Two-target flight test mission optimization planning method based on density algorithm

Abstract

The application discloses a two-target flight test task optimization planning method based on a DE algorithm, sets a population size, a maximum iteration number, a crossover probability and a mutation probability, randomly generates an initialization population with a size of n according to a coding rule, repairs a task execution vector coding of each individual in the population through a task execution logic repair mechanism, decodes the coding of the individual in the population and calculates fitness values of two targets in the population by using an idle time insertion technology, and then performs local search on 50% dominant individuals to obtain a population, generates new offspring individuals through lexicographic optimization selection for the population, performs a crossover operation on the population to generate offspring, and generates offspring through a mutation operation on the population; and the flight test cycle and task transfer cost are sequentially optimized under the constraint conditions of the flight test task and the test aircraft.

Description

Technical Field

[0001] This invention relates to a technology in the field of mission planning, specifically a two-objective flight test mission optimization planning method based on the Mimu algorithm. Background Technology

[0002] Developing an aircraft involves a series of processes, including project approval, preliminary design, detailed design, integration testing, prototype manufacturing, flight testing, and mass production. Each stage is essential. Among these, design and flight testing capabilities are core competencies of aircraft manufacturers. Currently, flight testing faces challenges such as high complexity, organizational difficulty, long testing cycles, and high safety risks. Flight testing has become a significant factor restricting model progress and development costs. The flight test plan directly determines the testing time and cost. A reasonable flight test mission plan can effectively allocate flight test resources, reduce testing risks, meet time requirements, and improve testing efficiency. Flight test mission planning is based on the model's mission requirements and characteristics, scientifically planning flight test missions to develop a reasonable and efficient flight test plan. The current method of manually planning flight test missions based on work experience conflicts with ensuring the development progress of large civil aircraft and saving development costs.

[0003] Flight test mission planning is an NP-hard problem, and there are two main types of solution methods for it. The first type is the exact method, which can find the optimal solution to a structured combinatorial optimization problem, such as branch and bound or mathematical programming. As a complete set of optimization algorithms, exact methods can guarantee finding the optimal solution for every finite-sized instance of the combinatorial optimization problem within a finite time. However, exact methods may require exponentially increased computation time when dealing with large-scale problems, resulting in low computational efficiency. Therefore, approximate methods that can obtain a solution to the problem in a short time can be used to solve the flight test mission planning problem. Among these, metaheuristic algorithms perform well in solving complex combinatorial optimization problems. However, they are prone to getting trapped in local optima, leading to premature convergence. Furthermore, in actual flight test mission planning, it is necessary to consider not only minimizing the flight test cycle but also optimizing multiple objectives. Summary of the Invention

[0004] This invention addresses the shortcomings of existing coding methods, such as their inability to solve the problems of aircraft arrangement and sequencing for test flight missions, their inability to handle logical constraints between test flight missions, their lack of local search operators for the target characteristics of test flight mission planning problems, and their inability to handle multiple targets. It proposes a two-objective test flight mission optimization planning method based on the Mimu algorithm, which optimizes the test flight cycle and mission transfer costs in an orderly manner while satisfying the constraints of the test flight mission and the test aircraft.

[0005] This invention is achieved through the following technical solution:

[0006] This invention relates to a two-objective flight test mission optimization planning method based on the MIC algorithm, comprising:

[0007] Step 1: Set the population size popsize, maximum number of iterations max_iter, crossover probability pc, and mutation probability pm;

[0008] Step 2: Set the iteration count G = 1, and randomly generate an initial population P of size popsize according to the encoding rules. G ;

[0009] Step 3: Repair the task execution vector encoding for each individual in the population using a task execution logic repair mechanism;

[0010] Step 4: Use idle time insertion technique to decode the encoding of individuals in the population and calculate the fitness values ​​of two targets in the population;

[0011] Step 5: Select 50% of the dominant individuals to perform a local search to obtain population T. G ;

[0012] Step 6: For population T G New offspring individuals N are generated through lexicographical optimization. G ;

[0013] Step 7: For population N G Perform a crossover operation to generate descendant C G ;

[0014] Step 8: For population C G The offspring M are generated through mutation operations. G ;

[0015] Step 9: If the number of iterations G has not reached the maximum number of iterations max_iter, then G = G + 1, and go to step 3; otherwise, output the result.

[0016] This invention relates to a two-objective flight test mission optimization planning system for implementing the above-mentioned method, comprising: a flight test database module, a flight test mission optimization algorithm module, a system support module, and a comprehensive display module. The flight test database module contains a set of flight test mission entities and a set of test aircraft entities. The flight test mission optimization algorithm module, based on the data provided by the flight test database module, calls the flight test mission planning algorithm to assign each flight test mission to a suitable test aircraft, sorts the assigned missions on the corresponding test aircraft, and outputs a Gantt chart of the flight test mission prototype network allocation. The system support module verifies whether the user has system access rights, inputs parameters selected by the user into some algorithms, and appropriately processes the input and output data of each module within the system to ensure they meet the required data types. The comprehensive display module provides a human-computer interaction interface for user login, algorithm parameter settings, and comprehensive result display functions.

[0017] The flight test mission entity set includes mission number, mission name, mission execution duration, set of preceding missions, mission preparation time, and set of compatible test aircraft; the test aircraft entity set includes test flight prototype number, test flight prototype deployment time, and test aircraft executable missions. Technical effect

[0018] Compared with existing technologies, this invention adopts a task repair mechanism to ensure that the logical relationship between test flight tasks is always satisfied during the population iteration process; by using a lexicographical optimization method, the test flight cycle and task transfer costs can be effectively optimized, and the optimization effect of the test flight cycle is improved by using idle time insertion technology for decoding. Attached Figure Description

[0019] Figure 1 This is a flowchart of the present invention;

[0020] Figure 2 This is a coding diagram from the embodiment;

[0021] Figure 3 This is a flowchart of the task execution logic repair mechanism in the embodiment;

[0022] Figure 4 This is a flowchart of the decoding process in the embodiment;

[0023] Figure 5 This is a flowchart of a local search process in the embodiment;

[0024] Figure 6 This is a flowchart illustrating the individual crossover process in the embodiment;

[0025] Figure 7 This is a flowchart illustrating individual variation in an example. Detailed Implementation

[0026] like Figure 1As shown in this embodiment, a two-target flight test mission optimization planning method based on the micro-matrix algorithm is involved, including:

[0027] Step 1: Set the population size, maximum number of iterations, crossover probability, and mutation probability for this embodiment;

[0028] Step 2: As Figure 2 As shown, a population is constructed and randomly initialized. The encoding of each individual's gene in this population consists of two parts: a flight assignment vector encoding and a task execution vector encoding, both with a length of n. Specifically: each gene in the task execution vector encoding represents a corresponding task number, and the entire vector constitutes the processing sequence for all tasks; the flight assignment vector encoding allocates compatible experimental machines according to the fixed task numbers.

[0029] For example, the task execution encoding vector <3, 5, 9, 6, 7, 4, 2, 8, 1, 10>, where 3, 5, 9, ..., 10 represent tasks 3, 5, 9, ..., 10. For the aircraft assignment encoding vector, each value in the sequence represents the aircraft assigned to the corresponding test flight task.

[0030] For example, Task 1 is assigned 1 testing machine, Task 2 is assigned 2 testing machines, and Task 3 is assigned 2 testing machines.

[0031] Step 3: As Figure 3 As shown, the task execution logic repair mechanism adjusts the task execution sequence encoding by using binary tree sorting and inorder traversal methods to meet the pre-constraints during task execution, specifically including:

[0032] Step 301: Analyze the logical relationships between flight test missions and establish a global logical matrix P to represent any two missions using breadth-first search, where: P ij =1 indicates that task i must be executed after task j, P ij =0 indicates that there is no logical relationship between tasks.

[0033] Step 302: For a task execution sequence encoding, construct a binary tree model, following the left-root-right order. That is, the task represented by the left subtree must be processed before the task represented by the root node of the subtree, and the task represented by the root node must be processed before the task represented by the right subtree. Therefore, performing an ordered traversal of the constructed binary tree transforms the original task execution sequence encoding into a task execution sequence encoding that satisfies the task execution logic constraints.

[0034] Step 4: As Figure 4 As shown, the idle time insertion technique is used to decode individual codes and calculate the fitness values ​​of two targets in the population. Specifically, this includes:

[0035] Step 401: Scan the values ​​i on the task execution vector from left to right to determine task i and its execution time t. i ;

[0036] Step 402: Find the corresponding executing aircraft k in the aircraft arrangement vector;

[0037] Step 403: Assign task i to aircraft k;

[0038] Step 4031: Check the free area on the pilot's aircraft K. in: The start time of the vacant area. This is the end time of the free area;

[0039] Step 4032: Calculate the earliest possible start time for task i on aircraft k. Among them: TP i c is the set of prerequisite tasks for task i. j The actual completion time of task j;

[0040] Step 4033: When task T i It can be inserted into a free area, that is If task i is assigned to this empty area, then task i will be assigned to the end of aircraft k, where: t i For task T i Execution time.

[0041] Step 404: Iterate through all subjects in the subject execution sequence according to steps 401 to 403 to complete the decoding.

[0042] Step 5: As Figure 5 As shown, 15% of dominant individuals are selected to perform a local search, specifically including:

[0043] Step 501: Randomly generate probability p;

[0044] Step 502:

[0045] a) When p < 0.2, randomly select two tasks from the task execution vector and swap them;

[0046] b) When p < 0.4, remove the randomly selected task from its position on the task execution vector and then insert it into another position;

[0047] c) When p < 0.6, randomly select a task and transfer it to another compatible test machine with lower utilization;

[0048] d) When p < 0.8, randomly select a task and transfer it to another compatible test machine with reduced task cost;

[0049] e) When p < 1, randomly select a task and randomly transfer it to another compatible test machine.

[0050] Step 6: Generate new offspring individuals through lexicographical optimization selection, specifically including:

[0051] Step 601: Randomly select two people from the population (each person has an equal probability of being selected), denoted as P. A and P B FTD stands for flight test cycle, and TTC stands for mission transfer cost.

[0052] Step 602: If FTD(P) A )≌FTD(P B If the FTD is smaller, then select the individual with the smaller FTD.

[0053] Step 603; If FTD(P) A )==FTD(P B If the TTC is smaller, then select the individual with the smaller TTC.

[0054] Step 604; Repeat step 601 until the new population reaches the original population size.

[0055] Step 7: As Figure 6 As shown, offspring are generated through crossover operations, specifically including:

[0056] Step 701: Assign the pilot assignment vector AAV to the two individuals i and j in the population. i With AAV j (i≠j) are used as parent sequence 1 and parent sequence 2 respectively;

[0057] Step 702: For each gene position in the subsequence, randomly generate a probability p. If p < 0.5, obtain the allele of parent 1; otherwise, obtain the allele of parent 2.

[0058] Step 703: Select the task execution vector (TSV) corresponding to two individuals i and j in the population. i With TSV j (i≠j) are used as parent sequence 1 and parent sequence 2 respectively;

[0059] Step 704: Randomly divide all tasks into two non-empty task sets G1 and G2. Specifically, first, a baseline probability p is randomly generated. For each task i, if rand < p, add task i to set G1; otherwise, add task i to set G2. If either G1 or G2 is empty, randomly select one task and add it to this empty set.

[0060] Step 705: Copy the gene corresponding to G1 from the parent generation to offspring 1, keeping the gene position unchanged. The remaining vacant positions in offspring 1 will be filled by genes from parent 2, with the same gene order as in parent 2;

[0061] Step 706: Similarly, the G2 gene of parent generation 2 remains in the same position in offspring generation 2, and the remaining empty positions are filled by the parent generation 1 gene of the same order belonging to G1 in parent generation 1.

[0062] Step 8: As Figure 7 As shown, offspring are generated through mutation operations, specifically including:

[0063] Step 801: Assign the pilot sequence AAV to individual i in the population. i As the parent generation sequence;

[0064] Step 802: Randomly select a task i from the parent sequence and assign it to another available aircraft;

[0065] Step 803: Transfer the task execution vector TSV corresponding to individual i in the population. i As the parent generation sequence;

[0066] Step 804: Randomly select a task i from the parent sequence, then randomly select another different task j (i≠j), and swap their positions.

[0067] Step 9: If the termination criterion is met, output the optimal solution obtained; otherwise, return to step 3.

[0068] This embodiment implements the algorithm for tasks ranging from 60 tasks × 3 aircraft to 500 tasks × 6 aircraft. The algorithm parameters are set as follows: population size popsize = 50, crossover probability pc = 0.8, mutation probability pm = 0.1. For instances 01-06, 07-12, and 13-18, the maximum number of iterations max_iter are 50, 100, and 150, respectively. The algorithm results are shown in the table below.

[0069] Compared with existing technologies, this method adopts a task repair mechanism to ensure that the logical relationship between test flight tasks is always satisfied during the population iteration process; it can effectively optimize the test flight cycle and task transfer cost through lexicographical optimization; it improves the optimization effect of the test flight cycle by using idle time insertion technology for decoding; and it improves the optimization effect of the algorithm by using local search operators.

[0070] This invention addresses the flight test mission planning problem by assigning each flight test mission to a suitable test aircraft and determining the execution order, thereby generating an efficient and feasible flight test plan. This invention is applicable to the formulation of flight test planning and implementation plans and has significant practical application value. This invention proposes a two-objective flight test mission optimization planning method based on a dense algorithm. First, it proposes an encoding method including aircraft arrangement vectors and mission execution vectors, and a task relationship repair mechanism based on a binary tree, which can ensure that the pre-condition logical constraints between flight test missions are met. Second, the proposed idle time insertion decoding technique improves the utilization rate of the test aircraft. Based on the importance of the objectives, the flight test cycle is optimized first, followed by optimization of mission transfer costs, which can efficiently solve the two-objective problem. During the evolutionary process, in addition to crossover and mutation operators, the local search operators for the two objectives of flight test mission planning can fully utilize the solution space, improving the algorithm's global search capability.

[0071] The above-described specific implementations can be partially adjusted by those skilled in the art in different ways without departing from the principles and purpose of the present invention. The scope of protection of the present invention is defined by the claims and is not limited to the above-described specific implementations. All implementation schemes within the scope of the claims are bound by the present invention.

Claims

1. A two-objective flight test mission optimization planning method based on the Mimu algorithm, characterized in that, include: Step 1: Set the population size Maximum number of iterations Crossover probability and mutation probability ; Step 2: Set the number of iterations Randomly generated according to the encoding rules, with a size of Initial population Specifically, this involves constructing and randomly initializing a population. The encoding of each individual's gene in this population comprises two parts: a flight assignment vector encoding and a task execution vector encoding, both of which have a length of [missing information]. In this context: each gene in the task execution vector encoding represents the corresponding task number, and the entire vector constitutes the processing sequence of all tasks; the aircraft arrangement vector encoding allocates compatible experimental aircraft according to the fixed sequence number of the task. Step 3: Repair the task execution vector encoding for each individual in the population using a task execution logic repair mechanism; The aforementioned task execution logic repair mechanism refers to adjusting the task execution sequence encoding using binary tree sorting and inorder traversal methods to meet the pre-constraints during task execution, specifically including: Step 301: Analyze the logical relationships between flight test missions and establish a global logical matrix to represent any two missions using breadth-first search. ,in: Indicates task Must be in the task Then execute, This indicates that there is no logical relationship between the tasks; Step 302: For a task execution sequence encoding, a binary tree model is established in the order of left-root-right, that is, the task represented by the left subtree must be processed before the task represented by the root node of the subtree, and the task represented by the root node must be processed before the task represented by the right subtree. Therefore, the constructed binary tree is traversed in an ordered manner to transform the original task execution sequence encoding into a task execution sequence encoding that can meet the requirements of the task execution logic constraints. Step 4: Use idle time insertion technique to decode the encoding of individuals in the population and calculate the fitness values ​​of two targets in the population; Step 5: Select the 15% of dominant individuals to perform a local search to obtain the population. ; Step 6: For the population New offspring individuals are generated through lexicographical optimization. ; Step 7: Population Perform crossover operation to generate offspring ; Step 8: Population Offspring are generated through mutation operations. ; Step 9: When the number of iterations... Maximum number of iterations not reached ,but If the condition is met, proceed to step 3; otherwise, output the result.

2. The two-target flight test mission optimization planning method based on the micro-matrix algorithm according to claim 1, characterized in that, Step 4 specifically includes: Step 401: Scan the values ​​on the task execution vector from left to right. Define the task and its execution time ; Step 402: Locate the corresponding executing aircraft in the aircraft allocation vector. ; Step 403: Transfer the task Assigned to pilot superior; Step 4031: Inspect the aircraft free space on ,in: The start time of the vacant area. This is the end time of the free area; Step 4032: Calculation Task piloting the aircraft Earliest start time ,in: For the task A collection of prerequisite tasks. For the task The actual completion time; Step 4033: When the task It can be inserted into a free area, that is The task will be completed at that time. Arrange the task in this available area; otherwise, the task will be... Arranged to fly the plane At the end, where: For the task Execution time; Step 404: Iterate through all subjects in the subject execution sequence according to steps 401 to 403 to complete the decoding.

3. The two-target flight test mission optimization planning method based on the micro-matrix algorithm according to claim 1, characterized in that, Step 5 specifically includes: Step 501: Randomly generate probabilities ; Step 502: a) When Two tasks are randomly selected from the task execution vector and swapped. b) When The randomly selected task is removed from its position on the task execution vector and then inserted into another position. c) When Randomly select a task and transfer it to another compatible testing machine with lower utilization; d) When Randomly select a task and transfer it to another compatible test machine with reduced task costs; e) When Randomly select a task and randomly transfer it to another compatible test machine.

4. The two-target flight test mission optimization planning method based on the micro-matrix algorithm according to claim 1, characterized in that, Step 6 specifically includes: Step 601: Randomly select two people from the population with equal probability, denoted as... and , The test flight cycle is represented by TTC, which is the cost of mission transfer. Step 602: If If so, select the individual with the smaller FTD; Step 603; if If so, select the individual with the smaller TTC; Step 604; Repeat step 601 until the new population reaches the original population size.

5. The two-target flight test mission optimization planning method based on the micro-matrix algorithm according to claim 1, characterized in that, Step 7 specifically includes: Step 701: Combine two individuals from the population The corresponding aircraft arrangement vector and These are respectively used as parent sequence 1 and parent sequence 2; Step 702: For each gene position in the subsequence, randomly generate a probability. If the allele is positive, then the allele of father 1 is obtained; otherwise, the allele of father 2 is obtained. Step 703: Select two individuals from the population The corresponding task execution vector and Each is treated as a parent sequence 1 and a parent sequence 2, respectively. Step 704: Randomly divide all tasks into two non-empty task sets. and Specifically, a baseline probability is first randomly generated. For each task Then the task Add to collection Otherwise, the task will be completed. Add to collection If it appears or If the set is empty, then randomly select one task and add it to the empty set; Step 705: Transfer the parent generation's The corresponding gene is copied into offspring 1, keeping the gene position unchanged; the remaining empty slots in offspring 1 will be filled by genes from parent 2, and the gene order will be the same as that in parent 2. Step 706: Similarly, parent generation 2... Genes retain the same position in F2, and the remaining empty positions are filled by genes with the same order in parent 1. The parent 1 gene is filled.

6. The two-target flight test mission optimization planning method based on the micro-matrix algorithm according to claim 1, characterized in that, Step 8 specifically includes: Step 801: Individuals in the population Corresponding flight arrangement sequence As the parent generation sequence; Step 802: Randomly select a task from the parent sequence. It will be assigned to another available aircraft; Step 803: Individuals in the population The corresponding task execution vector As the parent generation sequence; Step 804: Randomly select a task from the parent sequence. Then randomly select another different task. The positions of the two are swapped.

7. A two-objective flight test mission optimization planning system that implements the two-objective flight test mission optimization planning method based on the micro-matrix algorithm as described in any one of claims 1-6, characterized in that, include: The system comprises a flight test database module, a flight test mission optimization algorithm module, a system support module, and a comprehensive display module. Specifically: the flight test database module contains a built-in set of flight test mission entities and a set of test aircraft entities; the flight test mission optimization algorithm module, based on the data provided by the flight test database module, calls the flight test mission planning algorithm to assign each flight test mission to a suitable test aircraft, sorts the assigned missions on the corresponding test aircraft, and outputs a Gantt chart of the flight test mission prototype network allocation; the system support module verifies whether the user has system access permissions, inputs parameters selected by the user into some algorithms, and appropriately processes the input and output data of each module within the system to ensure they meet the required data types; the comprehensive display module provides a user-friendly interface for user login, algorithm parameter settings, and comprehensive result display.

8. The two-target flight test mission optimization planning system according to claim 7, characterized in that, The flight test mission entity set includes: mission number, mission name, mission execution duration, set of prerequisite missions, mission preparation time, and set of compatible test aircraft; the test aircraft entity set includes the test flight prototype number, test flight prototype deployment time, and test aircraft executable missions.

Images