Minimisation of nox emissions (MINNOX)

A method will be developed for the use of simulated 3D results of heat transfer from a few operating points for the improvement of models and parameter adjustment in 1D simulations in order to provide a fully predictive methodology to simulate NOx emissions for the entire engine map. Referring to result 15340 for environmental, economical and social impacts, This result is to be used for optimizing the engine in certain operation conditions given a certain emission target. Part of the method will be an engine-tuned one-dimensional model for heat transfer, and is a way to transfer CFD results obtained from the method described in result 15340 to a fast one-dimensional optimizing tool. Since CFD computations currently are too slow and costly to be used for this type of optimization, where thousands of simulations are carried out, such a method is needed to predict fuel consumption for a given emission target. Potential problems may, again, be deposit buildup on the piston, which may hamper the predictions. A method of describing such buildup would therefore be desirable.

The CFD code FIRE" is a multi purpose simulation software which can be used for different kind of applications. For the simulation of engines FIRE" provides special modules for the handling of the mesh movement as well as a number of mathematical models for the treatment of effects as fuel injection or combustion. The solution algorithm is based on the well known SIMPLE (Semi IMplicit Pressure Linked Equations) procedure. The code can handle computational cells with an arbitrary topology. Within the actual version of the program new sub-models from the ongoing project will be implemented. These sub-models make it possible to simulate the heat transfer on a higher level of accuracy then it has to be done up to now.

The CFD code KIVA-3 is a computer program for the numerical calculation of transient, two- and three dimensional, chemically reactive fluid flows with sprays. The CFD code STAR-CD is multi- purpose simulation software using unstructured grids, which can be used for different kind of applications. Within the actual version of the program new sub-models from the ongoing project will be implemented. These sub-models make it possible to simulate the heat transfer on a higher level of accuracy than has been done up to now. The relevant heat transfer subroutines are LAWALL(KESOLV) in case of KIVA-3, user-defined subroutine MODSWF, ROUGHW(SORKEP) in case of STAR-CD. Since both KIVA-3V and STAR-CD do not contain a local prescription of wall temperatures a CFD-FEM interface needs to be implemented. A new wall layer meshing strategy is envisaged that allows an adaptation of the mesh size next to wall dependent on local wall temperature and burnt gas temperature after flame arrival. In the standard wall function approach employed both in KIVA-3V and STAR-CD the variable-density affect in the wall boundary layer is left aside. New compressible wall functions have been analyzed and tested during the project. For given one-cylinder engine test cases and defined operating points (including the motored case) the computed heat fluxes have been compared with measured ones at defined sensor points across the cylinder head. The comparison of measured and computed heat flux data in a wide range of operation provides a data base for spark ignition engine heat transfer.

In the past the methods for generating metal temperature predictions for new engine designs was a considerably time consuming part of all CAE investigations. Within the framework of MinNOx a workflow was developed for the efficient and quick generation of 3D CFD meshes for both the coolant flow region as well as the gas side of a combustion engine. In addition the tools which are now available enable the user to create FE meshes for the structural parts automatically. Together with commercial FE codes like Abaqus the thermal analysis of gas water and metal can now be solved simultaneously resulting in more accurate temperature data specially for the combustion walls. This together with the improved turbulence modelling leads to highly accurate NOx predictions and enables the user to optimize e.g. combustion chamber shape, valve timings, intake flow swirl and tumble behaviour with regard to optimized emissions.

The purpose of this working package is to deliver measuring results of the surface temperature of the combustion chamber for 3-D simulations. With these measurements the models for heat transfer in 3-D-calculations could be controlled and calibrated. Exactly knowing of the local heat transfer is necessary to predict NOx emissions within 3D-calculations. An innovative surface temperature method using special surface thermocouples have been applied to the engine for measuring the local instantaneous heat flux. For measuring 121 surface thermocouples had been applied to a special cylinder head. Because of the extremely low mass of the sensors fast measuring, i.e. with a resolution of 1 °CA, was possible. The effect of the flame propagation on the local heat transfer had been intensively analysed under various operating conditions and also under various flow field conditions. This was realized with different intake duct geometry (tumble sheet). It had been necessary to apply an eligible engine to fit the existing cylinder head. The engine and the measuring equipment had build up and the measurements had been carried out. For every engine operating point a burn rate analysis had been calculated. Furthermore, heat flux at every sensor point on the combustion chamber surface had been calculated. The used measurement technology is also usable for temperature measurement at other engine components, for example fuel injectors, to determine the thermal load of such components. The few publications available where a comparable number of measuring points were used are published in the early 1980´s. In this time it was technically not possible to record all of the signals simultaneously. The only chance was to sample the signals, which made it impossible to evaluate single cycles and made an extreme measuring time necessary to collect all the data (several minutes). In this project, for the first time ever, it was possible to record more then 100 sensor positions (121) simultaneously with a resolution of 1°CA.

Simultaneous calculations of coolant flow and metal temperature prediction were performed. The results demonstrated that with an improved method the prediction of heat transfer mechanisms in a combustion engine can be performed faster and much more accurate. In combination with in cylinder flow calculations a more accurate prediction of NOx emissions is enabled. For the analysis work accurate CAD models and FE and CFD meshes for the above mentioned areas are required. The upfront mesh generation and boundary condition definition is performed in the AVL Workflow Manager environment. The simulation results together with measured data from a mass production engine helped to analyze the accuracy of heat tranport phenomena under various operating conditions. Tests and calculations were performed in a way that different physical effects could be separated such as in cylinder heat transfer, coolant pressure, heat transfer to the oil etc. The results will provided a deeper understanding of the accuracy of submodels in current programs such as Fluent as well as those developed within this project by other project partners.

In contrast to Wall Functions approach, integration of governing equations up to the wall has always been regarded as more accurate, but also much more computationally demanding due to the need to resolve near-wall layers with very fine computational grid. This fine grid clustering is the consequence of various damping functions used to account for wall and viscous/conductive effects on turbulence. The concept of elliptic relaxation (Durbin 1991) provides a more physical account of non-local wall-blocking effect on turbulence, flow and heat transfer, and makes it possible to achieve accurate integration with relatively mild grid clustering. We developed an improved version of the k-v2-f model and the new elliptic blending Reynolds stress model, which provide improvements in predicting flow and heat transfer in complex configurations. Both models can be combined with wall functions and thus be used in flows of complex geometry and structure where an a priori control of grid density is difficult to achieve.

An experimental facility was designed and constructed to enable measurements of fluid flow and heat transfer to be made in pulsating and swirling flow conditions. The facility is a stainless steel closed circuit water tunnel that can deliver flow rates up to 290lt/min. Periodic velocity perturbations are introduced to the flow by means of a rotating valve driven by a variable speed motor. The valve periodically blocks the passage of part of the flow generating a low-amplitude sinusoidal velocity perturbation with frequencies in the range of 3-60 Hz. The amplitude of the perturbations "delta u" is varied by means of control valves that adjust the relative amounts of perturbed/unperturbed flow through the rotating valve. The facility, has been used exclusively for MinNOx work and has generated benchmark data for pulsating and swirling flows. It is envisaged that the experimental facility will generate further research activity and will allow these investigations to be extended to other engineering applications. Both flow pulsations and swirl have a great potential as process enhancement mechanisms and it is expected that the present work will be continued and expanded so that our understanding of these unsteady and complex flows is improved while at the same time reliable experimental data are generated for validation and refinement of numerical methodologies for turbulence modelling.

The result will be an extension to the commercially available CFD code STAR-CD and possibly to other in-house CFD codes such as MERMAID, an extremely accurate and fast coupled finite-element CFD solver designed primarily for model evaluation. The result will enable us to compute in-cylinder effects that are generally not possible to accurately compute with CFD codes today, such as transient near-wall behaviour, effects of soot buildup and fuel-jet impingement on the wall, and heat transfer in boundary layers with very strong density gradients. The balance between fuel consumption and emission formation in diesel engines is currently and even more so in the future very delicate. Often very high injection pressures are needed to keep the soot formation at a low level, whereas NOx-reducing measures, such as Exhaust Gas Recirculation will increase soot formation and fuel consumption. Higher injection pressures will, in turn, increase the heat transfer to the piston. This will in itself affect NOx-formation and fuel efficiency. In addition it will make pistons vulnerable to temperature variation fatigue and thermal wear. To take all these effects into account it is very important that the submodel for each process is as valid and accurate as possible, otherwise this delicate balance will not be predicted. The result has the potential to be used to discover the potential of reducing NOx with known measures while at the same time avoid thermal wear on pistons and taking the isolating effects on soot layers on pistons into account. The plan is to use it as an everyday tool to guide in the design of pistons and combustion concepts. The end user of this result is typically an engineer in the field of computational analysis or combustion concept or piston design. The outcome of use of the result is in the end a more economical and environmentally friendly Diesel engine with impact on a vast amount of people. Potential problems using this the method or the result could be the uncertainty of determining the deposit buildup on the piston. A parameter, which, although included in the model, is very hard to guess or predict.

The result is a turbulence submodel that is sufficiently robust for use in complex flow computations within IC engines. Since the modelling close to the wall is crucial for reliable and accurate heat transfer predictions, the model must reproduce the near wall anisotropic behaviour of turbulence. Compared to the standard k-e model (which is the most widely used model for complex flows in IC engines) only one additional partial differential equation for the elliptic blending parameter gamma needs to the solved. This parameter quantifies the wall blocking effect and its solution is very fast, thus the additional computational overhead is minimised. Wall functions for the new variable, which result in coarser grid resolution close to the wall, are also developed. Turbulence anisotropy is accounted for away from the wall as well. The submodel does not use the distance from the wall and thus is applicable to complex geometries.

The high density variation at the wall is taken into account using a correction of the local heat transfer coefficient using a method developed at ERC (Wisconsin). New submodels for velocity-based and turbulent-kinetic energy based determination of heat transfer coefficient are implemented into the KIVA-3 code and as user-defined subroutine into STAR-CD. The high density correction yields higher heat fluxes (by factor 2-2.5) than the conventional constant-density approach. Differences observed between KIVA and STAR-CD are explainable by the different treatment of boundary conditions for turbulent kinetic energy. For both production and university research engines CFD simulations have been performed for a given set of operation points using the available set of models, regarding gas exchange, combustion process and local heat transfer, all these local phenomena considered to be relevant for NOx formation. Comparison with crank-angle resolved measurements of wall heat fluxes confirms the implementation of the ERC-compressibility correction approach. Global integral heat fluxes using Han-Reitz compressibility correction for incompressible wall functions with y+-values in the log-law validity range 10 < y+ < 100 correspond well with experimental data yielding slight over prediction in case of CFD-prediction. Example: O.P.:IMEP= 6bar: Over prediction by 13.6% (2000rpm), resp. 9.2% (3000rpm). A mesh-correction technique has been developed such that the near-wall distance of all cells next to the combustion chamber wall is moved into the range of validity for the application of wall functions supplied by the university partners (Delft, KCL). Heat transfer to components (piston, head, liner) may be properly described based on new dynamic mesh adaptation strategy to avoid y+-values outside the log-law validity range.

A large amount of test data was gathered from a production engine under various operating conditions. The measured data were used to establish a new method to derive metal temperature predictions using a combined method of CFD to describe the heat transfer phenomena from the fluids to the metal structure as well as FEM to calculate the metal temperature distribution. The new method is using the AVL workflow manager together with the CFD package FIRE. A very efficient method was developed to achieve high accuracy results in a reasonable time frame. The new process is to a large part automated thus enabling the user to perform CFD and FEM calculations by far more efficiently. This will lead to optimized engine designs with regard to thermal loading and at the same time improved combustion process.

In order to compute turbulent flows and heat transfer in complex configurations such as IC engines, much is to gain in computational effort and speed if the near-wall molecular layer can be bridged by pre-integrated analytical Wall Functions. Wall Functions make it possible to obtain computational solutions with relatively coarse numerical mesh without having to resolve very thin near-wall molecular layers, which otherwise require very fine mesh density and much larger computational effort. Typically, Wall Functions reduce the number of grid cells by more than 50%. Conventional Wall functions, developed almost 30 years ago, are used in almost all commercial CFD codes, but they are known to be erroneous in most complex flows because of inherent assumptions of turbulence energy equilibrium, which are very rare in practical situations. In this project we have developed new, advanced Wall Functions that are free from any conventional assumptions (.i.e. no log-law velocity distribution), thus accounting for all major physical processes in complex flows, such as pressure gradient, near-wall convection, strong variation of the energy dissipation rate, as well as variation of fluid properties in cases of large difference between the fluid and wall temperature.

A method will be developed for the use of simulated 3D results of heat transfer from a few operating points for the improvement of models and parameter adjustment in 1D simulations in order to provide a fully predictive methodology to simulate NOx emissions for the entire engine map.

The task of Chalmers University of Technology was the creation of an experimental database for a 2.02 l single cylinder D.I. diesel research engine. This database can be used for validation and verification of numerical models in order to improve NOx predictions. Due to the different requirements of the used measurement methods, the experimental work had to be divided into three different parts. Part 1: The research engine was instrumented with a thermo-piston. Two different thermo-pistons (Nr. 1 and Nr. 2) were used during the time of the project. Measurements were made after a stabilization period of around 5-10 minutes and they include: - Measuring of the local, crank angle resolved temperature at the piston surface and cylinder head. - Measuring of the exhaust emissions. - Conventional measurements (cylinder pressure, &), to allow the calculation of the heat loses to the coolant, oil, and exhaust gas. Part 2: The research engine was instrumented with a thermo-piston. An endoscope was inserted into the cylinder head to allow flame imaging. Due to problems with soot deposition on the protection window of the endoscope, the engine was first stabilized under motoring conditions and flame images were taken after a short stabilization time under firing conditions. - Determination of the sooting flame temperature, and the soot KL factor using two-colour pyrometry Part 3 The engine block and the cylinder head had to be changed to allow optical access through the liner and the cylinder head. Measurements were made under motoring conditions and include: - LDA measurements of the flow field in the combustion chamber.

Flow pulsations/oscillations are known to have a great effect on the flow field, turbulence and subsequently heat or mass transfer and can be considered as an active method of heat transfer augmentation. On the other hand, periodic pulsations naturally occur in many biological (eg. blood vessels) and engineering systems (reciprocating pumps, IC engines or pulse combustors). The thermal and fluid behaviour of oscillating flows and their enhancement potential appear to be a subject of controversy in the literature as some investigators report significant heat or mass transfer improvements and others no effect at all. The fundamental experiments of pulsating flows in the context of MinNOx have generated a set of benchmark experimental data that enables the scientific and industrial community to validate and further improve current numerical methodologies for the prediction of complex, unsteady flows and improve understanding of such flows. Similar data have been generated for swirling flows which are also particularly complex and of practical importance to many engineering applications. It is expected that the fundamental experiments carried out for MinNOx will open up new research opportunities for further systematic studies on the effects of flow pulsations/oscillations and swirl on transport processes so that the two technologies can be further explored and applied in process engineering.

Livrables

Partager cette page

Télécharger