A method for fault diagnosis of stirred tank impellers based on CFD and deep learning

Abstract

The present invention proposes a method for diagnosing agitator blade faults based on CFD and deep learning. The method comprises: photographing the flow field of the agitator blade at different times under different fracture conditions, processing the images using an image processing algorithm, and obtaining flow field trace images; modeling a normal blade using SCDM software and meshing the model; simulating the blade under normal and different fracture conditions using a CFD numerical simulation method based on the meshed model to obtain flow field trace image information under different conditions; constructing a deep learning model based on the flow field trace image information under different conditions and all simulation results; and importing the photographed upper liquid surface trace manifold into the deep learning model to calculate and diagnose the fault condition. The present invention is used to indirectly detect agitator blade faults and provides a certain reference and guidance for determining whether the agitator blade is operating normally in actual industrial processes.

Description

Technical Field

[0001] The present invention particularly relates to a method for diagnosing stirring tank blade faults based on CFD and deep learning. Background Art

[0002] 1. The present invention can provide a reference for whether the stirring tank blades are operating normally. By analyzing the traces of the flow field during the stirring process, it can indirectly analyze and judge whether the stirring tank blades are faulty without delaying production efficiency.

[0003] 2. The present invention can solve the problem of being unable to directly judge blade failure due to visual obstruction caused by turbid material during the blade stirring process, thereby improving the accuracy of subsequent predictions. Summary of the Invention

[0004] In view of the above situation, the main purpose of the present invention is to propose a stirrer blade fault diagnosis method based on CFD and deep learning to solve the above technical problems.

[0005] The present invention provides a method for diagnosing agitator blade faults based on CFD and deep learning, the method comprising the following steps:

[0006] Step 1: Use a high-speed camera to photograph the stirring tank blades, obtain flow field images at different times under different blade fracture conditions, use an image processing algorithm to process the flow field images at different times, and obtain flow field trace images, which are used as the input data set of the deep learning model;

[0007] Step 2: Use SCDM software to perform geometric 3D modeling on the normal blade to obtain a 3D model, and then mesh the 3D model. Based on the data extracted from the meshed 3D model, use CFD numerical simulation methods to simulate the working conditions of the blade under normal conditions and various fracture conditions to obtain flow field trace image information under normal conditions and various fracture conditions.

[0008] Step 3: Build a deep learning model based on the flow field trace image information under normal conditions and various fracture conditions;

[0009] Step 4: Import the upper liquid surface trace manifold captured by the high-speed camera into the deep learning model to calculate and obtain diagnostic data for fault diagnosis.

[0010] The present invention also proposes a stirred tank blade fault diagnosis system based on CFD and deep learning, the system comprising:

[0011] Data extraction module for:

[0012] A high-speed camera was used to capture images of the stirring tank blades at different times under different fracture conditions. These images were then processed using an image processing algorithm to obtain flow field trace images, which were then used as input data sets for the deep learning model.

[0013] Simulated flow field trace information module, used for:

[0014] SCDM software was used to perform geometric 3D modeling of the normal blade to obtain a 3D model, which was then meshed. Based on the data extracted from the meshed 3D model, the CFD numerical simulation method was used to simulate the blade's operating conditions under normal conditions and various fracture conditions, thereby obtaining flow field trace image information for each of the normal and fracture conditions.

[0015] Deep learning model building blocks for:

[0016] A deep learning model is constructed based on the flow field trace image information under normal conditions and various fracture conditions;

[0017] Prediction module, used to:

[0018] The upper liquid surface trace manifold captured by a high-speed camera is imported into the deep learning model to calculate and obtain diagnostic data for fault diagnosis.

[0019] Compared with the prior art, the present invention has the following beneficial effects:

[0020] 1. The present invention can provide a reference for whether the stirring tank blades are operating normally. By analyzing the traces of the flow field during the stirring process, it can indirectly analyze and judge whether the stirring tank blades are faulty without delaying production efficiency.

[0021] 2. The present invention can solve the problem of being unable to directly judge blade failure due to visual obstruction caused by turbid material during the blade stirring process, thereby improving the accuracy of subsequent predictions. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 This is a flow chart of a method for diagnosing stirring tank blade faults based on CFD and deep learning proposed in the present invention;

[0023] Figure 2 Schematic diagram of the geometric structure of the agitator tank for the agitator tank blade fault diagnosis method based on CFD and deep learning proposed in the present invention;

[0024] Figure 3 This is the overall system framework diagram of the stirring tank blade fault diagnosis method based on CFD and deep learning proposed in this invention. DETAILED DESCRIPTION

[0025] The following describes embodiments of the present invention in detail. Examples of the embodiments 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 intended only to explain the present invention and are not to be construed as limiting the present invention.

[0026] These and other aspects of the embodiments of the present invention will become clear 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 illustrate 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.

[0027] See also Figure 1 The embodiment of the present invention provides a method for diagnosing agitator blade faults based on CFD and deep learning, the method comprising the following steps:

[0028] Step 1: Use a high-speed camera to shoot the stirring tank blades, obtain flow field images at different times under different fracture conditions, use image processing algorithms to process the flow field images at different times, and obtain flow field trace images, which are used as the input data set of the deep learning model.

[0029] In step 1, a high-speed camera is used to photograph the stirred tank blades to obtain flow field images at different times under different fracture conditions. The image processing algorithm is used to enhance and extract the flow field trace images to obtain extracted flow field trace images, which are then used as the input dataset for the deep learning model. The specific steps are as follows:

[0030] Blades with different fracture conditions were selected for experiments;

[0031] Use a high-speed camera to take pictures of blades with different fracture conditions to obtain flow field images at different times;

[0032] Based on the flow field images at different times, the flow field traces are enhanced and extracted by the Canny edge detection method to obtain the flow field trace images;

[0033] The flow field trace images are used as input datasets for the deep learning model.

[0034] Step 2: Use SCDM software to perform geometric three-dimensional modeling on the normal blade to obtain a three-dimensional model, and mesh the three-dimensional model. Based on the data extracted from the meshed three-dimensional model, use the CFD numerical simulation method to simulate the working conditions of the blade under normal conditions and various fracture conditions to obtain the flow field trajectory image information under normal conditions and various fracture conditions respectively.

[0035] Please refer to Figure 2 In step 2, the SCDM software is used to perform geometric 3D modeling on the normal blade to obtain a 3D model, and the 3D model is meshed. Based on the data extracted from the meshed 3D model, the CFD numerical simulation method is used to simulate the working conditions of the blade under normal conditions and various fracture conditions to obtain flow field trace image information under normal conditions and various fracture conditions. The specific steps are as follows:

[0036] 2-1. Use SCDM software to perform geometric 3D modeling on a normal blade to obtain a geometric 3D model;

[0037] 2-2. Use ECEM software to mesh the geometric three-dimensional model, using structured mesh for the stationary domain and unstructured mesh for the rotating domain to obtain the meshed three-dimensional model;

[0038] 2-3. Based on the data extracted from the meshed 3D model, the governing equations are constructed, and the multiphase flow Euler model is constructed based on the governing equations.

[0039] 2-4. Set the parameters of the multiphase flow Euler model, set liquid water as the primary phase in the multiphase flow model, and set fixed particles as the secondary phase;

[0040] 2-5. Calculate the interaction force between particles and fluid based on the set parameters;

[0041] 2-6. Set the boundary conditions of the normal blade geometry 3D model: set the wall of the stirred tank as the fluid-solid coupling boundary, set the solution as the velocity inlet boundary, and the outlet as the pressure outlet boundary;

[0042] 2-7. Set the parameters of the reactor and reactants as the initial conditions of the geometric 3D model;

[0043] 2-8. Use the finite volume method to discretize the basic governing equations. Substitute the boundary conditions of the geometric three-dimensional model and the initial conditions of the particles and fluid into the discretized basic equations for a closed-form solution.

[0044] 2-9. Initialize the entire fluid domain, set the time step and start the calculation. Repeat the iterative calculation until the solved parameters converge and the model verification is completed.

[0045] 2-10. Based on the model verification, construct a three-dimensional geometric model of the mixing tank with different blade fracture conditions, and repeat steps 2-1 to 2-9 until the numerical simulation process of all the mixing tank geometric models is completed;

[0046] 2-11. Export all simulation results and obtain flow field trace image information under normal conditions and various fracture conditions;

[0047] In the steps of constructing the control equations based on the data extracted from the meshed 3D model and then constructing the multiphase flow Euler model based on the control equations, the control equations include the continuity equation and the momentum equation. The continuity equation is expressed as follows:

[0048] ;

[0049] in, is the Laplace operator, is the mesh porosity of the fluid, For the The volume fraction of the phase, For the The density of the phase, is the fluid velocity, For time, is the symbol of partial derivative;

[0050] The volume fraction of the main phase in the continuity equation is calculated by the closed equation, and the relationship between the corresponding process is:

[0051] ;

[0052] in, is the total number of phases;

[0053] The momentum equation is expressed as follows:

[0054] ;

[0055] in, is the static pressure, is the viscous stress tensor, is the acceleration due to gravity, is the surface tension, is the particle-fluid interaction force, is the density of the fluid;

[0056] The expression for the viscous stress tensor in the momentum equation is:

[0057] ;

[0058] in, is the feature tensor, is the dynamic viscosity of the fluid, is the spatial rate of change of velocity, is the transpose symbol;

[0059] The expression for surface tension in the momentum equation is:

[0060] ;

[0061] in, is the surface tension coefficient, is the curvature at the phase interface, For the The spatial rate of change of the volume fraction of the phase;

[0062] In the step of calculating the interaction force between particles and fluid based on the set parameters, the calculation relationship of the interaction force between particles and fluid is:

[0063] ;

[0064] in, is the drag force during the interaction between particles and fluid, is the pressure gradient force, is the viscous stress, is the Reynolds stress, is the capillary force, For Saffman lift, Lift for Magnus, is the interaction force between particles and fluid.

[0065] Step 3: Build a deep learning model based on the flow field trace image information under normal conditions and various fracture conditions;

[0066] In step 3, a deep learning model is constructed based on the flow field trajectory image information extracted from the experimental images and the flow field trajectory information under all simulation conditions. In order to take into account the cost issues and the sample imbalance caused by the lack of fault data, the fault data is supplemented by simulated data to make up for the imbalance between samples, so that the training data samples are rich and balanced.

[0067] Step 4: Import the upper liquid surface trace manifold captured by the high-speed camera into the deep learning model to calculate and obtain diagnostic data for fault diagnosis;

[0068] In step 4, the upper liquid surface trace manifold captured by a high-speed camera is input into the model to perform normal and fault classification and identification, and diagnose the fault condition.

[0069] See also Figure 3 , an embodiment of the present invention further provides a stirred tank blade fault diagnosis system based on CFD and deep learning, the system comprising:

[0070] Data extraction module for:

[0071] A high-speed camera was used to capture images of the stirring tank blades at different times under different fracture conditions. These images were then processed using an image processing algorithm to obtain flow field trace images, which were then used as input data sets for the deep learning model.

[0072] Simulated flow field trace information module, used for:

[0073] SCDM software was used to perform geometric 3D modeling of the normal blade to obtain a 3D model, which was then meshed. Based on the data extracted from the meshed 3D model, the CFD numerical simulation method was used to simulate the blade's operating conditions under normal conditions and various fracture conditions, thereby obtaining flow field trace image information for each of the normal and fracture conditions.

[0074] Deep learning model building blocks for:

[0075] A deep learning model is constructed based on the flow field trace image information under normal conditions and various fracture conditions;

[0076] Prediction module, used to:

[0077] The upper liquid surface trace manifold captured by a high-speed camera is imported into the deep learning model to calculate and obtain diagnostic data for fault diagnosis.

[0078] It should be understood that various components of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used: a discrete logic circuit having logic gate circuits for implementing logic functions on data signals, an application-specific integrated circuit having suitable combinational logic gate circuits, a programmable gate array (PGA), a field-programmable gate array (FPGA), etc.

[0079] Throughout this specification, reference to terms such as "one embodiment," "some embodiments," "examples," "specific examples," or "some examples" means that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0080] The above-described embodiments merely illustrate several implementations of the present invention, and while their descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art would be able to make numerous variations and improvements without departing from the spirit of the present invention, all of which fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be determined by the appended claims.

Claims

1. A method for fault diagnosis of agitator blades based on CFD and deep learning, characterized in that: The method comprises the following steps: Step 1: Use a high-speed camera to photograph the stirring tank blades, obtain flow field images at different times under different blade fracture conditions, use an image processing algorithm to process the flow field images at different times, and obtain flow field trace images, which are used as the input data set of the deep learning model; Step 2: Use SCDM software to perform geometric 3D modeling on the normal blade to obtain a 3D model, and then mesh the 3D model. Based on the data extracted from the meshed 3D model, use CFD numerical simulation methods to simulate the working conditions of the blade under normal conditions and various fracture conditions to obtain flow field trace image information under normal conditions and various fracture conditions. Step 3: Build a deep learning model based on the flow field trace image information under normal conditions and various fracture conditions; Step 4: Import the upper liquid surface trace manifold captured by the high-speed camera into the deep learning model to calculate and obtain diagnostic data for fault diagnosis.

2. The method for diagnosing agitator blade faults based on CFD and deep learning according to claim 1, characterized in that: In step 1, a high-speed camera is used to photograph the stirring tank blade to obtain flow field images at different times under different fracture conditions. An image processing algorithm is used to enhance and extract the flow field trace images to obtain extracted flow field trace images, and the flow field trace images are used as the input data set of the deep learning model. The specific steps are as follows: Blades with different fracture conditions were selected for experiments; Use a high-speed camera to take pictures of blades with different fracture conditions to obtain flow field images at different times; Based on the flow field images at different times, the flow field traces are enhanced and extracted by the Canny edge detection method to obtain the flow field trace images; The flow field trace images are used as input datasets for the deep learning model.

3. The method for diagnosing agitator blade faults based on CFD and deep learning according to claim 2, characterized in that: In step 2, SCDM software is used to perform geometric three-dimensional modeling on the normal blade to obtain a three-dimensional model, and the three-dimensional model is meshed. Based on the data extracted from the meshed three-dimensional model, the CFD numerical simulation method is used to simulate the working conditions of the blade under normal conditions and various fracture conditions, so as to obtain flow field trace image information under normal conditions and various fracture conditions respectively. The specific steps are as follows: 2-1. Use SCDM software to perform geometric 3D modeling on a normal blade to obtain a geometric 3D model; 2-2. Use ECEM software to mesh the geometric three-dimensional model, using structured mesh for the stationary domain and unstructured mesh for the rotating domain to obtain the meshed three-dimensional model; 2-3. Based on the data extracted from the meshed 3D model, the governing equations are constructed, and the multiphase flow Euler model is constructed based on the governing equations. 2-4. Set the parameters of the multiphase flow Euler model, set liquid water as the primary phase in the multiphase flow model, and set fixed particles as the secondary phase; 2-5. Calculate the interaction force between particles and fluid based on the set parameters; 2-6. Set the boundary conditions of the normal blade geometry 3D model: set the wall of the stirred tank as the fluid-solid coupling boundary, set the solution as the velocity inlet boundary, and the outlet as the pressure outlet boundary; 2-7. Set the parameters of the reactor and reactants as the initial conditions of the geometric 3D model; 2-8. Use the finite volume method to discretize the basic governing equations. Substitute the boundary conditions of the geometric three-dimensional model and the initial conditions of the particles and fluid into the discretized basic equations for a closed-form solution. 2-9. Initialize the entire fluid domain, set the time step and start the calculation. Repeat the iterative calculation until the solved parameters converge and the model verification is completed. 2-10. Based on the model verification, construct a three-dimensional geometric model of the mixing tank with different blade fracture conditions, and repeat steps 2-1 to 2-9 until the numerical simulation process of all the mixing tank geometric models is completed; 2-11. Export all simulation results and obtain flow field trace image information under normal conditions and various fracture conditions.

4. The method for diagnosing agitator blade faults based on CFD and deep learning according to claim 3, characterized in that: In the step of constructing the control equations based on the data extracted from the meshed 3D model and then constructing the multiphase flow Euler model based on the control equations, the control equations include the continuity equation and the momentum equation. The continuity equation is expressed as follows: ； in, is the Laplace operator, is the mesh porosity of the fluid, For the The volume fraction of the phase, For the The density of the phase, is the fluid velocity, For time, is the symbol of partial derivative; The volume fraction of the main phase in the continuity equation is calculated by the closed equation, and the relationship between the corresponding process is: ； in, is the total number of phases.

5. The method for diagnosing agitator blade faults based on CFD and deep learning according to claim 4, characterized in that: In the steps of constructing the control equations based on the data extracted from the meshed 3D model and then constructing the multiphase flow Euler model based on the control equations, the control equations include the continuity equation and the momentum equation. The momentum equation is expressed as follows: ； in, is the static pressure, is the viscous stress tensor, is the acceleration due to gravity, is the surface tension, is the particle-fluid interaction force, is the density of the fluid.

6. The method for diagnosing agitator blade faults based on CFD and deep learning according to claim 5, characterized in that: In the step of constructing the control equations based on the data extracted from the meshed 3D model and then building the multiphase flow Euler model based on the control equations, the control equations include the continuity equation and the momentum equation. The expression of the viscous stress tensor in the momentum equation is: ； in, is the feature tensor, is the dynamic viscosity of the fluid, is the spatial rate of change of velocity, is the transpose symbol; The expression for surface tension in the momentum equation is: ； in, is the surface tension coefficient, is the curvature at the phase interface, For the The spatial rate of change of the volume fraction of a phase.

7. The method for diagnosing agitator blade faults based on CFD and deep learning according to claim 6, characterized in that: In the step of calculating the interaction force between particles and fluid based on the set parameters, the calculation relationship of the interaction force between particles and fluid is: ； in, is the drag force during the interaction between particles and fluid, is the pressure gradient force, is the viscous stress, is the Reynolds stress, is the capillary force, For Saffman lift, Lift for Magnus, is the interaction force between particles and fluid.

8. A CFD and deep learning-based agitator blade fault diagnosis system, characterized by: The system applies a CFD and deep learning-based stirred tank blade fault diagnosis method as described in any one of claims 1 to 7, and the system includes: Data extraction module for: A high-speed camera was used to capture images of the stirring tank blades at different times under different fracture conditions. These images were then processed using an image processing algorithm to obtain flow field trace images, which were then used as input data sets for the deep learning model. Simulated flow field trace information module, used for: SCDM software was used to perform geometric 3D modeling of the normal blade to obtain a 3D model, which was then meshed. Based on the data extracted from the meshed 3D model, the CFD numerical simulation method was used to simulate the blade's operating conditions under normal conditions and various fracture conditions, thereby obtaining flow field trace image information for each of the normal and fracture conditions. Deep learning model building blocks for: A deep learning model is constructed based on the flow field trace image information under normal conditions and various fracture conditions; Prediction module, used to: The upper liquid surface trace manifold captured by a high-speed camera is imported into the deep learning model to calculate and obtain diagnostic data for fault diagnosis.

Images