High performance computing simulations to identify process parameter designs for profitable titanium machining
Academic Article
Overview
Research
Identity
Additional Document Info
Other
View All
Overview
abstract
2017 The Society of Manufacturing Engineers The objective of this paper was to conduct a study of the multi-level multi-variable design space in titanium machining through high performance computing (HPC) simulations that was otherwise too vast to be explored by physical experiments alone. For tool wear-based performance metrics, this resulted in a validated set of machining parameters for achieving profitable material removal rates (MRR) (optimized cost of processing and tooling) across multiple operational configurations, alloys, tool geometries, and process conditions. The approach was to include all machining-related variables and their distributions available within the software as inputs to finite-element models (FEM) of the machining process. The time intensiveness of conducting such large numbers of lengthy simulations was handled by wrapping Third Wave Systems AdvantEdge FEM machining simulation software with Dassault Systemes iSight to automate the building of experimental designs and their parallel execution on a HPC cluster. Results were analyzed using SimaFore software to identify key characteristics through bivariate analyses. A subset of simulations was validated through physical experiments, and these were in turn used to augment physically untested regions in the design space. Based on this, an MRR-based cost model for orthogonal turning was derived to drive optimal machining setups. This study showed the feasibility of integrating and automating a HPC loop involving the generation of suitable design of experiments (DOE), creating simulation jobs, deploying/executing it on a HPC cluster, and scripting outputs in a useful format. Besides highlighting the challenges in reading/transferring data across different software and in handling/compiling large amounts of data, this study also shed light on the need for benchmarking processoroperating systemsoftware combinations for computational efficiency.
