Method and system for predicting performance of solid waste-based super sulfate cement

Abstract

The invention provides a method and system for predicting the performance of solid waste-based supersulfate cement, and the method comprises the steps: sampling a solid waste-based supersulfate cement slurry sample, carrying out a hydration reaction under a sealed condition, and detecting and quantifying the hydration degree data of the solid waste-based supersulfate cement slurry sample in real time by using a low-field nuclear magnetic resonance method; constructing a hydration dynamic neural network model, and training a hydration dynamic neural network as a cement hydration dynamic model by taking the proportion data and the hydration time of the solid waste-based supersulfate cement slurry sample as input and taking hydration degree data corresponding to the hydration time as output; and utilizing the trained cement hydration kinetic model to test the hydration kinetic performance of the solid waste-based super sulfate cement paste to be tested. The method is used for predicting the performance of the solid waste-based super sulfate cement, and the method is convenient in experiment, high in universality, rapid in data processing and high in credibility.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of building materials, and relates to a solid waste-based supersulfated cement performance prediction method and system driven by low-field nuclear magnetic resonance data. BACKGROUND

[0002] Cement is one of the largest building materials in the civil construction industry. Traditional Portland cement production needs to go through the process of "two grinding and one burning", which results in about 0.8 tons of CO2 emission per ton of cement produced. Therefore, the production of traditional Portland cement is one of the main sources of global carbon emissions. In contrast, supersulfated cement mainly uses industrial solid waste as raw material and does not need calcination. It can be mixed and stirred with industrial solid waste. The development and application of supersulfated cement not only realizes "waste treatment with waste", reduces the consumption of natural resources, but also promotes the transformation of the cement industry to a circular economy. Supersulfated cement is expected to become a core option for green building materials and help achieve global carbon reduction targets.

[0003] Supersulfated cement, as a low-carbon cementitious material, mainly uses industrial solid waste as raw material (accounting for 75%~85%), and realizes hydration through the synergistic effect of sulfates and alkaline activators. It does not need high-temperature calcination, and the carbon emission is only 10% of that of traditional cement, and the energy consumption is reduced by 80%~90%. However, due to the complex composition of industrial solid waste, the performance of solid waste-based supersulfated cement is unstable, the early strength is low, and the carbonation resistance is poor, which limits the application of solid waste-based supersulfated cement. The cement hydration process is the core mechanism of the performance evolution of concrete materials. The accurate construction of the kinetics model is crucial for optimizing the concrete mix design, predicting the early strength development and durability. Therefore, the study of the hydration kinetics model of solid waste-based supersulfated cement is one of the main ways to solve the above problems.

[0004] The study of cement hydration kinetics model mainly focuses on two aspects: on the one hand, a suitable method for evaluating the hydration degree of cement needs to be found; on the other hand, an accurate model needs to be proposed to describe the mathematical relationship between the hydration degree of cement and time and other influencing factors.

[0005] The traditional methods for evaluating the hydration degree of cement at home and abroad include non-evaporable water (chemically combined water), hydration heat, and calcium hydroxide (CH) quantitative measurement. However, the existing technical methods are not suitable for evaluating the hydration degree of supersulfated cement, and the reasons are as follows:

[0006] For solid waste-based supersulfated cement, the amount of calcium hydroxide generated is very small, so the calcium hydroxide quantitative measurement method cannot be used.

[0007] For the hydration heat measurement method, the hydration products of supersulfated cement are very complex, including calcium vanadate, AFt, etc. The hydration heat of different hydration products is different, so the hydration heat measurement method cannot accurately describe the hydration degree.

[0008] The current mainstream non-evaporative water measurement method is thermal gravimetric analysis (TGA), which is complicated to operate, has large sample loss, and cannot be repeatedly measured on the same cement sample, so it cannot track the hydration degree in real time.

[0009] The existing cement hydration kinetics model for describing the mathematical relationship between cement hydration degree and time and other influencing factors has great limitations, such as NG-I-D model, Avrami model, Jander model, etc., which are all based on multiple processing assumptions. However, actual cement particles have multi-scale heterogeneity, which is difficult to dynamically respond to real-time influences of temperature, humidity and additives (such as water reducing agent and retarder), resulting in significant deviation between model prediction values and experimental data.

[0010] Therefore, how to provide a solid waste-based super-sulfate cement performance prediction method and system that ensures prediction accuracy while being simple and fast to operate is a problem that those skilled in the art need to solve. SUMMARY

[0011] Therefore, the present application provides a solid waste-based super-sulfate cement performance prediction method and system, which constructs a complex nonlinear dynamic model in cement hydration based on low-field nuclear magnetic data and hydration dynamics neural network, for predicting the performance of solid waste-based super-sulfate cement. The method is convenient to experiment, has strong universality, fast data processing and high reliability.

[0012] To achieve the above purpose, the present application adopts the following technical solutions:

[0013] The present application discloses a solid waste-based super-sulfate cement performance prediction method, comprising the following steps:

[0014] S1: sampling solid waste-based super-sulfate cement paste samples, and performing hydration reaction under sealed conditions, while using low-field nuclear magnetic resonance method to detect and quantify the hydration degree data of solid waste-based super-sulfate cement paste samples in real time;

[0015] S2: constructing a hydration dynamics neural network model, taking the mixing ratio data and hydration time of solid waste-based super-sulfate cement paste samples as input, and taking the hydration degree data corresponding to the hydration time as output to train the hydration dynamics neural network as a cement hydration dynamics model;

[0016] S3: using the trained cement hydration dynamics model to test the hydration dynamics performance of the solid waste-based super-sulfate cement paste to be tested.

[0017] Preferably, the step of hydrating the sampled solid waste-based supersulfated cement paste sample in a sealed condition comprises: hydrating in a closed container, wherein the closed space in the closed container is filled with the solid waste-based supersulfated cement paste sample.

[0018] Preferably, the step of detecting and quantifying the hydration degree data of the solid waste-based supersulfated cement paste sample in real time by using the low-field nuclear magnetic resonance method in S1 comprises:

[0019] detecting the T2 relaxation signal of the water in the supersulfated cement paste sample;

[0020] converting the T2 relaxation signal into a T2 relaxation time distribution curve by inverse Laplace transform;

[0021] calculating the reduction amount of the T2 relaxation time spectrum peak area of the supersulfated cement paste in a preset hydration time interval as the relative generation amount of chemically combined water;

[0022] quantifying the hydration degree data of the solid waste-based supersulfated cement paste sample according to the relative generation amount of the chemically combined water.

[0023] Preferably, the formula for quantifying the hydration degree data α(t) of the solid waste-based supersulfated cement paste sample is:

[0024]

[0025] In the formula, T2 peak total area data at the current time of the hydration test; T2 peak total area data at the start time of the hydration test; T2 peak total area data at the last time of the hydration test.

[0026] Preferably, the hydration kinetics neural network comprises an input layer, a hidden layer and an output layer; the activation function from the input layer to the hidden layer is a sigmod function, and the activation function from the hidden layer to the output layer is a Purelin function.

[0027] The application also provides a solid waste-based supersulfated cement performance prediction system according to the solid waste-based supersulfated cement performance prediction method.

[0028] The cement hydration kinetics model construction unit is configured to construct a hydration kinetics neural network model, train the hydration kinetics neural network as a cement hydration kinetics model by taking the mixing ratio data and hydration time of the solid waste-based supersulfated cement paste sample as input and taking the hydration degree data corresponding to the hydration time as output; wherein the hydration degree data is the hydration degree data of the sampled solid waste-based supersulfated cement paste sample in real time detected and quantified by using the low-field nuclear magnetic resonance method when the solid waste-based supersulfated cement paste sample is hydrated in a sealed condition.​

[0029] The solid waste-based supersulfated cement paste hydration kinetics performance test unit is used for testing the hydration kinetics performance of a solid waste-based supersulfated cement paste to be tested by using the cement hydration kinetics model trained.

[0030] Preferably, a sealed container is further included for performing the hydration reaction of the solid waste-based supersulfated cement paste sample, and a sealed space in the sealed container is filled with the solid waste-based supersulfated cement paste sample.

[0031] Compared with the prior art, the beneficial effects of the present application include:

[0032] The present application is based on low-field nuclear magnetic resonance technology and hydration kinetics neural network algorithm, and the hydration degree of different types of solid waste-based supersulfated cement is characterized by quantitatively analyzing the content of chemically combined water, and the hydration kinetics neural network is trained to fit the data to establish a hydration kinetics model suitable for different types of solid waste-based supersulfated cement. The method is simple and fast, the data is relatively accurate, and it is suitable for hydration performance analysis of various solid waste-based supersulfated cement, and the fitting error is less than almost all existing hydration kinetics mathematical models. BRIEF DESCRIPTION OF DRAWINGS

[0033] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments or the prior art description will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor on the basis of the provided drawings.

[0034] Figure 1 The flowchart of the solid waste-based supersulfated cement performance prediction method provided for the embodiments of the present application;

[0035] Figure 2 The schematic diagram of the sealed container provided for the embodiments of the present application;

[0036] Figure 3 The hydration kinetics neural network framework diagram provided for the embodiments of the present application. DETAILED DESCRIPTION

[0037] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0038] The first aspect of the embodiment of the present application provides a solid waste-based super-sulfate cement performance prediction method, as shown in the following steps. Figure 1

[0039] S1: sampling solid waste-based super-sulfate cement paste samples, carrying out hydration reaction under sealed conditions, and simultaneously using low-field nuclear magnetic resonance method to detect and quantify hydration degree data of the solid waste-based super-sulfate cement paste samples in real time;

[0040] S2: constructing a hydration dynamic neural network model, taking the mixing ratio data and hydration time of the solid waste-based super-sulfate cement paste samples as inputs, and taking the hydration degree data corresponding to the hydration time as outputs to train the hydration dynamic neural network as a cement hydration dynamics model;

[0041] S3: using the trained cement hydration dynamics model to test the hydration dynamics performance of the solid waste-based super-sulfate cement paste to be tested.

[0042] It should be noted that all hydration reactions of super-sulfate cement require water participation, and therefore the non-evaporable water measurement method is suitable for characterizing the hydration degree of super-sulfate cement, and the present application uses low-field nuclear magnetic resonance technology to measure the non-evaporable water amount. The low-field nuclear magnetic resonance technology can detect the T2 relaxation signal of the water in the super-sulfate cement paste, and the signal can be converted into a T2 relaxation time distribution curve graph through inverse Laplace transform, and the curve peak area represents the relative content of water in different states. With the progress of the hydration reaction, the free water in the super-sulfate cement paste gradually changes to physical and chemical combined water, and the T2 relaxation time distribution curve also changes with the progress of the hydration reaction, and so does the curve peak area. Due to the limitation of technical conditions, the ordinary low-field nuclear magnetic resonance instrument cannot detect the T2 relaxation signal of the chemical combined water, and therefore under sealed conditions, the reduction amount of the T2 relaxation time spectrum peak area of the super-sulfate cement paste is the relative generation amount of the chemical combined water. The low-field nuclear magnetic resonance technology is simple to operate, non-destructive, fast in detection speed, and suitable for hydration kinetics tracking.

[0043] In one embodiment, the step of sampling solid waste-based super-sulfate cement paste samples in S1 and carrying out hydration reaction under sealed conditions includes: carrying out hydration reaction in a closed container 2, and the closed space in the closed container 2 is filled with solid waste-based super-sulfate cement paste samples 1, and the closed space is sealed by a movable polytetrafluoroethylene piston 3.

[0044] In the present embodiment, glass containers are used to seal the cement paste, to isolate the cement from the exchange of moisture with the outside, and to ensure that the moisture loss amount monitored by the nuclear magnetic is equal to the generation amount of the chemical combined water. According to the research needs, solid waste-based super-sulfate cement pastes with different raw material mixing ratios and different water-cement ratios are configured, and are loaded into special glass containers as shown in Figure 2 . ​

[0045] The sealed container was placed in a low-field nuclear magnetic resonance analyzer for a 7-day test. A CPMG sequence was set in the instrument to automatically test the T2 relaxation signal every 15 minutes, and the T2 nuclear magnetic resonance spectrum was obtained using the inverse Laplace transform.

[0046] In one embodiment, the step of using a low-field nuclear magnetic resonance method to detect and quantify the hydration degree data of a solid waste-based supersulfate cement paste sample in real time in S1 includes:

[0047] Detection of T2 relaxation signal of water in supersulfate cement paste samples;

[0048] The T2 relaxation signal is converted into a T2 relaxation time distribution curve by inverse Laplace transform;

[0049] The decrease in the peak area of ​​the T2 relaxation time spectrum of the supersulfate cement paste within the preset hydration time interval is calculated as the relative generation amount of chemically bound water;

[0050] The hydration data of solid waste-based supersulfate cement paste samples were quantified based on the relative production of chemically bound water.

[0051] In one embodiment, the total area data of T2 peaks at different hydration times t are calculated based on the T2 nuclear magnetic resonance spectrum. , the formula for quantifying the hydration degree data α(t) of solid waste-based supersulfate cement paste samples is:

[0052] ;

[0053] Where, Indicates the total area data of T2 peak at the current moment of hydration test; Represents the total area data of the T2 peak at the start of the hydration test; Represents the total area data of the T2 peak at the last moment of the hydration test.

[0054] In one embodiment, the hydration dynamic neural network includes an input layer, a hidden layer, and an output layer. In this embodiment, a feedforward hydration dynamic neural network includes a linear output layer and at least one hidden layer including an activation function with a certain "squeezing" property (such as a sigmoid function). As long as the network has enough hidden units, it can approximate any continuous nonlinear function in a finite-dimensional space with arbitrary accuracy.

[0055] like Figure 3 As shown, the input layer is the input characteristic of solid waste-based supersulfate cement, is the weight of the i-th input feature and the j-th hidden layer neuron, For the hidden layer bias term (not embodied in the figure) of the jth neuron of the hidden layer m layers, respectively ; similarly is the weight of the jth input feature and the output layer , is the bias term (not embodied in the figure) of the jth neuron of the hidden layer , is the true value.

[0056] The input features of the input layer in the embodiment can include: water-cement ratio, silico-aluminate admixture content percentage, sulfate activator content percentage, alkaline activator content percentage, and hydration time. Among them, the water content (%) in the water-cement ratio, the sulfate activator content (%) in the sulfate activator content percentage, and the time (s) in the hydration time are not shown in the embodiment, but do not affect the expression of the parameters contained in the input features of the input layer. Figure 3 Figure 3 Figure 3 Figure 3

[0057] It should be noted that the sample data of the embodiment can be different types of solid waste-based supersulfated cement proportioning data and hydration water demand, wherein the silico-aluminate admixture, the sulfate activator and the alkaline activator can be various solid wastes, and if the solid waste-based supersulfated cement raw material is not a ternary system, the number of input features is adjusted accordingly according to the number of raw materials.

[0058] According to experience and literature, the number of neurons in the hidden layer is set to three times the number of input features, that is: The output layer is the hydration degree of cement, that is, the hydration degree.

[0059] In the embodiment, the activation function from the input layer to the hidden layer is the sigmod function, and the activation function from the hidden layer to the output layer is the Purelin function.

[0060] Therefore, taking the first neuron as an example, the mapping function relationship from the input layer to the hidden layer is:

[0061] ;

[0062] Among them, is the weight of the first input feature and the first hidden layer neuron, is the weight of the second input feature and the first hidden layer neuron, is the weight of the n-th input feature and the first hidden layer neuron. ​​​​are the 1st, 2nd and nth input layer neurons respectively; For the hidden layer The bias term of the first neuron in .

[0063] The mapping function relationship from the hidden layer to the output layer is:

[0064] ;

[0065] in, is the output layer; is the jth input feature and output layer The weight of is the jth neuron in the hidden layer; m is the number of neurons in the hidden layer The number of neurons; For the output layer The bias term.

[0066] Assuming there are k groups of data, the objective function J is set as:

[0067] ;

[0068] in, is the output layer feature obtained from the i-th group of data, is the true value corresponding to the i-th group of data.

[0069] The gradient descent method is used to optimize the parameters, and the back propagation gradient descent formula of the neural network is obtained after sorting (taking the kth group of data as an example):

[0070] ;

[0071] ;

[0072] ;

[0073] ;

[0074] in, is the objective function corresponding to the kth group of data; is the input feature of the kth group of data input to the i-th neuron in the input layer.

[0075] The second aspect of the embodiments of the present invention further discloses a code change impact intelligent analysis system for a solid waste-based supersulfate cement performance prediction method according to the first aspect of the embodiments, comprising:

[0076] The cement hydration kinetics model construction unit is configured to construct a hydration dynamics neural network model, take the mixing ratio data and hydration time of the solid waste-based supersulfated cement paste sample as input, take the hydration degree data corresponding to the hydration time as output, train the hydration dynamics neural network as the cement hydration kinetics model, and the hydration degree data is hydration degree data of the solid waste-based supersulfated cement paste sample in a sealed condition during a hydration reaction, which is detected and quantified in real time by using a low-field nuclear magnetic resonance method.

[0077] The solid waste-based supersulfated cement paste hydration kinetics performance test unit is configured to test the hydration kinetics performance of the solid waste-based supersulfated cement paste to be tested by using the trained cement hydration kinetics model.

[0078] In one embodiment, the sealed container 2 is further configured to perform the hydration reaction of the solid waste-based supersulfated cement paste sample, as shown in Figure 2 The sealed space in the sealed container 2 is filled with the solid waste-based supersulfated cement paste sample 1, and the sealed space is sealed by the movable polytetrafluoroethylene piston 3.

[0079] The second aspect of the embodiment is configured to perform all the steps of the first aspect of the embodiment.

[0080] The above describes the solid waste-based supersulfated cement performance prediction method and system provided by the present application in detail, the principle and implementation manner of the present application are described by using specific examples in the embodiment, and the above embodiment description is only used to help understand the method and core idea of the present application; meanwhile, for the general technical personnel in the art, the specific implementation manner and application range will be changed according to the idea of the present application, and the above description should not be understood as the limitation of the present application.

[0081] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined in the present embodiment can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown in the present embodiment, but will conform to the widest scope consistent with the principles and novel features disclosed in the present embodiment.

Claims

1. A method for predicting the performance of solid waste-based supersulfate cement, characterized in that: The steps include: S1: Sampling solid waste-based supersulfate cement paste samples, carrying out hydration reaction under sealed conditions, and simultaneously using low-field nuclear magnetic resonance method to detect and quantify the hydration degree data of solid waste-based supersulfate cement paste samples in real time; S2: Constructing a hydration dynamics neural network model, taking the mix ratio data and hydration time of the solid waste-based supersulfate cement paste sample as input, and taking the hydration degree data corresponding to the hydration time as output to train the hydration dynamics neural network as a cement hydration kinetics model; S3: Using the trained cement hydration kinetics model, the hydration kinetics performance of the solid waste-based supersulfate cement paste to be tested is tested.

2. The method for predicting properties of solid waste-based supersulfated cement according to claim 1, wherein: The step of sampling the solid waste-based supersulfate cement slurry sample and performing a hydration reaction under sealed conditions in S1 includes: performing a hydration reaction in a closed container, wherein the closed space in the closed container is filled with the solid waste-based supersulfate cement slurry sample.

3. The method for predicting properties of solid waste-based supersulfated cement according to claim 1, wherein: The step of using a low-field nuclear magnetic resonance method to detect and quantify the hydration degree data of the solid waste-based supersulfate cement paste sample in real time in S1 includes: Detection of T2 relaxation signal of water in supersulfate cement paste samples; Converting the T2 relaxation signal into a T2 relaxation time distribution curve graph by inverse Laplace transform; The decrease in the peak area of ​​the T2 relaxation time spectrum of the supersulfate cement paste within the preset hydration time interval is calculated as the relative generation amount of chemically bound water; The hydration degree data of the solid waste-based supersulfate cement paste samples were quantified according to the relative production of chemically bound water.

4. The method for predicting properties of solid waste-based supersulfated cement according to claim 3, wherein: The formula for quantifying the hydration degree data α(t) of solid waste-based supersulfate cement paste samples is: ； Where, Indicates the total area data of T2 peak at the current moment of hydration test; Represents the total area data of the T2 peak at the start of the hydration test; Represents the total area data of the T2 peak at the last moment of the hydration test.

5. The method for predicting properties of solid waste-based supersulfated cement according to claim 1, wherein: The hydration dynamic neural network includes an input layer, a hidden layer and an output layer; the activation function from the input layer to the hidden layer is a sigmoid function, and the activation function from the hidden layer to the output layer is a Purelin function.

6. A solid waste-based supersulfate cement performance prediction system according to a solid waste-based supersulfate cement performance prediction method according to any one of claims 1 to 5, characterized in that: include: A cement hydration kinetics model construction unit is used to construct a hydration dynamics neural network model, which takes the proportion data and hydration time of a solid waste-based supersulfate cement slurry sample as input and takes the hydration degree data corresponding to the hydration time as output to train the hydration dynamics neural network as a cement hydration kinetics model; wherein the hydration degree data is the hydration degree data of the solid waste-based supersulfate cement slurry sample when it undergoes a hydration reaction under sealed conditions, which is detected and quantitatively sampled in real time using a low-field nuclear magnetic resonance method; The solid waste-based supersulfate cement slurry hydration kinetics performance testing unit is used to test the hydration kinetics performance of the solid waste-based supersulfate cement slurry to be tested using the trained cement hydration kinetics model.

7. The solid waste-based supersulfate cement performance prediction system according to claim 6, characterized in that: It also includes a closed container for carrying out a hydration reaction of a solid waste-based supersulfate cement slurry sample, wherein the closed space in the closed container is filled with the solid waste-based supersulfate cement slurry sample.

Images