Search Results

This paper describes a computational and experimental effort to document the detailed flow field around a pickup truck. The major objective was to benchmark several different computational approaches through a series of validation simulations performed at Clemson University (CU) and overseen by those performing the experiments at the GM R&D Center. Consequently, no experimental results were shared until after the simulations were completed. This flow represented an excellent test case for turbulence modeling capabilities developed at CU. Computationally, three different turbulence models were employed. One steady simulation used the realizable k-ε model. The second approach was an unsteady RANS simulation, which included a turbulence closure model developed in-house. This simulation captured the unsteady shear layer rollup and breakdown over the front of the hood that was expected and seen in the experiments but unattainable with other off-the-shelf turbulence models.

This paper is a review of turbulence models and computational methods that have been produced at Clemson University's Advanced Computational Research Laboratory. The goal of the turbulence model development has been to create physics-based models that are economically feasible and can be used in a competitive environment, where turnaround time is a critical factor. Given this goal, all of the work has been focused on Reynolds-Averaged Navier-Stokes (RANS) simulations in the eddy-viscosity framework with the majority of the turbulence models having three transport equations in addition to mass, momentum, and energy. Several areas have been targeted for improvement in turbulence modeling for complex flows such as those found in motorsports aerodynamics: the effects of streamline curvature and rotation on the turbulence field, laminar-turbulent transition, and separated shear layer rollup and breakdown.

Part IV of this five-part paper provides an example case study using the recently developed robust CFD methodology and procedures presented in Part I. The first of the four classes of validation cases, documented in Part I, were analyzed here for the flow mechanisms responsible for total pressure losses in the entire intake system, including intake port, valve clearance, combustion chamber, and cylinder regions. Despite having over 5 million finite volumes, all grid meshes showed high quality, as signified by very low maximum and average skewness values of 0.76 and 0.32, respectively. Second order discretization scheme, unusually strict convergence criteria, and carefully enforced “grid independence” for all solutions were employed. To identify the physics of the flow through the intake valve region, four simulations corresponding to high valve lift (HL), medium valve lift (ML), low valve lift (LL) and a case in which the valve was removed, were completed.

Part I of this five-part paper presents a robust and comprehensive computational methodology, developed and validated for physically realistic and economically feasible results in racing. The methodology was applied, in Parts II-V, for the flow predictions and optimization in the entire intake cowl, complete intake manifold and the intake and exhaust valve regions of a V8 racecar engine. The validations had been performed through “blind” experimental tests for four intake and three exhaust valve cases, revealing consistently predicted flow rates within an acceptable error band. Additional validation was obtained on the Talladega racetrack where recommendations based on CFD predictions of intake cowl modifications lead to a significant time reduction of 0.286 seconds per lap. An original method to obtain a detailed electronic description of complex geometry from actual prototype hardware was developed.

Part V of the present five-part paper focuses on a research project designed to uncover new and innovative means of increasing airflow to the engine within NASCAR rules. Computational Fluid Dynamics (CFD) offers an alternative to the current “build-and-bust” technique reducing costs and time per design iteration, and provides the sponsor with a physics-based design tool with true predictive capability. Armed with a validated CFD based design system, the team could respond quickly to rule changes by analyzing new configurations through simulations avoiding the fabrication and track testing that is currently necessary. A robust and easy-to-use CFD methodology for this class of problems was developed and implemented to understand the flow physics and explore novel configurations, as described in Part I. Specifically, these problems involve external flow around a racecar traveling at 180 mph.

Part II of this five-part paper focuses on the flow field in the manifold of a V8 racecar engine with the use of the recently developed comprehensive, robust methodology presented in Part I. An exact electronic description of the computational domain for manifold was obtained using the methods described in Part I. Manifold flow was simulated for open and restricted engine configurations and for two unique pair of active runners, including cylinder pairs 1-8 and 3-4. Despite having over 11 million finite volumes, all grids are high quality, with maximum skewness of only 0.74. A second order discretization scheme was used along with unusually strict convergence criteria to obtain fully converged and grid independent solutions in all the cases presented here. The port entrance regions and the dividing walls between the paired runners are primarily responsible for the flow recirculation in the plenum chamber and for the flow separation inside the active runners.

Part V of this five-part paper investigates the flow field and the total pressure loss mechanisms for three valve lifts in the exhaust region of a V8 racecar engine using the robust, systematic computational methodology described in Part I. The replica of the engine geometry includes a cylinder, detailed combustion chamber, exhaust valve, valve seat, port, and “exhaust pipe”. A set of fully-converged and grid-independent solutions for the steady, time-averaged (or RANS), non-linear Navier-Stokes equations are obtained using dense and high quality grids, involving 2.1∼3.0 finite volumes, and unusually strict convergence criteria. Turbulence closure is attained via the realizable k-ε (RKE) model used in conjunction with the non-equilibrium wall function near-wall treatment. The validation presented in Part I showed that flow rate results from the “blind simulations” agree well with the experimental measurements.

Using a previously validated and documented CFD methodology, this research simulated the flow field in the intake region (inlet duct, plenum, ports, valves, and cylinder) involving the four cylinders (#1, #3, #4, #6) of a straight-six IC engine. Each cylinder was studied with its intake valves set at high, medium and low valve lifts. All twelve viscous 3-D turbulent flow simulation models had high density, high quality computational grids and complete domains. Extremely fine grid density were applied for every simulation up to 1,000,000 finite volume cells. Results for all the cases presented here were declared “fully converged” and “grid independent”. The relative magnitude of total pressure losses in the entire intake region and loss mechanisms were documented here. It was found that the total pressure losses were caused by a number of flow mechanisms.

Very large scale, 3D, viscous, turbulent flow simulations, involving 840,000 finite volume cells and the complete form of the time-averaged Navier-Stokes equations, were conducted to study the mechanisms responsible for total pressure losses in the entire intake system (inlet duct, plenum, ports, valves, and cylinder) of a straight-six diesel engine. A unique feature of this paper is the inclusion of physical mechanisms responsible for cylinder-to-cylinder variation of flows between different cylinders, namely, the end-cylinder (#1) and the middle cylinder (#3) that is in-line with the inlet duct. Present results are compared with cylinder #2 simulations documented in a recent paper by the Clemson group, Taylor, et al. (1997). A validated comprehensive computational methodology was used to generate grid independent and fully convergent results.

Computational fluid dynamics methods are applied to the intake regions of a diesel engine in the design stage at Caterpillar. Using a complete, tested and validated computational methodology, fully viscous 3-D turbulent flow simulations are performed for three valve lifts, with the goal of identifying and understanding the physics underlying loss in the intake regions of IC engines. The results of these simulations lead to several design improvements in the intake region. These improvements are made to the computational domain, and flow simulations are again performed at three different valve lifts. Improvements in overall total pressure loss of between 5% and 33% are found in the computed results between the original and modified (improved) domains. Physical mechanisms responsible for these improvements are documented in detail.

A computational methodology has been developed for loss prediction in intake regions of internal combustion engines. The methodology consists of a hierarchy of four major tasks: (1) proper computational modeling of flow physics; (2) exact geometry and high quality and generation; (3) discretization schemes for low numerical viscosity; and (4) higher order turbulence modeling. Only when these four tasks are dealt with properly will a computational simulation yield consistently accurate results. This methodology, which is has been successfully tested and validated against benchmark quality data for a wide variety of complex 2-D and 3-D laminar and turbulent flow situations, is applied here to a loss prediction problem from industry. Total pressure losses in the intake region (inlet duct, manifold, plenum, ports, valves, and cylinder) of a Caterpillar diesel engine are predicted computationally and compared to experimental data.