APPLICATION OF MODEL REDUCTION TECHNIQUES WITHIN CUMMINS INC.

摘要

For the past decade Cummins Inc. have increased the use of standard Finite Element Analysis (FEA) techniques to drive the design of its products. However, as FEA models are not scalable to the limits of hardware, running traditional FEA, especially on large High Horse Power (HHP) engine structures' assemblies, both reliably and within a reasonable time frame was found to be not possible. This led to carrying out numerous analyses with fewer parts and assumed boundary conditions. This strategy ignores effects due to system vibration of the assembly. To reduce the risk of failures on complex assemblies, high speed engines required a more accurate analytical prediction of modal stresses on a system level. To increase the capacity of running system level analyses, a structured approach was followed and the Model Reduction Techniques Functional Excellence mini team was set up to develop methods and train analysts. The team have been using Six Sigma tools to carry out voice of the customer interviews in order to define the analytical requirements for running models for large complex structures (＞20 million degree of freedom). This consists of brainstorming concepts to select solutions based on advanced analytical Substructuring techniques to best fit requirements. The benefits of the new process include a significant reduction in solve time, the ability to carry out system analysis, to follow an efficient working practice using a modular approach, to allow parallel processing globally and secure intellectual property rights when working with suppliers and customers of the Cummins Inc. products. This paper shares experience on applying model reduction techniques following a structured approach and highlights computing and training resources for an analysis team.