Rapid Response Control Design Method for the Starting Process of Aero-engines

Abstract

The present invention provides a rapid response control design method for the starting process of an aeroengine, belonging to the field of engine control. Using the N-dot control plan, an intelligent optimization algorithm is utilized to optimize the acceleration target curve, and then the controller is designed based on the active disturbance rejection control theory. Finally, the starting process time of the engine is minimized to achieve rapid response. The present invention can solve the problems that the time consistency and reliability of the transition state of an aeroengine are difficult to guarantee under traditional control methods. It is a control design method for the rapid response of the transition state of an aeroengine based on the N-dot control plan and active disturbance rejection control (ADRC), which can shorten the transition state adjustment time on the premise of ensuring the stable operation of the aeroengine.

Description

Technical Field

[0001] The present invention belongs to the field of engine control and relates to a control design method for realizing the rapid response of an aeroengine. Background Art

[0002] The starting process of an aeroengine is a very complex transient process, and the aeroengine system in this process has obvious non-linear and time-varying characteristics. Moreover, the starting performance directly affects the flight safety and use of the aircraft as well as the reliability and service life of the engine. Therefore, it is very important to conduct research on the starting process control.

[0003] Adopting the traditional starting control scheme, it is very difficult to ensure the time consistency and high reliability. When using a mechanical hydraulic control system for an aeroengine, since the regulator cannot be designed too complex, generally only an open-loop fuel supply regulation scheme can be selected, and starting over-temperature or hanging phenomena often occur, and the starting performance is restricted by various external factors. The currently commonly used acceleration control method is an open-loop control based on a fuel plan. This method is simple and easy to implement, but for different engines of the same batch or different life stages of the same engine, its acceleration performance is not consistent. In this regard, the literature has proposed to adopt a closed-loop control method based on the N-dot control plan; in the article "A Design Method of a N-dot Transient PI Control Law", scholars such as Wang Xi verified through simulation that adopting a closed-loop proportional integral (PI) control strategy of the rotor acceleration can ensure that the acceleration performance of the engine is not affected by component degradation. However, the traditional PI control strategy has a long response time and still has some limitations in practical applications, restricting the practicality of the model.

[0004] The particle swarm optimization (PSO) algorithm, as a bionic evolutionary algorithm, is proposed inspired by the behavior mechanism of biological groups in nature. The PSO algorithm is an evolutionary computing method proposed by American social psychologist J. Kennedy and electrical engineer R. Eberhart in 1995. The PSO algorithm originates from the research of artificial life, especially the imitation of the behavior mechanism of groups such as bird flocks and fish schools, and draws on the biological group model proposed by biologist F. Heppner. At the same time, the idea of evolutionary computing is also incorporated. The concept of the PSO algorithm is relatively simple, with few parameters to be adjusted, easy to be programmed and implemented, and it has no complex mathematical operations itself, and has low requirements for the speed and storage of computer hardware.

[0005] As a bionic algorithm, the PSO algorithm currently lacks a complete mathematical theoretical basis. However, as a new optimization algorithm, it has shown good application prospects in many fields. Therefore, in-depth research on the PSO algorithm is of great significance both theoretically and in practical applications. This invention patent utilizes the improved particle swarm algorithm, which has the advantages of simple implementation, good effect, and good versatility, to solve the problem of obtaining the time-optimal starting control under numerous constraints. First, the designed objective function, i.e., the starting process time, needs to be discretized. By ensuring the time-optimality of the starting process in each sub-time period, the optimal control plan for the entire starting process can be obtained. Since intelligent algorithms have better search capabilities in dealing with the performance optimization problems of aero-engines, the performance parameters of the engine converge faster. However, intelligent algorithms also have some disadvantages, such as being prone to falling into local optimal solutions and having high computational complexity. Therefore, it is necessary to analyze the applicability of the algorithm in combination with specific situations. Summary of the Invention

[0006] Aiming at the problems existing in the prior art, the present invention provides a control design method for realizing the fast response of the aero-engine transition state, which can meet the requirements of shortening the transition state adjustment time without surge and over-temperature during the starting process of the aero-engine.

[0007] In order to achieve the above object, the technical solution adopted by the present invention is as follows:

[0008] A control design method for the fast response control plan of the aero-engine transition state. First, an active disturbance rejection control (ADRC) controller is designed based on the N-dot control plan. Secondly, a competitive particle swarm intelligent control algorithm is constructed. Finally, it is applied to the model to obtain an ideal control result. The present invention can shorten the transition state adjustment time on the premise of ensuring the stable operation of the aero-engine. The specific steps are as follows:

[0009] Step 1: Design an ADRC controller based on the N-dot control plan

[0010] The characteristics of the N-dot control plan are as follows: The closed-loop control under the N-dot control plan can ensure the consistency of the acceleration performance of the aero-engine, respond to the transition state control scheme in a short time, and give full play to the potential of the engine. The starting process control is an important link in the aero-engine transition state control, and it is required to ensure that the engine does not surge, does not overheat, and shortens the starting time during this process. Based on the above research, the present invention adopts the active disturbance rejection control (ADRC) algorithm instead of the PI control algorithm and applies it to the starting process control of the aero-engine to construct a closed-loop control loop as Figure 1 shown. The specific steps are as follows:

[0011] (1) The open-loop nature of the open-loop fuel-air ratio control plan determines that it can only supply fuel to the aero-engine according to the original designed acceleration / deceleration plan, and cannot perform more flexible and detailed transient control based on the real-time state of the aero-engine. In addition, the open-loop fuel-air ratio control plan also has defects such as integral saturation and large performance impacts caused by component degradation. In this context, the more advantageous N-dot control plan has become one of the main methods for current transient control. The N-dot control plan ensures that engines with manufacturing tolerances and performance degradation achieve consistent transient performance by controlling the rotor acceleration of the engine.

[0012] (2) The active disturbance rejection controller (ADRC) mainly includes the following parts: the tracking differentiator (TD), the linear extended state observer (LESO), and the non-linear state error feedback (NLSEF). Its characteristics include: 1) Resolving the contradiction between the rapidity and overshoot of the control system by arranging the transient process. 2) Using the extended state observer to estimate the system disturbance in real time and perform disturbance compensation to enhance the robustness of the system. 3) Significantly improving the control function by adopting a linear state error combination method. The n-order structure diagram of the ADRC is as Figure 2 shown. The second-order tracking differential equation is as follows:

[0013]

[0014] In the formula, v0 is the reference setting value of the high-pressure shaft rotor acceleration N-dot, v1 is the tracking value of the reference input; v2 is the tracking value of the N-dot derivative; r is the speed factor, a physical quantity that determines the tracking speed; h is the sampling point of the system; h0 is the filtering factor; fhan(v1 - v0, v2, r, h0) is a non-linear function.

[0015] (3) Combining the N-dot control plan and the ADRC controller not only retains the characteristics of the N-dot control plan but also introduces the ADRC to enhance the strong anti-disturbance ability and strong robustness of the closed-loop control. Figure 3 is the ADRC closed-loop control structure designed based on the N-dot control plan in the aero-engine starting model.

[0016] Step 2: Construct the competitive particle swarm intelligent control algorithm

[0017] The described intelligent control algorithm is characterized by being able to solve the problem that traditional optimization algorithms have too many constraint conditions or multiple extreme values, making it impossible to solve the optimal solution.

[0018] (1) Considering that the traditional particle swarm optimization algorithm has a satisfactory optimization effect on small-scale particles, but its optimization effect on large-scale particles fails to meet the required optimization criteria. Inspired by the traditional particle swarm optimization algorithm, this invention patent utilizes a new competitive particle swarm optimization algorithm (CSO) for large-scale optimization. This algorithm is conceptually very different from the traditional particle swarm optimization algorithm. In the new competitive particle swarm optimization algorithm, when updating particles, neither the personal best position of each particle nor the global best position (or neighborhood best position) is involved. Instead, a pairwise competition mechanism is introduced. Under this mechanism, the particles that lose the competition will update their positions by learning from the winning particles.

[0019] (2) Without loss of generality, this invention considers the following minimization problem:

[0020] min f = f(X)

[0021] s.t. X ∈ χ

[0022] where χ ∈ R n is the feasible solution set, and n represents the dimension of the search space, that is, the number of decision variables.

[0023] To solve the above optimization problem, a population P(t) consisting of m particles is randomly initialized and iteratively updated, where m is called the population size and t is the generation index. Each particle has a two-dimensional position,

[0024] X i (t) = (x i,1 (t), x i,2 (t), …, x i,n (t)), representing the candidate solution of the above optimization problem, and an n-dimensional velocity vector, V i (t)(v i,1 (t), v i,2 (t), …, v i,n (t)). In each generation, the particles in P(t) are randomly assigned into m / 2 pairs (assuming the population size m is even), and then competition occurs between the two particles in each pair. As a result of each competition, the particle with better fitness (hereinafter referred to as the winner) will be directly passed to the next generation P(t + 1) of the population, while the particle that loses the competition (the loser) will update its position and velocity by learning from the winner. After learning from the winner, the loser will also be passed to the particle swarm P(t + 1). This means that each particle only participates in one competition. In other words, for a population size of m, m / 2 competitions occur, enabling all m particles to participate in one competition, and the positions and velocities of m / 2 particles will be updated.

[0025] Let X w,k (t), X l,k (t) and Vw,k (t), V l,k (t) is used to represent the positions and velocities of the winners and losers in the k-th round of the t-th generation, where k = 1, 2, …, m / 2. Therefore, the velocities of the particles that fail after the k-th competition will be updated using the following learning strategy:

[0026]

[0027] Meanwhile, the positions of the losers can be updated with the new velocities

[0028] X l,k (t + 1) = X l,k (t) + V l,k (t + 1)

[0029] where R1(k, t), R2(k, t), R3(k, t) ∈ [0, 1] n are three vectors randomly generated after the k-th competition and learning process in the t-th generation, is the average position value of the relevant particles, is the control parameter that affects

[0030] (3) Existing literature has proven the convergence of the search behavior of the proposed CSO and conducted an empirical analysis of its exploitation ability. The results show that the proposed CSO has a good balance in exploitation ability. Despite the simplicity of the algorithm, the existing experimental results show that on a wide range of large-scale optimization problems, the proposed CSO exhibits better overall performance than five state-of-the-art metaheuristic algorithms and can effectively solve problems with up to 5000 dimensions. Based on these advantages, this invention patent selects the CSO algorithm to solve the problem of optimizing the oil-gas ratio during the starting process. The flow of the competitive particle swarm optimization algorithm is as Figure 5 shown

[0031] Step 3: Combine the model to obtain the control result

[0032] (1) Optimization results of the CSO algorithm: Set the parameters in the optimization program in advance, and then set different prediction steps, iteration times, and the number of competitive particles respectively.

[0033] (1.1) Set the prediction step to 1 step, the iteration times to 60 times, and the number of competitive particles to 30. In this case, optimize the pre-set parameters, input the optimized oil-gas ratio curve into the starting model, and finally, the optimized high-pressure shaft ship speed rising curve can be obtained.

[0034] (1.2) The prediction steps are set to 1 or 2, the number of iterations is set to 60, and the number of competing particles is set to 30. This is used to illustrate the effect of the prediction steps on the optimization results. The results show that the more prediction steps there are, the farther away from the lower limit, the larger the surge margin, and the better the optimization effect.

[0035] (1.3) The number of prediction steps is set to 1, the number of iterations is set to 400, and the number of competitive particles is set to 200. This is used to illustrate the impact of the number of iterations and the number of particles on the optimization. The increase in the number of iterations and the number of particles will make the optimization results smoother and the effect better.

[0036] (2) Optimization results of the ADRC controller based on the N-dot control plan: The present invention uses the obtained high-pressure shaft speed curve as the control target value for closed-loop tracking control. As long as the optimized target curve can be tracked better, the limit will not be exceeded and the response time of the transition state will be shorter. In the comparative experiment, the N-dot-based ADRC algorithm can quickly control the error to zero, while the single PID algorithm and the N-dot-based PID algorithm need a period of time to stabilize to zero, especially at the steady-state point. The change is more drastic, which reflects the superior performance of the ADRC controller based on the N-dot control plan.

[0037] The beneficial effects of the present invention are:

[0038] Under the N-dot control plan, the consistency of the acceleration performance of the aircraft engine can be guaranteed, the control scheme of the transient state can be responded to in a short time, and the potential of the engine can be fully utilized. It can solve the problem that the traditional optimization algorithm has too many constraints or has multiple extreme values and cannot obtain the optimal solution.

[0039] Under the premise of safe and stable operation of the aero-engine, the present invention can shorten the starting time, thereby quickly reaching a specified rotation speed to achieve the thrust requirement of the engine, and further better adapt to the environment in which the engine is located. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 It is an ADRC closed-loop control loop based on the N-dot control plan;

[0041] Figure 2 It is the structure diagram of the active disturbance rejection controller;

[0042] Figure 3 It is an ADRC closed-loop control structure based on N-dot control plan;

[0043] Figure 4 This is the flow chart of particle swarm optimization algorithm;

[0044] Figure 5 This is the flow chart of competitive particle swarm optimization algorithm;

[0045] Figure 6 It is the internal structure of the controller;

[0046] Figure 7 It is a schematic diagram of the closed-loop logic structure;

[0047] Figure 8(a) is the comparison of the high-pressure shaft speed curve tracking performance; Figure 8(b) is the partial enlarged view at point A in Figure 8(a);

[0048] Figure 9 It is the comparison chart of the tracking error. Specific implementation mode

[0049] The present invention will be further described below through specific embodiments in conjunction with the accompanying drawings

[0050] A control design method for the rapid response control plan of an aero-engine during the transition state. First, design an ADRC controller based on the N-dot control plan; secondly, construct a competitive particle swarm intelligent control algorithm; finally, apply it to the model to obtain an ideal control result.

[0051] The first step: Design an ADRC controller based on the N-dot control plan:

[0052] The characteristics of the N-dot control plan are as follows: The closed-loop control under the N-dot control plan can ensure the consistency of the acceleration performance of the aero-engine, respond to the transition state control scheme in a short time, and give full play to the potential of the engine. The starting process control is an important link in the transition state control of the aero-engine, and it is required to ensure that the engine does not surge, does not overheat, and shortens the starting time during this process. Based on the above research, the present invention adopts the active disturbance rejection control (ADRC) algorithm to replace the PI control algorithm and applies it to the starting process control of the aero-engine, and constructs a closed-loop control loop as Figure 1 shown, and the specific steps are as follows:

[0053] (1) The open-loop attribute of the open-loop fuel-air ratio control plan determines that it can only supply fuel to the aero-engine according to the original designed acceleration / deceleration plan, and cannot perform more flexible and detailed transition state control according to the real-time state of the aero-engine. In addition, the open-loop fuel-air ratio control plan also has defects such as integral saturation and large influence of component degradation on performance. In such a background, the more advantageous N-dot control plan has become one of the main methods for transition state control at present. The N-dot control plan controls the rotor acceleration of the engine to ensure that the engines with manufacturing tolerances and performance degradation achieve the goal of consistent transition state performance.

[0054] (2) The Active Disturbance Rejection Controller (ADRC) mainly consists of the following parts: the Tracking Differentiator (TD), the Linear Extended State Observer (LESO), and the Nonlinear State Error Feedback (NLSEF). Its features include: 1) Resolving the contradiction between the rapidity and overshoot of the control system by arranging the transition process. 2) Using the Extended State Observer to estimate the system disturbance in real time and perform disturbance compensation to enhance the robustness of the system. 3) Adopting a linear state error combination method to significantly improve the control function. The n-order structure diagram of the ADRC is as shown in Figure 2 shown. The second-order tracking differential equation is as follows:

[0055]

[0056] where, v0 is the reference setting value of the high-pressure shaft rotor acceleration N-dot, v1 is the tracking value of the reference input; v2 is the tracking value of the derivative of N-dot; r is the speed factor, a physical quantity that determines the tracking speed; h is the sampling point of the system; h0 is the filtering factor; fhan(v1 - v0, v2, r, h0) is a nonlinear function.

[0057] (3) Combining the N-dot control plan and the ADRC controller not only retains the characteristics of the N-dot control plan but also introduces the ADRC to enhance the strong anti-disturbance ability and strong robustness of the closed-loop control. Figure 3 is the ADRC closed-loop control structure designed based on the N-dot control plan in the aero-engine starting model. In Figure 3 it can be seen that the bottom output line is connected to an interpolation module Turbofan, which switches at 5 s to simulate the starting motor driving. In Figure 3 the Controller is the controller that this invention focuses on introducing. It can integrate the ADRC controller and the PID controller based on the N-dot control plan for subsequent comparative experiments. First, look at the internal composition of the Controller. Its structure is as shown in Figure 6 shown. There is a MATLAB function inside the Controller, which divides the region of the high-pressure rotor speed difference. When the error between the output high-pressure rotor speed and the target value is greater than or less than a certain value, this invention makes the input of the controller a constant value, that is, the high-pressure shaft rotor acceleration is a constant value (the high-pressure shaft speed rises or falls at a certain slope), so as to ensure that the input of the controller is not too large to exceed certain limits during the engine operation; when the error is in the middle region, the error is processed by a multiple to make the controller respond quickly. Figure 6The speed of the medium and high pressure rotors obtains its acceleration value through the differential module, and is reduced by a multiple through an amplifier, where k = 1 / 100. Because the rotor acceleration is too sensitive, when the speed changes to a certain extent, the rotor acceleration may be very large. In order to ensure that ADRC can observe in time, the present invention performs a multiple scaling to facilitate control, and the input also performs the same operation. An open-loop oil-gas ratio of 0.00646369pps is superimposed on the output of the controller. Because the switch from interpolation to closed loop after 5s needs to be based on a certain oil-gas ratio, the open-loop oil-gas ratio replaces the oil-gas ratio value at the last moment before the switch. The deviation between the high-pressure shaft speed command and the feedback value is used as the input of the controller, and the output is the acceleration oil-gas ratio command. The fuel command is not directly transmitted to the engine but is transmitted to the engine after the optimized oil-gas ratio point is transmitted through the low-selection logic to realize the acceleration process control. The structural diagram is shown as follows Figure 7 shown.

[0058] Step 2: Constructing a competitive particle swarm intelligent control algorithm

[0059] The intelligent control algorithm is characterized in that it can solve the problem that the traditional optimization algorithm has too many constraints or has multiple extreme values and thus cannot find the optimal solution.

[0060] (1) Considering that the traditional particle swarm optimization algorithm has a good effect on the optimization of small-scale particles, it cannot meet the required optimization standards for the optimization of large-scale particles. Inspired by the traditional particle swarm optimization algorithm, the patent of this invention uses a new competitive particle swarm optimization algorithm (CSO) for large-scale optimization. This algorithm is conceptually very different from the traditional particle swarm optimization algorithm. In the new competitive particle swarm optimization algorithm, when updating particles, neither the personal best position of each particle nor the global best position (or the best position of the neighborhood) is involved. Instead, a pairwise competition mechanism is introduced, in which the particles that lose the competition will update their positions by learning from the winning particles.

[0061] (2) Without loss of generality, the present invention considers the following minimization problem:

[0062] minf=f(X)

[0063] stX∈χ

[0064] where χ∈R n is the set of feasible solutions, and n represents the dimension of the search space, that is, the number of decision variables.

[0065] To solve the above optimization problem, a population P(t) of m particles is randomly initialized and iteratively updated, where m is called the population size and t is the generation index. Each particle has a two-dimensional position,

[0066] X iP(t)=(x i,1 (t), x i,2 (t), …, x i,n (t)) represents the candidate solutions to the above optimization problem, as well as the n-dimensional velocity vector, V i (t)=(v i,1 (t), v i,2 (t), …, v i,n (t)). In each generation, the particles in P(t) are randomly assigned into m / 2 pairs (assuming the population size m is even), and then competition occurs between the two particles in each pair. As a result of each competition, the particle with better fitness (hereinafter referred to as the winner) will be directly passed on to the next generation P(t + 1) of the population, while the particle that loses the competition (the loser) will update its position and velocity by learning from the winner. After learning from the winner, the loser will also be passed on to the particle swarm P(t + 1). This means that each particle only participates in one competition. In other words, for a population size of m, m / 2 competitions occur, such that all m particles participate in one competition, and the positions and velocities of m / 2 particles will be updated.

[0067] Let X w,k (t) and X l,k (t) represent the positions of the winner and loser, and V w,k (t) and V l,k (t) represent the velocities of the winner and loser in the k-th round of competition in generation t, where k = 1, 2, …, m / 2. Therefore, the velocity of the particle that loses after the k-th competition will be updated using the following learning strategy:

[0068]

[0069] Meanwhile, the position of the loser can be updated with the new velocity

[0070] X l,k (t + 1)=X l,k (t)+V l,k (t + 1)

[0071] where R1(k, t), R2(k, t), R3(k, t) ∈ [0, 1] n are three vectors randomly generated after the k-th competition and learning process in generation t, is the average position value of the relevant particles, is the parameter that controls the influence.

[0072] The rotor acceleration schedule is the core of the acceleration controller, which is used to prevent compressor surge or excessive turbine inlet temperature caused by excessive rotor acceleration during the acceleration process. It is usually designed as a function of the corrected speed of the high-pressure rotor. Herein, the present invention optimizes the rotor acceleration schedule using the CSO algorithm, which can provide a basis for the design of subsequent control parameters. Consider the engine acceleration process: To increase the engine speed, the most direct approach is to increase the fuel-air ratio at a certain rate. The greater the fuel change rate, the faster the speed increases, and the shorter the acceleration process time. However, this may cause the engine to surge or overheat. During this acceleration process, the high-pressure rotor acceleration first increases and then decreases, that is, there is a maximum value for the rotor acceleration. Therefore, the design of the rotor acceleration schedule can be described as follows: Given the flight conditions, the process of starting the engine by the starter before the first-stage ignition is ignored because considering the role of the starter will make the optimization process too complex. For the time being, the process of starting the engine by the starter is not considered. Optimization starts from when the engine generates power. Under the conditions of ensuring the minimum high-pressure rotor surge margin limit and the maximum turbine inlet temperature limit, the CSO algorithm is used to make the high-pressure shaft speed reach the target value as soon as possible. Since the feedback value of the rotor acceleration cannot be directly measured, a differential processing method is adopted and smoothed by a filter.

[0073] (3) Existing literature has proven the convergence of the search behavior of the proposed CSO and conducted an empirical analysis of its exploitation ability. The results show that the proposed CSO has a good balance in exploitation ability. Although the algorithm is simple, the existing experimental results show that on a wide range of large-scale optimization problems, the proposed CSO exhibits better overall performance than five state-of-the-art metaheuristic algorithms and can effectively solve problems with up to 5000 dimensions. Based on these advantages, this invention patent selects the CSO algorithm to solve the problem of optimizing the fuel-air ratio during the starting process. The flow of the competitive particle swarm optimization algorithm is as Figure 5 shown.

[0074] The third step is to obtain the result by combining the model

[0075] (1) Optimization results of the CSO algorithm: Parameters are pre-set in the optimization program. The optimization duration is set to 7 s, the sampling time (simulation step) of the system is 0.1 s, the target value of the high-pressure shaft speed is 13230, the temperature limit is defined as 1100, the lower limit of the surge margin is set to 0.3, optimization starts from 17.3 s, the initial fuel-air ratio is set to 0.1036, the upper and lower bounds of the fuel-air ratio are 1.73 and 0.1010 respectively; the upper and lower bounds of the fuel-air ratio change amount are 0.05 and -0.05.

[0076] (1.1) Set the prediction step to 1 step, the number of iterations to 60 times, and the number of competing particles to 30. Optimize the preset parameters in this case. It is obtained that the oil-gas ratio does not exceed the upper bound of 1.73 and does not fall below the lower bound of 0.1010, meeting the optimization requirements. Input the optimized oil-gas ratio curve into the starting model, and finally, the optimized high-pressure shaft ship speed rising curve can be obtained.

[0077] (1.2) Set the prediction step to 1 step and 2 steps, the number of iterations to 60 times, and the number of competing particles to 30. To illustrate the influence of the prediction step on the optimization result, it is obtained that the more the prediction step, the farther away from the lower limit, the larger the surge margin, and the better the optimization effect.

[0078] (1.3) Set the prediction step to 1 step, the number of iterations to 400 times, and the number of competing particles to 200. To illustrate the influence of the number of iterations and the number of particles on the optimization, the obtained result is that the increase in the number of iterations and the number of particles will make the optimization result smoother and the effect better.

[0079] (2) Optimization results of the ADRC controller based on the N-dot control plan: The present invention uses the obtained high-pressure shaft speed curve as the control target value for closed-loop tracking control. As long as it can better track the optimized target curve, the limit will not exceed the limit and the response time of the transient state will be shorter. As shown in Figure 8, the tracking ability of PID is the worst, and the tracking ability of ADRC based on N-dot is the strongest. Through the enlarged part, the present invention finds that ADRC can almost perfectly track the target speed, while the effect of PID based on N-dot is relatively poor, but it is also much stronger than the control effect of a single PID. Finally, the present invention observes the performance of the three control algorithms through Figure 9 the change of the tracking error in. The present invention can find that the ADRC algorithm based on N-dot can quickly control the error to zero, while the single PID and PID based on N-dot need some time to stabilize to zero. Especially in the case of drastic changes at the steady-state point, the comparison effect is more obvious, highlighting the superior performance of the proposed control algorithm.

[0080] Matters not covered in the present invention are well-known technologies.

[0081] The above embodiments are only used to illustrate the technical concept and characteristics of the present invention. The purpose is to enable those who are familiar with this technology to understand the content of the present invention and implement it accordingly, and it cannot be used to limit the protection scope of the present invention. All equivalent changes or modifications made according to the spirit and essence of the present invention should be covered within the protection scope of the present invention.

Claims

1. A rapid response control design method for the starting process of an aero-engine, characterized in that This method can shorten the transition state regulation time on the premise of ensuring the stable operation of aeroengines. First, an active disturbance rejection control (ADRC) controller is designed based on the N-dot control plan. Second, a competitive swarm optimization (CSO) intelligent control algorithm is constructed. Finally, it is applied to the model to obtain the control results. It includes the following steps: Step 1: Design the ADRC controller based on the N-dot control plan The active disturbance rejection control (ADRC) algorithm is used to replace the traditional PI control algorithm and applied to the starting process control of aeroengines to construct a closed-loop control loop. The specific steps are as follows: (1) The N-dot control plan ensures the goal of consistent transition state performance for engines with manufacturing tolerances and performance degradation by controlling the rotor acceleration of the engine. (2) The active disturbance rejection controller ADRC mainly includes the following parts: the tracking differentiator TD, the linear extended state observer LESO, and the nonlinear state error feedback NLSEF. Its features are: 1) Solve the contradiction between the rapidity and overshoot of the control system by arranging the transition process; 2) Use the extended state observer to estimate the system disturbance in real time and perform disturbance compensation to enhance the robustness of the system; 3) Significantly improve the control function by adopting the linear state error combination method. The second-order tracking differential equation is as follows: , where is the reference set value of the high-pressure shaft rotor acceleration N-dot, is the tracking value of the reference input; is the tracking value of the N-dot derivative; r is the speed factor, a physical quantity that determines the tracking speed; is the sampling point of the system; is the filtering factor; is the non-linear function; (3) Combine the N-dot control plan and the ADRC controller, that is, retain the characteristics of the N-dot control plan and introduce ADRC to enhance the strong anti-disturbance ability and strong robustness of the closed-loop control. Step 2: Construct a competitive swarm optimization (CSO) intelligent control algorithm to solve the problem that traditional optimization algorithms have too many constraint conditions or multiple extreme values and thus cannot solve the optimal solution. The specific steps are as follows: (1) Adopt the new competitive swarm optimization (CSO) algorithm for large-scale optimization. When updating particles in this algorithm, neither the personal best position of each particle nor the global best position or neighborhood best position is involved. Instead, a pairwise competition mechanism is introduced. Under this mechanism, the losing particles will update their positions by learning from the winning particles. (2) The minimization problem is: , where is the feasible solution set, represents the dimension of the search space, that is, the number of decision variables; To solve the above optimization problem, a population P(t) containing particles is randomly initialized and iteratively updated, where is called the population size. Assume that the population size is an even number; is the generation exponent; Each particle has a two-dimensional position , representing a candidate solution to the above optimization problem, and a velocity vector in dimension ; in each generation, the particles in P(t) are randomly assigned to / 2 pairs, and then competition occurs between the two particles in each pair; as a result of each competition: the particle with better fitness, hereinafter referred to as the winner, will be directly passed to the next generation P(t + 1) of the population; while the particle that loses the competition, hereinafter referred to as the loser, will update its position and velocity by learning from the winner; after learning from the winner, the loser will also be passed to the particle swarm P(t + 1); that is, for the population size of / 2 competitions occur, such that all particles participate in one competition, and / 2 particles' positions and velocities will be updated; Respectively use 、 and 、 to represent the positions and velocities of the winners and losers in the nth round of the competition, where ; therefore, after the th competition, the velocities of the particles that failed will be updated using the following learning strategy: ， At the same time, the position of the loser is updated with the new velocity , where , , is the three vectors randomly generated after the th competition and learning process in the average position value of the relevant particles, is the parameter that controls the influence; (3) Select the CSO algorithm to solve the problem of optimizing the fuel-air ratio in the starting process. Step 3: Combine with the model to obtain the control results (1) Optimization results of the CSO algorithm: Set parameters in the optimization program in advance, and then set different prediction steps, iteration times, and the number of competitive particles respectively. Optimize the pre-set parameters, input the optimized fuel-air ratio curve into the starting model, and finally obtain the optimized high-pressure shaft ship speed rise curve. (2) Optimization results of the ADRC controller based on the N-dot control plan: Use the obtained high-pressure shaft speed curve as the control target value for closed-loop tracking control.

Images