Florian Schwaiger

• Master of Science
• Research Assistant at Technical University of Munich

This paper presents a method to assess a novel procedure for Air Traffic Controllers, enabled by predictions of a machine learning-based go-around prediction model, regarding the operational risk of separation infringements and traffic alarms. In a previous work, potentially novel procedures were elaborated in human-in-the-loop simulations with Air...

This paper presents a mission data-handling algorithm that provides a data link interface between a flight control computer (FCC) and a ground control station (GCS) using the protocol standard STANAG 4586, AEP-84 Volume II. A subset of defined messages is used for the transmission of mission and area definition data. Functions for decoding uplink m...

During system or software development, teams need to work with multiple tools which each have their use cases and strengths. For the combination of application life-cycle management (ALM) and model-based design (MBD) software, this paper presents a process for continuous requirement validation tests with complex multi-dimensional test parameters. I...

View Video Presentation: https://doi.org/10.2514/6.2022-1893.vid We present an open, modular toolbox for Matlab that implements an algorithm for Subset Simulation (SuS) with Markov-chain Monte Carlo (MCMC) sampling to estimate occurrence probabilities of rare events, available at http://go.tum.de/605579. The toolbox is built around an object-orient...

View Video Presentation: https://doi.org/10.2514/6.2022-1894.vid The certification of automatic landing systems requires demonstration of compliance with the requirements outlined by aviation authorities. In recent years, the certification authorities have shifted from prescriptive requirements to performance-based requirements. Compliance may be s...

View Video Presentation: https://doi.org/10.2514/6.2022-1868.vid We present the combination of Subset Simulation with refined, active learning Kriging meta-modeling to accelerate the estimation of rare event failure probabilities. To this end, we replace the computationally expensive limit state function by a more cost-effective Kriging surrogate m...

View Video Presentation: https://doi.org/10.2514/6.2022-1895.vid The development of modern flight control systems (FCS) is a highly iterative procedure. To ensure that modifications to the flight control laws do not entail adverse effects on the closed-loop behavior, we propose the integration of Optimal Control based test cases into the Continuous...

The safety and performance of aircraft systems must fulfill stringent requirements defined by certification authorities. Performance-based requirements are specified with a probability threshold. The probability of failure of the performance of the system must be below the threshold. Depending on the severity of the outcome, the probability thresho...

We present an object-oriented acceleration framework in Matlab to speed up estimating rare failure probabilities of simulation models in Simulink w.r.t. uncertain parameters. This framework bridges the gap between the execution of Simulink models and the generic implementation of the subset simulation algorithm, which we use to efficiently explore...

To support big data applications for flight data analysis, we present an improved method for the decoding of binary ARINC 717 recorder files. Our tool transcribes flight parameters from any given dataframe layout of the raw file into one of several tabular open data standards (e.g. .parquet, .csv). Our improved method cuts execution time by a facto...

This paper presents a machine learning algorithm trained to predict unstable approach events. Predictive modeling for unstable approaches (UA) forecasting needs a precursors analysis to determine the most important indicators (features) of aircraft instability. However, since the definition of aircraft instability is entirely dependent on the airli...

Airlines safety departments analyse aircraft data recorded on-board (FDM) to inspect safety occurrences. This activity relies on human experts to create a rule-based system that detects known safety issues, based on whether a small set of parameters exceed some predefined set of thresholds. However, rare events are the hardest to manually detect, a...

Techniques and lessons learned from SafeClouds.eu. Presents features of the decoding algorithm used to import QAR data and how flight data from is analysed for unstable approach scenarios.