A simplified numerical method to predict the structural limit of a diesel engine part through linear FEA

Many deterministic and fuzzy methods have been developed to optimise parallel computer performances through a rational meshing partition in order to balance working loads. This paper presents a method to examine simultaneous mechanical and thermal load application through a simplified use of Finite Element Analysis (FEM). An important aspect in determining a correctly-distributed stiffness matrix is the subdomain shape. Graph Theory can be a powerful tool to help solve problems of mesh rationalization. The application of thermal and mechanical loads and the consequent problem solution can be very complex, especially in the thermal field. In the field, engine temperature is nearly constant during a single-working cycle from engine start to stop, while pressure loads are variable with time. In this thermoelastic analysis, the temperature is commonly assumed to be constant with time: with non-linear behaviour of materials, the consequent stress can be evaluated by laborious resolution of strongly non-linear systems. The variational calculus is generally solvable with numerical methods such as that from Galiorkin and Ritz. Through these methods it is possible to subdivide each component into different volumes at nearly constant temperature. Then the non-linearity problems can be overridden and a linear equation system can be used with no relevant errors. The different volumes must be considered joined to each other through congruence constraints. The aim of this paper is tp optimize the mesh partition through graph theory coherent with the method of Galiorkin and Ritz, when the existence of a minimum computer-time function has been verified. This analysis is experimentally supported by tests on thermally-stressed components performed in the laboratories of the II Engineering Faculty in Forlì.