Blast furnace tuyere rotation area depth soft measurement method based on machine learning

Abstract

The invention provides a machine learning-based deep soft measurement method for a blast-furnace tuyere convolute area, which comprises the following steps of: constructing a three-dimensional simulation model of the blast-furnace tuyere convolute area by utilizing a fluid phase control equation and a particle motion control equation, and acquiring a representative temperature cloud picture of the blast-furnace tuyere convolute area in a simulation result by utilizing the three-dimensional simulation model; carrying out feature centralization on representative temperature cloud picture data obtained in a simulation result of the three-dimensional simulation model, substituting the obtained data after the feature centralization into the covariance matrix, carrying out eigenvalue decomposition on the covariance matrix by utilizing an eigenvalue decomposition formula to obtain eigenvalues and eigenvectors, and calculating the eigenvalues and the eigenvectors according to the eigenvalues and the eigenvectors; performing dimension reduction on the high-dimensional temperature cloud atlas according to the feature values and the feature vectors; on the basis of the feature data obtained by dimension reduction, constructing a strong regression model by using an AdaBoost algorithm in a multi-iteration training mode; and carrying out regression prediction on the depth of the blast-furnace tuyere convolution area by using a strong regression model constructed based on the feature data and an AdaBoost algorithm. According to the method, the strong regression model is constructed by using the AdaBoost algorithm, so that the workload of calculation is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of machine learning and computer numerical simulation, and in particular to a soft measurement method for blast furnace tuyere raceway depth based on machine learning. Background Art

[0002] Modern blast furnaces can be regarded as high-temperature and high-pressure multiphase reactors. In modern blast furnaces, it can be regarded as a high-temperature and high-pressure multiphase reactor, which involves reverse, co-current and cross-flow of gas, solid and liquid materials, accompanied by complex heat and mass transfer processes, chemical reactions and strong interactions. The interior of the blast furnace can be divided into several different physical zones from top to bottom according to the state and flow properties of the materials, including the block zone, soft melting zone, dripping zone, dead material column, tuyere whirlpool zone and hearth zone. Among the above-mentioned zones, the tuyere whirlpool zone located in the lower part of the blast furnace is particularly important. The mechanism of the formation of the tuyere whirlpool zone at the front edge of the blast furnace tuyere has always been valued by ironmaking workers, and they have conducted detailed research in many aspects, forming many theories and mathematical models on the formation mode, shape characteristics, heat transfer and mass transfer of the tuyere whirlpool zone.

[0003] At present, numerical simulation technology based on CFD-DEM coupling can evaluate the gas-solid flow effect by comparing different blast speeds and bed widths, and provide theoretical support for the optimization control of the actual vortex zone morphology. However, the calculation cost is high, and it usually takes several days or even weeks to establish and calculate a CFD-DEM model of a tuyere vortex zone.

[0004] At the same time, the feasibility of numerical simulation technology based on CFD-DEM coupling is limited by the quality and availability of computing resources, and a large amount of computing resources are required to perform simulations, which poses a challenge to researchers and engineers. Summary of the invention

[0005] In view of the above situation, the main purpose of the present invention is to propose a soft measurement method for the depth of blast furnace tuyere raceway based on machine learning to solve the above technical problems.

[0006] The present invention provides a blast furnace tuyere raceway depth soft measurement method based on machine learning, the method comprising the following steps:

[0007] Step 1: construct a three-dimensional simulation model of the blast furnace tuyere raceway based on the parameters and using the fluid phase control equation and the particle motion control equation, perform simulation and solve the fluid phase control equation and the particle motion control equation to obtain simulation results, and collect a representative temperature cloud map of the blast furnace tuyere raceway in the simulation results;

[0008] Step 2: Feature centering is performed on the representative temperature cloud map data obtained from the simulation results of the three-dimensional simulation model, and the obtained feature-centered data is substituted into the covariance matrix. The covariance matrix is ​​eigenvalue decomposed using the eigenvalue decomposition formula to obtain eigenvalues ​​and eigenvectors, and the high-dimensional temperature cloud map is reduced in dimension according to the eigenvalues ​​and eigenvectors;

[0009] Step 3: Based on the feature data obtained by dimensionality reduction, a strong regression model is constructed by using the AdaBoost algorithm through multiple iterative training;

[0010] Step 4: Use the strong regression model constructed based on feature data and AdaBoost algorithm to perform regression prediction on the depth of the blast furnace tuyere raceway zone.

[0011] The present invention also proposes a blast furnace tuyere raceway depth soft measurement method based on machine learning, the system comprising:

[0012] High-dimensional temperature cloud map building module:

[0013] A three-dimensional simulation model of the blast furnace tuyere raceway is constructed based on the parameters and by using the fluid phase control equation and the particle motion control equation, and simulation results are obtained by simulating and solving the fluid phase control equation and the particle motion control equation, and representative temperature cloud maps of the blast furnace tuyere raceway in the simulation results are collected;

[0014] High-dimensional temperature cloud image dimension reduction module:

[0015] Based on the representative temperature cloud map data obtained from the simulation results of the three-dimensional simulation model, feature centering is performed, the obtained feature-centered data is substituted into the covariance matrix, and the covariance matrix is ​​eigenvalue decomposed using the eigenvalue decomposition formula to obtain eigenvalues ​​and eigenvectors, and the high-dimensional temperature cloud map is reduced in dimension according to the eigenvalues ​​and eigenvectors;

[0016] Strong regression model building blocks:

[0017] The feature data obtained by dimensionality reduction is divided into a training data set and a test data set, a strong regression model is constructed based on the AdaBoost algorithm, and the data in the training data set is input into the strong regression model for iterative training;

[0018] Result prediction module:

[0019] The data in the test data set are input into the trained strong regression model to obtain the depth of the blast furnace tuyere raceway.

[0020] Compared with the prior art, the present invention has the following beneficial effects:

[0021] 1. The present invention utilizes the AdaBoost algorithm and constructs a strong regression model based on feature data. The algorithm trains a series of weak regressors in an iterative manner, adjusts the sample weight according to the performance of each weak regressor, and constructs a strong regressor based on the weight of the weak regressor and the weak regressor. Finally, the strong regression model constructed by the feature data and the AdaBoost algorithm is used to perform regression prediction on the depth of the blast furnace tuyere raceway. The AdaBoost algorithm itself does not require special selection or preprocessing of feature feature data, so it can be directly applied to feature data, reducing the workload of calculation.

[0022] 2. The present invention uses the AdaBoost algorithm. In the AdaBoost algorithm, the weights are iteratively trained using a normalization factor for normalizing the weights, and the sample weights are updated. The AdaBoost algorithm does not require a complex parameter adjustment process, only a weak regressor and the number of iterative training times need to be selected. The theory is relatively simple and relatively easy to implement.

[0023] Additional aspects and advantages of the present invention will be given in part in the following description and in part will be obvious from the following description or learned through embodiments of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 A flow chart of a method for soft-sensing the depth of a blast furnace tuyere raceway based on machine learning proposed by the present invention;

[0025] Figure 2 A representative temperature cloud diagram of a blast furnace tuyere raceway depth soft measurement method based on machine learning proposed by the present invention;

[0026] Figure 3 This is the overall framework diagram of a blast furnace tuyere raceway depth soft measurement method system based on machine learning proposed in the present invention. DETAILED DESCRIPTION

[0027] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be understood as limiting the present invention.

[0028] These and other aspects of the embodiments of the present invention will be apparent with reference to the following description and accompanying drawings. In these descriptions and accompanying drawings, some specific implementations of the embodiments of the present invention are specifically disclosed to represent some ways of implementing the principles of the embodiments of the present invention, but it should be understood that the scope of the embodiments of the present invention is not limited thereto.

[0029] See also Figure 1and Figure 2 , an embodiment of the present invention provides a blast furnace tuyere raceway depth soft measurement method based on machine learning, the method comprising the following steps:

[0030] Step 1: Based on the parameters and using the fluid phase control equation and the particle motion control equation, a three-dimensional simulation model of the blast furnace tuyere raceway is constructed, and simulation results are obtained by simulating and solving the fluid phase control equation and the particle motion control equation, and a representative temperature cloud map of the blast furnace tuyere raceway in the simulation results is collected. The method for obtaining the three-dimensional simulation model specifically includes the following steps:

[0031] In step 1, a three-dimensional simulation model of the blast furnace tuyere raceway is constructed based on parameters and using the fluid phase control equation and the particle motion control equation, and simulation results are obtained by simulating and solving the fluid phase control equation and the particle motion control equation, and a representative temperature cloud map of the blast furnace tuyere raceway in the simulation results is collected. The method for obtaining the three-dimensional simulation model specifically includes the following steps:

[0032] The fluid phase control equations include the continuity equation and the momentum equation. The continuity equation is constructed using parameters and is expressed as:

[0033]

[0034] Among them, ρ f is the gas density, ε f is the local porosity of the gas, u is the superficial velocity of the gas, is the Hamiltonian operator;

[0035] Using the parameters to construct the momentum equation, the momentum equation is expressed as:

[0036]

[0037] Where p is the gas pressure, F fp is the interaction force between gas particles per unit volume, g is the gravitational acceleration, T is the gas phase stress tensor, and t is the time step;

[0038] The particle motion control equation is constructed using the parameters and is expressed as:

[0039]

[0040] Where d is the diameter of the particle, m i is the mass of the ith particle, v i is the horizontal velocity of the ith particle, f e，ij is the elastic force between the i-th particle and the j-th particle, f d，ij is the viscous damping force between the i-th particle and the j-th particle, m ig is the gravity of the ith particle, I i is the moment of inertia of the ith particle, ω i is the rotation speed of the ith particle, T t，ij is the tangential moment between the i-th particle and the j-th particle, T r，ij is the rolling friction torque between the i-th particle and the j-th particle;

[0041] The three-dimensional simulation model constructed by the fluid phase control equation and the particle motion control equation is used to perform simulation calculations to obtain data, a temperature cloud map is made based on the calculated data, and a representative temperature cloud map when the blast furnace tuyere raceway zone is formed is selected.

[0042] Step 2: Feature centering is performed on the representative temperature cloud map data obtained from the simulation results of the three-dimensional simulation model, and the obtained feature-centered data is substituted into the covariance matrix. The covariance matrix is ​​eigenvalue decomposed using the eigenvalue decomposition formula to obtain eigenvalues ​​and eigenvectors, and the high-dimensional temperature cloud map is reduced in dimension according to the eigenvalues ​​and eigenvectors;

[0043] In step 2, feature centralization is performed based on representative temperature cloud map data obtained from the simulation results of the three-dimensional simulation model. The feature centralization is expressed as:

[0044] X_centered = X-μ;

[0045] Among them, X is the original data matrix, μ is the mean vector of each feature, and X_centered is the data after feature centering;

[0046] Substitute the obtained feature-centered data into the covariance matrix, and the covariance calculation equation is:

[0047]

[0048] Among them, X_centered[k] is the data after the k-th feature is centered, and X_centered[l] is the data after the l-th feature is centered;

[0049] The eigenvalue decomposition formula is used to decompose the covariance matrix to obtain the eigenvalues ​​and eigenvectors, and the high-dimensional temperature cloud map is reduced in dimension according to the eigenvalues ​​and eigenvectors:

[0050] The eigenvalue decomposition formula is:

[0051] C*v=λ*v;

[0052] Where C is the covariance matrix, v is the eigenvector, and λ is the eigenvalue;

[0053] According to the size of the eigenvalue, several eigenvectors are selected as principal components, and the data of the high-dimensional temperature cloud map are projected onto the selected principal components to obtain the data after dimensionality reduction, and the data after dimensionality reduction is composed of the principal components.

[0054] Step 3: Divide the feature data obtained by dimensionality reduction into a training data set and a test data set, build a strong regression model based on the AdaBoost algorithm, and input the data in the training data set into the strong regression model for iterative training;

[0055] In step 3, the feature data obtained by dimensionality reduction is divided into a training data set and a test data set, a strong regression model is constructed based on the AdaBoost algorithm, and the data in the training data set is input into the strong regression model for iterative training. The specific steps are as follows:

[0056] Assign equal sample weight to each data sample, and the process has the following relationship:

[0057]

[0058] Among them, w o is the oth sample weight, N is the number of samples;

[0059] The error of the weak regressor is calculated using the sample weights, and the corresponding process has the following relationship:

[0060]

[0061] Among them, h α (·) is the αth weak regressor, x β With y β are the βth sample features in the training data set, and ε is the error of the weak regressor;

[0062] The error of the weak regressor is used to calculate the weight of the weak regressor. The corresponding process has the following relationship:

[0063]

[0064] Among them, ε α is the error of the αth weak regressor, η α represents the weight of the αth weak regressor, and ln is the logarithmic function;

[0065] Based on the weights of the weak regressor, the normalization factor used to normalize the weights is calculated. The corresponding process has the following relationship:

[0066]

[0067] Among them, Z is the normalization factor used to normalize the weights;

[0068] The weights are iteratively trained based on the normalization factor used to normalize the weights, and the sample weights are updated. The corresponding process has the following relationship:

[0069]

[0070] Among them, w o+1 is the o+1th sample weight, exp is the exponential function;

[0071] After training, a strong regressor is constructed based on the weights of the weak regressor and the weak regressor. The corresponding process has the following relationship:

[0072]

[0073] Among them, H(·) is the final strong regressor and T is the total number of weak regressors.

[0074] Step 4: Input the data in the test data set into the trained strong regression model to obtain the depth of the blast furnace tuyere raceway.

[0075] See also Figure 3 , an embodiment of the present invention further provides a blast furnace tuyere raceway depth soft measurement system based on machine learning, the system comprising:

[0076] High-dimensional temperature cloud map building module:

[0077] A three-dimensional simulation model of the blast furnace tuyere raceway is constructed based on the parameters and by using the fluid phase control equation and the particle motion control equation, and simulation results are obtained by simulating and solving the fluid phase control equation and the particle motion control equation, and representative temperature cloud maps of the blast furnace tuyere raceway in the simulation results are collected;

[0078] High-dimensional temperature cloud image dimension reduction module:

[0079] Based on the representative temperature cloud map data obtained from the simulation results of the three-dimensional simulation model, feature centering is performed, the obtained feature-centered data is substituted into the covariance matrix, and the covariance matrix is ​​eigenvalue decomposed using the eigenvalue decomposition formula to obtain eigenvalues ​​and eigenvectors, and the high-dimensional temperature cloud map is reduced in dimension according to the eigenvalues ​​and eigenvectors;

[0080] Strong regression model building blocks:

[0081] The feature data obtained by dimensionality reduction is divided into a training data set and a test data set, a strong regression model is constructed based on the AdaBoost algorithm, and the data in the training data set is input into the strong regression model for iterative training;

[0082] Result prediction module:

[0083] The data in the test data set are input into the trained strong regression model to obtain the depth of the blast furnace tuyere raceway.

[0084] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0085] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0086] The above-mentioned embodiments only express several implementation methods of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention should be based on the attached claims.

Claims

1. A soft measurement method for blast furnace tuyere raceway depth based on machine learning, characterized in that: The method comprises the following steps: Step 1: construct a three-dimensional simulation model of the blast furnace tuyere raceway based on the parameters and using the fluid phase control equation and the particle motion control equation, perform simulation and solve the fluid phase control equation and the particle motion control equation to obtain simulation results, and collect a representative temperature cloud map of the blast furnace tuyere raceway in the simulation results; Step 2: Feature centering is performed on the representative temperature cloud map data obtained from the simulation results of the three-dimensional simulation model, and the obtained feature-centered data is substituted into the covariance matrix. The covariance matrix is ​​eigenvalue decomposed using the eigenvalue decomposition formula to obtain eigenvalues ​​and eigenvectors, and the high-dimensional temperature cloud map is reduced in dimension according to the eigenvalues ​​and eigenvectors; Step 3: Divide the feature data obtained by dimensionality reduction into training data set and test data set based on AdaB oo The st algorithm constructs a strong regression model and inputs the data in the training data set into the strong regression model for iterative training; Step 4: Input the data in the test data set into the trained strong regression model to obtain the depth of the blast furnace tuyere raceway.

2. The method for soft-sensing the depth of the blast furnace tuyere raceway based on machine learning according to claim 1, characterized in that: In the step 1, a three-dimensional simulation model of the blast furnace tuyere raceway is constructed based on the parameters and using the fluid phase control equation and the particle motion control equation, and simulation results are obtained by simulating and solving the fluid phase control equation and the particle motion control equation, and a representative temperature cloud map of the blast furnace tuyere raceway in the simulation results is collected. The method for obtaining the three-dimensional simulation model specifically includes the following steps: The fluid phase control equations include the continuity equation and the momentum equation. The continuity equation is constructed using parameters and is expressed as: Among them, ρ f is the gas density, ε f is the local porosity of the gas, u is the superficial velocity of the gas, is the Hamiltonian operator; Using the parameters to construct the momentum equation, the momentum equation is expressed as: Where p is the gas pressure, F fp is the interaction force between gas particles per unit volume, g is the gravitational acceleration, τ is the gas phase stress tensor, and t is the time step; The particle motion control equation is constructed using the parameters and is expressed as: Where d is the diameter of the particle, m i is the mass of the ith particle, v i is the horizontal velocity of the ith particle, f e，ij is the elastic force between the i-th particle and the j-th particle, f d，ij is the viscous damping force between the i-th particle and the j-th particle, m i g is the gravity of the ith particle, I i is the moment of inertia of the ith particle, ω i is the rotation speed of the ith particle, T t，ij is the tangential moment between the i-th particle and the j-th particle, T r，ij is the rolling friction torque between the i-th particle and the j-th particle; The three-dimensional simulation model constructed by the fluid phase control equation and the particle motion control equation is used to perform simulation calculations to obtain data, a temperature cloud map is made based on the calculated data, and a representative temperature cloud map when the blast furnace tuyere raceway zone is formed is selected.

3. The method for soft-sensing the depth of the blast furnace tuyere raceway based on machine learning according to claim 2 is characterized in that: In step 2, feature centralization is performed based on representative temperature cloud map data obtained from the simulation results of the three-dimensional simulation model. The feature centralization is expressed as: Xcentered=X-μ: Among them, X is the original data matrix, μ is the mean vector of each feature, and X _ Centered is the data after feature centering.

4. The method for soft-sensing the depth of the blast furnace tuyere raceway based on machine learning according to claim 3 is characterized in that: In step 2, the obtained feature-centered data is substituted into the covariance matrix, and the covariance calculation equation is: Among them, X_centered[k] is the data after the kth feature is centered, and X_centered[l] is the data after the lth feature is centered.

5. The method for soft-sensing the depth of the blast furnace tuyere raceway based on machine learning according to claim 4 is characterized in that: In step 2, the covariance matrix is ​​decomposed by eigenvalue decomposition using the eigenvalue decomposition formula to obtain eigenvalues ​​and eigenvectors, and the high-dimensional temperature cloud map is reduced in dimension according to the eigenvalues ​​and eigenvectors: The eigenvalue decomposition formula is: C*V=λ*V: Where C is the covariance matrix, V is the eigenvector, and λ is the eigenvalue; According to the size of the eigenvalue, several eigenvectors are selected as principal components, and the data of the high-dimensional temperature cloud map are projected onto the selected principal components to obtain the data after dimensionality reduction.

6. A blast furnace tuyere raceway depth soft measurement method based on machine learning according to claim 5, characterized in that, in the step 3, the feature data obtained by dimensionality reduction is divided into a training data set and a test data set, a strong regression model is constructed based on the AdaBoost algorithm, and the data in the training set is input into the strong regression model for iterative training, and the specific steps are as follows: Assign equal sample weight to each data sample, and the process has the following relationship: in, w o is the oth sample weight, N is the number of samples; The error of the weak regressor is calculated using the sample weights, and the corresponding process has the following relationship: Among them, h α (·) is the αth weak regressor, x β With y β are the βth sample features in the training data set, and ε is the error of the weak regressor; The error of the weak regressor is used to calculate the weight of the weak regressor. The corresponding process has the following relationship: Among them, ε α is the error of the αth weak regressor, η α represents the weight of the αth weak regressor, and ln is the logarithmic function; Based on the weights of the weak regressor, the normalization factor used to normalize the weights is calculated. The corresponding process has the following relationship: Among them, Z is the normalization factor used to normalize the weights; The weights are iteratively trained based on the normalization factor used to normalize the weights, and the sample weights are updated. The corresponding process has the following relationship: Among them, w o+1 is the o+1th sample weight, exp is the exponential function; After training, a strong regressor is constructed based on the weights of the weak regressor and the weak regressor. The corresponding process has the following relationship: Among them, H(·) is the final strong regressor and T is the total number of weak regressors.

7. The method for soft-sensing the depth of the blast furnace tuyere raceway based on machine learning according to claim 6 is characterized in that: In step 4, the data in the test data set is input into the trained strong regression model to obtain the depth of the blast furnace tuyere raceway.

8. A blast furnace tuyere raceway depth soft measurement system based on machine learning, characterized in that: The system applies a blast furnace tuyere raceway depth soft measurement method based on machine learning as described in any one of claims 1 to 7 above, and the system comprises: High-dimensional temperature cloud map building module: A three-dimensional simulation model of the blast furnace tuyere raceway is constructed based on the parameters and by using the fluid phase control equation and the particle motion control equation, and simulation results are obtained by simulating and solving the fluid phase control equation and the particle motion control equation, and representative temperature cloud maps of the blast furnace tuyere raceway in the simulation results are collected; High-dimensional temperature cloud image dimension reduction module: Based on the representative temperature cloud map data obtained from the simulation results of the three-dimensional simulation model, feature centering is performed, the obtained feature-centered data is substituted into the covariance matrix, and the covariance matrix is ​​eigenvalue decomposed using the eigenvalue decomposition formula to obtain eigenvalues ​​and eigenvectors, and the high-dimensional temperature cloud map is reduced in dimension according to the eigenvalues ​​and eigenvectors; Strong regression model building blocks: The feature data obtained by dimensionality reduction is divided into a training data set and a test data set, a strong regression model is constructed based on the AdaBoost algorithm, and the data in the training data set is input into the strong regression model for iterative training; Result prediction module: The data in the test data set are input into the trained strong regression model to obtain the depth of the blast furnace tuyere raceway.

Images