Geopolymer batching method, system and control method based on neural networks

Abstract

The application discloses a kind of geopolymer batching method, system and control method based on neural network.The scheme can rely on the polymer batching admixture prediction model after training, according to the input geopolymer engineering mechanics characteristic information, the corresponding each batching admixture is predicted, then using material control module and stirring module, respectively control feeding system and stirring system, material is transported and stirred and mixed, and the resistivity of batching mixture in stirring module is monitored in real time using resistivity detection device, and the resistivity of batching mixture measured in real time is input into feedback regulation module and is converted into resistivity distribution image, and the resistivity distribution image information obtained by conversion is fed back to the polymer batching admixture prediction model by feedback regulation module, and new batching admixture information is obtained by recalculation, and the batching in batching mixture is adjusted in real time, the production process can be controlled in real time, and batching admixture can be controlled more accurately.

Description

Technical Field

[0001] This invention relates to the field of geopolymer manufacturing, and more specifically to a method, system, and control method for geopolymer batching based on neural networks. Background Technology

[0002] Reducing material costs and resource consumption, as well as protecting the environment, are important trends in the development of building materials in the 21st century. Because the raw materials required for preparing geopolymers are low-cost, abundant, and the preparation process is simple, with low pollution and energy consumption, geopolymer materials are a new type of building material that aligns with the overall trend of structural material development and possesses enormous potential.

[0003] Geopolymers are a new type of inorganic cementitious material that uses industrial waste such as fly ash and slag as the main raw materials and reacts them with alkaline activators (such as sodium hydroxide and potassium hydroxide). It not only has cost-effectiveness, environmental friendliness, good mechanical properties, corrosion resistance and durability, but also can harden quickly in a short time, shortening the construction cycle, thus showing its unique advantages in the construction and engineering fields.

[0004] Specifically, its cost-effectiveness stems from the fact that its raw material is industrial waste. At the same time, the use of industrial waste reduces waste disposal problems and reduces reliance on traditional cement, thereby reducing energy consumption and emissions in the production process. In addition, it has good corrosion resistance and weather resistance, making it suitable for a variety of environmental conditions.

[0005] Under appropriate proportions and processing conditions, geopolymers can exhibit excellent mechanical properties, including compressive strength, tensile strength, and shear strength. Another significant advantage of geopolymers compared to traditional cement is their faster hardening time, thus shortening the construction cycle.

[0006] Generally, the preparation process of geopolymers includes three steps: raw material preparation, mixing, and curing. First, raw materials such as fly ash and slag are mixed with an alkaline activator. For mixing, existing technologies often use a rate control method, that is, a batching system that automatically controls the raw material rate to mix these raw materials in a certain proportion and under certain conditions to form a uniform slurry. Finally, the mixed slurry is cured under suitable conditions to form a solid material with certain mechanical properties.

[0007] Because geopolymers contain multiple components, their properties can be positively or negatively affected by differences in raw material ratios and preparation processes. However, existing batching methods lack real-time monitoring and feedback, which can easily lead to discrepancies between actual production products and theoretical values. This can result in products failing to meet target performance and hindering the stability of geopolymer quality. Therefore, precisely controlling the content of each ingredient is one of the key challenges in achieving the desired product.

[0008] Therefore, it is necessary to develop a precise ingredient dosing system or method that can accurately control the initial dosage of ingredients, and also provide real-time monitoring and feedback during the preparation process, so as to ensure stable product quality and achieve geopolymers that meet performance targets. Summary of the Invention

[0009] The purpose of this invention is to provide a method, system, and control method for batching geopolymers based on neural networks, in order to solve the technical problem that the batching system in the prior art lacks real-time monitoring and feedback during the batching process, which easily leads to a certain difference between the actual production product and the theoretical value, resulting in the product performance failing to reach the target value and making it difficult to ensure the quality stability of geopolymers.

[0010] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:

[0011] In a first aspect, the present invention discloses a method for preparing geopolymers based on neural networks, comprising the following steps:

[0012] S1. Obtain training sample instances of geopolymers;

[0013] S2. The polymer ingredient dosage prediction model is trained based on the training sample examples of geopolymers to obtain appropriate neural network weights and establish the trained polymer ingredient dosage prediction model. The trained polymer ingredient dosage prediction model is used to indicate the mapping relationship between the engineering mechanical properties of geopolymers and the dosage of each ingredient and the resistivity of geopolymers.

[0014] S3. Input the preset engineering mechanical properties of the geopolymer, and use the trained polymer batching dosage prediction model obtained in step S2 to analyze and obtain the initial dosage information of each batching and the resistivity of the target geopolymer.

[0015] S4. Based on the initial dosage information of each ingredient obtained in step S3, mix and stir each ingredient to obtain a mixture of ingredients;

[0016] S5. Measure the resistivity distribution image of the ingredient mixture in real time and feed the real-time measured resistivity distribution image of the ingredient mixture back to the trained polymer ingredient dosage prediction model.

[0017] S6. The polymer ingredient dosage prediction model determines whether the received resistivity distribution image meets the target requirements. If the target requirements are met, proceed to step S8; otherwise, proceed to step S7.

[0018] S7. The polymer ingredient dosage prediction model recalculates new ingredient dosage information based on the mapping relationship between the engineering mechanical properties of geopolymers and the resistivity of geopolymers; and adjusts the dosage of each ingredient according to the new ingredient dosage information, continues mixing and stirring, and repeats steps S5 to S6.

[0019] S8. Stop feeding and stirring, and obtain the final geopolymer.

[0020] This solution addresses the problem in existing batching systems that lack real-time monitoring and feedback during batching control. This lack of monitoring leads to discrepancies between actual and theoretical values ​​in the final product, resulting in unmet performance targets and inconsistent geopolymer quality. The neural network-based geopolymer batching method in this solution provides real-time monitoring and feedback during the batching process, enabling real-time control of the production process. This ensures the mechanical properties and stability of the resulting geopolymer, guaranteeing stable product quality and achieving target performance values.

[0021] As a preferred embodiment, in the polymer ingredient dosage prediction model, the mapping relationship between the engineering mechanical properties of geopolymers and the dosage of each ingredient and the resistivity of the geopolymer is expressed as follows:

[0022] Y1=a1X1+a2X2+a3X3+a4X4+a5X5+a6X6

[0023] Y2=b1X1+b2X2+b3X3+b4X4+b5X5+b6X6

[0024] Y3=c1X1+c2X2+c3X3+c4X4+c5X5+c6X6

[0025] Where Y1 is the unconfined compressive strength; Y2 is the tensile strength; Y3 is the shear strength; X1, a1, b1, and c1 are the alkali activator dosage and the corresponding unconfined compressive strength coefficient, tensile strength coefficient, and shear strength coefficient of the alkali activator, respectively.

[0026] X2, a2, b2, and c2 represent the slag powder content and the corresponding unconfined compressive strength, tensile strength, and shear strength coefficients, respectively; X3, a3, b3, and c3 represent the fly ash content and the corresponding unconfined compressive strength, tensile strength, and shear strength coefficients, respectively; X4, a4, b4, and c4 represent the water content and the corresponding unconfined compressive strength, tensile strength, and shear strength coefficients, respectively; X5, a5, b5, and c5 represent the admixture content and the corresponding unconfined compressive strength, tensile strength, and shear strength coefficients, respectively; X6, a6, b6, and c6 represent the sand content and the corresponding unconfined compressive strength, tensile strength, and shear strength coefficients, respectively.

[0027] Preferably, in step S1, the polymer ingredient dosage prediction model is one of the following: convolutional neural network, feedforward neural network, backpropagation neural network, recurrent neural network, and long short-term memory network; the loss function used to train the polymer ingredient dosage prediction model with geological polymer training sample instances is one of the following: cross-entropy loss function, mean squared error loss function, and log-likelihood loss.

[0028] Preferably, in step S5, the resistivity distribution image of the ingredient mixture is measured using the resistivity method, including the following steps:

[0029] S5.1 A series of electrodes are arranged in the ingredient mixture to form an electrode array for applying current to the ingredient mixture and measuring the voltage in the ingredient mixture;

[0030] S5.2 Apply current to the mixture of ingredients through an electrode array;

[0031] S5.3 Measure the voltage between different electrode pairs;

[0032] S5.4 By changing the combination of electrode pairs, different voltage data are collected to form a voltage data matrix;

[0033] S5.5. Using a mathematical inversion algorithm, the resistivity distribution in the mixture is calculated from the obtained voltage data matrix.

[0034] S5.6. Display the resistivity data obtained by inversion in the form of an image to form a resistivity distribution image of the ingredient mixture, and feed back the real-time measured resistivity distribution image of the ingredient mixture to the trained polymer ingredient dosage prediction model.

[0035] Preferably, before performing step S5.5, the collected voltage data needs to be preprocessed, such as through noise reduction and correction, to improve data quality.

[0036] Secondly, the present invention also discloses a neural network-based geopolymer batching system for implementing the neural network-based geopolymer batching method described above, including a central processing unit, a feeding system, a stirring system, and a resistivity detection device.

[0037] The feeding system includes a material storage chamber for storing the raw materials of each ingredient. Each raw material in the material storage chamber has an independent storage space, and each storage space has a material outlet. Each material outlet is equipped with an electrically controlled valve, and each electrically controlled valve is electrically connected to a central processing unit. The central processing unit can control the opening and closing of each electrically controlled valve and the degree of opening.

[0038] The mixing system includes a mixing container, and a stirrer is installed inside the mixing container for stirring the various ingredients input into the mixing container from the feeding system. The power supply terminal of the stirrer is electrically connected to the central processing unit, and the central processing unit can control the switch of the stirrer.

[0039] The stirring system is equipped with a sensing end of a resistivity detection device, which is electrically connected to the central processing unit. The resistivity detection device is used to monitor the real-time resistivity of the mixture in the mixing container, and the data transmission end of the resistivity detection device is used to feed back the real-time resistivity information of the mixture to the central processing unit.

[0040] Preferably, the central processing unit includes a polymer ingredient dosage prediction model module, a material control module, a stirring module, and a feedback adjustment module;

[0041] The polymer batching dosage prediction model module is used to receive the input geopolymer engineering mechanical properties information and the real-time measured resistivity distribution image information of the batching mixture fed back by the feedback adjustment module, and convert the geopolymer engineering mechanical properties information or the real-time measured resistivity distribution image information of the batching mixture into batching information, and transmit the batching information to the material control module.

[0042] The material control module is used to receive the batching information transmitted by the polymer batching dosage prediction model module, and control the feeding system to deliver each batching ingredient into the mixing system according to the batching information.

[0043] The stirring module is used to control the opening and closing of the stirrer to mix and stir the various ingredients to obtain a mixture of ingredients;

[0044] The feedback adjustment module is used to receive the resistivity data of the ingredient mixture measured in real time by the resistivity detection device, convert the resistivity data of the ingredient mixture into a resistivity distribution image, and feed the resistivity distribution image back to the polymer ingredient dosage prediction model module.

[0045] This solution's neural network-based geopolymer batching system can predict the corresponding dosage of each ingredient based on the input geopolymer engineering mechanical properties, using a trained polymer batching dosage prediction model. Then, a material control module and a mixing module control the feeding and mixing systems respectively to transport and mix the materials. Simultaneously, a resistivity detection device monitors the resistivity of the batching mixture in the mixing module in real time and sends the measured resistivity to a feedback adjustment module, converting it into a resistivity distribution image. This feedback adjustment module then feeds the converted resistivity distribution image back to the polymer batching dosage prediction model. Based on the mapping relationship between the geopolymer engineering mechanical properties and the geopolymer resistivity, new ingredient dosage information is recalculated. Finally, the material control and mixing modules control the feeding and mixing systems respectively to adjust the dosage of the ingredients in the batching mixture in real time. Once the required dosage is achieved, the mixing system is stopped.

[0046] Preferably, after receiving the resistivity data of the mixture in the mixing container monitored in real time by the resistivity detection device, the feedback module converts the original high-density data file into RES2DINV format, then performs inversion using least squares, outputs the inverted resistivity distribution image, and feeds the resistivity distribution image back to the polymer ingredient dosage prediction model module.

[0047] Thirdly, the present invention also discloses a control method for a geopolymer batching system based on a neural network, used to control the aforementioned geopolymer batching system based on a neural network, comprising the following steps:

[0048] A1. Input the preset geopolymer engineering mechanical properties into the polymer batching dosage prediction model, and obtain the initial dosage information of each ingredient and the target geopolymer resistivity by analyzing the polymer batching dosage prediction model.

[0049] A2. The initial dosage information of each ingredient obtained from the polymer dosage prediction model is transmitted to the material control module and the stirring module. The material control module controls the feeding system to transmit each ingredient, and the stirring module controls the stirrer to start working.

[0050] A3. When the mixture in the mixing container comes into contact with the sensing end of the resistivity detection device, the resistivity detection device starts to work, monitors the resistivity of the mixture in the mixing container in real time, and transmits the real-time monitored resistivity data to the feedback adjustment module.

[0051] A4. The feedback adjustment module converts the received real-time monitored resistivity data into a resistivity distribution image and transmits it to the polymer ingredient dosage prediction model.

[0052] A5. Determine whether the received resistivity distribution image meets the target requirements using the polymer ingredient dosage prediction model. If the target requirements are met, proceed to step A7; otherwise, proceed to step A6.

[0053] A6. The polymer batching dosage prediction model recalculates new batching dosage information based on the mapping relationship between the engineering mechanical properties of geopolymers and the resistivity of geopolymers. The new batching dosage information is then transmitted to the material control module, which controls the feeding system to transmit each batching component and repeats steps A3 to A5.

[0054] A7. The polymer batching dosage prediction model transmits the signal that the required dosage has been achieved to the material control module and the stirring module, and controls the feeding system and stirring system to stop working, thus obtaining the final geopolymer.

[0055] Preferably, the target requirement mentioned in step A5 refers to comparing the latest measured resistivity data transmitted by the feedback adjustment module with the target geological polymer resistivity in the polymer ingredient dosage prediction model. If the conformity between the measured resistivity data and the target geological polymer resistivity is greater than 95%, it is determined that the target requirement has been met; otherwise, it is determined that the target requirement has not been met.

[0056] The present invention has the following beneficial effects: The present invention establishes a mapping relationship between the engineering mechanical properties of geopolymers and the dosage of various ingredients and the resistivity of geopolymers through neural networks. It can perform real-time monitoring and feedback during the preparation process, and can control the production process in real time. It can more accurately control the dosage of ingredients to ensure the mechanical properties and stability of the obtained geopolymers, so as to ensure the product quality is stable and the geopolymers that achieve the performance target values ​​can be obtained. Attached Figure Description

[0057] To make the objectives, technical solutions, and advantages of the invention clearer, the invention will now be described in further detail with reference to the accompanying drawings, wherein:

[0058] Figure 1 This is a schematic diagram of the steps in the geopolymer batching method based on neural networks of the present invention;

[0059] Figure 2 This is a schematic diagram of the polymer ingredient dosage prediction model module of the present invention;

[0060] Figure 3 This is a schematic diagram of the workflow of the feedback adjustment module of the present invention;

[0061] Figure 4 This is a schematic diagram of the feeding system and stirring system of the present invention;

[0062] Figure 5 This is an example of a resistivity change image according to the present invention.

[0063] Figure labeling: 101, Polymer batching dosage prediction model module; 102, Material control module; 103, Stirring module; 104, Feedback adjustment module; 2, Feeding system; 201, Material storage chamber; 202, Material conveying channel; 3, Stirring system; 301, Mixing container; 302, Bottom surface; 304, Stirring paddle; 305, Rotating shaft; 303, Electrode; 306, DC resistivity meter; 4, Resistivity detection device. Detailed Implementation

[0064] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings.

[0065] It should be noted that similar reference numerals and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the figures, or the orientation or positional relationship commonly used when the product is in use. They are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. Furthermore, the terms "first," "second," and "third," etc., are only used to distinguish descriptions and should not be construed as indicating or implying relative importance. In addition, the terms "horizontal," "vertical," etc., do not indicate that the component is required to be absolutely horizontal or suspended, but can be slightly tilted. For example, "horizontal" simply means that its direction is more horizontal than "vertical," and does not mean that the structure must be completely horizontal, but can be slightly tilted. In the description of this invention, it should also be noted that, unless otherwise explicitly specified and limited, the terms "set," "install," "connect," and "link" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0066] This invention can be applied to the formulation of geopolymers, solving the technical problem that existing batching systems lack real-time monitoring and feedback during batching control, which can easily lead to discrepancies between the actual production product and the theoretical value, resulting in the product performance failing to meet the target value and making it difficult to ensure the quality stability of the geopolymer. The geopolymer batching method, system, and control method based on neural networks proposed in this invention can control the dosage of various ingredients by inputting the engineering mechanical properties of the target geopolymer, ultimately obtaining a geopolymer that meets the target value.

[0067] The technical advantages of the present invention regarding the neural network-based geopolymer batching method, system, and control method are as follows:

[0068] 1. This invention establishes a mapping relationship between the engineering mechanical properties of geopolymers and the dosage of various ingredients and the resistivity of geopolymers through neural networks. It can perform real-time monitoring and feedback during the preparation process, control the production process in real time, and control the dosage of ingredients more accurately to ensure the mechanical properties and stability of the obtained geopolymers, so as to ensure the product quality is stable and the geopolymers that achieve the performance target values.

[0069] 2. This invention utilizes resistivity distribution change images to visualize the uniformity of mixing of ingredients and the strength of the product, preventing under- or over-mixing and ensuring the quality of the geopolymer.

[0070] 3. This invention uses a feedback adjustment module to feed back the resistivity information of the batching mixture during the stirring process to the polymer batching dosage prediction model in real time. Based on the learned mapping relationship between the engineering mechanical properties of geopolymers and the dosage of various batchings and the resistivity of geopolymers, the polymer batching dosage prediction model can accurately supplement the missing batchings in real time, thereby minimizing resource waste and improving economic efficiency while ensuring product quality.

[0071] In a first aspect, based on the technical problems solved above, this invention discloses a method for preparing geopolymers based on neural networks, comprising the following steps:

[0072] S1. Obtain training sample instances of geopolymers;

[0073] S2. The polymer ingredient dosage prediction model is trained based on the training sample examples of geopolymers to obtain appropriate neural network weights and establish the trained polymer ingredient dosage prediction model. The trained polymer ingredient dosage prediction model is used to indicate the mapping relationship between the engineering mechanical properties of geopolymers and the dosage of each ingredient and the resistivity of geopolymers.

[0074] S3. Input the preset engineering mechanical properties of the geopolymer, and use the trained polymer batching dosage prediction model obtained in step S2 to analyze and obtain the initial dosage information of each batching and the resistivity of the target geopolymer.

[0075] S4. Based on the initial dosage information of each ingredient obtained in step S3, mix and stir each ingredient to obtain a mixture of ingredients;

[0076] S5. The resistivity distribution image of the ingredient mixture is measured in real time using the resistivity method, and the real-time measured resistivity distribution image of the ingredient mixture is fed back to the trained polymer ingredient dosage prediction model.

[0077] S6. The polymer ingredient dosage prediction model determines whether the received resistivity distribution image meets the target requirements. If the target requirements are met, proceed to step S8; otherwise, proceed to step S7.

[0078] S7. The polymer ingredient dosage prediction model recalculates new ingredient dosage information based on the mapping relationship between the engineering mechanical properties of geopolymers and the resistivity of geopolymers; and adjusts the dosage of each ingredient according to the new ingredient dosage information, continues mixing and stirring, and repeats steps S5 to S6.

[0079] S8. Stop feeding and stirring, and obtain the final geopolymer.

[0080] This solution addresses the problem in existing batching systems that lack real-time monitoring and feedback during batching control. This lack of monitoring leads to discrepancies between actual and theoretical values ​​in the final product, resulting in unmet performance targets and inconsistent geopolymer quality. The neural network-based geopolymer batching method in this solution provides real-time monitoring and feedback during the batching process, enabling real-time control of the production process. This ensures the mechanical properties and stability of the resulting geopolymer, guaranteeing stable product quality and achieving target performance values.

[0081] Specifically, in step S1, the training sample examples of the geopolymer include examples of the relationship between the engineering mechanical properties of the geopolymer and the dosage of each ingredient and the resistivity of the geopolymer.

[0082] Specifically, in step S1, examples of the relationship between the engineering mechanical properties of the geopolymer and the dosage of each ingredient and the resistivity of the geopolymer can be obtained from historical data, laboratory data, and engineering project design data.

[0083] Preferably, the dosage of the ingredients may include one or more of the following: alkali activator dosage, slag powder dosage, fly ash dosage, water dosage, admixture dosage, and sand dosage; the engineering mechanical properties of the geopolymer may include one or more of the following: unconfined compressive strength, tensile strength, and shear strength.

[0084] For a specific example of the embodiments of the present invention, please refer to Figure 2 In step S1, the mapping relationship between the engineering mechanical properties of the geopolymer and the dosage of various ingredients can be expressed as:

[0085] Y1=a1X1+a2X2+a3X3+a4X4+a5X5+a6X6 (1)

[0086] Y2=b1X1+b2X2+b3X3+b4X4+b5X5+b6X6 (2)

[0087] Y3=c1X1+c2X2+c3X3+c4X4+c5X5+c6X6 (3)

[0088] In equations (1) to (3), Y1 is the unconfined compressive strength; Y2 is the tensile strength; Y3 is the shear strength; X1, a1, b1, and c1 are the dosage of alkali activator and the corresponding unconfined compressive strength coefficient, tensile strength coefficient, and shear strength coefficient, respectively; X2, a2, b2, and c2 are the dosage of slag powder and the corresponding unconfined compressive strength coefficient, tensile strength coefficient, and shear strength coefficient, respectively; X3, a3, b3, and c3 are the dosage of fly ash and the corresponding unconfined compressive strength coefficient, tensile strength coefficient, and shear strength coefficient, respectively. The unconfined compressive strength coefficient, tensile strength coefficient, and shear strength coefficient are defined as follows: X4, a4, b4, and c4 represent the water dosage and its corresponding unconfined compressive strength coefficient, tensile strength coefficient, and shear strength coefficient; X5, a5, b5, and c5 represent the admixture dosage and its corresponding unconfined compressive strength coefficient, tensile strength coefficient, and shear strength coefficient; X6, a6, b6, and c6 represent the sand dosage and its corresponding unconfined compressive strength coefficient, tensile strength coefficient, and shear strength coefficient. The unconfined compressive strength coefficient, tensile strength coefficient, and shear strength coefficient corresponding to each ingredient dosage need to be predicted and determined using a trained polymer ingredient dosage prediction model.

[0089] Correspondingly, in step S1, the mapping relationship between the engineering mechanical properties of geopolymers and resistivity can be expressed as:

[0090]

[0091] In equation (4), R is the resistivity of the geopolymer; α is the correlation coefficient of tensile strength; β is the correlation coefficient of tensile strength; γ is the correlation coefficient of tensile strength.

[0092] Correspondingly, in step S1, the learning type used to train the polymer batching dosage prediction model using geological polymer training sample instances is batch learning. As the number of training iterations increases, the network error gradually decreases. The formula for the network error reduction process is as follows:

[0093] e(n+1)=e(n)+Δe (5)

[0094] In equation (5), e is the network error and n is the number of training iterations.

[0095] Preferably, in step S1, the polymer ingredient dosage prediction model is one of the following: CNN convolutional neural network, FNN feedforward neural network, BP backpropagation neural network, RNN recurrent neural network, or LSTM long short-term memory network. For the task of extracting information about the mapping relationship between different feature-related index parameters, neural network models such as CNN convolutional neural network, FNN feedforward neural network, BP backpropagation neural network, RNN recurrent neural network, and LSTM long short-term memory network have good information extraction and recognition capabilities, making them suitable for the extraction of the mapping relationship between the engineering mechanical properties of geopolymers and the dosage of various ingredients and the resistivity of geopolymers in this invention. On the other hand, preferably, the loss function used to train the polymer ingredient dosage prediction model using geopolymer training sample instances can be one of the following: CELoss cross-entropy loss function, MSE mean square error loss function, or LLLLoss log-likelihood loss function. These loss functions, used in conjunction with the aforementioned neural network model, can effectively balance improving the convergence speed of model training and the accuracy of training recognition, resulting in good training application effects.

[0096] Specifically, High-Density Resistivity Inversion (HDRI) is a geophysical exploration technique primarily used to measure and analyze the resistivity distribution of subsurface media. It involves arranging an electrode array on the surface or in a body of water, applying current, and measuring voltage to obtain information about the subsurface resistivity distribution. Its detection principle is based on the fact that when current flows through a medium, it is affected by the medium's resistivity. Through an inversion algorithm, the resistivity value of the medium can be calculated from the measured voltage data.

[0097] Preferably, in step S5, the resistivity distribution image of the ingredient mixture is measured using the resistivity method, including the following steps:

[0098] S5.1 A series of electrodes are arranged in the ingredient mixture to form an electrode array for applying current to the ingredient mixture and measuring the voltage in the ingredient mixture;

[0099] S5.2 Apply current to the mixture of ingredients through an electrode array;

[0100] S5.3 Measure the voltage between different electrode pairs;

[0101] S5.4 By changing the combination of electrode pairs, different voltage data are collected to form a voltage data matrix;

[0102] S5.5. Using a mathematical inversion algorithm, the resistivity distribution in the mixture is calculated from the obtained voltage data matrix.

[0103] S5.6. Display the resistivity data obtained by inversion in the form of an image to form a resistivity distribution image of the ingredient mixture, and feed back the real-time measured resistivity distribution image of the ingredient mixture to the trained polymer ingredient dosage prediction model.

[0104] Preferably, before performing step S5.5, the collected voltage data needs to be preprocessed, such as through noise reduction and correction, to improve data quality.

[0105] Specifically, in step S5.5, the resistivity method describes the distribution of conduction current in geopolymers under the influence of an artificially applied stable current field. Its principle can be expressed as follows:

[0106]

[0107] In equation (6), R is resistivity; U is electrode voltage; and I is current.

[0108] Specifically, in step S5.6, the resistivity distribution image obtained by the resistivity method displays parameters including depth, resistivity magnitude, and uniformity of distribution.

[0109] For example, please refer to Figure 5 The image shows resistivity measurements at different depths. The vertical axis represents the depth value, and the horizontal axis represents the resistivity value. This resistivity distribution image was obtained through six iterations during the resistivity inversion calculation, resulting in a root mean square error (RMS) of 1.26%. RMS error is an important indicator of the accuracy of the inversion results; a low RMS error (1.26%) indicates that the inversion calculation results are relatively accurate and reliable. The image also shows the resistivity distribution at different depths, with resistivity values ​​varying with depth.

[0110] Specifically, in step S5, the resistivity method adopts the high-density resistivity method, which uses multiple electrodes with a distance between each electrode ranging from 1 cm to 10 cm. Reducing the electrode spacing can improve the measurement resolution, making the resistivity distribution map more refined and able to reflect the subtle changes in the medium in more detail.

[0111] The method for preparing geopolymers based on neural networks disclosed in this invention has the following technical effects: The method for preparing geopolymers based on neural networks of this invention can perform real-time monitoring and feedback during the preparation process, and can control the production process in real time to ensure the mechanical properties and stability of the obtained geopolymers, so as to ensure the stable quality of the product and obtain geopolymers that meet the performance target values.

[0112] Secondly, please refer to Figure 4The present invention also discloses a geopolymer batching system based on a neural network for implementing the geopolymer batching method based on a neural network as described above, including a central processing unit, a feeding system 2, a stirring system 3, and a resistivity detection device 4.

[0113] The feeding system 2 includes a material storage chamber 201, which is used to store the raw materials of each ingredient. The material storage chamber 201 has an independent storage space corresponding to the raw materials of each ingredient. Each storage space is provided with a material outlet. Each material outlet is provided with an electrically controlled valve. Each electrically controlled valve is electrically connected to the central processing unit. The central processing unit can control the opening and closing and the opening degree of each electrically controlled valve.

[0114] The mixing system 3 includes a mixing container 301, and a stirrer is provided in the mixing container 301 for stirring the various ingredients input into the mixing container 301 from the feeding system 2. The power supply terminal of the stirrer is electrically connected to the central processing unit, and the central processing unit can control the switch of the stirrer.

[0115] The stirring system 3 is equipped with a sensing end of a resistivity detection device 4, which is electrically connected to the central processing unit. The resistivity detection device 4 is used to monitor the real-time resistivity of the mixture in the mixing container 301, and the data transmission end of the resistivity detection device 4 is used to feed back the real-time resistivity information of the mixture to the central processing unit.

[0116] Specifically, in order to reduce the number of pipe interfaces connected to the mixing container 301, a material conveying channel 202 is set up, and each material outlet is connected to the material conveying channel 202. The material conveying channel 202 is connected to the mixing container 301. The material conveying channel 202 pre-accommodates the raw materials of each ingredient and can also play a pre-mixing role.

[0117] Specifically, the mixing container 301 has a barrel-shaped structure, and the agitator is installed on the bottom surface 302 inside the mixing container 301. The agitator includes a motor, and a rotating shaft 305 is fixed to the output end of the motor. A stirring paddle 304 is fixed on the rotating shaft 305. The rotating shaft 305 can rotate with the rotation of the motor output end, thereby driving the stirring paddle 304 to rotate and stir the ingredients. The motor is electrically connected to the central processing unit, and the central processing unit can control the start and stop of the motor.

[0118] Specifically, the sensing end of the resistivity detection device 4 is an electrode 303, which is detachably installed on the top of the mixing container 301. One end of the electrode 303 extends into the mixing mixture and can contact the mixing mixture. The other end of the electrode 303 is electrically connected to a DC resistivity meter 306 for real-time monitoring of the resistivity of the mixing mixture in the mixing container 301.

[0119] For preference, please refer to Figure 1 The central processing unit includes a polymer ingredient dosage prediction model module 101, a material control module 102, a stirring module 103, and a feedback adjustment module 104.

[0120] The polymer batching dosage prediction model module 101 is used to receive the input geopolymer engineering mechanical properties information and the real-time measured resistivity distribution image information of the batching mixture fed back by the feedback adjustment module 104, and convert the geopolymer engineering mechanical properties information or the real-time measured resistivity distribution image information of the batching mixture into batching information, and transmit the batching information to the material control module 102.

[0121] The material control module 102 is used to receive the batching information transmitted by the polymer batching dosage prediction model module 101, and control the feeding system 2 to transport each batching material into the mixing system 3 according to the batching information.

[0122] The stirring module 103 is used to control the opening and closing of the stirrer to mix and stir the various ingredients to obtain a mixture of ingredients;

[0123] The feedback adjustment module 104 is used to receive the resistivity data of the ingredient mixture measured in real time by the resistivity detection device 4, convert the resistivity data of the ingredient mixture into a resistivity distribution image, and feed the resistivity distribution image back to the polymer ingredient dosage prediction model module 101.

[0124] This solution's neural network-based geopolymer batching system can predict the corresponding dosage of each ingredient based on the input geopolymer engineering mechanical properties, using a trained polymer batching dosage prediction model. Then, a material control module and a mixing module control the feeding and mixing systems respectively to transport and mix the materials. Simultaneously, a resistivity detection device monitors the resistivity of the batching mixture in the mixing module in real time and sends the measured resistivity to a feedback adjustment module, converting it into a resistivity distribution image. This feedback adjustment module then feeds the converted resistivity distribution image back to the polymer batching dosage prediction model. Based on the mapping relationship between the geopolymer engineering mechanical properties and the geopolymer resistivity, new ingredient dosage information is recalculated. Finally, the material control and mixing modules control the feeding and mixing systems respectively to adjust the dosage of the ingredients in the batching mixture in real time. Once the required dosage is achieved, the mixing system is stopped.

[0125] For details, please refer to Figure 3 The feedback module receives resistivity data of the mixture in the mixing container monitored in real time by the resistivity detection device, converts the original high-density data file into RES2DINV format, performs inversion using least squares, outputs the inverted resistivity distribution image, and feeds the resistivity distribution image back to the polymer ingredient dosage prediction model module.

[0126] The geopolymer batching system based on neural networks disclosed in this invention has the following technical effects: it can realize the above-mentioned geopolymer batching method based on neural networks, can perform real-time monitoring and feedback during the batching process, and can control the production process in real time to ensure the mechanical properties and stability of the obtained geopolymer, so as to ensure the product quality is stable and the geopolymer that meets the performance target value can be obtained.

[0127] Thirdly, please refer to Figure 1 The present invention also discloses a control method for a geopolymer batching system based on a neural network, used to control the aforementioned geopolymer batching system based on a neural network, comprising the following steps:

[0128] A1. Input the preset geopolymer engineering mechanical properties into the polymer batching dosage prediction model, and obtain the initial dosage information of each ingredient and the target geopolymer resistivity by analyzing the polymer batching dosage prediction model.

[0129] A2. The initial dosage information of each ingredient obtained from the polymer dosage prediction model analysis is transmitted to the material control module 102 and the stirring module 103. The material control module 102 controls the feeding system 2 to transmit each ingredient, and the stirring module 103 controls the stirrer to start working.

[0130] A3. Make the mixture in the mixing container 301 touch the sensing end of the resistivity detection device 4, and the resistivity detection device 4 starts to work, monitors the resistivity of the mixture in the mixing container 301 in real time, and transmits the real-time monitored resistivity data to the feedback adjustment module 104.

[0131] A4. The feedback adjustment module 104 converts the received real-time monitored resistivity data into a resistivity distribution image and transmits it to the polymer ingredient dosage prediction model.

[0132] A5. Determine whether the received resistivity distribution image meets the target requirements using the polymer ingredient dosage prediction model. If the target requirements are met, proceed to step A7; otherwise, proceed to step A6.

[0133] A6. The polymer batching dosage prediction model recalculates new batching dosage information based on the mapping relationship between the engineering mechanical properties of geopolymers and the resistivity of geopolymers; and transmits the new batching dosage information to the material control module 102, which controls the feeding system 2 to transmit each batching, and repeats steps A3 to A5.

[0134] A7. The signal that the polymer batching dosage prediction model has reached the required level is transmitted to the material control module 102 and the stirring module 103, and the feeding system 2 and the stirring system 3 are stopped to obtain the final geopolymer.

[0135] Preferably, the target requirement mentioned in step A5 refers to comparing the latest measured resistivity data transmitted by the feedback adjustment module with the target geological polymer resistivity in the polymer ingredient dosage prediction model. If the conformity between the measured resistivity data and the target geological polymer resistivity is greater than 95%, it is determined that the target requirement has been met; otherwise, it is determined that the target requirement has not been met.

[0136] The control method of the neural network-based geopolymer batching system disclosed in this scheme has the following technical effects: it can realize the neural network-based geopolymer batching method by controlling the above-mentioned neural network-based geopolymer batching system. The whole control process has simple logic, is easy to operate, has strong operability, and has high operation accuracy. It can ensure the mechanical properties and stability of the geopolymer and obtain the geopolymer that achieves the performance target value.

[0137] It is understood that the present invention has been described through some embodiments, and those skilled in the art will recognize that various changes or equivalent substitutions can be made to these features and embodiments without departing from the spirit and scope of the invention. Under the teachings of the present invention, modifications can be made to these features and embodiments to adapt to specific situations and materials without departing from the spirit and scope of the invention. The embodiments described in this invention are only a part of the embodiments of the invention, not all of them. The components of the embodiments of the invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. Therefore, the invention is not limited to the specific embodiments disclosed herein, and all other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the present invention.

Claims

1. A method for batching geopolymers based on neural networks, characterized in that, Includes the following steps: S1. Obtain training sample instances of geopolymers; S2. The polymer ingredient dosage prediction model is trained based on the training sample examples of geopolymers to obtain appropriate neural network weights and establish the trained polymer ingredient dosage prediction model. The trained polymer ingredient dosage prediction model is used to indicate the mapping relationship between the engineering mechanical properties of geopolymers and the dosage of each ingredient and the resistivity of geopolymers. In the polymer ingredient dosage prediction model, the mapping relationship between the engineering mechanical properties of the geopolymer and the dosage of each ingredient and the resistivity of the geopolymer is expressed as follows: ； ； ； in, It is the unconfined compressive strength; Tensile strength; Shear strength; , , , These are the dosage of alkali activator and the corresponding unconfined compressive strength coefficient, tensile strength coefficient, and shear strength coefficient of the alkali activator; , , , These are the slag powder content and the corresponding unconfined compressive strength coefficient, tensile strength coefficient, and shear strength coefficient of the slag powder; , , , These are the fly ash content and the corresponding unconfined compressive strength coefficient, tensile strength coefficient, and shear strength coefficient of fly ash; , , , These are the water dosage and the corresponding unconfined compressive strength coefficient, tensile strength coefficient, and shear strength coefficient for water; , , , These are the admixture dosage and the corresponding unconfined compressive strength coefficient, tensile strength coefficient, and shear strength coefficient of the admixture; , , , These are the sand content and the corresponding unconfined compressive strength coefficient, tensile strength coefficient, and shear strength coefficient of the sand; S3. Input the preset engineering mechanical properties of the geopolymer, and use the trained polymer batching dosage prediction model obtained in step S2 to analyze and obtain the initial dosage information of each batching and the resistivity of the target geopolymer. S4. Based on the initial dosage information of each ingredient obtained in step S3, mix and stir each ingredient to obtain a mixture of ingredients; S5. Measure the resistivity distribution image of the ingredient mixture in real time and feed the real-time measured resistivity distribution image of the ingredient mixture back to the trained polymer ingredient dosage prediction model. The process of measuring the resistivity distribution image of the mixture using the resistivity method includes the following steps: S5.1 A series of electrodes are arranged in the ingredient mixture to form an electrode array for applying current to the ingredient mixture and measuring the voltage in the ingredient mixture; S5.2 Apply current to the mixture of ingredients through an electrode array; S5.3 Measure the voltage between different electrode pairs; S5.4 By changing the combination of electrode pairs, different voltage data are collected to form a voltage data matrix; S5.

5. Using a mathematical inversion algorithm, the resistivity distribution in the mixture is calculated from the obtained voltage data matrix. S5.

6. Display the resistivity data obtained by inversion in the form of an image to form a resistivity distribution image of the ingredient mixture, and feed back the real-time measured resistivity distribution image of the ingredient mixture to the trained polymer ingredient dosage prediction model. S6. The polymer ingredient dosage prediction model determines whether the received resistivity distribution image meets the target requirements. If the target requirements are met, proceed to step S8; otherwise, proceed to step S7. S7. The polymer ingredient dosage prediction model recalculates new ingredient dosage information based on the mapping relationship between the engineering mechanical properties of geopolymers and the resistivity of geopolymers; and adjusts the dosage of each ingredient according to the new ingredient dosage information, continues mixing and stirring, and repeats steps S5 to S6. S8. Stop feeding and stirring, and obtain the final geopolymer.

2. The method for preparing geopolymers based on neural networks according to claim 1, characterized in that, In step S1, the polymer ingredient dosage prediction model is one of the following: convolutional neural network, feedforward neural network, backpropagation neural network, recurrent neural network, and long short-term memory network; the loss function used to train the polymer ingredient dosage prediction model with geological polymer training sample instances is one of the following: cross-entropy loss function, mean squared error loss function, and log-likelihood loss function.

3. The method for preparing geopolymers based on neural networks according to claim 1, characterized in that, Before performing step S5.5, the collected voltage data needs to be preprocessed, including noise reduction and correction, to improve data quality.

4. A neural network-based geopolymer batching system for implementing the neural network-based geopolymer batching method as described in claim 1, characterized in that, Includes a central processing unit, a feeding system, a mixing system, and a resistivity detection device; The feeding system includes a material storage chamber for storing the raw materials of each ingredient. Each raw material in the material storage chamber has an independent storage space, and each storage space has a material outlet. Each material outlet is equipped with an electrically controlled valve, and each electrically controlled valve is electrically connected to a central processing unit. The central processing unit can control the opening and closing of each electrically controlled valve and the degree of opening. The mixing system includes a mixing container, and a stirrer is installed inside the mixing container for stirring the various ingredients input into the mixing container from the feeding system. The power supply terminal of the stirrer is electrically connected to the central processing unit, and the central processing unit can control the switch of the stirrer. The stirring system is equipped with a sensing end of a resistivity detection device, which is electrically connected to the central processing unit. The resistivity detection device is used to monitor the real-time resistivity of the mixture in the mixing container, and the data transmission end of the resistivity detection device is used to feed back the real-time resistivity information of the mixture to the central processing unit. The central processing unit includes a polymer ingredient dosage prediction model module, a material control module, a stirring module, and a feedback adjustment module. The polymer batching dosage prediction model module receives input geopolymer engineering mechanical property information and real-time measured resistivity distribution image information of the batching mixture fed back by the feedback adjustment module. It then converts the geopolymer engineering mechanical property information or the real-time measured resistivity distribution image information of the batching mixture into batching information and transmits the batching information to the material control module. In the polymer batching dosage prediction model, the mapping relationship between the geopolymer engineering mechanical properties and the dosage of each batching ingredient and the geopolymer resistivity is expressed as follows: ； ； ； in, It is the unconfined compressive strength; Tensile strength; Shear strength; , , , These are the dosage of alkali activator and the corresponding unconfined compressive strength coefficient, tensile strength coefficient, and shear strength coefficient of the alkali activator; , , , These are the slag powder content and the corresponding unconfined compressive strength coefficient, tensile strength coefficient, and shear strength coefficient of the slag powder; , , , These are the fly ash content and the corresponding unconfined compressive strength coefficient, tensile strength coefficient, and shear strength coefficient of fly ash; , , , These are the water dosage and the corresponding unconfined compressive strength coefficient, tensile strength coefficient, and shear strength coefficient for water; , , , These are the admixture dosage and the corresponding unconfined compressive strength coefficient, tensile strength coefficient, and shear strength coefficient of the admixture; , , , These are the sand content and the corresponding unconfined compressive strength coefficient, tensile strength coefficient, and shear strength coefficient of the sand; The material control module is used to receive the batching information transmitted by the polymer batching dosage prediction model module, and control the feeding system to deliver each batching ingredient into the mixing system according to the batching information. The stirring module is used to control the opening and closing of the stirrer to mix and stir the various ingredients to obtain a mixture of ingredients; The feedback adjustment module is used to receive the resistivity data of the ingredient mixture measured in real time by the resistivity detection device, convert the resistivity data of the ingredient mixture into a resistivity distribution image, and feed the resistivity distribution image back to the polymer ingredient dosage prediction model module; wherein, measuring the resistivity distribution image of the ingredient mixture using the resistivity method includes the following steps: S5.1 A series of electrodes are arranged in the ingredient mixture to form an electrode array for applying current to the ingredient mixture and measuring the voltage in the ingredient mixture; S5.2 Apply current to the mixture of ingredients through an electrode array; S5.3 Measure the voltage between different electrode pairs; S5.4 By changing the combination of electrode pairs, different voltage data are collected to form a voltage data matrix; S5.

5. Using a mathematical inversion algorithm, the resistivity distribution in the mixture is calculated from the obtained voltage data matrix. S5.

6. Display the resistivity data obtained by inversion in the form of an image to form a resistivity distribution image of the ingredient mixture, and feed back the real-time measured resistivity distribution image of the ingredient mixture to the trained polymer ingredient dosage prediction model.

5. The neural network-based geopolymer batching system according to claim 4, characterized in that, After receiving the resistivity data of the mixture in the mixing container monitored in real time by the resistivity detection device, the feedback adjustment module converts the original high-density data file into RES2DINV format, then performs inversion using least squares, outputs the inverted resistivity distribution image, and feeds the resistivity distribution image back to the polymer ingredient dosage prediction model module.

6. A control method for a geopolymer batching system based on a neural network, used to control the geopolymer batching system based on a neural network as described in claim 4, characterized in that, Includes the following steps: A1. Input the preset geopolymer engineering mechanical properties into the polymer batching dosage prediction model, and obtain the initial dosage information of each ingredient and the target geopolymer resistivity by analyzing the polymer batching dosage prediction model. A2. The initial dosage information of each ingredient obtained from the polymer dosage prediction model is transmitted to the material control module and the stirring module. The material control module controls the feeding system to transmit each ingredient, and the stirring module controls the stirrer to start working. A3. When the mixture in the mixing container comes into contact with the sensing end of the resistivity detection device, the resistivity detection device starts to work, monitors the resistivity of the mixture in the mixing container in real time, and transmits the real-time monitored resistivity data to the feedback adjustment module. A4. The feedback adjustment module converts the received real-time monitored resistivity data into a resistivity distribution image and transmits it to the polymer ingredient dosage prediction model. A5. Determine whether the received resistivity distribution image meets the target requirements using the polymer ingredient dosage prediction model. If the target requirements are met, proceed to step A7; otherwise, proceed to step A6. A6. The polymer batching dosage prediction model recalculates new batching dosage information based on the mapping relationship between the engineering mechanical properties of geopolymers and the resistivity of geopolymers. The new batching dosage information is then transmitted to the material control module, which controls the feeding system to transmit each batching component and repeats steps A3 to A5. A7. The polymer batching dosage prediction model transmits the signal that the required dosage has been achieved to the material control module and the stirring module, and controls the feeding system and stirring system to stop working, thus obtaining the final geopolymer.

7. The control method for a geopolymer batching system based on a neural network according to claim 6, characterized in that, The achievement of the target requirement mentioned in step A5 refers to comparing the latest measured resistivity data transmitted by the feedback adjustment module with the target geological polymer resistivity in the polymer batching dosage prediction model. If the conformity between the measured resistivity data and the target geological polymer resistivity is greater than 95%, it is determined that the target requirement has been achieved; otherwise, it is determined that the target requirement has not been achieved.

Images