A machine learning-based cyclone separator optimization system and method

Abstract

This invention relates to the field of separation device technology, and discloses a cyclone separator optimization system and method based on machine learning. The system includes a user operation module, a numerical simulation module, and a deep learning module. The method includes: when a user terminal inputs cyclone separator parameters into a cloud platform, the system retrieves a cyclone separator parameter database; if no corresponding cyclone separator parameter is found in the database, simulation calculations are performed based on the user-input cyclone separator parameters, and a flow prediction model is established based on the simulation results; the flow prediction model makes predictions based on the simulation results; after the predictions are completed, the prediction results are compared; finally, the required optimized parameters for the cyclone separator are obtained, and the optimized parameters are transmitted to the user terminal through the cloud platform. This invention predicts the separation efficiency of cyclone separators with different designs, saving manpower, material resources, time, and costs in actual production.

Description

Technical Field

[0001] This invention relates to the field of separation device technology, and in particular to a cyclone separator optimization system and method based on machine learning. Background Technology

[0002] A cyclone separator utilizes the centrifugal force generated by the high-speed rotation of a dust-laden airflow inside to separate dust particles from the airflow and collect them on the separator wall. Under gravity, the particles eventually fall into the ash hopper, thus achieving dust removal. This type of dust removal equipment is widely used in industries such as steel metallurgy, chemical engineering, and thermal power generation due to its advantages of high dust removal efficiency, simple structure, low cost, convenient operation, and high reliability. In recent years, environmental problems have become increasingly prominent, especially since flue gas and dust pollution can cause various respiratory diseases. Therefore, the development of efficient cyclone separators is of great significance. Separation efficiency is one of the main performance characteristics of a cyclone separator, so most inventions focus on improving the separation efficiency of cyclone separators.

[0003] The separation efficiency of a cyclone separator is closely related to particle size, but is also affected by structural factors such as the separator cylinder diameter, cylinder length, and inlet height. Furthermore, cyclone separators require individual design for different operating scenarios. Existing design methods determine optimal design parameters by repeating a series of processes, including: collecting and analyzing initial process parameters; calculating actual process parameters under operating conditions; designing the cyclone single-tube distribution structure; designing and calculating the cyclone separator's structural dimensions; estimating the separator's pressure drop; estimating the separator's particle separation efficiency; and calculating the separator's wall thickness and strength verification. All these calculations are performed manually by the designer. This method is labor-intensive, time-consuming, has a long design cycle, low calculation accuracy, is prone to errors, and is costly. Moreover, this method typically requires designers with strong professional knowledge and extensive design experience to perform empirical design, making it unsuitable for everyone.

[0004] Prior art 1, Chinese patent application number: CN202322225931.9, relates to a cyclone separator dust removal and circulation system, including a cyclone separator, a fan, and a raw material mechanism. One end of the raw material mechanism has a feed inlet, and the bottom of the other end has a hard impurity outlet and a clean cottonseed outlet. The top of the other end of the raw material mechanism is connected to the inlet of the cyclone separator via an air outlet duct. The fan is connected to the part of the raw material mechanism near the hard impurity outlet via an air supply duct, and the top of the cyclone separator is connected to the fan inlet via a return air duct. The advantages are its simple structure and reasonable design. Although it achieves the circulation treatment of the cyclone separator exhaust gas, resulting in high cleanliness of the discharged gas, it is energy-saving and environmentally friendly, and the cost is low. However, the dust removal efficiency of the cyclone separator is low, especially when processing fine particles, making it difficult to achieve the ideal separation effect.

[0005] Prior art two, Chinese patent application number CN202410012033.1, discloses a cyclone separator, a cyclone separation system, and an operating method. The cyclone separator includes: a separator body with an inlet end and an outlet end; a riser pipe located at the outlet end and extending into the separator body; an adjusting component located inside the riser pipe and movably connected to it; and a moving component located outside the riser pipe and capable of moving the adjusting component. The moving component can move the adjusting component up and down within the separator body to adjust the separation efficiency of the separator body by changing the insertion depth of the adjusting component. While the above-mentioned cyclone separator can achieve online adjustment of particle escape rate and realizes the adjustment of the separator's separation efficiency, its simple and effective structure makes it suitable for different operating conditions and has good application value. However, under significant scale-up conditions, the performance of the cyclone separator may be negatively affected, leading to a decrease in separation efficiency.

[0006] Prior art three, Chinese patent application number CN202310923160.2, discloses a cyclone separator control method, system, and device adapted to flow rate changes, relating to the field of cyclone separator technology. The method includes: constructing a cyclone separator performance prediction model based on machine learning; inputting the gas flow rate at the cyclone separator inlet into the performance prediction model to predict the cyclone separator's separation efficiency; determining a flow control signal based on the difference between the predicted separation efficiency and the optimal separation efficiency; and controlling the motor frequency of the fan controller based on the flow control signal, thereby controlling the jet flow rate entering the cyclone separator, so that the cyclone separator is in the flow mode with optimal separation efficiency. Although this method can ensure that the separator is always in the flow mode with optimal separation efficiency, avoiding the problem of low cyclone separator separation efficiency caused by changes in flow parameters during the process, it lacks precise control of gas flow conditions, affecting the separation effect.

[0007] Currently, existing technologies 1, 2, and 3 suffer from low dust removal efficiency of cyclone separators, especially when handling fine particles, making it difficult to achieve ideal separation results; under significant scale-up effects, the performance of cyclone separators may be negatively affected, leading to a decrease in separation efficiency; and there is a lack of precise control over gas flow conditions, affecting the separation effect. To solve the above problems, this invention provides an optimization system and method for cyclone separators based on machine learning. Summary of the Invention

[0008] The main objective of this invention is to provide a machine learning-based cyclone separator optimization system and method to address the problems in the prior art, such as low dust removal efficiency of cyclone separators, especially when handling fine particles, making it difficult to achieve ideal separation results; the performance of cyclone separators may be negatively affected by significant scale-up effects, leading to a decrease in separation efficiency; and the lack of precise control over gas flow conditions, which affects the separation effect.

[0009] To achieve the above objectives, the present invention provides the following technical solution:

[0010] A machine learning-based cyclone separator optimization system includes:

[0011] The user operation module is responsible for handling user account registration upon first login. After successful login, users can input relevant parameters of the cyclone separator through the user terminal interface. Based on the user's input of cyclone separator parameters, the module retrieves data stored in the database. It then determines whether the corresponding cyclone separator parameter data is stored. If it is, the database data is retrieved and output to the user terminal. If not, the new data is imported into the numerical simulation module for calculation.

[0012] The parameters related to the cyclone separator include: cyclone separator design parameters, working environment, and operating parameters; the parameters of the cyclone separator include cyclone separator parameters, working environment, and operating parameters; the working environment includes working pressure, working temperature, humidity, and air density; the operating parameters include inlet velocity, particle size, particle density, particle viscosity, outlet pressure, and inlet pressure.

[0013] The numerical simulation module is responsible for receiving relevant parameters of the cyclone separator imported from the user terminal; drawing the geometric model of the cyclone separator using SolidWorks based on the design parameters of the cyclone separator; simulating the cyclone separator model and calculating the separation efficiency; and sending the separator simulation results and separation efficiency to the deep learning module.

[0014] The deep learning module is responsible for preprocessing simulation results and separation efficiency data to create a dataset; sorting the dataset according to timestamps; training the flow prediction model based on the sorted data; identifying the areas that need optimization in the flow prediction model by combining simulation results and prediction results; continuously iterating and adjusting the optimization areas to obtain the final cyclone separator parameters; and feeding the adjustments back to the user terminal.

[0015] As a further improvement of the present invention, the user operation module includes:

[0016] The registration submodule is responsible for collecting facial images of users to be registered, extracting the face features of the facial images, generating facial feature vectors, reducing the dimensionality of the facial vectors, calibrating the feature vectors, and if they meet the set conditions, inputting the feature vectors into the feature library and completing the registration.

[0017] If not, then the opposite is true;

[0018] The process involves comparing the similarity between the associated feature vector of the user to be registered and all key features in the feature database of the registered user. If the similarity meets the set conditions, a temporary identity is created. If the similarity does not meet the set conditions, identity authentication is performed. If identity authentication fails, the user's facial information is collected again.

[0019] The operation submodule is responsible for responding to data parameters initiated by the user terminal, retrieving the cyclone separator target database, determining whether the data parameters initiated by the user terminal correspond to the data parameters in the cyclone separator target database, checking whether there are corresponding data parameters in the cyclone separator target database, and outputting the judgment result to the data transmission submodule.

[0020] The data transmission submodule is responsible for outputting the data parameters to the user terminal if there are corresponding cyclone separator target database data parameters, and transmitting the cyclone separator parameters transmitted by the user terminal to the numerical simulation module if there are no corresponding cyclone separator target database data parameters.

[0021] The generation unit is responsible for generating control commands from data parameters in the target database that do not have a corresponding cyclone separator, processing the control commands, and returning if the command is incorrect, prompting the user terminal that the input control command is incorrect.

[0022] The command transmission unit is responsible for acquiring user terminal control commands and storing them; processing the transmitted data according to the control commands; and running the compilation device to compile the transmitted data.

[0023] The parsing unit is responsible for transmitting the compilation results to the numerical simulation module. During the transmission process, the data is sent to the numerical simulation module through the pin unit. The numerical simulation module performs control simulation based on the compilation results.

[0024] As a further improvement of the present invention, the numerical simulation module includes:

[0025] The first preprocessing submodule is responsible for preprocessing the relevant parameters of the cyclone separator imported from the user terminal, drawing the geometric model of the cyclone separator using SolidWorks based on the preprocessing results, preprocessing the relevant parameters of the cyclone separator, and meshing the cyclone separator model.

[0026] The simulation submodule is responsible for setting the relevant parameters of the simulation equipment based on the preprocessed cyclone separator parameters; simulating the gas-solid two-phase flow inside the cyclone separator to determine the flow mode of the cyclone separator; and calculating the separation efficiency based on the flow mode.

[0027] The simulation results transmission submodule is responsible for comparing the separation efficiency with that of historical experiments, evaluating the separation efficiency, and sending the simulation results, separation efficiency, and evaluation results to the deep learning module.

[0028] As a further improvement of the present invention, the first preprocessing submodule includes:

[0029] The drawing unit is responsible for verifying the relevant parameters of the cyclone separator imported from the user terminal, comparing the relevant parameters with standard data, verifying the accuracy and completeness of the relevant parameters, and drawing the geometric model of the cyclone separator based on the relevant data after verification.

[0030] The mesh generation unit is responsible for dividing the geometric model of the cyclone separator into multiple non-overlapping sub-regions, specifying the node positions of each sub-region and the control volume represented by each node; setting mesh parameters and generating the mesh.

[0031] The mesh verification unit is responsible for checking the mesh after it is generated. If the mesh is qualified, it is output to the simulation submodule. If the mesh is unqualified, the unqualified area is checked and the mesh is re-generated in that area.

[0032] As a further improvement of the present invention, the simulation submodule includes:

[0033] The flow determination unit is responsible for determining the flow type based on the target parameters of the cyclone separator after mesh generation.

[0034] Formula for determining flow type:

[0035]

[0036] Where ρ is the fluid density, U is the fluid flow velocity, L is the characteristic dimension, and μ is the fluid dynamic viscosity; for the cyclone separator of the present invention, when Re>4000, it is turbulent flow; otherwise, it is swirling flow.

[0037] The flow calculation unit is responsible for predicting the kinematic characteristics of the swirling flow within the three-dimensional spray gun using discretized Navier-Stokes equations, assuming the gas phase flow field is steady-state, compressible, and turbulent. When no other external forces act on the flow field, the continuity and momentum conservation equations for the swirling flow are as follows:

[0038]

[0039] Where u, ρ, P, and g represent the velocity vector, density, pressure, and gravitational acceleration of the gas phase in the computational domain, respectively. The symbol v represents the kinematic viscosity of the gas phase. t Indicates the viscosity of the gas flow;

[0040] Turbulent closure is performed using SST k-ω as the unknown term in the Navier-Stokes equations;

[0041] Computational Turbulence Function (k) g The formula is:

[0042]

[0043] Calculate the eddy current dissipation rate (ω) g The formula is:

[0044]

[0045] The separation efficiency unit is responsible for calculating the separation efficiency based on the above calculation results;

[0046] The formula for calculating the separation efficiency is as follows:

[0047]

[0048] Among them, M in M represents the mass of particles flowing into the cyclone separator inlet. out This refers to the mass of particles flowing out from the upper outlet of the cyclone separator.

[0049] As a further improvement of the present invention, the deep learning module includes:

[0050] The second preprocessing submodule is responsible for acquiring simulation results and separation efficiency, preprocessing the simulation results and separation efficiency to form a dataset, sorting the dataset according to timestamps to obtain a time series dataset, and dividing the time series dataset into training set, test set and validation set.

[0051] The flow prediction model submodule is responsible for training the flow prediction model by inputting time series data into the flow prediction model, taking environmental data information from the simulation results as input, and cyclone separator design parameter data as output, extracting the relationship between component deformation and changes in the surrounding environment, and finally obtaining the trained flow prediction model.

[0052] The first optimization submodule is responsible for comparing the simulation results with the prediction results, identifying the areas that need optimization in the flow prediction model, continuously adjusting the parameters of the optimization areas until the maximum number of iterations is reached or a qualified area is reached; and feeding back the adjusted optimization area parameters to the user terminal.

[0053] As a further improvement of the present invention, the second preprocessing submodule includes:

[0054] The third preprocessing unit is responsible for extracting relevant features, reducing the dimensionality and complexity of the data structure using dimensionality reduction devices, and normalizing the extracted data to ensure that the data have similar scale and distribution.

[0055] The feature extraction unit is responsible for extracting the timestamps of each data point, arranging the normalized data in order according to the timestamps to form time series data, and setting the time series data as a time series dataset.

[0056] The dataset partitioning unit is responsible for dividing the time series dataset into training, testing, and validation sets; and sending the divided training, testing, and validation sets of the time series dataset to the flow prediction model submodule 32.

[0057] As a further improvement of the present invention, the flow prediction model submodule includes:

[0058] The structural unit is responsible for inputting time series data into the flow prediction model and setting the flow prediction model architecture into an input layer, a hidden layer, and an output layer. The pipe diameter data from the simulation results is used as input, the hidden layer performs feature extraction and processing, and the cyclone separator design parameter data is used as output to train the flow prediction model. Iterative training is performed using minimum batch data.

[0059] The training unit is responsible for validating the process of training the cyclone separator design parameters using the validation set; if there is no improvement for several consecutive cycles, training is stopped; the relationship between the structure deformation and environmental parameters is analyzed; the influence of key pipe diameter parameters on the structure deformation is extracted, and a relevant relationship data table is generated;

[0060] The test model unit is responsible for adjusting the relevant hyperparameters of the flow prediction model based on the relevant relational data table, and finally obtaining the final flow prediction model.

[0061] As a further improvement of the present invention, the first optimized submodule includes:

[0062] The identification and optimization unit is responsible for comparing the simulation results with the prediction results, identifying the areas that need to be optimized in the flow prediction model, and adjusting these areas through continuous iteration. After each iteration, the simulation and prediction are repeated until the maximum number of iterations or a qualified area is reached.

[0063] The second optimization unit is responsible for transmitting the adjusted optimization area parameters to the user terminal through the cloud platform. After receiving the optimization parameters, the user terminal can further adjust the parameters or confirm the final cyclone separator according to the actual situation. The parameter adjustment includes: manual adjustment or re-requesting optimization.

[0064] The feedback unit is responsible for receiving the re-request for optimization initiated by the user terminal, retrieving the cyclone separator information from the database, and re-simulating and predicting the cyclone separator based on the target parameters given by the user until it is qualified.

[0065] To achieve the above objectives, the present invention also provides the following technical solution:

[0066] A machine learning-based cyclone separator optimization method:

[0067] First-time login requires account registration. After successful login, users can input relevant parameters of the cyclone separator through the user terminal interface. Based on the input parameters of the cyclone separator through the user terminal, the system retrieves the data stored in the database. It then determines whether the corresponding cyclone separator parameter data is stored. If it is, the database data is retrieved and output to the user terminal. If not, the new data is imported into the simulation model.

[0068] The parameters of the cyclone separator include cyclone separator parameters, working environment, and operating parameters; the working environment includes working pressure, working temperature, humidity, and air density; the operating parameters include inlet velocity, particle size, particle density, particle viscosity, outlet pressure, and inlet pressure.

[0069] Receive relevant parameters of the cyclone separator imported from the user terminal; draw the geometric model of the cyclone separator using SolidWorks according to the design parameters of the cyclone separator; simulate the cyclone separator model and calculate the separation efficiency; and preprocess the separator simulation results and separation efficiency.

[0070] A dataset is established based on the preprocessing results; the dataset is sorted according to timestamps, and the flow prediction model is trained based on the sorted data; the flow prediction model needs to be optimized by combining simulation results and prediction results; the optimization area is continuously adjusted iteratively to obtain the final cyclone separator parameters; the adjustments are fed back to the user terminal.

[0071] This invention can predict the separation efficiency of cyclone separators with different designs based on the production environment and operating parameters, providing suggestions for actual production and significantly saving manpower, materials, time, and costs. It does not require operators to have extensive knowledge of fluid mechanics or separator design; even those unfamiliar with cyclone separators can use the system to adjust design inputs to control relevant parameters of the design output, significantly reducing cyclone separator design time and workload. Through the integration of an online system and deep learning, real-time data updates are achieved, ensuring the richness, relevance, and timeliness of the data. Attached Figure Description

[0072] Figure 1 This is a functional module diagram of an embodiment of the cyclone separator optimization system based on machine learning of the present invention;

[0073] Figure 2 This is a structural diagram of a cyclone separator according to an embodiment of the machine learning-based cyclone separator optimization system of the present invention;

[0074] Figure 3 This is a functional module diagram of the user operation module of an embodiment of the cyclone separator optimization system based on machine learning of the present invention;

[0075] Figure 4 This is a schematic diagram of the user login process in an embodiment of the machine learning-based cyclone separator optimization system of the present invention.

[0076] Figure 5 This is a functional module diagram of the data transmission submodule of an embodiment of the machine learning-based cyclone separator optimization system of the present invention;

[0077] Figure 6 This is a functional module diagram of the numerical simulation module of an embodiment of the cyclone separator optimization system based on machine learning of the present invention;

[0078] Figure 7 This is an embodiment of the machine learning-based cyclone separator optimization system of the present invention. Figure 3 and Figure 5 Schematic diagram;

[0079] Figure 8 This is a functional module diagram of the first preprocessing submodule in an embodiment of the machine learning-based cyclone separator optimization system of the present invention;

[0080] Figure 9 This is a block diagram of a simulation of an embodiment of the cyclone separator optimization system based on machine learning of the present invention.

[0081] Figure 10This is a schematic diagram of the deep learning module in an embodiment of the cyclone separator optimization system based on machine learning of the present invention.

[0082] Figure 11 This is a schematic diagram of the deep learning module in an embodiment of the cyclone separator optimization system based on machine learning of the present invention.

[0083] Figure 12 This is a functional module diagram of the second preprocessing submodule in an embodiment of the cyclone separator optimization system based on machine learning of the present invention;

[0084] Figure 13 This is a functional module diagram of the flow prediction model submodule, which is an embodiment of the machine learning-based cyclone separator optimization system of the present invention.

[0085] Figure 14 This is a functional module diagram of the first optimization submodule of an embodiment of the cyclone separator optimization system based on machine learning of the present invention;

[0086] Figure 15 This is a flowchart illustrating the steps of an embodiment of the cyclone separator optimization method based on machine learning of the present invention.

[0087] Figure 16 This is a schematic diagram of the structure of an embodiment of the electronic device of the present invention;

[0088] Figure 17 This is a schematic diagram of the structure of one embodiment of the storage medium of the present invention. Detailed Implementation

[0089] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0090] The terms "first," "second," and "third" used in this invention are for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, a feature defined as "first," "second," or "third" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified. All directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of this invention are only used to explain the relative positional relationships and movements between components in a specific orientation (as shown in the figures). If the specific orientation changes, the directional indications also change accordingly. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or devices.

[0091] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a mutually exclusive, independent, or alternative embodiment. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0092] like Figure 1 As shown, this embodiment provides an example of a machine learning-based cyclone separator optimization system, which specifically includes:

[0093] User operation module 1 is responsible for handling user account registration upon first login. After successful login, users can input relevant parameters of the cyclone separator through the user terminal operation interface. Based on the user's input of relevant parameters of the cyclone separator, the module retrieves the data stored in the database. It determines whether the corresponding cyclone separator relevant parameter data is stored. If it is, the database data is retrieved and output to the user terminal. If not, the new data is imported into numerical simulation module 2 for calculation.

[0094] The parameters of the cyclone separator include cyclone separator parameters, working environment, and operating parameters; the working environment includes working pressure, working temperature, humidity, and air density; the operating parameters include inlet velocity, particle size, particle density, particle viscosity, outlet pressure, and inlet pressure; the cyclone separator design parameters are attached. Figure 2 ;

[0095] Numerical simulation module 2 is responsible for receiving relevant parameters of the cyclone separator imported from the user terminal; drawing the geometric model of the cyclone separator using SolidWorks based on the design parameters of the cyclone separator; simulating the cyclone separator model and calculating the separation efficiency; and sending the separator simulation results and separation efficiency to deep learning module 3.

[0096] Deep learning module 3 is responsible for preprocessing simulation results and separation efficiency data to establish a dataset; sorting the dataset according to timestamps; training the flow prediction model based on the sorted data; identifying the areas that need optimization in the flow prediction model by combining simulation results and prediction results; continuously iterating and adjusting the optimization areas to obtain the final cyclone separator parameters; and feeding the adjustments back to the user terminal.

[0097] Preferably, in this embodiment, the user operation module 1 ensures the uniqueness and security of user identity, provides an intuitive operation interface, facilitates user input and modification of parameters, quickly retrieves and displays historical data, improving efficiency; ensures data accuracy and consistency; directly displays historical data, reducing redundant calculations; and processes and analyzes new data. The numerical simulation module 2 ensures accurate data transmission; utilizes the powerful modeling capabilities of SolidWorks to ensure the accuracy of the geometric model; predicts the performance of the cyclone separator through numerical simulation; and provides basic data for subsequent data analysis and optimization. The deep learning module 3 ensures data quality and consistency; improves the accuracy and reliability of the model; accurately locates optimization points, improving design efficiency; achieves optimal design parameters through iterative optimization; and allows users to understand the optimization results in a timely manner, facilitating subsequent operations. The cyclone separator mainly consists of a dust-laden gas inlet pipe, a purified gas outlet pipe (exhaust pipe), a main body, an ash discharge pipe, and an ash hopper. Figure 2 In the diagram, a - inlet height, b - inlet width, D0 - cylinder diameter, D2 - ash discharge port diameter, α - semi-cone angle, h - cylinder height, Hh - cone height, and de - exhaust pipe diameter. The structural dimensions of the cyclone separator have a significant impact on its performance.

[0098] In summary, this invention has collectively promoted the automation and intelligence of cyclone separator design and optimization, improving design efficiency and performance. It can predict the separation efficiency of cyclone separators with different designs based on the production environment and various operating parameters, providing suggestions for actual production and significantly saving manpower, resources, time, and costs. Operators do not need extensive expertise in fluid mechanics or separator design; even those unfamiliar with cyclone separators can use this system to adjust design inputs to control relevant design output parameters, significantly reducing cyclone separator design time and workload. The integration of an online system and deep learning enables real-time data updates, ensuring the richness, relevance, and timeliness of the data.

[0099] Furthermore, such as Figure 3 As shown, user operation module 1 specifically includes:

[0100] The registration submodule 11 is responsible for collecting the face images of users to be registered, extracting the face features of the face images, generating face feature vectors, reducing the dimensionality of the face vectors, calibrating the feature vectors, and if the set conditions are met, inputting the feature vectors into the target database and completing the registration.

[0101] If not, then the opposite is true;

[0102] The process involves comparing the similarity between the associated feature vector of the user to be registered and all key features in the feature database of the registered user. If the similarity meets the set conditions, a temporary identity is created. If the similarity does not meet the set conditions, identity authentication is performed. If identity authentication fails, the user's facial information is collected again.

[0103] F(x) = W·X + b

[0104] Where: F(x) represents the output face feature vector, which is a multi-dimensional vector containing multiple features extracted from the face image; W is a weight matrix that contains the transformation rules between the face image and the face feature vector; X is the input face image data, which can be a matrix of pixel values ​​or data after dimensionality reduction through some method (such as PCA); and b is a bias vector used to adjust the overall position of the output feature vector.

[0105] Introducing some nonlinear transformations and activation functions, as follows:

[0106] F(x) = σ(W·X + b)

[0107] σ represents the activation function, which performs a nonlinear transformation on the result of multiplying the weight matrix and the input image.

[0108] Furthermore, this process can be extended to multi-layer networks:

[0109] F(x)=σ(W2·σ(W1·X+b1)+b2)

[0110] In the equation, W1 and W2 are weight matrices of different layers, and b1 and b2 are bias vectors of different layers;

[0111] Operation submodule 12 is responsible for responding to data parameters initiated by the user terminal to retrieve the cyclone separator target database, determining the data parameters in the cyclone separator target database corresponding to the data parameters initiated by the user terminal, judging whether there are corresponding data parameters in the cyclone separator target database, and outputting the judgment result to data transmission submodule 13;

[0112] The data transmission submodule 13 is responsible for outputting the data parameters to the user terminal if there are corresponding cyclone separator target database data parameters, and transmitting the cyclone separator parameters transmitted by the user terminal to the numerical simulation module 2 if there are no corresponding cyclone separator target database data parameters.

[0113] Preferably, in this embodiment, the registration submodule 11 is responsible for collecting the facial image of the user to be registered, extracting the facial features of the facial image, generating a facial feature vector, and reducing the dimensionality of the facial vector to generate a further facial feature vector. Next, the feature vector is calibrated. If it meets the set conditions, the feature vector is input into the feature library, and registration is completed. If it does not meet the conditions, a similarity comparison and identity authentication are performed. The uniqueness and security of the user's identity are improved through a multi-step verification process, enhancing the system's accuracy and reliability. The operation submodule 12 is responsible for responding to data parameters initiated by the user terminal, retrieving the cyclone separator target database, determining the data parameters corresponding to the cyclone separator target database, and checking if there are corresponding data parameters in the cyclone separator target database. The judgment result is output to the data transmission submodule 13. The accuracy and timeliness of the data are improved through an efficient database query and data transmission mechanism, enhancing the system's response speed and data processing capabilities. The data transmission submodule 13 is responsible for outputting the data parameters to the user terminal if the corresponding cyclone separator target database exists; otherwise, it transmits the cyclone separator parameters from the user terminal to the numerical simulation module 2. This flexible data transmission mechanism ensures accurate data transfer and processing, improving the overall system efficiency and user experience (see appendix for details). Figure 4 ).

[0114] In summary, this embodiment has jointly constructed an efficient, accurate, and secure face recognition and data processing system.

[0115] Furthermore, such as Figure 5 As shown, the data transmission submodule 13 specifically includes:

[0116] The generation unit 131 is responsible for generating control commands from data parameters in the target database that do not have a corresponding cyclone separator, processing the control commands, and returning if the command is incorrect, and prompting the user terminal that the input control command is incorrect.

[0117] Command transmission unit 132 is responsible for acquiring user terminal control commands and storing the control commands; processing the transmitted data according to the control commands; and running a compilation device to compile the transmitted data.

[0118] The parsing unit 133 is responsible for transmitting the compilation results to the numerical simulation module 2; during the transmission process, the transmission data is sent to the numerical simulation module 2 through the pin unit; the numerical simulation module 2 performs control simulation based on the compilation results.

[0119] Preferably, in this embodiment, the generation unit 131 generates and processes control commands to ensure their accuracy and effectiveness. If an instruction is incorrect, the module returns and prompts the user terminal that the input control command is incorrect, thereby improving the system's reliability and the accuracy of user operation. The command transmission unit 132 acquires and stores the user terminal's control commands, and then processes the transmitted data according to the commands. The transmitted data is compiled by running a compilation device to ensure data integrity and consistency. The parsing unit 133 transmits the compilation results to the numerical simulation module 2 and sends the transmitted data to the numerical simulation module 2 via a pin unit; the numerical simulation module 2 performs control simulation based on the compilation results, thereby achieving precise control and simulation of the system.

[0120] In summary, the generation unit ensures the accuracy and reliability of instructions; the command transmission unit ensures the integrity and consistency of data; and the parsing unit enables precise control and simulation of the system.

[0121] Furthermore, such as Figure 6 As shown, numerical simulation module 2 specifically includes:

[0122] The first preprocessing submodule 21 is responsible for preprocessing the relevant parameters of the cyclone separator imported from the user terminal, drawing the geometric model of the cyclone separator using SolidWorks based on the preprocessing results, preprocessing the relevant parameters of the cyclone separator, and meshing the cyclone separator model.

[0123] The simulation submodule 22 is responsible for setting the relevant parameters of the simulation equipment based on the preprocessed cyclone separator parameters; simulating the gas-solid two-phase flow inside the cyclone separator to determine the flow mode of the cyclone separator; and calculating the separation efficiency based on the flow mode.

[0124] The simulation result transmission submodule 23 is responsible for comparing the separation efficiency with the separation efficiency of historical experiments, evaluating the separation efficiency, and sending the simulation results, separation efficiency, and evaluation results to the deep learning module 3.

[0125] Preferably, in this embodiment, the first preprocessing submodule 21 preprocesses the relevant parameters of the cyclone separator imported from the user terminal, ensuring the accuracy and consistency of the parameters. The preprocessed parameters are used for subsequent geometric model drawing and mesh generation, providing reliable basic data for subsequent simulation. Furthermore, drawing the geometric model of the cyclone separator using SolidWorks provides a visual representation of the separator's structure, facilitating subsequent analysis and optimization. The simulation submodule 22 sets the relevant parameters of the simulation equipment based on the preprocessed cyclone separator parameters and uses computational fluid dynamics software to simulate the gas-solid two-phase flow inside the cyclone separator. During the simulation, the flow mode of the cyclone separator is determined, and the separation efficiency is calculated based on the flow mode. This helps in understanding the working principle and performance of the cyclone separator, providing a basis for optimized design. The simulation result transmission submodule 23 sends the simulation results, separation efficiency, and evaluation results to the deep learning module 3. By comparing the separation efficiency with historical experimental results, the separation efficiency is evaluated, thereby optimizing the design and performance of the cyclone separator. Artificial intelligence technology, especially deep learning, is used to predict and optimize the final quality of products, improving design efficiency and accuracy. (See appendix for details on the principles.) Figure 7 )

[0126] In summary, each module mainly focuses on data preprocessing, simulation, result evaluation, and optimization design. Through these technical means, the design efficiency and performance of cyclone separators can be effectively improved.

[0127] Furthermore, such as Figure 8 As shown, the first preprocessing submodule 21 specifically includes

[0128] The drawing unit 211 is responsible for verifying the relevant parameters of the cyclone separator imported by the user terminal, comparing the relevant parameters with standard data, verifying the accuracy and completeness of the relevant parameters, and drawing the geometric model of the cyclone separator based on the relevant data after verification.

[0129] Mesh generation unit 212 is responsible for dividing the geometric model of the cyclone separator into multiple non-overlapping sub-regions, specifying the node positions of each sub-region and the control volume represented by each node; setting mesh parameters and generating the mesh.

[0130] The mesh verification unit 213 is responsible for checking the mesh after it is generated; if the mesh is qualified, it outputs the mesh to the simulation submodule 22; if the mesh is unqualified, it checks the unqualified area and re-meshes the area.

[0131] Preferably, in this embodiment, the drawing unit 211 compares the relevant parameters of the cyclone separator imported by the user terminal with standard data to ensure the accuracy and completeness of the parameters. This provides a reliable data foundation for subsequent geometric model drawing. Based on the verified relevant data, the geometric model of the cyclone separator is drawn. The accuracy and consistency of the model provide an accurate geometric foundation for subsequent simulation analysis. The meshing unit 212 divides the geometric model of the cyclone separator into multiple non-overlapping sub-regions, with clear node positions and control volumes for each sub-region. This improves the accuracy and computational efficiency of the mesh; by setting appropriate mesh parameters, a high-quality mesh is generated. The uniformity and consistency of the mesh improve the accuracy of the simulation results. The purpose of meshing is to divide the computational domain into small discrete units, i.e., mesh cells, to facilitate numerical calculations of fluid flow. After discretizing the computational domain, physical quantities within each mesh cell, such as velocity, pressure, and temperature, can be calculated using numerical methods. Simultaneously, meshing can also describe the complexity of the flow field; complex flow field phenomena such as curvature, suspended objects, and boundary layers can be described through reasonable meshing. After generating the mesh, mesh verification unit 213 checks the mesh quality to ensure it is up to standard. This includes checking parameters such as mesh distortion and orthogonality. If the mesh is substandard, the substandard areas are checked, and the mesh for those areas is re-generated. The overall mesh quality is assessed to avoid simulation result deviations caused by localized mesh quality issues.

[0132] In summary, this embodiment improves the accuracy and reliability of cyclone separator simulation analysis through steps such as parameter verification, geometric model drawing, mesh generation, and mesh verification. This not only enhances the precision of the simulation results but also optimizes the efficiency of computational resource utilization.

[0133] Furthermore, such as Figure 9 As shown, simulation 22 specifically includes:

[0134] Flow determination unit 221 is responsible for determining the flow type based on the target parameters of the cyclone separator after grid division;

[0135] Formula for determining flow type:

[0136]

[0137] Where ρ is the fluid density, U is the fluid flow velocity, L is the characteristic dimension, and μ is the fluid dynamic viscosity; for the cyclone separator of the present invention, when Re>4000, it is turbulent flow; otherwise, it is swirling flow.

[0138] The flow calculation unit 222 is responsible for predicting the kinematic characteristics of the swirling flow within the three-dimensional spray gun using discretized Navier-Stokes equations, assuming the gas phase flow field is steady-state, compressible, and turbulent. When no other external forces act on the flow field, the continuity and momentum conservation equations of the swirling flow are as follows:

[0139]

[0140] Where u, ρ, P, and g represent the velocity vector, density, pressure, and gravitational acceleration of the gas phase in the computational domain, respectively. The symbol v represents the kinematic viscosity of the gas phase, and vt represents the viscosity of the gas flow.

[0141] Turbulent closure is performed using SST k-ω as the unknown term in the Navier-Stokes equations;

[0142] The formula for calculating turbulence energy (kg) is:

[0143]

[0144] The formula for calculating the eddy current dissipation rate (ωg) is:

[0145]

[0146] In the formula, ρ g The density of a fluid (unit: kg / m³) 3 ), k g u represents turbulent kinetic energy (unit: J / kg). j u represents the j-th component of velocity (in m / s). j The turbulence generation term is represented by W / kg, β represents the eddy current dissipation rate and the turbulent kinetic energy ratio, and ω represents the turbulence generation term. g Eddy dissipation rate (unit: s) -1 ), μ g μ represents turbulent viscosity (unit: Pa·s). t σ represents turbulent stress (unit: Pa). k Prandtl number, representing turbulent kinetic energy, is used to control the eddy viscosity coefficient, α. g P represents the eddy current dissipation rate and the turbulent kinetic energy ratio. a β represents the eddy current dissipation term (unit: W / kg). b σ represents the self-dissipation coefficient of eddy current dissipation rate. a Prandtl number represents the eddy dissipation rate, and F1 represents an empirical coefficient in the eddy dissipation rate model.

[0147] The separation efficiency unit 223 is responsible for calculating the separation efficiency based on the above calculation results;

[0148] The formula for calculating the separation efficiency η is as follows:

[0149]

[0150] Among them, M in Mout is the mass of particles flowing into the cyclone separator inlet, and Mout is the mass of particles flowing out of the cyclone separator outlet.

[0151] Preferably, the turbulence in this embodiment is a highly random, instantaneous flow state, with the flow velocity and direction changing constantly. The energy dissipation process of turbulence involves the separation of large eddies into many smaller eddies. For these very small eddies, energy dissipation occurs, converting kinetic energy into internal energy. Accurately simulating turbulent flow requires a vast amount of computation and is extremely time-consuming. In engineering applications, we typically focus on the average flow state of the fluid over a period of time. Therefore, time-averaging is used in engineering calculations to obtain the time-averaged solution. In the numerical simulation, discretized Navier-Stokes equations are used to predict the motion characteristics of the swirling flow within the three-dimensional spray gun, assuming the gas phase flow field is steady-state, compressible, and turbulent. When no other external forces act on the flow field, the turbulence model plays a crucial role in the accuracy of predicting the motion characteristics of the swirling flow within the spray gun. Large eddy simulation (LES) and direct numerical simulation (DNS) are two numerical models that can accurately simulate fluid flow. However, due to their high computational cost, both of these models typically require substantial computational resources, and it is difficult to set sufficient grid resolution to adequately assess the turbulent motion characteristics within industrial equipment. The Reynolds-averaged Navier-Stokes (RANS) equations substitute the average and random components of velocity and pressure into the Navier-Stokes equations to predict gas-phase turbulent flow. As two representative flow models within RANS, the Reynolds stress model (RSM) and the shear stress transport (SST) k-ω model are often used to predict the dynamic characteristics of swirling flows. The RSM model demonstrates accurate predictive capabilities in solving axial and swirling motions, Reynolds stress anisotropy, streamline curvature, and swirling flow effects. However, because the RSM involves solving six stress transport equations and one turbulent dissipation equation, its computational cost is significantly higher than that of the SST k-ω model, which only requires solving two equations. The SST model combines the advantages of the k-ε and k-ω methods in predicting free flow and viscous sublayers. In summary, this model, while maintaining low computational cost, also possesses the ability to simulate rapid strain flows and track the flow characteristics of complex swirling systems. This system uses SST k-ω as the unknown term in the Navier-Stokes equations for turbulent closure. Turbulent kinetic energy (kg) and eddy current dissipation rate (ωg) are two important variables in the SST k-ω model. The simulation calculation program runs on a server with the software configured as described above.

[0152] In summary, this embodiment calculates the separation efficiency based on the calculation results. The technology provides a large number of working conditions with different particle sizes, different design parameters, and different feeding conditions, and imports each working condition into the numerical simulation module for calculation, and evaluates the effect of the calculation results for each working condition.

[0153] Furthermore, such as Figure 10 As shown, Deep Learning Module 3 specifically includes:

[0154] The second preprocessing submodule 31 is responsible for acquiring simulation results and separation efficiency, preprocessing the simulation results and separation efficiency to form a dataset, sorting the dataset according to timestamps to obtain a time series dataset, and dividing the time series dataset into training set, test set and validation set.

[0155] The flow prediction model submodule 32 is responsible for training the flow prediction model based on time series data input, using environmental data information from simulation results as input and cyclone separator design parameter data as output, extracting the relationship between component deformation and changes in the surrounding environment, and finally obtaining the trained flow prediction model.

[0156] The first optimization submodule 33 is responsible for comparing the simulation results with the prediction results, identifying the areas that need to be optimized in the flow prediction model, continuously adjusting the parameters of the optimization areas until the maximum number of iterations is reached or the qualified area is reached; and feeding back the adjusted optimization area parameters to the user terminal.

[0157] Preferably, in this embodiment, the second preprocessing submodule 31 obtains simulation results and separation efficiency, and preprocesses the data to form a dataset. Then, the dataset is sorted according to timestamps to obtain a time-series dataset. This helps ensure the temporal consistency of the data and provides a reliable data foundation for subsequent model training. The time-series dataset is divided into training, testing, and validation sets. This helps evaluate the model's performance on different datasets and its generalization ability. Through steps such as data cleaning, data integration, data reduction, and data transformation, the quality and usability of the data are improved. For example, using a sliding window algorithm to extract time-series data and construct the dataset required for model training can effectively extract key information and reduce redundant data. The flow prediction model submodule 32 uses environmental data information from the simulation results as input and cyclone separator design parameter data as output to train the flow prediction model. This combination helps the model better understand the relationship between environmental changes and design parameters, thereby improving prediction accuracy. The flow prediction model is trained by extracting the relationship between component deformation and changes in the surrounding environment. The training method in this embodiment helps the model learn complex nonlinear relationships and improves prediction accuracy. The first optimization submodule 33 compares the simulation results with the prediction results to identify the areas where the flow prediction model needs optimization. The comparison method in this embodiment helps optimize the model's performance in practical applications. The parameters of the optimization area are continuously adjusted until the maximum number of iterations or a qualified area is reached. This adjustment method in this embodiment helps improve the model's adaptability under different operating conditions and enhances its robustness. The adjusted optimization area parameters are fed back to the user terminal, allowing the user to understand the model's optimization progress in real time and make further adjustments and optimizations based on the feedback. (For detailed principles, please refer to the appendix.) Figure 11 )

[0158] In summary, this embodiment mainly focuses on data preprocessing, model training and optimization, and parameter adjustment and feedback, which helps to improve the accuracy and robustness of the model, thereby better serving practical applications.

[0159] Furthermore, such as Figure 12 As shown, the second preprocessing submodule 31 specifically includes:

[0160] The third preprocessing unit 311 is responsible for extracting relevant features, reducing the dimensionality and complexity of the data structure using a dimensionality reduction device, and normalizing the extracted data so that the data have similar scale and distribution.

[0161] The feature extraction unit 312 is responsible for extracting the timestamps of each data point, arranging the normalized data in sequence according to the timestamps to form time series data, and setting the time series data as a time series dataset.

[0162] The dataset partitioning unit 313 is responsible for partitioning the time series dataset into a training set, a test set, and a validation set; and sending the partitioned training set, test set, and validation set of the time series dataset to the flow prediction model submodule 32.

[0163] Preferably, the third preprocessing unit 311 in this embodiment effectively reduces data complexity by reducing dimensionality, making data processing more efficient and convenient. This improves the speed and efficiency of subsequent model training. Normalization processes data to have similar scales and distributions, eliminating the influence of different features on their dimensions, and improving data comparability and training efficiency. The feature extraction unit 312 extracts the timestamps of each data point and arranges them sequentially according to the timestamps to form time series data, which helps to capture the time dependencies of the data and provides a foundation for subsequent time series analysis. Setting the time series data as a time series dataset facilitates subsequent training and prediction of the time series model. The dataset partitioning unit 313 divides the time series dataset into a training set, a test set, and a validation set. The time period of the test set is after the time period of the training set to simulate the prediction situation in real-world scenarios. Data partitioning should ensure that the data distribution of each subset is as consistent as possible to ensure that the model can perform well under various conditions. By partitioning the dataset, the generalization ability of the model can be better evaluated, and the model can perform well even on unseen data.

[0164] In summary, this embodiment improves data processing efficiency and model training efficiency by reducing data dimensionality and normalizing the data; it provides a foundation for time series analysis by extracting timestamps and setting up time series datasets; and it enhances the model's generalization ability and prediction accuracy by rationally dividing the dataset.

[0165] Furthermore, such as Figure 13 As shown, the flow prediction model submodule 32 specifically includes:

[0166] Structural unit 321 is responsible for inputting time series data into the flow prediction model, setting the flow prediction model architecture as an input layer, a hidden layer, and an output layer; using the pipe diameter data from the simulation results as input, the hidden layer performs feature extraction and processing, and the cyclone separator design parameter data as output for training the flow prediction model, using minimum batch data for iterative training;

[0167] Training unit 322 is responsible for validating the process of training the cyclone separator design parameters using the validation set; if there is no improvement for several consecutive cycles, training is stopped; the relationship between the structure deformation and environmental parameters is analyzed; the influence of key pipe diameter parameters on the structure deformation is extracted, and a relevant relationship data table is generated;

[0168] Test model unit 323 is responsible for adjusting the relevant hyperparameters of the flow prediction model based on the relevant relational data table, and finally obtaining the final flow prediction model.

[0169] Preferably, the structural unit 321 in this embodiment, by setting up a flow prediction model architecture with an input layer, a hidden layer, and an output layer, can effectively extract and process features from the pipe diameter data of the simulation results. This improves the model's understanding and prediction capabilities for complex flow phenomena. Iterative training using minimum batch data can accelerate the model's convergence speed and improve its generalization ability. By using cyclone separator design parameter data as output, the model can more accurately predict parameters, thereby optimizing the design process. The training unit 322 uses a validation set to validate the model, allowing for dynamic adjustment of hyperparameters during training, ensuring consistent model performance across different datasets. If the model performance no longer improves within several consecutive cycles, training automatically stops to avoid overfitting and improve the model's stability and reliability. By analyzing the relationship between deformation and environmental parameters, the influence of environmental factors on deformation can be better understood, providing a basis for subsequent design and optimization. The testing model unit adjusts the relevant hyperparameters of the flow prediction model according to the relevant relationship data table, further optimizing the model's performance and improving prediction accuracy. Through the above steps, the final flow prediction model has high accuracy and generalization ability, and can be effectively applied in practical engineering.

[0170] In summary, this embodiment mainly focuses on improving the model's feature extraction capabilities, accelerating the training process, optimizing design parameter prediction, dynamically adjusting hyperparameters, analyzing the impact of environmental parameters, and ultimately improving the model's prediction accuracy and reliability.

[0171] Furthermore, such as Figure 14 As shown, the first optimization submodule 33 specifically includes:

[0172] The identification and optimization unit 331 is responsible for comparing the simulation results with the prediction results, identifying the area that needs to be optimized in the flow prediction model, and adjusting the area through continuous iteration. After each iteration, the simulation and prediction are repeated until the maximum number of iterations or the qualified area is reached.

[0173] The second optimization unit 332 is responsible for transmitting the adjusted optimization area parameters to the user terminal through the cloud platform. After receiving the optimization parameters, the user terminal can further adjust the parameters or confirm the final cyclone separator according to the actual situation. The parameter adjustment includes: manual adjustment or re-requesting optimization.

[0174] Feedback unit 333 is responsible for receiving the re-request optimization operation initiated by the user terminal, retrieving the cyclone separator information from the database, and resimulating and predicting the cyclone separator according to the target parameters given by the user until it is qualified.

[0175] Preferably, the optimization unit 331 in this embodiment adjusts the structural parameters of the cyclone separator through continuous iteration to improve its separation efficiency and reduce operating pressure drop. Specifically, this unit compares simulation results with prediction results, identifies areas requiring optimization, and performs multiple iterative adjustments on these areas until a qualified area is reached or the maximum number of iterations is achieved. The second optimization unit 332 efficiently transmits the optimized parameters to the user terminal via a cloud platform, enabling the user to further adjust the parameters or confirm the final cyclone separator design based on actual conditions. The cloud platform's data transmission management can quickly and reliably handle large-scale data transmission demands, reducing data transmission latency and network bandwidth consumption. The feedback unit 333, when the user terminal initiates a re-request for optimization, allows the system to quickly retrieve cyclone separator information from the database and re-simulate and predict based on the user's revised target parameters until a qualified standard is achieved. The flexibility and adaptability of the cyclone separator design allow for rapid adjustment and optimization according to user needs.

[0176] In summary, this embodiment mainly demonstrates that it significantly improves the performance and design efficiency of the cyclone separator through iterative optimization, efficient data transmission, and a rapid feedback mechanism.

[0177] Furthermore, such as Figure 15 As shown, this embodiment also provides a machine learning-based cyclone separator optimization method. In this embodiment, a machine learning-based cyclone separator design and optimization method is applied to a machine learning-based cyclone separator design and optimization system as described in the above embodiment. This machine learning-based cyclone separator optimization method specifically includes the following steps:

[0178] Step S1: First-time login requires account registration. After successful login, input the relevant parameters of the cyclone separator through the user terminal operation interface; retrieve the data stored in the database based on the cyclone separator parameters input by the user terminal; determine whether the corresponding cyclone separator parameter data is stored; if it is, retrieve the database data and output it to the user terminal; if not, import the new data into the simulation model.

[0179] The parameters of the cyclone separator include cyclone separator parameters, working environment, and operating parameters; the working environment includes working pressure, working temperature, humidity, and air density; the operating parameters include inlet velocity, particle size, particle density, particle viscosity, outlet pressure, and inlet pressure.

[0180] Step S2: Receive relevant parameters of the cyclone separator imported from the user terminal; draw the geometric model of the cyclone separator using SolidWorks according to the design parameters of the cyclone separator; simulate the cyclone separator model and calculate the separation efficiency; and preprocess the separator simulation results and separation efficiency.

[0181] Step S3: Establish a dataset based on the preprocessing results; sort the dataset according to timestamps, and train the flow prediction model based on the sorted data; combine simulation results and prediction results to identify the areas that need to be optimized in the flow prediction model; continuously iterate and adjust the optimization areas to obtain the final cyclone separator parameters; and feed the adjustments back to the user terminal.

[0182] Preferably, in step S1, initial login requires account registration. After successful login, users input relevant parameters of the cyclone separator through the user terminal interface. User identity verification and accurate parameter input provide a foundation for subsequent data processing and simulation. Based on the cyclone separator parameters input by the user terminal, the system retrieves data stored in the database and determines whether the corresponding cyclone separator parameter data is stored. If it matches, the database data is retrieved and output to the user terminal; otherwise, the new data is imported into the simulation model. The use of the database enables rapid data retrieval and storage of new data, improving system response speed and data management efficiency. Step S2 receives the cyclone separator parameters imported from the user terminal and uses SolidWorks to draw a geometric model of the cyclone separator based on its design parameters. Through 3D modeling technology, precise design and visualization of the cyclone separator are achieved. The cyclone separator model is simulated to calculate the separation efficiency, and the simulation results and separation efficiency are preprocessed. Through numerical simulation and modeling techniques, the performance of the cyclone separator is evaluated, providing data support for subsequent optimization. Step S3 establishes a dataset based on the preprocessing results, sorts the dataset by timestamp, trains a flow prediction model using the sorted data, and identifies the areas requiring optimization in the flow prediction model by combining simulation and prediction results. The optimization areas are then iteratively adjusted to obtain the final cyclone separator parameters, which are then fed back to the user terminal. Through the establishment of the dataset and the training of the flow prediction model, continuous optimization and feedback of the cyclone separator performance are achieved, improving the system's adaptability and performance.

[0183] In summary, this embodiment achieves a comprehensive improvement and optimization of cyclone separator performance through the synergistic effect of multiple technical aspects, including data storage and retrieval, geometric model drawing and simulation, data preprocessing and model training, structural optimization and iterative adjustment, and user feedback and terminal interaction.

[0184] like Figure 16As shown, this embodiment provides an embodiment of an electronic device 8, which includes a processor 81 and a memory 82 coupled to the processor 81.

[0185] The memory 82 stores program instructions for implementing the layout method of the chemical pump body processing equipment of any of the above embodiments.

[0186] The processor 81 is used to execute program instructions stored in the memory 82 for the layout of chemical pump body processing equipment.

[0187] The processor 81 can also be referred to as a CPU (Central Processing Unit). The processor 81 may be an integrated circuit chip with signal processing capabilities. The processor 81 can also be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. A general-purpose processor can be a microprocessor or any conventional processor.

[0188] Furthermore, Figure 17 This is a schematic diagram of the structure of a storage medium according to an embodiment of this application. The storage medium 9 of this embodiment stores program instructions 91 capable of implementing all the methods described above. These program instructions 91 can be stored in the storage medium in the form of a software product, including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks, or terminal devices such as computers, servers, mobile phones, and tablets.

[0189] In the several embodiments provided by this invention, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection between apparatuses or units through some interfaces, and may be electrical, mechanical, or other forms.

[0190] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated units described above can be implemented in hardware or as software functional units. The above are merely embodiments of the present invention and do not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

[0191] The specific embodiments of the invention have been described in detail above, but these are merely examples, and the invention is not limited to the specific embodiments described above. For those skilled in the art, any equivalent modifications or substitutions to the invention are also within the scope of this invention. Therefore, all equivalent transformations, modifications, and improvements made without departing from the spirit and principles of this invention should be included within its scope.

Claims

1. A cyclone separator optimization system based on machine learning, characterized in that, The machine learning-based cyclone separator optimization system includes: The user operation module is responsible for handling user account registration upon first login. After successful login, users can input relevant parameters of the cyclone separator through the user terminal interface. Based on the user's input of relevant parameters of the cyclone separator, the module retrieves data stored in the database. It then determines whether the corresponding cyclone separator parameter data is stored. If so, the database data is retrieved and output to the user terminal. If not, the new data is imported into the numerical simulation module for calculation. The parameters related to the cyclone separator include: cyclone separator design parameters, working environment, and operating parameters; the working environment includes working pressure, working temperature, humidity, and air density; the operating parameters include inlet velocity, particle size, particle density, particle viscosity, outlet pressure, and inlet pressure. The numerical simulation module is responsible for receiving relevant parameters of the cyclone separator imported from the user terminal; drawing the geometric model of the cyclone separator using SolidWorks based on the design parameters of the cyclone separator; simulating the cyclone separator model and calculating the separation efficiency; and sending the separator simulation results and separation efficiency to the deep learning module. The deep learning module is responsible for preprocessing simulation results and separation efficiency data to build a dataset; sorting the dataset according to timestamps; training the flow prediction model based on the sorted data; identifying the areas that need optimization in the flow prediction model by combining simulation results and prediction results; iteratively adjusting the optimization areas to obtain the final cyclone separator parameters; and feeding the adjustments back to the user terminal. The numerical simulation module includes: The first preprocessing submodule is responsible for preprocessing the relevant parameters of the cyclone separator imported from the user terminal, drawing the geometric model of the cyclone separator using SolidWorks based on the preprocessing results, preprocessing the relevant parameters of the cyclone separator, and meshing the cyclone separator model. The simulation submodule is responsible for setting the relevant parameters of the simulation equipment based on the preprocessed cyclone separator parameters; simulating the gas-solid two-phase flow inside the cyclone separator to determine the flow mode of the cyclone separator; and calculating the separation efficiency based on the flow mode. The simulation results transmission submodule is responsible for comparing the separation efficiency with that of historical experiments, evaluating the separation efficiency, and sending the simulation results, separation efficiency, and evaluation results to the deep learning module.

2. The cyclone separator optimization system based on machine learning according to claim 1, characterized in that, The user operation module includes: The registration submodule is responsible for collecting facial images of users to be registered, extracting the face features of the facial images, generating facial feature vectors, reducing the dimensionality of the facial vectors, calibrating the feature vectors, and if the set conditions are met, inputting the feature vectors into the feature database and completing the registration. If the set conditions are not met, the feature vector will not be input into the feature database, and the registration will not be completed. The process involves comparing the similarity between the associated feature vector of the user to be registered and all key features in the feature database of the registered user. If the similarity meets the set conditions, a temporary identity is created. If the similarity does not meet the set conditions, identity authentication is performed. If identity authentication fails, the user's facial information is collected again. The operation submodule is responsible for responding to data parameters initiated by the user terminal, retrieving the cyclone separator target database, determining the data parameters in the cyclone separator target database corresponding to the data parameters initiated by the user terminal, checking whether there are corresponding data parameters in the cyclone separator target database, and outputting the judgment result to the data transmission submodule. The data transmission submodule is responsible for outputting the data parameters to the user terminal if there are corresponding cyclone separator target database data parameters, and transmitting the cyclone separator parameters transmitted by the user terminal to the numerical simulation module if there are no corresponding cyclone separator target database data parameters. The generation unit is responsible for generating control commands from data parameters in the target database that do not have a corresponding cyclone separator, processing the control commands, and returning if the command is incorrect, prompting the user terminal that the input control command is incorrect. The command transmission unit is responsible for acquiring user terminal control commands and storing them; processing the transmitted data according to the control commands; and running the compilation device to compile the transmitted data. The parsing unit is responsible for transmitting the compilation results to the numerical simulation module. During the transmission process, the data is sent to the numerical simulation module through the pin unit. The numerical simulation module performs control simulation based on the compilation results.

3. The cyclone separator optimization system based on machine learning according to claim 1, characterized in that, The first preprocessing submodule includes: The drawing unit is responsible for verifying the relevant parameters of the cyclone separator imported from the user terminal, comparing the relevant parameters with standard data, verifying the accuracy and completeness of the relevant parameters, and drawing the geometric model of the cyclone separator based on the relevant data after verification. The mesh generation unit is responsible for dividing the geometric model of the cyclone separator into multiple non-overlapping sub-regions, defining the node positions of each sub-region, the control volume represented by each node, setting mesh parameters, and generating the mesh. The mesh verification unit is responsible for checking the mesh after it is generated. If the mesh is qualified, it is output to the simulation submodule. If the mesh is unqualified, the unqualified area is checked and the mesh is re-generated in that area.

4. The cyclone separator optimization system based on machine learning according to claim 1, characterized in that, The simulation submodule includes: The flow determination unit is responsible for determining the flow type based on the target parameters of the cyclone separator after mesh generation. Formula for determining flow type: Where ρ is the fluid density, U is the fluid velocity, L is the characteristic dimension, and μ is the fluid dynamic viscosity; for a cyclone separator, when Re > 4000, it is turbulent flow; otherwise, it is steady flow. The flow calculation unit is responsible for predicting the kinematic characteristics of the swirling flow within the three-dimensional spray gun using the discretized Navier-Stokes equations, assuming that the gas phase flow field is steady-state, compressible, and turbulent. When no other external forces act on the flow field, the continuity and momentum conservation equations of the swirling flow are as follows: in, ρ, P and These represent the velocity vector, density, pressure, and gravitational acceleration vector of the gas phase in the computational domain, respectively; the symbol v represents the kinematic viscosity of the gas phase, v t Indicates the viscosity of the gas flow; Turbulent closure is performed using SST k-ω as the unknown term in the Navier-Stokes equations; Calculate the turbulent kinetic energy k g The formula is: Calculate the eddy current dissipation rate ω g The formula is: In the formula, ρ g k represents the density of the fluid. g U represents turbulent kinetic energy. j u represents the j-th component of velocity. j ω represents the turbulence generation term, β represents the eddy current dissipation rate and the turbulent kinetic energy ratio, and ω represents the turbulence generation term. g μ represents the eddy current dissipation rate. g μ represents turbulent viscosity. t σ represents turbulent stress. k Prandtl number, representing turbulent kinetic energy, is used to control the eddy viscosity coefficient, α. g P represents the eddy current dissipation rate and the turbulent kinetic energy ratio. a β represents the eddy current dissipation term. b σ represents the self-dissipation coefficient of eddy current dissipation rate. a Prandtl number represents the eddy dissipation rate, and F1 represents an empirical coefficient in the eddy dissipation rate model. The separation efficiency unit is responsible for calculating the separation efficiency based on the calculation results. The formula for calculating the separation efficiency is as follows: Among them, M in M represents the mass of particles flowing into the cyclone separator inlet. out This refers to the mass of particles flowing out from the upper outlet of the cyclone separator.

5. The cyclone separator optimization system based on machine learning according to claim 1, characterized in that, The deep learning module includes: The second preprocessing submodule is responsible for acquiring simulation results and separation efficiency, preprocessing the simulation results and separation efficiency to form a dataset, sorting the dataset according to timestamps to obtain a time series dataset, and dividing the time series dataset into training set, test set and validation set. The flow prediction model submodule is responsible for training the flow prediction model by inputting time series data into the flow prediction model, taking environmental data information from the simulation results as input, and cyclone separator design parameter data as output, extracting the relationship between component deformation and changes in the surrounding environment, and finally obtaining the trained flow prediction model. The first optimization submodule is responsible for comparing the simulation results with the prediction results, identifying the areas that need optimization in the flow prediction model, continuously adjusting the parameters of the optimization areas until the maximum number of iterations is reached or a qualified area is reached; and feeding back the adjusted optimization area parameters to the user terminal.

6. The cyclone separator optimization system based on machine learning according to claim 5, characterized in that, The second preprocessing submodule includes: The third preprocessing unit is responsible for extracting relevant features, reducing the dimensionality and complexity of the data structure using dimensionality reduction devices, and normalizing the extracted data to ensure that the data have similar scale and distribution. The feature extraction unit is responsible for extracting the timestamps of each data point, arranging the normalized data in order according to the timestamps to form time series data, and setting the time series data as a time series dataset. The dataset partitioning unit is responsible for dividing the time series dataset into training, test, and validation sets, and then sending the divided training, test, and validation sets of the time series dataset to the flow prediction model submodule.

7. The cyclone separator optimization system based on machine learning according to claim 6, characterized in that, The flow prediction model submodule includes: The structural unit is responsible for inputting time series data into the flow prediction model and setting the flow prediction model architecture as an input layer, a hidden layer, and an output layer. The pipe diameter data from the simulation results is used as input, the hidden layer performs feature extraction and processing, and the cyclone separator design parameter data is used as output for training the flow prediction model. Iterative training is performed using minimum batch data. The training unit is responsible for validating the process of training the cyclone separator design parameters using the validation set; if the model no longer improves within several consecutive cycles, the training is stopped; the relationship between the structure deformation and environmental parameters is analyzed; the influence of key pipe diameter parameters on the structure deformation is extracted, and a relevant relationship data table is generated; The test model unit is responsible for adjusting the relevant hyperparameters of the flow prediction model based on the relevant relational data table, and finally obtaining the final flow prediction model.

8. The cyclone separator optimization system based on machine learning according to claim 5, characterized in that, The first optimization submodule includes: The identification and optimization unit is responsible for comparing the simulation results with the prediction results, identifying the areas that need to be optimized in the flow prediction model, and adjusting these areas through continuous iteration. After each iteration, the simulation and prediction are repeated until the maximum number of iterations or a qualified area is reached. The second optimization unit is responsible for transmitting the adjusted optimization area parameters to the user terminal through the cloud platform. After receiving the optimization area parameters, the user terminal can further adjust the parameters or confirm the final cyclone separator according to the actual situation. The parameter adjustment includes: manual adjustment or re-requesting optimization. The feedback unit is responsible for receiving the re-request for optimization initiated by the user terminal, retrieving the optimization area parameters from the database, and resimulating and predicting the cyclone separator model based on the target parameters given by the user until it meets the requirements.

9. A machine learning-based cyclone separator optimization method, applied to a machine learning-based cyclone separator optimization system as described in any one of claims 1 to 8, characterized in that, The machine learning-based cyclone separator optimization method includes: First-time login requires account registration. After successful login, users can input relevant parameters of the cyclone separator through the user terminal interface. Based on the input parameters of the cyclone separator through the user terminal, the system retrieves the data stored in the database. It then determines whether the corresponding cyclone separator parameter data is stored. If yes, the database data is retrieved and output to the user terminal. If no, the new data is imported into the numerical simulation module. The relevant parameters of the cyclone separator include the cyclone separator design parameters, working environment, and operating parameters; the working environment includes working pressure, working temperature, humidity, and air density; the operating parameters include inlet velocity, particle size, particle density, particle viscosity, outlet pressure, and inlet pressure. Receive relevant parameters of the cyclone separator imported from the user terminal; draw the geometric model of the cyclone separator using SolidWorks according to the design parameters of the cyclone separator; simulate the cyclone separator model and calculate the separation efficiency; and preprocess the separator simulation results and separation efficiency. A dataset is established based on the preprocessing results; the dataset is sorted according to timestamps, and the flow prediction model is trained based on the sorted data; the flow prediction model needs to be optimized by combining simulation results and prediction results; the optimization area is continuously adjusted iteratively to obtain the final cyclone separator parameters; the adjustments are fed back to the user terminal.

Images