Abstract
Aerodynamic force parameter identification is the main measure of aerodynamic numerical simulation and wind tunnel test results verification. The wide use of intelligent control in the control field provides a new way to resolve the problem for system identification. Due to the global search optimization ability of intelligent algorithm, it can effectively improve precision and efficiency of aircraft aerodynamic force parameter identification. This paper gives a brief review of system identification method based on intelligent algorithm and its application in aircraft aerodynamic force parameters identification, including genetic algorithm, particle swarm optimization and artificial colony algorithm, etc. In this paper, the research work will provide useful technology reference for the aircraft aerodynamic force parameter identification.
Introduction
To ensure the safety and performance of modern aircraft, engineers require highly precise mathematical models of aerodynamic forces. Aerodynamic parameter identification is the process of extracting these models from experimental data, typically obtained via wind tunnel tests or flight trials. Traditionally, this field has been dominated by the Maximum Likelihood Method (MLM), which treats identification as an optimization problem aimed at minimizing the discrepancy between predicted and observed outputs.
However, as aircraft design moves toward high Mach numbers and unsteady flight regimes, traditional gradient-based methods like MLM face significant hurdles. This paper reviews the emergence of "intelligent algorithms"—bio-inspired optimization techniques—that offer a robust alternative for modeling complex, nonlinear aerodynamic behavior.
The Limitations of Traditional Identification
Traditional gradient-based optimization methods, while computationally efficient, suffer from three primary flaws in the context of modern aerodynamics:
- Sensitivity to Initial Values: These algorithms require a starting guess. If the initial estimate is far from the truth, the algorithm often converges on a "local optimum" rather than the global best solution.
- Mathematical Rigidity: Methods like MLM require the objective function to be continuous and differentiable. Complex aerodynamic phenomena often produce "pathological" landscapes where derivatives are difficult or impossible to calculate.
- Complexity Scaling: In regimes involving high degrees of coupling between parameters, traditional methods often become unstable or fail to converge entirely.
Intelligent Algorithms for Aerodynamic Modeling
Intelligent algorithms address these flaws by mimicking natural processes to explore the solution space more effectively. Three primary algorithms are currently at the forefront of this research:
Genetic Algorithm (GA)
Inspired by Darwinian evolution, GA represents aerodynamic parameters as "genes" within a population of candidate solutions. Through iterative cycles of selection, crossover (breeding), and mutation, the population evolves. Because GA searches from a group of points rather than a single starting point, it is significantly less likely to be trapped by local optima.
Particle Swarm Optimization (PSO)
PSO mimics the social foraging behavior of bird flocks or fish schools. Each candidate solution is a "particle" moving through the search space, adjusted based on its own historical success and the success of its neighbors. PSO is particularly valued for its effectiveness in handling non-differentiable and multimodal functions.
Artificial Bee Colony (ABC)
Simulating the organized foraging of honeybees, the ABC algorithm employs three classes of agents: employed bees, onlooker bees, and scout bees. This structure creates a sophisticated balance between exploiting known high-quality solutions and exploring new areas of the search space, making it highly resistant to premature stagnation.
Methodological Framework
The application of these algorithms generally follows a standardized identification loop:
- Excitation: An input signal (such as a "3211" pulse, step, or frequency scan) is applied to the aircraft.
- Observation: The aircraft's response—including flight speed, angle of attack, sideslip angle, and pitching velocity—is recorded.
- Optimization: The intelligent algorithm iteratively adjusts the model parameters to maximize a fitness function, which is typically defined as the reciprocal of the error between the model and the flight data.
Insights and Comparative Performance
The shift from local gradient-based climbing to global population-based searching represents a paradigm shift in system identification. These algorithms are derivative-free, meaning they only require the value of the objective function, providing immense flexibility for discontinuous aerodynamic models.
Research indicates that Genetic Algorithms are particularly robust when dealing with "serious coupling," where multiple parameters influence each other simultaneously. Meanwhile, the Artificial Bee Colony algorithm has demonstrated a superior ability to escape local optima compared to earlier stochastic methods. However, these benefits come at a cost: intelligent algorithms are computationally "heavy" and require significantly more processing power than traditional MLM due to the need to evaluate entire populations of solutions.
Conclusion and Future Directions
While intelligent algorithms offer a more reliable path to accurate aerodynamic models, their computational overhead remains a barrier to real-time application. To address this, the authors suggest two primary paths forward:
- Hybrid Approaches: Utilizing an intelligent algorithm for a broad global search to locate the optimal "neighborhood," followed by a traditional gradient-based method for rapid, precise local refinement.
- Parallel Computing: Leveraging the naturally parallel structure of population-based algorithms to distribute the computational load across multiple processors.
By combining the global search capabilities of bio-inspired algorithms with the local efficiency of traditional methods, engineers can achieve a new level of precision in aircraft aerodynamic force parameter identification.
Summary updated