Session: 01-05-02: Applications of Artificial Intelligence/Machine Learning for Aerospace Structures

Paper Number: 137441

137441 - Designing Robust Cross-Barrier Communication Using Adaptive Support Vector Machines 

There exist certain aerospace applications where it is desirable to protect sensitive electronics from electromagnetic and/or radiation energy, (for example solar flares or deep space solar radiation) by completely encasing them in a Faraday cage. However, it is still necessary to transmit data between the two sides of the physical barrier without compromising the structure, and hence the shielding effectiveness, of the barrier. One solution to this challenge is to encode the data as mechanical waves transmitted through the barrier itself, thereby using it as the communication channel.

Unfortunately, this solution can be susceptible to new disruptions that may interrupt the data stream, such as mechanical vibrations induced by extreme environments or by adjacent components like pumps, fans, and thrusters. Judicious design of the signal processing scheme used for communication can help to mitigate these risks. However, when the design process involves computationally intensive simulations and a large number of design variables, comprehensively evaluating the entire range of design possibilities becomes prohibitive.

In this work, we cast the problem of data communication disruption as a binary classification problem and apply machine learning. Support Vector Machines (SVMs) are used with a bespoke adaptive scheme to classify the sets of communication schemes in which data disruption (defined as bit errors in the digital stream being transmitted) is or is not expected to occur. The approach is tested on a model of a representative barrier transduction system using standard single-carrier digital communications techniques, such as phase-shift keying and quadrature amplitude modulation, without error correction codes. The adaptive SVM approach is shown to be effective at predicting which communication schemes are robust to disruptive environmental conditions, while only requiring exploration of a small subset of the entire design space.

SNL is managed and operated by NTESS under DOE NNSA contract DE-NA0003525

Presenting Author: Cameron Mccormick Sandia National Laboratories

Presenting Author Biography: Cam earned his Ph.D. in Acoustics from Penn State University, where his research centered around shape optimization of so-called "Acoustic Black Holes". Prior to that, he received an MSc in Sound and Vibration Studies from the University of Southampton, where he investigated the application of active Helmholtz resonators to acoustic metamaterial design. He has been at Sandia National Laboratories since 2022.