Vehicle CFD Team at VTEC Automotive Fan Noise Modelling using STAR-CCM+ Anders Tenstam Volvo Technology AB Anders Tenstam, Volvo Technology AB STAR European Conference March 22-23 2010 Vehicle CFD
Vehicle CFD Team at VTEC Volvo Technology – a business unit within the Volvo Group AB Volvo Business Areas • Volvo Technology is the center for innovation, research and development in the Volvo Group • Construction Financial The customer base is Renault Trucks Mack Trucks Volvo Trucks Nissan Diesel Buses Volvo Penta Volvo Aero Equipment Services limited, focusing on the Business Units Volvo Group, Volvo Cars, selected suppliers and public Volvo 3P bodies • Secures hard & soft product Volvo Powertrain & process innovation for superior end customer solutions Volvo Parts • Established 1969 • Locations: Gothenburg, Lyon, Greensboro, Volvo Technology Chesapeake, Hagerstown, Los Angeles • Volvo Logistics ~500 employees Volvo Information Technology Anders Tenstam, Volvo Technology AB STAR European Conference March 22-23 2010 Vehicle CFD
Vehicle CFD Team at VTEC Vehicle CFD Team - Main Focus : Support Volvo Group with Methods Development & Resources in: - Aerodynamics - UTM (Heat management & Fan operation) Buses - Aeroacoustics - Climate / HVAC - Internal flows - Fuel Cell Technology - In-cylinder model development Trucks Construction Equipment Anders Tenstam, Volvo Technology AB STAR European Conference March 22-23 2010 Vehicle CFD
Vehicle CFD Team at VTEC Understanding Fan Noise – the Motive: • Fan noise is one of the main contributors to noise pollution from heavy-duty vehicles. • The overall noise level from a heavy-duty vehicle is an entity controlled by legislation. • Given a certain discharge flow, the link between hydraulic efficiency for a fan and emitted noise is quite strong. Hence a quiet fan has also normally low losses. • A large truck fan at high load can consume 30-35kW of power from the crank shaft, a substantial portion of the mechanical power from the engine. Anders Tenstam, Volvo Technology AB STAR European Conference March 22-23 2010 Vehicle CFD
Vehicle CFD Team at VTEC The Project - Goal: • To predict and understand noise sources using CFD simulations. • To investigate the propagation of noise sources to far-field; directivity. • Increase the in-house understanding of fan noise, i.e. what is important in low noise emission fan design? The Project - Facts: • The project was sponsored by the Volvo Group Key Technology Comittee and performed during 2008-2009 • The numerical part of the project was performed by staff from Volvo Technology and Volvo Aero Corporation. • Masurements were made by the NVH lab at Volvo 3P. Anders Tenstam, Volvo Technology AB STAR European Conference March 22-23 2010 Vehicle CFD
Vehicle CFD Team at VTEC The Project - Scope: • To predict the major noise sources in two fan installations of an Articulated hauler: Low-range frequencies from LES High-range broad-band signatures from URANS B. Conventional heavy-duty fan A: In-house developed AH fan at VCE • To predict propagated noise using acoustics analogies • Propose ways to further reduce noise in the installations Anders Tenstam, Volvo Technology AB STAR European Conference March 22-23 2010 Vehicle CFD
Vehicle CFD Team at VTEC The Project – Scope contd. • The computational time frame for each geometry must not exceed 1 week. This way, the developed method may allow for fan noise calculations to be fit into product development chain. • Perform LES simulations of each fan, and extract noise sources and frequency spectrum up to threshold given by maximum model size. • Propagate sources to far-field by a FWH (Ffowcs-Williams Hawkins) integration routine. • Perform URANS simulations to allow for additional broad-band character noise prediction • Analyse and Compare with measured data Anders Tenstam, Volvo Technology AB STAR European Conference March 22-23 2010 Vehicle CFD
Vehicle CFD Team at VTEC The Computational Model Model facts: Region #Cells Fan Region 800 000 Outer Region 1 900 000 Radiator 100 000 TOTAL: 2 800 000 Anders Tenstam, Volvo Technology AB STAR European Conference March 22-23 2010 Vehicle CFD
Vehicle CFD Team at VTEC Results – URANS Simulation (Broadband Sources): Volume Noise Sources (Proudman): Logarithmised values (Isosurface corresponding to 100 dB): AP AP dB 10 log [W/m3] P ref Where 3 5 U U c AP 0 5 L a 0 AP = Acoustic power [W/m3] ε = constant ρ0 = density [kg/m3] Derived from k, ε a0 = speed of sound [m/s] U = Turbulent velocity scale L = Turbulent length scale This equation is implemented in STAR-CCM+ 4.06.011 Similar regions are depicted by adopting the Curle Expression for surface sources Suction side (upstream) Pressure side (downstream) Anders Tenstam, Volvo Technology AB STAR European Conference March 22-23 2010 Vehicle CFD
Vehicle CFD Team at VTEC Results – Broadband Sources contd.: • Broadband sources are strong in the tip region, especially for fan B. This leads to blade-to-blade interaction, which is a strong noise contributor. • For the less noisy fan (A), additional sources exist at mid-radius due to partial separation (improvement potential). • Tip leakage leads also to performance degradation (improvement potential). Anders Tenstam, Volvo Technology AB STAR European Conference March 22-23 2010 Vehicle CFD
Vehicle CFD Team at VTEC Results – LES Simulation (Narrow Band Sources): Near blade surface: max. 2 kHz • Blade wakes: max. 1 kHz Simulations run for 2-3 revolutions for quasi-stationary behaviour • Pressure data is collected for 2 revolutions Maximum theoretically resolved frequency (Nyqvist Frequency): fmax = 1/(2* t ) t = sampling time (s) where Maximum attainable frequency is however limited by the grid spacing (Mesh cut- off Frequency). Derived from URANS simulations as a relation between turbulent velocity scales and length scales Minimum resolved frequency: (and also the frequency resolution to be captured by the Discrete Fourier Transform) is the inverse of the interval length T fmin = f =1/( N* t ) = 1 / T where N = number of samples • For each surface cell, a discrete Fourier Transform is performed, and data is mapped back to the model surface directly or summed up by frequency bands (3rd octave or octave) using ProAm+shell scripts Anders Tenstam, Volvo Technology AB STAR European Conference March 22-23 2010 Vehicle CFD
Vehicle CFD Team at VTEC Results – LES Simulation (Narrow Band Sources): RMS values of pressure series (logarithmised): • Distinctly lower sources are present for the less noisy fan (A) • Sources are mostly concentrated to area near tips, these are linked to the tip vortex. • Stator (shroud) interaction levels are stronger for fan B. • Frequency decomposition reveals strongest contributions around BPF. Anders Tenstam, Volvo Technology AB STAR European Conference March 22-23 2010 Vehicle CFD
Vehicle CFD Team at VTEC Results – LES Simulation (Narrow Band Sources) – Moved Fan: RMS values of pressure series (logarithmised): When the A type fan is moved axially, a substantial increase in surface pressure fluctuations is encountered! Fan A in Nominal position Fan A moved partially out of shroud Anders Tenstam, Volvo Technology AB STAR European Conference March 22-23 2010 Vehicle CFD
Vehicle CFD Team at VTEC Results – LES Simulation (Propagation - FWH): Time series in a listener Directivity (power spectrum), BPF marked with arrow location (on axis) Fan A Fan A Fan A Fan A moved Fan A moved Beside fan: Fan B Fan B High levels at frequencies above BPF Fan A, moved Fan A On axis (+): Fan A Downstream Upstream Fan A moved Fan A moved Fan A: BPF Fan B Fan B dominates. Fan B: 0.5*BPF Fan B dominates on pressure side (fan/shroud interaction)! Anders Tenstam, Volvo Technology AB STAR European Conference March 22-23 2010 Vehicle CFD
Vehicle CFD Team at VTEC Results – Summation over whole observer sphere, A-weighted, third octave band filtered. Experimental results from echo chamber: +20 Fan A moved (PWL) Fan B (PWL) +10 Fan A (PWL) Fan A Measured (SPL) 0 Fan B Measured (SPL) -10 Conclusions: • BPF is well resolved for all fans -20 • Broadband character above BPF is less well represented, the contribution over ~800-1000 Hz is exaggerated (probably caused by the tip vortex and blade-to-blade interaction. Overall, the results are promising. -30 Remarks: • Different results definitions required for simulations (PWL) and -40 experiments (SPL). • Fan A position was not specified to 100% in experiments. -50 • Due to Sliding mesh interpolation routine, simulations were slow. No significant speedup detected above 4 CPU:s on single quadcore machine. A significant speedup should be expected if this is resolved or the model can then be made finer (problem could be parallelized more). Anders Tenstam, Volvo Technology AB STAR European Conference March 22-23 2010 Vehicle CFD
Recommend
More recommend
