Model approach keeps cars quiet

Reducing noise from car exhaust manifolds, or improving the efficiencies of turbochargers and their associated ducts, are complex engineering problems. They require the use of software with multiple capabilities, plus a significant amount of human intelligence. Modelling exhaust manifold impingement noise is certainly a challenging exercise, as Hui Zhao of Jaguar Cars confirmed at the recent Altair CAE Technology Conference in Gaydon. “It requires an FSI [Fluid Structure Interaction] simulation -- and the mechanism behind this phenomenon is not completely understood,” he said. Of particular concern to luxury car makers such as Jaguar, the problem is described as an impulsive noise in the range 2200 to 4500 Hz, generated under certain speed and load conditions and thought to result from exhaust gases impinging on the manifold walls. It can be reduced by changing the mass, stiffness and damping properties of the manifold, or the flow structure of the exhaust gas. This requires modelling of both the fluid flow and the mechanical structure of the manifold – as well as the interaction between the two. Zhao used Radioss CFD, which has both an FSI and a CAA (Computational Acoustic Analysis) capability, and constructed a hexahedral mesh for the fluid path and a shell mesh for the structure. At the inlets, boundary conditions were derived from a 1D engine simulation code, Wave. Outlet conditions required the addition of ‘beam elements’ representing down pipes, after-treatment devices and silencers. Once the flow and interaction were solved, the surface velocity at the structure could be processed, using Rad2Noise software to calculate the sound pressure level at predefined locations. This could then be correlated with real-world sound measurements in a semi-anechoic chamber. Zhao reported up to 10dB differences between simulated and measured sound levels, which he thought represented a good correlation. Academics at the event felt that a 10dB difference was not really satisfactory, and thought the discrepancies were more likely to have arisen from not taking sound measurements in a proper anechoic chamber, rather than from errors in the software modelling itself. In the end, Zhao found that noise would be significantly reduced simply by making parts of the manifold wall slightly thicker. At the same event, Warren Seeley, of Ford Dunton Research and Engineering Centre, described an even more challenging piece of work to optimise the compressor inlet duct of a turbocharger. This is a piece of pipe with two bends in it. It runs between the air filter and the turbocharger, with a cross section that has to change along its length for reasons that would appear to arise from the shape of the rest of the engine. “We were looking for a quicker method that was more robust that we could apply to a real engine system,” said Seeley. “Get it wrong and the turbocharger runs inefficiently. I have even seen the blades break.” The solution was to take an initial CAD design, mesh it using HyperMesh, analyse it via Star-CD, parameterise its key features with HyperMorph and then shape optimise it by means of HyperStudy. The exact methodology is very complex, but the outcome, according to a written paper distributed at the conference, is an increase in the amount of flow rotation immediately upstream of the impeller eye of 330%. “We are now going to apply it to water jackets and intake manifolds,” added Seeley. Pointers * Software exists to enable the analysis of complex interactions between fluids and solids, and the consequent generation of noise * It is also possible to optimise fluid-carrying components to best achieve design criteria in a systematic way, without having to undertake full, repeated simulations