Multi-Objective Design Optimization For Performance Improvement Of A Turbocharger Using Arbitrary Shape Deformation

The increasing demand for enhanced turbocharger compressor performance becomes insistently important for the improvement of impeller and volute design.It is desirable that an automobile turbocharger would have high efficiency and wider operation range due to the wide range of engine operating conditions.This,however,is a challenge for designers considering the limitations of aerodynamic and structural design parameters.Over the years,several research works have focused on obtaining high efficiency and boost pressure for centrifugal compressors using design optimization.Geometrical design and parameterization of the impeller blades have mainly been the use of ruled surfaces with Bezier or B-spline curves.Similarly for the volutes,smooth surface generation via Non-uniform Rational B-spline surfaces(NURBS)in the circumferential direction have been coupled with genetic algorithms for optimization.Novel shape parameterization techniques applied to geometries either by mesh morphing or freeform deformation of geometrical surfaces have shown to have unique benefits of shape perturbation of complex geometries whiles maintaining the their topology.The use of lattice connected control points in a control volume provides strong local morphing effect and the direct manipulation of surface smoothness in a three dimensional form.Furthermore,the advent of deep learning and data mining for design space exploration of an optimization task is key to obtaining the best design compromise for all objective functions.It provides a holistic representation not only to design improvements and trade-offs but also new approaches to the conventional design.This becomes even more important when novel design techniques which have indirect relationship between design parameters and objective functions are used.On the contrary,very few research works have employed data mining methods,especially coupled with arbitrary shape deformation(ASD),to study the design space particularly for centrifugal compressors and volutes.This research work seeks to bridge this gap.First and foremost,the original compressor design was tested experimentally and numerical models were used to predict the performance via computational fluid dynamics(CFD)and computational structural mechanics(CSM).Fast response temperature and pressure sensitive paint(TSP/PSP)technique was used to measure the blade surface temperature and pressure distributions at different operating points.Numerical simulations were subsequently validated using the experimental results.The results showed that the blade surface temperatures of the numerical results were higher than the experiment and the distribution turned to increase streamwise for increasing mass flow at maximum speed(N).Larger distribution area was observed for the experiment compared to the numerical results.At different speeds,the impeller blade surface temperature for both numerical and experiments increased with increasing speed.The trend shows an over-prediction of the numerical surface temperatures compared to experimental results.For blade surface pressures,the pressure distribution is similar for compressor speeds of 0.27N to 0.45N compared to experiments which showed an increasing trend.In all,the numerical results were in good agreement with experiments although there were some discrepancies.Secondly,aerodynamic optimization of the turbocharger compressor impeller using arbitrary surface design and genetic algorithm,taking into consideration structural constraints,showed a percentage increase in isentropic efficiency,pressure ratio and work input coefficient of 1.64%,6.34%and 7.5%at peak values for 0.81N.The flow mechanism revealed that at peak efficiency operating point,the increase in efficiency and pressure ratio was attributed to the increase in meridional velocity in the impeller passages.Nevertheless at operating point two,the efficiency of the optimum decreased compared with the baseline design.It was found that the magnitude of the relative flow angle increased for the optimum blade and rapidly built up towards the mid-section of the impeller.This resulted in incidence loss at the active flow zone of the impeller inlet.The combine effect of inlet recirculation and the incidence loss at the leading edge of the optimum impeller increased its passage entropy and consequently reduced the efficiency of the impeller.Thirdly,aerodynamic optimization of the turbocharger volute using arbitrary surface deformation and multi-objective design optimization was performed with the focus of reducing flow induced distortion and increasing operating range.The results showed that the performance of the optimized volute increased in isentropic efficiency and total pressure ratio from medium to low flow rates at design and mid-speed but marginal improvement at low speed for optimum design compared with basic design.Stable operating range significantly increased from 56.3%to 65.5%at maximum speed.Furthermore,changes in static pressure recovery and total pressure loss was mainly due to the increase in radial velocity and decrease in the absolute flow angle at the volute inlet at medium and low flow rates for the optimum compared with baseline.Arbitrary shape deformation parameterization was successful in improving the performance of the volute.Lastly,data mining was employed to gain some insights into design trade-offs and association rules.Similar optimization process as that of the impeller was used.The cluster zones from the self-organizing maps(SOM)for the impeller Design of Experiments(Do E)revealed that the best trade-off between efficiency and pressure ratio can be obtained from zones 2,4 and 5.The neuron cluster zones for the volute showed that high values of static pressure recovery and equivalent low values of total pressure loss were found in zones 4,5 and 8.The zones with the highest static pressure recovery values were 4 and 8.For the engineering variables,the area ratio design variable close to the tongue and area ratio at90~0 circumferential location had significant effect on static pressure recovery and corresponding total pressure loss.For the overhang angles showed a moderate effect on performance objectives.For the arbitrary shape deformation design variables,control points in the z direction of the third control volume was found to be the most influential design variable on static pressure recovery and corresponding total pressure loss.