A dynamic simulation method for fuel system

Abstract

The present invention belongs to the field of fuel simulation / emulation technology for flight simulators, and specifically relates to a method for dynamic simulation of a fuel system, aiming to solve the problems of insufficient simulation accuracy and dynamic response distortion of the fuel system of existing aircraft simulators. The present invention obtains flight attitude, acceleration, and fuel tank parameters, constructs an equivalent liquid level prediction model, and determines the fuel pump supply status. When the fuel supply is normal, a state space model containing cavitation effects and compressibility corrections is used to calculate the ejection flow rate; and a multi-task neural network is used to predict the dynamic compensation factor and the liquid level correction model. When the fuel supply is limited, the fuel pump output flow rate is obtained based on the flight attitude angle. Finally, the ejection flow rate, fuel pump output flow rate, and liquid level height are used as key parameters for dynamic simulation of the fuel system. The present invention improves the accuracy of the dynamic response modeling of the fuel pump, enhances its adaptability to changes in flight attitude, and improves the simulation authenticity and training effect.

Description

Technical Field

[0001] The invention belongs to the technical field of fuel simulation / emulation of flight simulators, and in particular relates to a dynamic simulation method for a fuel system. Background Art

[0002] In flight training simulators, the accuracy of fuel system simulation directly impacts the pilot's perception of aircraft status and operational judgment. Currently, existing fuel system modeling methods are mostly based on static parameter settings or empirical formulas, lacking in-depth characterization of the dynamic operating characteristics of the fuel pump, particularly in modeling the pressure-flow relationship.

[0003] The main limitations are as follows:

[0004] 1. Imperfect fuel pump simulation: Existing models typically simplify the fuel pump as a constant flow or constant pressure source, ignoring the nonlinear coupling relationship between pressure and flow under different operating conditions. This fails to truly reflect the dynamic response characteristics of the fuel pump as load changes during actual flight.

[0005] Second, the lack of a coordinated control mechanism for the ejector pump: As a critical flow-regulating device in the fuel system, the ejector pump generates a flow-inducing effect based on the main pump's output pressure. However, existing simulation systems fail to establish an effective dynamic flow model for the ejector pump, making it difficult to simulate its flow-inducing capacity under varying pressure differentials and its impact on the overall fuel cycle.

[0006] 3. Inadequate modeling of flight attitude effects: When an aircraft performs complex maneuvers, such as steep dives, rolls, or climbs, the distribution of fuel within the tank changes significantly, leading to fluctuations in the fuel level, changes in the fuel supply path, and even partial interruptions in fuel flow. Existing models fail to fully account for the impact of flight attitude changes on the fuel level, fuel pump inlet pressure, and fuel supply continuity, resulting in distorted fuel system simulations.

[0007] 4. Lack of dynamic compensation mechanism: Under non-steady-state conditions (such as unstable fuel pump supply and cavitation), traditional models lack effective dynamic correction methods and are unable to adapt to the uncertain changes of the fuel system in complex flight environments, limiting the robustness and adaptability of the simulation system.

[0008] Therefore, there is an urgent need for a high-precision fuel system simulation method that can comprehensively consider the dynamic characteristics of the fuel pump, the synergistic effect of the ejector pump, and the influence of flight attitude, so as to improve the authenticity and reliability of the flight training simulator. Summary of the Invention

[0009] To address the aforementioned problems in the prior art, namely, the lack of modeling of the dynamic pressure-flow relationship of the fuel pump, the coordinated control of the ejector pump, and the coupling effect of flight attitude in the existing aircraft simulator fuel system simulation, resulting in insufficient simulation accuracy and distorted dynamic response, the first aspect of the present invention provides a fuel system dynamic simulation method, comprising the following steps:

[0010] The aircraft's current flight attitude angle and three-axis acceleration data are obtained. Combined with the fuel tank's geometric parameters and fuel physical properties, an equivalent liquid level prediction model based on inertial field reconstruction is constructed. The equivalent liquid level is calculated based on the equivalent liquid level prediction model.

[0011] judging whether the fuel pump is in a fuel supply condition according to the equivalent liquid level;

[0012] When the fuel pump meets the fuel conditions, the dynamic injection flow rate is calculated based on the state space discretization model based on the electric pump working pressure and the injection pump inlet pressure;

[0013] Based on the flight attitude angle and the fuel tank geometric parameters, a multi-task multi-layer perceptron model is used to predict the dynamic compensation factor and the liquid level height to obtain a predicted value. The predicted value is used to correct the state space discretization model to calculate a corrected dynamic ejection flow rate;

[0014] When the fuel pump does not have fuel conditions, the ejector pump does not work when the fuel pump has no output flow; based on the flight attitude angle, the fuel pump output flow that meets the non-steady-state working condition is obtained;

[0015] The fuel system is dynamically simulated by taking the ejection flow rate, the fuel pump output flow rate, and the equivalent liquid level as key parameters.

[0016] In some preferred embodiments, the equivalent liquid level prediction model is:

[0017] ;

[0018] in, Indicates the A tank at a time The effective oil suction port liquid level height; is the nonlinear geometric mapping function of the tank structure, For the A tank at a time The volume of fuel in the tank, is the flight attitude angle, are the geometric parameters of the fuel tank; is the integral term of the fuel sloshing effect, is the acceleration, is the rate of change of acceleration, is the fuel temperature, For the Fuel volume per tank; is the fuel transfer coupling term among multiple fuel tanks; For the Fuel transfer flow per tank; is the auxiliary correction term of the neural network.

[0019] In some preferred embodiments, the integral term of the fuel slosh effect is:

[0020] ;

[0021] in, is the function of fuel viscosity changing with temperature; is the attenuation coefficient of fuel volume; is the second derivative of acceleration; is the time interval; is the differential of the integral variable;

[0022] The fuel transfer coupling term between multiple fuel tanks is:

[0023] ;

[0024] in, From the fuel tank Flow to the fuel tank Fuel flow rate; For fuel tank cross-sectional area; is the time step; To sum the index variables, from 1 to .

[0025] In some preferred embodiments, the method for determining whether the fuel pump is in a fuel supply condition based on the equivalent liquid level is as follows:

[0026] When the equivalent liquid level height is higher than the fuel pump suction port height, the fuel pump is determined to be in a fuel supply condition; otherwise, the fuel pump is considered to be in a fuel supply condition.

[0027] In some preferred embodiments, the state space discretization model is:

[0028] ;

[0029] in, For the ejector pump at time Output flow rate; For the The pressure of the electric pump, For the The inlet pressure of the ejector pump, For the The resistance term of the fuel fluid characteristics, is the additional resistance term due to fuel compressibility; To introduce the cavitation effect judgment factor; is the efficiency factor; is the operating status factor; is the environmental feedback factor; is the learning adjustment factor; is the flow influence coefficient, is the exponential parameter; is the time interval, For quantity.

[0030] In some preferred embodiments, the additional resistance term of the fuel compressibility is:

[0031] ;

[0032] in, Indicates fuel density; is the speed of sound; Indicates the rate of change of fuel volume with pressure;

[0033] The cavitation effect judgment factor is:

[0034] ;

[0035] in, is the cavitation probability function based on the Bernoulli equation and the Reynolds number, is the vapor pressure of the fuel, is the viscosity of the fuel;

[0036] The learning regulator for:

[0037] ;

[0038] in, is the learning rate; For the moment The measured flow rate of the ejector pump; For the moment Predicted flow rate of the ejector pump.

[0039] In some preferred embodiments, the multi-task multi-layer perceptron model comprises a shared feature extraction layer and multiple task branch output layers;

[0040] The shared feature extraction layer is at least two layers of fully connected neural networks, each layer using a ReLU activation function;

[0041] The task branch output layer is used to predict different target variables and uses a linear activation function to output at least two physical related variables, including a dynamic compensation factor. , Liquid level prediction value .

[0042] In some preferred embodiments, a training phase is further included, wherein the multi-task multi-layer perceptron model is jointly optimized using a composite loss function, which includes multiple task loss terms and a physical consistency term:

[0043] The composite loss function is:

[0044] ;

[0045] in, is the prediction error of the dynamic compensation factor; is the liquid level prediction error; They are 、 The weight coefficient of is the physical constraint weight; is the cross-sectional area of ​​the fuel tank; is the rate of change of the predicted value of the liquid level; For the moment Flow rate into the ejector pump; For the moment Flow rate out of the ejector pump.

[0046] In some preferred embodiments, upper and lower limit constraints are set for the dynamic compensation factor.

[0047] In some preferred embodiments, the predicted value modifies the state space discretization model as follows:

[0048] ;

[0049] in, is the predicted value of liquid level height.

[0050] Beneficial effects of the present invention:

[0051] By establishing a nonlinear relationship model between the electric pump operating pressure, the ejector pump inlet pressure, and the ejector flow rate, the dynamic response of the pump under different operating conditions is truly reflected, compensating for the errors caused by simplifying the fuel pump as a constant source in traditional simulations.

[0052] By introducing a state-space discretization model and combining the cavitation effect judgment factor with the fuel compressibility correction term, we can accurately simulate the ejector pump's drainage capacity under different pressure differential conditions and its impact on the system's circulation flow, thus improving the integrity of the fuel system simulation.

[0053] By integrating flight attitude angle and three-axis acceleration data to construct an equivalent liquid level prediction model, we can effectively capture the distribution changes of fuel in the fuel tank, thereby affecting the fuel pump supply status and system pressure fluctuations, and enhancing the simulation's adaptability to complex flight scenarios.

[0054] A multi-task neural network model is introduced for dynamic compensation prediction, and a composite loss function is used to ensure that the prediction results are consistent with physical laws, thereby improving the model's stability and generalization capabilities.

[0055] By dynamically simulating the flow rate and liquid level changes of the fuel system throughout the flight, pilots are provided with operational feedback that is closer to the real flight environment, which helps to improve the quality and efficiency of flight training. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Other features, objects and advantages of the present application will become more apparent upon reading the detailed description of non-limiting embodiments made with reference to the following drawings:

[0057] Figure 1 The present invention is a flowchart of the steps of a fuel system dynamic simulation method. DETAILED DESCRIPTION

[0058] The present application will be further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are intended only to illustrate the relevant invention and are not intended to limit the invention. It should also be noted that, for ease of description, only portions relevant to the relevant invention are shown in the accompanying drawings.

[0059] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in this application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0060] In order to more clearly illustrate a fuel system dynamic simulation method of the present invention, the following is combined with Figure 1 Each step in the embodiment of the present invention is described in detail.

[0061] The present invention proposes a fuel system dynamic simulation method, see Figure 1 , the method comprises the following steps:

[0062] Obtain the aircraft's current flight attitude angle (pitch angle, roll angle) and three-axis acceleration data, combine the fuel tank's geometric parameters (shape, volume, fuel intake port location) and fuel physical parameters (density, viscosity, temperature), build an equivalent liquid level prediction model based on inertial field reconstruction, and calculate the equivalent liquid level based on the equivalent liquid level prediction model;

[0063] In this embodiment, the equivalent liquid level prediction model is:

[0064] ;

[0065] in, Indicates the A tank at a time The effective oil suction port liquid level height; is the nonlinear geometric mapping function of the tank structure, For the A tank at a time The volume of fuel in the tank, is the flight attitude angle, are the geometric parameters of the fuel tank; is the integral term of the fuel sloshing effect, is the acceleration, is the rate of change of acceleration, is the fuel temperature, For the Fuel volume per tank; is the fuel transfer coupling term among multiple fuel tanks; For the Fuel transfer flow per tank; is the auxiliary correction term of the neural network;

[0066] The geometric nonlinear mapping function It includes a lookup table method or interpolation algorithm based on the fuel tank shape parameters, which is used to automatically select the corresponding liquid level calculation model according to different types of fuel tanks (such as rectangular, trapezoidal, conical, with baffles, etc.), and calibrate it through experimental data or finite element analysis. Specifically, it obtains the fuel tank number, current fuel volume , tilt angle , a set of oil tank structural parameters (such as length, width, height, oil suction port position, inclined wall);

[0067] Get the type of the fuel tank according to the obtained fuel tank number;

[0068] Based on the obtained tilt angle, a lookup table relationship or interpolation function is established between the fuel volume and the liquid level in the zero-tilt state to obtain the liquid level. For the tilted fuel tank, finite element analysis or analytical formulas are used to simulate the fuel distribution. For example, the liquid level of a rectangular fuel tank in a tilted state is calculated using the following formula:

[0069] ;

[0070] In order to reflect the sloshing of fuel in the fuel tank during the maneuvering of the flight simulator, an integral term of the fuel sloshing effect is set. The integral term of the fuel sloshing effect is:

[0071] ;

[0072] in, is the function of fuel viscosity changing with temperature; is the attenuation coefficient of fuel volume; is the second derivative of acceleration; is the time interval; is the differential of the integral variable;

[0073] The fuel transfer coupling term between multiple fuel tanks is:

[0074] ;

[0075] in, From the fuel tank Flow to the fuel tank Fuel flow rate; For fuel tank cross-sectional area; is the time step; To sum the index variables, from 1 to ;

[0076] The neural network auxiliary correction term Real-time corrections are performed using a lightweight MLP or multi-layer Transformer network. The network inputs include flight attitude angle, three-axis acceleration, fuel temperature, fuel volume, and current fuel level predictions. The output is an error compensation value to improve the accuracy and robustness of the model under complex operating conditions.

[0077] Based on the equivalent liquid level, the method for determining whether the fuel pump is in a fuel supply condition is as follows:

[0078] When the equivalent liquid level is higher than the fuel pump suction port height, the fuel pump is deemed to be able to supply fuel; otherwise, the fuel pump is deemed to be unable to supply fuel. Furthermore, it is necessary to further check whether there is a fuel system fault and whether the system power supply is normal. If either a fuel system fault or power supply abnormality is true, the fuel pump is deemed to be unable to supply fuel regardless of whether the equivalent liquid level meets the standard.

[0079] When the fuel pump meets the fuel conditions, the dynamic injection flow rate is calculated based on the state space discretization model based on the electric pump output pressure and the injection pump inlet pressure;

[0080] In this embodiment, the state space discretization model is:

[0081] ;

[0082] in, For the ejector pump at time Output flow rate; For the Working pressure of an electric pump, For the The inlet pressure of the ejector pump, that is, the pressure at the bottom of the tank, For the The resistance term of the fuel fluid characteristics, is the additional resistance term due to fuel compressibility; To introduce the cavitation effect judgment factor; is the efficiency factor; is the operating status factor; is the environmental feedback factor; is the learning adjustment factor; is the flow influence coefficient, is the exponential parameter; is the time interval, For quantity;

[0083] The additional resistance term for fuel compressibility is:

[0084] ;

[0085] in, Indicates fuel density; is the speed of sound; Indicates the rate of change of fuel volume with pressure, obtained through experimental fitting, and is used to correct the delayed response phenomenon during fuel transfer under high pressure differential conditions;

[0086] The cavitation effect judgment factor is:

[0087] ;

[0088] in, It is a cavitation probability function based on the Bernoulli equation and Reynolds number, which is used to limit the output flow of the ejector pump under extreme pressure difference to prevent performance degradation caused by cavitation. is the cavitation probability function based on the Bernoulli equation and the Reynolds number, is the vapor pressure of the fuel, is the viscosity of the fuel;

[0089] The cavitation probability function adopts the empirical formula:

[0090] ;

[0091] The learning regulator for:

[0092] ;

[0093] in, is the learning rate, which is used to fine-tune the model parameters according to the error between the actual measured flow and the model predicted flow, thereby improving the long-term operation accuracy; For the moment The measured flow rate of the ejector pump; For the moment Predicted flow rate of the ejector pump;

[0094] The method further includes: using a multi-task multi-layer perceptron model to predict a dynamic compensation factor and a liquid level height based on the fuel flow rate, the flight attitude angle, and the geometric parameters of the fuel tank to obtain a predicted value, and using the predicted value to correct the state space discretization model to obtain a corrected dynamic ejection flow rate, thereby providing accurate ejection flow rate prediction;

[0095] In this embodiment, the multi-task multi-layer perceptron model includes a shared feature extraction layer and multiple task branch output layers;

[0096] The shared feature extraction layer is at least two layers of fully connected neural networks, each layer using a ReLU activation function;

[0097] The task branch output layer is used to predict different target variables and uses a linear activation function to output at least two physical related variables, including a dynamic compensation factor. , Liquid level prediction value ;

[0098] The method further includes a training phase, wherein the multi-task multi-layer perceptron model is jointly optimized using a composite loss function, which includes multiple task loss terms and a physical consistency term:

[0099] The composite loss function is:

[0100] ;

[0101] in, is the prediction error of the dynamic compensation factor; is the liquid level prediction error; They are 、 The weight coefficient of is the physical constraint weight; is the cross-sectional area of ​​the fuel tank; is the rate of change of the predicted value of the liquid level; For the moment Flow rate into the ejector pump; For the moment Flow rate out of the ejector pump;

[0102] The physical consistency term calculates the time derivative of the liquid level height by automatic differentiation or numerical difference and compares it with the fuel system dynamics equation;

[0103] The differential form of the relationship between the rate of change of the liquid level and the inflow and outflow of fuel is:

[0104] ;

[0105] The rate of change of the liquid level output by the model is calculated by automatic differentiation and brought into the consistency term for physical embedding:

[0106] ;

[0107] It also includes setting upper and lower limit constraints on the dynamic compensation factor to prevent its value from exceeding the physical boundary of the fuel system;

[0108] The predicted value modifies the state space discretization model as follows:

[0109] ;

[0110] in, is the predicted value of liquid level height;

[0111] When the fuel pump is not fuel-ready, the ejector pump will not operate when the fuel pump has no output flow (specifically, in extreme flight attitudes, such as high-angle dives or sharp turns). Significant displacement of fuel will occur within the fuel tank, causing fuel to accumulate at the front or rear of the tank. This attitude coupling effect will affect the coverage of the fuel pump suction port, causing suction ports that could otherwise draw fuel to be unable to extract sufficient fuel, or suction ports that could not originally draw fuel to be able to draw fuel. Consequently, the output pressure of the fuel pump will be significantly affected, manifested as the fuel pump output pressure dynamically adjusting with decreasing or increasing flow. In this case, the output flow of the fuel pump may be significantly reduced or even completely stopped, causing the ejector pump to not operate when the fuel pump has no output flow, affecting the fuel pump's fuel supply efficiency). Based on the flight attitude angle, the fuel pump output flow rate that meets the non-steady-state operating conditions can be obtained; for example, this can be obtained through dynamic modeling and mapping of the attitude angle. This is a prior art and will not be described in detail here.

[0112] The fuel system is dynamically simulated using the ejection flow rate, the fuel pump output flow rate, and the equivalent liquid level as key parameters. This comprehensively considers the dynamic characteristics of the fuel pump, the synergistic effect of the ejection pump, and the impact of flight attitude, thereby improving the authenticity and reliability of the flight training simulator.

[0113] Although the various steps in the above embodiment are described in the above-mentioned order, those skilled in the art will understand that in order to achieve the effect of this embodiment, different steps do not have to be executed in such an order. They can be executed simultaneously (in parallel) or in a reverse order. These simple changes are within the scope of protection of the present invention.

[0114] Those skilled in the art should be able to appreciate that the modules and method steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two, and the programs corresponding to the software modules and method steps can be placed in random access memory (RAM), internal memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disks, removable disks, CD-ROMs, or any other form of storage medium known in the art. In order to clearly illustrate the interchangeability of electronic hardware and software, the above description has generally described the components and steps of each example according to their functions. Whether these functions are performed in electronic hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0115] The terms "first", "second", etc. are used to distinguish similar objects, rather than to describe or indicate a particular order or sequence.

[0116] The term "comprise" or any other similar term is intended to cover non-exclusive inclusion such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed or inherent to such process, method, article, or apparatus.

[0117] Thus far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art may make equivalent changes or substitutions to the relevant technical features, and the technical solutions after such changes or substitutions will fall within the scope of protection of the present invention.

Claims

1. A fuel system dynamic simulation method, characterized in that: The method comprises the following steps: Acquiring the aircraft's current flight attitude angle and triaxial acceleration data, and combining the fuel tank's geometric parameters and fuel physical properties to construct an equivalent liquid level prediction model based on inertial field reconstruction, and calculating the equivalent liquid level based on the equivalent liquid level prediction model; judging whether the fuel pump is in a fuel supply condition according to the equivalent liquid level; When the fuel pump meets the fuel conditions, the dynamic injection flow rate is calculated based on the state space discretization model based on the electric pump working pressure and the injection pump inlet pressure; Based on the flight attitude angle and the fuel tank geometric parameters, a multi-task multi-layer perceptron model is used to predict the dynamic compensation factor and the liquid level height to obtain a predicted value. The predicted value is used to correct the state space discretization model to calculate a corrected dynamic ejection flow rate; When the fuel pump does not have fuel conditions, the ejector pump does not work when the fuel pump has no output flow; based on the flight attitude angle, the fuel pump output flow that meets the non-steady-state working condition is obtained; Performing dynamic simulation on the fuel system using the ejection flow rate, the fuel pump output flow rate, and the equivalent liquid level as key parameters; The equivalent liquid level prediction model is: ； in, Indicates the A tank at a time The effective oil suction port liquid level height; is the nonlinear geometric mapping function of the tank structure, For the A tank at a time The volume of fuel in the tank, is the flight attitude angle, are the geometric parameters of the fuel tank; is the integral term of the fuel sloshing effect, is the acceleration, is the rate of change of acceleration, is the fuel temperature, For the Fuel volume per tank; is the fuel transfer coupling term among multiple fuel tanks; For the Fuel transfer flow per tank; is the auxiliary correction term of the neural network; The state space discretization model is: ； in, For the ejector pump at time Output flow rate; For the Working pressure of an electric pump, For the The inlet pressure of the ejector pump, that is, the pressure at the bottom of the tank, For the The resistance term of the fuel fluid characteristics, is the additional resistance term due to fuel compressibility; To introduce the cavitation effect judgment factor; is the efficiency factor; is the operating status factor; is the environmental feedback factor; is the learning adjustment factor; is the flow influence coefficient, is the exponential parameter; is the time interval, For quantity.

2. A fuel system dynamic simulation method according to claim 1, characterized in that: The integral term of the fuel slosh effect is: ； in, is the function of fuel viscosity changing with temperature; is the attenuation coefficient of fuel volume; is the second derivative of acceleration; is the time interval; is the differential of the integral variable; The fuel transfer coupling term between multiple fuel tanks is: ； in, From the fuel tank Flow to the fuel tank Fuel flow rate; For fuel tank cross-sectional area; is the time step; To sum the index variables, from 1 to .

3. A fuel system dynamic simulation method according to claim 1, characterized in that: The method for judging whether the fuel pump has the fuel supply condition according to the equivalent liquid level is as follows: When the equivalent liquid level height is higher than the fuel pump suction port height, the fuel pump is determined to be in a fuel supply condition; otherwise, the fuel pump is considered to be in a fuel supply condition.

4. A fuel system dynamic simulation method according to claim 1, characterized in that: The additional resistance term for fuel compressibility is: ； in, Indicates fuel density; is the speed of sound; Indicates the rate of change of fuel volume with pressure; The cavitation effect judgment factor is: ； in, is the cavitation probability function based on the Bernoulli equation and the Reynolds number, is the vapor pressure of the fuel, is the viscosity of the fuel; The learning regulator for: ； in, is the learning rate; For the moment The measured flow rate of the ejector pump; For the moment Predicted flow rate of the ejector pump.

5. A fuel system dynamic simulation method according to claim 1, characterized in that: The multi-task multi-layer perceptron model includes a shared feature extraction layer and multiple task branch output layers; The shared feature extraction layer is at least two layers of fully connected neural networks, each layer using a ReLU activation function; The task branch output layer is used to predict different target variables and uses a linear activation function to output at least two physical related variables, including a dynamic compensation factor. , Liquid level prediction value .

6. A fuel system dynamic simulation method according to claim 5, characterized in that: The method further includes a training phase, wherein the multi-task multi-layer perceptron model is jointly optimized using a composite loss function, which includes multiple task loss terms and a physical consistency term: The composite loss function is: ； in, is the prediction error of the dynamic compensation factor; is the liquid level prediction error; They are 、 The weight coefficient of is the physical constraint weight; is the cross-sectional area of ​​the fuel tank; is the rate of change of the predicted value of the liquid level; For the moment Flow rate into the ejector pump; For the moment Flow rate out of the ejector pump.

7. A fuel system dynamic simulation method according to claim 6, characterized in that: It also includes setting upper and lower limit constraints on the dynamic compensation factor.

8. A fuel system dynamic simulation method according to claim 1, characterized in that: The predicted value modifies the state space discretization model as follows: ； in, is the predicted value of liquid level height, is the dynamic compensation factor, Indicates the fuel density, is the acceleration due to gravity.

Images