[Paper Review] Implementation of real-time moving horizon estimation for robust air data sensor fault diagnosis in the RECONFIGURE benchmark

Yiming Wan, Tamás Keviczky|arXiv (Cornell University)|Jun 4, 2016

Fault Detection and Control Systems3 citations

TL;DR

This paper presents a real-time moving horizon estimation (MHE) approach for robust fault diagnosis of air data sensors (airspeed and angle-of-attack) in the RECONFIGURE benchmark, using a constrained MHE formulation with wind bounds to enhance fault sensitivity while maintaining disturbance robustness. The method achieves sub-1s detection delays and estimation errors below 1.85 kts for VCAS and 0.82 deg for AOA across diverse flight conditions, validated via industrial-compatible implementation using Airbus SAO library.

ABSTRACT

This paper presents robust fault diagnosis and estimation for the calibrated airspeed and angle-of-attack sensor faults in the RECONFIGURE benchmark. We adopt a low-order longitudinal model augmented with wind dynamics. In order to enhance sensitivity to faults in the presence of winds, we propose a constrained residual generator by formulating a constrained moving horizon estimation problem and exploiting the bounds of winds. The moving horizon estimation problem requires solving a nonlinear program in real time, which is challenging for flight control computers. This challenge is addressed by adopting an efficient structure-exploiting algorithm within a real-time iteration scheme. Specific approximations and simplifications are performed to enable the implementation of the algorithm using the Airbus graphical symbol library for industrial validation and verification. The simulation tests on the RECONFIGURE benchmark over different flight points and maneuvers show the efficacy of the proposed approach.

Motivation & Objective

• To address the challenge of detecting and isolating simultaneous multiple faults in triplex-redundant air data sensors, which triplex voting cannot handle.
• To improve fault sensitivity in the presence of wind disturbances without compromising robustness.
• To implement a computationally intensive constrained MHE-based fault diagnosis method on flight control computers using industrial-standard graphical symbol libraries.
• To validate the method on the high-fidelity RECONFIGURE benchmark across diverse flight points and maneuvers.
• To enable industrial validation and verification (V&V) of advanced optimization-based FDI methods in an Airbus-compliant environment.

Proposed method

• Formulates a constrained residual generator as a moving horizon estimation (MHE) problem, incorporating bounds on wind dynamics to enhance fault sensitivity.
• Solves the nonlinear program using an interior-point sequential quadratic programming (IP-SQP) strategy within a real-time iteration scheme with fixed iteration count for predictable computational cost.
• Employs a structure-exploiting Riccati-based method to efficiently solve the linearized Karush-Kuhn-Tucker (KKT) system at each time step.
• Applies specific algorithmic approximations and simplifications to enable implementation using the Airbus SAO graphical symbol library, which lacks native matrix operation blocks.
• Uses a fixed number of iterations per sample to ensure real-time feasibility, accepting suboptimal solutions for computational speed.
• Integrates wind state estimation into the MHE framework to maintain observability even when air data sensors fail.

Experimental results

Research questions

• RQ1 Can a constrained MHE formulation improve fault sensitivity for simultaneous AOA and VCAS sensor faults while maintaining robustness to wind disturbances?
• RQ2 How can a computationally intensive nonlinear program for MHE be implemented in real time on flight control computers with limited computational resources?
• RQ3 What level of fault detection delay and estimation error can be achieved using MHE-based FDI under realistic wind and fault profiles in the RECONFIGURE benchmark?
• RQ4 Can the MHE-based FDI method be successfully implemented using the Airbus SAO graphical symbol library for industrial V&V?
• RQ5 How does the method perform under extreme wind shear and multiple fault types, such as bias, oscillation, and runaway?

Key findings

• The method achieves a maximum absolute estimation error (AEE) of 0.82 deg for angle-of-attack and 1.85 kts for calibrated airspeed across all fault scenarios, except in Scenario 2-F where all three VCAS sensors fail.
• Detection delays are mostly below 0.45 s, with the longest delay of 0.92 s occurring in Scenario 6-F due to a trade-off between false alarm avoidance and detection speed.
• In Scenario 7-F, despite strong wind shear (peak amplitude 30.87 kts), the method maintains satisfactory fault detection and estimation performance.
• The real-time computational cost of the SAO-implementation is 5.8 ms per sample, confirming feasibility for flight control computer deployment.
• The VCAS estimation error in Scenario 2-F is unacceptably high (74.36 kts max) because all three VCAS sensors fail and horizontal wind becomes unobservable.
• The method successfully isolates faulty sensors and maintains low estimation errors in all scenarios except when multiple VCAS sensors fail, highlighting the importance of sensor redundancy for wind observability.