Statistical Dependence Analyses of Operational Flight Data Used for Landing Reconstruction Enhancement

Lukas Höhndorf, Thomas Nagler, Phillip Koppitz, Claudia Czado, Florian HolzapfelarXiv 2022 · DOI · arXiv

Abstract

The Rauch-Tung-Striebel (RTS) smoother is a cornerstone of state estimation in aviation, used to enhance the physical coherence and resolution of recorded flight data. This paper proposes a method to improve aircraft landing reconstruction by refining the statistical assumptions regarding measurement noise. Traditionally, noise is assumed to be constant and independent; however, this study demonstrates that noise characteristics change over time. By implementing an iterative framework that estimates time-varying covariance matrices via kernel smoothing and analyzes residual structures using copula models, the researchers significantly minimize errors in operational flight data. Results show improved accuracy in 83% of test cases, providing a more reliable basis for Flight Data Monitoring (FDM).

Introduction

Flight Data Monitoring (FDM) is essential for airline safety and efficiency, relying on data from Quick Access Recorders (QAR). However, raw QAR data is often noisy, low-resolution, or physically inconsistent. To address these issues, engineers use the Rauch-Tung-Striebel (RTS) smoother, a state-estimation algorithm that "cleans" data by fitting it to a physical model of aircraft dynamics. Standard RTS applications typically assume that measurement noise is "white noise" with a constant, diagonal covariance matrix. In reality, noise levels fluctuate during landing, and errors across different sensors are often correlated. Ignoring these dependencies leads to suboptimal flight path reconstructions.

Methodology

The authors propose a two-stage iterative framework to move beyond static noise assumptions:

  1. Initial Pass: A standard RTS smoother (based on an Extended Kalman Filter) is run using a fixed, diagonal measurement noise covariance matrix ($R$).
  2. Noise Estimation: Residuals from the first pass are used to estimate a time-varying covariance matrix ($R_k$) via kernel smoothing. This moving average approach with a Gaussian kernel captures how noise levels evolve throughout the landing.
  3. Refined Pass: The RTS smoother is re-run using the dynamic $R_k$, with correlation limits applied to maintain numerical stability.
  4. Copula Analysis: To understand remaining errors, copula models are employed to analyze the dependence structure between variables independently of their marginal distributions, helping identify potential deficiencies in the underlying physical model.
System model architecture illustrating the integration of measured inputs and noise into the mathematical state-space model.
The recursive prediction and correction cycle of the Kalman filter, which forms the basis for the RTS smoother.

Datasets and Results

The study utilized operational flight data from 24 civil aircraft landings, including parameters such as GPS position, barometric and radio altitudes, attitude angles, and Instrument Landing System (ILS) deviations. The analysis focused on the critical landing phase from 1000 feet above ground level to the airport apron.

Key findings include:

  • Heteroscedasticity: Noise variance is not static; it fluctuates significantly during the landing sequence, proving that a fixed noise model is insufficient.
  • Improved Accuracy: The refined smoother outperformed the standard pass in 83% of cases, producing trajectories that are more physically plausible.
  • Model Diagnostics: Copula analysis revealed strong, non-Gaussian dependencies between attitude angles and rotational rates, suggesting areas where the physical transition models can be further improved.

Conclusion

Moving beyond the assumption of constant, independent noise significantly enhances the quality of aircraft landing reconstruction. By iteratively refining noise covariance and analyzing residual dependencies, this approach provides a more accurate and robust framework for Flight Data Monitoring. Future work will focus on automating parameter selection and further refining the physical models based on the identified statistical dependencies.

Summary updated

Abstract

The RTS smoother is widely used for state estimation and it is utilized here to increase the data quality with respect to physical coherence and to increase resolution. The purpose of this paper is to enhance the performance of the RTS smoother to reconstruct an aircraft landing using on board recorded data only. Thereby, errors and uncertainties of operational flight data (e.g. altitude, attitude, position, speed) recorded during flights of civil aircraft are minimized. These data can be used for subsequent analyses in terms of flight safety or efficiency, which is commonly referred to as Flight Data Monitoring (FDM). Statistical assumptions of the smoother theory are not always verified during application but (consciously or not) assumed to be fulfilled. These assumptions can hardly be verified prior to the smoother application, however, they can be verified using the results of an initial smoother iteration and modifications of specific smoother characteristics can be suggested. This project specifically verifies assumptions on the measurement noise characteristics. Variance and covariance of the measurement noise can be checked after the initial smoother application. It is discovered that these characteristics change over time and should be accounted for with a time varying covariance matrix. This sequence of matrices is estimated by kernel smoothing and replaces an initially assumed fixed and diagonal covariance matrix used for the first smoother run. The results of this second smoother iteration are mostly improved compared to the initial iteration, i.e. the errors are significantly reduced. Subsequently, the remaining dependence structures of the residuals of the second smoother iteration can be captured by copula models. Their interpretation is useful for a revision of the physical model utilized by the RTS smoother.
BibTeX
@misc{hohndorf2022statistical,
  author = {Lukas Höhndorf and Thomas Nagler and Phillip Koppitz and Claudia Czado and Florian Holzapfel},
  title = {{Statistical Dependence Analyses of Operational Flight Data Used for Landing Reconstruction Enhancement}},
  year = {2022},
  eprint = {2206.09809},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2206.09809},
}
Figure