Valve electric actuating mechanism design parameter optimization method based on deep learning

Abstract

The invention discloses a valve electric actuating mechanism design parameter optimization method based on deep learning, and the method comprises the steps: carrying out the simulation analysis of a valve electric actuating mechanism under different design parameters through finite element calculation, and generating sample data of output rotating speed and stroke time; constructing a deep learning model based on the sample data, and predicting an output rotating speed and travel time; constructing a multi-objective optimization model, and optimizing the output rotating speed and the travel time by using an MOEA / D algorithm to obtain a Pareto optimal solution; and determining a weight coefficient between the targets by using an AHP-entropy weight method, and finally carrying out sorting decision on the Pareto optimal solution by using a TOPSIS method to obtain an optimal scheme. According to the invention, based on deep learning, a parameter combination with optimized output rotating speed and travel time is searched through multi-objective optimization; the water hammer pressure peak value in the valve closing process is reduced, the response speed of the system is maximized, and the performance of the valve electric actuating mechanism is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of electric valves, and more specifically to a method for optimizing design parameters of valve electric actuators based on deep learning. Background Art

[0002] Electric actuators and valves are key components in fluid control systems. Electric actuators are devices that receive electrical signals and convert them into mechanical motion. They control the flow of fluids by driving valves. Valves are devices used to regulate fluid flow and pressure. Their closed state directly affects the operating efficiency and stability of the fluid system.

[0003] However, during the valve closing process, water hammer effect may occur due to the sudden stop or rapid change of direction of the fluid. The "water hammer effect" is an instantaneous high-pressure wave that propagates in the pipeline, causing impact and damage to the system. In order to reduce the damage to the valve caused by the "water hammer effect" generated when the valve is closed, some advanced electric actuators have a segmented speed regulation function, which performs fast closing when the stroke range is 100% to 20%; and slow closing when the stroke range is 20% to 0%. Although advanced electric actuators have reduced the water hammer effect through segmented speed regulation, these methods also ignore the importance of valve response speed; in the current existing technology, the design of valve electric actuator parameters has not yet considered the optimization of the output speed and stroke time of the electric actuator.

[0004] Therefore, in order to further reduce the water hammer pressure peak during valve closing and maximize the response speed of the system, it is necessary to optimize the output speed and travel time of the electric actuator to ensure that the water hammer effect is minimized while meeting the system response speed, thereby improving the reliability and safety of the entire system. Summary of the invention

[0005] In view of this, the present invention provides a valve electric actuator design parameter optimization method based on deep learning that solves at least some of the above-mentioned technical problems. The method uses multi-objective optimization based on deep learning to find a parameter combination that optimizes the output speed and travel time. The method can reduce the water hammer pressure peak during the valve closing process, maximize the response speed of the system, and improve the performance of the valve electric actuator.

[0006] To achieve the above object, the technical solution adopted by the present invention is:

[0007] An embodiment of the present invention provides a method for optimizing design parameters of a valve electric actuator based on deep learning, the method comprising:

[0008] S1. Use finite element calculation method to simulate and analyze the valve electric actuator under different design parameters to generate sample data of output speed and travel time;

[0009] S2. Building a deep learning model based on the sample data to predict the output speed and travel time of the valve electric actuator;

[0010] S3. Construct a multi-objective optimization model and use the MOEA / D optimization algorithm to optimize the output speed and travel time of the valve electric actuator to obtain the Pareto optimal solution; use the AHP-entropy weight method to determine the weight coefficients between each objective, and finally use the TOPSIS method to sort the Pareto optimal solution and obtain the optimal solution for the design parameters of the valve electric actuator.

[0011] Optionally, in step S1, the design parameters include:

[0012] Material properties: elastic modulus, density, Poisson’s ratio;

[0013] Geometric parameters: valve diameter, wall thickness, stroke length;

[0014] Operating conditions: working pressure, working temperature;

[0015] Electric actuator parameters: motor power, rated speed, torque.

[0016] Optionally, in step S1, comsol multiphysics software is used to perform simulation under different design parameters.

[0017] Optionally, in step S2, the deep learning model constructed includes:

[0018] Input layer: There are n input nodes in total;

[0019] Hidden layer: A multi-layer fully connected neural network is used, each layer contains a preset number of neurons and uses the ReLU activation function;

[0020] Output layer: predicts output speed and travel time, with 2 output nodes in total.

[0021] Optionally, the deep learning model is trained using stochastic gradient descent or Adam optimizer, and the loss function is mean square error.

[0022] Optionally, in step S3, the multi-objective optimization model constructed includes an objective function:

[0023] Minimize output speed error: f1 = |Vtarget-Voutput|

[0024] Minimize travel time error: f2 = |Ttarget-Toutput|

[0025] Where f1 represents the minimum output speed error, V target Indicates the target speed, V output represents the output speed; f2 represents the minimum travel time error, T target represents the target travel time, T output Indicates the output travel time.

[0026] Optionally, in step S3, the calculation formula of the MOEA / D optimization algorithm is:

[0027]

[0028] In the formula, λ i is the weight vector, z i is the reference point, f i (x) is the objective function value.

[0029] Optionally, in step S3, the weight coefficients between the objectives are determined using the AHP-entropy weight method, where:

[0030] The AHP method is:

[0031] Establish a hierarchical model, including the target layer, criterion layer and solution layer; construct a judgment matrix and calculate the subjective weight of each criterion; calculate the eigenvector of the judgment matrix and perform normalization; calculate the consistency ratio:

[0032]

[0033] in, λ max is the maximum eigenvalue of the judgment matrix, n is the order of the judgment matrix, and RI is the random consistency index;

[0034] The entropy weight method is:

[0035]

[0036] in, r ij is the value of the i-th solution under the j-th criterion, j is the number of solutions;

[0037] Determine the objective weight of each criterion:

[0038]

[0039] Combining the subjective weight of the AHP method, the comprehensive weight is obtained:

[0040]

[0041] Among them, α is the weight coefficient.

[0042] Optionally, in step S3, the TOPSIS method is used to perform a ranking decision on the Pareto optimal solution, and the specific process includes:

[0043] Construct a decision matrix: Take each solution in the Pareto optimal solution set as a decision plan and establish a decision matrix;

[0044] Standardized decision matrix: Use vector normalization method to standardize the decision matrix:

[0045]

[0046] Among them, x ij is the value of the i-th solution under the j-th criterion;

[0047] Calculate the weighted standardized decision matrix: Combine the weights obtained by the AHP-entropy weight method to calculate the weighted standardized decision matrix:

[0048]

[0049] Among them, v ij is the weighted normalized decision matrix;

[0050] Determine the positive and negative ideal solutions:

[0051] Positive ideal solution: the maximum value of each criterion:

[0052]

[0053] Negative ideal solution: Minimum of each criterion:

[0054]

[0055] In the formula, is the nth positive ideal solution, is the nth negative ideal solution, J is the set of all solutions;

[0056] Compute the Euclidean distance of each solution to the positive and negative ideal solutions:

[0057]

[0058] In the formula, is the positive ideal solution of the jth solution, is the negative ideal solution of the jth solution;

[0059] Calculate closeness:

[0060]

[0061] in, represents the Euclidean distance of the positive ideal solution, represents the Euclidean distance of the negative ideal solution;

[0062] Sorting: Sort the solutions according to the closeness Ci and select the best solution.

[0063] Compared with the prior art, the present invention has at least the following beneficial effects:

[0064] The present invention provides a method for optimizing design parameters of valve electric actuators based on deep learning, which uses multi-objective optimization based on deep learning to find the optimal parameter combination of output speed and travel time; it helps to reduce the water hammer pressure peak during valve closing, maximize the response speed of the system, and improve the performance of the valve electric actuator.

[0065] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description and the accompanying drawings.

[0066] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0068] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention.

[0069] Figure 1 A schematic flow chart of a method for optimizing design parameters of a valve electric actuator based on deep learning provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0070] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. In addition, various serial numbers are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0071] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0072] like Figure 1 As shown, the embodiment of the present invention discloses a method for optimizing design parameters of a valve electric actuator based on deep learning, and the method mainly includes:

[0073] S1. Through the finite element calculation method, the valve electric actuator under different design parameters is simulated and analyzed to generate a series of representative sample data of output speed and travel time;

[0074] S2. Use these sample data to build a deep learning model to predict the output speed and travel time of the valve electric actuator;

[0075] S3. Construct a multi-objective optimization model and use the MOEA / D optimization algorithm to optimize the output speed and travel time of the valve electric actuator to obtain the Pareto optimal solution; use the AHP-entropy weight method to determine the weight coefficients between each objective, and finally use the TOPSIS method to sort the Pareto optimal solution and obtain the optimal solution for the design parameters of the valve electric actuator.

[0076] The specific implementation methods of the method of the present invention are described in detail below:

[0077] 1. Finite element calculation technology:

[0078] In a specific embodiment, the design parameters include:

[0079] Material properties: elastic modulus, density, Poisson's ratio;

[0080] Geometric parameters: valve diameter, wall thickness, stroke length;

[0081] Operating conditions: working pressure, working temperature;

[0082] Electric actuator parameters: motor power, rated speed, torque;

[0083] Simulation analysis:

[0084] In this embodiment, Comsol is preferably used to perform simulation under different design parameters; the output speed and travel time are extracted as simulation results.

[0085] 2. Deep learning model construction:

[0086] In this embodiment, the sample data for constructing the deep learning model is: output speed and travel time data obtained through finite element analysis.

[0087] Model structure:

[0088] Input layer: includes material properties, geometric parameters, operating conditions and electric actuator parameters, with a total of n input nodes.

[0089] Hidden layer: A multi-layer fully connected neural network is used, each layer contains a certain number of neurons, and uses the ReLU activation function. For example, a three-layer hidden layer structure:

[0090] The first hidden layer: 64 neurons, activation function ReLU;

[0091] Second hidden layer: 32 neurons, activation function ReLU;

[0092] The third hidden layer: 16 neurons, activation function ReLU;

[0093] Output layer: predicts output speed and travel time, with 2 output nodes in total.

[0094] Training process: Stochastic gradient descent (SGD) or Adam optimizer is preferably used for training, and the loss function is mean square error (MSE).

[0095] 3. Multi-objective optimization model:

[0096] 1. Multi-objective optimization model:

[0097] Objective function:

[0098] Minimize output speed error: f1=|V target -V output |

[0099] Minimize travel time error: f2 = |T target -T output |

[0100] Where f1 represents the minimum output speed error, V target Indicates the target speed, V output represents the output speed; f2 represents the minimum travel time error, T target represents the target travel time, T output Indicates the output travel time.

[0101] Constraints:

[0102] In this embodiment, physical and operational range constraints on design parameters are imposed.

[0103] 2.MOEA / D algorithm:

[0104] Initialization: Generate an initial population with a population size of N; divide the weight vector and construct sub-problems.

[0105] Evolutionary process: For each sub-problem, select solutions in the neighborhood for crossover and mutation to generate new solutions; calculate the objective function value of the new solution.

[0106] In this embodiment, the Chebyshev decomposition method is preferably used to update the solution in the neighborhood. The formula of the Chebyshev decomposition method is:

[0107]

[0108] Among them, λ i is the weight vector, z i is the reference point, f i (x) is the objective function value.

[0109] Pareto optimal solution: Finally, the Pareto frontier solution set is obtained, which is the optimal solution for multi-objective optimization.

[0110] 3.AHP-Entropy Weight Method:

[0111] AHP method:

[0112] In this embodiment, a hierarchical model is established, including a target layer, a criterion layer, and a solution layer. The judgment matrix can be constructed by the expert scoring method to calculate the relative weight of each criterion. The eigenvector of the judgment matrix is ​​calculated and normalized. The consistency ratio (CR) is calculated:

[0113]

[0114] in, λ max is the maximum eigenvalue of the judgment matrix, n is the order of the judgment matrix, and RI is the random consistency index.

[0115] Entropy Weight Method:

[0116]

[0117] in, r ij is the value of the ith solution under the jth criterion, and j is the number of solutions.

[0118] Determine the objective weight of each criterion:

[0119]

[0120] Combining the subjective weight of the AHP method, the comprehensive weight is obtained:

[0121]

[0122] Among them, α is the weight coefficient.

[0123] 4.TOPSIS ranking decision:

[0124] Construct a decision matrix: Take each solution in the Pareto optimal solution set as a decision plan and establish a decision matrix.

[0125] Standardized decision matrix: Use vector normalization method to standardize the decision matrix:

[0126]

[0127] Among them, x ij is the value of the ith solution under the jth criterion.

[0128] Calculate the weighted standardized decision matrix: Combine the weights obtained by the AHP-entropy weight method to calculate the weighted standardized decision matrix:

[0129]

[0130] Determine the positive and negative ideal solutions:

[0131] Positive ideal solution: the maximum value of each criterion:

[0132]

[0133] Negative ideal solution: Minimum of each criterion:

[0134]

[0135] Calculate the distance of each solution to the positive and negative ideal solutions:

[0136] Compute the Euclidean distance of each solution to the positive and negative ideal solutions:

[0137]

[0138] Calculate closeness:

[0139]

[0140] Sorting: Sort the solutions according to their closeness and select the best one.

[0141] From the description of the above embodiments, those skilled in the art can know that the present invention provides a method for optimizing the design parameters of valve electric actuators based on deep learning, and uses Comsol Multiphysics software to simulate the output speed and travel time of the valve electric actuator to generate a series of sample data. Then a deep learning model is constructed. Using the MOEA / D algorithm, a multi-objective optimization model is established. In this model, in order to reduce the water hammer pressure peak during valve closing and maximize the response speed of the system, the output speed and travel time are set as optimization targets. In order to ensure the fairness and accuracy of the multi-objective optimization model, the AHP-entropy weight method is used to determine the weight coefficients between the various targets. The AHP-entropy weight method combines subjective judgment and objective data analysis, and can fully consider subjective and objective factors, making the weight coefficient more reasonable. Finally, the TOPSIS method is used to rank the Pareto optimal solution.

[0142] The present invention uses multi-objective optimization based on deep learning to find a parameter combination that optimizes the output speed and travel time; it is beneficial to reduce the water hammer pressure peak during valve closing, maximize the response speed of the system, and improve the performance of the valve electric actuator.

[0143] It should be noted that the word "comprising" does not exclude the presence of components or steps not listed in the claims. The word "a" or "an" preceding a component does not exclude the presence of a plurality of such components. The invention can be implemented by means of hardware comprising several distinct components, and by means of a suitably programmed computer.

[0144] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0145] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for optimizing design parameters of valve electric actuator based on deep learning, characterized in that: The method includes: S1. Use finite element calculation method to simulate and analyze the valve electric actuator under different design parameters to generate sample data of output speed and travel time; S2. Building a deep learning model based on the sample data to predict the output speed and travel time of the valve electric actuator; S3. Construct a multi-objective optimization model and use the MOEA / D optimization algorithm to optimize the output speed and travel time of the valve electric actuator to obtain the Pareto optimal solution; and use the AHP-entropy weight method to determine the weight coefficients between each objective; finally, use the TOPSIS method to sort the Pareto optimal solution and obtain the optimal solution for the design parameters of the valve electric actuator.

2. A method for optimizing design parameters of a valve electric actuator based on deep learning according to claim 1, characterized in that: In step S1, the design parameters include: Material properties: elastic modulus, density, Poisson’s ratio; Geometric parameters: valve diameter, wall thickness, stroke length; Operating conditions: working pressure, working temperature; Electric actuator parameters: motor power, rated speed, torque.

3. The method for optimizing design parameters of a valve electric actuator based on deep learning according to claim 1, characterized in that: In the step S1, comsol multiphysics software is used to perform simulation under different design parameters.

4. The method for optimizing design parameters of a valve electric actuator based on deep learning according to claim 1, characterized in that: In step S2, the deep learning model constructed includes: Input layer: There are n input nodes in total; Hidden layer: A multi-layer fully connected neural network is used, each layer contains a preset number of neurons and uses the ReLU activation function; Output layer: predicts output speed and travel time, with 2 output nodes in total.

5. A method for optimizing design parameters of a valve electric actuator based on deep learning according to claim 4, characterized in that: The deep learning model is trained using stochastic gradient descent or Adam optimizer, and the loss function is mean square error.

6. The method for optimizing design parameters of a valve electric actuator based on deep learning according to claim 1, characterized in that: In step S3, the multi-objective optimization model constructed includes the objective function: Minimize output speed error: f1=|V target -V output | Minimize travel time error: f2 = |T target -T output | Where f1 represents the minimum output speed error, V target Indicates the target speed, V output represents the output speed; f2 represents the minimum travel time error, T target represents the target travel time, T output Indicates the output travel time.

7. The method for optimizing design parameters of a valve electric actuator based on deep learning according to claim 1, characterized in that: In step S3, the calculation formula of the MOEA / D optimization algorithm is: Where λ is the weight vector, i is the i-th weight vector, z i is the reference point, x is the decision vector, and f i (x) is the objective function value, m is the maximum number of operations, g te is the minimum value.

8. The method for optimizing design parameters of a valve electric actuator based on deep learning according to claim 7, characterized in that: In step S3, the weight coefficients between the objectives are determined using the AHP-entropy weight method, where: The AHP method is: Establish a hierarchical model, including the target layer, criterion layer and solution layer; construct a judgment matrix and calculate the subjective weight of each criterion; calculate the eigenvector of the judgment matrix and perform normalization; calculate the consistency ratio: in, λ max is the maximum eigenvalue of the judgment matrix, n is the order of the judgment matrix, and RI is the random consistency index; The entropy weight method is: Among them, H j is the information entropy, r ij is the value of the i-th solution under the j-th criterion, j is the number of solutions; Determine the objective weight of each criterion: Combining the subjective weight of the AHP method, the comprehensive weight is obtained: Among them, α is the weight coefficient.

9. A method for optimizing design parameters of a valve electric actuator based on deep learning according to claim 8, characterized in that: In step S3, the TOPSIS method is used to rank the Pareto optimal solution, and the specific process includes: Construct a decision matrix: Take each solution in the Pareto optimal solution set as a decision plan and establish a decision matrix; Standardized decision matrix: Use vector normalization method to standardize the decision matrix: Among them, x ij is the value of the i-th solution under the j-th criterion; Calculate the weighted standardized decision matrix: Combine the weights obtained by the AHP-entropy weight method to calculate the weighted standardized decision matrix: Among them, v ij is the weighted normalized decision matrix; Determine the positive and negative ideal solutions: Positive ideal solution: the maximum value of each criterion: Negative ideal solution: Minimum of each criterion: In the formula, is the nth positive ideal solution, is the nth negative ideal solution, J is the set of all solutions; Compute the Euclidean distance of each solution to the positive and negative ideal solutions: In the formula, is the positive ideal solution of the jth solution, is the negative ideal solution of the jth solution; Calculate closeness: in, represents the Euclidean distance of the positive ideal solution, represents the Euclidean distance of the negative ideal solution; Sorting: Sort the solutions according to the closeness Ci and select the best solution.

Images