Within the department of Distributed Exhaust Systems (NXDC) at Scania CV AB, thermo-mechanical simulations are used for analyzing thermal behaviour and fatigue of silencers. In the design optimization process, design engineers are more interested to look at trends between different design changes where the accuracy is not equally important. Therefore the purpose of this thesis work is to implement the thermo-mechanical simulation in CATIA V5 and adapt the current method to shorten simulation time and implement a method for the design optimization process.

Three investigation fields were explored, called Module A, B and C. Module A contains investigation of material data, steady-state and transient solver and post-processing alternatives in CATIA V5. Module B focuses on mesh sizes, geometry simplifications and solver-settings. Module C contains investigation of plasticity correction models.

Module A showed that the material data was already linearized and that further simplifications of the data might affect results significant and not save much time. A steady-state solver in thermal simulation and transient solver in mechanical simulation gave roughly similar results as the transient method. Steady-state solver in both thermal and mechanical simulation ended up in higher stresses than transient simulation. When investigating a design change of the silencer, trends between the designs were not similar towards the transient simulation. Several other variables were investigated to see if the trends in design changes could be analyzed with a steady-state simulation, but without good results. The problems with post-processing in CATIA was solved.

Module B showed that a coarser mesh for non-relevant parts and finer mesh for relevant parts, were design changes have been made, were useful. The geometry of the silencer could not be simplified in a general way. The total step time within solver-settings can be decreased for the mechanical simulation, together with increased maximum increment size. Number of steps in the simulation could be decreased from four to two and still see critical areas. Plasticity corrections in Module C were investigated but without any useful results.

Useful methods from each module were implemented in design changes to evaluate the accuracy of each module. Combining these ended up in a final method, where two approaches were presented. The first approach is for a first evaluation of concept and takes around 2-3h to simulate locally, CPU with 4 cores. The second approach is for more specific design changes and is a bit more accurate than the first approach and took around 4-13 hours locally depending on which settings are used. This can be compared to the original method that takes roughly around 18 hours with 48 cores.