Convergent Science

Knocking Prediction

Abnormal combustion is an important issue to consider in the design of spark ignited engines. Knocking is one type of abnormal combustion involving the spontaneous ignition of the end gases ahead of the flame and the subsequent transmission of pressure waves throughout the combustion chamber.  Knocking can cause engine damage, result in significant noise, and negatively affect performance.  Knocking can occur in any spark ignited engine, yet can be an especially prevalent in modern down-sized boosted engines (Heywood).

 

Modeling Turbulent Flame Speed

CONVERGE™ CFD software is ideally suited to model IC engine spray, turbulence, combustion, and emission formation for cases with and without knocking.  For all spark ignited cases, one key element is to model the propagation of the turbulent flame (deflagration). CONVERGE™ currently has two methods for modeling this, allowing the user to choose the accuracy/speed tradeoff they desire:

 

Flame modeling approach #1: G-equation model

The G-equation model uses empirical relationships to specify the turbulent flame speed based upon various factors which might include species concentrations, turbulence levels and temperature.  The beauty of the G-equation approach for turbulent flame tracking is that it runs very quickly.  The downsides are that it often shows grid sensitivity and due to its empirical nature, frequently requires model tuning.

 

Flame modeling approach #2: Detailed Chemistry

The second approach CONVERGE™ offers to model turbulent premixed flames utilizes the SAGE detailed chemistry solver which is an integrated standard feature of CONVERGE™.  With SAGE, the chemical kinetics in every cell can be considered to determine an overall reaction rate. Convergent Science continues to invest heavily in making SAGE run as fast as possible. Two recent developments include the multi-zone enhancement to SAGE (developed in collaboration with Lawrence Livermore National Lab) and the dynamic mechanism reduction.  The computational overhead associated with detailed chemistry has been greatly reduced with these enhancements to SAGE.

 

Unfortunately, solving for the detailed chemistry will not predict the flame speed accurately unless the mesh is sufficiently fine to resolve the turbulent flame thickness (usually on the order of 1 mm for a typical gasoline engine).   Resolving the entire domain with this resolution is unrealistic as the mesh count would be prohibitive.  A much better approach is to utilize the Adaptive Mesh Refinement (AMR) capability in CONVERGE™ which enhances the mesh locally in areas of large gradients in field variables such as temperature and velocity. 

 

Therefore, SAGE and AMR can be used in conjunction to calculate the turbulent flame speed directly from first principles (heat release and diffusion) without ever having to specify an empirical flame speed correlation.  This approach automatically adjusts the flame speed as conditions (equivalence ratio, rpm, temperature, etc) in the engine change.  This approach is also grid convergent in that as the mesh is refined, the solution converges to a specific result.  The mesh insensitivity and lack of tuning of this approach makes it very accurate.  However, the run times associated with the detailed chemistry/AMR approach will likely be longer than that of the G-equation.  With that said, as SAGE continues to be sped up, the difference in run times between the two approaches will diminish.

 

Knocking1

Figure 1: Contours of temperature showing the flame front, as predicted by detailed chemistry, for a spark ignited engine. Of note is Adaptive Mesh Refinement (AMR) resolving the turbulent flame thickness.

 

 

Regardless of which flame speed approach is used (G-equation or detailed chemistry), SAGE can be activated ahead of the flame front to predict the end-gas autoignition via low temperature kinetics.  

A representative case is shown below, showing the flame front development of a spark ignited engine as computed directly by detailed chemistry.  The locations of auto-ignition (as predicted by SAGE in the end-gas) are shown in light (weak knock) and dark (strong knock) blue.   

 

 

Video showing the turbulent flame front (red) and locations of auto-ignition (blue) for a spark ignited engine as modeled by CONVERGE™ CFD software. Courtesy Chrysler Group LLC. 

 

 

Natural Gas Engine Autoignition

This exact approach can also be used to predict the auto-ignition behavior for a high BMEP, medium bore, natural gas engine as seen in Figure 2:

 

Knocking3

Figure 2: Flame propagation with multiple autoignition events (flame is a red isosurface, H2O2 is a black isosurface). Courtesy Prometheus Applied Technologies.

 

 

Knocking Index

When knocking is expected, monitor points can be placed throughout the CONVERGE™ domain to study any pressure fluctuations which might occur as a result of the rapid release of energy due to autoignition of the end gases. Representative monitor points and pressure signal for a gasoline engine are shown in Figure 3:

 

Knocking4

Figure 3: Point pressure minus average cylinder pressure (both actual and FFT filtered) at a monitoring point for a spark ignited engine operating at 3000 rpm

 

 With this data, the knocking index can be readily calculated and the sensitivity to spark advance can be determined. A typical result of the knocking index is shown in Figure 4:

 

Knocking5

Figure 4: Knock index versus spark advance at monitor points (as well as the average value) for a spark ignited engine operating at 3000 rpm

 

 

Lube Oil Abnormal Combustion

It should be noted that abnormal combustion can also be caused by the auto-ignition of lubricating oil in the cylinder of higher BMEP natural gas engines (Yasueda, et al).  CONVERGE™ can readily simulate such phenomena by utilizing the SAGE chemistry solver as shown in Figure 5:

 

Knocking6 Knocking7

Figure 5: Lube oil autoignition locations (left) and pressure signals for a high BMEP natural gas engine. Courtesy Prometheus Applied Technologies.

 

Summary

In summary, CONVERGE™ is the code of choice to model any internal combustion engine, with or without knocking.   In addition to eliminating all user meshing time, CONVERGE™ offers a powerful suite of models for spray, turbulence, combustion and emission formation to handle any engine type.  For spark ignited engines, CONVERGE™ offers the user options for tracking the flame front, depending upon what level of accuracy and run time they desire.  The SAGE chemistry solver can be used to determine both the flame location as well as the autoignition of the end gases and lube oil droplets. 

Use CONVERGE™ and never make a mesh again.

 

Bibliography:

A Method for Predicting Knock in Gas Engines by Meas of Chemical Precursors from Detailed Chemistry CFD, Emmanuella Sotiropoulou, Jessica Harral, Dr. Luigi Tozzi, Prometheus Applied Technologies, LLC, Fort Collins, CO, USA.

Predicting Autoignition Caused by Lubricating Oil in Gas Engines, Shinji Yasueda, Ph.D., GDEC, Japan. Emmanuella Sotiropoulou, M.S., Prometheus Applied Technologies, LLC, USA. Luigi Tozzi, Ph.D., Prometheus Applied Technologies,LLC, USA, CIMAC Congress 2013, Shanghai.

Heywood, J. B. Internal Combustion Engine Fundamentals. London: McGraw-Hill, 1988.

Share This Page