Enhancing Computational Efficiency of PSU-WOPWOP through MPI based Parallelization
Open Access
- Author:
- Daher, Abdallah
- Graduate Program:
- Aerospace Engineering
- Degree:
- Master of Science
- Document Type:
- Master Thesis
- Date of Defense:
- March 28, 2025
- Committee Members:
- Kenneth Steven Brentner, Thesis Advisor/Co-Advisor
Eric Greenwood, Committee Member
Amy Pritchett, Program Head/Chair
Alan Richard Wagner, Committee Member - Keywords:
- Aeroacoustics
MPI
PSU-WOPWOP
Parallelization
Noise Prediction - Abstract:
- As rotorcraft designs, particularly electric Vertical Take-Off and Landing (eVTOL) aircraft, become more complex, the need for accurate and efficient noise prediction has become a critical challenge. Aeroacoustic simulations, such as those performed using the PSU-WOPWOP code, require significant computational resources due to the high-fidelity models involved in predicting rotorcraft noise. Traditional methods relying on serial computation are no longer sufficient to handle the increasing complexity and scale of these simulations. This thesis focuses on enhancing the computational efficiency of PSU-WOPWOP through the development and implementation of a time-parallel algorithm that is based on the unique physics of acoustic propagation. Time parallelization, the core contribution of this work, addresses the inefficiencies in long-duration noise simulations by dividing the total simulation time into smaller segments. These segments are processed concurrently on separate processors, leading to substantial reductions in overall computation time. Unlike source and observer parallelization, which are limited by the number of independent noise sources or observers, time parallelization offers a scalable solution that can be applied to any simulation involving aperiodic flight conditions and complex rotor interactions. Performance analysis of the algorithm demonstrates significant improvements in both speedup and efficiency, particularly for long simulation times. However, the study also reveals challenges related to communication overhead and synchronization between processors, which limit scalability as the number of processors increases. Despite these limitations, the time-parallel algorithm achieves remarkable performance gains, reducing computation time without sacrificing accuracy.