Method for predicting compressive strength of bacillus mucilaginosus cemented cyanidation tailings
By combining the response surface method and machine learning model, establish a data set and use the LSTM neural network to predict, optimize the parameters and form the GUI user page, solve the problem of poor predicting effects when dealing with complex non -linear relations systems, realize the poor effects, realize the realization of poor effects, realize the realization of the effect, realize the realization of the effect, realize the realization of the effect, realize it, and achieve realization. The efficient predictive prediction of glue Bacteria gel tail residue compressive strength.
Patent Information
- Application Number
- CN202411398982.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-09
- Publication Date
- 2025-01-21
AI Technical Summary
The prior art cannot accurately capture the complex interactions between variables when processing systems with complex nonlinear relationships, and the prediction effect is not ideal for data beyond the experimental scope.
Using the method of combining the response surface method and the machine learning model, by sharing the response surface method data, a data set of the machine learning model is established, standardized, and randomly divided into a training set and a test set, and the LSTM neural network model is used for prediction, and through The genetic algorithm optimizes parameters and finally forms a GUI visual user page.
The prediction accuracy and efficiency of the compressive strength of cemented cyanide tailings cemented by Bacillus glioides are significantly improved, making it easy for non-technical users to use this tool, which has important practical significance.
Smart Images
Figure CN119337717A_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the technical field of environmental engineering and biotechnology, and specifically refers to a method for predicting the compressive strength of cyanidation tailings cemented by Bacillus mucilaginosus. Background Art
[0002] Gold mineral resources are important materials to ensure the stable development of my country's economic construction and national economy. The cyanide gold extraction method has a series of advantages such as simple process, high gold recovery rate, low cost, strong adaptability to ore, and the ability to produce gold locally. After a hundred years, it still occupies an absolute dominant position in the current gold industry.
[0003] However, a large amount of cyanide waste residue is often produced after cyanide gold extraction, which brings harm to the surrounding environment and the health of animals and plants. Therefore, it is very necessary to develop a green, clean, and low-cost method for harmless treatment and disposal of cyanide tailings. The treatment and disposal of cyanide tailings mainly involves the recovery and extraction of valuable metals, the production of building materials, and the filling of tailings in goafs. At present, filling cyanide tailings into underground goafs is an environmentally friendly, reliable, and mainstream treatment method for cyanide tailings. The carbonic anhydrase-producing characteristics of Bacillus subtilis induce the production of carbonate deposition, which can cement cyanide tailings to form a solid body, which can reduce environmental pollution and reduce costs.
[0004] The higher the strength of the above-mentioned cyanide tailings, the lower the permeability, which helps to prevent groundwater infiltration and the spread of pollutants. However, high-strength cementation may have an impact on microbial ecology. For example, excessive use of binders may inhibit the growth of Bacillus colloids, affecting its role in the ecosystem, and may also increase treatment costs, increase operational difficulty and risks. Therefore, it is very important to optimize the independent variable parameters of Bacillus colloids cemented cyanide tailings.
[0005] Based on the above problems, we can provide a method to predict the compressive strength of cyanide tailings cemented by Bacillus colloids to solve the above problems. In addition, the development of a graphical user interface (GUI) enables non-technical users to use this tool conveniently, which has important practical significance. Summary of the invention
[0006] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides a method for predicting the compressive strength of cemented cyanide tailings by Bacillus subtilis, which is used to solve the difficulty of controlling the compressive strength of cemented cyanide tailings, and the problem that the response surface method cannot accurately capture the complex interactions between variables when processing systems with complex nonlinear relationships, and the prediction effect is not ideal for data beyond the experimental range.
[0007] The technical solutions adopted in this application are as follows:
[0008] This scheme provides a method for predicting the compressive strength of cyanide tailings cemented by Bacillus colloids, comprising the following steps:
[0009] (1) Share the response surface methodology data, including the calcium source type, calcium source addition amount, bacterial powder addition amount, and moisture content of the cemented specimens, and establish a data set for the machine learning model;
[0010] (2) Standardize the data set to obtain a preprocessed sample data set;
[0011] (3) randomly dividing the pre-processed sample data set into a training set and a test set, and using a machine learning model to predict the compressive strength prediction model of Bacillus colloids cemented cyanidation tailings;
[0012] (4) Using the optimization algorithm, find the optimal parameters of the compressive strength of the cyanide tailings cemented by Bacillus sp.
[0013] (5) A visualization page is formed according to the prediction model for the compressive strength of Bacillus cyanidation tailings cemented by Bacillus cyanidation tailings to develop a prediction system for the compressive strength of Bacillus cyanidation tailings cemented by Bacillus cyanidation tailings.
[0014] As a further preferred technical solution, in step (1), response surface analysis is performed on the basis of the orthogonal test results, and three factors that have a greater impact on the compressive strength of the sample are selected: calcium source type, calcium source addition amount, and water content as the factors for investigation. The response surface method can not only establish a regression model with high calculation accuracy, but also show the influence of the influencing factors and their interactions through qualitative analysis, and can more accurately optimize the experimental conditions and the ratio of each factor level.
[0015] As a further preferred technical solution, in step (1), the Box-Behken module in Design-Expert 12 software is used to perform response surface analysis based on the orthogonal test results. Three factors that have a greater impact on the compressive strength of the sample are selected: calcium source type, calcium source addition amount, and moisture content as the investigation factors.
[0016] As a further preferred technical solution, in step (1), the response surface methodology and the machine learning model comprehensively consider the solidification performance of the cyanide tailings cemented by Bacillus subtilis and the cost control of the technical implementation, and calcium lactate is used as the cementing calcium source.
[0017] As a further preferred technical solution, in step (1), the default calcium source type in the data set is calcium lactate, the calcium source addition amount and moisture content are used as input variables, the compressive strength is used as the output, the data is read using origin 2022 software, and a data set for the machine learning model is established.
[0018] As a further preferred technical solution, in the step (2), a tool StandardScaler in the scikit-learn library is used to standardize the data set, and the input data (calcium source addition amount and moisture content) and the output data (compressive strength) are converted into a standard normal distribution to ensure that each variable is within the same scale range.
[0019] As a further preferred technical solution, in the step (3), the train_test_split function in the scikit-learn library is used to divide the standardized data set into a training set and a test set, and the processed data set is divided into 80% of the training set and 20% of the test set, and a long short-term memory (LSTM) neural network algorithm model is adopted.
[0020] As a further preferred technical solution, in step (3), the trained LSTM model needs to be evaluated by determining the coefficient of determination (R 2 ), mean absolute error (MAE), mean absolute deviation (AAD) and other indicators are used to measure the prediction performance of the model.
[0021] As a further preferred technical solution, in the step (4), a genetic algorithm based on natural selection and genetic mechanism is used to perform parameter optimization.
[0022] As a further preferred technical solution, in the step (5), a GUI visual user interface is used to achieve rapid prediction and intuitive display of the compressive strength of the cyanidation tailings cemented by Bacillus subtilis.
[0023] The beneficial effects achieved by the present invention using the above scheme are as follows:
[0024] 1. Use machine learning models to predict and optimize the bonding effect under different variable conditions, significantly improving the accuracy and efficiency of predictions.
[0025] 2. The GUI visual user page design enables non-technical users to use this tool easily. The visual user page can quickly predict and intuitively display the compressive strength of Bacillus cereus cemented cyanidation tailings, which has important practical significance. BRIEF DESCRIPTION OF THE DRAWINGS
[0026] Figure 1 This is a graph of error changes during the training process of the machine learning model provided in this embodiment;
[0027] Figure 2 The number of hidden neurons in the model training process provided in this embodiment is proportional to R 2 and schematic diagram of the impact of AAD;
[0028] Figure 3A scatter plot of the test data and training data provided in this embodiment;
[0029] Figure 4 This is a fitness curve diagram during the evolution process provided by this embodiment. DETAILED DESCRIPTION
[0031] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention.
[0032] Example
[0033] In order to solve the problem that existing technologies such as response surface analysis methods cannot accurately capture the complex interactions between variables when dealing with systems with complex nonlinear relationships, and the prediction effect is not ideal for data beyond the experimental range, the present invention provides a method for predicting the compressive strength of Bacillus colloids cemented cyanidation tailings based on machine learning.
[0034] Specifically, as shown in FIG5 , the method steps are:
[0035] Step 1, data collection:
[0036] (1) The Box-Behken module in Design-Expert 12 software was used to perform response surface analysis based on the orthogonal test results. The three factors that have the greatest impact on the compressive strength of the specimens were selected as the investigation factors: calcium source type, calcium source addition amount, and moisture content. The data were exported using origin software.
[0037] (2) A total of 1024 data points were collected, each of which included two input variables (calcium source addition amount and moisture content) and one output variable (unconfined compressive strength).
[0038] The second step is data standardization:
[0039] (1) Use the pandas library to read the response surface data in CSV format.
[0040] (2) Use the StandardScaler object to calculate the mean and standard deviation of the input data and output data.
[0041] (3) The calcium source addition amount and moisture content were standardized to satisfy the standard normal distribution (mean = 0, standard deviation = 1).
[0042] (4) The output data (unconfined compressive strength) is converted into a two-dimensional array and then standardized.
[0043] The normalization formula is shown in Equation 1 below:
[0044]
[0045] Among them, x is the original data, μ is the mean of the data, σ is the standard deviation of the data, and x′ is the standardized data.
[0046] The third step is data division:
[0047] Use the train_test_split function in the sklearn.model_selection module to divide the dataset into training and test sets with a ratio of 80% training set and 20% test set.
[0048] Step 4: Machine learning model construction:
[0049] (1) Select the LSTM (Long Short-Term Memory) neural network model to process the data, which can capture the long-term dependencies in time series data and is suitable for processing engineering data with time series characteristics.
[0050] (2) In the model construction, the input layer is the standardized calcium source addition amount and water content data; the LSTM layer is an LSTM layer with 10 hidden units; the fully connected layer 1 connects the output of the LSTM layer to a fully connected layer containing 10 neurons; the fully connected layer 2 outputs the prediction result through a fully connected layer containing 1 neuron.
[0051] Step 5: Machine learning model training:
[0052] (1) During the model training process, the mean square error (MSE) is used as the loss function, the Adam optimizer is selected as the optimizer, and the learning rate is 0.01. The error changes during the training process are as follows: Figure 1 shown.
[0053] (2) The definition of the MSE loss function is as shown in Equation 2:
[0054]
[0055] Where n is the number of samples, y i is the ith actual value, is the i-th predicted value.
[0056] (3) Use the ReLU activation function to enhance the nonlinear representation ability of the model. The design formula of the ReLU activation function is shown in Equation 3:
[0057] ReLU(x)=max(0,x)#(3)
[0058] Step 6: Evaluation of machine learning models:
[0059] (1) Using R 2and mean absolute error (AAD) as evaluation indicators. The number of hidden neurons in the model training process has an impact on R 2 The impact of AAD Figure 2 As shown, the scatter plots of test and training are as follows Figure 3 shown.
[0060] (2)R 2 The formula is shown in equation 4 below:
[0061]
[0062] Where n is the number of samples, y i is the ith actual value, is the ith predicted value, is the mean of the observations.
[0063] (3) The calculation formula of AAD is shown in the following formula 5:
[0064]
[0065] Where n is the number of samples, y i is the ith actual value, is the i-th predicted value.
[0066] Step 7: Genetic algorithm optimization:
[0067] (1) Randomly generate individuals in the initial population in the range of 0-30, use tournament selection to select individuals with better performance as parents, and introduce new gene combinations and random changes through Blend crossover (cxBlend) and Gaussian mutation (mutGaussian) to increase the diversity of the population.
[0068] (2) The initial population is defined to contain 50 individuals. The population is continuously evolved through selection, crossover, and mutation. After 40 generations of iteration, an optimal combination of calcium source content and water content is found. The fitness curve of the evolution process is as follows: Figure 4 shown.
[0069] Step 8, GUI user page:
[0070] (1) Using the modern appearance options provided by the customtkinter library, the system appearance mode and blue theme were created.
[0071] (2) A variety of custom controls are used, including frames (CTkFrame), labels (CTkLabel), input boxes (CTkEntry), and buttons (CTkButton) for page layout.
[0072] (3) The predicted results are inversely normalized to obtain the actual value, which is then compared with the median of the output data to determine the compressive strength.
[0073] It should be noted that although the embodiments of the present invention have been shown and described, for ordinary technicians in this field, without departing from the purpose of the invention, structural methods and embodiments similar to the technical solution are designed without creativity and should all fall within the scope of protection of the present invention.
Claims
1. A method for predicting the compressive strength of cyanide tailings cemented by Bacillus colloids, characterized in that: The following steps are involved: (1) Share the response surface methodology data, including the calcium source type, calcium source addition amount, bacterial powder addition amount, and moisture content of the cemented specimens, and establish a data set for the machine learning model; (2) Standardize the data set to obtain a preprocessed sample data set; (3) randomly dividing the pre-processed sample data set into a training set and a test set, and using a machine learning model to predict the compressive strength prediction model of Bacillus colloids cemented cyanidation tailings; (4) Using the optimization algorithm, find the optimal parameters of the compressive strength of the cyanide tailings cemented by Bacillus sp. (5) A visualization page is formed according to the prediction model for the compressive strength of Bacillus cyanidation tailings cemented by Bacillus cyanidation tailings to develop a prediction system for the compressive strength of Bacillus cyanidation tailings cemented by Bacillus cyanidation tailings.
2. A method for predicting the compressive strength of cyanide tailings cemented by Bacillus colloids according to claim 1, characterized in that: In the step (1), response surface analysis is performed on the basis of orthogonal test results and three test factors with the greatest influence on the compressive strength of the sample are selected: calcium source type, calcium source addition amount and water content as the investigation factors.
3. A method for predicting the compressive strength of cyanide tailings cemented by Bacillus colloids according to claim 2, characterized in that: In the step (1), according to the response surface test results, the order of influence of each test factor on the unconfined compressive strength of the cemented specimen is: calcium source type > calcium source addition amount > water content.
4. A method for predicting the compressive strength of cyanide tailings cemented by Bacillus colloids according to claim 3, characterized in that: In the step (1), the response surface methodology and the machine learning model comprehensively consider the solidification performance of the cyanide tailings cemented by Bacillus subtilis and the cost control of the technical implementation, and calcium lactate is used as the cementing calcium source.
5. A method for predicting the compressive strength of cyanide tailings cemented by Bacillus colloids according to claim 4, characterized in that: In the step (1), the data set defaults to calcium lactate as the calcium source type, the calcium source addition amount and water content are used as input variables, and the compressive strength is used as the output variable. The data is read using software, and a data set for the machine learning model is established.
6. The method for predicting the compressive strength of cyanide tailings cemented by Bacillus colloids according to claim 5, characterized in that: In the step (2), the data set is standardized, and the input data: calcium source addition amount and moisture content, and the output data: compressive strength are converted into a standard normal distribution to ensure that each variable is within the same scale range.
7. The method for predicting the compressive strength of cyanide tailings cemented by Bacillus colloids according to claim 6, characterized in that: In the step (3), the standardized data set is divided into 80% training set and 20% test set, and a long short-term memory neural network algorithm model is adopted.
8. The method for predicting the compressive strength of cyanide tailings cemented by Bacillus colloids according to claim 7, characterized in that: In step (3), the trained LSTM model needs to be evaluated by using the coefficient of determination (R 2 ), mean absolute error (MAE) and / or average absolute deviation (AAD) indicators to measure the prediction performance of the model.
9. The method for predicting the compressive strength of cyanide tailings cemented by Bacillus colloids according to claim 1, characterized in that: In the step (4), a genetic algorithm based on natural selection and genetic mechanism is used to optimize parameters.
10. The method for predicting the compressive strength of cyanide tailings cemented by Bacillus mucilaginosus according to claim 1, characterized in that: In the step (5), a GUI visual user page is used to realize the rapid prediction and intuitive display of the compressive strength of the cyanidation tailings cemented by Bacillus subtilis.