Fluid system dynamic behavior prediction method based on multi-parameter influence and neural network model

Abstract

The invention relates to a fluid system dynamic behavior prediction method based on multi-parameter influence and a neural network model, belongs to the technical field of fluid system prediction, and solves the problem of low prediction precision caused by the fact that the influence of multiple factors is not comprehensively considered in fluid behavior prediction in the prior art. The method comprises the steps that dynamic behaviors of a fluid system are analyzed, multi-parameter influences are comprehensively considered, and a phase function model and a displacement function model of the fluid system are constructed; carrying out numerical solution on the dynamic response of the fluid system under complex excitation based on the phase function model and the displacement function model, and constructing a first training sample set and a second training sample set; training a first neural network model for solving a phase prediction value of the fluid system displacement in real time by using the first training sample set, and training a second neural network model for solving a prediction value of the fluid system displacement in real time by using the second training sample set; and utilizing the first neural network model and the second neural network model to predict the dynamic behavior of the fluid system.

Description

Technical Field

[0001] The present invention relates to the technical field of fluid system prediction, and particularly to a method for predicting the dynamic behavior of a fluid system based on multi-parameter influence and a neural network model. Background Art

[0002] In the field of fluid dynamics, the technology for predicting and controlling the dynamic behavior of objects in complex fluid environments has gone through multiple development stages. Early research mainly relied on theoretical analysis and experimental verification, and described and predicted the dynamic behavior of objects by constructing simplified physical models. However, these models have great limitations in dealing with complex fluid environments and multi-parameter influences, resulting in insufficient prediction accuracy.

[0003] With the rapid development of computer technology, numerical simulation methods have gradually become an important means for studying fluid dynamics. Numerical methods such as the finite difference method and the finite element method are widely used to solve fluid dynamic equations. These methods have improved the prediction accuracy to a certain extent, but still face challenges in terms of computational efficiency and accuracy when dealing with problems of high nonlinearity and complex boundary conditions.

[0004] In recent years, the rise of artificial intelligence technology has provided a new opportunity for the research of fluid dynamics. Machine learning, especially deep learning technology, has demonstrated powerful capabilities in pattern recognition, feature extraction, and prediction. Researchers have begun to explore the combination of artificial intelligence algorithms and physical models to overcome the limitations of traditional methods. By combining the accuracy of physical models with the generalization ability of artificial intelligence algorithms, not only can the prediction accuracy of fluid dynamic behavior be improved, but also the computational efficiency can be significantly enhanced, making it possible to predict and control dynamic behavior in complex fluid environments.

[0005] Currently, research hotspots such as the dynamic behavior modeling of fluid systems based on multi-parameter fusion, a hybrid computing framework combining numerical solution and artificial intelligence algorithms, and the solution of dynamic deformation coefficients and phase functions under complex excitations are leading the development direction of fluid dynamics prediction and control technology. These methods not only enrich the theory of fluid dynamics but also provide strong support for engineering technology applications.

[0006] However, in the field of fluid dynamics, the prediction and control of the dynamic behavior of objects in complex fluid environments remain a technical challenge. Existing technologies usually have difficulty comprehensively and accurately considering the influence of multiple excitation units on the dynamic behavior of objects, especially in complex scenarios involving fluid-solid interaction, vibration analysis, and dynamic system response evaluation. Traditional models often ignore the interaction between multiple parameters, resulting in limited prediction accuracy. At the same time, as the complexity of fluid systems increases, pure numerical methods also face huge challenges in terms of computational efficiency. In addition, how to organically combine artificial intelligence algorithms with complex model solutions is also an urgent problem at present.

[0007] In summary, how to develop a dynamic behavior prediction method that can comprehensively consider the influence of multiple factors and improve the prediction accuracy and calculation efficiency has become an urgent problem to be solved in the field of fluid dynamics. Summary of the Invention

[0008] In view of the above analysis, the embodiments of the present invention aim to provide a fluid system dynamic behavior prediction method based on multi-parameter influence and neural network model to solve the problem of low prediction accuracy caused by the failure to comprehensively consider the influence of multiple factors in the existing fluid behavior prediction.

[0009] The present invention discloses a fluid system dynamic behavior prediction method based on multi-parameter influence and neural network model, and the method includes:

[0010] Analyze the dynamic behavior of the fluid system, comprehensively consider the influence of multiple parameters, and construct a phase function model and a displacement function model of the fluid system;

[0011] Based on the phase function model and the displacement function model, numerically solve the dynamic response of the fluid system under complex excitation, construct a first training sample set composed of the corresponding relationships between each experimental time, the amplitude of the fluid system displacement, and the solved value of the phase of the fluid system displacement, and construct a second training sample set composed of the corresponding relationships between each experimental time, the amplitude of the fluid system displacement, the solved value of the phase of the fluid system displacement, and the solved value of the fluid system displacement;

[0012] Use the first training sample set to train a first neural network model for real-time solving the predicted value of the phase of the fluid system displacement, and use the second training sample set to train a second neural network model for real-time solving the predicted value of the fluid system displacement;

[0013] Use the passed first neural network model and second neural network model to predict the dynamic behavior of the fluid system.

[0014] On the basis of the above solution, the present invention has also made the following improvements:

[0015] Further, for the prediction of the dynamic behavior of the fluid system using the passed first neural network model and second neural network model, execute:

[0016] During the actual operation process, input the prediction time and the amplitude of the fluid system displacement at the corresponding prediction time into the passed first neural network model, and the first neural network model predicts and outputs the predicted value of the phase of the fluid system displacement at the corresponding prediction time;

[0017] Input the prediction time, the amplitude of the fluid system displacement corresponding to the prediction time, and the predicted value of the phase of the fluid system displacement into the second neural network model that has passed the training. The second neural network model predicts and outputs the predicted value of the fluid system displacement corresponding to the corresponding prediction time.

[0018] Furthermore, the phase function model of the fluid system is expressed as:

[0019]

[0021] where s(t) and θ(t) respectively represent the amplitude function and the phase function of the displacement function x(t) of the fluid system, t represents time; δ represents the object perturbation parameter, b is the offset constant, and k represents the vibration coefficient of the object; m n , γ n respectively represent the deformation coefficient m n of the nth excitation unit, the elastic recovery time γ n , n = 1 to N, N represents the total number of excitation units; ω represents the system frequency; x represents the displacement of the object in the fluid system.

[0022] Furthermore, the displacement function model of the fluid system is expressed as:

[0023] x(t) = s(t)·cos(θ(t)) + b (2).

[0024] Furthermore, construct the first training sample set in the following manner:

[0025] Define the range of the experimental time, construct a time vector; according to the amplitude of the fluid system displacement corresponding to each experimental time in the time vector, construct an amplitude function vector; determine the system parameters that affect the dynamic behavior of the fluid system; and the training control parameters of the neural network model;

[0026] Initialize the phase function vector theta as a zero vector with the same length as the time vector;

[0027] For the experimental time t in the time vector, according to the amplitude of the fluid system displacement and the phase function model of the fluid system at the experimental time t, iteratively solve the phase solution value theta_t of the fluid system displacement and store it in the corresponding position in the phase function vector theta;

[0028] Construct the first training sample set composed of the corresponding relationships between each experimental time, the amplitude of the fluid system displacement, and the phase solution value of the fluid system displacement.

[0029] Furthermore, construct the second training sample set in the following manner:

[0030] Initialize the displacement function vector x as a zero vector with the same length as the time vector;

[0031] For the experimental time t in the time vector, substitute the current experimental time t, the amplitude s_t of the fluid system displacement, and the solved value theta_t of the phase of the fluid system displacement into the displacement function model of the fluid system to obtain the solved value x_t of the fluid system displacement at the experimental time t, and store it at the corresponding position in the displacement function vector x;

[0032] Construct a second training sample set composed of the corresponding relationships between each experimental time, the amplitude of the fluid system displacement, the solved value of the phase of the fluid system displacement, and the solved value of the fluid system displacement.

[0033] Furthermore, for the iterative solution of the solved value theta_t of the phase of the fluid system displacement, execute:

[0034] Extract the amplitude s_t of the fluid system displacement at the experimental time t from the amplitude function vector;

[0035] Set the initial phase guess theta_guess and the iterative vector theta_prev of all zeros; where the iterative vector theta_prev is used to sequentially store the theta_guess obtained in each iteration;

[0036] In each iteration process, substitute the theta_guess obtained in the previous iteration into the phase function model of the fluid system, and use the numerical integration method to calculate the first integral term integral1 and the second integral term integral2 in the phase function model of the fluid system respectively, which are the integral values corresponding to different integral intervals;

[0037] Substitute the first integral term, the second integral term, and the current phase guess theta_guess into the phase function model of the fluid system, update theta_guess, and store the updated theta_guess in the iterative vector theta_prev;

[0038] Based on the iterative tolerance tol, check whether the latest stored theta_guess in the iterative vector theta_prev converges. If it converges before reaching the maximum number of iterations max_iter, store the last theta_guess when theta_prev converges as the solved value theta_t of the phase of the fluid system displacement at the experimental time t.

[0039] Furthermore, for the iterative solution of the solved value theta_t of the phase of the fluid system displacement, it also executes:

[0040] If it does not converge and the maximum number of iterations max_iter is not reached, jump to the next iteration process.

[0041] Furthermore, train a first neural network model for real-time solving the phase prediction value of the fluid system displacement, and execute:

[0042] Use each experimental time and the amplitude of the fluid system displacement in the first training sample set as inputs, and use the corresponding phase solution value of the fluid system displacement as the output to train the first neural network model for real-time solving the phase prediction value of the fluid system displacement, and obtain a first neural network model that passes the training.

[0043] Furthermore, train a second neural network model for real-time solving the prediction value of the fluid system displacement, and execute:

[0044] Use each experimental time, the amplitude of the fluid system displacement, and the phase solution value of the fluid system displacement in the second training sample set as inputs, and use the corresponding solution value of the fluid system displacement as the output to train the second neural network model for real-time solving the prediction value of the fluid system displacement, and obtain a second neural network model that passes the training.

[0045] Compared with the prior art, the present invention can achieve at least one of the following beneficial effects:

[0046] The fluid system dynamic behavior prediction method based on multi-parameter influence and neural network model provided by the present invention has the following advantages:

[0047] (1) The present invention provides a fluid system dynamic behavior modeling method based on multi-parameter influence. By comprehensively considering parameters such as the solid state characteristics of an object, vibration coefficient, deformation coefficient of the excitation unit, and elastic recovery time, and combining the phase and amplitude functions that change with time, an accurate phase function model and displacement model suitable for the fluid system are constructed. This method significantly improves the modeling accuracy of the model for the dynamic behavior of objects in complex fluid environments.

[0048] (2) Solving the displacement function and phase function affected by multiple excitation units: For the dynamic deformation problem of an object in a fluid system under complex excitations, the present invention proposes a method for solving the phase function and displacement function of the dynamic deformation coefficient of an object affected by multiple excitation units. Through the iterative solution algorithm, an accurate characterization of the dynamic deformation characteristics of the object under complex excitations is achieved, providing a powerful tool for fluid dynamics research.

[0049] (3) Hybrid Framework of Artificial Intelligence and Numerical Solution: The present invention innovatively combines artificial intelligence methods with numerical solution algorithms to construct a method for solving complex nonlinear equations in the fluid field based on a hybrid computing framework of numerical solution and neural network. Through the learning and prediction of the neural network on the numerical solution results, the efficient calculation and accurate prediction of the fluid dynamic behavior are realized, significantly improving the calculation efficiency and prediction ability.

[0050] In the present invention, the above technical solutions can also be combined with each other to achieve more preferred combination schemes. Other features and advantages of the present invention will be described in the subsequent description, and some advantages can be made obvious from the description, or understood by implementing the present invention. The objectives and other advantages of the present invention can be achieved and obtained through the content specifically pointed out in the description and the drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] The drawings are only for the purpose of showing specific embodiments and are not considered as a limitation to the present invention. Throughout the drawings, the same reference signs represent the same components;

[0052] Figure 1 FIG. is a flowchart of a method for predicting the dynamic behavior of a fluid system based on multi-parameter influence and neural network model provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0053] The preferred embodiments of the present invention will be specifically described below with reference to the drawings. The drawings form a part of this application and are used together with the embodiments of the present invention to explain the principle of the present invention, but are not used to limit the scope of the present invention.

[0054] A specific embodiment of the present invention discloses a method for predicting the dynamic behavior of a fluid system based on multi-parameter influence and neural network model, and the flowchart is as Figure 1 shown.

[0055] Step S1: Analyze the dynamic behavior of the fluid system, comprehensively consider the influence of multiple parameters, and construct a phase function model and a displacement function model of the fluid system.

[0056] A fluid system refers to a system composed of a fluid (liquid or gas) and a container or pipeline containing the fluid, and is used for transporting, controlling, and distributing the fluid. By analyzing the dynamic behavior of the fluid system, the parameters involved in the multi-physical effects of the fluid system can be determined, which are specifically described as follows.

[0057] Object perturbation parameter δ: δ>0, which is used to describe the degree of interference of the outside world on the object in the fluid system.

[0058] Deformation coefficient and vibration coefficient k of the object: They can reflect the deformation and vibration characteristics of the object when it is stressed.

[0059] The deformation coefficient m of the excitation unit (n = 1 to N) n and the elastic recovery time γ n , to describe its mechanical response characteristics, where N represents the total number of excitation units. In a fluid system, an excitation unit refers to a component or device that can generate an excitation, provide energy or a signal to the fluid system, and thus cause a system response. Exemplarily, there can be various excitation units in a fluid system, and the following are some common types: mechanical excitation units (such as pumps, fans, and compressors), electromagnetic excitation units (such as solenoid valves, fluid power pumps), heating excitation units (such as heaters, coolers), fluid power excitation units (such as nozzles, turbines), external force excitation units (such as shakers, gravity).

[0060] The system frequency ω, which is used to describe the vibration or flow frequency of the fluid system.

[0061] As a common fluid system, a hydraulic control system is a device that uses the pressure and flow rate of a fluid (usually hydraulic oil) to control the movement of mechanical components. A hydraulic control system typically includes components such as a hydraulic pump, hydraulic valves, actuators (such as hydraulic cylinders or hydraulic motors), and a fuel tank. By adjusting the opening of the hydraulic valves and the displacement of the pump, the speed and force of the actuator can be precisely controlled. In a hydraulic control system, the solid-state characteristics of an object are reflected in that, in a hydraulic control system, the deformation coefficient and vibration coefficient of the actuator (such as the piston and cylinder block of a hydraulic cylinder) have an important impact on the accuracy and stability of the system. In a hydraulic control system, the excitation units can be components such as hydraulic pumps and hydraulic valves, and their mechanical characteristics (such as deformation coefficient and elastic recovery time) need to be considered in the transient response and stability analysis of the system. In particular, the fast response and precise control of hydraulic valves are crucial for the performance of the system.

[0062] The torque converter in a hydrodynamic transmission system, as a common fluid system, is a device that uses the kinetic energy of a fluid (usually oil) to transmit torque and adjust the rotational speed. The torque converter mainly consists of an impeller, a turbine, and a stator. These components usually have complex blade shapes to optimize hydrodynamic performance. In the torque converter, the fluid is accelerated by the impeller and then transmitted to the turbine, driving the turbine to rotate; the stator is used to adjust the fluid flow field to improve efficiency. In the torque converter, the solid-state characteristics of an object are reflected in: 1) Deformation coefficient: The impeller, turbine, and stator in the torque converter may undergo minor deformations due to fluid pressure and centrifugal force during high-speed rotation. This deformation can be minimized through advanced materials science and manufacturing processes, but it is still a factor that needs to be considered in system design. 2) Vibration coefficient: Due to the instability of fluid dynamics and the interaction of mechanical components, the torque converter may generate vibrations. These vibrations can be reduced through optimized design and the use of damping materials. In the torque converter, the mechanical characteristics of the excitation unit are reflected in: 1) Deformation coefficient: In the torque converter, the torsional stiffness (i.e., the deformation coefficient) of the input shaft and output shaft has an important impact on the performance of the system. They need to be able to withstand the transmitted torque and maintain a certain rigidity. 2) Elastic recovery time: Although the torque converter itself is not an elastic element, the fluid and some elastic seals in the system may have elastic recovery characteristics. These characteristics need to be considered in transient response and stability analysis.

[0063] In a fluid system, the displacement of an object is often affected by multiple factors jointly, including the flow characteristics of the fluid, the physical properties of the object, and external excitation, etc. Traditional displacement modeling methods often ignore the interaction between these complex factors, resulting in limited model prediction accuracy. To solve this problem, this embodiment provides a displacement modeling method that can comprehensively consider the influence of multiple factors. This method constructs an accurate displacement model applicable to fluid systems by comprehensively considering the solid-state characteristics of the object (including the deformation coefficient of the object's solid-state characteristics and the vibration coefficient related to solid state), the vibration coefficient, the deformation coefficient of the excitation unit, and the elastic recovery time, etc., in combination with the phase and amplitude functions that change with time. This displacement model can be applied to displacement prediction and control in fluid dynamics, especially suitable for occasions involving fluid-solid interaction, vibration analysis, and dynamic system response evaluation, and can effectively improve the accuracy and reliability of displacement prediction. Next, the displacement modeling method provided in this embodiment will be described as follows.

[0064] (1) Parameter definition and initialization

[0065] Define the displacement x(t) of the object in the fluid system, where t represents time, the object velocity and the object perturbation parameter δ.

[0066] Introduce the vibration coefficient k of the object (the vibration coefficient related to the solid state of the object).

[0067] For the nth excitation unit, define its deformation coefficient m n and the elastic recovery time γ n , where n = 1 to N.

[0068] Set the system frequency ω.

[0069] (2) Phase function construction

[0070] Model the displacement function x(t) of the fluid system as a cosine function, whose amplitude varies with time and is described by the amplitude function s(t); its phase varies with time, that is, the phase function θ(t) of the fluid system.

[0071] Construct a phase function model that is related to the vibration coefficient k and takes into account the influence of the excitation unit. Specifically, the phase function model of the fluid system is expressed as:

[0072]

[0073] where b is the offset constant, x represents the displacement of the object in the fluid system, and Ω is the equivalent frequency term.

[0074]

[0075] Using the obtained θ(t), construct the displacement function model of the fluid system as:

[0076] x(t) = s(t)·cos(θ(t)) + b = s(t)·cos(Ω·t) + b (3)

[0077] This displacement model comprehensively considers parameters such as the solid state characteristics of the object, the vibration coefficient, the deformation coefficient of the excitation unit, and the elastic recovery time, and can accurately describe the change of the displacement of the object in the fluid system with time.

[0078] Step S2: Based on the phase function model and the displacement function model, numerically solve the dynamic response of the fluid system under complex excitation, construct a first training sample set composed of the correspondence between each experimental time, the amplitude of the fluid system displacement and the solved value of the phase of the fluid system displacement, and construct a second training sample set composed of the correspondence between each experimental time, the amplitude of the fluid system displacement, the solved value of the phase of the fluid system displacement and the solved value of the fluid system displacement.

[0079] In this embodiment, complex excitation is used to simulate the combined action of multiple excitation units on a fluid system. Preferably, the complex excitation may come from multiple excitation sources (excitation units), each having different mechanical properties and time dependencies. The excitation sources of the fluid system can be divided into internal excitation sources and external excitation sources. Common internal excitation sources include turbulent excitation, unsteady flow excitation, vortex shedding excitation, etc. These excitation sources are usually closely related to the flow state of the fluid (such as velocity, pressure, density, etc.), and have obvious mechanical properties and time dependencies. Common external excitation sources include mechanical vibration, electromagnetic field variation, temperature variation, etc. The mechanical properties of each excitation source are different, mainly reflected in frequency characteristics, amplitude characteristics, phase characteristics, etc. The time dependencies of each excitation source are different, which reflect the variation law of the excitation signal with time, such as periodic excitation, random excitation, transient excitation, etc. Traditional analytical methods often have difficulty dealing with such problems, especially when facing highly nonlinear and implicit dependency relationships, they become even more powerless. To solve this problem, this embodiment provides a method that can flexibly and accurately solve the dynamic response of a fluid system under complex excitation. Specifically, this embodiment provides an iterative numerical solution algorithm for calculating the variation of the phase with time and amplitude, and the variation of the displacement with time, amplitude, and phase of an object in a fluid system under the action of multiple complex excitations. This algorithm is implemented through the following steps.

[0080] Step S21: Define the range of the experimental time, construct a time vector; construct an amplitude function vector according to the amplitude of the displacement of the fluid system corresponding to each experimental time in the time vector; determine the system parameters that affect the dynamic behavior of the fluid system; and the training control parameters of the neural network model. The specific description is as follows.

[0081] Time vector T: Represents the time range to be solved.

[0082] Amplitude function vector s: Can be a constant or a function of time, representing the amplitude change of the object displacement. Specifically, each experimental time in the time vector corresponds to a value of the amplitude function, which is used as the amplitude of the displacement of the fluid system at the corresponding experimental time, and the constructed amplitude function vector has the same length as the time vector. In the specific implementation process, the displacement data of the object in the fluid system at different time points can be measured and modeled as a cosine function to obtain the amplitude that varies with time.

[0083] System parameters delta, k, m_n, gamma_n, omega, b that affect the dynamic behavior of the fluid system: Respectively represent the object perturbation parameter, vibration coefficient, deformation coefficient of the excitation unit, elastic recovery time of the excitation unit, system frequency, and offset constant.

[0084] Number of excitation units N: Represents the number of excitation sources acting on the object.

[0085] Training control parameters of the neural network model: iteration tolerance tol and maximum number of iterations max_iter, which are used to control the accuracy and efficiency of iterative solution.

[0086] Step S22: Initialize the phase function vector theta and the displacement function vector x as zero vectors with the same length as the time vector respectively.

[0087] Step S23: For the experimental time t in the time vector, according to the amplitude of the fluid system displacement and the phase function model of the fluid system at the experimental time t, iteratively solve the phase solution value theta_t of the fluid system displacement and store it in the corresponding position in the phase function vector theta. The specific iterative process is as follows:

[0088] Step S231: Extract the amplitude s_t of the fluid system displacement at the experimental time t from the amplitude function vector s.

[0089] Step S232: Set the initial phase guess theta_guess and the iterative vector theta_prev of all zeros; among them, the iterative vector theta_prev is used to sequentially store the theta_guess obtained in each iteration.

[0090] Step S233: In each iteration process, substitute the theta_guess obtained in the previous iteration into the phase function model of the fluid system (the initial phase guess theta_guess is used in the first iteration), and use the numerical integration method to calculate the first integral term integral1 and the second integral term integral2 in the phase function model of the fluid system respectively, which correspond to the integral values of different integral intervals.

[0091]

[0092] Step S234: Substitute the first integral term, the second integral term and the current phase guess theta_guess into the phase function model of the fluid system, update theta_guess, and store the updated theta_guess in the iterative vector theta_prev. That is, substitute the current phase guess theta_guess as θ(t) on the right side of the equation into formula (1), and take the θ(t) obtained on the left side of the formula as the updated theta_guess. In particular, when sin(theta_guess) is zero, the algorithm will report an error or take special handling measures to avoid the error of dividing by zero.

[0093] Step S235: Taking the iterative tolerance tol as the judgment basis, check whether the latest stored theta_guess in the iterative vector theta_prev converges (if the change amount between the latest stored theta_guess and the theta_guess obtained in the previous iteration is less than the iterative tolerance tol, it converges; otherwise, it does not converge). If it does not converge and the maximum number of iterations max_iter has not been reached, jump to step S233; if it converges before the maximum number of iterations max_iter is reached, store the last theta_guess when theta_prev converges as the solution value theta_t of the phase of the fluid system displacement at the experimental time t; if it still does not converge after reaching the maximum number of iterations max_iter, issue a warning.

[0094] Step S24: For the experimental time t in the time vector, substitute the current experimental time t, the amplitude s_t of the fluid system displacement, and the solution value theta_t of the phase of the fluid system displacement into the displacement function model of the fluid system to obtain the solution value x_t of the fluid system displacement at the experimental time t, and store it at the corresponding position in the displacement function vector x.

[0095] Repeat step S23 and step S24, and the solution value of the phase of the fluid system displacement and the solution value of the fluid system displacement at all experimental times can be obtained, so as to construct a first training sample set composed of the corresponding relationships between each experimental time, the amplitude of the fluid system displacement, and the solution value of the phase of the fluid system displacement, and construct a second training sample set composed of the corresponding relationships between each experimental time, the amplitude of the fluid system displacement, the solution value of the phase of the fluid system displacement, and the solution value of the fluid system displacement.

[0096] Step S3: Use the first training sample set to train the first neural network model for real-time solving the phase prediction value of the fluid system displacement, and use the second training sample set to train the second neural network model for real-time solving the prediction value of the fluid system displacement.

[0097] Preferably, in the specific implementation process, a suitable neural network structure can be designed according to the complexity of the specific problem, such as a multi-layer perceptron (MLP), a convolutional neural network (CNN), or a recurrent neural network (RNN), etc. Considering the characteristics of time series data, RNN or its variants (such as LSTM, GRU) are more suitable.

[0098] Step S31: Use each experimental time and the amplitude of the fluid system displacement in the first training sample set as inputs, and use the corresponding solution value of the phase of the fluid system displacement as the output to train the first neural network model for real-time solving the phase prediction value of the fluid system displacement, and obtain the first neural network model that has passed the training.

[0099] Step S32: Use the experimental time, the magnitude of the fluid system displacement, and the solved phase value of the fluid system displacement in the second training sample set as inputs, and use the solved value of the corresponding fluid system displacement as the output to train a second neural network model for real-time solving the predicted value of the fluid system displacement, obtaining a second neural network model that passes the training.

[0100] During the training process of the neural network model, continuously adjust the network parameters (such as weights and biases) to minimize the error between the predicted value and the solved value.

[0101] Step S4: Use the first neural network model and the second neural network model that pass the training to predict the dynamic behavior of the fluid system.

[0102] Preferably, during the actual operation, input the prediction time and the magnitude of the fluid system displacement at the corresponding prediction time into the first neural network model that passes the training, and the first neural network model predicts and outputs the predicted value of the phase of the fluid system displacement at the corresponding prediction time; input the prediction time, the magnitude of the fluid system displacement at the corresponding prediction time, and the predicted value of the phase of the fluid system displacement into the second neural network model that passes the training, and the second neural network model predicts and outputs the predicted value of the fluid system displacement at the corresponding prediction time.

[0103] Preferably, in this embodiment, the predicted value of the phase of the fluid system at the prediction time and the predicted value of the displacement together serve as the prediction result of the dynamic behavior of the fluid system at the corresponding prediction time.

[0104] The hybrid computing framework provided in this embodiment combines the accuracy of the numerical solution method and the efficiency of the neural network, and has the following significant advantages:

[0105] Improve computing efficiency: Through the rapid prediction of the numerical solution results by the neural network, the time and resource consumption of repeated calculations are reduced.

[0106] Enhance prediction ability: The neural network can capture complex patterns and trends in the data, thereby improving the accuracy and generalization ability of the prediction.

[0107] Expand the application scope: This method is not only applicable to problems in the fluid field, but can also be extended to other fields that require the solution of complex nonlinear equations.

[0108] The method provided in this embodiment can be widely applied to fields such as fluid dynamics research, fluid-structure interaction analysis, and engineering structure vibration control. By accurately solving the dynamic deformation coefficient and phase function of the object, it can provide engineers with more accurate and reliable prediction and control means, thereby optimizing the structural design and improving the system performance.

[0109] Those skilled in the art can understand that all or part of the processes for implementing the methods of the above embodiments can be completed by instructing relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium. Among them, the computer-readable storage medium is a disk, an optical disc, a read-only memory, a random access memory, etc.

[0110] As described above, only the preferred specific embodiments of the present invention are provided, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed by the present invention should be covered by the protection scope of the present invention.

Claims

1. A method for predicting the dynamic behavior of a fluid system based on multi-parameter influence and neural network model, characterized in that: The method comprises: Analyze the dynamic behavior of the fluid system, comprehensively consider the influence of multiple parameters, and construct the phase function model and displacement function model of the fluid system; Based on the phase function model and the displacement function model, the dynamic response of the fluid system under complex excitation is numerically solved, and a first training sample set consisting of the corresponding relationship between each experimental time, the amplitude of the fluid system displacement and the phase solution value of the fluid system displacement is constructed, and a second training sample set consisting of the corresponding relationship between each experimental time, the amplitude of the fluid system displacement, the phase solution value of the fluid system displacement and the solution value of the fluid system displacement is constructed; Using the first training sample set to train a first neural network model for solving the phase prediction value of the fluid system displacement in real time, and using the second training sample set to train a second neural network model for solving the prediction value of the fluid system displacement in real time; The dynamic behavior of the fluid system is predicted using the trained first neural network model and the second neural network model.

2. The method for predicting dynamic behavior of a fluid system based on multi-parameter influence and neural network model according to claim 1 is characterized in that: The dynamic behavior prediction of the fluid system is performed by using the trained first neural network model and the second neural network model, and executing: In the actual operation process, the predicted time and the amplitude of the fluid system displacement corresponding to the predicted time are input into the trained first neural network model, and the first neural network model predicts and outputs the phase prediction value of the fluid system displacement corresponding to the predicted time; The prediction time, the amplitude of the fluid system displacement corresponding to the prediction time, and the phase prediction value of the fluid system displacement are input into the trained second neural network model, and the second neural network model predicts and outputs the prediction value of the fluid system displacement corresponding to the prediction time.

3. The method for predicting dynamic behavior of a fluid system based on multi-parameter influence and neural network model according to claim 2 is characterized in that: The phase function model of the fluid system is expressed as: Where s(t) and θ(t) represent the amplitude function and phase function of the displacement function x(t) of the fluid system, respectively; t represents time; δ represents the object disturbance parameter, b is the offset constant, and k represents the vibration coefficient of the object; m n , γ n Respectively represent the deformation coefficient m of the nth excitation unit n , elastic recovery time γ n , n=1~N, N represents the total number of excitation units; ω represents the system frequency; x represents the displacement of the object in the fluid system.

4. The method for predicting dynamic behavior of a fluid system based on multi-parameter influence and neural network model according to claim 3 is characterized in that: The displacement function model of the fluid system is expressed as: x(t)=s(t)·cos(θ(t))+b (2).

5. The method for predicting dynamic behavior of a fluid system based on multi-parameter influence and neural network model according to claim 4, characterized in that: The first training sample set is constructed in the following way: Define the range of experimental time and construct a time vector; construct an amplitude function vector according to the amplitude of the fluid system displacement corresponding to each experimental time in the time vector; determine the system parameters that affect the dynamic behavior of the fluid system; and the training control parameters of the neural network model; Initialize the phase function vector theta to a zero vector with the same length as the time vector; For the experimental time t in the time vector, according to the amplitude of the fluid system displacement at the experimental time t and the phase function model of the fluid system, the phase solution value theta_t of the fluid system displacement is iteratively solved and stored in the corresponding position in the phase function vector theta; A first training sample set consisting of the corresponding relationship between each experimental time, the amplitude of the fluid system displacement and the phase solution value of the fluid system displacement is constructed.

6. The method for predicting dynamic behavior of a fluid system based on multi-parameter influence and neural network model according to claim 5, characterized in that: The second training sample set is constructed in the following way: Initialize the displacement function vector x to a zero vector with the same length as the time vector; For the experimental time t in the time vector, the current experimental time t, the amplitude s_t of the fluid system displacement and the phase solution value theta_t of the fluid system displacement are brought into the displacement function model of the fluid system to obtain the solution value x_t of the fluid system displacement at the experimental time t, which is stored in the corresponding position in the displacement function vector x; A second training sample set is constructed, which is composed of the corresponding relationship between each experimental time, the amplitude of the fluid system displacement, the phase solution value of the fluid system displacement and the solution value of the fluid system displacement.

7. The method for predicting dynamic behavior of a fluid system based on multi-parameter influence and neural network model according to claim 6, characterized in that: The iterative solution for the phase solution value theta_t of the fluid system displacement is performed: Extract the amplitude s_t of the fluid system displacement at the experimental time t from the amplitude function vector; Set the initial phase guess theta_guess and the all-zero iteration vector theta_prev; the iteration vector theta_prev is used to sequentially store theta_guess obtained in each iteration; In each iteration, the theta_guess obtained in the previous iteration is brought into the phase function model of the fluid system, and the first integral term integral1 and the second integral term integral2 in the phase function model of the fluid system are calculated using the numerical integration method, corresponding to the integral values ​​of different integral intervals. Bring the first integral term, the second integral term and the current phase guess theta_guess into the phase function model of the fluid system, update theta_guess, and store the updated theta_guess in the iteration vector theta_prev; Based on the iteration tolerance tol, check whether the latest theta_guess stored in the iteration vector theta_prev converges. If it does not converge before the maximum number of iterations max_iter, store the last theta_guess when theta_prev converges as the phase solution value theta_t of the fluid system displacement at the experimental time t.

8. The method for predicting dynamic behavior of a fluid system based on multi-parameter influence and neural network model according to claim 7, characterized in that: The iterative solution for the phase solution value theta_t of the fluid system displacement also performs: If it does not converge and the maximum number of iterations max_iter is not reached, it jumps to the next iteration process.

9. The method for predicting dynamic behavior of a fluid system based on multi-parameter influence and neural network model according to claim 8, characterized in that: Train the first neural network model to solve the phase prediction value of the fluid system displacement in real time, and execute: The experimental time and the amplitude of the fluid system displacement in the first training sample set are taken as input, and the corresponding phase solution value of the fluid system displacement is taken as output, and a first neural network model for solving the phase prediction value of the fluid system displacement in real time is trained to obtain a trained first neural network model.

10. The method for predicting dynamic behavior of a fluid system based on multi-parameter influence and neural network model according to claim 9, characterized in that: Train a second neural network model to solve the predicted value of the fluid system displacement in real time, and execute: The experimental time, amplitude of the fluid system displacement, and phase solution value of the fluid system displacement in the second training sample set are used as input, and the corresponding solution value of the fluid system displacement is used as output to train a second neural network model for real-time solution of the predicted value of the fluid system displacement, thereby obtaining a trained second neural network model.

Images