Hybrid, physics-informed AI methods for real-time prediction of sound fields for active noise cancellation in generators and vehicles
Short description:
The overall objective of this work is to reduce the acoustic signatures of generators and vehicles caused by the operation of internal combustion engines, ventilation and cooling components, and electrical components. A predictive control system for active noise cancellation is being developed that forecasts sound fields in order to precisely control compensation speakers. At its core is a PIML (Physics-Informed Machine Learning) approach that combines a physical understanding of source, path, and radiation mechanisms with learning-based patterns derived from operational and sensor data. This results in robust, real-time predictions and reference signals for targeted noise reduction. The work covers the development of the hybrid AI methodology, the integration of real and virtual sensor data, and validation in operation – with a focus on measurably reduced noise.
Content and time schedule:
- Specification and development of the hybrid AI methodology
- Definition and use of real and virtual sensor technology
- Implementation on a provided control unit
- Data input and customization for demonstrators
- Conducting all necessary preliminary, development, concept, system, and demonstrator tests
- Preparation of the written form
Start:
immediately
Duration:
approx. 3 years
Head of institute:
Univ.-Prof. Dr.-Ing. Christian Trapp, +43 (316) 873-30000, Bitte Javascript aktivieren!
Supervisor:
Dipl.-Ing. Dr.techn. Christian Frühwirth, +43 (316) 873-30037, Bitte Javascript aktivieren!