Weld Fatigue Life Variational Assessment for Exhaust System Using Monte Carlo Approach

<div class="section abstract"><div class="htmlview paragraph">Parts in automotive exhaust assembly are joined to each other using welding process. When the exhaust is subjected to dynamic loads, most of these weld joints experience high stresses. Hence it should be ensured that the exhaust assembly is designed to meet the requirements of exhaust durability for the estimated life of the vehicle. We also know that all parts used in manufacturing of exhaust system have inherent variations with respect to sheet metal thickness, dimensions and shape. Some parts like flex coupling and isolators have high variations in their stiffness based on their material and manufacturing processes. This all leads to a big challenge to ensure that the exhaust system meets the durability targets on a vehicle manufactured with all these variations.</div><div class="htmlview paragraph">This works aims to evaluate the statistical spread in weld life of an exhaust with respect to inherent variations of its components. For the purpose of variational analysis, a Design of Experiments (DOE) is done where various input parameters are varied between known lower and upper bounds. Bounds for most of the parameters are available as manufacturing tolerances specified in the drawings. Similar is true for stiffness of the bushes and the flex couplings. Some parts may undergo change of dimension during operation due to high operational temperatures. These changes were incorporated using morphing technique on the Finite Element models.</div><div class="htmlview paragraph">A structural solver was used to first evaluate stress under dynamic load. The stress values with material fatigue-life data are then fed into fatigue solver to evaluate life for weld elements. A Multi-Disciplinary Optimization (MDO) framework was used to manage the process seamlessly. Also, indirect Monte Carlo approach was used to take large number of sample points on the response surface method (RSM) to execute prescriptive statistics. This process is very fast compared to direct Monte Carlo approach which may require large number of CAE evaluations. The process developed can be used by exhaust system analysts and designers for design of a robust system with adequate weld life, even when subjected to variations in different parameters.</div></div>