Neural network-based concrete mix proportion optimization method and system

Abstract

The application relates to the technical field of concrete application, and discloses a concrete mix proportion optimization method and system based on a neural network, which comprises the following steps: generating multiple groups of performance and factor test data according to standard performance parameter information, basic main factor characteristics and an orthogonal test mode of target concrete; determining key main factor characteristics according to all the performance and factor test data; constructing and training a converged concrete mix proportion analysis model according to a double-hidden layer architecture scheme, an input / output node configuration scheme, a weight tilt scheme, a network drop rate parameter and a learning rate parameter; and obtaining optimal mix proportion results of the target concrete based on expected performance parameter information of the target concrete and the concrete mix proportion analysis model. It can be seen that the application can improve the accuracy and reliability, efficiency and convenience of the concrete mix proportion, reduce the use amount of cement under the guarantee of the performance of the concrete, and improve the compressive strength of the concrete while ensuring the slump of the concrete.

Description

Technical Field

[0001] This invention relates to the field of concrete application technology, and in particular to a method and system for optimizing concrete mix proportions based on neural networks. Background Technology

[0002] With the continuous development of the construction industry, modern construction projects have put forward higher requirements for the performance and quality of concrete. Among them, the concrete mix proportion directly determines the key indicators such as concrete strength, durability, workability and economy. Therefore, determining the concrete mix proportion has become a major research focus.

[0003] Currently, the main method for determining concrete mix proportions relies on the subjective judgment of workers based on their accumulated experience. However, this subjective approach is susceptible to various influences, leading to discrepancies in mix proportions and processing times even under identical conditions for the same application scenario. This results in low accuracy and efficiency in concrete mix proportion determination. Therefore, providing a new method for optimizing concrete mix proportions to improve their accuracy and efficiency is of paramount importance. Summary of the Invention

[0004] This invention provides a method and system for optimizing concrete mix proportions based on neural networks, which can improve the accuracy and reliability of concrete mix proportion determination, as well as the efficiency and convenience of concrete mix proportion determination. This allows for a reduction in cement usage while ensuring concrete performance, and an increase in concrete compressive strength while ensuring that the concrete slump meets predetermined standards.

[0005] To address the aforementioned technical problems, the first aspect of this invention discloses a method for optimizing concrete mix proportions based on neural networks, the method comprising:

[0006] Determine the standard performance parameter information of the target concrete, including slump parameter and compressive strength parameter and their corresponding standard parameter value range;

[0007] The fundamental characteristics that can affect concrete performance are identified, and multiple sets of performance and factor test data are generated based on the standard performance parameter information, the fundamental characteristics, and the determined orthogonal test method.

[0008] Based on all the performance and factor test data, the key principal characteristics of the target concrete were determined, including sand ratio characteristics, cement content characteristics, and water-reducing agent characteristics.

[0009] Based on the determined dual-hidden-layer architecture scheme, input / output node configuration scheme, weight skew scheme, network dropout rate parameter, and learning rate parameter, a converged concrete mix proportion analysis model is constructed and trained.

[0010] Based on the determined expected performance parameters of the target concrete and the concrete mix proportion analysis model, the optimal mix proportion of the target concrete is obtained.

[0011] As an optional implementation, in the first aspect of the present invention, the basic key characteristics include sand ratio characteristics, crushed stone ratio characteristics, cement content characteristics, water-cement ratio characteristics, and water-reducing agent characteristics; the test data for each of the aforementioned performance and factors includes first test data for standard performance parameters and second test data for the basic key characteristics that are matched thereto.

[0012] And, the determination of the key principal characteristics of the target concrete based on all the performance and factor test data includes:

[0013] Based on all the performance and factor test data, the first target sub-primary factor feature that meets the preset sensitivity conditions is selected from all the basic sub-primary factor features;

[0014] Based on all the performance and factor test data, select second target sub-key features that meet the preset significance conditions from all the first target sub-key features;

[0015] Based on all the second target sub-key characteristics, the key key characteristics of the target concrete are determined.

[0016] As an optional implementation, in the first aspect of the present invention, the step of selecting a first target sub-primary factor feature that meets a preset sensitivity condition from all basic sub-primary factor features based on all the performance and factor test data includes:

[0017] Based on all the performance and factor test data, determine the first sensitivity index result for each basic sub-key characteristic at the slump level and the second sensitivity index result at the compressive strength level.

[0018] Based on the first sensitivity index result of each of the basic sub-primary features, the first sensitivity ranking is determined, and based on the second sensitivity index result of each of the basic sub-primary features, the second sensitivity ranking is determined.

[0019] Based on the first sensitivity ranking and the second sensitivity ranking, a first target sub-primary feature that meets the preset ranking conditions is determined from all the basic sub-primary features.

[0020] As an optional implementation, in the first aspect of the present invention, the step of selecting second target sub-key features that meet preset significance conditions from all the first target sub-key features based on all the performance and factor test data includes:

[0021] Based on all the performance and factor test data, determine the significance index results for each of the first target sub-key characteristics;

[0022] Based on the significance index results of each of the first target sub-primary features, the second target sub-primary features that meet the preset significance threshold conditions are determined from all the first target sub-primary features.

[0023] As an optional implementation, in the first aspect of the present invention, before constructing and training a converged concrete mix proportion analysis model based on the determined dual hidden layer architecture scheme, input / output node configuration scheme, weight skew scheme, network dropout rate parameter, and learning rate parameter, the method further includes:

[0024] Based on the sand ratio characteristics, the cement content characteristics, and the water-reducing agent characteristics, the model input node information is determined, and based on the slump parameter and the compressive strength parameter, the model output node information is determined.

[0025] Based on the model input node information and the model output node information, determine the input and output node configuration scheme for the concrete mix proportion analysis model;

[0026] Based on the preset Bayesian optimization method, the number of first nodes for the first hidden layer and the number of second nodes for the second hidden layer are determined, and the network dropout rate parameter and learning rate parameter are determined.

[0027] Based on the number of the first node and the number of the second node, a dual-hidden-layer architecture scheme is determined, wherein the first hidden layer precedes the second hidden layer, and the number of the first node is greater than the number of the second node;

[0028] Based on the determined first weight value for the slump parameter and the second weight value for the compressive strength parameter, a weight tilting scheme is determined, wherein the first weight value is less than the second weight value.

[0029] As an optional implementation, in the first aspect of the present invention, the method further includes:

[0030] Based on the optimal mix proportion results and the concrete mix proportion analysis model, the quantitative characteristic contribution results are determined;

[0031] Based on the quantified feature contribution results, a three-dimensional response surface is generated. The three-dimensional response surface is used to demonstrate the interaction effects of sand ratio feature-cement quantity feature and cement quantity feature-water reducing agent feature.

[0032] As an optional implementation, in the first aspect of the present invention, the method further includes:

[0033] Determine the actual application scenario information of the target concrete;

[0034] Based on the optimal mix proportion results and the actual application scenario information, determine whether the target concrete meets the preset conditions for additional additives;

[0035] When it is determined that the target concrete meets the conditions for the additional added material, the additional required functions of the target concrete are determined based on the optimal mix proportion results and the actual application scenario information; based on the additional required functions and the optimal mix proportion results, the types of additional materials to be added to the target concrete and their corresponding addition and mixing methods and material proportion schemes are determined.

[0036] Based on the optimal mix proportion results, the type of additional material and its addition and mixing method and material proportion scheme, determine whether the target concrete meets the preset conditions for further optimization of the mix proportion;

[0037] When it is determined that the target concrete does not meet the conditions for further optimization of the mixing ratio, the final manufacturing plan corresponding to the target concrete is generated based on the optimal mix ratio result, the type of additional material and its addition and mixing method and material ratio scheme.

[0038] When it is determined that the target concrete meets the conditions for further optimization of the mix ratio, the reverse performance impact on the target concrete is analyzed based on the type of additional material, its addition and mixing method, and the material proportioning scheme. Based on the reverse performance impact, the optimal mix ratio result is adjusted and updated accordingly to obtain the adjusted and updated optimal mix ratio result. Based on the adjusted and updated optimal mix ratio result, the type of additional material, its addition and mixing method, and the material proportioning scheme, the final manufacturing scheme corresponding to the target concrete is generated.

[0039] As an optional implementation, in the first aspect of the present invention, determining whether the target concrete meets the preset conditions for additional additives based on the optimal mix proportion result and the actual application scenario information includes:

[0040] Based on the optimal mix ratio, the actual performance type and effect are determined, and based on the actual application scenario information, the required performance type and effect are determined.

[0041] Determine whether the actual performance type and effect completely encompass the required performance type and effect;

[0042] When it is determined that the actual performance type and effect completely encompass the required performance type and effect, it is determined that the target concrete does not meet the preset additional material conditions.

[0043] When it is determined that the actual performance type and effect do not fully encompass the required performance type and effect, the target concrete is determined to meet the preset additional material conditions.

[0044] As an optional implementation, in the first aspect of the present invention, determining whether the target concrete meets the preset conditions for further optimization of the mix ratio based on the optimal mix proportion result, the type of additional material and its addition and mixing method and material proportion scheme includes:

[0045] Based on the optimal mix proportion results, the type of additional material and its addition and mixing method and material proportion scheme, it is determined whether the target concrete meets the preset performance weakening condition;

[0046] When it is determined that the target concrete meets the condition that the performance is weakened, it is determined that the target concrete meets the preset condition for further optimization of the mixing ratio.

[0047] When it is determined that the target concrete does not meet the condition for weakened performance, it is determined that the target concrete does not meet the preset condition for further optimization of the mixing ratio.

[0048] A second aspect of this invention discloses a concrete mix design optimization system based on a neural network, the system comprising:

[0049] The performance parameter determination module is used to determine the standard performance parameter information of the target concrete. The standard performance parameter information includes slump parameter and compressive strength parameter and their corresponding standard parameter value ranges.

[0050] The first feature determination module is used to determine the fundamental key features that can affect concrete performance.

[0051] The test data generation module is used to generate multiple sets of performance and factor test data based on the standard performance parameter information, the basic main factor characteristics, and the determined orthogonal test method.

[0052] The second feature determination module is used to determine the key principal features of the target concrete based on all the performance and factor test data. The key principal features include sand ratio features, cement content features, and water-reducing agent features.

[0053] The model building and training module is used to build and train a converged concrete mix proportion analysis model based on the determined double hidden layer architecture scheme, input and output node configuration scheme, weight tilt scheme, network dropout rate parameter and learning rate parameter.

[0054] The mix proportion determination module is used to obtain the optimal mix proportion result of the target concrete based on the determined expected performance parameter information of the target concrete and the concrete mix proportion analysis model.

[0055] As an optional implementation, in the second aspect of the present invention, the basic key characteristics include sand ratio characteristics, crushed stone ratio characteristics, cement content characteristics, water-cement ratio characteristics, and water-reducing agent characteristics; the test data for each of the aforementioned performance and factors includes first test data for standard performance parameters and second test data for the basic key characteristics that are matched thereto.

[0056] Furthermore, the second feature determination module determines the key principal features of the target concrete based on all the performance and factor test data in the following specific ways:

[0057] Based on all the performance and factor test data, the first target sub-primary factor feature that meets the preset sensitivity conditions is selected from all the basic sub-primary factor features;

[0058] Based on all the performance and factor test data, select second target sub-key features that meet the preset significance conditions from all the first target sub-key features;

[0059] Based on all the second target sub-key characteristics, the key key characteristics of the target concrete are determined.

[0060] As an optional implementation, in a second aspect of the present invention, the method by which the second feature determination module selects a first target sub-primary factor feature that meets a preset sensitivity condition from all basic sub-primary factor features based on all the performance and factor test data specifically includes:

[0061] Based on all the performance and factor test data, determine the first sensitivity index result for each basic sub-key characteristic at the slump level and the second sensitivity index result at the compressive strength level.

[0062] Based on the first sensitivity index result of each of the basic sub-primary features, the first sensitivity ranking is determined, and based on the second sensitivity index result of each of the basic sub-primary features, the second sensitivity ranking is determined.

[0063] Based on the first sensitivity ranking and the second sensitivity ranking, a first target sub-primary feature that meets the preset ranking conditions is determined from all the basic sub-primary features.

[0064] As an optional implementation, in a second aspect of the present invention, the method by which the second feature determining module selects second target sub-key features that meet preset significance conditions from all the first target sub-key features based on all the performance and factor test data specifically includes:

[0065] Based on all the performance and factor test data, determine the significance index results for each of the first target sub-key characteristics;

[0066] Based on the significance index results of each of the first target sub-primary features, the second target sub-primary features that meet the preset significance threshold conditions are determined from all the first target sub-primary features.

[0067] As an optional implementation, in a second aspect of the invention, the system further includes:

[0068] The model parameter determination module is used to determine the model input node information based on the sand ratio characteristics, cement content characteristics, and water-reducing agent characteristics, and to determine the model output node information based on the slump parameter and compressive strength parameter, before the model construction and training module constructs and trains a converged concrete mix proportion analysis model according to the determined double hidden layer architecture scheme, input and output node configuration scheme, weight tilting scheme, network dropout rate parameter, and learning rate parameter. Based on the model input node information and the model output node information, the module determines the input and output node configuration scheme for the concrete mix proportion analysis model. Based on a preset Bayesian optimization method, it determines the number of first nodes for the first hidden layer and the number of second nodes for the second hidden layer, and determines the network dropout rate parameter and learning rate parameter. Based on the number of first nodes and the number of second nodes, it determines the double hidden layer architecture scheme, where the first hidden layer precedes the second hidden layer, and the number of first nodes is greater than the number of second nodes. Based on the determined first weight value for the slump parameter and the second weight value for the compressive strength parameter, it determines the weight tilting scheme, where the first weight value is less than the second weight value.

[0069] As an optional implementation, in a second aspect of the invention, the system further includes:

[0070] The response surface determination module is used to determine the quantitative feature contribution result based on the optimal mix proportion result and the concrete mix proportion analysis model; and to generate three-dimensional response surface content based on the quantitative feature contribution result. The three-dimensional response surface content is used to display the interaction effect of sand ratio feature-cement quantity feature and cement quantity feature-water reducing agent feature.

[0071] As an optional implementation, in a second aspect of the invention, the system further includes:

[0072] The judgment module is used to determine the actual application scenario information of the target concrete; based on the optimal mix proportion result and the actual application scenario information, it determines whether the target concrete meets the preset additional additive conditions;

[0073] An additional material determination module is used to determine the additional required functions of the target concrete based on the optimal mix proportion result and the actual application scenario information when the judgment module determines that the target concrete meets the conditions for adding additional materials; and to determine the type of additional material to be added to the target concrete and its corresponding addition and mixing method and material ratio scheme based on the additional required functions and the optimal mix proportion result.

[0074] The judgment module is also used to determine whether the target concrete meets the preset conditions for further optimization of the mixing ratio based on the optimal mix ratio result, the type of additional material and its addition and mixing method and material ratio scheme;

[0075] The manufacturing scheme determination module is used to generate the final manufacturing scheme corresponding to the target concrete based on the optimal mix proportion result, the type of additional material and its addition mixing method and material proportion scheme when the judgment module determines that the target concrete does not meet the conditions for further optimization of the mixing ratio.

[0076] The mix proportion determination module is further configured to, when the judgment module determines that the target concrete meets the conditions for further optimization of the mix proportion, analyze the adverse performance impact on the target concrete based on the type of additional material, its addition and mixing method, and the material proportion scheme; and perform corresponding adjustment and update operations on the optimal mix proportion result based on the adverse performance impact to obtain the adjusted and updated optimal mix proportion result.

[0077] The manufacturing scheme determination module is also used to generate the final manufacturing scheme corresponding to the target concrete based on the adjusted and updated optimal mix proportion results, the type of additional material and its addition mixing method and material proportion scheme.

[0078] As an optional implementation, in a second aspect of the present invention, the method by which the judging module judges whether the target concrete meets the preset conditions for additional additives based on the optimal mix proportion result and the actual application scenario information specifically includes:

[0079] Based on the optimal mix ratio, the actual performance type and effect are determined, and based on the actual application scenario information, the required performance type and effect are determined.

[0080] Determine whether the actual performance type and effect completely encompass the required performance type and effect;

[0081] When it is determined that the actual performance type and effect completely encompass the required performance type and effect, it is determined that the target concrete does not meet the preset additional material conditions.

[0082] When it is determined that the actual performance type and effect do not fully encompass the required performance type and effect, the target concrete is determined to meet the preset additional material conditions.

[0083] As an optional implementation, in a second aspect of the present invention, the method by which the judging module determines whether the target concrete meets the preset conditions for further optimization of the mixing ratio based on the optimal mix proportion result, the type of additional material and its addition and mixing method and material proportion scheme specifically includes:

[0084] Based on the optimal mix proportion results, the type of additional material and its addition and mixing method and material proportion scheme, it is determined whether the target concrete meets the preset performance weakening condition;

[0085] When it is determined that the target concrete meets the condition that the performance is weakened, it is determined that the target concrete meets the preset condition for further optimization of the mixing ratio.

[0086] When it is determined that the target concrete does not meet the condition for weakened performance, it is determined that the target concrete does not meet the preset condition for further optimization of the mixing ratio.

[0087] A third aspect of this invention discloses another concrete mix design optimization system based on a neural network, the system comprising:

[0088] Memory containing executable program code;

[0089] A processor coupled to the memory;

[0090] The processor calls the executable program code stored in the memory to execute the concrete mix proportion optimization method based on neural networks disclosed in the first aspect of the present invention.

[0091] The fourth aspect of the present invention discloses a computer storage medium storing computer instructions, which, when invoked, are used to execute the neural network-based concrete mix proportion optimization method disclosed in the first aspect of the present invention.

[0092] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0093] In this embodiment of the invention, standard performance parameters of the target concrete are determined, including slump parameters and compressive strength parameters and their corresponding standard parameter value ranges; fundamental key characteristics that can affect concrete performance are determined, and multiple sets of performance and factor test data are generated based on the standard performance parameters, the fundamental key characteristics, and the determined orthogonal experimental method; based on all the performance and factor test data, key key characteristics of the target concrete are determined, including sand ratio, cement content, and water-reducing agent characteristics; based on the determined double hidden layer architecture scheme, input / output node configuration scheme, weight tilt scheme, network dropout rate parameter, and learning rate parameter, a converged concrete mix proportion analysis model is constructed and trained; based on the determined expected performance parameters of the target concrete and the concrete mix proportion analysis model, the optimal mix proportion result of the target concrete is obtained. It is evident that this invention can determine the key characteristics of the target concrete based on multiple sets of performance and factor test data, obtain a converged concrete mix proportion analysis model based on a series of determined model construction and training parameters, and obtain the optimal mix proportion result of the target concrete based on the expected performance parameter information and the concrete mix proportion analysis model. This is beneficial to improving the comprehensiveness and rationality of the method for determining the optimal concrete mix proportion result, thereby improving the accuracy and efficiency of concrete mix proportion optimization, which in turn improves the accuracy and reliability of the determined concrete mix proportion, improves the efficiency and convenience of determining the concrete mix proportion, and further reduces the amount of cement used while ensuring concrete performance, and improves the compressive strength of concrete while ensuring that the concrete slump meets the established standard. Attached Figure Description

[0094] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0095] Figure 1 This is a schematic flowchart of a concrete mix proportion optimization method based on neural networks disclosed in an embodiment of the present invention;

[0096] Figure 2 This is a schematic flowchart of another concrete mix proportion optimization method based on neural networks disclosed in an embodiment of the present invention;

[0097] Figure 3 This is a schematic diagram of a concrete mix proportion optimization system based on a neural network disclosed in an embodiment of the present invention;

[0098] Figure 4This is a schematic diagram of another concrete mix proportion optimization system based on neural networks disclosed in an embodiment of the present invention;

[0099] Figure 5 This is a schematic diagram of another concrete mix proportion optimization system based on neural networks disclosed in an embodiment of the present invention;

[0100] Figure 6 This is a schematic diagram of the network architecture of a concrete mix proportion analysis model disclosed in an embodiment of the present invention. Detailed Implementation

[0101] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0102] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this invention are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, apparatus, product, or end that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or ends.

[0103] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0104] This invention discloses a concrete mix design optimization method and system based on neural networks. It can determine the key characteristics of the target concrete based on multiple sets of performance and factor test data. Based on a series of determined model construction and training parameters, a converged concrete mix design analysis model is obtained. Based on expected performance parameter information and the concrete mix design analysis model, the optimal mix design result of the target concrete is obtained. This improves the comprehensiveness and rationality of the method for determining the optimal concrete mix design result, thereby improving the accuracy and efficiency of concrete mix design optimization. This, in turn, improves the accuracy and reliability of the determined concrete mix design, increases the efficiency and convenience of determining the concrete mix design, and further reduces the amount of cement used while ensuring concrete performance. It also improves the compressive strength of the concrete while ensuring that the concrete slump meets the predetermined standard. Detailed descriptions follow.

[0105] Example 1

[0106] Please see Figure 1 , Figure 1 This is a schematic flowchart of a concrete mix proportion optimization method based on a neural network, as disclosed in an embodiment of the present invention. Wherein, Figure 1 The described method can be applied to a neural network-based concrete mix design optimization system. The device may include a server, which may be a local server or a cloud server; this embodiment of the invention is not limited to this. Figure 1 As shown, the concrete mix design optimization method based on neural networks includes the following operations:

[0107] 101. Determine the standard performance parameters of the target concrete. The standard performance parameters include slump parameters and compressive strength parameters and their corresponding standard parameter value ranges.

[0108] Optionally, the standard parameter range for slump is, for example, 160-200 mm, but this embodiment of the invention does not limit it.

[0109] Optionally, the standard parameter range for compressive strength is, for example, ≥50MPa, but this embodiment of the invention does not limit it.

[0110] 102. Determine the fundamental characteristics that can affect concrete performance, and generate multiple sets of performance and factor test data based on standard performance parameter information, fundamental characteristics, and the determined orthogonal test method.

[0111] Optionally, the value ranges of factors corresponding to the basic main characteristics, such as sand ratio, crushed stone ratio, cement content, water-cement ratio, and water-reducing agent characteristics, should cover a reasonable range that can achieve a slump of 160-200mm. The number of test groups (e.g., L) 25 (5 5(e.g.,) it is necessary to ensure that the sample data can cover a sufficient range of performance fluctuations, but this embodiment of the invention does not impose any limitations.

[0112] Optionally, multiple sets of performance and factor test data are provided. For example, 25 sets of test data are provided, including 5 factors (i.e., sand ratio, crushed stone ratio, cement content, water-cement ratio, and water-reducing agent) and 2 performance parameters (i.e., slump and compressive strength). Furthermore, one part is training set data (orthogonal test data), and the other part is validation set data (used for early shutdown and overparameter testing). This embodiment of the invention does not limit the scope of the invention.

[0113] 103. Based on all performance and factor test data, determine the key principal characteristics of the target concrete. The key principal characteristics include sand ratio characteristics, cement content characteristics, and water-reducing agent characteristics.

[0114] Optionally, a range analysis can be performed based on all performance and factor test data to calculate the range. Analysis of variance (P<0.05) was used to screen out key factors (i.e. key principal characteristics) that have a "large and significant impact". This embodiment of the invention is not limited to these factors.

[0115] Optionally, for example: by analyzing the average strength corresponding to different levels of cement content in multiple sets of performance and factor test data, key factors are determined. "Key influencing factors" refer to the mix proportion parameters (i.e., corresponding sand ratio, crushed stone ratio, cement content, water-cement ratio, and water-reducing agent) that have a significant regulatory effect on the performance (slump, compressive strength) of C50 concrete. This embodiment of the invention does not limit these parameters.

[0116] 104. Based on the determined double hidden layer architecture scheme, input and output node configuration scheme, weight tilt scheme, network dropout rate parameter and learning rate parameter, construct and train a converged concrete mix proportion analysis model.

[0117] Optionally, the architecture diagram corresponding to the concrete mix proportion analysis model can be referenced, but is not limited to, the diagram attached to the instruction manual. Figure 6 As shown, the embodiments of the present invention are not limited.

[0118] Optionally, a dual-hidden-layer architecture scheme can be used. For example, a dual-hidden-layer structure of "32 hidden layers → 16 hidden layers" can be set. This embodiment of the invention is not limited to this.

[0119] Optionally, the input / output node configuration scheme is illustrated as follows: set "3 inputs" as input nodes, specifically including: sand ratio input layer, cement quantity input layer and water-reducing agent input layer; set "2 outputs" as output nodes, specifically including: slump output and compressive strength output. This embodiment of the invention is not limited.

[0120] Optionally, a weighting scheme can be used, for example: Design a weighted loss function for the network architecture (Loss = 0.7 × MSE strength + 0.3 × MSE collapse). By weighting (e.g., prioritizing compressive strength), the model's predicted strength achievement rate can be improved. Further, the weighted loss function can be specifically as follows:

[0121]

[0122] in, The output of the model is the predicted concrete strength. This represents the actual measured value of the concrete strength. The output of the model is the predicted concrete slump. The actual measured value of concrete slump is used, and the embodiments of the present invention are not limited to this.

[0123] Optionally, network dropout rate and learning rate parameters can be illustrated as follows: Optimal hyperparameters (such as the number of hidden layer nodes, dropout rate, and learning rate) can be searched using Bayesian optimization. Furthermore, the hidden layers can employ the ReLU activation function and be configured with Dropout regularization to specifically address the problem of identifying "nonlinear inflection points" in concrete data and suppress random interference in experimental data through regularization. ReLU addresses the "gradient vanishing" problem in neurons, while Dropout addresses the "overfitting" problem. The two work together to ensure the network's ability to fit the nonlinear performance of concrete. This embodiment of the invention is not limited in its specific implementation.

[0124] 105. Based on the determined expected performance parameters of the target concrete and the concrete mix proportion analysis model, the optimal mix proportion of the target concrete is obtained.

[0125] Optional, feasible domain definition, for example: based on engineering experience, set the sand ratio (36%-40%) and cement content (440-500 kg / m³). 3 ) and other parameter boundaries, to avoid the optimization results exceeding the construction feasibility, the embodiments of the present invention do not impose limitations.

[0126] Optionally, for example: based on the concrete mix proportion analysis model that has been trained and converged, the mix proportion with the maximum strength (i.e., the corresponding expected performance parameter information) (i.e., the optimal mix proportion result) is solved under the slump constraint. This embodiment of the invention is not limited.

[0127] Optionally, a neural network can be used as the response surface model, and the Adam algorithm can be used to search for the maximum strength mix proportion under constraints. Further, for example, the constraint optimization uses the gradient descent algorithm to search for the optimal solution within the feasible region of sand ratio (1.37, 1.45), cement content (0.88, 1.00), and water-reducing agent (0.0122, 0.0146) to obtain the optimal mix proportion result. This embodiment of the invention is not limited.

[0128] As can be seen, the concrete mix proportion optimization method based on neural networks described in this embodiment of the invention can determine the key principal characteristics of the target concrete based on multiple sets of performance and factor test data, obtain a converged concrete mix proportion analysis model based on a series of determined model construction and training parameters, and obtain the optimal mix proportion result of the target concrete based on the expected performance parameter information and the concrete mix proportion analysis model. This is beneficial to improving the comprehensiveness and rationality of the method for determining the optimal concrete mix proportion result, thereby improving the accuracy and efficiency of concrete mix proportion optimization, which in turn improves the accuracy and reliability of the determined concrete mix proportion, improves the efficiency and convenience of determining the concrete mix proportion, and further reduces the amount of cement used while ensuring concrete performance, and improves the compressive strength of concrete while ensuring that the concrete slump meets the established standard.

[0129] In an optional embodiment, the basic principal characteristics include sand ratio characteristics, crushed stone ratio characteristics, cement content characteristics, water-cement ratio characteristics, and water-reducing agent characteristics; the test data for each performance and factor include first test data for standard performance parameters and second test data for the basic principal characteristics that are matched thereto.

[0130] Furthermore, the key principal characteristics of the target concrete determined based on all performance and factor test data may include:

[0131] Based on all performance and factor test data, the first target sub-key feature that meets the preset sensitivity conditions is selected from all basic sub-key features;

[0132] Based on all performance and factor test data, second target sub-key features that meet the preset significance conditions are selected from all first target sub-key features;

[0133] Based on all the secondary target sub-key characteristics, the key key characteristics of the target concrete are determined.

[0134] Optionally, the above-mentioned satisfaction of the preset sensitivity conditions can be determined by the range value of the basic sub-primary factor characteristics, and the embodiments of the present invention are not limited thereto.

[0135] Optionally, the above-mentioned satisfaction of the preset significance condition can be determined by the difference value of the basic sub-primary factor features, and the embodiments of the present invention are not limited thereto.

[0136] Optionally, all second target sub-key features may be identified as key key features of the target concrete; however, this embodiment of the invention does not impose any limitations on this.

[0137] As can be seen, this optional embodiment can determine the key characteristics of the target concrete through sensitivity screening conditions and significance screening conditions, which is conducive to improving the comprehensiveness, rationality and progressiveness of the key characteristics determination method, and further conducive to improving the diversity, flexibility and pertinence of the screening conditions used to screen out key characteristics, thereby improving the accuracy and reliability of the determined key characteristics.

[0138] In another optional embodiment, the process of selecting the first target sub-primary factor feature that meets the preset sensitivity conditions from all basic sub-primary factor features based on all performance and factor test data may include:

[0139] Based on all performance and factor test data, the results of the first sensitivity index for each basic sub-key characteristic at the slump level and the second sensitivity index for the compressive strength level were determined.

[0140] Based on the results of the first sensitivity index for each basic sub-primary factor feature, the first sensitivity ranking is determined, and based on the results of the second sensitivity index for each basic sub-primary factor feature, the second sensitivity ranking is determined.

[0141] Based on the first sensitivity ranking and the second sensitivity ranking, the first target sub-primary feature that meets the preset ranking conditions is determined from all basic sub-primary features.

[0142] Optionally, the results of the first and second sensitivity indicators may include, but are not limited to, the range; furthermore, they can be obtained through... This formula is used for range calculation, and the embodiments of the present invention are not limited thereto.

[0143] Optionally, in the range analysis, Ri represents the range of the i-th factor, reflecting the magnitude of the influence of the i-th factor on concrete performance (such as slump, compressive strength, etc.). The larger the range, the more significant the regulatory effect of the factor on performance. : The performance mean of the i-th factor at the j-th level, where i: factor index (i=1,2,…,n, where n is the total number of factors), for example, in the optimization of C50 concrete mix proportion, i can correspond to factors such as sand ratio (i=1), cement content (i=2), water-reducing agent (i=3); j: level index (j=1,2,…,m, where m is the number of levels), for example, if the sand ratio is set to 5 levels (34%, 36%, 38%, 40%, 42%), then j=1 corresponds to 34%, j=2 corresponds to 36%, and so on; The maximum value of the water average for the i-th factor, and the water average for all factors of the i-th factor. The maximum value represents the performance of the factor at its optimal level. The minimum value of the average value of all water values ​​of the i-th factor is taken as the minimum value of the average value of all water values ​​of the i-th factor, which represents the performance of the factor at the worst level. This embodiment of the invention does not limit the specific value.

[0144] Further optional, the above-mentioned preset sorting conditions can be illustrated as follows: from all basic sub-primary factors, the top three basic sub-primary factors with the highest range values ​​are determined as the first target sub-primary factors. That is, for the slump factor: cement content (C) > crushed stone ratio (B) > sand ratio (A); for the strength factor: cement content (C) > water-reducing agent (E) > sand ratio (A). Therefore, the first target sub-primary factors are determined to include cement content, crushed stone ratio, sand ratio, and water-reducing agent. This embodiment of the invention does not limit the specific factors.

[0145] Optionally, by comparing the R values ​​of each factor i The value can determine the priority of performance regulation, but this embodiment of the invention does not limit it.

[0146] As can be seen, this optional embodiment can determine the first sensitivity ranking of each basic sub-primary feature to the slump level and the second sensitivity ranking to the compressive strength level, and then determine the first target sub-primary feature that meets the preset ranking conditions. This is beneficial to improving the comprehensiveness and rationality of the method for determining the first target sub-primary feature, and further beneficial to improving the diversity, flexibility and pertinence of the consideration levels (i.e. slump level and compressive strength level) used to determine the principal factor feature, thereby improving the accuracy and reliability of the determined first target sub-primary feature.

[0147] In another optional embodiment, the process of selecting second target sub-key features that meet preset significance conditions from all first target sub-key features based on all performance and factor test data may include:

[0148] Based on all performance and factor test data, determine the significance index results for each primary characteristic of the first target sub-sub ...

[0149] Based on the significance index results of each first target sub-primary feature, second target sub-primary features that meet the preset significance threshold conditions are determined from all first target sub-primary features.

[0150] Optionally, the significance index results can be represented by the variance value, which is not limited in this embodiment of the invention.

[0151] Optionally, the above-mentioned conditions satisfying the preset significance threshold are illustrated by, for example: determining the first target sub-primary feature whose variance value is greater than or equal to the preset variance threshold from all first target sub-primary features, and using it as the second target sub-primary feature. This embodiment of the invention does not limit this.

[0152] As can be seen, this optional embodiment can determine the significance index result of each first target sub-primary feature and then determine the second target sub-primary feature through a preset significance threshold condition. This is beneficial to improving the comprehensiveness and rationality of the method for determining the second target sub-primary feature, and thus to improving the accuracy and reliability of the determined second target sub-primary feature.

[0153] In yet another optional embodiment, before constructing and training a converged concrete mix proportion analysis model based on the determined dual hidden layer architecture scheme, input / output node configuration scheme, weight skew scheme, network dropout rate parameter, and learning rate parameter, the method may further include the following operations:

[0154] Based on the characteristics of sand ratio, cement content, and water-reducing agent, the model input node information is determined, and based on the slump parameter and compressive strength parameter, the model output node information is determined.

[0155] Based on the model input node information and model output node information, determine the input and output node configuration scheme for the concrete mix proportion analysis model;

[0156] Based on the preset Bayesian optimization method, the number of first nodes for the first hidden layer and the number of second nodes for the second hidden layer are determined, and the network dropout rate parameter and learning rate parameter are determined.

[0157] Based on the number of the first node and the number of the second node, a dual-hidden-layer architecture scheme is determined, with the first hidden layer preceding the second hidden layer, and the number of the first node being greater than the number of the second node;

[0158] Based on the determined first weight value for the slump parameter and the second weight value for the compressive strength parameter, a weight tilting scheme is determined, where the first weight value is less than the second weight value.

[0159] Optionally, a first weight value for the slump parameter and a second weight value for the compressive strength parameter may be determined based on actual engineering needs (i.e., priority objects), but this embodiment of the invention does not impose any limitations.

[0160] Optionally, examples can be given regarding the Bayesian optimization method, number of nodes, dropout rate parameter, and learning rate parameter: Bayesian search space: number of nodes (16, 32, 64), dropout rate (0.1, 0.4), learning rate (1e-4, 1e-2). This embodiment of the invention does not limit these parameters.

[0161] As can be seen, this optional embodiment can provide a series of methods for determining model construction training parameters, which is beneficial to improving the comprehensiveness, integrity, and rationality of the method for determining model construction training parameters, thereby improving the accuracy and reliability of the determined model construction training parameters. In addition, it provides a specific method for determining the input and output node configuration scheme, which is beneficial to improving the accuracy and reliability of the determined input and output node configuration scheme, thereby simplifying the model complexity and avoiding interference from non-critical factors on model prediction. Furthermore, the proposed dual-hidden-layer architecture scheme with the two hidden layers being a decreasing structure is beneficial to reducing model prediction errors, improving the application accuracy and reliability of the concrete mix proportion analysis model, and conforming to the nonlinear decay characteristics of concrete performance. In addition, the proposed weight tilting scheme is beneficial to improving the training and application fit and relevance of the model.

[0162] In yet another optional embodiment, the method may further include the following operations:

[0163] Determine the actual application scenario information for the target concrete;

[0164] Based on the optimal mix proportion results and actual application scenario information, determine whether the target concrete meets the preset conditions for additional additives.

[0165] When it is determined that the target concrete meets the conditions for adding additional materials, the additional required functions of the target concrete are determined based on the optimal mix proportion results and actual application scenario information; based on the additional required functions and the optimal mix proportion results, the types of additional materials to be added to the target concrete and their corresponding addition and mixing methods and material proportion schemes are determined.

[0166] Based on the optimal mix proportion results, the types of additional materials and their addition and mixing methods and material proportion schemes, determine whether the target concrete meets the preset conditions for further optimization of the mix proportion;

[0167] When it is determined that the target concrete does not meet the conditions for further optimization of the mix ratio, the final manufacturing plan corresponding to the target concrete is generated based on the optimal mix ratio result, the type of additional material and its addition mixing method and material ratio scheme.

[0168] When it is determined that the target concrete meets the conditions for further optimization of the mix ratio, the reverse performance impact on the target concrete is analyzed based on the type of additional material, its addition and mixing method, and the material proportion scheme. Based on the reverse performance impact, the optimal mix ratio result is adjusted and updated accordingly to obtain the adjusted and updated optimal mix ratio result. Based on the adjusted and updated optimal mix ratio result, the type of additional material, its addition and mixing method, and the material proportion scheme, the final manufacturing scheme corresponding to the target concrete is generated.

[0169] Optionally, the above determination of whether the target concrete meets the preset conditions for additional added substances can be illustrated by the following examples: For instance, in actual application scenarios, in addition to requirements for the target concrete in terms of slump performance and compressive strength performance, it is also required that the target concrete exhibits an unconventional color. In this case, it is necessary to add pigments or other substances to the target concrete. Furthermore, it is required that the target concrete has an extraordinary antioxidant function. In this case, it is necessary to add additional substances to the target concrete to enhance the antioxidant function. In such cases, it is determined that the preset conditions for additional added substances are met. Conversely, the same applies. This embodiment of the invention does not impose any limitations.

[0170] Optionally, the above-mentioned determination of the additional material types to be added to the target concrete, along with their corresponding mixing methods and proportions, based on additional functional requirements and optimal mix proportions, may include:

[0171] Based on the additional required functionality, determine the types of substances that can realize that additional required functionality;

[0172] Based on the optimal mix proportion results, the material types that will not conflict with the current material of the target concrete are determined from all available material types, and these are used as additional material types to be added to the target concrete.

[0173] Based on the current composition of the target concrete, determine the appropriate addition and mixing method for the type of additional material.

[0174] Determine the impact of the optimal mixing ratio on the conventional functional formulation of additional substance types, and based on the impact, additional required functions, and the performance realization mixing rules of the determined additional substance types, determine the corresponding substance mixing scheme for the additional substance types.

[0175] Optionally, the above-mentioned determination of whether the target concrete meets the preset mixing ratio optimization conditions can be illustrated as follows: if the slump performance and compressive strength performance of the target concrete are weakened due to the reaction between the added substances and the original concrete materials or special addition methods, then the initially determined optimal mix proportion results need to be adjusted to achieve both slump performance and compressive strength performance as well as the additional functional requirements corresponding to the actual application scenario information. In this case, the target concrete is considered to meet the preset mixing ratio optimization conditions, and vice versa. This embodiment of the invention does not limit the scope of the invention.

[0176] As can be seen, this optional embodiment can perform the corresponding final manufacturing scheme generation operation for the results of satisfying the conditions of additional added materials and the results of satisfying the conditions of further optimization of the mixing ratio. This is conducive to improving the comprehensiveness and rationality of the final manufacturing scheme generation method, and thus conducive to improving the diversity, flexibility and pertinence of the final manufacturing scheme generation method. In this way, it is conducive to improving the accuracy and reliability of the generated final manufacturing scheme. In addition, it is also conducive to improving the timeliness, accuracy and efficiency of adjusting and updating the optimal mixing ratio results.

[0177] In another optional embodiment, the above-mentioned determination of whether the target concrete meets the preset conditions for additional additives based on the optimal mix proportion results and actual application scenario information may include:

[0178] Based on the optimal mix ratio, determine the actual performance type and effect, and based on the actual application scenario information, determine the required performance type and effect;

[0179] Determine whether the actual performance type and effect fully encompass the required performance type and effect;

[0180] When it is determined that the actual performance type and effect fully encompass the required performance type and effect, it is determined that the target concrete does not meet the preset additional material conditions.

[0181] When it is determined that the actual performance type and effect do not fully encompass the required performance type and effect, the target concrete is confirmed to meet the preset conditions for additional additives.

[0182] Optionally, the above determination of whether the actual performance type and effect fully encompass the required performance type and effect may include:

[0183] Determine whether the actual performance type fully includes the required performance type to obtain a first judgment result; if the first judgment result is negative, determine that the actual performance type and effect do not fully include the required performance type and effect; if the first judgment result is positive, determine whether the actual performance effect fully includes the required performance effect to obtain a second judgment result; if the second judgment result is positive, determine that the actual performance type and effect fully include the required performance type and effect; if the second judgment result is negative, determine that the actual performance type and effect do not fully include the required performance type and effect.

[0184] As can be seen, this optional embodiment can determine the result of satisfying the additional material conditions by determining the inclusion of the actual performance type and effect with the required performance type and effect, which is conducive to improving the comprehensiveness and rationality of the method for determining the result of satisfying the additional material conditions, and thus conducive to improving the accuracy and reliability of the determined result of satisfying the additional material conditions.

[0185] In another optional embodiment, the above-mentioned determination of whether the target concrete meets the preset conditions for further optimization of the mix proportion based on the optimal mix proportion results, the type of additional material and its addition and mixing method and material proportion scheme may include:

[0186] Based on the optimal mix proportion results, the type of additional materials and their addition and mixing methods and material proportion schemes, determine whether the target concrete meets the preset performance weakening conditions.

[0187] When it is determined that the target concrete meets the condition of weakened performance, the target concrete is determined to meet the preset condition of further optimization of the mixing ratio.

[0188] When it is determined that the target concrete does not meet the performance weakening condition, it is determined that the target concrete does not meet the preset mixing ratio optimization condition.

[0189] Optionally, the above-mentioned performance weakening condition is satisfied for example: whether adding additional material types and their mixing methods and material proportions will cause the slump performance and / or compressive strength performance of the target concrete to deteriorate based on the initial optimal mix proportion. If so, the preset performance weakening condition is satisfied; otherwise, the preset performance weakening condition is not satisfied. This embodiment of the invention does not limit this.

[0190] As can be seen, this optional embodiment can determine the result of satisfying the condition for further optimization of the mixing ratio by determining the result of satisfying the condition for weakened performance, which is beneficial to improving the comprehensiveness and rationality of the method for determining the result of satisfying the condition for further optimization of the mixing ratio, and thus beneficial to improving the accuracy and reliability of the result of satisfying the condition for further optimization of the mixing ratio.

[0191] Example 2

[0192] Please see Figure 2 , Figure 2 This is a schematic flowchart of another concrete mix proportion optimization method based on neural networks disclosed in an embodiment of the present invention. Figure 2 The described method can be applied to a neural network-based concrete mix design optimization system, wherein the system may include a server, which may be a local server or a cloud server; this embodiment of the invention is not limited thereto. Figure 2 As shown, the concrete mix design optimization method based on neural networks includes the following operations:

[0193] 201. Determine the standard performance parameters of the target concrete. The standard performance parameters include slump parameters and compressive strength parameters and their corresponding standard parameter value ranges.

[0194] 202. Determine the fundamental characteristics that can affect concrete performance, and generate multiple sets of performance and factor test data based on standard performance parameter information, fundamental characteristics, and the determined orthogonal test method.

[0195] 203. Based on all performance and factor test data, determine the key principal characteristics of the target concrete. The key principal characteristics include sand ratio characteristics, cement content characteristics, and water-reducing agent characteristics.

[0196] 204. Based on the determined double hidden layer architecture scheme, input and output node configuration scheme, weight tilt scheme, network dropout rate parameter and learning rate parameter, construct and train a converged concrete mix proportion analysis model.

[0197] 205. Based on the determined expected performance parameters of the target concrete and the concrete mix proportion analysis model, the optimal mix proportion of the target concrete is obtained.

[0198] 206. Based on the optimal mix proportion results and the concrete mix proportion analysis model, determine the quantitative characteristic contribution results.

[0199] 207. Based on the contribution results of the quantified features, generate three-dimensional response surface content. The three-dimensional response surface content is used to show the interaction effect between sand ratio feature and cement content feature, and cement content feature and water-reducing agent feature.

[0200] In this embodiment of the invention, for other descriptions of steps 201-207, please refer to the other detailed descriptions of steps 101-105 in Embodiment 1. These descriptions will not be repeated in this embodiment of the invention.

[0201] As can be seen, the embodiments of the present invention can determine the key principal characteristics of the target concrete based on multiple sets of performance and factor test data, obtain a converged concrete mix proportion analysis model based on a series of determined model construction and training parameters, and obtain the optimal mix proportion result of the target concrete based on the expected performance parameter information and the concrete mix proportion analysis model. This is beneficial to improving the comprehensiveness and rationality of the method for determining the optimal concrete mix proportion result, thereby improving the accuracy and efficiency of concrete mix proportion optimization, which in turn improves the accuracy and reliability of the determined concrete mix proportion, improves the efficiency and convenience of determining the concrete mix proportion, further reduces the amount of cement used while ensuring concrete performance, and improves the compressive strength of concrete while ensuring that the concrete slump meets the established standard. In addition, it can also provide a method for determining the three-dimensional response surface content for sand ratio characteristics, cement content characteristics, and water-reducing agent characteristics, which is beneficial to improving the diversity and flexibility of the intelligent functions of this system, and improving the convenience and efficiency of viewing the contribution of different characteristics.

[0202] Example 3

[0203] Please see Figure 3 , Figure 3 This is a schematic diagram of a concrete mix proportion optimization system based on a neural network, as disclosed in an embodiment of the present invention. Figure 3 The described system may include a server, which may be a local server or a cloud server; this embodiment of the invention does not limit the scope. Figure 3 As shown, the neural network-based concrete mix design optimization system may include:

[0204] The performance parameter determination module 301 is used to determine the standard performance parameter information of the target concrete. The standard performance parameter information includes slump parameter and compressive strength parameter and their corresponding standard parameter value range.

[0205] The first feature determination module 302 is used to determine the basic key features that can affect the performance of concrete.

[0206] The test data generation module 303 is used to generate multiple sets of performance and factor test data based on standard performance parameter information, basic main factor characteristics and determined orthogonal test methods.

[0207] The second feature determination module 304 is used to determine the key principal features of the target concrete based on all performance and factor test data. The key principal features include sand ratio, cement content and water-reducing agent characteristics.

[0208] The model building and training module 305 is used to build and train a converged concrete mix proportion analysis model based on the determined double hidden layer architecture scheme, input and output node configuration scheme, weight tilt scheme, network dropout rate parameter and learning rate parameter.

[0209] The mix proportion determination module 306 is used to obtain the optimal mix proportion result of the target concrete based on the determined expected performance parameter information of the target concrete and the concrete mix proportion analysis model.

[0210] It is evident that implementation Figure 3The described neural network-based concrete mix design optimization system can determine the key characteristics of the target concrete based on multiple sets of performance and factor test data. Based on a series of determined model construction and training parameters, a converged concrete mix design analysis model is obtained. Based on the expected performance parameters and the concrete mix design analysis model, the optimal mix design result of the target concrete is obtained. This improves the comprehensiveness and rationality of the method for determining the optimal concrete mix design result, thereby improving the accuracy and efficiency of concrete mix design optimization. This, in turn, improves the accuracy and reliability of the determined concrete mix design, increases the efficiency and convenience of determining the concrete mix design, and further reduces the amount of cement used while ensuring concrete performance. It also improves the compressive strength of the concrete while ensuring that the concrete slump meets the established standards.

[0211] In an optional embodiment, the basic principal characteristics include sand ratio characteristics, crushed stone ratio characteristics, cement content characteristics, water-cement ratio characteristics, and water-reducing agent characteristics; the test data for each performance and factor include first test data for standard performance parameters and second test data for the basic principal characteristics that are matched thereto.

[0212] Furthermore, the second feature determination module 304 determines the key principal features of the target concrete based on all performance and factor test data in the following specific ways:

[0213] Based on all performance and factor test data, the first target sub-key feature that meets the preset sensitivity conditions is selected from all basic sub-key features;

[0214] Based on all performance and factor test data, second target sub-key features that meet the preset significance conditions are selected from all first target sub-key features;

[0215] Based on all the secondary target sub-key characteristics, the key key characteristics of the target concrete are determined.

[0216] It is evident that implementation Figure 4 The described system can identify the key characteristics of target concrete through sensitivity screening and significance screening conditions. This helps to improve the comprehensiveness, rationality and progressiveness of the key characteristic identification method, and in turn, it helps to improve the diversity, flexibility and pertinence of the screening conditions used to identify key characteristics, thereby improving the accuracy and reliability of the identified key characteristics.

[0217] In another optional embodiment, the second feature determination module 304 selects the first target sub-primary factor feature that meets the preset sensitivity condition from all basic sub-primary factor features based on all performance and factor test data. Specifically, this includes:

[0218] Based on all performance and factor test data, the results of the first sensitivity index for each basic sub-key characteristic at the slump level and the second sensitivity index for the compressive strength level were determined.

[0219] Based on the results of the first sensitivity index for each basic sub-primary factor feature, the first sensitivity ranking is determined, and based on the results of the second sensitivity index for each basic sub-primary factor feature, the second sensitivity ranking is determined.

[0220] Based on the first sensitivity ranking and the second sensitivity ranking, the first target sub-primary feature that meets the preset ranking conditions is determined from all basic sub-primary features.

[0221] It is evident that implementation Figure 4 The described system can also determine the first sensitivity ranking of each basic sub-primary feature for the slump level and the second sensitivity ranking for the compressive strength level, and then determine the first target sub-primary feature that meets the preset ranking conditions. This helps to improve the comprehensiveness and rationality of the method for determining the first target sub-primary feature, and in turn, it helps to improve the diversity, flexibility and pertinence of the consideration levels (i.e. slump level and compressive strength level) used to determine the principal factor feature, thereby helping to improve the accuracy and reliability of the determined first target sub-primary feature.

[0222] In another optional embodiment, the second feature determination module 304 selects second target sub-key features that meet preset significance conditions from all first target sub-key features based on all performance and factor test data in the following specific ways:

[0223] Based on all performance and factor test data, determine the significance index results for each primary characteristic of the first target sub-sub ...

[0224] Based on the significance index results of each first target sub-primary feature, second target sub-primary features that meet the preset significance threshold conditions are determined from all first target sub-primary features.

[0225] It is evident that implementation Figure 4 The described system can also determine the significance index results of each first target sub-primary causal feature and then determine the second target sub-primary causal feature by setting a significance threshold condition. This helps to improve the comprehensiveness and rationality of the determination method of the second target sub-primary causal feature, and thus helps to improve the accuracy and reliability of the determined second target sub-primary causal feature.

[0226] In yet another alternative embodiment, such as Figure 4 As shown, the system may also include:

[0227] The model parameter determination module 307 is used to determine the model input node information based on sand ratio characteristics, cement content characteristics, and water-reducing agent characteristics, and to determine the model output node information based on slump parameters and compressive strength parameters, before the model construction and training module 305 constructs and trains a converged concrete mix proportion analysis model based on the determined double hidden layer architecture scheme, input and output node configuration scheme, weight tilting scheme, network dropout rate parameters, and learning rate parameters. Based on the model input and output node information, the module determines the input and output node configuration scheme for the concrete mix proportion analysis model. Based on a preset Bayesian optimization method, it determines the number of first nodes for the first hidden layer and the number of second nodes for the second hidden layer, and determines the network dropout rate parameters and learning rate parameters. Based on the number of first and second nodes, it determines the double hidden layer architecture scheme, with the first hidden layer preceding the second hidden layer and the number of first nodes greater than the number of second nodes. Based on the determined first weight value for the slump parameter and the second weight value for the compressive strength parameter, it determines the weight tilting scheme, with the first weight value less than the second weight value.

[0228] It is evident that implementation Figure 4 The described system also provides a series of methods for determining model construction training parameters, which helps improve the comprehensiveness, integrity, and rationality of the parameter determination methods, thereby improving the accuracy and reliability of the determined model construction training parameters. In addition, it provides specific methods for determining input and output node configuration schemes, which helps improve the accuracy and reliability of the determined input and output node configuration schemes, thereby simplifying model complexity and avoiding interference from non-critical factors in model prediction. Furthermore, it proposes a dual-hidden-layer architecture scheme with the two hidden layers having a decreasing structure, which helps reduce model prediction errors and improve the application accuracy and reliability of concrete mix proportion analysis models, conforming to the nonlinear decay characteristics of concrete performance. In addition, it proposes a weight tilting scheme, which helps improve the training and application fit and relevance of the model.

[0229] In yet another alternative embodiment, such as Figure 4 As shown, the system may also include:

[0230] The response surface determination module 308 is used to determine the quantitative feature contribution results based on the optimal mix proportion results and the concrete mix proportion analysis model; based on the quantitative feature contribution results, it generates three-dimensional response surface content, which is used to display the interaction effect of sand ratio feature-cement quantity feature and cement quantity feature-water reducing agent feature.

[0231] It is evident that implementation Figure 4The described system can also provide a method for determining the content of three-dimensional response surfaces for sand ratio characteristics, cement quantity characteristics, and water-reducing agent characteristics, which is conducive to improving the diversity and flexibility of the intelligent functions of this system, and to improving the convenience and efficiency of viewing the contribution of different characteristics.

[0232] In yet another alternative embodiment, such as Figure 4 As shown, the system may also include:

[0233] The judgment module 309 is used to determine the actual application scenario information of the target concrete; based on the optimal mix proportion results and the actual application scenario information, it determines whether the target concrete meets the preset conditions for additional additives.

[0234] The additional material determination module 310 is used to determine the additional required functions of the target concrete based on the optimal mix proportion results and actual application scenario information when the judgment module 309 determines that the target concrete meets the conditions for adding additional materials; and to determine the type of additional material to be added to the target concrete and its corresponding addition and mixing method and material ratio scheme based on the additional required functions and the optimal mix proportion results.

[0235] The judgment module 309 is also used to determine whether the target concrete meets the preset conditions for further optimization of the mixing ratio based on the optimal mix proportion results, the type of additional material and its addition mixing method and material proportion scheme.

[0236] The manufacturing scheme determination module 311 is used to generate the final manufacturing scheme corresponding to the target concrete based on the optimal mix proportion result, the type of additional material and its addition mixing method and material proportion scheme when the judgment module 309 determines that the target concrete does not meet the conditions for further optimization of the mixing ratio.

[0237] The mix proportion determination module 306 is also used to analyze the reverse performance impact on the target concrete based on the type of additional material, its addition and mixing method, and the material proportion scheme when the judgment module 309 determines that the target concrete meets the conditions for further optimization of the mix proportion; based on the reverse performance impact, the optimal mix proportion result is adjusted and updated accordingly to obtain the adjusted and updated optimal mix proportion result.

[0238] The manufacturing scheme determination module 311 is also used to generate the final manufacturing scheme corresponding to the target concrete based on the adjusted and updated optimal mix proportion results, additional material types and their addition mixing methods and material proportion schemes.

[0239] It is evident that implementation Figure 4The described system can also perform corresponding final manufacturing scheme generation operations for the results of additional material addition conditions and the results of further optimization of mixing ratio conditions. This is conducive to improving the comprehensiveness and rationality of the final manufacturing scheme generation method, and thus to improving the diversity, flexibility and pertinence of the final manufacturing scheme generation method. This, in turn, is conducive to improving the accuracy and reliability of the generated final manufacturing scheme. In addition, it is also conducive to improving the timeliness, accuracy and efficiency of adjusting and updating the optimal mixing ratio results.

[0240] In another optional embodiment, the determination module 309 determines whether the target concrete meets the preset conditions for additional additives based on the optimal mix proportion results and actual application scenario information. Specifically, this includes:

[0241] Based on the optimal mix ratio, determine the actual performance type and effect, and based on the actual application scenario information, determine the required performance type and effect;

[0242] Determine whether the actual performance type and effect fully encompass the required performance type and effect;

[0243] When it is determined that the actual performance type and effect fully encompass the required performance type and effect, it is determined that the target concrete does not meet the preset additional material conditions.

[0244] When it is determined that the actual performance type and effect do not fully encompass the required performance type and effect, the target concrete is confirmed to meet the preset conditions for additional additives.

[0245] It is evident that implementation Figure 4 The described system can also determine the result of satisfying additional material conditions by comparing the actual performance type and effect with the inclusion of the required performance type and effect. This helps to improve the comprehensiveness and rationality of the method for determining the result of satisfying additional material conditions, and thus helps to improve the accuracy and reliability of the determined result of satisfying additional material conditions.

[0246] In another optional embodiment, the determination module 309 determines whether the target concrete meets the preset conditions for further optimization of the mix ratio based on the optimal mix ratio result, the type of additional material and its addition mixing method and material proportion scheme. Specifically, this includes:

[0247] Based on the optimal mix proportion results, the type of additional materials and their addition and mixing methods and material proportion schemes, determine whether the target concrete meets the preset performance weakening conditions.

[0248] When it is determined that the target concrete meets the condition of weakened performance, the target concrete is determined to meet the preset condition of further optimization of the mixing ratio.

[0249] When it is determined that the target concrete does not meet the performance weakening condition, it is determined that the target concrete does not meet the preset mixing ratio optimization condition.

[0250] It is evident that implementation Figure 4 The described system can also determine the result of satisfying the condition for further optimization of the mixing ratio by determining the result of satisfying the weakened performance condition. This helps to improve the comprehensiveness and rationality of the method for determining the result of satisfying the condition for further optimization of the mixing ratio, and thus helps to improve the accuracy and reliability of the result of satisfying the condition for further optimization of the mixing ratio.

[0251] Example 4

[0252] Please see Figure 5 , Figure 5 This is a schematic diagram of another concrete mix proportion optimization system based on a neural network disclosed in an embodiment of the present invention. Figure 5 The described system may include a server, which may be a local server or a cloud server; this embodiment of the invention does not limit the scope. Figure 5 As shown, the system may include:

[0253] Memory 401 storing executable program code;

[0254] Processor 402 coupled to memory 401;

[0255] Furthermore, it may also include an input interface 403 coupled to the processor 402 and an output interface 404;

[0256] The processor 402 calls the executable program code stored in the memory 401 to execute the steps in the neural network-based concrete mix proportion optimization method described in Embodiment 1 or Embodiment 2.

[0257] Example 5

[0258] This invention discloses a computer storage medium that stores a computer program for electronic data interchange, wherein the computer program causes a computer to execute the steps in the neural network-based concrete mix proportion optimization method described in Embodiment 1 or Embodiment 2.

[0259] Example 6

[0260] This invention discloses a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program, and the computer program is operable to cause a computer to perform the steps in the neural network-based concrete mix proportion optimization method described in Embodiment 1 or Embodiment 2.

[0261] The device embodiments described above are merely illustrative. The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; that is, they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0262] Through the detailed description of the above embodiments, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, including read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electrically-Erasable Programmable Read-Only Memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

[0263] Finally, it should be noted that the concrete mix proportion optimization method and system based on neural networks disclosed in the embodiments of the present invention are merely preferred embodiments of the present invention, and are only used to illustrate the technical solutions of the present invention, not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for optimizing concrete mix proportions based on neural networks, characterized in that, The method includes: Determine the standard performance parameter information of the target concrete, including slump parameter and compressive strength parameter and their corresponding standard parameter value range; The fundamental characteristics that can affect concrete performance are identified, and multiple sets of performance and factor test data are generated based on the standard performance parameter information, the fundamental characteristics, and the determined orthogonal test method. The fundamental characteristics include sand ratio, crushed stone ratio, cement content, water-cement ratio, and water-reducing agent characteristics. Each set of performance and factor test data includes first test data for the standard performance parameter and second test data for the corresponding fundamental characteristics. Based on all the performance and factor test data, the first target sub-primary factor feature that meets the preset sensitivity conditions is selected from all the basic sub-primary factor features; Based on all the performance and factor test data, select second target sub-key features that meet the preset significance conditions from all the first target sub-key features; Based on all the second target sub-key characteristics, the key key characteristics of the target concrete are determined, including sand ratio characteristics, cement content characteristics, and water-reducing agent characteristics. Based on the determined dual-hidden-layer architecture scheme, input / output node configuration scheme, weight skew scheme, network dropout rate parameter, and learning rate parameter, a converged concrete mix proportion analysis model is constructed and trained. Based on the determined expected performance parameters of the target concrete and the concrete mix proportion analysis model, the optimal mix proportion of the target concrete is obtained. Furthermore, before constructing and training a converged concrete mix proportion analysis model based on the determined dual-hidden-layer architecture scheme, input / output node configuration scheme, weight skew scheme, network dropout rate parameter, and learning rate parameter, the method further includes: Based on the sand ratio characteristics, the cement content characteristics, and the water-reducing agent characteristics, the model input node information is determined, and based on the slump parameter and the compressive strength parameter, the model output node information is determined. Based on the model input node information and the model output node information, determine the input and output node configuration scheme for the concrete mix proportion analysis model; Based on the preset Bayesian optimization method, the number of first nodes for the first hidden layer and the number of second nodes for the second hidden layer are determined, and the network dropout rate parameter and learning rate parameter are determined. Based on the number of the first node and the number of the second node, a dual-hidden-layer architecture scheme is determined, wherein the first hidden layer precedes the second hidden layer, and the number of the first node is greater than the number of the second node; Based on the determined first weight value for the slump parameter and the second weight value for the compressive strength parameter, a weight tilting scheme is determined, wherein the first weight value is less than the second weight value.

2. The concrete mix proportion optimization method based on neural networks according to claim 1, characterized in that, The step of selecting the first target sub-primary factor feature that meets the preset sensitivity conditions from all basic sub-primary factor features based on all the performance and factor test data includes: Based on all the performance and factor test data, determine the first sensitivity index result for each basic sub-key characteristic at the slump level and the second sensitivity index result at the compressive strength level. Based on the first sensitivity index result of each of the basic sub-primary features, the first sensitivity ranking is determined, and based on the second sensitivity index result of each of the basic sub-primary features, the second sensitivity ranking is determined. Based on the first sensitivity ranking and the second sensitivity ranking, a first target sub-primary feature that meets the preset ranking conditions is determined from all the basic sub-primary features. And, the step of selecting second target sub-key features that meet preset significance conditions from all the first target sub-key features based on all the performance and factor test data includes: Based on all the performance and factor test data, determine the significance index results for each of the first target sub-key characteristics; Based on the significance index results of each of the first target sub-primary features, the second target sub-primary features that meet the preset significance threshold conditions are determined from all the first target sub-primary features.

3. The concrete mix design optimization method based on neural networks according to claim 1, characterized in that, The method further includes: Based on the optimal mix proportion results and the concrete mix proportion analysis model, the quantitative characteristic contribution results are determined; Based on the quantified feature contribution results, a three-dimensional response surface is generated. The three-dimensional response surface is used to demonstrate the interaction effects of sand ratio feature-cement quantity feature and cement quantity feature-water reducing agent feature.

4. The concrete mix proportion optimization method based on neural networks according to any one of claims 1-3, characterized in that, The method further includes: Determine the actual application scenario information of the target concrete; Based on the optimal mix proportion results and the actual application scenario information, determine whether the target concrete meets the preset conditions for additional additives; When it is determined that the target concrete meets the conditions for the additional added material, the additional required functions of the target concrete are determined based on the optimal mix proportion results and the actual application scenario information; based on the additional required functions and the optimal mix proportion results, the types of additional materials to be added to the target concrete and their corresponding addition and mixing methods and material proportion schemes are determined. Based on the optimal mix proportion results, the type of additional material and its addition and mixing method and material proportion scheme, determine whether the target concrete meets the preset conditions for further optimization of the mix proportion; When it is determined that the target concrete does not meet the conditions for further optimization of the mixing ratio, the final manufacturing plan corresponding to the target concrete is generated based on the optimal mix ratio result, the type of additional material and its addition and mixing method and material ratio scheme. When it is determined that the target concrete meets the conditions for further optimization of the mix ratio, the reverse performance impact on the target concrete is analyzed based on the type of additional material, its addition and mixing method, and the material proportioning scheme. Based on the reverse performance impact, the optimal mix ratio result is adjusted and updated accordingly to obtain the adjusted and updated optimal mix ratio result. Based on the adjusted and updated optimal mix ratio result, the type of additional material, its addition and mixing method, and the material proportioning scheme, the final manufacturing scheme corresponding to the target concrete is generated.

5. The concrete mix proportion optimization method based on neural networks according to claim 4, characterized in that, The step of determining whether the target concrete meets the preset conditions for additional additives based on the optimal mix proportion results and the actual application scenario information includes: Based on the optimal mix ratio, the actual performance type and effect are determined, and based on the actual application scenario information, the required performance type and effect are determined. Determine whether the actual performance type and effect completely encompass the required performance type and effect; When it is determined that the actual performance type and effect completely encompass the required performance type and effect, it is determined that the target concrete does not meet the preset additional material conditions. When it is determined that the actual performance type and effect do not fully encompass the required performance type and effect, the target concrete is determined to meet the preset additional material conditions. And, the step of determining whether the target concrete meets the preset conditions for further optimization of the mix ratio based on the optimal mix ratio result, the type of additional material and its addition and mixing method and material proportion scheme includes: Based on the optimal mix proportion results, the type of additional material and its addition and mixing method and material proportion scheme, it is determined whether the target concrete meets the preset performance weakening condition; When it is determined that the target concrete meets the condition that the performance is weakened, it is determined that the target concrete meets the preset condition for further optimization of the mixing ratio. When it is determined that the target concrete does not meet the condition for weakened performance, it is determined that the target concrete does not meet the preset condition for further optimization of the mixing ratio.

6. A concrete mix design optimization system based on neural networks, characterized in that, The system includes: The performance parameter determination module is used to determine the standard performance parameter information of the target concrete. The standard performance parameter information includes slump parameter and compressive strength parameter and their corresponding standard parameter value ranges. The first feature determination module is used to determine the fundamental key features that can affect concrete performance. The test data generation module is used to generate multiple sets of performance and factor test data based on the standard performance parameter information, the basic main characteristics, and the determined orthogonal test method; the basic main characteristics include sand ratio characteristics, crushed stone ratio characteristics, cement content characteristics, water-cement ratio characteristics, and water-reducing agent characteristics; each set of performance and factor test data includes first test data for the standard performance parameters and second test data for the basic main characteristics that match them; The second feature determination module is used to: select first target sub-primary features that meet preset sensitivity conditions from all basic sub-primary features based on all the performance and factor test data; select second target sub-primary features that meet preset significance conditions from all the first target sub-primary features based on all the performance and factor test data; and determine the key primary features of the target concrete based on all the second target sub-primary features, wherein the key primary features include sand ratio features, cement content features, and water-reducing agent features. The model building and training module is used to build and train a converged concrete mix proportion analysis model based on the determined double hidden layer architecture scheme, input and output node configuration scheme, weight tilt scheme, network dropout rate parameter and learning rate parameter. The mix proportion determination module is used to obtain the optimal mix proportion result of the target concrete based on the determined expected performance parameter information of the target concrete and the concrete mix proportion analysis model; The model parameter determination module is used to determine the model input node information based on the sand ratio characteristics, cement content characteristics, and water-reducing agent characteristics, and to determine the model output node information based on the slump parameter and compressive strength parameter, before the model construction and training module constructs and trains a converged concrete mix proportion analysis model according to the determined double hidden layer architecture scheme, input and output node configuration scheme, weight tilting scheme, network dropout rate parameter, and learning rate parameter. Based on the model input node information and the model output node information, the module determines the input and output node configuration scheme for the concrete mix proportion analysis model. Based on a preset Bayesian optimization method, it determines the number of first nodes for the first hidden layer and the number of second nodes for the second hidden layer, and determines the network dropout rate parameter and learning rate parameter. Based on the number of first nodes and the number of second nodes, it determines the double hidden layer architecture scheme, where the first hidden layer precedes the second hidden layer, and the number of first nodes is greater than the number of second nodes. Based on the determined first weight value for the slump parameter and the second weight value for the compressive strength parameter, it determines the weight tilting scheme, where the first weight value is less than the second weight value.

7. A concrete mix design optimization system based on neural networks, characterized in that, The system includes: Memory containing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the concrete mix proportion optimization method based on neural networks as described in any one of claims 1-5.

8. A computer storage medium, characterized in that, The computer storage medium stores computer instructions, which, when invoked, are used to execute the neural network-based concrete mix proportion optimization method as described in any one of claims 1-5.

Images