CN116759031B - Design method of sludge ash concrete material mixing ratio based on ANN - Google Patents
Design method of sludge ash concrete material mixing ratio based on ANN Download PDFInfo
- Publication number
- CN116759031B CN116759031B CN202311002324.4A CN202311002324A CN116759031B CN 116759031 B CN116759031 B CN 116759031B CN 202311002324 A CN202311002324 A CN 202311002324A CN 116759031 B CN116759031 B CN 116759031B
- Authority
- CN
- China
- Prior art keywords
- super
- ann
- model
- training
- ann model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 67
- 239000010802 sludge Substances 0.000 title claims abstract description 67
- 238000002156 mixing Methods 0.000 title claims abstract description 46
- 238000013461 design Methods 0.000 title claims abstract description 27
- 239000000463 material Substances 0.000 title claims abstract description 25
- 238000012549 training Methods 0.000 claims abstract description 68
- 238000012360 testing method Methods 0.000 claims abstract description 38
- 230000008569 process Effects 0.000 claims abstract description 18
- 239000002994 raw material Substances 0.000 claims abstract description 17
- 238000012795 verification Methods 0.000 claims abstract description 11
- 238000010276 construction Methods 0.000 claims abstract description 3
- 238000005457 optimization Methods 0.000 claims description 47
- 238000010606 normalization Methods 0.000 claims description 23
- 125000003275 alpha amino acid group Chemical group 0.000 claims description 18
- 150000001875 compounds Chemical class 0.000 claims description 18
- 230000006870 function Effects 0.000 claims description 18
- 239000000203 mixture Substances 0.000 claims description 18
- 238000011156 evaluation Methods 0.000 claims description 12
- 238000004364 calculation method Methods 0.000 claims description 9
- 239000004568 cement Substances 0.000 claims description 7
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 7
- 238000011282 treatment Methods 0.000 claims description 5
- VEXZGXHMUGYJMC-UHFFFAOYSA-M Chloride anion Chemical compound [Cl-] VEXZGXHMUGYJMC-UHFFFAOYSA-M 0.000 claims description 4
- 239000003513 alkali Substances 0.000 claims description 4
- 239000003638 chemical reducing agent Substances 0.000 claims description 4
- 239000003795 chemical substances by application Substances 0.000 claims description 4
- 230000008014 freezing Effects 0.000 claims description 4
- 238000007710 freezing Methods 0.000 claims description 4
- 230000010220 ion permeability Effects 0.000 claims description 4
- 239000011159 matrix material Substances 0.000 claims description 4
- 239000004575 stone Substances 0.000 claims description 4
- 239000002562 thickening agent Substances 0.000 claims description 4
- 230000004913 activation Effects 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 3
- 239000004576 sand Substances 0.000 claims description 3
- 238000013528 artificial neural network Methods 0.000 description 79
- 230000000694 effects Effects 0.000 description 8
- 238000010586 diagram Methods 0.000 description 2
- 238000003912 environmental pollution Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 229910052799 carbon Inorganic materials 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 238000001035 drying Methods 0.000 description 1
- 238000004836 empirical method Methods 0.000 description 1
- 239000010419 fine particle Substances 0.000 description 1
- 239000010813 municipal solid waste Substances 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000004064 recycling Methods 0.000 description 1
- 230000001105 regulatory effect Effects 0.000 description 1
- 239000010865 sewage Substances 0.000 description 1
- 238000010200 validation analysis Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16C—COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
- G16C60/00—Computational materials science, i.e. ICT specially adapted for investigating the physical or chemical properties of materials or phenomena associated with their design, synthesis, processing, characterisation or utilisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/0985—Hyperparameter optimisation; Meta-learning; Learning-to-learn
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02W—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO WASTEWATER TREATMENT OR WASTE MANAGEMENT
- Y02W30/00—Technologies for solid waste management
- Y02W30/50—Reuse, recycling or recovery technologies
- Y02W30/91—Use of waste materials as fillers for mortars or concrete
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computing Systems (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Computer Hardware Design (AREA)
- Biomedical Technology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Geometry (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Mathematical Physics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Feedback Control In General (AREA)
Abstract
The invention discloses a design method of sludge ash concrete material mixing ratio based on ANN, which comprises the following steps: collecting and constructing an initial data set of raw material mixing proportion and mechanical property parameters of sludge ash concrete; randomly dividing an initial data set into a training set and a testing set; performing iterative training construction based on the training set to obtain an ANN model, verifying the ANN model according to the testing set, outputting the constructed ANN model if verification is passed, and re-executing the training process to optimize the ANN model if verification is not passed until the ANN model is passed; and inputting the performance index of the sludge ash concrete material to be prepared into an ANN model which is based on verification, and outputting the mixing ratio of the sludge ash concrete raw materials through the ANN model. The invention can solve the problems of long time consumption, large workload and the like of the existing concrete mixing proportion design method.
Description
Technical Field
The invention relates to the technical field of concrete mix proportion design. In particular to a design method of sludge ash concrete material mixing ratio based on ANN.
Background
Sludge ash refers to ash obtained by drying, burning and other treatments of sludge generated in the sewage treatment process. The sludge ash has high activity and fine particles, has certain cement activity, can replace part of cement or aggregate in concrete, and achieves the purposes of recycling resources and reducing environmental pollution. Has important significance for realizing urban solid waste utilization, reducing environmental pollution, reducing the consumption of cement and realizing the power-assisted double-carbon target.
However, the properties of the sludge ash are complex, and the influence of the sludge ash on the performances of the concrete such as strength, durability and the like is obviously different from that of the traditional concrete material. In order to ensure the stability and reliability of the sludge ash concrete, reasonable mix proportion design is necessary. Most of the existing mix proportion design methods are mass methods or volume methods, but the methods take longer time and have larger workload, and cannot meet the mix proportion design requirements of sludge ash concrete. In this context, it is necessary to explore new, fast, efficient design methods.
Disclosure of Invention
Therefore, the invention aims to solve the technical problems of long time consumption, large workload and the like of the existing sludge ash concrete material mixing ratio design method by providing the design method of the sludge ash concrete material mixing ratio based on ANN (Artificial Neural Network ).
In order to solve the technical problems, the invention provides the following technical scheme:
a design method of sludge ash concrete material mixing ratio based on ANN comprises the following steps:
step (1), collecting and constructing an initial data set of raw material mixing proportion and mechanical property parameters of sludge ash concrete;
step (2), randomly dividing the initial data set into a training set and a testing set;
step (3), performing iterative training by using a training set to construct an ANN model, and verifying the ANN model obtained by training according to a testing set to test the fitting condition of the mixing ratio output by the ANN model and the actual mixing ratio; outputting the constructed ANN model if the verification is passed, and re-executing the training process to optimize the ANN model if the verification is not passed until the ANN model is passed;
and (4) inputting the performance index of the sludge ash concrete material to be prepared into an ANN model passing verification, and outputting the mixing ratio of the sludge ash concrete raw materials through the ANN model.
The method for designing the mixing ratio of the sludge ash concrete material based on the ANN comprises the following steps of:
step (3-1), taking the mixing ratio of sludge ash concrete raw materials as an output variable and taking mechanical property parameters as an input variable, and respectively carrying out normalization treatment on the input variable and the output variable;
step (3-2), determining a loss function of the ANN model, adjusting internal parameters of the ANN model based on a gradient descent optimization algorithm, and adopting a Bayesian optimization algorithm to adaptively adjust and select the super parameters of the ANN model so as to reduce the loss function of the ANN model in training iteration to an acceptable level;
a weight w and a bias b are set in each layer of neurons of the ANN; for the known input x and the known output y, the neural network continuously adjusts w and b in training to fit the relation between the existing y and x, a simplified relation of y=wx+b is obtained at the end of training, and w and b are huge parameter matrixes trained by a ann model; the invention adopts a gradient descent algorithm, namely an algorithm for optimizing internal parameters of a model, belongs to a model parameter optimization algorithm, and the adjusted model parameters are w in a model expression y=wx+b. The Bayesian super-parameter optimization algorithm is an algorithm for selecting and setting the super-parameters of the ANN model, belongs to an automatic parameter adjustment algorithm, can replace methods such as manual parameter adjustment, grid search and the like, is convenient, takes short time, can determine the optimal super-parameters, and can solve the problems that the parameter adjustment by an empirical method is time and labor consuming and the super-parameters which enable the model performance to be optimal cannot be obtained; the super parameters include: super parameters such as iteration times, network layer number and the like which are set when modeling is needed;
step (3-3), verifying the trained ANN model by using the test set, inputting the input variable of the test set into the trained ANN model, mapping the output result to a real interval by using a Min-Max inverse normalization algorithm, and comparing the output result with the output variable of the test set;
and (3-4) evaluating the training result of the ANN model by using the root mean square error and the decision coefficient, outputting the training model as the ANN model if the evaluation target is met, and re-executing the training process until the requirement of the evaluation target is met if the evaluation target is not met.
In the design method of the sludge ash concrete material mixing ratio based on ANN, in the step (3-1), the Min-Max normalization method is adopted to normalize the input variable and the output variable, and the calculation formula of the normalization is as follows:
（1）；
in the formula (1), the components are as follows,xthe values that are not normalized are represented,representing the value after normalization;maxrepresents the maximum value in the same batch of data, and min represents the minimum value in the same batch of data;
in step (3-2), the loss function of the ANN model is:
（2）；
in the formula (2), the amino acid sequence of the compound,nrepresenting the total number of training set samples,representing the predicted value of the mix given by the model, < + >>Representing the normalized actual mix ratio;lossvalues in the range of 0 to 0.05 indicate that the model loss function is reduced to an acceptable level;
in the step (3-3), the Min-Max inverse normalization algorithm has a calculation formula:
（3）；
in the formula (3), the amino acid sequence of the compound,xthe values that are not normalized are represented,representing the value after normalization;maxrepresents the maximum value in the same batch of data,minrepresenting the minimum value in the same batch of data;
in step (3-4), root mean square errorRMSEDetermining coefficientsR 2 Is calculated by the formula of (2)The method comprises the following steps:
（4）；
（5）；
in the formulas (4) and (5),nrepresenting the total number of samples of the test set,representing the inverse normalized predicted value,/->Representing the actual value +.>Representation ofnAn average of actual values of the individual test set samples; when root mean square errorRMSELess than or equal to 0.01 and determining coefficientsR 2 When the training model is more than or equal to 0.95, the training model is shown to meet the requirements; root mean square errorRMSEThe closer the value of (2) is to 0, the better the fitting effect of the model is, and the coefficient is determinedR 2 The closer to 1 the value of (c) indicates the better the fitting of the model. Thus, if the root mean square errorRMSELarger values or determining coefficientsR 2 If the value is smaller, the training process should be adjusted and re-executed until the root mean square errorRMSEDetermining coefficientsR 2 Meets the requirements.
In the method for designing the sludge ash concrete material mixing ratio based on the ANN, in the step (3-2), the super parameters of the ANN model are adaptively adjusted and selected by adopting a Bayes super parameter optimization algorithm so as to improve the modeling precision of the ANN model, and finally, the sludge ash concrete mixing ratio ANN design model with high precision is obtained; the specific tuning method comprises the following steps:
firstly, designing or adjusting an optimization space Θ of each super parameter in an ANN model according to experience or performance of model test so as to reduce operation complexity; the optimization space can be adjusted according to the predicted performance of the subsequent model;
secondly, setting an objective function of super parameter tuningh(θ)：
（6）；
In the formula (6), the amino acid sequence of the compound,nrepresenting the total number of samples,representing predicted values +.>The actual value is represented by a value that is,θrepresenting the super-parameters;
then, solving by using a Bayes super-parameter optimization method to obtain the leadh(θ) Super-parameters with minimum values; the solving process is expressed as:
（7）；
in the formula (7), the amino acid sequence of the compound,θ* The optimal super-parameters to be searched for by the Bayes super-parameter optimization algorithm,θfor the input hyper-parameters, Θ is the set parameter space,nrepresenting the total number of samples,representing predicted values +.>Representing the actual value.
The method for searching the optimal super-parameters by using the Bayes super-parameter optimization algorithm comprises the following steps of:
objective functionh(θ) Obeys gaussian distribution, i.e.:
（8）；
in the formula (8), the amino acid sequence of the compound,μ(θ) Is thath(θ) Is used for the average value of (a),O(θ,θ(v) ish(θ) Is used for the co-variance matrix of (a),O(θ,θ’) The initial value of (2) is expressed as:
（9）；
in the Bayes super-parameter optimization process, covariance matrixO(θ,θ’) Will change with the training iteration, and as the training iteration proceeds, assume that the super parameter input at step t+1 isθ t+1 ThenO(θ,θ’) The values of (2) are expressed as:
（10）；
thus, the objective functionh(θ) The posterior probability of (2) is calculated by the following formula:
（11）；
in the formula (11), the amino acid sequence of the compound,Din order to observe the value of the value,μ t+1 (θ) Is t+1st steph(θ) Mean, sigma of 2 t+1 (θ) Is t+1st steph(θ) Is a variance of (2);
after posterior probability is obtained, searching the space near the last super parameter according to posterior distribution by a Bayes super parameter optimization algorithm, wherein the searching method comprises the following steps:
（12）；
zeta in (12) t+1 Is a constant, set to 0.01;θ t+1 is the super parameter of the selected t+1 step;
through continuous iterative search, the optimal super-parameters in a given super-parameter optimization space Θ can be determined by using a Bayesian super-parameter optimization algorithm, and the set of super-parameters can enable the ANN model to obtain an optimal training result.
According to the design method of the sludge ash concrete material mixing ratio based on the ANN, the ANN model super parameters regulated by the Bayesian super parameter optimization algorithm comprise the number of hidden layers, the training iteration times, an optimizer, the batch sample size, the learning rate, the activation function and the discarding method ratio.
In the method for designing the mixing ratio of the sludge ash concrete material based on the ANN, in the step (1), an initial data set of the mixing ratio of the sludge ash concrete raw material and the mechanical property is obtained by collecting public data materials and/or experimental modes.
In the design method of the sludge ash concrete material mixing ratio based on ANN, in the step (2), the data in the initial data set are randomly divided into a training set and a testing set according to the proportion of 7:3.
In the design method of the mixing proportion of the sludge ash concrete material based on the ANN, in the step (1), the sludge ash concrete raw material comprises sludge ash, cement, broken stone, sand, a water reducing agent, a thickening agent, an expanding agent and water; the mechanical properties include compressive strength, flexural strength, slump and durability.
The design method of the mixing ratio of the sludge ash concrete material based on ANN has the durability of one or two or more of chloride ion permeability coefficient, freezing resistance and alkali aggregate reaction.
The technical scheme of the invention has the following beneficial technical effects:
1. because the influence of the sludge ash on the concrete performance is complex, the prior mixing proportion design means is long in time consumption and large in workload, and the complex mixing proportion design requirement of the sludge ash concrete is difficult to meet. The mix proportion design model based on ANN and Bayesian super-parameter optimization provided by the invention is not only far higher in prediction precision than the prior art, but also higher in intelligent and automation level, and is more suitable for engineering practice.
2. According to the invention, the design is carried out based on the performance index of the sludge ash concrete, the ANN artificial neural network method is used for modeling the performance parameters such as the mixing ratio and the compressive strength of various admixtures of the sludge ash concrete, and the Bayesian super-parameter optimization algorithm is used for optimizing a plurality of super parameters of the ANN model, so that the fitting capacity of the ANN model can be fully exerted compared with the manual optimization method based on experience, and the optimal design model is obtained.
3. According to the invention, an ANN is used for constructing a blending ratio regression prediction model of each raw material of the sludge ash concrete, the traditional modes such as a quality method, a volume method and a mathematical relation are replaced, up to 14 variables can be modeled, and a complex nonlinear relation between the performance index of the sludge ash concrete and the blending ratio is better fitted, so that the blending ratio design which is closer to the preset performance index is obtained.
Drawings
FIG. 1 is a schematic diagram of a model training process when an ANN model is constructed in an embodiment of the invention;
FIG. 2 is a schematic diagram of a model test process when an ANN model is constructed in an embodiment of the present invention;
FIG. 3 is a graph of model predictive value versus actual value based on a validation set in an embodiment of the invention.
Detailed Description
The ANN-based sludge ash concrete material mixing proportion design method comprises the following steps:
step (1), collecting and constructing an initial data set of raw material mixing proportion and mechanical property parameters of sludge ash concrete; in the embodiment, the raw material mixing ratio and the mechanical property data of the sludge ash concrete are obtained by collecting network public data and experimental acquisition, and an initial data set is constructed.
Step (2), randomly dividing the initial data set into a training set and a testing set; the values of compressive strength, flexural strength, slump, durability (durability includes chloride ion permeability coefficient, freezing resistance, alkali aggregate reaction) in the initial data set are used as input variables, and the mixing proportion of sludge ash, cement, crushed stone, sand, water reducer, thickener and expanding agent is used as output variable; the initial data set is randomly divided into a training set and a testing set according to the proportion of 7:3, the training set is used for training a model to obtain the best fitting effect, and the testing set is used for verifying and evaluating the calculating effect of the training model on the sludge ash concrete mixing proportion.
Performing iterative training by using a training set to construct an ANN model, verifying the ANN model obtained by training according to a test set, outputting the constructed ANN model if verification is passed, and re-executing the training process to optimize the ANN model if verification is not passed until the ANN model is passed; namely: an ANN model is built, and the model is trained based on training set iteration, so that the model can output a result which is more and more close to an output variable according to the numerical value of the input variable of the training set. In order to obtain a super-parameter combination which enables the ANN model to have better performance, the embodiment uses a Bayesian super-parameter optimization algorithm to carry out self-adaptive optimization on the model super-parameters such as iteration times, learning rate and the like. And verifying the ANN model based on the optimal parameter combination by using the test set, and evaluating the training effect of the model by using the root mean square error and the decision coefficient as evaluation indexes.
Specifically, the construction method of the ANN model comprises the following steps:
step (3-1), taking the mixing ratio of sludge ash concrete raw materials as an output variable and taking mechanical property parameters as an input variable, and respectively carrying out normalization treatment on the input variable and the output variable; the Min-Max normalization method is adopted to normalize the input variable and the output variable, and the calculation formula of the normalization is as follows:
（1）；
in the formula (1), the components are as follows,xthe values that are not normalized are represented,representing the value after normalization;maxrepresents the maximum value in the same batch of data,minrepresenting the minimum value in the same batch of data.
Step (3-2), determining a loss function of the ANN model, adjusting internal parameters of the ANN model based on a gradient descent optimization algorithm, and adopting a Bayesian optimization algorithm to adaptively adjust and select the super parameters of the ANN model so as to reduce the loss function of the ANN model in training iteration to an acceptable level; the loss function of the ANN model is:
（2）；
in the formula (2), the amino acid sequence of the compound,nrepresenting the total number of training set samples,representing the predicted value of the mix given by the model, < + >>Representing the normalized actual mix ratio;lossvalues in the range of 0 to 0.05 indicate that the model loss function is reduced to an acceptable level;
the ANN model super parameters optimized by the Bayesian super parameter optimization algorithm comprise the number of hidden layers, the training iteration times, an optimizer, batch sample size, learning rate, activation function and discarding method ratio; the method for optimizing the internal super parameters of the ANN model by using the Bayesian super parameter optimization algorithm comprises the following steps:
firstly, designing or adjusting an optimization space Θ of each super parameter in an ANN model according to the performance of experience or model test;
secondly, setting an objective function of super parameter tuningh(θ)：
（6）；
In the formula (6), the amino acid sequence of the compound,nrepresenting the total number of samples,representing predicted values +.>Representation ofThe actual value of the current,θrepresenting the super-parameters;
then, solving by using a Bayes super-parameter optimization method to obtain the leadh(θ) Super-parameters with minimum values; the solving process is expressed as:
（7）；
in the formula (7), the amino acid sequence of the compound,θ* The optimal super-parameters to be searched for by the Bayes super-parameter optimization algorithm,θfor the input hyper-parameters, Θ is the set parameter space,nrepresenting the total number of samples,representing predicted values +.>Representing the actual value.
Step (3-3), verifying the trained ANN model by using the test set, inputting the input variable of the test set into the trained ANN model, mapping the output result to a real interval by using a Min-Max inverse normalization algorithm, and comparing the output result with the output variable of the test set; the Min-Max inverse normalization algorithm has the following formula:
（3）；
in the formula (3), the amino acid sequence of the compound,xthe values that are not normalized are represented,representing the value after normalization;maxrepresents the maximum value in the same batch of data,minrepresenting the minimum value in the same batch of data;
and (3-4) evaluating the training result of the ANN model by using the root mean square error and the decision coefficient, outputting the training model as the ANN model if the evaluation target is met, and re-executing the training process until the requirement of the evaluation target is met if the evaluation target is not met. Mean squareRoot errorRMSEDetermining coefficientsR 2 The calculation formula of (2) is as follows:
（4）；
（5）；
in the formulas (4) and (5),nrepresenting the total number of samples of the test set,representing the inverse normalized predicted value,/->Representing the actual value +.>Representation ofnAn average of actual values of the individual test set samples; when root mean square errorRMSELess than or equal to 0.01 and determining coefficientsR 2 And when the training model is greater than or equal to 0.95, the training model is indicated to meet the requirements.
The method for searching the optimal super-parameters by the Bayes super-parameter optimization algorithm comprises the following steps:
objective functionh(θ) Obeys gaussian distribution, i.e.:
（8）；
in the formula (8), the amino acid sequence of the compound,μ(θ) Is thath(θ) Is used for the average value of (a),O(θ,θ(v) ish(θ) Is used for the co-variance matrix of (a),O(θ,θ’) The initial value of (2) is expressed as:
（9）；
as training iterations progress, assume that the super-parameters entered at step t+1 areθ t+1 ThenO(θ,θ’) The values of (2) are expressed as:
（10）；
thus, the objective functionh(θ) The posterior probability of (2) is calculated by the following formula:
（11）；
in the formula (11), the amino acid sequence of the compound,Din order to observe the value of the value,μ t+1 (θ) Is t+1st steph(θ) Mean, sigma of 2 t+1 (θ) Is t+1st steph(θ) Is a variance of (2);
after posterior probability is obtained, searching the space near the last super parameter according to posterior distribution by a Bayes super parameter optimization algorithm, wherein the searching method comprises the following steps:
（12）；
zeta in (12) t+1 Is a constant, set to 0.01;θ t+1 is the super parameter of the selected t+1 step;
by constant iterative search, the optimal superparameter in a given superparameter optimization space Θ is determined.
In the process of constructing the ANN model, super parameters such as the number of hidden layers, the iteration times, the learning rate and the like of the ANN model are not required to be preset, the super parameters are adaptively determined in the training process by a Bayesian super parameter optimization algorithm, and only an optimization space of the super parameters is required to be preset; the optimization space given in this example is shown in table 1.
TABLE 1 super parameter optimization space
Before the data is input into the model, min-Max normalization processing is carried out; after the model outputs the result, the predicted value in the true value domain can be obtained by inverse normalization calculation.
The training data in the initial data set collected in this embodiment is 84 sets, and the remaining 36 sets are used as test sets to verify the performance of the model after training (since there are more data, this embodiment does not enumerate the data). In the model training process, the compression strength in the training data is usedX 1 ) Break strength%X 2 ) Slump ofX 3 ) Chloride ion permeability coefficient [ ]X 4 ) Freezing resistanceX 5 ) Reaction of alkali aggregateX 6 ) Six groups of variables are taken as input variables of the model, and the model is trained to output sludge ashY 1 ) CementY 2 ) Crushing stoneY 3 ) SandY 4 ) The water isY 5 ) Water reducing agentY 6 ) ThickenerY 7 ) Expansion agentY 8 ) Eight sets of variables.
After the training is finished, the super-parameter set obtained by Bayes super-parameter optimization in the training can be checked. The super parameter set in this example is shown in table 2.
Table 2 super parameter set obtained by Bayes super parameter optimization
Finally, 36 groups of test data are used for verifying the performance of the model, and root mean square error is usedRMSE) And determining coefficient [ ]R 2 ) And evaluating the model fitting effect by the two evaluation indexes. It is known from the calculation that,RMSE=0.00263，R 2 = 0.98221. The model gives a comparison of the predicted value with the actual value as shown in fig. 3.
The embodiment of the inventionThe sludge ash concrete mixing proportion modeling is obtained in the test setRMSE=0.00263，R 2 Test results of = 0.98221. According toRMSEAnd R is 2 The formula and meaning of (c) can be known,RMSEthe closer the value of (2) is to 0, R 2 The closer to 1 the value of (c) is, the higher the fitting degree of the predicted value given by the model to the actual value is. Therefore, as can be seen from the evaluation indexes, the ANN model constructed in the embodiment fits the sludge ash concrete performance indexes and the mixture ratio well, which shows that the ANN model fully learns the rules between the sludge ash concrete performance indexes and the mixture ratio, and can give the required mixture ratio according to the expected performance index values. The comparison of the predicted value and the actual value in fig. 2 can also be obtained by: and the ANN is combined with a Bayes super-parameter optimization method to give a match ratio regression prediction result very close to the actual value.
And (4) storing the verified ANN model, and when the method is applied to engineering, firstly inputting the expected sludge ash concrete performance index value to be achieved, and then calculating the ANN model according to the trained parameters so as to output the mixing ratio. And (5) preparing the sludge ash concrete according to the mixing ratio.
According to the experimental test of the sludge ash concrete performance index of the sludge ash concrete prepared according to the mixing ratio output by the ANN model, the performance of the sludge ash concrete prepared by adopting the mixing ratio output by the ANN model is basically consistent with the expected performance index to be achieved, and the engineering application requirements are completely met.
In other embodiments, the types of the raw materials of the sludge ash concrete can be adjusted as output variables according to actual conditions, or the types of the indexes of the mechanical property data can be adjusted as input variables according to application indexes of the sludge ash concrete, so that a new ANN model is constructed.
It is apparent that the above examples are given by way of illustration only and are not limiting of the embodiments. Other variations or modifications of the above teachings will be apparent to those of ordinary skill in the art. It is not necessary here nor is it exhaustive of all embodiments. While the obvious variations or modifications which are extended therefrom remain within the scope of the claims of this patent application.
Claims (4)
1. The design method of the sludge ash concrete material mixing ratio based on the ANN is characterized by comprising the following steps of:
step (1), collecting and constructing an initial data set of raw material mixing proportion and mechanical property parameters of sludge ash concrete; the sludge ash concrete raw materials comprise sludge ash, cement, broken stone, sand, a water reducing agent, a thickening agent, an expanding agent and water; mechanical properties include compressive strength, flexural strength, slump and durability;
step (2), randomly dividing the initial data set into a training set and a testing set;
performing iterative training by using a training set to construct an ANN model, verifying the ANN model obtained by training according to a test set, outputting the constructed ANN model if verification is passed, and re-executing the training process to optimize the ANN model if verification is not passed until the ANN model is passed;
step (4), inputting performance indexes of the sludge ash concrete material to be prepared into an ANN model passing verification, and outputting the mixing ratio of the sludge ash concrete raw materials through the ANN model;
in the step (3), the construction method of the ANN model comprises the following steps:
step (3-1), taking the mixing ratio of sludge ash concrete raw materials as an output variable and taking mechanical property parameters as an input variable, and respectively carrying out normalization treatment on the input variable and the output variable;
in the step (3-1), a Min-Max normalization method is adopted to normalize the input variable and the output variable, and a calculation formula of the normalization is as follows:
（1）；
in the formula (1), the components are as follows,indicating non-normalized values, ++>Representing the value after normalization; />Represents the maximum value in the same batch of data, +.>Representing the minimum value in the same batch of data;
step (3-2), determining a loss function of the ANN model, adjusting internal parameters of the ANN model based on a gradient descent optimization algorithm, and adopting a Bayesian optimization algorithm to adaptively adjust and select the super parameters of the ANN model so as to reduce the loss function of the ANN model in training iteration to an acceptable level;
in step (3-2), the loss function of the ANN model is:
（2）；
in the formula (2), the amino acid sequence of the compound,representing the total number of training set samples, +.>Representing the predicted value of the mix given by the model, < + >>Representing the normalized actual mix ratio;lossvalues in the range of 0 to 0.05 indicate that the model loss function is reduced to an acceptable level;
step (3-3), verifying the trained ANN model by using the test set, inputting the input variable of the test set into the trained ANN model, mapping the output result to a real interval by using a Min-Max inverse normalization algorithm, and comparing the output result with the output variable of the test set;
in the step (3-3), the Min-Max inverse normalization algorithm has a calculation formula:
（3）；
in the formula (3), the amino acid sequence of the compound,indicating non-normalized values, ++>Representing the value after normalization; />Represents the maximum value in the same batch of data, +.>Representing the minimum value in the same batch of data;
step (3-4), evaluating the training result of the ANN model by using root mean square error and a decision coefficient, outputting the training model as the ANN model if the evaluation target is met, and re-executing the training process if the evaluation target is not met until the requirement of the evaluation target is met;
in step (3-4), root mean square errorRMSEDetermining coefficientsR 2 The calculation formula of (2) is as follows:
（4）；
（5）；
in the formulas (4) and (5),representing the total number of test set samples, +.>Representing the inverse normalized predicted value,/->Representing the actual value +.>Representation->An average of actual values of the individual test set samples; when root mean square errorRMSELess than or equal to 0.01 and determining coefficientsR 2 When the training model is more than or equal to 0.95, the training model is shown to meet the requirements;
in the step (3-2), adopting a Bayes super-parameter optimization algorithm to carry out self-adaptive adjustment and selection on the super parameters of the ANN model; the ANN model super parameters optimized by the Bayesian super parameter optimization algorithm comprise the number of hidden layers, the training iteration times, an optimizer, batch sample size, learning rate, activation function and discarding method ratio; the specific tuning method comprises the following steps:
firstly, designing or adjusting an optimization space Θ of each super parameter in an ANN model according to the performance of experience or model test;
secondly, setting an objective function of super parameter tuningh(θ)：
（6）；
In the formula (6), the amino acid sequence of the compound,representing the total number of samples->Representing predicted values +.>The actual value is represented by a value that is,θrepresenting the super-parameters;
then, solving by using a Bayes super-parameter optimization method to obtain the leadh(θ) Super-parameters with minimum values; the solving process is expressed as:
（7）；
in the formula (7), the amino acid sequence of the compound,θ* The optimal super-parameters to be searched for by the Bayes super-parameter optimization algorithm,θfor the input hyper-parameters, Θ is the set parameter space,representing the total number of samples->Representing predicted values +.>Representing the actual value;
the method for searching the optimal super-parameters by the Bayes super-parameter optimization algorithm comprises the following steps:
objective functionh(θ) Obeys gaussian distribution, i.e.:
（8）；
in the formula (8), the amino acid sequence of the compound,μ(θ) Is thath(θ) Is used for the average value of (a),O(θ,θ(v) ish(θ) Is used for the co-variance matrix of (a),O(θ,θ’) The initial value of (2) is expressed as:
（9）；
as training iterations progress, assume that the super-parameters entered at step t+1 areθ t+1 ThenO(θ,θ’) The values of (2) are expressed as:
（10）；
thus, the objective functionh(θ) The posterior probability of (2) is calculated by the following formula:
（11）；
in the formula (11), the amino acid sequence of the compound,Din order to observe the value of the value,μ t+1 (θ) Is t+1st steph(θ) Mean, sigma of 2 t+1 (θ) Is t+1st steph(θ) Is a variance of (2);
after posterior probability is obtained, searching the space near the last super parameter according to posterior distribution by a Bayes super parameter optimization algorithm, wherein the searching method comprises the following steps:
（12）；
zeta in (12) t+1 Is a constant, set to 0.01;θ t+1 is the super parameter of the selected t+1 step;
by constant iterative search, the optimal superparameter in a given superparameter optimization space Θ is determined.
2. The method for designing the mix proportion of sludge ash concrete materials based on ANN according to claim 1, wherein in the step (1), the initial data set of the mix proportion of sludge ash concrete raw materials and mechanical properties is obtained by collecting public data and/or experimental means.
3. The method for designing a mix ratio of sludge ash concrete materials based on ANN according to claim 1, wherein in step (2), the data in the initial data set is randomly divided into a training set and a test set according to a ratio of 7:3.
4. The method for designing the mix ratio of the ANN-based sludge ash concrete material according to claim 1, wherein the durability is one or two or more of chloride ion permeability coefficient, freezing resistance and alkali aggregate reaction.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311002324.4A CN116759031B (en) | 2023-08-10 | 2023-08-10 | Design method of sludge ash concrete material mixing ratio based on ANN |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311002324.4A CN116759031B (en) | 2023-08-10 | 2023-08-10 | Design method of sludge ash concrete material mixing ratio based on ANN |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116759031A CN116759031A (en) | 2023-09-15 |
| CN116759031B true CN116759031B (en) | 2024-01-26 |
Family
ID=87953545
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311002324.4A Active CN116759031B (en) | 2023-08-10 | 2023-08-10 | Design method of sludge ash concrete material mixing ratio based on ANN |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116759031B (en) |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107633301A (en) * | 2017-08-28 | 2018-01-26 | 广东工业大学 | The training method of testing and its application system of a kind of BP neural network regression model |
| WO2021007812A1 (en) * | 2019-07-17 | 2021-01-21 | 深圳大学 | Deep neural network hyperparameter optimization method, electronic device and storage medium |
| CN113059570A (en) * | 2021-04-09 | 2021-07-02 | 华南理工大学 | Human-robot cooperative control method based on human body dynamic arm strength estimation model |
| CN113554148A (en) * | 2021-06-07 | 2021-10-26 | 南京理工大学 | BiLSTM voltage deviation prediction method based on Bayesian optimization |
| CN115169714A (en) * | 2022-07-08 | 2022-10-11 | 福州大学 | Urban subway station entrance and exit passenger flow prediction method |
| CN115206463A (en) * | 2022-06-30 | 2022-10-18 | 重庆茂侨科技有限公司 | Machine learning-based intelligent design method for concrete mix proportion |
| CN115564155A (en) * | 2022-10-11 | 2023-01-03 | 云南电网有限责任公司电力科学研究院 | Distributed wind turbine generator power prediction method and related equipment |
| CN116418000A (en) * | 2022-10-31 | 2023-07-11 | 河北工业大学 | Bayesian state estimation method for unobservable distribution network based on embedded tide neural network |
| CN116423924A (en) * | 2023-03-31 | 2023-07-14 | 广东佛斯伯智能设备有限公司 | Paperboard line wet end production speed model construction method and production speed prediction method |
-
2023
- 2023-08-10 CN CN202311002324.4A patent/CN116759031B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107633301A (en) * | 2017-08-28 | 2018-01-26 | 广东工业大学 | The training method of testing and its application system of a kind of BP neural network regression model |
| WO2021007812A1 (en) * | 2019-07-17 | 2021-01-21 | 深圳大学 | Deep neural network hyperparameter optimization method, electronic device and storage medium |
| CN113059570A (en) * | 2021-04-09 | 2021-07-02 | 华南理工大学 | Human-robot cooperative control method based on human body dynamic arm strength estimation model |
| CN113554148A (en) * | 2021-06-07 | 2021-10-26 | 南京理工大学 | BiLSTM voltage deviation prediction method based on Bayesian optimization |
| CN115206463A (en) * | 2022-06-30 | 2022-10-18 | 重庆茂侨科技有限公司 | Machine learning-based intelligent design method for concrete mix proportion |
| CN115169714A (en) * | 2022-07-08 | 2022-10-11 | 福州大学 | Urban subway station entrance and exit passenger flow prediction method |
| CN115564155A (en) * | 2022-10-11 | 2023-01-03 | 云南电网有限责任公司电力科学研究院 | Distributed wind turbine generator power prediction method and related equipment |
| CN116418000A (en) * | 2022-10-31 | 2023-07-11 | 河北工业大学 | Bayesian state estimation method for unobservable distribution network based on embedded tide neural network |
| CN116423924A (en) * | 2023-03-31 | 2023-07-14 | 广东佛斯伯智能设备有限公司 | Paperboard line wet end production speed model construction method and production speed prediction method |
Non-Patent Citations (1)
| Title |
|---|
| 张丹丹.基于深度神经网络技术的高分遥感图像处理及应用.北京：中国宇航出版社,2020,第94-96页. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116759031A (en) | 2023-09-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111985796B (en) | Method for predicting concrete structure durability based on random forest and intelligent algorithm | |
| CN109242236B (en) | PMU primary frequency modulation load prediction method based on correlation full-connection neural network and LSTM | |
| Bachir et al. | Using artificial neural networks approach to estimate compressive strength for rubberized concrete | |
| CN110435009B (en) | Intelligent design method for concrete production mix proportion | |
| CN110598929B (en) | Wind power nonparametric probability interval ultrashort term prediction method | |
| CN109359737B (en) | BP neural network prediction analysis method for recycled aggregate pervious concrete | |
| CN115847596B (en) | Production debugging control method and system for gypsum block | |
| CN112069656A (en) | Durable concrete mix proportion multi-objective optimization method based on LSSVM-NSGAII | |
| Chen et al. | Optimization of high-performance concrete mix ratio design using machine learning | |
| CN116187731A (en) | Dynamic optimization method of permeable brick compression molding production line based on neural network | |
| CN116759031B (en) | Design method of sludge ash concrete material mixing ratio based on ANN | |
| CN111709585A (en) | Air conditioner load prediction method and device and storage medium | |
| CN117935988A (en) | Method for predicting compressive strength of recycled coarse aggregate concrete based on support vector regression | |
| Mahmood et al. | Enhancing compressive strength prediction in self-compacting concrete using machine learning and deep learning techniques with incorporation of rice husk ash and marble powder | |
| CN112800673A (en) | Method for predicting blasting block degree based on SA-GA-BP algorithm | |
| CN111144051A (en) | Geopolymer concrete strength prediction method based on support vector machine and improved cuckoo algorithm | |
| CN113011091A (en) | Automatic-grouping multi-scale light-weight deep convolution neural network optimization method | |
| Mehmannavaz et al. | Research Article Presenting an Appropriate Neural Network for Optimal Mix Design of Roller Compacted Concrete Dams | |
| Kellouche et al. | Comparative study of different machine learning approaches for predicting the compressive strength of palm fuel ash concrete | |
| Haghighi et al. | Optimizing mix proportions of reactive powder concrete using group method of data handling and genetic programming | |
| Vasudevan | Concrete Classification Using Machine Learning Techniques | |
| CN117283743B (en) | Resin production molding processing flow prediction control system and method | |
| CN116844645B (en) | Gene regulation network inference method based on multi-view layered hypergraph | |
| Martynov et al. | Method for calculating the composition of cellular concrete | |
| Zhang et al. | Research on Hydraulic Concrete Strength Prediction Model Based on ANN-RBF Neural Network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |