A DLH-IHBA-based adaptive cycle engine control law optimization method

Abstract

The present application provides a method for optimizing the control law of an adaptive cycle engine based on DLH-IHBA, which belongs to the field of aviation engine control technology. The method includes establishing an adaptive cycle engine model and determining optimization variables based on the adaptive cycle engine model; determining constraints and an objective function based on the optimization variables for each control mode of the adaptive cycle engine, wherein the control modes include a steady-state control mode and a transitional state control mode, and the steady-state control modes include a minimum fuel consumption control mode, a maximum thrust control mode, and a minimum turbine pre-temperature control mode; based on the objective function and constraints, a multi-dimensional learning hunting strategy is used to integrate the honey badger optimization algorithm to optimize the control law of each control mode, obtain the optimal control variables, and use the tent mapping as the initial particle swarm position mapping. The present application scheme improves the convergence speed of engine control law optimization, improves the optimization effect and engine performance.

Description

Technical Field

[0001] The present application relates to the field of aero-engine control technology, and in particular to a method for optimizing control laws of an adaptive cycle engine based on DLH-IHBA. Background Art

[0002] Aircraft engines, as the core power units of aircraft, are of great significance in both civilian and military applications. Optimizing aircraft engine performance under various flight conditions and operating states is crucial to reducing fuel consumption and minimizing environmental impact. Engine control laws can optimize engine performance parameters (such as thrust and fuel consumption) by adjusting parameters such as the main fuel flow rate, afterburner fuel flow rate, and tail nozzle area. However, despite significant research achievements in engine control laws and performance optimization in recent years, performance degradation caused by factors such as wear and fouling during long-term operation remains a challenge. When performance declines, existing control laws are often unable to adapt, which in turn affects the engine's performance potential, reducing fuel efficiency and adaptability.

[0003] To address this challenge, a performance-optimal control (PSC) method has been proposed to optimize the engine operating point under real-world conditions. PSC optimizes constraints to find the optimal control variable, thereby achieving optimal performance while ensuring safe operation. In this context, the core of PSC lies in the selection of onboard adaptive models and optimization algorithms. Linear programming (LP) algorithms and intelligent optimization algorithms are commonly used as PSC algorithms. However, while LP algorithms can find optimal solutions in some cases, they are prone to getting stuck in local minima, while intelligent optimization algorithms often suffer from poor real-time performance, limiting the practical application of PSC. Furthermore, the practical application of PSC is limited by the fact that it requires multiple iterations of the onboard engine model to find the optimal control variable. This process is time-consuming and cannot meet the real-time requirements of aircraft engine control systems. In recent years, metaheuristic optimization algorithms have attracted attention due to their unique solution mechanisms. For example, particle swarm optimization, grasshopper optimization, and whale optimization have been widely used for optimizing complex systems. In particular, the Honey Badger Algorithm (HBA) represents a metaheuristic optimization algorithm inspired by the foraging behavior of honey badgers. The HBA has attracted attention for its clear algorithmic structure, simple implementation, and excellent stability. However, like other metaheuristic optimization algorithms, the HBA suffers from a lack of global exploration capabilities, slow convergence, low accuracy, and a tendency to fall into local optima. Therefore, finding a strategy to address the performance degradation of aircraft engines during long-term operation is of great significance to control law optimization technology. Summary of the Invention

[0004] In view of this, an embodiment of the present application provides an adaptive cycle engine control law optimization method based on DLH-IHBA, which at least partially solves the problems in the prior art that engines in long-term service cannot be adaptively adjusted after component degradation and traditional optimization algorithms have low optimization accuracy.

[0005] The present application provides a method for optimizing an adaptive cycle engine control law based on DLH-IHBA, the method comprising:

[0006] establishing an adaptive cycle engine model, and determining optimization variables based on the adaptive cycle engine model;

[0007] determining constraints and an objective function based on optimization variables for each adaptive cycle engine control mode, wherein the control mode includes a steady-state control mode and a transient control mode, wherein the steady-state control mode includes a minimum fuel consumption control mode, a maximum thrust control mode, and a minimum turbine inlet temperature control mode;

[0008] Based on the objective function and the constraints, a multidimensional learning hunting strategy is adopted to integrate the honey badger optimization algorithm to optimize the control law of each control mode to obtain the optimal control variables; wherein, the multidimensional learning hunting strategy is integrated with the honey badger optimization algorithm, including the tent mapping as the initial particle swarm position mapping.

[0009] According to a specific implementation of the embodiment of the present application, the optimization variables are:

[0010] ,

[0011] The constraints in the process of optimizing the control law of each control mode are described as follows:

[0012] ,

[0013] Among them, u is the optimization variable, n c is the compressor speed, π T is the turbine pressure drop ratio, W fb is the afterburner fuel flow, A 63 is the area of ​​the rear duct ejector, α f is the fan guide blade angle, α cdfs is the CDFS guide vane angle, α c is the compressor guide vane angle, A 13 Select valve area for mode, A 125 is the area of ​​the front duct ejector, A 225 is the inlet area of ​​the mixing zone before the outer duct, T4 is the total temperature before the high-pressure turbine, n f is the fan speed, SM f is the fan surge margin, SMcdfs is the CDFS surge margin, SM c is the compressor surge margin, P3 is the compressor outlet pressure, subscript min is the minimum value, and subscript max is the maximum value.

[0014] According to a specific implementation of the embodiment of the present application, the objective function and constraint conditions of the minimum fuel consumption control mode are expressed as follows:

[0015] ,

[0016] Among them, f(u) is the objective function, g i’ (u) is the inequality constraint, h k’ (u) is the equality constraint, minsfc is the minimum fuel consumption rate, st is the constraint, i' represents the i'th inequality constraint, and k' represents the k'th equality constraint;

[0017] The objective function and constraints of the maximum thrust control mode are expressed as follows:

[0018] ,

[0019] Wherein, max F is the maximum engine thrust;

[0020] The objective function and constraints of the minimum turbine pre-temperature control mode are expressed as follows:

[0021] ,

[0022] Among them, min T4 is the lowest total temperature before the high-pressure turbine.

[0023] According to a specific implementation of an embodiment of the present application, the objective function of the transition state control mode adopts an adaptive variable weight coefficient method to transform the dual objective function and multi-inequality constraint problem into an unconstrained optimization problem. The objective function expression of the transition state control mode is:

[0024] ,

[0025] Among them, minJ(k,σ) is the objective function of the transition state control mode, ω1 and ω2 are the weights of the objective function, n c (k) is the speed at the current moment k, T4(k) is the total temperature before the high-pressure turbine at the current moment k, n c,obj is the target speed value, T 4,obj is the target value of the total temperature before the high-pressure turbine, σ is an infinite positive number, and g(x) is the cost function.

[0026] According to a specific implementation of the embodiment of the present application, the multi-dimensional learning hunting strategy is integrated with the honey badger optimization algorithm to optimize the control law of each control mode, including:

[0027] Construct the initial population, maximum number of iterations, variables, spatial dimension of optimization variables and the boundary of the search space, initialize the position of the honey badger, and use the tent map as the initial particle swarm position map;

[0028] Defining the olfactory intensity of prey;

[0029] Calculate the density factor of the quasi-cosine law variation;

[0030] Based on the density factor and the olfactory intensity, updating the position of the honey badger in the digging phase and the position of the honey badger in the honey collecting phase;

[0031] Based on the multi-dimensional learning hunting strategy, the honey badger position is updated to improve the quality of the search individuals and increase the search ability, and the current best position and the current best fitness value are obtained;

[0032] A greedy strategy is used to interfere with the current best position to obtain the final global optimal position and the best fitness value.

[0033] According to a specific implementation of the embodiment of the present application, the mathematical expression of the Tent mapping is:

[0034] ,

[0035] Among them, μ is the chaos parameter, i is the population number, j is the chaotic variable number, and the chaotic sequence is randomly generated between (0,1). For the i-th population and the j-th chaotic variable, a chaotic sequence is randomly generated between (0,1). Randomly generate a chaotic sequence between (0,1) for the i-th population and the j+1-th chaotic variable;

[0036] The expression of the olfactory intensity is:

[0037] ,

[0038] Among them, I i is the olfactory intensity of the prey, r5 is a random number between 0 and 1, S is the source intensity, d i represents the distance between the prey and the honeypot, x i is the location of the honeypot, x prey is the current position of the prey, indicating the best position currently found in the search space;

[0039] The expression of the density factor of the quasi-cosine law change is:

[0040] ,

[0041] Among them, α is the density factor of the quasi-cosine law change, ω is the traditional density factor, t is the current iteration number, t max is the maximum number of iterations.

[0042] According to a specific implementation of an embodiment of the present application, the action expression of the honey badger in the digging stage is:

[0043] ,

[0044] Among them, x new is the updated position of the honey badger, β is the ability of the honey badger to obtain food, F is the flag for changing the search direction, r1, r2 and r3 are three different random numbers between 0 and 1;

[0045] The action expression of the honey badger in the honey collecting stage is:

[0046] ,

[0047] Among them, r4 is a random number.

[0048] According to a specific implementation of an embodiment of the present application, updating the honey badger position based on a multi-dimensional learning hunting strategy includes:

[0049] Construct a radius matrix based on the original position of the honey badger and the new position of the honey badger;

[0050] Construct a neighborhood matrix based on the radius matrix and the Euclidean distance between the current individual and the candidate individuals;

[0051] New individuals are generated by learning from multiple neighborhood matrices, where the dth dimension of each new individual is updated based on the dth dimension of a randomly selected individual, and the new individual is the position of the honey badger after being updated based on the multidimensional learning hunting strategy.

[0052] According to a specific implementation of the embodiment of the present application, the radius matrix expression is: ,

[0053] The neighborhood matrix expression is:

[0054] ,

[0055] The expression of the generated new individual is:

[0056] ,

[0057] Among them, Radiusi(t) is the radius matrix, Neighbouri(t) is the neighborhood matrix, xi (t) is the current individual, x j (t) is the candidate individual, x new (t+1) is the new position of individual i at the t+1th iteration, D is the dimension of the optimization variable space, N is the initial population, x i-DLH,j (t+1) is the new individual generated, x i,d (t) is the updated position of individual i in the dth dimension at the tth iteration, x n,d (t) is the individual i in the tth iteration based on the randomly selected individual in the dth dimension, x r,d (t) is the reference position of individual i in the dth dimension at the tth iteration, and rand is a random function that generates a random number between 0 and 1.

[0058] According to a specific implementation of the embodiment of the present application, the expression of the greedy strategy is:

[0059] ,

[0060] Among them, x i (t+1) is the updated position of individual i at the t+1th iteration, x i-new (t+1) is the position of individual i after the t+1th iteration without executing the multi-dimensional learning hunting strategy, f(x i-new ) is x i-new The corresponding objective function, x i-DLH (t+1) is the position of individual i after the t+1th iteration after the multi-dimensional learning hunting strategy is updated, f(x i-DLH ) is x i-DLH The corresponding objective function.

[0061] Beneficial effects:

[0062] The adaptive cycle engine control law optimization method based on DLH-IHBA in the embodiments of this application introduces tent mapping into the honey badger algorithm and combines it with a multidimensional learning hunting strategy to improve the algorithm's global convergence, thereby achieving the lowest fuel consumption rate and thus improving fuel efficiency; achieving maximum thrust when power demand is maximum to meet specific flight requirements; maintaining the turbine inlet temperature at a minimum level to ensure the turbine's service life; and ensuring that engine system parameters do not exceed limit values ​​during transient processes to achieve the shortest acceleration time. The optimization method of this application improves convergence speed and optimization effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0064] Figure 1 Flowchart of a method for optimizing a control law of an adaptive cycle engine based on DLH-IHBA according to an embodiment of the present invention;

[0065] Figure 2 A schematic diagram of a multi-objective optimization principle according to an embodiment of the present invention;

[0066] Figure 3 This is a flowchart of a multi-dimensional learning hunting strategy integrated with honey badger optimization according to an embodiment of the present invention. DETAILED DESCRIPTION

[0067] The embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0068] The following describes the embodiments of the present application through specific examples, and those skilled in the art can easily understand other advantages and effects of the present application from the contents disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. The present application can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present application. It should be noted that, in the absence of conflict, the features in the following embodiments and embodiments can be combined with each other. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without making creative work are within the scope of protection of this application.

[0069] It should be noted that various aspects of the embodiments within the scope of the appended claims are described below. It should be apparent that the aspects described herein can be embodied in a wide variety of forms, and any specific structure and / or function described herein is merely illustrative. Based on this application, it should be understood by those skilled in the art that an aspect described herein can be implemented independently of any other aspect, and two or more of these aspects can be combined in various ways. For example, any number of aspects described herein can be used to implement an apparatus and / or practice a method. In addition, other structures and / or functionalities other than one or more of the aspects described herein can be used to implement this apparatus and / or practice this method.

[0070] It should also be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present application. The illustrations only show components related to the present application and are not drawn according to the number, shape and size of components in actual implementation. In actual implementation, the type, quantity and proportion of each component can be changed at will, and the component layout type may also be more complicated.

[0071] Additionally, in the following description, specific details are provided to provide a thorough understanding of the examples. However, one skilled in the art will appreciate that the aspects described can be practiced without these specific details.

[0072] The embodiment of the present application provides a method for optimizing the control law of an adaptive cycle engine based on DLH-IHBA. Figures 1 to 3 Provide a detailed description.

[0073] In one embodiment, referring to Figure 1 , the method comprises the following steps:

[0074] establishing an adaptive cycle engine model (ACE onboard adaptive model), and determining optimization variables based on the adaptive cycle engine model;

[0075] Determining constraints and an objective function based on optimization variables for each adaptive cycle engine control mode, wherein the control modes include a steady-state control mode and a transient control mode. The steady-state control mode includes a minimum fuel consumption control mode, a maximum thrust control mode, and a minimum turbine inlet temperature control mode. The transient control mode includes an acceleration control mode and a deceleration control mode.

[0076] Based on the objective function and the constraints, a multidimensional learning hunting strategy fused with a honey badger optimization algorithm (DLH-IBHA) is used to optimize the control law of each control mode to obtain the optimal control variables for output; wherein, the multidimensional learning hunting strategy fused with a honey badger optimization algorithm includes a tent mapping as the initial particle swarm position mapping.

[0077] In this embodiment, a multidimensional learning harvesting strategy is integrated with the honey badger optimization algorithm (DLH-IBHA). The multidimensional learning harvesting strategy (DLH) is integrated with the honey badger algorithm (HBA). The honey badger algorithm is a heuristic optimization algorithm based on the behavior of natural organisms. It simulates the indomitable and random search characteristics of the honey badger when searching for food. This algorithm is a metaheuristic algorithm used to solve complex optimization problems.

[0078] The adaptive control law optimization method proposed in this embodiment does not directly optimize engine control parameters. Instead, it optimizes controlled parameters (such as speed and pressure ratio) and other adjustable parameters. Compared to directly optimizing control parameters, adjusting controlled parameters offers the following advantages: First, because engines are susceptible to external factors (such as flight altitude and wind gusts), direct optimization of control parameters requires real-time adjustment based on the environment and flight conditions. Existing optimization algorithms struggle to simultaneously meet both accuracy and real-time requirements. Second, because engine performance parameters cannot be directly measured, conventional control systems cannot directly control them. This embodiment proposes the DLH-IHBA optimization algorithm, which improves upon the Honey Badger optimization algorithm and employs a multidimensional learning and hunting strategy for optimizing steady-state and transient control laws for an adaptive cycle engine (ACE). This algorithm aims to enhance control law optimization and improve engine performance. Simulation results demonstrate that the DLH-IHBA converges faster and achieves better optimization results than several classic intelligent optimization algorithms. It effectively reduces the ACE's specific fuel consumption and minimum turbine inlet temperature, increases maximum thrust, and shortens acceleration time by 50%.

[0079] Furthermore, for the adaptive cycle engine model studied in this embodiment, the optimization variables are:

[0080] , (1)

[0081] The constraints in the process of optimizing the control law of each control mode are described as follows:

[0082] , (2)

[0083] Among them, u is the optimization variable, n c is the compressor speed, π T is the turbine pressure drop ratio, W fb is the afterburner fuel flow, A 63is the area of ​​the rear duct ejector, α f is the fan guide blade angle, α cdfs is the CDFS guide vane angle, α c is the compressor guide vane angle, A 13 Select valve area for mode, A 125 is the area of ​​the front duct ejector, A 225 is the inlet area of ​​the mixing zone before the outer duct, T4 is the total temperature before the high-pressure turbine, n f is the fan speed, SM f is the fan surge margin, SM cdfs is the CDFS surge margin, SM c is the compressor surge margin, P3 is the compressor outlet pressure, subscript min is the minimum value, and subscript max is the maximum value.

[0084] The expression of the above constraints can be expressed as:

[0085] , (3)

[0086] Among them, g i’ (u) is the inequality constraint, and i' represents the i'th inequality constraint.

[0087] In one embodiment, in the minimum fuel consumption control mode, the equality constraint emphasizes achieving the lowest fuel consumption rate while maintaining constant thrust. Comprehensive adjustments are required to improve the efficiency of the fan, compressor, and inlet, increase engine flow, and thus achieve thrust improvement. By solving nonlinear constraints, a suitable combination of control variables is found to achieve the desired engine thrust and operate at the lowest fuel consumption rate. The objective function and constraints of the minimum fuel consumption control mode are expressed as:

[0088] , (4)

[0089] Among them, f(u) is the objective function, g i’ (u) is the inequality constraint, h k’ (u) is the equality constraint, minsfc is the minimum fuel consumption rate, st is the constraint, i' represents the i'th inequality constraint, and k' represents the k'th equality constraint;

[0090] The maximum thrust control mode is primarily used during high-thrust demand phases, such as takeoff and accelerated climb. In this mode, the goal is to achieve maximum thrust while ensuring safe engine operation. At the engine operating point, the global maximum thrust is located at the intersection of the fan surge margin and the minimum safety margin of the compressor surge margin. The objective function and constraints of the maximum thrust control mode are expressed as:

[0091] , (5)

[0092] Wherein, max F is the maximum engine thrust;

[0093] The minimum turbine inlet temperature control mode is mainly used in high Mach number flight conditions. In this mode, the goal is to reduce the turbine inlet temperature as much as possible to extend the engine service life and reduce infrared radiation while ensuring that the thrust remains unchanged. This mode mainly reduces the engine turbine temperature by reducing the main fuel flow rate, and improves the engine pressure ratio, efficiency and flow rate by comprehensively adjusting parameters such as the tail nozzle nozzle area, the fan guide vane angle and the compressor guide vane angle, thereby increasing the thrust and ultimately keeping the thrust basically constant. The objective function and constraints of the minimum turbine inlet temperature control mode are expressed as:

[0094] , (6)

[0095] Among them, min T4 is the lowest total temperature before the high-pressure turbine.

[0096] In specific implementation, the optimization of the minimum fuel consumption control mode and the optimization of the minimum turbine inlet temperature mode are both multi-objective optimization. The goal of the minimum fuel consumption control mode optimization is to achieve the desired thrust value while minimizing the fuel consumption rate; the goal of the minimum turbine inlet temperature mode optimization is to achieve the desired thrust value while minimizing the turbine inlet temperature. The principle of multi-objective optimization is as follows Figure 2 As shown in the figure, through step-by-step iterative optimization, target 1 or target 2 is optimized. By selecting appropriate weights W=[W1,W2], where W1 is the weight of target 1 (lowest fuel consumption rate, lowest turbine inlet temperature), and W2 is the weight of target 2 (unchanged thrust), the comprehensive performance of target 1 and target 2 is optimized.

[0097] Combining formulas (4) to (6), the optimization of the adaptive control law of the intelligent engine can be expressed as the following nonlinear programming problem:

[0098] , (7)

[0099] In order to solve constrained optimization problems, a penalty function can be introduced.

[0100] , (8)

[0101] Here, F(u,σ) is the objective function that introduces a penalty function, and σ1 and σ2 are infinite positive numbers. This transforms the original constrained problem into an unconstrained one. During the intelligent optimization of engine adaptive control laws, an intelligent optimization algorithm is required to find the optimal combination of optimization parameters to achieve optimal engine performance.

[0102] In one embodiment, the objective function of the transition state control mode adopts an adaptive variable weight coefficient method to transform the dual objective function and multi-inequality constraint problem into an unconstrained optimization problem. The objective function expression of the transition state control mode is:

[0103] ,

[0104] Among them, minJ(k,σ) is the objective function of the transition state control mode, ω1 and ω2 are the weights of the objective function, n c (k) is the speed at the current moment k, T4(k) is the total temperature before the high-pressure turbine at the current moment k, n c,obj is the target speed value, T 4,obj is the target value of the total temperature before the high-pressure turbine, σ is an infinite positive number, and g(x) is the cost function.

[0105] In specific implementation, for the transient control mode, the key to the engine's acceleration performance is to achieve the shortest acceleration time, which can be expressed by the following expression:

[0106] , (9)

[0107] Among them, t ac is the acceleration time, I is the rotor moment of inertia, n is the rotor speed, n max is the maximum speed, n idle is the speed at slow speed, ΔH ax is the residual power of the turbine during the acceleration process. In order to achieve the best acceleration control in each control cycle, the high-pressure rotor speed can be used as the objective function:

[0108] , (10)

[0109] Where t' is the integration time.

[0110] The total temperature T4 before the high-pressure turbine can also effectively improve the turbine power. In order to fully reflect the energy changes during engine operation, the total temperature T4 before the high-pressure turbine and the high-pressure rotor speed are simultaneously used as optimization targets. The discrete objective function is defined using a linear weighted method:

[0111] , (11)

[0112] Among them, ω1 and ω2 are the weights of the objective function, n c (k) is the speed at the current moment k, T4(k) is the total temperature before the high-pressure turbine at the current moment k, n c,obj is the target speed value, T 4,obj is the target value of the total temperature before the high-pressure turbine.

[0113] By adopting the adaptive variable weight coefficient method, the dual objective function and multi-inequality constraint problem is transformed into an unconstrained optimization problem:

[0114] , (12)

[0115] This approach helps achieve maximum turbine surplus power, reduces acceleration time, and ensures high specific thrust.

[0116] In one embodiment, referring to Figure 3 The multi-dimensional learning hunting strategy is integrated with the honey badger optimization algorithm to optimize the control law of each control mode, including:

[0117] Construct the initial population, maximum number of iterations, variables, spatial dimension of optimization variables and the boundary of the search space, initialize the position of the honey badger, and use the tent map as the initial particle swarm position map;

[0118] Defining the olfactory intensity of prey;

[0119] Calculate the density factor of the quasi-cosine law variation;

[0120] Based on the density factor and the olfactory intensity, updating the position of the honey badger in the digging phase and the position of the honey badger in the honey collecting phase;

[0121] Based on the multi-dimensional learning hunting strategy, the honey badger position is updated to improve the quality of the search individuals and increase the search ability, and the current best position and the current best fitness value are obtained;

[0122] A greedy strategy is used to interfere with the current best position to obtain the final global optimal position and the best fitness value.

[0123] According to a specific implementation of the embodiment of the present application, the mathematical expression of the Tent mapping is:

[0124] ,

[0125] Among them, μ is the chaos parameter, i is the population number, j is the chaotic variable number, and the chaotic sequence is randomly generated between (0,1). For the i-th population and the j-th chaotic variable, a chaotic sequence is randomly generated between (0,1). Randomly generate a chaotic sequence between (0,1) for the i-th population and the j+1-th chaotic variable;

[0126] The expression of the olfactory intensity is:

[0127] ,

[0128] Among them, I iis the olfactory intensity of the prey, r5 is a random number between 0 and 1, S is the source intensity, d i represents the distance between the prey and the honeypot, x i is the location of the honeypot, x prey is the current position of the prey, indicating the best position currently found in the search space;

[0129] The expression of the density factor of the quasi-cosine law change is:

[0130] ,

[0131] Among them, α is the density factor of the quasi-cosine law change, ω is the traditional density factor, t is the current iteration number, t max is the maximum number of iterations.

[0132] According to a specific implementation of an embodiment of the present application, the action expression of the honey badger in the digging stage is:

[0133] ,

[0134] Among them, x new is the updated position of the honey badger, β is the ability of the honey badger to obtain food, F is the flag for changing the search direction, r1, r2 and r3 are three different random numbers between 0 and 1;

[0135] The action expression of the honey badger in the honey collecting stage is:

[0136] ,

[0137] Among them, r4 is a random number.

[0138] According to a specific implementation of an embodiment of the present application, updating the honey badger position based on a multi-dimensional learning hunting strategy includes:

[0139] Construct a radius matrix based on the original position of the honey badger and the new position of the honey badger;

[0140] Construct a neighborhood matrix based on the radius matrix and the Euclidean distance between the current individual and the candidate individuals;

[0141] New individuals are generated by learning from multiple neighborhood matrices, where the dth dimension of each new individual is updated based on the dth dimension of a randomly selected individual, and the new individual is the position of the honey badger after being updated based on the multidimensional learning hunting strategy.

[0142] According to a specific implementation of the embodiment of the present application, the radius matrix expression is: ,

[0143] The neighborhood matrix expression is:

[0144] ,

[0145] The expression of the generated new individual is:

[0146] ,

[0147] Among them, Radiusi(t) is the radius matrix, Neighbouri(t) is the neighborhood matrix, x i (t) is the current individual, x j (t) is the candidate individual, x new (t+1) is the new position of individual i at the t+1th iteration, D is the dimension of the optimization variable space, N is the initial population, x i-DLH,j (t+1) is the new individual generated, x i,d (t) is the updated position of individual i in the dth dimension at the tth iteration, x n,d (t) is the individual i in the tth iteration based on the randomly selected individual in the dth dimension, x r,d (t) is the reference position of individual i in the dth dimension at the tth iteration, and rand is a random function that generates a random number between 0 and 1.

[0148] According to a specific implementation of the embodiment of the present application, the expression of the greedy strategy is:

[0149] ,

[0150] Among them, x i (t+1) is the updated position of individual i at the t+1th iteration, x i-new (t+1) is the position of individual i after the t+1th iteration without executing the multi-dimensional learning hunting strategy, f(x i-new ) is x i-new The corresponding objective function, x i-DLH (t+1) is the position of individual i after the t+1th iteration after the multi-dimensional learning hunting strategy is updated, f(x i-DLH ) is x i-DLH The corresponding objective function.

[0151] In specific implementation, the flowchart of multi-dimensional learning hunting strategy integrated with honey badger optimization algorithm (DLH-IHBA) DLH-IHBA is as follows Figure 3 Through this method, DLH-IHBA overcomes the defects of the original HBA, further improves the optimization speed and solution accuracy, and provides an effective strategy for solving complex problems.

[0152] Step 1: Initialization

[0153] Construct the initial population N and the maximum number of iterations t max , variables, spatial dimensions, and upper and lower search range boundaries, and initialize the position of the honey badger. In order to improve the speed and accuracy of the optimization algorithm, it is necessary to improve the uniformity of the initial population in the search space. This application uses Tent mapping as the initial particle swarm position mapping. Compared with the random generation method, the chaotic sequence has better randomness, ergodicity and regularity. The uniform distribution characteristics of Tent mapping help to improve the optimization speed and solution accuracy of the algorithm. The mathematical expression of Tent mapping is shown as follows:

[0154] , (13)

[0155] Among them, the chaos parameter μ∈(0,2] is proportional to the chaos, i and j are the population number and the chaotic variable number respectively. To randomly generate chaotic sequences between (0,1).

[0156] Since the Tent mapping is highly sensitive to the selection of initial values, this embodiment uses a variety of different initial values ​​to generate corresponding chaotic sequences and converts them into the search space of each individual. In the DLH-IHBA algorithm, the position of the i-th honey badger is expressed as:

[0157] , (14)

[0158] Among them, lb and ub are search spaces respectively. lower and upper limits.

[0159] The position of the i-th honey badger in the D-dimensional space is expressed as:

[0160] , (15)

[0161] Where D is the spatial dimension of the optimization variable.

[0162] Step 2: Definition of the olfactory intensity of the prey

[0163] The olfactory intensity is related to the distance between the prey and the i-th honey badger, I i Indicates the intensity of the prey's sense of smell. High intensity indicates fast movement. i It is defined by the following formula:

[0164] , (16)

[0165] Among them, I i is the olfactory intensity of the prey, r5 is a random number between 0 and 1, S is the source intensity, d i represents the distance between the prey and the honeypot, x i is the location of the honeypot, x preyis the current position of the prey, indicating the best position currently found in the search space.

[0166] Step 3: Update Location

[0167] The core idea of ​​the DLH-IHBA algorithm is to continuously update the prey location, search among a limited number of N honey badgers, and update the current location x of the prey. prey , which is the solution to the problem. This method divides the honey badger's foraging behavior into a digging phase and a nectar-gathering phase. The digging phase refers to the honey badger's behavior of digging holes in search of food, while the nectar-gathering phase refers to the honey badger's behavior of following a guide hummingbird to find a hive. During the digging phase, the honey badger's movements resemble a heart shape, which can be expressed by the following formula:

[0168] , (17)

[0169] Among them, x new is the updated position of the honey badger, β is the ability of the honey badger to obtain food (β≥1, β=6 is usually set), r1, r2 and r3 are three different random numbers between 0 and 1, and F is the flag for changing the search direction. F is determined by the following formula:

[0170] , (18)

[0171] α is the density factor. In the traditional honey badger optimization algorithm, the density factor α is tr Determined by the following formula:

[0172] , (19)

[0173] Where C is a constant equal to 2, t max is the maximum number of iterations.

[0174] The traditional density factor changes very gently, which reduces the convergence speed of the algorithm. Therefore, the present invention proposes a density factor that changes according to the quasi-cosine law, and its mathematical model is as follows:

[0175] , (20)

[0176] Among them, α is the density factor of the quasi-cosine law change, ω is the traditional density factor, t is the current iteration number, t max is the maximum number of iterations.

[0177] The density factor α of the improved quasi-cosine law decreases rapidly in the early iterations, ensuring that the population searches a wide range of spaces and accelerating the algorithm's convergence. In the middle stages of the iteration, when α drops to a smaller value, a new incremental step size is assigned, allowing the population to find the global optimal solution again over a wide range, helping the algorithm escape the local optimum. In the later stages of the iteration, the value of α decreases again significantly, allowing the population to quickly approach the global optimum. When the value of α is very small, the population is able to conduct a deep local search.

[0178] During the honey collection phase, the honey badger follows the guide hummingbird to find the hive, which can be simulated as follows:

[0179] .(twenty one)

[0180] Step 4: DLH strategy update location

[0181] Traditional HBA has low convergence and eventually leads to the dilemma of local optimality. Therefore, DLH is introduced to improve the quality of search individuals and increase the search capability.

[0182] DLH generates the following replacement individuals:

[0183] ,(twenty two)

[0184] The radius matrix Radius is generated by subtracting the distance between the original and new positions, and the neighborhood matrix is ​​based on the current individual x i (t) and alternative individual x j (t) is constructed by the Euclidean distance between them.

[0185] ,(twenty three)

[0186] By learning from many neighborhood matrices, DLH generates new individuals x i-DLH,j (t+1), where the dth dimension of each new individual is based on a randomly selected individual x n,d (t) is updated in the dth dimension.

[0187] ,(twenty four)

[0188] Recalculate the individual fitness value and retain the current best position and the current best fitness value.

[0189] Finally, a greedy strategy is used to interfere with the current optimal position, evaluate the fitness value of the current optimal individual, and retain individuals that are more valuable for population position updates. Its mathematical model is as follows:

[0190] . (25)

[0191] When the current number of iterations does not satisfy t <t max , and finally returns the best fitness value and the global optimal position.

[0192] The embodiments of this application address the issues of long-serving engines failing to adapt after component degradation and the low optimization accuracy of traditional optimization algorithms. They propose an intelligent optimization method for adaptive cycle engine (ACE) control laws based on an improved Honey Badger algorithm. This method introduces Tent mapping and the quasi-cosine law, combined with a multidimensional learning hunting strategy, to enhance the algorithm's global convergence. The method aims to achieve the lowest fuel consumption rate, thereby improving fuel efficiency; achieve maximum thrust when power demand is highest to meet specific flight requirements; maintain the turbine inlet temperature at a minimum level to ensure turbine service life; and ensure that engine system parameters do not exceed their limit values ​​during transient processes, achieving the shortest acceleration time. The specific contributions of this invention are as follows:

[0193] (1) A multi-dimensional learning hunting strategy fused with honey badger algorithm (DLH-IHBA) is proposed. Tent mapping is used as the initial particle swarm position mapping to improve the optimization speed and solution accuracy of the algorithm. The density factor of the algorithm is improved by using the quasi-cosine law, so that it decreases rapidly in the early iterations, accelerating the convergence speed. A multi-dimensional learning hunting strategy is introduced to improve the individual quality, enhance the search process, and balance the exploration and exploitation phases.

[0194] (2) The DLH-IHBA algorithm and a variety of classic intelligent optimization algorithms were used to conduct numerical simulation verification of the intelligent optimization of the ACE control law to verify the feasibility of the proposed method. As optimization variables, the steady-state control mode and the transition-state control mode are optimized separately.

[0195] The following describes the adaptive cycle engine control law optimization method based on DLH-IHBA of the present application with a specific embodiment.

[0196] The DLH-IHBA algorithm and a variety of classic intelligent optimization algorithms are used to perform numerical simulations to verify the feasibility of the proposed method. , respectively optimize the steady-state control mode and the transition state control mode, where n c is the compressor speed, π T is the turbine pressure ratio, A8 is the main duct nozzle outlet area, A 38 is the Flade duct nozzle area, α fl is the Flade guide vane angle, α f is the fan guide blade angle, α cdfs is the CDFS guide vane angle, α c is the compressor guide vane angle.

[0197] (1) Minimum fuel consumption control mode

[0198] At low Mach numbers, the fuel consumption rate of turbojet engines is higher than that of turbofan engines, and turbofan engines are not suitable for high Mach number flight conditions due to their large bypass ratio. r =78000N, H is the altitude, Ma is the Mach number) as an example to verify the minimum fuel consumption control method of the engine, and the three-outer-turbofan mode of ACE is used for verification.

[0199] The optimization objective function is set as: , the minimum fuel consumption is reduced by 1.26%, F r is the reference thrust, F n The actual thrust.

[0200] (2) Maximum thrust control mode

[0201] In this mode, the supersonic cruise point (H=10km,Ma=1.5,n c =100%) as an example, the maximum thrust control mode of ACE is simulated. The optimization objective function is set to fitness=10000 / F n The maximum thrust is increased by 0.35%.

[0202] (3) Minimum turbine pre-temperature control mode

[0203] ACE's minimum turbine inlet temperature control is usually used in high altitude and high Mach flight. The goal of the minimum turbine inlet temperature control is to ensure that the airflow temperature in front of the turbine does not exceed the maximum allowable temperature of the turbine material, thereby protecting the turbine. r =80000N) as an example, the method proposed in this invention is used to optimize the minimum turbine pre-temperature control mode. The optimization objective function is set to fitness = (F r -F n ) 2 +T4. The minimum turbine inlet temperature is reduced by 0.23%.

[0204] (4) Transition state control mode

[0205] The acceleration mode is a typical transitional control mode. This mode simulates and verifies the acceleration process of the ACE's three-peripheral working mode at an altitude of 18 km and an altitude of 0.8 Ma, with the fan percentage speed accelerating from 82% to 98%.

[0206] The present invention provides a novel adaptive cycle engine control law optimization method based on the DLH-IHBA. This method improves upon the Honey Badger optimization algorithm and employs a multidimensional learning and hunting strategy for optimizing the steady-state and transitional control laws of an adaptive cyclic engine (ACE). The goal is to enhance control law optimization and improve engine performance. Simulation results demonstrate that the DLH-IHBA algorithm converges faster and achieves better optimization results than several traditional intelligent optimization algorithms. It effectively reduces the ACE's fuel consumption and minimum turbine inlet temperature, increases maximum thrust, and shortens acceleration time by 50%.

[0207] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. A method for optimizing control law of an adaptive cycle engine based on DLH-IHBA, characterized in that: The method comprises: establishing an adaptive cycle engine model, and determining optimization variables based on the adaptive cycle engine model; determining constraints and an objective function based on optimization variables for each adaptive cycle engine control mode, wherein the control mode includes a steady-state control mode and a transient control mode, wherein the steady-state control mode includes a minimum fuel consumption control mode, a maximum thrust control mode, and a minimum turbine inlet temperature control mode; Based on the objective function and the constraints, a multi-dimensional learning hunting strategy fused with a honey badger optimization algorithm is used to optimize the control law of each control mode to obtain the optimal control variable; wherein, the multi-dimensional learning hunting strategy fused with the honey badger optimization algorithm includes using the tent map as the initial particle swarm position map; The multi-dimensional learning hunting strategy is integrated with the honey badger optimization algorithm to optimize the control law of each control mode, including: Construct the initial population, maximum number of iterations, variables, spatial dimension of optimization variables and the boundary of the search space, initialize the position of the honey badger, and use the tent map as the initial particle swarm position map; Defining the olfactory intensity of prey; Calculate the density factor of the quasi-cosine law variation; Based on the density factor and the olfactory intensity, updating the position of the honey badger in the digging phase and the position of the honey badger in the honey collecting phase; Based on the multi-dimensional learning hunting strategy, the honey badger position is updated to improve the quality of the search individuals and increase the search ability, and the current best position and the current best fitness value are obtained; Use a greedy strategy to interfere with the current best position to obtain the final global optimal position and the best fitness value; The method of updating the honey badger position based on the multi-dimensional learning hunting strategy includes: Construct a radius matrix based on the original position of the honey badger and the new position of the honey badger; Construct a neighborhood matrix based on the radius matrix and the Euclidean distance between the current individual and the candidate individuals; New individuals are generated by learning from multiple neighborhood matrices, where the dth dimension of each new individual is updated based on the dth dimension of a randomly selected individual, and the new individual is the position of the honey badger after being updated based on the multidimensional learning hunting strategy.

2. The adaptive cycle engine control law optimization method based on DLH-IHBA according to claim 1 is characterized in that: The optimization variables are: ， The constraints in the process of optimizing the control law of each control mode are described as follows: ， Among them, u is the optimization variable, n c is the compressor speed, π T is the turbine pressure drop ratio, W fb is the afterburner fuel flow, A 63 is the area of ​​the rear duct ejector, α f is the fan guide blade angle, α cdfs is the CDFS guide vane angle, α c is the compressor guide vane angle, A 13 Select valve area for mode, A 125 is the area of ​​the front duct ejector, A 225 is the inlet area of ​​the mixing zone before the outer duct, T4 is the total temperature before the high-pressure turbine, n f is the fan speed, SM f is the fan surge margin, SM cdfs is the CDFS surge margin, SM c is the compressor surge margin, P3 is the compressor outlet pressure, subscript min is the minimum value, and subscript max is the maximum value.

3. The adaptive cycle engine control law optimization method based on DLH-IHBA according to claim 2 is characterized in that: The objective function and constraints of the minimum fuel consumption control mode are expressed as follows: ， Among them, f(u) is the objective function, g i’ (u) is the inequality constraint, h k’ (u) is the equality constraint, min sfc is the minimum fuel consumption rate, st is the constraint, i' represents the i'th inequality constraint, k' represents the k'th equality constraint, I' represents the total number of inequality constraints, and K' represents the total number of equality constraints; The objective function and constraints of the maximum thrust control mode are expressed as follows: ， Wherein, max F is the maximum engine thrust; The objective function and constraints of the minimum turbine pre-temperature control mode are expressed as follows: ， Among them, min T4 is the lowest total temperature before the high-pressure turbine.

4. The adaptive cycle engine control law optimization method based on DLH-IHBA according to claim 3 is characterized in that: The objective function of the transition state control mode adopts the adaptive variable weight coefficient method to transform the dual objective function and multi-inequality constraint problem into an unconstrained optimization problem. The objective function expression of the transition state control mode is: ， Among them, minJ(k,σ) is the objective function of the transition state control mode, ω1 and ω2 are the weights of the objective function, n c (k) is the compressor speed at the current time k, T4(k) is the total temperature before the high-pressure turbine at the current time k, n c,obj is the target value of compressor speed, T 4,obj is the target value of the total temperature before the high-pressure turbine, σ is an infinite positive number, and g(x) is the cost function.

5. The adaptive cycle engine control law optimization method based on DLH-IHBA according to claim 1, characterized in that: The mathematical expression of the Tent mapping is: ， Among them, μ is the chaos parameter, i is the population number, j is the chaotic variable number, and the chaotic sequence is randomly generated between (0,1). For the i-th population and the j-th chaotic variable, a chaotic sequence is randomly generated between (0,1). Randomly generate a chaotic sequence between (0,1) for the i-th population and the j+1-th chaotic variable; The expression of the olfactory intensity is: ， Among them, I i is the olfactory intensity of the prey, r5 is a random number between 0 and 1, S is the source intensity, d i represents the distance between the prey and the honeypot, x i is the location of the honeypot, x prey is the current position of the prey, indicating the best position currently found in the search space; The expression of the density factor of the quasi-cosine law change is: ， Among them, α is the density factor of the quasi-cosine law change, ω is the traditional density factor, t is the current iteration number, t max is the maximum number of iterations.

6. The adaptive cycle engine control law optimization method based on DLH-IHBA according to claim 5, characterized in that: The action expression of the honey badger in the digging stage is: ， Among them, x new is the updated position of the honey badger, β is the ability of the honey badger to obtain food, F is the flag for changing the search direction, r1, r2 and r3 are three different random numbers between 0 and 1; The action expression of the honey badger in the honey collecting stage is: ， Among them, r4 is a random number.

7. The adaptive cycle engine control law optimization method based on DLH-IHBA according to claim 6, characterized in that: The radius matrix expression is: ， The neighborhood matrix expression is: ， The expression of the generated new individual is: ， Among them, Radiusi(t) is the radius matrix, Neighbouri(t) is the neighborhood matrix, x i (t) is the current individual, x j (t) is the candidate individual, x new (t+1) is the new position of individual i at the t+1th iteration, D is the dimension of the optimization variable space, N is the initial population, x i-DLH,j (t+1) is the new individual generated, x i,d (t) is the updated position of individual i in the dth dimension at the tth iteration, x n,d (t) is the individual i in the tth iteration based on the randomly selected individual in the dth dimension, x r,d (t) is the reference position of individual i in the dth dimension at the tth iteration, and rand is a random function that generates a random number between 0 and 1.

8. The adaptive cycle engine control law optimization method based on DLH-IHBA according to claim 7, characterized in that: The expression of the greedy strategy is: ， Among them, x i (t+1) is the updated position of individual i at the t+1th iteration, x i-new (t+1) is the position of individual i after the t+1th iteration without executing the multi-dimensional learning hunting strategy, f(x i-new ) is x i-new The corresponding objective function, x i-DLH (t+1) is the position of individual i after the t+1th iteration after the multi-dimensional learning hunting strategy is updated, f(x i-DLH ) is x i-DLH The corresponding objective function.

Images