expressive power.

[0044] W C Let b be the weight matrix of the candidate cell states. C The bias vector of the candidate cell state; combined with the "historical hidden state h" t-1 "and "Current input x t "Through the weight matrix W C Bias b C With tanh activation, generate "new information candidates for the current time step". ", will subsequently be related to the output i of the input gate t Multiplication, jointly updating cell state C t The model selectively retains old memories and integrates them with new, important information to form a comprehensive memory for the current time step. Based on newly received data (such as a sudden increase in the rate of temperature rise), the model determines that this is a new pattern that needs to be memorized (possibly corresponding to the arrival of a hydration exothermic peak) and generates a candidate "new memory". This controls the writing intensity of this new memory; this memory may encode abstract temporal features such as "concrete is in the most intense stage of hydration and heat release," but this feature itself is a data pattern rather than a physical quantity.

[0045] Output gate o t It is a vector with values ​​between (0,1), activated by the Sigmoid function:

[0046] ;

[0047] Among them W o Let b be the weight matrix of the output gate. o The bias vector of the output gate; o t Determining cell state C t How much information will be output to the hidden state h at the current time step? t (Element-wise, meaning each dimension independently controls the output ratio). The closer the value is to 1, the more cell state information is output for the corresponding dimension; the closer it is to 0, the less cell state information is output for the corresponding dimension.

[0048] The formula for updating the hidden state is:

[0049] ;

[0050] For element-wise multiplication (Hadamard Product), for two vectors o of the same dimension t and tanh(C t Multiplying element by element achieves the effect of "filtering information proportionally" (i.e., each element in \(ot\) is multiplied by tanh(C)). t (Corresponding element's "on / off" control); Output gate o t As a "weight controller," it uses element-wise multiplication to transform the cell state C after the tanh transformation. t Select some information from the data and generate the hidden state h for the current time step. t For example, to predict the internal point temperature, the output gate might learn to "mask" noise in memory that is sensitive to surface temperature but not to internal temperature; hidden state h t It is the "selective output" of LSTM to "long-term memory (cell state)" that retains the key information of cell state while adapting to the needs of subsequent calculations or tasks.

[0051] 3) Final prediction layer

[0052] The high-level abstract features h extracted by LSTM t Mapped to a specific physical quantity—future temperature predictions. :

[0053] ;

[0054] Among them W y Let b be the weight matrix of the prediction layer. y The bias vector of the prediction layer; This is the model's final prediction output; for example:

[0055] ;

[0056] The model outputs the predicted temperature values ​​of all sensors for the next m time steps. .

[0057] Furthermore, the LSTM model also includes model training and loss function settings:

[0058] Loss function: The mean squared error (MSE) is used to measure the difference between the predicted value and the true value;

[0059] ;

[0060] N is the number of sensors, and m is the prediction time step. Let be the actual temperature measured by the i-th sensor at time step t+j. The predicted temperature measured by the i-th sensor at time step t+j (calculated from the model input time step t to the predicted value at output time step t+j) is backpropagated through a time-mapping algorithm, and the loss function is minimized using an optimizer (such as Adam). This updates all parameters.

[0061] .

[0062] Preferably, the low-heat, high-crack-resistant concrete in S3 comprises the following raw materials:

[0063] Cementitious materials include: ordinary Portland cement, grade 42.5, 200-230 kg / m³. 3 It provides the main strength, but controlling its dosage is key to reducing the heat of hydration; Grade II fly ash, 90-110 kg / m³ 3 Core admixtures: ① Physical filler to improve particle size distribution; ② Secondary hydration reaction, consuming Ca(OH)2 to generate low-alkalinity CSH gel, optimizing interface structure; ③ Significantly reduces early hydration heat peak and rate; S95 grade mineral powder, 70-90 kg / m³ 3 Core admixtures: ① Micro-aggregate effect and pozzolanic effect, significantly reducing heat of hydration; ② Improved later-stage strength and durability; ③ Reduced chloride ion permeability of concrete;

[0064] Aggregate sand, including: fine aggregate, 700-750 kg / m³ 3 The aggregate forms the framework of the concrete, and stable, high-quality aggregate is the foundation for reducing shrinkage; coarse aggregate, 1020-1080 kg / m³ 3 It consists of 5-25mm continuously graded crushed stone;

[0065] Chemical admixtures, including: polycarboxylate-based high-performance water-reducing agents, 4.0-5.5 kg / m³. 3 Based on solids content, it features: ① High water reduction rate ≥25%, ensuring workability at low water-cement ratios; ② Slow-release slump retention, suitable for long-distance transportation and long-term pouring of large-volume concrete; ③ Low shrinkage, helping to reduce chemical shrinkage; Fiber: Polypropylene fiber, 0.6-0.9 kg / m³ 3 Length: 12-19mm. During the plastic stage and early hardening stage of concrete, millions of fibers are randomly distributed within the matrix, forming a three-dimensional support network. This effectively inhibits the generation and development of plastic shrinkage cracks and drying shrinkage cracks, improving the toughness of concrete. Water content for mixing: 155-165kg / m³ 3 Controlling water consumption is key to ensuring strength and reducing porosity;

[0066] Key proportioning parameters:

[0067] Water-cement ratio (W / B): 0.38-0.42. A lower water-cement ratio is a prerequisite for ensuring high strength and high durability; Total amount of cementitious material: 360-430 kg / m³ 3 While ensuring strength, the amount of cement used is effectively controlled by admixtures; the total amount of admixtures accounts for 40%-50% of the mass of cementitious materials, which is the core means to achieve "low heat"; the sand ratio is 40%-42%, which is finely adjusted according to the specific conditions of sand and gravel to ensure good workability and density.

[0068] Preferably, the intelligent scheduling in S4 specifically involves: establishing a linkage mechanism between noise monitoring data and the operating status of construction equipment; when the real-time monitoring value of site boundary noise approaches 70 dB(A), the system automatically sends an alarm to the dispatch center and recommends or implements strategies such as suspending some high-noise operations, adjusting equipment positions, or activating backup noise reduction measures.

[0069] Furthermore, the intelligent scheduling includes the following noise propagation optimization algorithm:

[0070] This algorithm aims to minimize the impact of construction noise on the surrounding environment through modeling and optimization.

[0071] 1) Noise propagation model ;

[0072] L p,AB The sound pressure level (in dB) produced by noise source A at receiving point B describes the noise intensity contributed by a specific noise source A (such as static pressure piles, rotary drilling rigs, or other construction equipment) at receiving point B; L w,A The sound power level (in dB) of noise source A is the "sound emission capability" of noise source A itself, and is a core sound source parameter; it can be obtained through on-site testing and is an inherent property of the noise source (independent of propagation distance, medium, etc.); r AB The distance from noise source A to receiving point B (in meters) is a key factor affecting sound pressure level attenuation. The greater the distance, the more significant the sound energy diffusion and the lower the sound pressure level. α represents the sound pressure level attenuation term (in dB) caused by the geometric diffusion of spherical waves. When noise propagates in the form of spherical waves, the sound energy decreases with the square of the distance, corresponding to a logarithmic decrease in sound pressure level with distance (the sound pressure level decreases by 20 dB for every 10-fold increase in distance); α is the air absorption coefficient (in dB / m), the ability of air to absorb sound waves, which is related to the sound wave frequency, air temperature, humidity, etc. (the higher the frequency and the drier the air, the stronger the absorption); αr ABThis is the sound pressure level attenuation term caused by air absorption (in dB). When sound waves propagate in the air, sound energy is absorbed due to air molecule friction, heat conduction, etc., causing the sound pressure level to decrease linearly with the propagation distance (α is the attenuation per unit distance, multiplied by the distance r). AB (Total attenuation) The additional attenuation (in dB) of the sound pressure level caused by the sound barrier (if present), i.e., the insertion loss (dB) of the sound barrier (if present), is related to the barrier height h. b The sound absorption coefficient β of the material and the relative position of the sound source / receiving point are related. The sound barrier reduces the sound pressure level at the receiving point by blocking and reflecting noise. It needs to be calculated based on factors such as the height, length and material of the barrier.

[0073] In summary, this formula quantitatively describes the "sound pressure level attenuation process when a single noise source A propagates to the receiving point B", and is the basis for calculating the total sound pressure level by superimposing multiple sound sources (the total sound pressure level requires superposition of the energy of each noise source's \(L{p,AB}\)).

[0074] 2) Superposition of multiple noise sources

[0075] The total sound pressure level at a receiving point is the sum of the contributions from all sound sources:

[0076] ;

[0077] The total sound pressure level (in dB) generated by all n noise sources at the receiving point B describes the total noise intensity at the receiving point B (such as a field boundary monitoring point) when multiple noise sources act simultaneously, reflecting the superposition effect of noise.

[0078] Optimization based on genetic algorithms:

[0079] The noise control problem is defined as an optimization problem and solved using a genetic algorithm.

[0080] Decision variable, project duration:

[0081] ;

[0082] t k s is the start time of operation for device k; k The working position of device k (selected from the available workstations);

[0083] Objective function:

[0084] ;

[0085] w1 and w2 are weighting coefficients used to balance noise control and construction efficiency;

[0086] Constraints, Noise:

[0087] Daytime ≤70dB; Nighttime ≤55dB; inter-device operation logic constraints (e.g., device A must start after device B has finished); devices cannot occupy the same space at the same time;

[0088] Genetic algorithm operation:

[0089] Selection: Roulette wheel selection method, solutions with higher fitness (smaller f(X)) have a greater chance of proliferating;

[0090] Crossover: Single-point crossing, where partial gene sequences of two parent chromosomes are exchanged to produce a new individual;

[0091] Mutation: A random change to the start time or location of a device with a small probability;

[0092] 3) Multi-objective collaborative optimization algorithm

[0093] This is the system's top-level command center, responsible for balancing and coordinating the sometimes conflicting goals of temperature control and noise control;

[0094] Problem Modeling

[0095] Objective 1, Temperature Quality:

[0096] ;

[0097] Objective 2, Noise Impact:

[0098] ;

[0099] Objective 3, Economic Costs:

[0100] ;

[0101] Where t is the time step, B is the noise receiving point, and Y is the collaborative decision variable, including maintenance plan, equipment scheduling plan, etc., which is the core variable affecting various costs. Different Y values ​​will lead to changes in costs such as temperature control, noise control, and construction delays.

[0102] f3(Y) is the total cost function related to construction scheme Y, which comprehensively measures the sum of various costs during the construction process. It is the objective that needs to be minimized in the optimization problem and is used to evaluate the economics of different construction schemes; C temp Temperature control costs refer to the costs incurred during construction for controlling temperature (such as concrete curing temperature control, equipment heat dissipation temperature control, etc.), including the purchase / rental of temperature control equipment, energy consumption, and manual monitoring expenses; C noiseThe cost of noise control refers to the costs incurred during construction for noise control (such as setting up sound barriers, using low-noise equipment, and implementing noise reduction measures during nighttime construction), including the investment in noise reduction equipment and the labor and material costs for implementing the measures; C delay Costs for project delays include the costs incurred due to project delays caused by construction plan Y, such as increased management fees, contract penalties, and equipment downtime losses.

[0103] In summary, this formula is a comprehensive quantitative model of the total construction cost. The core logic is as follows: the construction plan Y will simultaneously affect three types of costs: temperature control, noise control, and construction delay. The total cost f3(Y) is the linear sum of these three types of costs. The optimization objective is to minimize the total cost by adjusting Y. It is suitable for scenarios involving the economic efficiency of construction plans and collaborative optimization of multiple constraints (such as selecting the construction plan with the lowest cost under the premise of meeting temperature and noise standards).

[0104] Preferably, the multi-objective collaborative optimization model in S5 is solved using a non-dominated sorting genetic algorithm with an elitist strategy, and the output is a set of Pareto optimal construction schemes that balance crack resistance, noise control, cost and schedule, for decision-makers to choose from.

[0105] Furthermore, the non-dominated sorting genetic algorithm (NSGA-II) with an elitist strategy is used to find the Pareto optimal solution set;

[0106] Non-dominated sorting: Stratify the solution population so that the solution that is no worse than other solutions on any of the objectives f1, f2, f3, and is better on at least one objective is ranked first (Rank1).

[0107] Crowding calculation: Within the same Rank, calculate the density of each solution in the target space, and prioritize retaining solutions in sparse regions to maintain population diversity;

[0108] Cooperative constraint handling:

[0109] Hard constraint: The internal and external temperature difference ΔT(t) between the concrete core and surface ≤ 25℃, noise L p,total,B ≤ Regulatory limit; solutions that do not meet the limit are directly eliminated;

[0110] Soft constraints / penalties: For example, if high-noise work at night (such as earthwork excavation) greatly interferes with concrete curing and temperature monitoring personnel, then a penalty term P is added to the objective function. night ;

[0111] decision making:

[0112] The algorithm will eventually output a set of Pareto optimal solutions; project decision-makers can choose a final implementation plan from this set of solutions based on the current project's priorities (whether to focus more on absolute quality or on community relations).

[0113] Through this complex, multi-layered algorithm system, we can transform traditional experience-based construction management into a data-driven, model-predictive, and intelligently optimized precise control process, thereby achieving optimal synergy between the project's intrinsic quality and the impact of the external environment.

[0114] This invention is the first to incorporate concrete temperature control and construction noise reduction into a unified algorithm framework, achieving synergistic improvement in quality and environmental protection through multi-objective optimization. This technology can be extended to large-scale projects such as underground engineering and water conservancy projects, which is in line with the development trend of green construction and intelligent construction and has broad prospects for industrial application. The algorithm system has been calibrated for the No. 1 zero-carbon energy station project in the North Zone of Taiyuan Wusu International Airport, and can be adapted to other engineering scenarios through transfer learning in the future.

[0115] Furthermore, the application of the non-dominated sorting genetic algorithm with an elitist strategy in engineering decision-making includes the following process:

[0116] I. Algorithm Objectives

[0117] Under the premise of satisfying all hard constraints (temperature difference, cracks, noise), a series of construction schemes that make the best trade-off between the two conflicting objectives of total cost and total construction period are found. The set of these schemes is called the Pareto optimal solution set, and its mapping in the objective space is called the Pareto front.

[0118] II. NSGA-II Solution Process

[0119] 1) Initialize the population

[0120] Operation: Randomly generate Y initial "construction plans", each plan including:

[0121] Concrete mix design number (discrete value); Curing start time (continuous value); Number of static pressure pile drivers used (integer value); Segmentation number for skip-pile method (discrete value); Schedule of high-noise operation periods (discrete value).

[0122] 2) Evaluating individuals

[0123] Operation: For each Y in the population, calculate its objective function value and constraint violation degree, input the scheme Y into the surrogate model (such as LSTM temperature prediction model, noise propagation model) and simulate its execution consequences;

[0124] Calculate the target values ​​f1, f2, f3 and their constraint violation degrees;

[0125] 3) Non-dominated sorting

[0126] The entire population is stratified according to Pareto dominance;

[0127] According to the conditions, if and only if:

[0128] Y1 is no worse than Y2;

[0129] At least on one objective, Y1 is significantly better than Y2;

[0130] The dominance relationship is derived: Option Y1 dominates Option Y2;

[0131] Rank 1: Identify all individuals that are not dominated by any other individual; this is called the first non-dominated layer.

[0132] Rank 2: From the remaining individuals, find all the non-dominated individuals, which is called the second non-dominated layer;

[0133] Repeat this process until all individuals have been stratified;

[0134] The smaller the Rank value, the better the solution; the set of Rank 1 is an approximation of the Pareto optimal solution set we are looking for.

[0135] 4) Crowding Calculation

[0136] Within the same non-dominated layer, calculate the density of each individual in its objective function space;

[0137] Calculation: For each objective function, calculate the difference between the function values ​​of the individual and the two adjacent individuals on both sides, and sum them; the larger this value is, the more "open" the area around the individual is.

[0138] To maintain population diversity and prevent all solutions from crowding into a small area on the Pareto front, highly crowded individuals (in sparsely distributed areas) will be preferentially preserved.

[0139] 5) Selection, crossover, and mutation

[0140] Selection: Based on non-dominated ranking and crowding, a binary tournament selection method is adopted; that is, two individuals are randomly selected, and the one with the smaller rank is selected first; if the ranks are the same, the one with the larger crowding is selected; this ensures that excellent and diverse genes are inherited.

[0141] Crossover: Randomly select two "parent" individuals, exchange some of their genes, and produce new "offspring" individuals; this is equivalent to combining the advantages of two construction schemes.

[0142] Mutation: With a small probability, a gene value of an individual is randomly changed; this introduces new possibilities and helps to escape local optima.

[0143] 6) Elite Strategy

[0144] The parent and offspring populations are merged into a single population of size 2N.

[0145] The non-dominated sorting and crowding calculation are recalculated for this merged population, and then only the top N individuals are retained as the parents for the next iteration.

[0146] This ensures that the outstanding individuals (elites) in the previous generation are not destroyed, thereby accelerating convergence and ensuring that the quality of the solution set does not decrease monotonically;

[0147] III. Constraint Handling

[0148] NSGA-II typically employs the "constraint dominance" principle to handle hard constraints:

[0149] If both individuals satisfy all constraints, then the comparison is performed according to the original Pareto dominance relationship;

[0150] If one individual satisfies the constraints while another violates them, the individual that satisfies the constraints is always considered superior.

[0151] If both individuals violate the constraints, the individual with the smaller degree of constraint violation is considered the better one;

[0152] IV. Final Output and Application

[0153] When the algorithm reaches the maximum number of iterations, all individuals in the first non-dominated layer (Rank 1) will be output, which is the Pareto optimal solution set;

[0154] Each solution in this set represents a unique trade-off between cost and time; no solution can be further improved without compromising one of the objectives.

[0155] Decision support: From the Pareto optimal solution set, make the final decision based on the actual priorities of the current project.

[0156] If funds are tight: choose the option with the lowest cost (even if the construction period may be slightly longer);

[0157] If the owner requires urgent completion: choose the option with the shortest construction period (even if it is more expensive);

[0158] Unless otherwise specified: Choose an equilibrium point located in the middle of the Pareto front;

[0159] By applying NSGA-II, the system will be upgraded from "being able to provide a few feasible solutions" to "being able to systematically provide a series of globally optimal trade-off solutions", thereby making the decision-making process more scientific and visualized, rather than relying on empiricism.

[0160] Compared with the prior art, the beneficial effects of the present invention are:

[0161] 1. This invention employs a triple control mechanism of materials, monitoring, and prediction, resulting in a lower maximum temperature rise and smaller internal and external temperature difference during the curing process. This ensures crack widths of <0.2mm or even no cracks, significantly improving the structural quality and long-term safety, and fundamentally enhancing the crack resistance and durability of large-volume concrete.

[0162] Material-based heat reduction: The low-heat, high-crack-resistant concrete mix design through the S3 step significantly increases the amount of admixtures such as fly ash and mineral powder (accounting for 40%-50% of cementitious materials), replacing part of the cement. This reduces the total heat of hydration and the rate of heat release from the source, weakening the intensity of the heat source that causes temperature stress.

[0163] Intelligent prediction, prevention is better than cure: A temperature field intelligent prediction model based on LSTM, its cell state C t It can learn and memorize the long-term hydration heat release law of concrete, while the hidden state h t It captures short-term dynamic changes in temperature and can predict temperature values ​​at multiple future time steps in advance. This allows the system to proactively issue warnings and suggest maintenance measures (such as covering with an insulation layer in advance) when the temperature difference is close to but does not exceed a threshold, realizing a paradigm shift from passive response to proactive intervention and completely solving the problem of lag in traditional maintenance.

[0164] 2. This invention upgrades noise reduction from management measures to embedded intelligent scheduling and control, achieving precise, intelligent, and interference-minimized noise control:

[0165] Source control and propagation blocking: In complex geological conditions, low-noise technologies such as static pressure piles are applied to directly reduce noise at the source. Simultaneously, modular mobile sound barriers are deployed to effectively block noise propagation paths.

[0166] Intelligent scheduling optimizes spatiotemporal layout: A noise propagation optimization algorithm based on genetic algorithm is introduced, which incorporates noise monitoring data, equipment operation logic, spatiotemporal constraints, etc. into a mathematical model. By superimposing multi-source noise, the boundary noise is accurately calculated, and the operation time and spatial location of high-noise equipment are intelligently planned. It automatically avoids sound-sensitive periods (such as night) and areas, thereby maximizing construction efficiency while meeting stringent environmental protection standards. This invention can stably control the boundary noise within the limits of ≤70dB(A) during the day and ≤55dB(A) at night, avoiding environmental penalties and community disputes, and realizing green and harmonious construction.

[0167] 3. Break down the barriers between quality and environmental protection to achieve optimal overall decision-making.

[0168] This invention employs a multi-objective collaborative optimization algorithm. The model acknowledges the complex and even contradictory relationships between quality (temperature difference, cracks), environmental protection (noise), cost, and construction period. Through advanced algorithms such as non-dominated sorting and congestion calculation, it performs a global search for all possible combinations of construction parameters. Following collaborative constraint processing, the model simultaneously considers temperature quality objectives and noise impact objectives, seeking to minimize the total cost. This invention, through system collaboration, avoids high repair costs and construction delays caused by quality issues, as well as downtime losses due to noise problems, ultimately achieving a globally optimal solution with the lowest total project cost and shortest total construction period.

[0169] 4. This invention deeply integrates modern information technologies such as IoT monitoring, big data analysis, artificial intelligence prediction, and operations research optimization into the traditional civil engineering construction process, constructing a complete data closed loop of "monitoring-analysis-decision-execution". This transforms construction management from relying on experienced workers' experience to relying on data and scientific algorithms, improving the accuracy, reliability, and efficiency of decision-making. It represents the inevitable direction for the construction industry to transform and upgrade towards industrialization, digitalization, and intelligence. Attached Figure Description

[0170] Figure 1 This is a process flow diagram of the intelligent pouring and curing method for large-volume concrete used in pile-raft foundation construction proposed in this invention. Detailed Implementation

[0171] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with existing known technologies. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0172] I. Design of the Invention

[0173] The research on quality control during pouring or curing and construction noise reduction is not only directly related to the quality and safety of the project itself and the harmony and stability of the surrounding social environment, but also an inevitable path for the construction industry to respond to the call for green construction, enhance its technological content, and achieve sustainable development. It plays an indispensable and key role in promoting technological progress in the industry.

[0174] Meanwhile, quality control during pouring or curing and noise reduction during construction are not two independent issues, but rather a unified system that influences each other and requires coordinated optimization. High-noise operations not only disturb residents but may also directly interfere with the quality and monitoring of temperature-controlled curing. Conversely, if unexpected events occur during temperature-controlled curing, it may create additional noise sources for nighttime emergency repairs, mainly in the following aspects:

[0175] 1) Construction noise control may interfere with the monitoring system:

[0176] Sound and vibration pollution: Strong sound waves are accompanied by mechanical vibrations, especially from hammering piles, blasting, and heavy vehicle traffic. This vibration can interfere with high-precision temperature sensors, causing momentary jumps or drifts in readings, affecting data quality, and even triggering false warnings. Electromagnetic interference: The starting and stopping of high-power electrical equipment (such as welding machines and large pile drivers) may generate electromagnetic pulses, interfering with the signal transmission of wireless sensors.

[0177] 2) Potential interference of construction noise control with maintenance operations:

[0178] Personnel Intervention Obstacles: Maintenance operations (such as covering with insulation, checking for leaks, and adjusting sprinkler heads) require technicians to approach or even enter the maintenance area. If high-decibel, high-risk work is being carried out nearby (such as pile driving), it poses a safety hazard, forcing maintenance work to be delayed or simplified, affecting maintenance quality. Equipment Conflicts: The access routes of high-noise equipment (such as concrete mixer trucks and pump trucks) may conflict with the layout of maintenance pipelines (sprinkler pipes, cables), potentially crushing or severing the pipelines and causing maintenance system failure. Interference with Data Interpretation: Without coupled analysis, an abnormal temperature sensor reading caused by vibration may be misinterpreted as an abnormal heat of hydration within the concrete, leading to incorrect maintenance decisions.

[0179] Based on the above mechanism, the following research and experimental methods were designed:

[0180] For large-volume concrete projects, material testing methods are required. Indoor experiments should be conducted to optimize the mix proportion of low-heat, high-performance concrete incorporating admixtures such as fly ash and mineral powder, thereby reducing the peak heat of hydration. A coupled temperature-stress field model should be established using physical simulation and numerical simulation to predict the temperature and stress development patterns under different curing conditions. Crucially, the Internet of Things (IoT) monitoring method should be introduced. Wireless temperature sensors should be embedded in the physical project to collect real-time internal concrete temperature data. Based on big data analysis, an intelligent early warning and feedback system should be built to dynamically optimize heat preservation and moisture retention curing strategies, thus elevating traditional experience-based curing to data-driven precision control. For construction noise reduction, a comparative technical analysis method should be used to comprehensively evaluate the applicability, economy, and reliability of low-noise technologies such as static pressure piles and rotary drilling. Acoustic measurement and modeling methods should be employed to monitor the noise reduction effects of different equipment and barriers on-site, establishing a sound propagation model to optimize the design and layout of sound barriers. Simultaneously, operations research methods from management science should be incorporated to scientifically plan the temporal and spatial layout of high-noise operations, minimizing the impact on sound-sensitive areas. Ultimately, the research in both areas needs to be integrated and verified through field trials to construct a closed loop of "monitoring-analysis-decision-execution" in actual engineering, forming scalable technical guidelines and standard processes. The technical route is as follows: Figure 1 As shown.

[0181] II. Project Schedule of this Invention

[0182] Implementation Location: Construction Project Department of Energy Storage Tank Project, No. 1 Zero-Carbon Energy Station, North District, Shanxi Road & Bridge Municipal Engineering Co., Ltd.

[0183] Schedule: November 2025: Preliminary preparation and scheme design phase, focusing on completing literature reviews and technical surveys in two main directions, clarifying the selection and installation scheme of the intelligent temperature control system, assessing the applicability of noise reduction technology, and developing a detailed concrete mix design test plan and noise reduction effect simulation scheme. December 2025: Core testing and data collection phase, conducting on-site concrete pouring tests, embedding sensors and monitoring temperature field changes in real time, and simultaneously implementing the skip-pour method process verification; at the same time, conducting noise comparison tests between static pressure piles and traditional processes, collecting actual acoustic data under different noise reduction measures, and comprehensively collecting first-hand data on curing and noise reduction. January 2026: Data analysis and results summary phase, processing and analyzing temperature and noise monitoring data, evaluating the measured effects of optimization measures, completing the writing of the phase research report, and developing a draft of standardized operating guidelines applicable to subsequent projects based on the research results, laying the foundation for the promotion and application of the technology.

[0184] III. Basic Maintenance Methods

[0185] The intelligent pouring and curing method for large-volume concrete used in pile-raft foundation construction includes the following steps:

[0186] S1: Construct an integrated monitoring system by arranging a wireless temperature sensor array inside the concrete structure and a noise monitoring instrument at the construction site boundary to achieve synchronous acquisition of temperature field and acoustic environment data.

[0187] S2: Establish a data-driven intelligent prediction model for concrete temperature field. Based on the time-series data collected in S1, predict the future development of the temperature field and generate early warning of excessive temperature difference and intelligent maintenance suggestions accordingly.

[0188] The LSTM model consists of the following three-layer architecture:

[0189] 1) Engineering data interface (input and output)

[0190] Input vector x t :

[0191]

[0192] 2) Internal computation of LSTM (mathematical abstraction process)

[0193] Cell state: ;

[0194] Forgot Gate Output: ;

[0195] Input gate output: ;

[0196] Candidate cell status: ;

[0197] Output gate: ;

[0198] The formula for updating the hidden state is:

[0199] ;

[0200] 3) Final prediction layer

[0201] The high-level abstract features h extracted by LSTM t Mapped to a specific physical quantity—future temperature predictions. :

[0202] ;

[0203] like: The model outputs the predicted temperature values ​​from all sensors for the next m time steps. .

[0204] Loss function: The loss function is minimized through backpropagation using a time-based algorithm and an optimizer (such as Adam). This updates all parameters.

[0205] .

[0206] S3: Applying low-heat, high-crack-resistant concrete preparation technology, by compounding large-volume admixtures, high-performance additives and fibers, the heat of hydration is reduced and crack resistance is improved from the material source.

[0207] Low-heat, high-crack-resistant concrete comprises the following raw materials:

[0208] Cementitious materials include: ordinary Portland cement, grade 42.5, 200-230 kg / m³. 3 Grade II fly ash, 90-110 kg / m³ 3 S95 grade mineral powder, 70-90 kg / m³ 3 ;

[0209] Aggregate sand, including fine aggregate, 700-750 kg / m³ 3 Coarse aggregate, 1020-1080 kg / m³ 3 It consists of 5-25mm continuously graded crushed stone;

[0210] Chemical admixtures, including polycarboxylate-based high-performance water-reducing agents, 4.0-5.5 kg / m³ 3 Polypropylene fiber, 0.6-0.9 kg / m 3 Length: 12-19mm; Water for mixing: 155-165kg / m 3 ;

[0211] Key mixing parameters: Water-cement ratio (W / B): 0.38-0.42; Total amount of cementitious material: 360-430 kg / m³ 3 Total admixture content: 40%-50% of the mass of cementitious materials; Sand ratio: 40%-42%.

[0212] S4: Implement a low-noise construction technology system, including the application of static pressure piles under complex geological conditions, the deployment of modular mobile sound barriers, and the intelligent scheduling of high-noise operations based on noise monitoring data from S1.

[0213] Noise propagation optimization algorithm:

[0214] 1) Noise propagation model ;

[0215] 2) Superposition of multiple noise sources

[0216] ;

[0217] Optimization based on genetic algorithms:

[0218] Decision variable, project duration: ;

[0219] Objective function: ;

[0220] Constraints, Noise: Daytime ≤70dB; Nighttime ≤55dB; inter-device operation logic constraints (e.g., device A must start after device B has finished); devices cannot occupy the same space at the same time;

[0221] Genetic algorithm operations: selection, crossover, mutation;

[0222] 3) Multi-objective collaborative optimization algorithm

[0223] Objective 1, Temperature Quality:

[0224] ;

[0225] Objective 2, Noise Impact:

[0226] ;

[0227] Objective 3, Economic Costs:

[0228] ;

[0229] Construction scheme Y will simultaneously affect three types of costs: temperature control, noise control, and construction delay. The total cost f3(Y) is a linear sum of these three types of costs. The optimization objective is to minimize the total cost by adjusting Y. This approach is suitable for scenarios involving the economic efficiency of construction schemes and collaborative optimization of multiple constraints.

[0230] S5: Construct a multi-objective collaborative optimization model with constraints such as concrete internal and external temperature difference ≤25℃, crack width <0.2mm, and field boundary noise ≤70dB(A) during the day and ≤55dB(A) at night. The objective functions are minimizing the total project cost and minimizing the total construction period. Solve for the optimal combination of construction parameters to achieve global optimization of quality, environmental protection, cost, and construction period. This includes non-dominated sorting, congestion calculation, collaborative constraint processing, and decision-making process.

[0231] IV. Implementation and Verification of Maintenance Methods

[0232] Based on the above basic maintenance methods, the following embodiments and comparative examples are designed:

[0233] Example 1

[0234] Materials: Low-heat, high-crack-resistant mix design (cement 210kg / m³) 3 fly ash 105kg / m³ 3 Mineral powder 85kg / m 3 Fiber 0.8kg / m 3 (Water-to-binder ratio 0.40).

[0235] Maintenance: Activate the LSTM-based intelligent predictive maintenance system (early warning threshold 23℃, control threshold 25℃).

[0236] Noise reduction: Static pressure piles + modular mobile sound barriers + intelligent scheduling system based on real-time monitoring are used.

[0237] Optimization: The construction parameters were globally optimized using a multi-objective collaborative optimization model (NSGA-II).

[0238] Example 2

[0239] Materials: Adopting a reinforced low-heat, high-crack-resistance mix design (cement 200kg / m³) 3 fly ash 110kg / m³ 3 Mineral powder 90kg / m 3 Fiber 0.6kg / m 3 (Water-to-binder ratio 0.38); the rest is the same as in Example 1.

[0240] Example 3

[0241] Materials: Standard ready-mixed concrete mix design (cement 230kg / m³) 3 fly ash 90kg / m³ 3 70kg / m³ of mineral powder 3 Fiber 0.9kg / m 3 (Water-to-binder ratio 0.42); the rest is the same as in Example 1.

[0242] Comparative Example 1

[0243] Materials: Standard ready-mixed concrete mix design (cement 320kg / m³) 3 70kg / m³ of mineral powder 3 Fiber 0.8kg / m 3 (Water-to-binder ratio 0.42); the rest is the same as in Example 1.

[0244] Comparative Example 2

[0245] Materials: Standard ready-mixed concrete mix design (cement 320kg / m³) 3 70kg / m³ of mineral powder 3 (Water-to-binder ratio 0.42); the rest is the same as in Example 1.

[0246] Comparative Example 3

[0247] Maintenance: Traditional maintenance methods are used (regular watering and covering based on experience); the rest is the same as in Example 1.

[0248] Comparative Example 4

[0249] Maintenance: Activate the LSTM-based intelligent predictive maintenance system (early warning threshold 22℃, control threshold 24℃); Noise reduction: replace static pressure piles with impact piles; the rest is the same as in Example 1.

[0250] Comparative Example 5

[0251] Maintenance: LSTM-based intelligent predictive maintenance system is enabled (early warning threshold 22℃, control threshold 24℃); Noise reduction: No intelligent scheduling; the rest is the same as in Example 1.

[0252] V. Performance Analysis

[0253] The curing processes and product performance obtained in Examples 1-3 and Comparative Examples 1-4 were evaluated, as shown in Table 1 below:

[0254] Table 1. Effects of Formula and Algorithm on Maintenance Effect

[0255]

[0256] The curing processes and product performance obtained from Examples 1-3 and Comparative Examples 1-4 were evaluated, as shown in Table 1 below: Maximum temperature rise and maximum internal and external temperature difference: Comparing Example 1 with Comparative Example 1, it is evident that the material is the foundation for temperature control. Furthermore, Example 2 achieved the lowest temperature rise through the maximum amount of admixture; Comparative Example 1 experienced the highest temperature rise due to high cement content. Comparing Example 1 with Comparative Example 1 also demonstrates that intelligent curing is the core of temperature difference control. Additionally, Example 3 achieved the lowest temperature difference through precise LSTM control; Comparative Example 1 experienced the largest temperature difference due to uncontrolled heat source.

[0257] Crack resistance: Comparison of Example 1 and Comparative Example 2 shows that the material and intelligent scheduling work together to ensure no cracks and prevent fiber deterioration: Examples 1 / 3 have no cracks under system protection; Comparative Example 2 has the most severe cracks in the absence of fibers.

[0258] Noise control: Comparing Example 1 and Comparative Example 4, it is shown that noise source control is the key, and the comprehensive measures in Example 1 have the best effect; Comparative Example 4 seriously exceeded the standard due to the uncontrolled noise source of the impact pile; It also shows that intelligent scheduling can ensure compliance at night: Example 1 strictly controls noise through intelligent scheduling; Comparative Example 4 and Comparative Example 5 both failed to meet the strict nighttime limits for different reasons.

[0259] Temperature Field Control: Comparing Example 1 and Comparative Example 3, Comparative Example 3 disables LSTM intelligent prediction and adopts traditional timed, experience-based curing. Traditional curing is passive and experience-driven, resulting in severe lag and inaccuracy. It cannot predict upcoming temperature changes within the concrete and can only take action after problems (such as surface whitening and cracking, or excessive temperature differences shown by the thermometer) occur, which is too late. In contrast, the LSTM model in Example 1 is "actively predictive" and "data-driven," and its internal C... t (Cellular state) has memorized long-term hydration heat release patterns, h t (Hidden state) captures short-term dynamic changes, enabling forward-looking decisions. The failure of Comparison Example 3 highlights the necessity of a paradigm shift from "experience-based maintenance" to "intelligent maintenance."

[0260] Overall Cost and Duration: Comparing Example 1 and Comparative Example 2, it is shown that collaborative optimization can achieve economic efficiency: Example 1 is globally optimal; Comparative Example 2 has the highest post-treatment cost and longest duration due to quality issues.

[0261] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for intelligent pouring and curing of large-volume concrete for pile-raft foundation construction, characterized in that, Includes the following steps: S1: Construct an integrated monitoring system by arranging a wireless temperature sensor array inside the concrete structure and a noise monitoring instrument at the construction site boundary to achieve synchronous acquisition of temperature field and acoustic environment data. S2: Establish a data-driven intelligent prediction model for concrete temperature field. Based on the time series data collected in S1, predict the future development of temperature field and generate early warning of excessive temperature difference and intelligent maintenance suggestions accordingly. S3: Applying low-heat, high-crack-resistant concrete preparation technology, by compounding large-volume admixtures, high-performance additives and fibers, the heat of hydration is reduced and crack resistance is improved from the material source. S4: Implement a low-noise construction technology system, including the application of static pressure piles under complex geological conditions, the deployment of modular mobile sound barriers, and the intelligent scheduling of high-noise operations based on noise monitoring data from S1. S5: Construct a multi-objective collaborative optimization model with constraints such as concrete internal and external temperature difference ≤25℃, crack width <0.2mm, and field boundary noise ≤70dB(A) during the day and ≤55dB(A) at night. The objective functions are minimizing the total project cost and minimizing the total construction period. Solve for the optimal combination of construction parameters to achieve global optimization of quality, environmental protection, cost and construction period.

2. The intelligent pouring and curing method for large-volume concrete used in pile-raft foundation construction as described in claim 1, characterized in that, The intelligent prediction model in S2 is a Long Short-Term Memory (LSTM) network model. Its inputs include historical temperature sequences, ambient temperature and humidity, and static characteristics of concrete mix proportions. The output is the predicted internal temperature of the concrete for a specific future time window. When the predicted internal and external temperature difference exceeds 23°C, an early warning is triggered. When it exceeds 25°C, an active curing control command is triggered.

3. The intelligent pouring and curing method for large-volume concrete used in pile-raft foundation construction as described in claim 2, characterized in that, The LSTM model includes the following three-layer architecture: 1) Engineering data interface input vector x t : T N,t : Temperature measurement of the Nth sensor at time step t; this is the most central input; Ta t : Ambient temperature at time step t; RH t : Relative humidity of the environment at time step t; S c : Static feature vector, containing: c: cement dosage, w / c: water-cement ratio, pfa: fly ash content, raft thickness, sensor burial depth; 2) Internal computation of LSTM The LSTM model includes a forget gate, an input gate, and an output gate. The formula for the cell state at time step t is: ； f t The output of the forget gate is a vector with values ​​between (0,1), activated by the sigmoid function: ； The Sigmoid function is: h t-1 x represents the LSTM hidden state at the previous time step t-1. t Given the input sequence elements at the current time step t, concatenate the two columns into a long vector, which is [h] t-1 ,x t ] is used as the input for the forget gate; W f Let b be the weight matrix of the forget gate. f This is the bias vector for the forget gate; C t-1 It is the cell state at the previous time step (historical memory). i t The output value of the input gate is a vector with values ​​between (0,1), activated by the Sigmoid function: ； Among them W i Let b be the weight matrix of the forget gate. i This is the bias vector for the forget gate; The candidate cell state is a vector with values ​​between (-1, 1), activated by the tanh function: ； Where tanh is the hyperbolic tangent activation function, the formula is: W C Let b be the weight matrix of the candidate cell states. C is the bias vector for the candidate cell state; Output gate o t It is a vector with values ​​between (0,1), activated by the Sigmoid function: ； Among them W o Let b be the weight matrix of the output gate. o This is the bias vector for the output gate; The formula for updating the hidden state is: ； Element-wise multiplication; 3) Final prediction layer ； Among them W y Let b be the weight matrix of the prediction layer. y The bias vector of the prediction layer; This is the final prediction output of the model.

4. The intelligent pouring and curing method for large-volume concrete used in pile-raft foundation construction as described in claim 3, characterized in that, The LSTM model also includes model training and loss function settings: Loss function: ； N is the number of sensors, and m is the prediction time step. Let be the actual temperature measured by the i-th sensor at time step t+j. For the predicted temperature measured by the i-th sensor at time step t+j, the loss function is minimized through backpropagation using a time-based algorithm and an optimizer (such as Adam). This updates all parameters. 。 5. The intelligent pouring and curing method for large-volume concrete used in pile-raft foundation construction as described in claim 1, characterized in that, The low-heat, high-crack-resistant concrete in S3 comprises the following raw materials: Cementitious materials, including: Ordinary Portland cement, grade 42.5, 200-230 kg / m³ 3 It provides the main strength, but controlling its dosage is key to reducing the heat of hydration; Grade II fly ash, 90-110 kg / m³ 3 ; S95 grade mineral powder, 70-90 kg / m³ 3 ; Aggregate sand, including: Fine aggregate, 700-750 kg / m³ 3 ; Coarse aggregate, 1020-1080 kg / m³ 3 ; Chemical admixtures, including: Polycarboxylate-based high-performance water-reducing agent, 4.0-5.5 kg / m³ 3 In terms of solid content; Polypropylene fiber, 0.6-0.9 kg / m 3 Length: 12-19mm; Water for mixing: 155-165 kg / m³ 3 .

6. The intelligent pouring and curing method for large-volume concrete used in pile-raft foundation construction as described in claim 1, characterized in that, The intelligent scheduling in S4 specifically involves establishing a linkage mechanism between noise monitoring data and the operating status of construction equipment. When the real-time monitoring value of site boundary noise approaches 70dB, the system automatically sends an alarm to the dispatch center and recommends or executes strategies such as suspending some high-noise operations, adjusting equipment positions, or activating backup noise reduction measures.

7. The intelligent pouring and curing method for large-volume concrete used in pile-raft foundation construction as described in claim 6, characterized in that, The intelligent scheduling includes the following noise propagation optimization algorithm: 1) Noise propagation model ; L p,AB Let L be the sound pressure level produced by noise source A at receiving point B, describing the noise intensity contributed by a specific noise source A at receiving point B; w,A The sound power level of noise source A; r AB Let A be the distance from noise source A to receiving point B. The sound pressure level attenuation term is caused by the geometric diffusion of spherical waves; α is the air absorption coefficient; αr AB This is the sound pressure level attenuation term caused by air absorption; Added attenuation to the sound pressure level caused by the sound barrier; 2) Superposition of multiple noise sources The total sound pressure level produced by all n noise sources A at receiver point B is: ; Optimization based on genetic algorithms: The noise control problem is defined as an optimization problem and solved using a genetic algorithm. Decision variable, project duration: ； t k s is the start time of operation for device k; k Let k be the operating position of device k; Objective function: ； w1 and w2 are weighting coefficients; Constraints, Noise: Daytime ≤70dB; at night ≤55dB; Equipment room operation logic constraints; Devices cannot occupy the same space at the same time; Genetic algorithm operation: Selection: Roulette wheel selection method, solutions with higher fitness (smaller f(X)) have a greater chance of proliferating; Crossover: Single-point crossing, where partial gene sequences of two parent chromosomes are exchanged to produce a new individual; Mutation: A random change to the start time or location of a device with a small probability; 3) Multi-objective collaborative optimization algorithm This is the system's top-level command center, responsible for balancing and coordinating the sometimes conflicting goals of temperature control and noise control; Problem Modeling Objective 1, Temperature Quality: ； Objective 2, Noise Impact: ； Objective 3, Economic Costs: ； Where t is the time step, B is the noise receiving point, and Y is the collaborative decision variable, including maintenance plan and equipment scheduling plan; f3(Y) is the total cost function related to construction scheme Y, C temp For temperature control costs, C noise For noise control costs, C delay Costs related to construction delays.

8. The intelligent pouring and curing method for large-volume concrete used in pile-raft foundation construction as described in claim 1, characterized in that, The multi-objective collaborative optimization model in S5 is solved using a non-dominated sorting genetic algorithm with an elite strategy. The output is a set of Pareto optimal construction schemes that balance crack resistance, noise control, cost, and construction period, which are then selected by decision-makers.

9. The intelligent pouring and curing method for large-volume concrete used in pile-raft foundation construction as described in claim 8, characterized in that, The non-dominated sorting genetic algorithm with an elitist strategy includes the following process: Non-dominated sorting: Stratify the solution population so that the solution that is no worse than other solutions on any of the objectives f1, f2, f3, and is better on at least one objective is ranked first (Rank1). Crowding calculation: Within the same Rank, calculate the density of each solution in the target space, and prioritize retaining solutions in sparse regions to maintain population diversity; Cooperative constraint handling: Hard constraint: The internal and external temperature difference ΔT(t) between the concrete core and surface ≤ 25℃, noise L p,total,B ≤70dB or ≤55dB; solutions that do not meet these requirements are directly eliminated. Soft constraint / penalty term: If high-noise operations at night greatly interfere with concrete curing and temperature monitoring personnel, then a penalty term P is added to the objective function. night ; Decision: The algorithm will eventually output a set of Pareto optimal solutions.

10. The intelligent pouring and curing method for large-volume concrete used in pile-raft foundation construction as described in claim 8, characterized in that, The application of the non-dominated sorting genetic algorithm with an elitist strategy in engineering decision-making includes the following process: I. Algorithm Objective Under the premise of satisfying all hard constraints including temperature difference, cracks, and noise, a series of construction schemes that make the best trade-off between the two conflicting objectives of total cost and total construction period are found. The set of these schemes is called the Pareto optimal solution set, and its mapping in the objective space is called the Pareto front. II. NSGA-II Solution Process 1) Initialize the population Y initial "construction plans" are randomly generated. Each plan includes the concrete mix proportion number, curing start time, number of static pressure pile drivers used, number of blocks for skip-pile method, and time period arrangement for high-noise operation. 2) Evaluating individuals For each Y in the population, calculate its objective function value and constraint violation degree, input the scheme Y into the surrogate model, and simulate its execution consequences; calculate the objective values ​​f1, f2, f3 and their constraint violation degrees; 3) Non-dominated sorting The entire population is stratified according to Pareto dominance; According to the conditions, if and only if: Y1 is no worse than Y2; At least on one objective, Y1 is significantly better than Y2; The dominance relationship is determined as follows: Option Y1 dominates Option Y2; Rank 1: Identify all individuals that are not dominated by any other individual; this is called the first non-dominated layer. Rank 2: From the remaining individuals, find all the non-dominated individuals, which is called the second non-dominated layer; Repeat this process until all individuals have been stratified; 4) Crowding Calculation Within the same non-dominated layer, calculate the density of each individual in its objective function space; Calculation: For each objective function, calculate the difference in function values ​​between the individual and its two adjacent individuals, and sum them; individuals with higher crowding will be retained first. 5) Selection, crossover, and mutation Selection: Based on non-dominated ranking and crowding, a binary tournament selection method is adopted; that is, two individuals are randomly selected, and the one with the smaller rank is selected first; if the ranks are the same, the one with the larger crowding is selected. Crossover: Randomly select two "parent" individuals, exchange some of their genes, and produce new "offspring" individuals; Mutation: A small probability of randomly changing the value of a certain gene in an individual; 6) Elite Strategy The parent and offspring populations are merged into a single population of size 2N. The non-dominated sorting and crowding calculation are recalculated for this merged population, and then only the top N individuals are retained as the parents for the next iteration. III. Constraint Handling NSGA-II typically uses the constraint dominance principle to handle hard constraints: If both individuals satisfy all constraints, then the comparison is performed according to the original Pareto dominance relationship; If one individual satisfies the constraints while another violates them, the individual that satisfies the constraints is always considered superior. If both individuals violate the constraints, the individual with the smaller degree of constraint violation is considered the better one; IV. Final Output and Application Once the algorithm reaches its maximum number of iterations, all individuals in the first non-dominated layer (Rank1) will be output, which is the Pareto optimal solution set. Decision support: Make the final decision based on the actual priority of the current project within the Pareto optimal solution set.

Images