Multi-electric aeroengine afterburner fuel control method and control device

Abstract

This invention discloses an afterburning fuel control method for multi-electric aero-engines, belonging to the field of aerospace engine control technology. Addressing the engine control instability caused by the dead-zone characteristics of the afterburning fuel actuator during experiments with multi-electric aero-engines using electric pumps for fuel metering, this invention designs an afterburning fuel closed-loop control loop based on a neural network inverse model and an improved μ-correction adaptive control method. This invention also discloses an afterburning fuel control device for multi-electric aero-engines. Simulation results show that the control method of this invention can effectively avoid the dead-zone region of the actuator and achieve a smooth transition of engine state throughout the entire operating range of the afterburning fuel actuator, effectively improving the stability of the control system. Furthermore, the improved μ-correction adaptive control algorithm proposed in this control method has better dynamic characteristics than the traditional μ-correction adaptive control algorithm.

Description

Technical Field

[0001] This invention relates to a method and device for controlling afterburning fuel in a multi-electric aero-engine, belonging to the field of aerospace engine control technology. Background Technology

[0002] With the increasingly widespread application of rapidly developing power electronics technology in the aerospace field, the concept of multi-electric engines has emerged. Multi-electric engines replace traditional hydraulic or mechanical control systems with electric drive systems, significantly reducing system weight and fuel consumption while improving engine maintainability and reliability. The first step in developing multi-electric engines is to implement a fuel system with an electric motor-driven fuel pump. Traditional aero-engine fuel systems primarily employ a mechanical-hydraulic structure. During engine operation, the rotor drives the accessory gearbox's drive gear to rotate, which in turn drives the main fuel pump and oil pump to provide the engine with the necessary fuel and lubricating oil. This makes the fuel pump's operating speed dependent on the engine's operating speed. In actual engine operation, the main fuel pump often provides more fuel than the system requires, necessitating the addition of a return fuel system. The complex mechanical-hydraulic structure increases engine weight, and the high temperature of the fuel after use, when returning to the fuel tank via the return fuel system, causes the fuel temperature to rise, threatening system stability. The fuel system of the electric motor uses an electric motor to drive a fuel pump for fuel supply. The electric pump itself also serves as a fuel metering device. The engine controller sends instructions to the electric fuel pump controller via an electronic controller based on the actual fuel required. The electric fuel pump controller then adjusts the motor speed to accurately supply the required amount of fuel to the engine. Due to this precise fuel supply, the engine no longer needs a return fuel system or to extract power from the engine rotor. The use of an electric fuel pump greatly simplifies the system's structural complexity, reduces its weight, lowers fuel consumption, and improves its operational stability [Thomas RE. Performance and weight impact of electric environmental control system and more electric engine on citation CJ2. In: 45th AIAA aerospace sciences meeting and exhibit, Reno, 8–11 January 2007, AIAA 2007-1395.].

[0003] Numerous scholars have conducted research on the fuel system configuration and characteristics of the More Electric Engine (MEE). The literature [Morioka, N. and Oyori, H., "Fuel Pump System Configuration for the More Electric Engine," SAE Technical Paper 2011-01-2563, 2011, https: / / doi.org / 10.4271 / 2011-01-2563.] studies fuel system configuration schemes for MEEs. After comparing several schemes, a fuel supply scheme using a permanent magnet synchronous motor to drive a fixed displacement gear pump was ultimately determined. The motor speed is controlled by a motor controller, and the precise fuel flow required by the engine is delivered. This scheme can replace the traditional fuel pump system without affecting the engine's structural design, and improves engine efficiency while reducing fuel consumption. The literature [Sun Ruhui, Zhang Tianhong, Lei Yankai. Research on flow control of electric fuel pump with compensation [J / OL]. Propulsion Technology: 1-12 [2022-11-16]. DOI:10.13675 / j.cnki.tjjs.22010025.] proposes a novel flow feedback system, which improves the accuracy of fuel metering by controlling the motor speed and verifies the feasibility of simplifying the structure of the MEE fuel pump system. Scholars have focused more on numerical simulations of the characteristics of electric fuel systems. The literature [Wei R, Ye Z. Experimental and numerical analysis of fluid-solid-thermal coupling on electric fuel pump[J]. Proceedings of the Institution of Mechanical Engineers, Part G: Journal of Aerospace Engineering, 2021, 235(11): 1427-1440.] designed an axially partitioned fuel cooling shell for electric pumps to address the motor temperature rise problem. Under different operating conditions, the cooling characteristics of the fuel cooling shell were calculated using CFD numerical simulations and experiments. The results show that the axially partitioned electric pump cooling shell can effectively reduce the motor temperature rise problem. The literature [Skawinski G. Fuel pump motor-drive systems for moreelectric aircraft[D]. University of Bath, 2010.] studied different motor drive control strategies and proposed using an improved open-loop / closed-loop strategy to replace the current control method of the electric centrifugal pump, reducing fuel flow overshoot and fuel pressure fluctuations.As a key component of the actuators in multi-electric aero-engines, the performance changes of the electric motor-driven fuel pump supply system inevitably affect the engine's operating state. For afterburning fuel systems, to improve efficiency, a zoned fuel supply method is usually adopted. Using an electric motor-driven fuel pump to measure fuel in zones will make the engine's afterburning characteristics more complex, and the nonlinearity of the afterburning fuel system may even affect the stability of the engine control system. However, there is still little research on the integrated modeling of afterburning fuel systems / engines and afterburning state control. The literature [Zhu Qing, Zhong Jialong, Jiang Yihe. High-altitude simulation test study on an improved afterburning scheme for a turbojet engine [J]. Gas Turbine Test and Research, 2000(02):23-27+62.] established a dynamic simulation program for a certain type of hybrid turbofan engine, studied the changes in the characteristics of various engine components during the dynamic process of engine afterburning on / off, and focused on the instability problem caused by sudden changes in afterburning fuel quantity. In [Liu Shuoshuo. Research on Afterburning Characteristics and Control Strategy of Turbofan Engine Based on Integrated Simulation [D]. Harbin Engineering University, 2021. DOI:10.27060 / d.cnki.ghbcu.2021.001534.], an integrated simulation platform for the exhaust nozzle adjustment device / afterburning fuel supply system and the turbofan engine body was established, and the application of linear active disturbance rejection control algorithm in exhaust nozzle area control during afterburning was carried out. The above studies on engine afterburning process control either neglect the influence of the afterburning fuel actuator or simplify the afterburning fuel actuator, failing to accurately reflect the impact of the nonlinear characteristics of the multi-electric afterburning fuel actuator on the engine afterburning characteristics, thus having limited reference value. Summary of the Invention

[0004] The technical solution to be solved by the present invention is to overcome the shortcomings of the prior art and provide a method for controlling afterburning fuel in multi-electric aero engines, which can achieve rapid, smooth and stable transition control of afterburning fuel.

[0005] The present invention specifically adopts the following technical solution:

[0006] A method for controlling afterburning fuel in a multi-electric aircraft engine, wherein the multi-electric aircraft engine achieves fuel supply through a zoned fuel supply afterburning fuel actuator; comprising the following steps:

[0007] Using flight altitude H, Mach number Ma, and thrust command F ref As input, the steady-state value W of the afterburner fuel flow rate of the multi-electric aero-engine is obtained through a pre-trained neural network inverse model. faB The training data used by the neural network inverse model is the steady-state data of the multi-electric aero-engine without afterburner fuel actuator;

[0008] According to the inlet airflow W a2The on-state threshold and saturation threshold of each fuel zone of the afterburner fuel actuator are adaptively adjusted, based on the steady-state value W of the afterburner fuel flow. faB The amplitude u of the afterburner fuel actuator is determined according to the following formula. max :

[0009] Among them, W fmin_i W fmax_i These represent the on-th and saturation thresholds of the i-th fuel zone from low to high, respectively, where N is the total number of fuel zones;

[0010] Based on the steady-state value W of the afterburner fuel flow rate faB The amplitude u of the afterburner fuel actuator max And the difference between the actual thrust F and the thrust command F ref The deviation is corrected by an improved μ-correction adaptive controller to obtain the afterburner fuel command W of the afterburner fuel actuator. fCmd The control input u of the improved μ-correction adaptive controller c (t) Specifically as follows:

[0011]

[0012] In the formula, u lin (t) represents the control input of conventional linear parameter adaptive control, k x (t), k r (t) is the adaptive gain, μ is the design constant, and Δu c (t) represents the control input u c The defect of (t).

[0013] Based on the same inventive concept, the following technical solutions can also be obtained:

[0014] A multi-electric aircraft engine afterburner fuel control device, wherein the multi-electric aircraft engine achieves fuel supply through a zoned fuel supply afterburner fuel actuator; the control device includes:

[0015] The neural network inverse model is used to calculate flight altitude H, Mach number Ma, and thrust command F. ref As input, the steady-state value W of the afterburner fuel flow rate of the multi-electric aero-engine is obtained. faB The training data used by the neural network inverse model is the steady-state data of the multi-electric aero-engine without afterburner fuel actuator;

[0016] The amplitude adjustment module is used to adjust the amplitude based on the inlet airflow W of the air intake. a2 The on-state threshold and saturation threshold of each fuel zone of the afterburner fuel actuator are adaptively adjusted, based on the steady-state value W of the afterburner fuel flow.faB The amplitude u of the afterburner fuel actuator is determined according to the following formula. max :

[0017] Among them, W fmin_i W fmax_i These represent the on-th and saturation thresholds of the i-th fuel zone from low to high, respectively, where N is the total number of fuel zones;

[0018] An improved μ-correction adaptive controller is used to adjust the afterburner fuel flow rate based on the steady-state value W. faB The amplitude u of the afterburner fuel actuator max The afterburning fuel command W of the afterburning fuel actuator is obtained by considering the deviation between the actual thrust and the thrust command. fCmd The control input u of the improved μ-correction adaptive controller c (t) Specifically as follows:

[0019]

[0020] In the formula, u lin (t) represents the control input of conventional linear parameter adaptive control, k x (t), k r (t) is the adaptive gain, μ is the design constant, and Δu c (t) represents the control input u c The defect of (t).

[0021] Preferably, the inverse neural network model includes an input layer with 3 nodes, a hidden layer with 10 nodes, an output layer with 1 node, and a normalization module and an inverse normalization module for normalizing and inverse normalizing the input data and output data, respectively.

[0022] Preferably, the inlet airflow W of the air intake duct a2 Both the actual thrust F and the actual thrust are estimated values.

[0023] More preferably, the estimated values ​​of the inlet airflow and actual thrust are obtained through a pre-trained dynamic neural network; the dynamic neural network is based on the inlet airflow W... a2 And the estimated value of the actual thrust F For the output, the following nine engine parameters are used as inputs: current step and the previous two steps' data: fan physical speed N1, compressor physical speed N2, main combustion chamber fuel flow rate W. fb The exhaust nozzle throat area is A8, and the afterburner fuel flow rate is W. faEngine inlet total pressure P2, engine inlet total temperature T2, high-pressure compressor outlet total pressure P3, afterburner inlet total temperature T6.

[0024] Compared with the prior art, the technical solution of the present invention has the following beneficial effects:

[0025] This invention can effectively avoid the problem of the afterburner fuel actuator getting stuck in the fuel dead zone, and achieve a smooth transition of engine fuel, thrust and other state variables. Compared with traditional μ-correction adaptive control, this invention can significantly improve the dynamic response of engine thrust. In the numerical simulation of thrust step, it reduces the thrust response adjustment time by 2.3s and has no overshoot throughout the process. Attached Figure Description

[0026] Figure 1 Block diagram of the afterburner fuel actuator for a multi-electric engine;

[0027] Figure 2 Model of a fuel-assisted afterburner;

[0028] Figure 3 This is a schematic diagram of a turbofan engine.

[0029] Figure 4 Schematic diagram of the closed-loop control loop for direct thrust of afterburning fuel;

[0030] Figure 5 Here is a structural diagram of the thrust estimator;

[0031] Figure 6 Simulation of the afterburning fuel actuator (H=0, Ma=1.14);

[0032] Figure 7 Simulation of afterburner fuel actuator (H=11, Ma=1.2);

[0033] Figure 8 Schematic diagram of the dead zone characteristics of an afterburning fuel actuator;

[0034] Figure 9 A schematic diagram of the dynamic response of the integrated model of the afterburner fuel actuator / engine;

[0035] Figure 10 This is a schematic diagram of the CMAC control system structure;

[0036] Figure 11 This is a schematic diagram of the inverse model structure of a neural network;

[0037] Figure 12 This is a schematic diagram of the amplitude adjustment module structure;

[0038] Figure 13 This is a schematic diagram illustrating the principle of amplitude adjustment by the actuator.

[0039] Figure 14 A schematic diagram of the improved μ-correction adaptive controller structure;

[0040] Figure 15 This is a schematic diagram of the training error of the thrust estimator.

[0041] Figure 16 This is a schematic diagram of the test error of the thrust estimator;

[0042] Figure 17 Simulation results comparing the performance of the PID control algorithm and the CMAC system;

[0043] Figure 18 Simulation results comparing the effects of traditional μ-correction adaptive control and the improved μ-correction adaptive control. Detailed Implementation

[0044] To address the shortcomings of existing technologies, this invention proposes a solution through simulation to establish a comprehensive model of the afterburning fuel actuator / multi-electric aero-engine, explore the impact of the actuator's dead-zone characteristics on engine performance, and design an afterburning fuel closed-loop control system (CMAC) based on a neural network inverse model and an improved μ-correction adaptive control method for the afterburning fuel control loop. This aims to resolve the impact of the actuator's dead-zone characteristics on the control system and achieve rapid, smooth, and stable transition control of the afterburning fuel.

[0045] The specific technical solution proposed in this invention is as follows:

[0046] A method for controlling afterburning fuel in a multi-electric aircraft engine, wherein the multi-electric aircraft engine achieves fuel supply through a zoned fuel supply afterburning fuel actuator; comprising the following steps:

[0047] Using flight altitude H, Mach number Ma, and thrust command F ref As input, the steady-state value W of the afterburner fuel flow rate of the multi-electric aero-engine is obtained through a pre-trained neural network inverse model. faB The training data used by the neural network inverse model is the steady-state data of the multi-electric aero-engine without afterburner fuel actuator;

[0048] According to the inlet airflow W a2 The on-state threshold and saturation threshold of each fuel zone of the afterburner fuel actuator are adaptively adjusted, based on the steady-state value W of the afterburner fuel flow. faB The amplitude u of the afterburner fuel actuator is determined according to the following formula. max :

[0049] Among them, W fmin_i W fmax_iThese represent the on-th and saturation thresholds of the i-th fuel zone from low to high, respectively, where N is the total number of fuel zones;

[0050] Based on the steady-state value W of the afterburner fuel flow rate faB The amplitude u of the afterburner fuel actuator max And the difference between the actual thrust F and the thrust command F ref The deviation is corrected by an improved μ-correction adaptive controller to obtain the afterburner fuel command W of the afterburner fuel actuator. fCmd The control input u of the improved μ-correction adaptive controller c (t) Specifically as follows:

[0051]

[0052] In the formula, u lin (t) represents the control input of conventional linear parameter adaptive control, k x (t), k r (t) is the adaptive gain, μ is the design constant, and Δu c (t) represents the control input u c The defect of (t).

[0053] Based on the same inventive concept, the following technical solutions can also be obtained:

[0054] A multi-electric aircraft engine afterburner fuel control device, wherein the multi-electric aircraft engine achieves fuel supply through a zoned fuel supply afterburner fuel actuator; the control device includes:

[0055] The neural network inverse model is used to calculate flight altitude H, Mach number Ma, and thrust command F. ref As input, the steady-state value W of the afterburner fuel flow rate of the multi-electric aero-engine is obtained. faB The training data used by the neural network inverse model is the steady-state data of the multi-electric aero-engine without afterburner fuel actuator;

[0056] The amplitude adjustment module is used to adjust the amplitude based on the inlet airflow W of the air intake. a2 The on-state threshold and saturation threshold of each fuel zone of the afterburner fuel actuator are adaptively adjusted, based on the steady-state value W of the afterburner fuel flow. faB The amplitude u of the afterburner fuel actuator is determined according to the following formula. max :

[0057] Among them, W fmin_i W fmax_i These represent the on-th and saturation thresholds of the i-th fuel zone from low to high, respectively, where N is the total number of fuel zones;

[0058] An improved μ-correction adaptive controller is used to adjust the afterburner fuel flow rate based on the steady-state value W. faB The amplitude u of the afterburner fuel actuator max The afterburning fuel command W of the afterburning fuel actuator is obtained by considering the deviation between the actual thrust and the thrust command. fCmd The control input u of the improved μ-correction adaptive controller c (t) Specifically as follows:

[0059]

[0060] In the formula, u lin (t) represents the control input of conventional linear parameter adaptive control, k x (t), k r (t) is the adaptive gain, μ is the design constant, and Δu c (t) represents the control input u c The defect of (t).

[0061] To facilitate public understanding, the technical solution of the present invention will be described in detail below through a specific embodiment and in conjunction with the accompanying drawings:

[0062] To improve afterburner fuel combustion efficiency, aero-engine afterburners typically employ a zoned fuel supply system. The afterburner fuel actuator in this embodiment is as follows: Figure 1 As shown, it contains three zones: Zone I, Zone II, and Zone III. In the diagram, P2 and T2 represent the total temperature and total pressure at the intake outlet, respectively; W... fCmd The total afterburner fuel flow command transmitted from the engine control system; W fRef Provide additional fuel flow commands to each fuel zone; n pRef This refers to the motor speed command in the motor speed control circuit; n p W represents the actual motor speed. f W represents the actual fuel output of each fuel zone. fA The total fuel output by the afterburner fuel actuator is represented by the subscripts "I", "II", and "III", which represent fuel zones I, II, and III, respectively. Each fuel zone has a basically the same structure, consisting of an electric pump (including a permanent magnet synchronous motor, a motor controller, and a fixed displacement gear pump), a pressure sensor, a fuel metering controller, and equivalent nozzles. The electric gear pump has two functions: one is to pressurize the fuel as usual, and the other is to act as a metering pump to meter the fuel in each zone of the afterburner. The electric pump changes the speed of the gear pump by controlling the motor speed to achieve accurate fuel metering, as shown in equation (1). Due to the introduction of the metering pump, there is no need to add a return fuel system to the fuel system.

[0063] Q f=2πKzm 2 bn p η v (1)

[0064] In the formula, Q f The output fuel quantity of the electric pump, m is the gear module, z is the number of teeth, b is the tooth width, and n is the gear diameter. p For the gear pump speed, K is taken as 1.0875, η v The volumetric efficiency is determined by experiments and is generally between 0.7 and 0.95.

[0065] The design parameters of the electric pump used in this embodiment are shown in Table 1. Due to the limitation of the motor speed range, the fuel pump fuel metering range has a maximum value Q. max and minimum value Q min (Q max Q min >0 kg / s). The fuel pump used in this embodiment has a metering range of 0.03 to 0.75 kg / s. When the required fuel quantity is within the fuel metering range, the fuel pump can accurately measure the required fuel quantity, but when the fuel quantity is less than the minimum value Q of the metering range... min or greater than the maximum value Q in the measurement interval max At this time, the fuel pump cannot output the specified fuel, as shown in equation (2).

[0066]

[0067] In the formula, W fRef This indicates the amount of fuel required.

[0068] Maximum value Q of fuel metering range for electric pump max and minimum value Q min The presence of this will introduce significant nonlinearity into the afterburner fuel actuator.

[0069] Table 1 Main parameters of electric pumps

[0070]

[0071]

[0072] Figure 1 In the table, the fuel pump characteristic table reflects the fuel output W of the fuel pump. f With fuel pump speed n pThe relationship between the inlet and outlet pressure difference ΔP and the characteristic data can usually be obtained through experiments. A pressure sensor is installed at the outlet of the electric fuel pump to monitor the gear pump outlet pressure in real time. The fuel output flow rate is obtained by interpolation using a flow-speed-pressure characteristic data table summarized from experimental data. The error between the output flow rate and the set flow rate is fed back to the fuel metering controller, which controls the speed of the electric fuel pump to ensure that the actual fuel supply matches the set fuel supply.

[0073] To ensure the combustion efficiency of the afterburner, it is essential to rationally design the fuel distribution rules for each zone of the afterburner. A good afterburner fuel distribution rule can improve the combustion efficiency of the afterburner while ensuring a uniform temperature distribution at the combustion chamber outlet. Literature has conducted performance tests on two fuel distribution schemes under different residual gas coefficients for the afterburner. The results show that the iso-oxygen content method of fuel distribution design can effectively improve the performance of the afterburner. To ensure that each zone of the afterburner operates near the optimal residual gas coefficient, the afterburner fuel distribution rules adopted in this embodiment are shown in equations (3) to (5).

[0074] 1) When W fCmd <30%×W fmax Only enable partition I:

[0075]

[0076] 2) When 30% × W fmax ≤W fCmd <60%×W fmax Simultaneously enable partitions I and II:

[0077]

[0078] 3) When W fCmd ≥60%×W fmax Simultaneously enable partitions I, II, and III:

[0079]

[0080] In the formula, W fmax The theoretical maximum fuel quantity that does not exceed the afterburner shutdown boundary under the current intake airflow conditions is calculated using the formula shown in equation (6).

[0081] W fmax =far max ×W a2 (6)

[0082] In the formula, far max The oil-gas ratio at the oil-rich boundary can be obtained by interpolation from the afterburner characteristic diagram; W a2This refers to the airflow rate at the air intake, which varies with flight conditions (altitude H, Mach number Ma).

[0083] Based on the modeling scheme for afterburning fuel actuators proposed above, a simulation model of the afterburning fuel actuator was established using AMESim and Simulink. First, separate fuel sub-region (region I, region II, region III) models were built on the AMESim platform, as follows: Figure 2 As shown in (a), a fixed displacement gear pump is simulated using a fixed displacement pump 1 and a variable throttle valve 5, where the fixed displacement pump 1 simulates the theoretical oil supply of the gear pump, and the variable throttle valve 5 simulates the leakage flow of the gear pump. A first-order inertial element 3 simulates the transfer function of the motor 2, and a limiting element 4 simulates the controllable speed range of the motor. A pressure sensor 6 is used to monitor the outlet pressure of the gear pump. An overflow valve 7 simulates an equivalent nozzle for adjusting the outlet pressure of the gear pump. A PID controller 9 is the motor speed controller, and module 8 is the flow-speed-pressure conversion function.

[0084] Export each fuel zone model in FMU format, and build an overall model of the afterburner fuel actuator on the Simulink platform, such as... Figure 2 As shown in (b), the module includes a fuel distribution rule module (Equations (3-5)), the I, II, and III zone FMU models, and a summation module. The summation module is used to calculate the total fuel flow rate W in the afterburner. fA As shown in equation (7).

[0085] W fA =W fRef_I +W fRef_II +W fRef_III (7)

[0086] The aerodynamic-thermodynamic model of the turbofan engine is established using the component method. Its structure is as follows: Figure 3 As shown, the components of a turbofan engine include the intake manifold, fan, compressor, combustion chamber, high-pressure turbine, low-pressure turbine, bypass duct, mixing chamber, afterburner, and exhaust nozzle. The afterburner fuel actuator's influence is mainly reflected in the fuel supply characteristics during engine afterburning, which in turn affects other engine parameters.

[0087] Direct thrust control has received widespread attention due to its superior performance. The block diagram of the afterburning fuel direct thrust closed-loop control loop is shown below. Figure 4 As shown.

[0088] The engine's thrust requirement is determined based on flight altitude, Mach number, and throttle position. The goal of the control system is to ensure that the engine's output thrust tracks the required thrust, while maintaining stable, oscillating operation of the entire system. During flight, the engine thrust F and the inlet airflow W... a2Neither of these parameters can be directly measured; therefore, they are estimated using a dynamic neural network to obtain the thrust estimate. As the actual thrust feedback quantity, and simultaneously, the estimated airflow rate. The command is transmitted to the afterburner fuel actuator to calculate the fuel quantity instructions for each fuel zone, as shown in equations (2-5). The dynamic neural network structure is as follows: Figure 5 As shown.

[0089] This dynamic neural network thrust estimator takes the current step (k) and the data from the previous two steps (k-1, k-2) of nine engine parameters as input. The inputs for the first two steps are represented by two delay cycles. The nine parameters are: fan physical speed N1, compressor physical speed N2, main combustion chamber fuel flow rate W. fb The exhaust nozzle throat area is A8, and the afterburner fuel flow rate is W. fa Engine inlet total pressure P2, engine inlet total temperature T2, high-pressure compressor outlet total pressure P3, afterburner inlet total temperature T6; neural network output is the estimated engine thrust. And estimate the inlet airflow of the air intake. Use function T DNN The inputs and outputs of the thrust estimator are described as follows:

[0090]

[0091] At ground point H=0, Ma=1.14, a dynamic performance simulation analysis was performed on the afterburner fuel actuator established above. The fuel command was set as ramp command and step command respectively. The simulation results are as follows: Figure 6 As shown.

[0092] By the ramp response of the actuator ( Figure 6 As shown in (a)), with the increase of fuel command, the afterburner fuel actuator gradually opens and outputs fuel in the order of I→II→III. When the previous zone reaches its maximum fuel supply, the next zone opens. When the fuel command decreases, the afterburner fuel actuator gradually closes in the order of III→II→I. It is worth noting that when a certain zone is opened or closed, the total fuel output W... fA There are fluctuations to some extent, and it is not strictly linear. This is due to the disturbance to the total fuel output caused when a new fuel supply range is added or removed from the system.

[0093] Step response of the actuator ( Figure 6 As can be seen from (b) in the figure, in most cases, the output fuel of the afterburner fuel actuator can track the command fuel W. fA =W fCmd However, steady-state errors exist in certain fuel supply ranges. For example, in the diagram, when the fuel command is W... fCmdWhen the fuel consumption is 0.59 (unit: kg / s, the same below), the actual fuel output W of the afterburner fuel actuator is... fA =0.5625≠W fCmd There is a steady-state error. The reasons are analyzed as follows:

[0094] Under flight conditions of H=0 and Ma=1.14, the theoretical maximum fuel quantity not exceeding the afterburner shutdown boundary is:

[0095] W fmax =far max ×W a2 =1.875 (9)

[0096] In the formula, W is the inlet airflow of the intake duct. a2 This can be calculated from the engine model. According to the afterburner fuel distribution rules, as shown in equations (3-5), the fuel supply amounts for each zone are as follows:

[0097]

[0098] Due to Zone II fuel quantity directive W fRef_II =0.275 is less than the minimum value Q of the electric pump's metering range. min (Q min =0.03), the Zone II fuel pump cannot output the specified fuel, and its actual output fuel W f_II =0, thus causing the total output fuel W of the afterburner fuel actuator to be 0. fA There is a steady-state error.

[0099] Similarly, when fuel command W fCmd When the value is 1.15, the electric pump in Zone III cannot measure the output Q. min A fuel quantity of less than 0.03 causes the afterburner fuel actuator to actually output fuel W. fA =1.125≠W fCmd There is a steady-state error. This paper defines the fuel supply intervals where the output fuel cannot track the fuel command as the dead zone intervals of the afterburner fuel actuator. From the above analysis, it can be seen that there is a dead zone operating interval for each fuel zone activated, and the length of the dead zone interval ΔW fdz It should be equal to the minimum measurable fuel value Q in each zone. min In this article, ΔW fdz =0.03, which is mainly due to the minimum value of the electric pump's metering range.

[0100] Next, at an altitude of H = 11 km and Ma = 1.2, a dynamic performance simulation analysis of the afterburner fuel actuator was conducted, and the results are as follows. Figure 7As shown. Similar to the ground point, the output fuel exhibits steady-state error in certain fuel supply ranges, i.e., dead zones, and near these dead zones, the fuel ramp response shows significant fluctuations. It is worth noting that due to changes in flight conditions, the theoretical maximum fuel flow rate W in the afterburner... fmax With the inlet air flow rate W a2 As can be seen from equations (3) to (6), the starting and ending points of the fuel flow dead zone will also change.

[0101] In summary, for the afterburner fuel actuator described in this paper, there are a total of 3 fuel dead zones, distributed from 0 to Q after each zone is activated. min (Q min =0.03) within the fuel range. When fuel command W fCmd When the actuator is in the dead zone, the actual fuel output W fa Unable to track fuel commands, resulting in steady-state error. Figure 8 Table 2 summarizes the dead zone characteristics of afterburning fuel actuators.

[0102] Table 2 Dead Zone Characteristics of Multi-Electrical Fuel Actuators

[0103]

[0104]

[0105] To further verify the impact of the nonlinear characteristics of the afterburning fuel actuator on the engine, a combined model of the afterburning fuel actuator / engine was tested under ground conditions of H=0 and Ma=1.14. Figure 9 The simulation of the variable throttle lever shown in (a) is as follows: Figure 9 As shown in (b) to (f) in the simulation. During the simulation, the afterburner fuel was used... Figure 4 The closed-loop control shown uses a traditional proportional-integral control controller.

[0106] Depend on Figure 9 Simulation results show that, in most cases, the afterburner fuel control system can ensure good engine stability and dynamic performance. During the process of pushing the throttle lever from 78° to 85°, the total temperature T at the afterburner combustion chamber outlet... 75 The thrust F increases steadily with the increase of the throttle lever angle PLA, such as Figure 9 As shown in (b) and (c) in the figure. However, at t = 29s, T 75 F and F fluctuated slightly. Figure 9 As shown in (d), at this time, the afterburner I zone reaches the maximum fuel supply limit, and the afterburner fuel actuator II zone is activated. The change in the dynamic characteristics of the actuator causes fluctuations in the total output fuel, which in turn causes fluctuations in the engine state.

[0107] When the throttle lever angle PLA = 92°, T 75 The throttle levers F and F exhibited sustained vibrations of approximately 0.5% and 0.3% respectively. This was due to the engine thrust command F at that throttle lever angle. ref The corresponding demand for enhanced fuel W fCmd Located in the dead zone of the afterburner fuel actuator, afterburner zone III cannot output the given fuel, thus causing T 75 And the continuous shaking of F.

[0108] The afterburning fuel actuator also poses a threat to the stability of the core engine control system, such as... Figure 9 As shown in (e) and (f) in the figure, when the required afterburning fuel is located in the dead zone of the actuator, both the engine fan speed and compressor speed exhibit slight fluctuations. The simulation results show that the nonlinear characteristics of the afterburning fuel actuator severely affect the stability and dynamic steady-state performance of the engine control system, making research on the stability control of afterburning fuel systems of great significance.

[0109] Based on the above analysis, in order to improve the control performance degradation that occurs in the dead zone of the afterburner fuel actuator by traditional PI control, this invention combines the advantages of neural network inverse control and μ-correction adaptive control, and proposes a composite correction adaptive controller (CMAC) design method for the first time for the afterburner fuel closed-loop control loop.

[0110] μ-correction adaptive control is a type of model reference adaptive control with resistance to input saturation, based on Lyapunov stability design. It can adjust adaptive control parameters to ensure that the dynamic output of the controlled object tracks the desired trajectory given by the reference model as much as possible when the characteristics of the controlled object change. Furthermore, it can provide a stable adaptive rate even when the actuator has a maximum amplitude limit. This invention designs a μ-correction adaptive controller based on the integrated model of the afterburning fuel actuator / engine established above.

[0111] (1) State-space equations of the controlled object considering input saturation

[0112] The state-space equation of the controlled object is given by the following differential equation:

[0113]

[0114] Where x∈R n is the state vector of the controlled object. A is a matrix with unknown parameters. b is a constant vector, and λ is an unknown constant with a known sign.

[0115] For an input vector u(t) with saturation properties, it can be described as follows:

[0116]

[0117] Where u c (t) is the control input, u max >0 defines the saturation amplitude of the actuator.

[0118] Substituting equation (12) into equation (11), we obtain the state variable equation of the controlled object, as shown in equation (13):

[0119]

[0120] where Δu(t)=u(t)-u c (t) indicates that due to the amplitude saturation constraint of the actuator, u max (t) caused control defects.

[0121] The state value x(t) of the controlled object selected in this paper is the total temperature F at the outlet of the afterburner chamber of the turbofan engine. n The input value u(t) is the amount of afterburning fuel W of the turbofan engine. fa .

[0122] Choose a constant δ (0 < δ) max ), and define u δ max =u max -δ. Control defects can be represented as:

[0123] Δu(t)=Δu c (t)+Δ sat (t) (14)

[0124] in

[0125]

[0126] The μ-corrected control law is defined as follows:

[0127]

[0128] Where u lin (t) represents conventional linear parameter adaptive control, k x (t)∈R n ,k r (t)∈R is the adaptive gain, and μ is the design constant.

[0129] Substituting equations (6) and (8) into equation (5), the closed-loop dynamic equation is as follows:

[0130]

[0131] ​

[0132] Δu lin (t) defines the linear parameter adaptive signal u lin The defect of (t).

[0133] (2) Reference model with instruction correction

[0134] Based on the state variable model given in equation (17), the following adaptive reference model is established:

[0135]

[0136] Comparing the parameters of (19) and (17), when the two models are perfectly matched, we will get:

[0137]

[0138] Where k x * ,k r * ,k u * It is the value when the reference model and the actual model are precisely matched.

[0139] When the input of the controlled object does not exceed the amplitude limit of the actuator, Δu lin =0, at this point μ correction degenerates into conventional adaptive control. When the input of the controlled object exceeds the amplitude limit of the actuator, Δu lin ≠0, the input instructions of the reference model are adaptively modified according to equation (19). At this time, the saturation of the actuator is avoided by modifying the instructions.

[0140] (3) μ Correction control law

[0141] Define the error tracking signal as: e(t) = x(t) - x m (t), whose dynamic equation can be written as:

[0142]

[0143] Where, Δk x (t)=k x (t)-k x * ,Δk r (t)=k r (t)-k r * ,Δk u (t)=k u (t)-k u * This represents the adaptive parameter error.

[0144] Consider the following adaptive control law:

[0145]

[0146] Where γ x =γ x T >0,γ r >0,γ u >0 is the adaptive parameter adjustment speed coefficient.

[0147] In traditional μ-correction controllers, the actuator amplitude is often constant, which effectively solves the anti-saturation control problem of fixed and singular actuator amplitude. However, for the multi-electric afterburner fuel actuator established in this paper, there are multiple fuel dead zones, and the specific range of the dead zone interval changes with flight conditions. To avoid the actuator getting stuck in the dead zone operating range and to fully utilize the function of each fuel zone, the actuator amplitude needs to be adaptively changed under different flight conditions and throttle conditions. Traditional μ-correction controllers cannot meet the requirements. Therefore, this invention combines a neural network inverse model to develop a neural network inverse-adaptive control system with adaptively adjustable actuator amplitude, called the Composite Correction Adaptive Control System (CMAC).

[0148] The CMAC system is essentially a neural network inverse controller. It uses a static BP neural network to approximate the inverse of the controlled object, and then connects an improved μ-correction controller in parallel with the static neural network to compensate for the static neural network's errors, thus forming a dynamic neural network inverse controller. The structure of the CMAC system is as follows: Figure 10 As shown, it consists of three parts: a neural network inverse model, an amplitude adjustment module, and an improved μ-correction adaptive controller.

[0149] (1) Neural Network Inverse Model

[0150] First, steady-state data of the afterburning state of the integrated model of the afterburning fuel actuator / engine is collected within the wide envelope, and then measured using flight altitude H, Mach number Ma, and thrust command F. ref The steady-state value W of the afterburner fuel quantity serves as the input to the neural network. faB As the output of the neural network, the neural network with the required envelope range is trained. The data required for training the inverse model of the neural network is collected based on a PI closed-loop system. It should be noted that during the training data collection process, the controlled object is the engine, and no afterburner fuel actuator is added. A total of 1615 sets of steady-state data were selected within the envelope. Due to the relatively small amount of data, a single hidden layer can generally achieve satisfactory training results. After trial and error, a three-layer neural network 3-10-1 was obtained, as shown below. Figure 11 As shown. To avoid data overload caused by different input data magnitudes and to adapt to the training requirements of neural networks, the input and output variables D of the neural network inverse model are...i Normalize to 0 to 1, as shown in equation (23).

[0151]

[0152] In the formula, D max D represents the maximum value of each variable. min The minimum value of each variable. These are the normalized variables. Accordingly, in practical use, the normalized variables of the neural network output need to be denormalized.

[0153] Use function T BP The inverse model of the neural network is described as shown in equation (24).

[0154]

[0155] Since the inverse model of the neural network is a static neural network, both the input and output are the data from the current step (k). Thrust command F ref Based on the flight altitude, Mach number, and throttle position, the CMAC control system can obtain the steady-state value W of the afterburner fuel flow corresponding to the current throttle position under any flight condition within the engine envelope. faB Because the dead-zone characteristics of the afterburner actuator are not considered during the data acquisition process, W faB For ideal afterburner fuel steady-state output.

[0156] (2) Amplitude Adjustment Module

[0157] The amplitude adjustment module can adaptively adjust the amplitude of the actuator in real time according to the current working status. The structure diagram is as follows: Figure 12 As shown, the module input is the ideal steady-state value W of the afterburner fuel. faB Inlet airflow W a2 The output is the actuator amplitude u. max Actuator amplitude u max The calculation method is as shown in equations (25-28) and Figure 13 As shown:

[0158] 1) When W faB ≤W fmin_I Do not open any partitions:

[0159] u max =0 (25)

[0160] 2) When W fmin_I <W faB ≤W fmin_II At this point, only partition I is enabled:

[0161] u max =W fmax_I(26)

[0162] 3) When W fmin_II <W faB ≤W fmin_III Simultaneously enable partitions I and II:

[0163] u max =W fmax_II (27)

[0164] 4) When W faB >W fmin_III Simultaneously enable partitions I, II, and III:

[0165] u max =W fmax_III (28)

[0166] In the formula, W fmin_Ⅰ W fmax_Ⅰ W fmin_ⅠI W fmax_ⅠI W fmax_ⅠII The calculation method is shown in Table 2.

[0167] The CMAC control system can adaptively adjust the actuator amplitude u through this module. max On the one hand, as flight conditions change, the estimated inlet flow rate of the air intake... Adaptively adjust the limiting parameters W for each fuel zone fmin_Ⅰ W fmax_Ⅰ W fmin_ⅠI W fmax_ⅠI W fmax_ⅠII On the other hand, by using the ideal steady-state value W of the afterburner fuel faB By comparing the parameters with the limits of each fuel zone, the number of fuel zones that need to be activated is determined, and then the u is adjusted accordingly. max Correspondence with fuel zone limits: As W faB The increase of W faB When the actuator is in the dead zone, the actuator amplitude u is limited. max To prevent the actuator from getting stuck in the dead zone, the maximum value of the previous partition is used. faB As you continue to increase the fuel dead zone and enter the next normal operating range, u max Adjust the fuel supply to the maximum value for the next zone and start fuel supply for the next zone.

[0168] (3) Improved μ correction controller

[0169] Improved μ-correction adaptive controller structure as follows Figure 14 As shown, unlike traditional μ-correction adaptive control, its control input amplitude u maxIt is not fixed; its value is calculated by the amplitude adjustment module and adjusted in real time according to flight conditions and throttle position. On the other hand, improvements to u c The calculation formula for (t) is based on formula (16), with the addition of the neural network output W. faB As shown in equation (29).

[0170]

[0171] Improved u c (t) The calculation formula (29) consists of two parts: 1) the output W of the inverse model of the neural network. faB ;2) Output u of the traditional μ-corrected adaptive control law lin (t)+μΔu c (t). Output W of the inverse neural network model faB u c (t) It quickly adjusts to the steady-state value corresponding to the current control command, which can significantly improve the dynamic performance of the traditional μ-correction adaptive controller; while the output u of the traditional μ-correction adaptive controller... lin (t)+μΔu c (t) is used to compensate for the construction error of the neural network inverse system, thereby improving the steady-state accuracy of the controller and eliminating the effect of actuator saturation. The specific compensation mechanism is as follows: when there is a dynamic steady-state error between the engine output and the control command, the μ-corrected adaptive controller will output a compensating control quantity, making the engine thrust output track the command input. Therefore, the neural network inverse model output W... faB The introduction of this technology can significantly improve the dynamic response speed of traditional μ-correction controllers, while the output u of traditional μ-correction controllers... lin (t)+μΔu c (t) is the output W of the inverse model of the neural network. faB The compensation amount can eliminate the steady-state error of the neural network inverse model. It can be expected that the improved μ-correction controller will have superior dynamic and steady-state performance.

[0172] To verify the effectiveness of the CMAC system, an integrated model of the afterburning fuel actuator / turbofan engine was used as the controlled object, and numerical simulation verification of the control system was carried out based on the MATLAB / Simulink simulation platform, including two parts: neural network inverse model accuracy simulation and afterburning control simulation.

[0173] The accuracy of the neural network inverse model directly affects the control performance of the CMAC system. Figure 15 The training results of the estimator are shown. The relative error of all data is less than 7%, the relative deviation of more than 99% of the data is less than 0.2%, and the average relative error is 2.11×10-6, indicating that the training results of the training set have high accuracy. Figure 16The test results of the neural network inverse model on the test dataset are shown. From Figure 16 As can be seen, the fitting results of the neural network are satisfactory.

[0174] To verify the effectiveness of the control system of this invention, PID control algorithm and CMAC system were used as controllers to perform closed-loop control on the integrated model of afterburner fuel actuator / turbofan engine. The operating points were selected as H=0, Ma=1.14, and the throttle lever PLA changed as follows: Figure 17 As shown in (a) above, the simulation results are as follows: Figure 17 As shown in (b) to (h) in the diagram.

[0175] from Figure 17 As shown in (c) to (f), when the engine demand thrust command causes the afterburner fuel to enter the dead zone of the afterburner fuel actuator, the afterburner fuel flow will fluctuate continuously in the dead zone under the traditional PID control mode. The CMAC control system modifies the engine demand thrust command, and the modified command can effectively prevent the actuator from entering the dead zone. When the demand thrust command further increases to the normal operating range of another fuel zone, the reference model output command again tracks the actual demand thrust command, and the new fuel zone opens normally, realizing a smooth transition of engine thrust, afterburner fuel, and other states, effectively improving the performance stability of the fuel mechanism when operating in the dead zone.

[0176] like Figure 17 As shown in (g) and (h), both CMAC control and PID control can achieve stable thrust control within the normal operating range of the fuel actuator. During the thrust increase or decrease process, PID control will exhibit a slight overshoot, while CMAC control system can ensure no overshoot throughout the entire process and the adjustment time is also reduced compared to PID control.

[0177] To further verify the advantages of the improved μ-correction control algorithm proposed in the CMAC control system, throttle lever step and ramp simulations were performed using both traditional μ-correction adaptive control and the improved μ-correction control. Figure 18 As shown in (a), the simulation results are as follows Figure 18 As shown in (b) to (f) in the diagram.

[0178] Compared with the traditional μ-correction adaptive control, the improved μ-correction control shows a significant advantage in thrust dynamic response performance. Figure 18 In the throttle lever step process shown in (c) and (d), the adjustment time was shortened by 2.3s and 1.5s respectively. Figure 18In the throttle ramp adjustment process shown in (e) and (f), the adjustment time was shortened by 0.42s and 0.6s respectively, significantly improving engine maneuverability. Meanwhile, by Figure 18 As can be seen from (c) and (d), during the throttle lever step operation, the improved μ-correction control can completely eliminate the overshoot phenomenon of the traditional μ-correction control, which can effectively prevent overheating at the afterburner outlet. The overall dynamic simulation results verify that the improved μ-correction control has better control performance than the traditional μ-correction control when the engine state changes.

Claims

1. A method for controlling afterburner fuel of a more-electric aircraft engine, the more-electric aircraft engine being supplied with fuel by a zoned fuel actuator for afterburner fuel; characterized in that, Comprising the following steps: with the flight altitude H, the Mach number Ma, the thrust command F ref As input, a steady state value W of the afterburner fuel flow of the multi-electric aero-engine is obtained by a pre-trained neural network inverse model faB The training data used by the neural network inverse model is steady state data of the multi-electric aero-engine without afterburner fuel actuator According to the air flow W a2 Adaptively adjusting the on threshold and the saturation threshold of each fuel partition of the afterburning fuel actuator, and according to the afterburning fuel flow steady state value W faB , the amplitude u of the afterburning fuel actuator is determined according to the following formula max : W fmax_i respectively represent the on threshold and the saturation threshold of the ith fuel partition from low to high for the force fuel actuator, and N is the total number of fuel partitions. According to the steady value of the afterburner fuel flow W faB , the amplitude of the afterburner fuel actuator u max , and the deviation between the actual thrust F and the thrust command F ref , the afterburner fuel command W fCmd of the afterburner fuel actuator is obtained by an improved μ-modified adaptive controller, the control input u c of which is specifically as follows: where u lin (t) denotes the control input of the conventional linear parameter-varying control, k x (t), k r (t) is the adaptive gain, μ is a design constant, Δu c (t) is the control input u c (t) of the conventional linear parameter-varying control.

2. The method of afterburner fuel control for a more-electric aeroengine of claim 1, wherein, The neural network inverse model comprises an input layer with 3 nodes, a hidden layer with 10 nodes, an output layer with 1 node, and a normalization module and a denormalization module for normalizing and denormalizing input data and output data respectively.

3. The method of afterburner fuel control for a more-electric aeroengine of claim 1, wherein, The intake port inlet air flow rate W a2 And the actual thrust F are estimated values.

4. The method of afterburner fuel control for a more-electric aeroengine of claim 1, wherein, The intake inlet air flow and the actual thrust are estimated by a pre-trained dynamic neural network, which takes as input the current and previous two steps of the following nine engine parameters: fan physical speed N1, compressor physical speed N2, main combustor fuel flow W a2 , and the estimated values of the actual thrust F , and the estimated values of the actual thrust F fb , and the estimated values of the actual thrust F fa , and the estimated values of the actual thrust F fb , and the estimated values of the actual thrust F fa , and the estimated values of the actual thrust F fb , and the estimated values of the actual thrust F fa , and the estimated values of the actual thrust F fb , and the estimated values of the actual thrust F fa , and the estimated values of the actual thrust F fb , and the estimated values of the actual thrust F fa , and the estimated values of the actual thrust F fb , and the estimated values of the actual thrust F fa , and the estimated values of the actual thrust F fb , and the estimated values of the actual thrust F fa , and the estimated values of the actual thrust F fb , and the estimated values of the actual thrust F fa , and the estimated values of the actual thrust F fb , and the estimated values of the actual thrust F fa , and the estimated values of the actual thrust F fb , and the estimated values of the actual thrust F fa , 5. A control device for afterburner fuel of a multi-electric aero-engine, the multi-electric aero-engine being supplied with fuel by an afterburner fuel actuator of zoned fuel supply; characterized in that, The control device comprises: a neural network inverse model for obtaining, as an output, a steady state value W of the afterburner fuel flow of the multi-electric aero-engine as a function of the flight altitude H, the Mach number Ma, the thrust command F ref as an input, the steady state value W of the afterburner fuel flow of the multi-electric aero-engine faB ; the training data used by the neural network inverse model being steady state data of the multi-electric aero-engine without afterburner fuel actuator An amplitude adjustment module is configured to determine an amplitude u of the afterburner fuel actuator according to the intake inlet air flow W a2 adaptively adjusts the on threshold and the saturation threshold of each fuel partition of the afterburner fuel actuator, and determines the amplitude u of the afterburner fuel actuator according to the afterburner fuel flow steady state value W faB , according to the following formula max : W fmax_i respectively represent the on threshold and the saturation threshold of the ith fuel partition from low to high for the force fuel actuator, and N is the total number of fuel partitions. Improved mu-recovery adaptive controller for obtaining a reheat fuel command W faB , the amplitude of the reheat fuel actuator u max and the deviation between actual thrust and thrust command, to obtain a reheat fuel command W fCmd for the reheat fuel actuator; the control input u c of the improved mu-recovery adaptive controller is given by t) as follows: where u lin (t) is the control input, k x (t) is the control input, k r (t) is the control input, k c (t) is the control input, k c (t) is the control input, k 6. The afterburner fuel control system for a more-electric aircraft engine of claim 5, wherein, The neural network inverse model comprises an input layer with 3 nodes, a hidden layer with 10 nodes, an output layer with 1 node, and a normalization module and a denormalization module for normalizing and denormalizing input data and output data respectively.

7. The afterburner fuel control system for a more-electric aircraft engine of claim 5, wherein, The intake port inlet air flow rate W a2 And the actual thrust F are estimated values.

8. The afterburner fuel control system for a more-electric aircraft engine of claim 7, wherein, The estimated values of the inlet air flow and the actual thrust are estimated by a pre-trained dynamic neural network; the dynamic neural network is trained with the estimated values of the inlet air flow W a2 and the actual thrust F as output and with the current step and the previous two steps of the following nine engine parameters as input: the fan physical speed N1, the compressor physical speed N2, the main combustion chamber fuel flow W fb , the nozzle throat area A8, the afterburner fuel flow W fa , the engine inlet total pressure P2, the engine inlet total temperature T2, the high pressure compressor outlet total pressure P3, the afterburner inlet total temperature T6.

Images