High-Performance Computing Probabilistic Fracture Mechanics Implementation for Gas Turbine Rotor Disks On Distributed Architectures Including Graphics Processing Units (GPUS)

Abstract

\n We present an efficient Monte Carlo (MC) based probabilistic fracture mechanics simulation implementation on heterogeneous high-performance (HPC) architectures including CPUs and GPUs for large heavy-duty gas turbine rotor components for the energy sector. A reliable probabilistic risk quantification requires simulating millions to billions of MC samples. We apply a modified Runge-Kutta algorithm to solve numerically the fatigue crack growth for this large number of cracks for varying initial crack sizes, locations, material and service conditions. This compute intensive simulation was demonstrated to perform efficiently and scalable on parallel and distributed architectures with hundreds of CPUs utilizing Message Passing Interface (MPI). In this work, we include GPUs in parallelization strategy. We develop a load distribution scheme to share one or more GPUs on compute nodes distributed over network. We detail technical challenges and strategies in performing the simulations on GPUs efficiently. We show that the key computation of the modified Runge-Kutta integration step speeds up over two orders of magnitude on a typical GPU compared to a single threaded CPU supported by use of GPU textures for efficient interpolation of multi-dimensional tables. We demonstrate weak and strong scaling of our GPU implementation, i.e., that we can efficiently utilize large number of GPUs/CPUs to solve for more MC samples, or reduce the computational turnaround time, respectively. On seven different GPUs spanning four generations, our probabilistic fracture mechanics simulation tool ProbFM achieves speedups ranging from 16.4x to 47.4x compared to single threaded CPU implementation.