Hydrocyclone wall surface wear prediction method based on CFD and machine learning

Abstract

The invention provides a hydrocyclone wall surface wear prediction method and system based on CFD and machine learning, and the method comprises the steps: obtaining wear state indication data through a wear sensor; constructing a three-dimensional model of the hydrocyclone, and carrying out gridding treatment on the constructed hydrocyclone; constructing a control equation and a hydrocyclone fluid motion equation based on data extracted by the model after gridding processing, and simulating and solving the three-dimensional simulation model based on the control equation and the hydrocyclone fluid motion equation; and constructing a feature project based on data obtained after simulation solution, training a machine learning model based on the constructed feature project, and performing real-time and future wall surface wear prediction by using the trained model. The method can be used for predicting the wear condition of the wall surface of the hydrocyclone and guiding the industrial production to replace equipment in time so as to improve the production efficiency.

Description

Technical Field

[0001] The present invention relates to a technical field, and more particularly to a method for predicting hydrocyclone wall wear based on CFD and machine learning. Background Art

[0002] Hydrocyclone is one of the most important classification equipment in industrial production. It has the characteristics of small size, low cost and high separation efficiency. It is widely used in metallurgy, chemical industry, petroleum and other industries. The medium in the hydrocyclone makes a strong rotation under the action of the centrifugal field, causing impact and friction on the wall of the hydrocyclone, resulting in a certain degree of wall wear. The wear problem is the main failure form of the hydrocyclone, which will lead to the decline of equipment performance, increase maintenance costs, and even affect production safety. At the same time, its sorting efficiency will also be reduced, resulting in serious waste of resources and bringing serious economic losses to industrial production.

[0003] In terms of monitoring and maintenance of hydrocyclone wall wear, traditional methods usually rely on the subjective judgment and experience of operators, which makes wear prediction less accurate and easily affected by operator skills and subjective factors. At the same time, regular inspections are often not enough to capture rapid changes in wear, resulting in wear problems being ignored or discovered at a late stage, increasing maintenance costs. Moreover, traditional methods cannot provide real-time wall wear prediction, which is very important for avoiding equipment damage and improving production efficiency. Summary of the invention

[0004] In view of the above situation, the main purpose of the present invention is to propose a hydrocyclone wall wear prediction method based on CFD and machine learning to solve the above technical problems.

[0005] The present invention provides a method for predicting hydrocyclone wall wear based on CFD and machine learning, the method comprising the following steps: Step 1, obtaining data indicating wear status; A three-dimensional model of the hydrocyclone is constructed based on the wear state indicator data, and an inlet, a boundary layer, and a transition layer of the three-dimensional model of the hydrocyclone are meshed to obtain a meshed three-dimensional model of the hydrocyclone; Step 2: using the data derived from the gridded three-dimensional model of the hydrocyclone, and using the derived data to construct the control equation and the hydrocyclone fluid motion equation; The three-dimensional simulation model is simulated and solved based on the control equation and the fluid motion equation in the hydrocyclone to obtain solution data; the solution data and the wear state indication data are used as training data; Step 3: Based on the data obtained after the simulation solution, feature engineering is constructed in combination with the data collected by the wear sensor. Based on the feature engineering, the machine learning model is trained in combination with the training data to obtain a trained model; Use the trained model to predict real-time and future wall wear.

[0006] The present invention also proposes a hydrocyclone wall wear prediction system based on CFD and machine learning, the system comprising: Model building and mesh processing modules for: Acquiring data indicating wear status; A three-dimensional model of the hydrocyclone is constructed based on the wear state indicator data, and an inlet, a boundary layer, and a transition layer of the three-dimensional model of the hydrocyclone are meshed to obtain a meshed three-dimensional model of the hydrocyclone; Equation building blocks for: Using the data derived from the gridded three-dimensional model of the hydrocyclone, the control equations and the hydrocyclone fluid motion equations are constructed using the derived data; The three-dimensional simulation model is simulated and solved based on the control equation and the fluid motion equation in the hydrocyclone to obtain solution data; the solution data and the wear state indication data are used as training data; Learning model training module for: Based on the data obtained after simulation and solution, feature engineering is constructed in combination with the data collected by the wear sensor. Based on feature engineering, the machine learning model is trained in combination with the training data to obtain a trained model. Use the trained model to predict real-time and future wall wear.

[0007] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention combines computational fluid dynamics (CFD) and machine learning techniques, uses wear sensor data and physical models, and comprehensively considers the impact of multiple factors on hydrocyclone wall wear. This comprehensive approach can improve prediction accuracy, thereby effectively guiding equipment maintenance and replacement decisions in industrial production.

[0008] 2. Compared with traditional detection methods, the present invention has higher efficiency and accuracy. By predicting the wall wear of the hydrocyclone, effective decision support can be provided for industrial production. Timely replacement of damaged equipment can avoid production interruptions and quality problems, thereby improving overall production efficiency and product quality.

[0009] 3. Since this method is based on physical models and machine learning models, it can be applied to hydrocyclones of different types and sizes, has certain transferability, and can be applied to a variety of other industrial fields.

[0010] 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

[0011] Figure 1 A flow chart of a method for predicting hydrocyclone wall wear based on CFD and machine learning proposed by the present invention; Figure 2 A schematic diagram of the installation of a wear sensor for a hydrocyclone wall wear prediction method based on CFD and machine learning proposed by the present invention; Figure 3 This is the overall framework of a hydrocyclone wall wear prediction system based on CFD and machine learning proposed in the present invention. DETAILED DESCRIPTION

[0012] 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.

[0013] 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.

[0014] See also Figure 1 and Figure 2 The embodiment of the present invention provides a method for predicting the wall wear of a hydrocyclone based on CFD and machine learning, the method comprising the following steps: Step 1, obtaining data indicating wear status; A three-dimensional model of the hydrocyclone is constructed based on the wear state indicator data, and an inlet, a boundary layer, and a transition layer of the three-dimensional model of the hydrocyclone are meshed to obtain a meshed three-dimensional model of the hydrocyclone; In step 1, a slight change in the wear of the wall will also cause a change in resistance. Therefore, the wear sensor uses the change in resistance to quantitatively represent the state of wall wear. The installation positions of the wear sensor are at the inlet, column section and cone section of the hydrocyclone. The specific steps for meshing the three-dimensional simulation model are as follows: Mesh the 3D simulation model to generate unstructured tetrahedral and hexahedral meshes; The inlet area, transition area and boundary layer area in the three-dimensional simulation model are meshed and refined.

[0015] In this step, the parameters collected and processed by the wear sensor are: Resistance value: As the degree of wall wear increases, the resistance of the wall will change; Resistance change rate: The wear rate of the wall surface can be estimated by continuously detecting the change of resistance value and calculating the resistance change rate; Voltage output: Resistance sensors usually detect the resistance change of the wall caused by wear through voltage output, which is used to analyze and predict the wear state of the wall; Temperature change: The wall surface will cause local temperature rise during the wear process. The resistance sensor detects the change of resistance in the temperature rise area and indirectly detects the temperature change. The temperature change is used as one of the indicators for detecting wear. Resistance mapping: Multiple resistance sensors are installed at different locations on the wall to create a three-dimensional resistance monitoring system to identify the distribution of wall wear and predict the wear rate in different areas.

[0016] Step 2: using the data derived from the gridded three-dimensional model of the hydrocyclone, and using the derived data to construct the control equation and the hydrocyclone fluid motion equation; The three-dimensional simulation model is simulated and solved based on the control equation and the fluid motion equation in the hydrocyclone to obtain solution data; the solution data and the wear state indication data are used as training data; In step 2, the data derived from the gridded three-dimensional model of the hydrocyclone are used to construct the control equation and the hydrocyclone fluid motion equation; wherein the control equation includes the mass conservation equation, the momentum conservation equation and the turbulent energy equation, and the expression of the mass conservation equation is: ; in, is the fluid density, is the fluid velocity with x-, y- and z-direction components, is the Hamiltonian operator, is the sign of the partial derivative of density with respect to time; The expression of the momentum conservation equation is: ; in, is the static pressure, is gravity, For external force, is the stress tensor; The expression of the turbulent energy equation is: ; in, is the empirical coefficient, is the turbulent kinetic energy, is the turbulent dissipation rate, is the turbulent Prandtl number of phase k, is the turbulent kinetic energy caused by the mean velocity gradient, is the turbulent kinetic energy caused by buoyancy, is the contribution of wave expansion in compressible turbulence to the overall dissipation rate, is the default value, is the dynamic viscosity, is the turbulent viscosity, For the case where the turbulent kinetic energy is transported along with the movement of the fluid in the i direction, is the diffusion rate of turbulent kinetic energy in the j direction, is the average velocity of the fluid in the i direction; The data derived from the gridded hydrocyclone three-dimensional model are used to construct the control equation and the hydrocyclone fluid motion equation; the hydrocyclone fluid motion equation includes the continuity equation of the mixture, and the expression of the continuity equation of the mixture is as follows: ; ; in, is the mass-average velocity, is the density of the mixture, is the volume fraction of k phase, is the mass-average velocity of phase k, is the mixture density of phase k, is a mixed term, is the phase number, F is the force, is the viscosity of the mixture, is the recommended speed of the secondary phase k, T is the transposed sign; The control equation expression is: ;

[0017] in, and is the water speed, and is the fluctuating water velocity calculated by the equation, i, j = 1, 2, 3, , and are Cartesian coordinate components, i, j, k = 1, 2, 3, p is the static pressure, is the viscosity of water, is the convective transport term, is the turbulent diffusion term, is the fluctuating static pressure, is the stress generating term, is the pressure strain term, and All are Kronecker's second-order tensor identifiers; In this step, the three-dimensional simulation model is simulated and solved based on the control equation and the fluid motion equation in the hydrocyclone. The steps are as follows: Determine reaction mechanisms based on fluid motion in 3D simulation models; Set the inlet flow rate and add the fluid phase, air phase, and particle phase as initial conditions; After setting the total time step, the three-dimensional simulation model is solved based on the set initial conditions and the total time step. The solution data includes: the wall velocity gradient, pressure distribution, temperature distribution, turbulence intensity, etc. of the hydrocyclone during the movement of the mixed phase fluid.

[0018] Step 3: Based on the data obtained after the simulation solution, feature engineering is constructed in combination with the data collected by the wear sensor. Based on the feature engineering, the machine learning model is trained in combination with the training data to obtain a trained model; Use the trained model to predict real-time and future wall wear; In step 3, feature engineering is constructed based on feature data extracted from the solved data and the data collected by the wear sensor, and the machine learning model is trained using feature engineering. The specific steps for training the machine learning model using feature engineering are as follows: Extract data from feature engineering, clean and process the data, and obtain processed data; Select the model to use for training; Divide the processed data and training data into training set, test set and validation set; Use the training set to train the learning model and continuously adjust the model parameters; Use the validation set to evaluate the performance of the model and adjust the model’s hyperparameters as needed; Use the test set to validate the model and evaluate its performance; The verification is passed, the training is completed, and the trained model is obtained; Finally, the trained and verified model is used to predict real-time and future wall wear.

[0019] In this step, the data in feature engineering includes: resistance difference, pressure distribution, turbulence distribution, temperature distribution, flow velocity gradient, inlet flow, timestamp and other related data.

[0020] In this step, the machine learning model includes one or more combinations of a random forest regression algorithm, a decision tree regression algorithm, a support vector machine regression algorithm, and a time series prediction model algorithm.

[0021] See also Figure 3 The embodiment of the present invention also provides a hydrocyclone wall wear prediction system based on CFD and machine learning, the system comprising: Model building and mesh processing modules for: Acquiring data indicating wear status; A three-dimensional model of the hydrocyclone is constructed based on the wear state indicator data, and an inlet, a boundary layer, and a transition layer of the three-dimensional model of the hydrocyclone are meshed to obtain a meshed three-dimensional model of the hydrocyclone; Equation building blocks for: Using the data derived from the gridded three-dimensional model of the hydrocyclone, the control equations and the hydrocyclone fluid motion equations are constructed using the derived data; The three-dimensional simulation model is simulated and solved based on the control equation and the fluid motion equation in the hydrocyclone to obtain solution data; the solution data and the wear state indication data are used as training data; Learning model training module for: Based on the data obtained after simulation and solution, feature engineering is constructed in combination with the data collected by the wear sensor. Based on feature engineering, the machine learning model is trained in combination with the training data to obtain a trained model. Use the trained model to predict real-time and future wall wear.

[0022] 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, multiple 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.

[0023] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", 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 can be combined in any one or more embodiments or examples in a suitable manner.

[0024] The above-mentioned embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it 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, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. A method for predicting hydrocyclone wall wear based on CFD and machine learning, characterized in that: The method comprises the following steps: Step 1, obtaining data indicating wear status; A three-dimensional model of the hydrocyclone is constructed based on the wear state indicator data, and an inlet, a boundary layer, and a transition layer of the three-dimensional model of the hydrocyclone are meshed to obtain a meshed three-dimensional model of the hydrocyclone; Step 2: using the data derived from the gridded three-dimensional model of the hydrocyclone, and using the derived data to construct the control equation and the hydrocyclone fluid motion equation; The three-dimensional simulation model is simulated and solved based on the control equation and the fluid motion equation in the hydrocyclone to obtain solution data; the solution data and the wear state indication data are used as training data; Step 3: Based on the data obtained after the simulation solution, feature engineering is constructed in combination with the data collected by the wear sensor. Based on the feature engineering, the machine learning model is trained in combination with the training data to obtain a trained model; Use the trained model to predict real-time and future wall wear.

2. The method for predicting hydrocyclone wall wear based on CFD and machine learning according to claim 1, characterized in that: In step 1, the specific steps of meshing the constructed hydrocyclone three-dimensional simulation model are as follows: Mesh the 3D simulation model to generate unstructured tetrahedral and hexahedral meshes; The inlet area, transition area and boundary layer area in the three-dimensional simulation model are meshed and refined.

3. The method for predicting hydrocyclone wall wear based on CFD and machine learning according to claim 2, characterized in that: In step 2, the data derived from the gridded three-dimensional model of the hydrocyclone is used to construct the control equation and the hydrocyclone fluid motion equation; wherein the control equation includes the mass conservation equation, the momentum conservation equation and the turbulent energy equation, and the expression of the mass conservation equation is: ； in, is the fluid density, is the fluid velocity with x-, y- and z-direction components, is the Hamiltonian operator, is the sign of the partial derivative of density with respect to time; The expression of the momentum conservation equation is: ； in, is the static pressure, is gravity, For external force, is the stress tensor.

4. The method for predicting hydrocyclone wall wear based on CFD and machine learning according to claim 3 is characterized in that: In step 2, the data derived from the gridded three-dimensional model of the hydrocyclone is used to construct the control equation and the hydrocyclone fluid motion equation, wherein the control equation includes the mass conservation equation, the momentum conservation equation and the turbulent energy equation, and the expression of the turbulent energy equation is: ； in, is the empirical coefficient, is the turbulent kinetic energy, is the turbulent dissipation rate, is the turbulent Prandtl number of phase k, is the turbulent kinetic energy caused by the mean velocity gradient, is the turbulent kinetic energy caused by buoyancy, is the contribution of wave expansion in compressible turbulence to the overall dissipation rate, is the default value, is the dynamic viscosity, is the turbulent viscosity, For the case where the turbulent kinetic energy is transported along with the movement of the fluid in the i direction, is the diffusion rate of turbulent kinetic energy in the j direction, is the average velocity of the fluid in the i direction.

5. According to the method for predicting the wall wear of a hydrocyclone based on CFD and machine learning in claim 4, in step 2, the data derived from the three-dimensional model of the hydrocyclone after gridding is used, and the control equation and the fluid motion equation of the hydrocyclone are constructed using the derived data, wherein: The fluid motion equation of the hydrocyclone includes the continuity equation of the mixture. The expression of the continuity equation of the mixture is as follows: ； ； in, is the mass-average velocity, is the density of the mixture, is the volume fraction of k phase, is the mass-average velocity of phase k, is the mixture density of phase k, is a mixed term, is the phase number, F is the force, is the viscosity of the mixture, is the recommended speed of the secondary phase k, and T is the transposed sign.

6. The method for predicting hydrocyclone wall wear based on CFD and machine learning according to claim 5, characterized in that: In step 2, the data derived from the gridded three-dimensional model of the hydrocyclone is used to construct the control equation and the hydrocyclone fluid motion equation; wherein the control equation is expressed as: ； ； in, and is the water speed, and is the fluctuating water velocity calculated from Eq. , and are Cartesian coordinate components, p is the static pressure, is the viscosity of water, is the convective transport term, is the turbulent diffusion term, is the fluctuating static pressure, is the stress generating term, is the pressure strain term, and Both are Kronecker's second-order tensor identifiers.

7. The method for predicting hydrocyclone wall wear based on CFD and machine learning according to claim 6, characterized in that: In step 2, the three-dimensional simulation model is simulated and solved based on the control equation and the fluid motion equation in the hydrocyclone, and the steps are as follows: Determine reaction mechanisms based on fluid motion in 3D simulation models; Set the inlet flow rate and add the fluid phase, air phase, and particle phase as initial conditions; After setting the total time step, the three-dimensional simulation model is solved based on the set initial conditions and total time step, and using the control equations and the fluid motion equations in the hydrocyclone.

8. The method for predicting hydrocyclone wall wear based on CFD and machine learning according to claim 7, characterized in that: In step 3, based on the data obtained after the simulation solution, feature engineering is constructed in combination with the data collected by the wear sensor. Based on the feature engineering, the machine learning model is trained in combination with the training data to obtain a trained model. The specific steps of training the machine learning model using feature engineering are as follows: Extract data from feature engineering, clean and process the data, and obtain processed data; Select the model to use for training; Divide the processed data and training data into training set, test set and validation set; Use the training set to train the learning model and continuously adjust the model parameters; Use the validation set to evaluate the performance of the model and adjust the model’s hyperparameters as needed; Use the test set to validate the model and evaluate its performance; The model verification is passed, the training is completed, and the trained model is obtained.

9. A hydrocyclone wall wear prediction system based on CFD and machine learning, characterized in that: The system applies a hydrocyclone wall wear prediction method based on CFD and machine learning as described in any one of claims 1 to 8 above, and the system comprises: Model building and mesh processing modules for: Acquiring data indicating wear status; A three-dimensional model of the hydrocyclone is constructed based on the wear state indicator data, and an inlet, a boundary layer, and a transition layer of the three-dimensional model of the hydrocyclone are meshed to obtain a meshed three-dimensional model of the hydrocyclone; Equation building blocks for: Using the data derived from the gridded three-dimensional model of the hydrocyclone, the control equations and the hydrocyclone fluid motion equations are constructed using the derived data; The three-dimensional simulation model is simulated and solved based on the control equation and the fluid motion equation in the hydrocyclone to obtain solution data; the solution data and the wear state indication data are used as training data; Learning model training module for: Based on the data obtained after simulation and solution, feature engineering is constructed in combination with the data collected by the wear sensor. Based on feature engineering, the machine learning model is trained in combination with the training data to obtain a trained model. Use the trained model to predict real-time and future wall wear.

Images