Aircraft Test Data Generation for Multi-Threshold Disturbance States
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing black box optimization testing methods for aircraft flight stability under disturbances or system faults fail to adequately address multi-modal multi-objective optimization problems, particularly in identifying disturbance variable combinations that simultaneously cause multiple flight state indexes to reach unsafe thresholds, due to limitations in topology structure definition and solution diversity.
Innovation Solution
A multi-modal multi-objective particle swarm optimization algorithm based on adaptive resonance topology network is employed to iteratively update particle positions, utilizing an adaptive resonance topology network to learn the distribution of Pareto optimal solutions without predefined topology, ensuring convergence and exploring decision space diversity, and using a testing data archive to avoid duplicates.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If classical niching method is used to locate multiple optimal solutions, then solution diversity is improved, but parameter sensitivity increases making the method extremely sensitive to parameter values
Solution Approach 1:
The patent transforms the classical niching method by changing from fixed parameter-based niching to topology-based adaptive resonance. The resonance parameter σ is dynamically adjusted based on the distribution of non-dominated solutions rather than being fixed, and the topology structure adapts as new solutions are discovered, making the method less sensitive to initial parameter settings while maintaining solution diversity
Solution Approach 2:
The patent introduces dynamic topology adaptation where the network structure evolves continuously as new non-dominated solutions are found. The topology parameters (number of nodes, connections) are not fixed but adapt dynamically based on the distribution of solutions in the decision space, allowing the method to automatically adjust to different problem characteristics without manual parameter tuning
2Ease of operation
If self-organizing mapping network is used to provide neighborhood information, then local search capability is improved, but predefined topology information is required which limits applicability
Solution Approach 1:
The patent eliminates the need for preliminary topology definition by using adaptive resonance theory. Instead of pre-defining the network structure, the topology is built incrementally as solutions are discovered, with nodes and connections formed dynamically based on the actual distribution of non-dominated solutions in the decision space
Solution Approach 2:
The adaptive resonance network automatically adjusts its own topology structure based on the input data distribution. The network serves itself by dynamically creating nodes and connections as new non-dominated solutions are identified, without requiring external specification of topology parameters or prior knowledge of the problem structure
3Ease of manufacture
If circular topology is used to form stable niching, then parameter definition is simplified, but true topological structure of particle distribution cannot be reflected
Solution Approach 1:
The patent replaces the static circular topology with a dynamic adaptive topology that evolves as solutions are discovered. The network structure continuously adjusts to match the actual distribution of non-dominated solutions in the decision space, accurately reflecting the true topological relationships without being constrained by a predetermined circular pattern
Solution Approach 2:
The patent changes the topology parameters dynamically based on the distribution of non-dominated solutions. Instead of using a fixed circular arrangement, the number of nodes, their positions, and connection patterns are all adjusted to match the actual solution distribution, providing accurate topological representation while maintaining implementation simplicity
Data Source
AI summary
A multi-modal multi-objective testing data generation method based on topology adaptive resonance theory is disclosed. The method considers that that the test aircraft under what kind of external disturbance or its own system fault conditions will be close to multiple index thresholds of unreliable or unsafe state at the same time as a black box optimization testing problem. The testing data is generated through optimization algorithms to test the aircraft state under different disturbance conditions. This type of black box optimization problem belongs to multi-modal multi-objective optimization problems. The testing problem can be solved using a multi-modal multi-objective particle swarm optimization algorithm based on adaptive resonance topology network. The external disturbance variable parameter combination encountered by the aircraft is regarded as the particle position, and the optimal parameter combination is found by iteratively updating the particle position.