Model-Based Detection and Isolation of Rudder Faults for a Small UAS

Raghu Venkataraman, Peter SeilerAIAA 2015 · DOI

TL;DR

This paper presents a model-based fault detection and isolation (FDI) filter designed to identify and locate stuck rudder faults in small unmanned aircraft systems (UASs). By combining analytical redundancy with a split rudder design, the system aims to enhance the reliability of UASs to meet stringent safety standards.

What Problem Does This Paper Address?

The increasing integration of Unmanned Aircraft Systems (UASs) into national airspace necessitates higher reliability, comparable to manned aircraft. Current UASs often rely on single-string components, making them vulnerable to catastrophic failures. While hardware redundancy is a solution, it's often limited by size, weight, and power constraints in small UASs. This paper addresses the challenge of improving UAS reliability by proposing a method that leverages analytical redundancy – using mathematical models to detect faults – in conjunction with a physically redundant component (a split rudder) to mitigate the impact of rudder failures. Specifically, it focuses on detecting and isolating "stuck rudder" faults, where one or both rudder surfaces become fixed in a certain position.

Methods

The core of the paper is a model-based FDI filter. This filter operates by comparing the real-time behavior of the aircraft (measured outputs and control commands) with the predicted behavior from a high-fidelity simulation model.

Here's a breakdown of the key methods:

  • Model-Based Comparison: The FDI filter uses a physics-based model of the aircraft's flight dynamics. This model, when fed the same control inputs as the real aircraft, generates expected outputs.
  • Residual Generation: The difference between the real aircraft's outputs and the model's predicted outputs creates "residuals." These residuals are expected to be near zero under normal conditions.
  • Detection Logic: The detection stage analyzes these residuals for transient deviations that are characteristic of rudder faults. It uses flags based on the signs of residuals from multiple sensors (airdata and IMU) to identify specific fault patterns. This logic is designed to be robust to noise, turbulence, and wind gusts.
  • Isolation and Estimation: Once a fault is detected, the isolation and estimation stage pinpoints which rudder (top or bottom) is faulty and estimates the magnitude of the fault. This is achieved by analyzing the residuals of control commands, particularly the aileron command, which exhibits a distinct steady-state behavior depending on which rudder is stuck.
  • Robust Controller: A classical flight control law is also designed to be robust to rudder faults. This controller uses integral action on sideslip angle to maintain stability even when a rudder is stuck, which, while improving robustness, makes fault detection more challenging.
  • Split Rudder Design: The aircraft is physically modified with a split rudder, allowing for some yaw control authority even if one part of the rudder is stuck. This provides a degree of hardware redundancy.

Datasets

The paper does not explicitly mention using pre-existing datasets for training or evaluation. Instead, it relies on:

  • High-Fidelity Simulation Environment: A custom, documented simulation environment for the Ultra Stick 120 aircraft, built using Matlab/Simulink, is used. This environment includes models for aircraft dynamics, actuators, sensors, and environmental effects.
  • Simulated Fault Injection: Faults (stuck rudders) are injected into the simulation to generate data for evaluating the FDI filter's performance.

Key Insights and Ideas

  • Analytical Redundancy as a Complement to Hardware: The paper demonstrates that analytical redundancy, through model-based FDI, can significantly enhance the reliability of UASs, especially when hardware redundancy is limited.
  • Physics-Based Fault Modeling is Crucial: Understanding the underlying flight dynamics of how a fault manifests is key to designing an effective FDI filter. The paper emphasizes using flight dynamics principles to architect the filter.
  • Transient Response is Key for Detection: Steady-state behavior alone is insufficient to detect stuck rudder faults due to the controller's compensatory actions. The FDI filter leverages the transient deviations in sensor readings.
  • Aileron Command as an Isolation Indicator: While rudder faults might produce similar initial responses in other states, the steady-state aileron command provides a unique signature to distinguish between top and bottom rudder faults.
  • Trade-off Between Robustness and Detectability: Making the controller more robust to faults can make those faults harder to detect. The FDI filter's design balances this by looking for specific, consistent fault signatures.

Key Results

  • Robustness to Disturbances: The FDI filter successfully rejected wind gusts and was robust to model uncertainty and commanded maneuvers, avoiding false alarms.
  • Fault Detection: The filter was able to detect stuck rudder faults with a detection time as low as 0.44 seconds for a 10° fault.
  • Fault Isolation and Estimation: The filter accurately isolated the faulty rudder and estimated the fault magnitude. For a +10° bottom rudder fault, the estimated fault was +9.4°, resulting in a small estimation error.
  • Fault Magnitude Impact: Larger injected faults generally led to faster detection times.
  • Undetectable Fault Range: A range of faults between -7° and +7° was found to be undetectable by the current FDI logic, highlighting a limitation.

Limitations and Future Work

  • Undetectable Fault Range: The current FDI filter has a range of small faults that it cannot reliably detect. Future work could involve adjusting thresholds or the number of consecutive samples required for detection to cover this range, though this might impact robustness.
  • Simulation-Based Evaluation: All results presented are based on simulations. The authors state that future work will involve validating these simulation results through actual flight tests.
  • Limited Fault Types: The study focused specifically on stuck rudder faults. Expanding the FDI filter to cover other fault modes (e.g., partial failures, actuator failures) is a natural next step.
  • Monte Carlo Simulations: The authors acknowledge that for a more thorough characterization of performance and robustness, Monte Carlo simulations would be beneficial to account for the stochastic nature of residuals.

Summary updated

Abstract

New regulatory safety standards will soon require unmanned aircraft systems to meet high levels of reliability. There is potential to increase the reliability of such systems without necessarily increasing the number of hardware components. This paper motivates a mix of physical and analytical redundancy in order to increase the system-level reliability of a small unmanned aircraft. The aircraft discussed in this paper has a split rudder for fault-tolerant control. Hardware faults, such as a stuck rudder, need to be detected and isolated in real-time in order for the controller to be reconfigured. In this paper, flight dynamics principles are used to design a model-based filter for detecting and isolating stuck faults in the split rudder of the aircraft. A classical controller is developed in order to make the aircraft robust to stuck rudder faults. The performance and robustness of the filter is evaluated, in closed-loop, through high fidelity simulations. The results in this paper highlight the potential for increasing the reliability of safety-critical aviation systems through analytical redundancy.
BibTeX
@inproceedings{venkataraman2015model,
  author = {Raghu Venkataraman and Peter Seiler},
  title = {{Model-Based Detection and Isolation of Rudder Faults for a Small UAS}},
  booktitle = {AIAA},
  year = {2015},
  doi = {10.2514/6.2015-0857},
}
Figure