Data Acquisition

3:00 PM  – Toolkit Spectrum – An Innovative Solution for Chapter 10 Data Analysis
Alexis Chatail-Rigolleau, Safran Data Systems

RF data analysis in Chapter 10 files traditionally requires replaying them on recorders, even for basic tasks such as assessing the file label or performing frequency spectrum analysis. However, input files can be poorly labeled or may require significant transfer times to the recorder. The Chapter 10 Toolkit addresses these challenges by providing a portable software-based solution for recorded data representation. The Toolkit offers a fast and comprehensive solution to streamlining Chapter 10 file analysis via waterfall spectrum display, packet data and header visualization, channel extraction tools, and specific packet search functions. Additionally, the Toolkit integrates a dynamic sampled representation of the recorded signal and cutting-edge AI-based prediction tool to analyze input data, predict signal waveforms and other characteristics. By leveraging advance algorithms, this predictive tool aids in the visualization, evaluation, and interpretation of test telemetry data.

3:20 PM – Enhancing Flight Safety Training with AI-Generated Telemetry Data for  Mission Readiness
Benjamin Hennessy, Raytheon Australia

Obtaining simulated or pre-recorded telemetry data is often restricted due to classification or proprietary constraints, limiting its use for training, pre-mission workups and flight safety preparation. In many cases operators only encounter real telemetry data at the first live test event. The use of AI driven tools allows for the generation of realistic flight safety values, enabling the production of simulated telemetry data streams. These values can be manipulated to represent a range of realistic flight conditions, providing the necessary complexity and situation evolution seen in high dynamic failure modes. These data can therefore be used to enhance training scenarios, by ensuring that flight safety personnel are prepared for a range of complex and realistic failure conditions. In addition, other AI-based tools have enabled historical paper records (e.g. Apollo 11 and past missions) to be digitized in simulated telemetry IRIG 106 Chapter 4 Pulse Code Modulation (PCM) streams and then utilized for system checkout and technician training.

3:40 PM – Enhancing Structural Health Monitoring Data Acquisition While Ensuring  Compatibility With Existing Systems
Pat Quinn, Curtis Wright; Manuel Jesus Justicia Aldos, Natalia Esteve Ferrer, and Jaime García Alonso, Airbus Defence and Space

Operational Monitoring Structural Health Monitoring (SHM) systems fulfil Fatigue Monitoring purposes of in-service aircraft. The deployment of this technology on Airbus’ military aircraft fleet helps identifying potential issues before they happen. This paper discusses the use case and techniques enhancing the Operational SHM on-board data acquisition providing  Damage Detection capabilities, in particular event diagnosis (i.e., mechanical impacts diagnosis). Specifically, gathering data from accelerometers and Piezoelectric sensors, acquired via high-speed acquisition cards able to maintain compatibility with the existing Operational SHM systems on-board and providing diagnosis information about events that could potentially affect the structural integrity of the aircraft.

4:00 PM – IADS In The Cloud
Michael Jones, Curtiss-Wright Defense Solutions, IADS

This paper will evaluate the use of IADS telemetry processing software in a cloud environment. The evaluation process will utilize the products available via Microsoft Azure services at several data centers across the continental United States. The information presented will include areas of success and areas of concern. The review will reveal the potential benefits of cloud computing for distributed test operations (DTO).

4:20 PM- Unlocking Secure Telemetry Analysis: Generating High-Quality Provisional Labels with Autoencoders and K-Means Clustering
Kenneth Call, USAF

Secure environments often hinder machine learning applications, particularly when labeling sensitive telemetry data is required. This paper presents a case study involving glyph classification, motivated by the desire to extract textual information from classified flight videos. Faced with inadequate commercial OCR solutions and infeasible manual labeling for a custom glyph classifier, we turned to the power of unsupervised learning. By combining autoencoders and K-means clustering, our method learns a low-dimensional latent space from extracted glyphs, where K-means effectively groups similar glyphs to generate highly accurate provisional labels. This approach reduced final labeling effort significantly – from an estimated 200 hours to just 8. While data sensitivity necessitates demonstrating this technique using the MNIST dataset, its core principles and efficiency gains directly translate to secure analyses, enabling AI-driven insights even within constrained data environments.