A method and system for optimizing the mix ratio of ultra-early-strength concrete based on high-dimensional multi-objective optimization

Abstract

The present invention relates to the technical field of concrete mix ratio optimization, and proposes an ultra-early-strength concrete mix ratio optimization method and system based on high-dimensional multi-objective optimization, which comprises the following steps: obtaining an experimental data set of ultra-early-strength concrete mix ratio; establishing a multi-layer feedforward neural network model, inputting the experimental data set into the multi-layer feedforward neural network model and training it to full convergence based on an error back propagation algorithm to obtain a nonlinear mapping relationship between the mix ratio corresponding to the experimental data set and the 6-hour compressive strength; constructing a high-dimensional multi-objective optimization model according to the nonlinear mapping relationship between the mix ratio corresponding to the experimental data set and the 6-hour compressive strength; solving the high-dimensional multi-objective optimization model to obtain a Pareto optimal solution set, and outputting the Pareto optimal solution set as an ultra-early-strength concrete mix ratio optimization scheme, which can effectively improve the efficiency of the ultra-early-strength concrete mix ratio optimization work and simultaneously ensure the optimal comprehensive benefit of the ultra-early-strength concrete mix ratio.

Description

Technical Field

[0001] The present invention relates to the technical field of concrete mix ratio optimization, and more specifically, to a method and system for optimizing the mix ratio of ultra-early-strength concrete based on high-dimensional multi-objective optimization. Background Art

[0002] Ultra-early-strength concrete is a special type of concrete that incorporates admixtures into conventional concrete formulas to accelerate chemical reactions, allowing the concrete to reach a certain strength in the shortest possible time to meet engineering requirements. It is generally considered ultra-early-strength concrete when its one-day strength exceeds 50% of the design strength grade and its three-day strength reaches 90% to 100% of the design strength grade, shortening the concrete maturation period from 28 days to 3 days. Ultra-early-strength concrete has broad application prospects in the maintenance of projects such as highways, bridges, and seaports. It can be used for both wartime emergency repairs and rapid peacetime repair and construction, such as bridge closures and highway or airport runway repairs.

[0003] Compared with ordinary concrete, ultra-early strength concrete has special requirements in terms of compressive and tensile strength, stability, durability and workability. The proportion of various raw materials (such as water-cement ratio) has a decisive influence on the performance of ultra-early strength concrete. The current preparation method of ultra-early strength concrete mainly relies on experience and a large number of repeated experiments to continuously adjust the quality of the formula. In the proportioning experiment, it is generally necessary to make multiple batches of specimens for compressive and tensile strength testing. These experiments usually require a large amount of raw materials (such as water, cement, sand and gravel, admixtures, etc.), human resources and working time. Summary of the Invention

[0004] In order to overcome the current defects of ultra-early-strength concrete mix ratio optimization that relies on experience and a large number of repeated experiments, resulting in high resource consumption and low work efficiency, the present invention provides an ultra-early-strength concrete mix ratio optimization method based on high-dimensional multi-objective optimization, and an ultra-early-strength concrete mix ratio optimization system based on high-dimensional multi-objective optimization.

[0005] In order to solve the above technical problems, the technical solutions of the present invention are as follows:

[0006] A method for optimizing the mix ratio of ultra-early-strength concrete based on high-dimensional multi-objective optimization comprises the following steps:

[0007] S1. Obtain the experimental data set of ultra-early strength concrete mix proportions;

[0008] S2. Establish a multi-layer feedforward neural network model, input the experimental data set into the multi-layer feedforward neural network model and train it based on the error back propagation algorithm until it fully converges, to obtain a nonlinear mapping relationship between the ratio corresponding to the experimental data set and the 6-hour compressive strength;

[0009] S3. Construct a high-dimensional multi-objective optimization model based on the nonlinear mapping relationship between the ratio corresponding to the experimental data set and the 6-hour compressive strength;

[0010] S4. Solve the high-dimensional multi-objective optimization model to obtain a Pareto optimal solution set, and output the Pareto optimal solution set as an ultra-early strength concrete mix ratio optimization solution.

[0011] In this technical solution, the mix proportion of ultra-early-strength concrete is improved based on an artificial neural network model and a high-dimensional multi-objective optimization algorithm. A multi-layer feedforward neural network model is used to obtain a nonlinear mapping relationship between the mix proportion and the 6-hour compressive strength, and the nonlinear mapping relationship is further constructed into a high-dimensional multi-objective optimization model. The high-dimensional multi-objective optimization model is solved using an optimization algorithm, and the Pareto optimal solution is obtained as the mix proportion optimization result output, which can effectively improve the efficiency of the ultra-early-strength concrete mix proportion optimization work.

[0012] Furthermore, the present invention also proposes an ultra-early-strength concrete mix ratio optimization system based on high-dimensional multi-objective optimization, which is applied to the above-mentioned ultra-early-strength concrete mix ratio optimization method.

[0013] The ultra-early-strength concrete mix ratio optimization system proposed in the present invention based on high-dimensional multi-objective optimization includes:

[0014] Data acquisition module, used to obtain experimental data sets of ultra-early strength concrete mix proportions;

[0015] A neural network module, comprising a multi-layer feedforward neural network model for iterative training using the experimental data set collected by the data acquisition module, and for generating a nonlinear mapping relationship between the ultra-early strength concrete mix ratio and the 6-hour compressive strength corresponding to the experimental data set;

[0016] A multi-objective optimization module includes a high-dimensional multi-objective optimization model and an optimization solution unit; the high-dimensional multi-objective optimization model is constructed or updated according to the nonlinear mapping relationship between the proportion corresponding to the experimental data set and the 6-hour compressive strength; the optimization solution unit is used to solve the high-dimensional multi-objective optimization model using an optimization algorithm to obtain a Pareto optimal solution set, and output the Pareto optimal solution set as an ultra-early strength concrete proportion optimization scheme.

[0017] Compared with the prior art, the beneficial effect of the technical solution of the present invention is: the present invention improves the mix ratio of ultra-early strength concrete based on an artificial neural network model and a high-dimensional multi-objective optimization algorithm, and uses the nonlinear mapping relationship between the mix ratio obtained by fitting the multi-layer feedforward neural network model and the 6-hour compressive strength to construct a high-dimensional multi-objective optimization model, and then uses the optimization algorithm to perform the optimization solution, and the Pareto optimal solution set obtained by the solution is output as the mix ratio optimization result, which can effectively improve the efficiency of the ultra-early strength concrete mix ratio optimization work, and at the same time ensure that the comprehensive benefits of the ultra-early strength concrete mix are optimal. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 This is a flow chart of the ultra-early strength concrete mix ratio optimization method based on high-dimensional multi-objective optimization in Example 1.

[0019] Figure 2 This is a flow chart of the multi-layer feedforward neural network model training in Example 2.

[0020] Figure 3 This is the architecture diagram of the BP neural network model of Example 3.

[0021] Figure 4 Schematic diagram of BP neural network linear regression in Example 3.

[0022] Figure 5 This is the main loop flow chart of the Borg algorithm in Example 3.

[0023] Figure 6 A parallel coordinate graph with a Pareto optimal solution set is drawn for Example 3.

[0024] Figure 7 This is an architecture diagram of the ultra-early-strength concrete mix ratio optimization system based on high-dimensional multi-objective optimization in Example 4. DETAILED DESCRIPTION

[0025] The accompanying drawings are for illustrative purposes only and are not to be construed as limiting this patent;

[0026] In order to better illustrate this embodiment, some parts in the drawings may be omitted, enlarged, or reduced, and do not represent the actual product size;

[0027] It is understandable to those skilled in the art that some well-known structures and descriptions thereof may be omitted in the drawings.

[0028] The technical solution of the present invention is further described below with reference to the accompanying drawings and embodiments.

[0029] Example 1

[0030] This embodiment proposes a method for optimizing the mix ratio of ultra-early-strength concrete based on high-dimensional multi-objective optimization. Figure 1, which is a flow chart of the ultra-early strength concrete mix ratio optimization method based on high-dimensional multi-objective optimization in this embodiment.

[0031] The ultra-early-strength concrete mix ratio optimization method based on high-dimensional multi-objective optimization proposed in this embodiment includes the following steps:

[0032] S1. Obtain the experimental dataset of ultra-early strength concrete mix proportions.

[0033] S2. Establish a multi-layer feedforward neural network model, input the experimental data set into the multi-layer feedforward neural network model and train it based on the error back propagation algorithm until it fully converges, and obtain a nonlinear mapping relationship between the ratio corresponding to the experimental data set and the 6-hour compressive strength.

[0034] S3. Based on the nonlinear mapping relationship between the ratio corresponding to the experimental data set and the 6-hour compressive strength, a high-dimensional multi-objective optimization model is constructed.

[0035] S4. Solve the high-dimensional multi-objective optimization model to obtain a Pareto optimal solution set, and output the Pareto optimal solution set as an ultra-early strength concrete mix ratio optimization solution.

[0036] In an optional embodiment, the experimental data set of ultra-early-strength concrete mix ratio is obtained by collecting all ultra-early-strength concrete mix ratio experimental data of local laboratories, and optionally, by targeted crawling of ultra-early-strength concrete mix ratio data publicly available on the Internet using a focused web crawler technology.

[0037] Among them, the experimental data of ultra-early strength concrete mix ratio includes the amount of cementitious materials, coarse aggregate, fine aggregate, liquid and admixtures, as well as the corresponding 6-hour compressive strength.

[0038] In an optional embodiment, the experimental data set obtained in step S1 will continue to expand over time, and the multi-layer feedforward neural network model will be iteratively trained using the experimental data set, and the number of layers in the network model and the number of neurons in each layer will be repeatedly adjusted until a training result with stable convergence and reasonable error is obtained. In this way, a nonlinear mapping relationship from the proportion of various raw materials of concrete to the 6-hour compressive strength can be established in a more accurate manner. That is, as long as the amount of each raw material is given, the early compressive strength of ultra-early strength concrete can be inferred.

[0039] In this embodiment, the mix ratio of ultra-early-strength concrete is improved based on an artificial neural network model and a high-dimensional multi-objective optimization algorithm, wherein a multi-layer feedforward neural network model is used to obtain a nonlinear mapping relationship between the mix ratio and the 6-hour compressive strength, and the nonlinear mapping relationship is further constructed into a high-dimensional multi-objective optimization model. The high-dimensional multi-objective optimization model is solved by an optimization algorithm, and the Pareto optimal solution is obtained as the mix ratio optimization result output, which can effectively improve the efficiency of the ultra-early-strength concrete mix ratio optimization work and ensure that the comprehensive benefits of the ultra-early-strength concrete mix are optimal.

[0040] Example 2

[0041] This embodiment improves upon the ultra-early-strength concrete mix ratio optimization method proposed in Example 1, and further proposes an ultra-early-strength concrete mix ratio optimization method based on high-dimensional multi-objective optimization.

[0042] The ultra-early-strength concrete mix ratio optimization method based on high-dimensional multi-objective optimization proposed in this embodiment includes the following steps:

[0043] S1. Obtain the experimental dataset of ultra-early strength concrete mix proportions.

[0044] In this embodiment, the experimental data set is obtained by collecting all ultra-early strength concrete mix ratio experimental data of local laboratories and / or by using a focused web crawler technology to crawl the ultra-early strength concrete mix ratio data publicly available on the Internet.

[0045] S2. Establish a multi-layer feedforward neural network model, input the experimental data set into the multi-layer feedforward neural network model and train it based on the error back propagation algorithm until it fully converges, and obtain a nonlinear mapping relationship between the ratio corresponding to the experimental data set and the 6-hour compressive strength.

[0046] The multi-layer feedforward neural network model in this embodiment includes an input layer, at least one hidden layer, and an output layer. Each layer in the multi-layer feedforward neural network model contains a number of neurons (nodes), and each layer is interconnected through neurons, and each layer uses the output of the previous layer as its input.

[0047] In addition, there is evidence that a two-layer (i.e., only one hidden layer) feedforward neural network with enough neurons can fit any data with arbitrary accuracy, and feedforward neural networks are good at solving the fitting problem of nonlinear functions.

[0048] In an optional embodiment, the number of neurons included in the input layer is equal to the number of raw materials in the ultra-early-strength concrete mix. The number of neurons included in the output layer is equal to the number of key performance parameters in the ultra-early-strength concrete mix. The number of hidden layers and the number of neurons in each layer are adjusted based on the training results of the multi-layer feedforward neural network model.

[0049] Furthermore, the step of inputting the experimental data set into a multi-layer feedforward neural network model and training the model to sufficient convergence based on an error back propagation algorithm includes:

[0050] S2.1. Normalize the data in the experimental dataset and map all data to the interval [-1, 1].

[0051] S2.2. Set the training parameters of the multi-layer feedforward neural network model.

[0052] In an optional embodiment, the training parameters include data classification method (i.e., how to divide the training set, validation set and test set), training algorithm (different types of error back propagation algorithms), neural network performance evaluation method and gradient calculation method.

[0053] S2.3. Input the normalized experimental data set into the multi-layer feedforward neural network model, and iteratively train the multi-layer feedforward neural network model according to the training parameters until a stable convergence training result is obtained; wherein, the number of hidden layers and the number of neurons in each layer are adjusted according to the results of each iterative training.

[0054] like Figure 2 FIG. 1 is a flowchart of a multi-layer feedforward neural network model training process according to the present embodiment.

[0055] S3. Based on the nonlinear mapping relationship between the ratio corresponding to the experimental data set and the 6-hour compressive strength, a high-dimensional multi-objective optimization model is constructed.

[0056] In this embodiment, the step of constructing a high-dimensional multi-objective optimization model includes determining decision variables, objective functions, and constraints.

[0057] The decision variables include the proportions of various raw materials for ultra-early strength concrete; the objective function includes key performance indicators for achieving ultra-early strength concrete; and the constraints include upper and lower limits of the proportions of various raw materials for ultra-early strength concrete, as well as limiting factors for preparing ultra-early strength concrete.

[0058] In an optional embodiment, the key performance indicators of ultra-early strength concrete in the objective function of the high-dimensional multi-objective optimization model include: minimizing the preparation cost per unit weight, minimizing the greenhouse gas emissions per unit weight, minimizing the water-cement ratio W / C, and maximizing the 6-hour compressive strength σ of the ultra-early strength concrete mix. 6h .

[0059] The expression of the objective function is as follows:

[0060]

[0061] Subject to: 30%≤SP≤40%;

[0062] Where n is the number of raw materials; U i represents the market price of raw material i (￥ / g), K i represents the amount of raw material i (g), C i represents the carbon emission factor of raw material i (kg CO2eq / t); m water Indicates the water consumption of the current ultra-early strength concrete mix (g), m cement Indicates the cement content of the current ultra-early strength concrete mix (g); They represent the lower and upper limits of the amount of raw material i, and SP represents the sand ratio.

[0063] Among them, the 6-hour compressive strength σ of the ultra-early strength concrete mix is 6h Obtained by fitting a multi-layer feedforward neural network model. Constraints include upper and lower limits for the amounts of various raw materials and the sand ratio (SP), which needs to be controlled between 30% and 40% to ensure the workability and stability of ultra-early strength concrete.

[0064] S4. Solve the high-dimensional multi-objective optimization model to obtain a Pareto optimal solution set, and output the Pareto optimal solution set as an ultra-early strength concrete mix ratio optimization solution.

[0065] This embodiment uses an optimization algorithm that can effectively search for at least four independent objectives included in the high-dimensional multi-objective optimization model to solve the high-dimensional multi-objective optimization model.

[0066] Among them, the metaheuristic algorithm designed based on Darwin's "theory of evolution" has strong adaptability (performs well when solving optimization problems in different fields and types). When the algorithm has a module that can efficiently deal with "dominant resistance", this type of algorithm is suitable for solving most high-dimensional multi-objective optimization problems.

[0067] Furthermore, when applying an optimization algorithm to solve a specific high-dimensional, multi-objective optimization problem, it's necessary to first configure the relevant parameters. For example, in a genetic algorithm, it's necessary to specify the population size, the number of generations, and the probabilities of crossover and mutation operations. This guides the algorithm to search the solution space of the optimization problem according to these parameters, ultimately obtaining the Pareto optimal solution set for the high-dimensional, multi-objective optimization problem.

[0068] In one alternative, the Pareto optimal solution set is plotted on a parallel coordinate graph for visualization, and the mix ratio with the best overall benefits is selected as the output of the ultra-early-strength concrete mix ratio optimization solution. Alternatively, a quantitative multi-attribute decision analysis method, such as the Top-and-Shorts Method (TOPSIS), can be used to screen out the ultra-early-strength concrete mix ratio solution with the best overall benefits.

[0069] Example 3

[0070] This example applies the ultra-early-strength concrete mix optimization method proposed in Example 2 to prepare ultra-early-strength concrete for rapid pavement repair. This method addresses the design requirements of reducing preparation cost, greenhouse gas emissions, shrinkage, and improving 6-hour compressive strength. Minimizing preparation cost refers to minimizing the amount of cement, aggregate, and various admixtures in the concrete mix while satisfying the sand ratio constraint.

[0071] In this embodiment, the steps of applying the ultra-early-strength concrete mix ratio optimization method to prepare ultra-early-strength concrete for rapid road repair include:

[0072] Step 1: Based on the 75 acquired ultra-early-strength concrete experimental mix data (including original mix experiment records and data resources captured by the Focus Web Crawler), a dataset of ultra-early-strength concrete experimental mixes was constructed. This dataset includes the raw material dosages required to prepare ultra-early-strength concrete specimens and the corresponding compressive strengths measured after 6 hours of curing. The raw materials include coarse aggregate, fine aggregate, cementitious materials, liquid, and admixtures.

[0073] As concrete preparation experiments continue and target site data is continuously updated, the above experimental data set will continue to expand, and ultimately improve the nonlinear mapping accuracy of the BP neural network model.

[0074] Step 2: According to the raw material type of ultra-early strength concrete, a three-layer BP neural network model consisting of 1 input layer, 1 hidden layer and 1 output layer is established.

[0075] like Figure 3 FIG. 1 is a diagram showing the architecture of the BP neural network model of this embodiment.

[0076] Among them, the input layer (Input) contains 11 neurons; the hidden layer (Hidden) contains 5 neurons, and the number of neurons in the hidden layer is obtained by repeatedly training the BP neural network; the output layer (Output) contains 1 neuron.

[0077] The 11 neurons in the input layer correspond to the amounts of 11 raw materials, including water, cement, sand, gravel, water reducer, silica fume (or mineral powder), retarder 1, retarder 2, water glass, sodium hydroxide (NaOH), and accelerator. The single neuron in the output layer corresponds to the 6-hour compressive strength of the ultra-early-strength concrete specimen. Each neuron in each layer is connected to each neuron in the next layer via a weight W and a bias b.

[0078] When training the BP neural network, the original data is first normalized, that is, all data are mapped to the interval [-1, 1] using formula (1). Then, the training parameters of the BP neural network are set as follows: select "random grouping" for data classification; select "Levenberg-Marquardt back propagation method" for training algorithm; select "normalized mean square error" for neural network performance evaluation method; and select "default gradient algorithm" for gradient calculation method. After setting the above parameters, the BP neural network is trained until it fully converges.

[0079]

[0080] Where x is the original experimental data; x′ is the normalized experimental data; and x′∈R, x min =min(x),x max =max(x)

[0081] The BP neural network linear regression diagram obtained in this embodiment is as follows Figure 4 As shown. Figure 4 It can be seen that no matter the training set (Training), validation set (Validation) or test set (Test), the fitting value of the BP neural network is in good agreement with the measured 6-hour compressive strength of ultra-early strength concrete, and the correlation coefficient R is greater than 0.8.

[0082] Step 3: Establish a high-dimensional multi-objective optimization model of the embodiment according to formula (2). The decision variables are the amounts of 11 raw materials required to prepare ultra-early strength concrete (unit: g), and the objective functions are minimizing the preparation cost per unit weight (unit: ￥ / g), minimizing the greenhouse gas emission equivalent per unit weight (unit: g-CO2-e / g), minimizing the water-cement ratio W / C (unit: %), and maximizing the 6-hour compressive strength σ 6h (Unit: MPa).

[0083]

[0084] Subject to: 30%≤SP≤40%

[0085] Among them, the water-cement ratio W / C is the main factor affecting the shrinkage rate of ultra-early strength concrete. The larger the water-cement ratio, the greater the shrinkage rate, which in turn destroys the consistency with the original pavement concrete; the σ corresponding to a certain ratio is 6h It is obtained by fitting the BP neural network.

[0086] The constraints include the upper and lower limits of the amounts of various raw materials and the sand ratio SP, among which the sand ratio SP needs to be controlled at 30% to 40% to ensure the workability and stability of ultra-early strength concrete.

[0087] Step 4: Select the Borg algorithm as the tool to solve the above high-dimensional multi-objective optimization model.

[0088] The Borg algorithm has the following three characteristics:

[0089] (1) Adopting a ranking method based on ∈-dominance.

[0090] To ensure convergence and diversity of the solution set, we use ∈-dominance as a new evaluation criterion to compare and sort the solution set, building on Pareto dominance. We establish an external solution archive to store the current optimal non-dominated solution. Here, ∈ represents the minimum resolution of a target in the sorting process (i.e., differences smaller than ∈ are ignored), effectively controlling the size of the Pareto solution set (overcoming "dominance resistance").

[0091] (2) It has adaptive population size and two restart mechanisms.

[0092] The first approach is to establish a metric, ∈-progress, to measure search progress based on ∈-dominance. During the search process, if the evolutionary distance of a solution is less than ∈, the original search is halted and a new search is initiated. Another approach is to determine whether to activate a new search by judging the γ value (the ratio of the population size to the size of the external solution archive). If this ratio is greater than 4 (i.e., the current population size exceeds 4 times the size of the optimal solution set), a restart is triggered (reinitialization in the solution space with the original population size).

[0093] (3) Equipped with multiple search operators for joint optimization.

[0094] During the optimization process, the Borg algorithm uses six operators with different search performances. Based on the quality of the offspring produced by each operator, a feedback mechanism is established to automatically adjust the application ratio of each operator, ensuring that operators with outstanding search performance have a greater probability of producing offspring individuals, thereby improving the quality of the solution set.

[0095] like Figure 5 As shown in Figure 1, it is the main loop flow chart of the Borg algorithm. Figure 5 The selection of the six operators in the figure is represented by dashed and solid lines. The solid line indicates the selected operator, while the dashed line indicates the operator to be selected. Based on the adaptive multi-operator feedback mechanism, one operator is selected for genetic manipulation at a time. The chance of each operator being selected depends on the survival of its offspring in the external archive.

[0096] Figure 5The current selection simulates the binary crossover operator and polynomial mutation for genetic operations. This process uniformly and randomly selects 1 parent individual from the archive and selects k-1 parent individuals from the population through the league selection method. After the operator with k parent individuals generates a child individual, it is passed as a decision variable to the high-dimensional multi-objective optimization model (which includes calling the BP neural network to obtain σ 6h ) and calculate the objective function values ​​of the current offspring individuals. The fitness of the objective function values ​​is checked. Offspring individuals that do not meet the retention criteria are eliminated, while offspring individuals that meet the survival criteria are retained and saved to the population and external archives respectively.

[0097] During Borg's optimization process, the program periodically checks the progress of the ∈-progress search and the value of γ. If a restart condition is met, the main loop stops and the restart mechanism is invoked. After the restart is complete, the main loop resumes. The entire optimization process repeats these steps until the termination condition is met.

[0098] Given that the Borg algorithm is highly adaptable, this embodiment uses the default parameter configuration of the algorithm to solve the high-dimensional multi-objective optimization model established in step 3, and the number of evaluations of the objective function is set to 10,000.

[0099] Step 5: After optimization calculation, Borg finally stores 19 Pareto optimal solutions in the external file. The above Pareto optimal solutions are plotted in the parallel coordinates diagram. The parallel coordinate diagram with the Pareto optimal solution set is as follows: Figure 6 shown.

[0100] Depend on Figure 6 The Pareto optimal solution shown in the figure shows that the preparation cost of ultra-early-strength concrete has a high degree of synergy with greenhouse gas emissions and 6-hour compressive strength, but has a contradictory relationship with the water-cement ratio. That is, the lower the preparation cost, the higher the water-cement ratio, which in turn leads to an increase in the shrinkage rate of ultra-early-strength concrete, affecting the consistency of the pavement in the later stage of repair and posing a greater risk of cracking.

[0101] By comparing the above Pareto optimal solutions, we can find a representative solution with the best comprehensive benefits (see Figure 6 This solution is close to the optimal level in terms of preparation cost and greenhouse gas emissions, and has the highest 6-hour compressive strength and a median water-cement ratio.

[0102] It can be seen from the above examples that the present invention can quickly find the raw material ratio scheme with the best comprehensive benefits that comprehensively considers the design requirements of four aspects: ultra-early strength concrete preparation cost, greenhouse gas emissions, shrinkage rate, and early compressive strength. By preparing test specimens in the laboratory, the measured 6-hour compressive strength was 57.31 MPa, and the relative error between the BP neural network estimated value and the measured value was 6.5%, which has high accuracy. The above examples prove that the concrete mix design method based on neural network and high-dimensional multi-objective optimization proposed in the present invention can significantly improve the preparation efficiency of ultra-early strength concrete.

[0103] Example 4

[0104] This embodiment proposes an ultra-early-strength concrete mix ratio optimization system based on high-dimensional multi-objective optimization, and applies the ultra-early-strength concrete mix ratio optimization method proposed in Example 1 or Example 2. Figure 7 , which is an architecture diagram of the ultra-early strength concrete mix ratio optimization system of this embodiment.

[0105] The ultra-early-strength concrete mix ratio optimization system proposed in this embodiment based on high-dimensional multi-objective optimization includes:

[0106] The data acquisition module 100 is used to obtain an experimental data set of ultra-early strength concrete mix proportions.

[0107] In an optional embodiment, the data acquisition module 100 collects data by collecting all ultra-early strength concrete mix ratio experimental data of a local laboratory, and / or collects data by using a focused web crawler technology to capture ultra-early strength concrete mix ratio data publicly available on the Internet.

[0108] The neural network module 200 includes a multi-layer feedforward neural network model 210, which is used to perform iterative training using the experimental data set collected by the data acquisition module 100, and to generate a nonlinear mapping relationship between the ultra-early strength concrete mix ratio and the 6-hour compressive strength corresponding to the experimental data set.

[0109] In an optional embodiment, the multi-layer feedforward neural network model 210 includes an input layer, at least one hidden layer, and an output layer; wherein the input layer, hidden layer, and output layer are sequentially connected through neurons.

[0110] The multi-objective optimization module 300 includes a high-dimensional multi-objective optimization model 310 and an optimization solution unit 320 .

[0111] The high-dimensional multi-objective optimization model 310 is constructed or updated according to the nonlinear mapping relationship between the ratio corresponding to the experimental data set and the 6-hour compressive strength.

[0112] The optimization solution unit 320 is used to solve the high-dimensional multi-objective optimization model using an optimization algorithm to obtain a Pareto optimal solution set, and output the Pareto optimal solution set as an ultra-early strength concrete mix ratio optimization solution.

[0113] In an optional embodiment, the high-dimensional multi-objective optimization model 310 includes decision variables, objective functions, and constraints. The decision variables include the ratios of various raw materials for ultra-early-strength concrete; the objective function includes key performance indicators for achieving ultra-early-strength concrete; and the constraints include upper and lower limits for the ratios of various raw materials for ultra-early-strength concrete, as well as limiting factors for producing ultra-early-strength concrete.

[0114] Furthermore, the key performance indicators for achieving ultra-early strength concrete in the objective function include: minimizing the preparation cost per unit weight, minimizing the greenhouse gas emissions per unit weight, minimizing the water-cement ratio W / C, and maximizing the 6-hour compressive strength σ of the ultra-early strength concrete mix. 6h .

[0115] During the specific implementation process, the data acquisition module 100 obtains the experimental data set of ultra-early strength concrete mix by collecting all ultra-early strength concrete mix experimental data from the local laboratory, and by using the focused web crawler technology to capture the ultra-early strength concrete mix data publicly available on the Internet, and transmits it to the neural network module 200 and the multi-objective optimization module 300.

[0116] The trained multi-layer feedforward neural network model 210 in the neural network module 200 fits the input experimental data set data, outputs the nonlinear mapping relationship between the ultra-early strength concrete mix ratio and the 6-hour compressive strength corresponding to the experimental data set, and transmits it to the multi-objective optimization module 300.

[0117] The multi-objective optimization module 300 updates the high-dimensional multi-objective optimization model 310 according to the experimental data set data it receives and the nonlinear mapping relationship between the ultra-early strength concrete mix ratio and the 6-hour compressive strength corresponding to the experimental data set.

[0118] The optimization solution unit 320 calls the high-dimensional multi-objective optimization model 310, uses an optimization algorithm to solve the multi-objective optimization model, obtains a Pareto optimal solution set, and outputs the Pareto optimal solution set as an ultra-early strength concrete mix ratio optimization solution.

[0119] In an optional embodiment, the system also includes a visualization module 400, which is used to plot the Pareto optimal solution set output by the multi-objective optimization module 300 in a parallel coordinate diagram for visualization, so that technicians can select the ratio with the best comprehensive benefits from the Pareto optimal solution set as the ultra-early strength concrete ratio optimization solution.

[0120] In an optional embodiment, the system further includes a decision analysis module 500, which is used to adopt a quantitative multi-attribute decision analysis method to select a mix ratio with the best comprehensive benefits from the Pareto optimal solution set as the ultra-early strength concrete mix ratio optimization solution.

[0121] The same or similar reference numerals correspond to the same or similar components;

[0122] The terms used in the drawings to describe positional relationships are for illustrative purposes only and should not be construed as limiting this patent;

[0123] Obviously, the above embodiments of the present invention are merely examples for the purpose of clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. Those skilled in the art will appreciate that other variations or modifications can be made based on the above description. It is not necessary and impossible to enumerate all embodiments here. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the claims of the present invention.

Claims

1. A method for optimizing the mix ratio of ultra-early-strength concrete based on high-dimensional multi-objective optimization, characterized in that: The following steps are involved: S1. Obtain the experimental data set of ultra-early strength concrete mix proportions; S2. Establish a multi-layer feedforward neural network model, input the experimental data set into the multi-layer feedforward neural network model and train it based on the error back propagation algorithm until it fully converges, to obtain a nonlinear mapping relationship between the ratio corresponding to the experimental data set and the 6-hour compressive strength; S3. Construct a high-dimensional multi-objective optimization model based on the nonlinear mapping relationship between the ratio corresponding to the experimental data set and the 6-hour compressive strength; The steps of constructing a high-dimensional multi-objective optimization model include determining decision variables, objective functions, and constraints; wherein the decision variables include the ratios of various raw materials for ultra-early-strength concrete; the objective function includes key performance indicators for achieving ultra-early-strength concrete; and the constraints include upper and lower limits for the ratios of various raw materials for ultra-early-strength concrete, as well as limiting factors for preparing ultra-early-strength concrete. The key performance indicators for achieving ultra-early strength concrete in the objective function include: minimizing the preparation cost per unit weight Cost , minimizing greenhouse gas emissions per unit weight GHG , minimize the water-cement ratio W / C , maximize the 6-hour compressive strength of ultra-early strength concrete mix σ 6h ; Wherein, the expression of the objective function is as follows: Subject to : ，30%≤ SP ≤40%； Where, n is the number of raw materials; U i Indicates raw materials i The corresponding market unit price, K i Indicates raw materials i Dosage, C i Indicates raw materials i Carbon emission factors; m water Indicates the water consumption of the current ultra-early strength concrete mix ratio, m cement Indicates the cement content of the current ultra-early strength concrete mix; Represents raw materials i The lower and upper limits of dosage, SP Indicates sand ratio; S4. Solving the high-dimensional multi-objective optimization model to obtain a Pareto optimal solution set, and outputting the Pareto optimal solution set as an optimization solution for ultra-early strength concrete mix ratio; In the step S4, the high-dimensional multi-objective optimization model is solved by using an optimization algorithm that can effectively search for at least four independent objectives contained in the high-dimensional multi-objective optimization model.

2. The method for optimizing the mix ratio of ultra-early-strength concrete according to claim 1, wherein: In step S1, the experimental data set is obtained by collecting all ultra-early-strength concrete mix ratio experimental data from local laboratories, and / or by using a focused web crawler technology to capture ultra-early-strength concrete mix ratio data publicly available on the Internet.

3. The method for optimizing the mix ratio of ultra-early-strength concrete according to claim 1, wherein: In step S2, the multi-layer feedforward neural network model includes an input layer, at least one hidden layer, and an output layer; wherein the input layer, hidden layer, and output layer are sequentially connected through neurons.

4. The method for optimizing the mix ratio of ultra-early-strength concrete according to claim 3, wherein: The number of neurons contained in the input layer is equal to the types of raw materials in the ultra-early strength concrete mix; The number of neurons contained in the output layer is equal to the number of key performance parameters in the ultra-early strength concrete mix; The number of hidden layers and the number of neurons in each layer are adjusted according to the training results of the multi-layer feedforward neural network model.

5. The method for optimizing the mix ratio of ultra-early-strength concrete according to claim 4, wherein: In step S2, the step of inputting the experimental data set into a multi-layer feedforward neural network model and training it to sufficient convergence based on an error back propagation algorithm includes: S2.

1. Normalize the data in the experimental dataset; S2.

2. Setting the training parameters of the multi-layer feedforward neural network model; S2.

3. Input the normalized experimental data set into the multi-layer feedforward neural network model, and iteratively train the multi-layer feedforward neural network model according to the training parameters until a stable convergence training result is obtained; wherein, the number of hidden layers and the number of neurons in each layer are adjusted according to the results of each iterative training.

6. The method for optimizing the mix ratio of ultra-early strength concrete according to any one of claims 1 to 5, wherein: The step S4 further includes the following steps: plotting the Pareto optimal solution set in a parallel coordinate diagram for visualization, or using a quantitative multi-attribute decision analysis method to select a mix ratio with the best comprehensive benefits from the Pareto optimal solution set as an output of an ultra-early strength concrete mix ratio optimization solution.

7. An ultra-early-strength concrete mix ratio optimization system based on high-dimensional multi-objective optimization, applied to the ultra-early-strength concrete mix ratio optimization method according to any one of claims 1 to 5, characterized in that: include: Data acquisition module, used to obtain experimental data sets of ultra-early strength concrete mix proportions; A neural network module, comprising a multi-layer feedforward neural network model for iterative training using the experimental data set collected by the data acquisition module, and for generating a nonlinear mapping relationship between the ultra-early strength concrete mix ratio and the 6-hour compressive strength corresponding to the experimental data set; A multi-objective optimization module includes a high-dimensional multi-objective optimization model and an optimization solution unit; the high-dimensional multi-objective optimization model is constructed or updated according to the nonlinear mapping relationship between the proportion corresponding to the experimental data set and the 6-hour compressive strength; the optimization solution unit is used to solve the high-dimensional multi-objective optimization model using an optimization algorithm to obtain a Pareto optimal solution set, and output the Pareto optimal solution set as an ultra-early strength concrete proportion optimization scheme.

Images