Design method for exhaust diffuser of high-altitude test bed

Abstract

The invention discloses a design method for an exhaust diffuser of a high-altitude test bed. The design method comprises the following steps that firstly, the inlet area and the outlet area of the exhaust diffuser are solved based on a one-dimensional thermodynamic equation; 2, based on an existing high-altitude test bed, establishing an equal-proportion two-dimensional simulation model, and verifying the reliability of the simulation model; 3, the reflux quantity at an inlet of the exhaust diffuser is obtained; and 4, solving the distance between the inlet of the exhaust diffuser and the outlet of the exhaust nozzle and the diffusion angle when the injection coefficient is maximum on the premise of meeting the constraint conditions through a multi-task deep learning model. The advantages of a one-dimensional thermodynamic equation and numerical simulation are combined, and the exhaust diffuser at the front end of the exhaust system is designed for a high-altitude test bed of an existing exhaust system.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of high-altitude test stand exhaust diffuser design, and particularly relates to a high-altitude test stand exhaust diffuser design method. BACKGROUND

[0002] The exhaust diffuser is one of the key components of the high-altitude test stand exhaust system, which converts part of the kinetic energy of the engine exhaust gas into pressure energy, and simulates the engine exhaust back pressure and environmental pressure conditions. It can reduce the speed, temperature and noise of the high-temperature and high-speed exhaust gas of the test engine, so that the exhaust gas can be smoothly discharged into the atmosphere. The design of the exhaust diffuser is crucial to the reasonable configuration of the test engine: it is directly related to the simulated altitude of the high-altitude test stand, the working range and energy saving. Therefore, the design of the exhaust diffuser must be considered comprehensively.

[0003] For the design of the exhaust diffuser, the current main target is to improve the entrainment coefficient of the diffuser, and there are mainly two design methods: 1) using one-dimensional gas ejector flow characteristic equation to design the structure of the diffuser; and 2) using three-dimensional numerical calculation method to design the structure. The diffuser design method based on aerodynamic thermodynamic calculation is simple and fast, but the precision is poor; the three-dimensional numerical calculation method is complex and time-consuming, but the effect is good. However, the two design methods are both aimed at improving the entrainment coefficient of the diffuser for structural design, and do not further consider whether the designed diffuser will cause damage to the test engine due to backflow of the exhaust gas. SUMMARY

[0004] In order to overcome the above-mentioned shortcomings of the prior art, in order to meet the needs of high-altitude test of different types of engines, the present application provides a high-altitude test stand exhaust diffuser design method, which combines the advantages of one-dimensional thermodynamic equation and numerical simulation simulation, and designs the exhaust diffuser at the front end of the exhaust system of the existing high-altitude test stand.

[0005] The present application provides a high-altitude test stand exhaust diffuser design method, which comprises the following steps: Step 1: solving the inlet area and outlet area of the exhaust diffuser based on one-dimensional thermodynamic equation; Step 2: based on the existing high-altitude test stand, an equal-proportion two-dimensional simulation model is established to verify the reliability of the simulation model; Step 3: selecting the diffuser angle and the distance between the exhaust diffuser inlet and the tail nozzle outlet; combining the exhaust diffuser inlet and outlet areas obtained in step 1, building a simulation model for calculation, and using UDF to write related functions to obtain the backflow amount at the inlet of the exhaust diffuser; Step four: based on the simulation values, the relationship between the distance between the exhaust diffuser inlet and the tail nozzle outlet, the expansion angle of the exhaust diffuser, the entrainment coefficient, and the backflow amount is found out respectively; and according to the constraint condition that the backflow amount is less than or equal to 5%, a multi-task deep learning model is established, and through the multi-task deep learning model, the distance between the exhaust diffuser inlet and the tail nozzle outlet and the expansion angle of the exhaust diffuser are obtained when the entrainment coefficient is maximum under the premise of meeting the constraint condition.

[0006] Further, the step one specifically comprises the following steps: Step 1-1, known parameters: test engine total exhaust temperature , total static temperature , specific heat ratio of ejector flow , universal gas constant ; Based on the following formula (1), formula (2), formula (3), formula (4), formula (5): (1) the gas flow parameters of the test engine tail nozzle outlet gas or ejector flow are obtained: outlet Mach number , velocity coefficient , aerodynamic function and , and the aerodynamic function of mass flow and gas property constant ; (2) the gas flow parameters in the high-altitude cabin are obtained: Mach number , velocity coefficient , aerodynamic function and , and gas property constant : (1); (2); (3); (4); (5); (6); Step 1-2, based on step 1-1, the target entrainment coefficient is given, the ejector area ratio is obtained based on the following formula, and the ejector inlet area is further determined: (7); (8); Step 1-3, the exhaust diffuser contraction / expansion ratio is obtained. (1) Calculate the specific heat ratio of the mixed gas flow , constant pressure specific heat , total temperature , velocity coefficient : (9); (10); (11); (12); (13); (14); (2) Based on the formula in step 1-1, calculate the aerodynamic function of the mixed gas and and the aerodynamic function of the mass flow and the gas characteristic constant ; (3) According to the isobaric condition of the isobaric ejector mixing chamber, calculate the total pressure of the mixed gas and the contraction / expansion ratio of the exhaust diffuser , and then calculate the outlet diameter of the exhaust diffuser: (15); (16); (17)。

[0007] Further, the two-dimensional simulation model mainly consists of a high-altitude test chamber, an exhaust diffuser and an engine.

[0008] Further, the boundary conditions of the two-dimensional simulation model are set as follows: set the inlet boundary condition: adopt the pressure inlet condition, set the total pressure and total temperature of the gas flow; set the outlet boundary condition: adopt the pressure outlet condition, set the total pressure and total temperature of the gas flow; set the axisymmetric condition: set the corresponding side as the symmetry axis.

[0009] Further, the calculation method of the two-dimensional simulation model is set as follows: adopt the steady, implicit solver based on density; select the standard k-epsilon turbulence model; the calculation of pressure and velocity adopts the Implicit algorithm.

[0010] Further, the reliability of the simulation model in step two is verified by comparing the high-altitude chamber pressure in the test data with the chamber pressure calculated by the simulation model.

[0011] Further, in the step three, the diffuser angle is selected with a gradient of a degrees, and the distance between the exhaust diffuser inlet and the nozzle outlet is selected with a gradient of b mm.

[0012] Further, the specific method for establishing the multi-task deep learning model comprises the following steps: S01: data preparation and preprocessing: Obtain a historical test data set, which contains key input parameters: distance to nozzle, ejector expansion angle; and key output performance parameters: backflow, forward flow, and calculated ejector coefficient; use the standardizer StandardScaler to standardize the input and output data, convert them to a data distribution with a mean of 0 and a variance of 1, and randomly divide the data set into a training set and a test set in proportion; S02: constructing and training a multi-task deep learning model: Construct a multi-task deep learning model, which adopts a shared bottom network combined with an independent top network architecture, including: Input layer: receives the standardized distance and expansion angle two-dimensional features and converts them into Pytorch tensors.

[0013] Shared hidden layer: composed of multiple fully connected layers and ReLU activation functions, used to extract shared high-level features from input parameters; Multi-task output layer, including: backflow prediction branch: starting from shared features, output the predicted value of backflow through an independent fully connected network; structure same as backflow branch, output the predicted value of forward flow; ejector coefficient prediction branch: structure same as backflow branch, output the predicted value of ejector coefficient; Backflow rate calculation: during the forward propagation of the model, the predicted values of backflow and forward flow are used to calculate the backflow rate in real time; Train the model using the training set data, and the loss function is the sum of the mean square errors between the predicted values and the true values of each task; use the Adam optimizer to iteratively optimize the model parameters until the loss function converges; S03: define the comprehensive objective function and constraint conditions: The user inputs the backflow weight coefficient and the ejector coefficient weight coefficient , wherein + = 1; the above weights represent the relative importance of the designer to reduce backflow and improve the ejector coefficient; Objective function A: defined as A = * backflow weight coefficient * The return flow, maximizing A means seeking the optimum of the comprehensive performance in the sense of weighting; Constraint: the return flow rate is less than or equal to 5%. This is a hard constraint, and any solution that does not meet this condition is an infeasible solution; S04: Global search and optimization solution: Generate high-resolution parameter grid points within the specified parameter range; use the multi-task model trained in step S02 to predict the return flow, forward flow, entrainment coefficient and return flow rate of each point in the entire parameter space; According to the constraint condition defined in S103, filter out all grid points with a return flow rate of less than or equal to 5% (feasible region); in the feasible region, calculate the value of the objective function A corresponding to each point, and find the point with the maximum A value, which corresponds to the optimal parameter combination of the distance and expansion angle; S05: Output optimal results: the system outputs the optimal solution, including: The optimal distance from the nozzle and the expansion angle of the ejector; the predicted return flow, forward flow, entrainment coefficient, return flow rate and objective function A value under this configuration.

[0014] The working principle of the present application is: a design method suitable for high-altitude test stand exhaust diffuser, which combines one-dimensional gas thermodynamics equation and numerical simulation, is a fast and accurate design method that balances the entrainment coefficient and the exhaust flow rate of the diffuser. In this design method, the design of the exhaust diffuser based on the existing exhaust system is considered to meet the needs of different engine models for high-altitude testing. For this, the outlet of the designed exhaust diffuser is a known parameter, and the one-dimensional gas thermodynamics equation is used to calculate the inlet area of the diffuser under a certain entrainment coefficient. The outlet and inlet area structure parameters are used as input parameters of the simulation model in the numerical simulation of the diffuser, and the tilt angle between the outlet and inlet of the diffuser is used as the variable, and the return flow and the entrainment coefficient are used as the two constraint targets to develop an exhaust diffuser optimization design method. This method avoids the damage to the test engine caused by the backflow of the exhaust diffuser due to the design deficiency of the exhaust diffuser.

[0015] For the structure design method of the exhaust diffuser for high-altitude test stand, the method learns from the design of one-dimensional thermodynamics equation and numerical simulation simulation, and uses one-dimensional thermodynamics equation design and numerical simulation simulation. Through the combination of the two, the rapidity and accuracy are met. The particularity of the design method of the present application lies in the combination of the two methods, which not only greatly reduces the calculation amount of the design, but also improves the accuracy of the design.

[0016] Compared with the prior art, the method has the beneficial effects that:

[0017] Compared with the prior art, the method has the beneficial effects that: 1. The method for designing the exhaust diffuser of the high-altitude test stand is provided, which combines the thermodynamic equation solving and numerical simulation, ensures the rapidity and accuracy of the exhaust diffuser design, and realizes the precise design of the exhaust diffuser matched with different engines.

[0018] 2. The method for designing the exhaust diffuser is provided, which simultaneously considers the entrainment coefficient and the backflow of the high-temperature fuel gas. BRIEF DESCRIPTION OF DRAWINGS

[0019] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiment description.

[0020] Figure 1 is a flow chart of the method for designing the exhaust diffuser of the high-altitude test stand.

[0021] Figure 2 is a schematic diagram of a two-dimensional simulation model in the method for designing the exhaust diffuser of the high-altitude test stand.

[0022] Figure 3 is a schematic diagram of a two-dimensional simulation model of the exhaust diffuser inlet in the method for designing the exhaust diffuser of the high-altitude test stand.

[0023] Figure 4 is a flow chart of the method for designing the exhaust diffuser of the high-altitude test stand.

[0024] Figure 5 is a visualization result diagram of the method for designing the exhaust diffuser of the high-altitude test stand, wherein (a) is a backflow prediction surface diagram, (b) is an entrainment coefficient prediction surface diagram, (c) is a backflow rate distribution and constraint boundary diagram, and (d) is a feasible region diagram of the target function A distribution. DETAILED DESCRIPTION

[0025] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0026] like Figure 1 As shown in the figure, an embodiment of the present invention proposes a design method for an exhaust diffuser on a high-altitude test stand, which includes the following steps: Step 1: Solve for the inlet and outlet areas of the exhaust diffuser based on the one-dimensional thermodynamic equation; Specifically, the steps include: To facilitate theoretical calculation and analysis, this embodiment also defines the symbols for the tailpipe airflow and high-altitude cabin airflow parameters, as shown in the table below: Table 1. Definition of symbols for tailpipe airflow and high-altitude cabin airflow parameters: Step 1.1, Known parameters: Total exhaust temperature of the test engine Total static temperature Specific heat ratio of ejector gas Universal gas constant ; Based on the following equations (1), (2), (3), (4), and (5): (1) Calculate the gas flow parameters of the exhaust gas (ejector flow) at the outlet of the test engine's tailpipe: outlet Mach number speed coefficient aerodynamic function and and the aerodynamic function of mass flow rate and gas characteristic constants ; (2) Calculate the airflow parameters inside the high-altitude cabin: Mach number speed coefficient aerodynamic function and and gas characteristic constants : Step 1.2, based on Step 1.1, given the target ejection coefficient, calculate the ejector area ratio using the following formula. Further determine the ejector inlet area : Step 1.3, Calculate the exhaust diffuser contraction / expansion ratio. .

[0027] (1) Obtain the specific heat ratio of the mixed gas flow , the constant pressure specific heat , the total temperature , the velocity coefficient : (2) Based on the formula in step 1.1, obtain the aerodynamic functions of the mixed gas and , and the aerodynamic functions of the mass flow and the gas property constant .

[0028] (3) According to the isobaric condition of the mixing chamber of the isobaric ejector, obtain the total pressure of the mixed gas and the contraction / expansion ratio of the exhaust diffuser , and then obtain the outlet diameter of the exhaust diffuser.

[0029] Step two: Based on the existing high-altitude test bed, establish a two-dimensional simulation model in proportion, and verify the reliability of the simulation model; The two-dimensional simulation model mainly consists of a high-altitude test chamber, an exhaust diffuser, and an engine. Since the cross-sectional shape and space size of the test chamber have little effect on the performance of the ejector, it is simplified as a cylinder with a length of 3m and a diameter of 2.66m. The engine nozzle has a diameter of 0.268m. The exhaust diffuser is divided into a diffuser section and a straight section. The diffuser section is a cone with an inlet diameter of 0.34m, an outlet diameter of 0.79m, and a length of 1.34m. The straight section is a cylinder with a length of 6.6m and a diameter of 0.766m. The model is a central symmetric body, and the gravity effect is not considered. The model is reduced from three dimensions to two-dimensional axisymmetric surface, as shown in Figure 2 .

[0030] The boundary conditions of the simulation model are set as follows: 1) Inlet boundary condition: pressure inlet condition is adopted, and the total pressure and total temperature of the gas flow are set; 2) Outlet boundary condition: pressure outlet condition is adopted, and the total pressure and total temperature of the gas flow are set; 3) Axisymmetric condition: as shown in Figure 2 , the corresponding edges are set as symmetric axes.

[0031] The calculation method of the simulation model is set as follows: (1) Adopt the steady, implicit solver based on density; (2) Select the standard k-epsilon turbulence model; (3) Use Implicit algorithm for pressure and velocity.

[0032] The simulation model is verified by comparing the cabin pressure in the high-altitude cabin in the test data and the cabin pressure calculated by the simulation model. The comparison of simulation data and test data is shown in Table 2. As can be seen from Table 2, the maximum relative error of the cabin pressure calculated by the simulation and the test value is 2.18%, which meets the requirement of 5% in engineering, verifying the accuracy of the simulation model, and the model can be used for further research.

[0033] Step three: select the diffuser angle [x1, x2] and the distance [y1, y2] between the exhaust diffuser inlet and the nozzle outlet, where the diffuser angle is selected in a gradient of a degrees, and the distance between the exhaust diffuser inlet and the nozzle outlet is selected in a gradient of b mm. Based on the exhaust diffuser outlet and inlet area obtained in step one, a simulation model is built for calculation, and UDF is used to write related functions to obtain the backflow at the exhaust diffuser inlet.

[0034] Step four: based on the simulation values, the relationship between the distance between the exhaust diffuser inlet and the nozzle outlet, the diffuser angle and the injection coefficient, and the backflow is found out, and according to the constraint condition of backflow ≤ 5%, the distance between the exhaust diffuser inlet and the nozzle outlet and the diffuser angle that make the injection coefficient maximum under the premise of meeting the constraint condition are obtained.

[0035] Example: Assumed known parameters: 1) Nozzle outlet gas parameters (injection gas parameters): engine exhaust flow rate 11 kg / s, exhaust total temperature 1800 K, total static temperature 1650 K, injection gas specific heat ratio 1.3, total pressure 150 kPa, nozzle radius 0.075 m, and universal gas constant 8.314 kJ / (mol·K).

[0036] 2) High-altitude cabin gas flow parameters (injected gas parameters): total temperature 300 K, total static temperature 280 K, injection gas specific heat ratio 1.4, and initial total pressure 12 kPa.

[0037] 3) Injection coefficient, which is taken as 0.15. When designing the ejector, different target injection coefficients can be selected.

[0038] According to the calculation formula in step one, the inlet diameter of the diffuser is 0.346 m, and the outlet diameter is 0.783 m, and the correctness of the simulation model is verified in step two. This paper mainly shows the optimization process of step three and step four.

[0039] The combination of the diffuser angle [0°, 30°] and the distance between the exhaust diffuser inlet and the tail nozzle outlet [0mm, 300mm] is selected, wherein the diffuser angle is selected with a gradient of 2 degrees, and the distance between the exhaust diffuser inlet and the tail nozzle outlet is selected with a gradient of 50 mm. The simulation model is built and calculated in combination with the exhaust diffuser outlet and inlet area obtained in step one, and the relevant functions are written using UDF to obtain the backflow at the exhaust diffuser inlet, as shown in Figure 3

[0040] Using the verified simulation model, the calculation results are shown in Table 3. Based on the existing small amount of simulation and experimental data, a kind of ejector parameter optimization method and system based on deep learning multi-objective constraint optimization is used, which can quickly and accurately find the optimal distance and expansion angle parameter combination that maximizes the comprehensive performance target function (A = * Ejector coefficient - * Backflow) under the given constraint condition (backflow rate ≤ 5%), and gives the predicted performance reference value, thereby significantly reducing the number of tests, reducing the development cost, and realizing the intelligent trade-off of performance.

[0041] Table 3 Simulation results table The prior art mainly relies on physical experiments and physical model simulation (CFD), and the present application first uses a data-driven deep learning model to construct the complex nonlinear mapping relationship between parameters and performance. This is a completely new digital and intelligent design paradigm. The present application establishes a multi-task deep learning model to accurately predict the backflow, positive flow and ejector coefficient, and uses the internal mathematical relationship (backflow rate) between them as a constraint condition to construct a comprehensive target function A by weighting, realizing the collaborative optimization of multiple objectives. The implementation process is shown in Figure 4 , which mainly includes the following steps: S01: Data preparation and preprocessing: Obtain the historical test data set, which should include the key input parameters: distance from the tail nozzle, ejector expansion angle; and the key output performance parameters: backflow, positive flow, and calculated ejector coefficient. Use the standardizer StandardScaler to standardize the input and output data, and convert it to a data distribution with a mean of 0 and a variance of 1 to speed up the training convergence process of the subsequent deep learning model. Then, the data set is randomly divided into training set and test set in proportion (such as 8:2).

[0042] S02: Build and train multi-task deep learning model: ​A multi-task deep learning model, ConstrainedOptimizationModel, is constructed. This model adopts a shared bottom network combined with independent top networks architecture: (1) Input layer: receives standardized distance and expansion angle two-dimensional features and converts them into Pytorch tensors.

[0043] (2) Shared hidden layer: composed of multiple fully connected layers (32->16->8 neurons) and ReLU activation functions, used to extract shared high-level features from input parameters.

[0044] (3) Multi-task output layer, including backflow prediction branch: starting from shared features, an independent fully connected network outputs the predicted value of backflow. The structure is the same as the backflow branch, which outputs the predicted value of forward flow. Ejector coefficient prediction branch: structure same as backflow branch, output the predicted value of ejector coefficient.

[0045] (4) Backflow rate calculation: during the forward propagation of the model, the predicted values of backflow and forward flow are used to calculate the backflow rate (backflow / (backflow+forward flow)) in real time.

[0046] The model is trained using the training set data, and the loss function is the sum of the mean square error (MSE) between the predicted values and the true values of each task. The Adam optimizer is used to iteratively optimize the model parameters until the loss function converges.

[0047] S03: Define the comprehensive objective function and constraints: The user inputs two weight coefficients (backflow weight) and (ejector coefficient weight), where + = 1. The weights represent the relative importance of the designer to reduce backflow and improve ejector coefficient.

[0048] Objective function A: defined as A = * ejector coefficient - * backflow. Maximizing A means seeking the optimal comprehensive performance in the weighted sense.

[0049] Constraints: backflow rate ≤ 5%. This is a hard constraint, and any solution that does not meet this condition is an infeasible solution.

[0050] S04: Global search and optimization solution: Generate high-resolution parameter grid points within the specified parameter range (distance: 0~300; expansion angle: 0~30°). Use the trained multi-task model S02 to predict the return flow, positive flow, entrainment coefficient, and backflow rate of each point in the entire parameter space.

[0051] According to the constraint condition defined in S103, filter out all grid points with backflow rate ≤5% (feasible region). Then, in the feasible region, calculate the value of the objective function A corresponding to each point, and find the point with the maximum A value. The distance and expansion angle corresponding to this point are the optimal parameter combination.

[0052] S05: Output the optimal result: The system outputs the optimal solution, including: the optimal distance from the tail nozzle and the ejector expansion angle. The predicted return flow, positive flow, entrainment coefficient, backflow rate, and the value of the objective function A under this configuration.

[0053] These results provide engineers with directly usable design references. Figure 5 a、 Figure 5 b shows the spatial position relationship between the predicted data and the actual data; Figure 5 d shows that the optimal solution (star point in the figure) is located within the backflow rate constraint boundary (red contour line) and has the maximum value of the objective function A. Figure 5 c shows the optimal solution (star point in the figure) is located within the backflow rate constraint boundary (red contour line) and has the maximum value of the objective function A.

[0054] The above is only a specific embodiment of the present application, but the protection scope of the present application is not limited thereto. Any changes or replacements within the technical scope disclosed by the present application can be easily thought of by those skilled in the art, and should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A design method for an exhaust diffuser for a high-altitude test stand, used on a test stand, characterized in that, Includes the following steps: Step 1: Solve for the inlet and outlet areas of the exhaust diffuser based on the one-dimensional thermodynamic equation; Step 2: Based on the existing high-altitude test stand, establish a scaled two-dimensional simulation model to verify the reliability of the simulation model; Step 3: Combine the selected diffuser angle and the distance between the exhaust diffuser inlet and the tailpipe outlet; combine the exhaust diffuser inlet and outlet areas obtained in Step 1 to build a simulation model for calculation, and use UDF to write the relevant functions to obtain the return flow at the exhaust diffuser inlet; Step 4: Based on simulation values, find the relationship between the distance between the exhaust diffuser inlet and the tail nozzle outlet, the exhaust diffuser expansion angle and ejector coefficient, and the return flow rate; and establish a multi-task deep learning model based on the constraint that the return flow rate is ≤5%. Through the multi-task deep learning model, find the distance between the exhaust diffuser inlet and the tail nozzle outlet and the diffusion angle that maximize the ejector coefficient under the premise of satisfying the constraint.

2. The design method for an exhaust diffuser for a high-altitude test stand according to claim 1, characterized in that, Step one specifically includes the following steps: Step 1-1, Known parameters: Total exhaust temperature of the test engine Total static temperature Specific heat ratio of ejector gas Universal gas constant ; Based on the following equations (1), (2), (3), (4), and (5): (1) Calculate the gas flow parameters of the exhaust gas or ejector gas at the tail nozzle of the test engine: exit Mach number speed coefficient aerodynamic function and and the aerodynamic function of mass flow rate and gas characteristic constants ; (2) Calculate the airflow parameters inside the high-altitude cabin: Mach number speed coefficient aerodynamic function and and gas characteristic constants : (1)； (2)； (3)； (4)； (5)； (6)； Step 1-2: Based on Step 1-1, given the target ejection coefficient, calculate the ejector area ratio using the following formula. Further determine the ejector inlet area : (7)； (8)； Steps 1-3: Calculate the contraction / expansion ratio of the exhaust diffuser. ; (1) Calculate the specific heat ratio of the mixed airflow. Specific heat at constant pressure Total temperature speed coefficient : (9)； (10)； (11)； (12)； (13)； (14)； (2) Based on the formula in step 1-1, calculate the aerodynamic function of the gas mixture. and and the aerodynamic function of mass flow rate and gas characteristic constants ; (3) Calculate the total pressure after mixing based on the isobaric conditions of the mixing chamber of the isobaric ejector. and the exhaust diffuser's contraction / expansion ratio Then, the outlet diameter of the exhaust diffuser can be calculated: (15)； (16)； (17)。 3. The design method for an exhaust diffuser for a high-altitude test stand according to claim 1, characterized in that, The two-dimensional simulation model mainly consists of a high-altitude test chamber, an exhaust diffuser, and an engine.

4. The design method for an exhaust diffuser for a high-altitude test stand according to claim 1, characterized in that, The boundary conditions of the two-dimensional simulation model are set as follows: Inlet boundary conditions: pressure inlet conditions are adopted, and the total pressure and total temperature of the airflow are set; Outlet boundary conditions: pressure outlet conditions are adopted, and the total pressure and total temperature of the airflow are set; Axisymmetry conditions: the corresponding edges are set as axes of symmetry.

5. The design method for an exhaust diffuser for a high-altitude test stand according to claim 1, characterized in that, The calculation method for the two-dimensional simulation model is set as follows: a density-based steady, implicit solver is used; the standard k-epsilon turbulence model is selected; and the implicit algorithm is used to calculate pressure and velocity.

6. The design method for an exhaust diffuser for a high-altitude test stand according to claim 1, characterized in that, The reliability of the simulation model in step two is specifically verified by comparing the high-altitude cabin pressure in the experimental data with the cabin pressure calculated by the simulation model to verify whether the simulation model is accurate.

7. The design method for an exhaust diffuser for a high-altitude test stand according to claim 1, characterized in that, In step three, the diffuser angle is selected with a degree as a gradient, and the distance between the exhaust diffuser inlet and the tailpipe outlet is selected with a mm as a gradient.

8. The design method for an exhaust diffuser for a high-altitude test stand according to claim 1, characterized in that, The specific method for establishing a multi-task deep learning model includes the following steps: S01: Data Preparation and Preprocessing Obtain a historical test dataset containing key input parameters: distance to the tail nozzle and ejector expansion angle; and key output performance parameters: return flow rate, forward flow rate, and calculated ejector coefficient. Use the StandardScaler to standardize the input and output data, transforming them into a data distribution with a mean of 0 and a variance of 1. Randomly divide the dataset into training and test sets according to the proportion. S02: Building and training multi-task deep learning models: Construct a multi-task deep learning model that employs an architecture combining a shared bottom-level network with independent top-level networks, including: Input layer: Receives standardized two-dimensional features of distance and spread angle, and converts them into PyTorch tensors; Shared hidden layer: Consists of multiple fully connected layers and ReLU activation functions, used to extract shared high-level features from the input parameters; The multi-task output layer includes: a return flow prediction branch: starting from shared features, it outputs the predicted value of the return flow through an independent fully connected network; the structure is the same as the return flow branch, and it outputs the predicted value of the forward flow; and an ejector coefficient prediction branch: the structure is the same as the return flow branch, and it outputs the predicted value of the ejector coefficient. Return rate calculation: During the forward propagation of the model, the return rate is calculated in real time using the predicted values ​​of return flow and forward flow. The model is trained using the training set data, and the loss function is the sum of the mean squared errors between the predicted values ​​and the true values ​​for each task. The Adam optimizer is used to iteratively optimize the model parameters until the loss function converges. S03: Define the comprehensive objective function and constraints: Users input the return flow weight coefficient according to design requirements. and the weighting coefficient of the entrainment coefficient ,in + =1; The above weights represent the relative importance the designer places on reducing the return flow and increasing the ejection coefficient; Objective function A: defined as A = * Gravitation coefficient - * For return flow, maximizing A means seeking the optimal overall performance in a weighted sense; Constraint: Reflux rate ≤ 5%; This is a hard constraint, and any solution that does not meet this condition is infeasible. S04: Global Search and Optimization Solution: Generate high-resolution parameter grid points within the specified parameter range; use the multi-task model trained in step S02 to predict the entire parameter space and obtain the predicted values ​​of return flow, forward flow, ejection coefficient and return rate for each point. Based on the constraints defined in S103, all grid points with a reflux rate ≤ 5% (feasible region) are selected. Within the feasible region, the value of the objective function A corresponding to each point is calculated, and the point that maximizes the value of A is found. The distance and spread angle corresponding to this point are the optimal parameter combination. S05: Output Optimal Result: The system outputs the optimal solution, including: The optimal distance from the tailpipe and ejector expansion angle; the predicted backflow rate, forward flow rate, ejector coefficient, backflow rate, and objective function A value under this configuration.

Images