Single Plane Manifold

A new year and a new set of results...

As with the dual plane manifold, I have created a CAD model that lets me simulate and analyse the air flow down individual runners of the single plane manifold:

The model was run under the same conditions as the dual plane manifold i.e. each inlet valve is opened in sequence and a pressure drop applied to approximate the piston velocity. The simulation is then interrogated for the mass flow into each cylinder. The mass flow variation against the average for each cylinder is shown below:

The average mass flow was 18% higher in the single plane manifold when compared to the dual plane design indicating that the engine would be able to breathe more easily and hence produce more power with this type of manifold (unfortunately it does not represent an 18% gain in power due to the complex flow conditions present when the engine is drawing air in sequence with the firing order, not the simple steady state conditions of the simulation). However, it can be seen that the cylinder-to-cylinder variation is still in the order of the dual plane design (-20% / +10% variation) indicating that this manifold is also not ideal for a port fuel injection application. Either manifold is suited to a carburettor or throttle body injection since the air-fuel ratio is established at the throttle plate rather than the valve.

Given the above information and a distinct lack of a readily available stock port fuel injected manifolds on this side of the pond (I didn't like the look of it much anyway!), I have decided to look at the possibility of fabricating my own. The CAD work I have done to date suggests a twin-plenum design would be the best for packaging around the known hard points. Watch this space...

Dual plane manifold

In parallel to the manufacture and testing of the Megasquirt controller, I have been working on the CFD analysis of options for the intake manifold. Having established my cylinder head model had at least a passing resemblance to reality in this post, I have moved on to creating a simple engine model comprising the manifold, two cylinder heads and outlet pipes representing the cylinders:

This model was then run 8 times, each with a different inlet valve being open. These were run with a constant pressure drop representing a high RPM running condition and maximum valve lift. This is the condition where runner mass flow variation will be most obvious. Once the simulations were complete, I interrogated the results and normalised them against the average mass flow for all 8 cylinders:

The chart indicates that the variation in inlet runner geometry produces quite a variation in mass flow into the different cylinders (from -11% to +15% from the average). This is not an issue with a carburettor providing the fuel because the mixing is done at the throttle plate so each cylinder gets an appropriate air fuel ratio. However, the fuel injection system will work on the total mass air flow being split evenly between the cylinders and fuel accordingly resulting in overly rich and overly lean cylinders. This is not conducive to a happy engine, so an alternative manifold is required. Onto the single plane design then...

CFD Model Calibration

In order to have any confidence in the CFD analysis I will be undertaking, I need to check that the model has some bearing on reality. I have the flow bench data for the cylinder head I am intending to use so I want to know how well my CAD model matches up with the empirical information. To achieve this, I have created a 3D model of the cylinder head and set it up in a virtual flow bench i.e. I added a radiused inlet port and a pipe to represent the cylinder.

Then it was a simple task of adding the pressure drop according to the real flow numbers I have and running the simulation for a number of different valve lift values. The results are as follows:

As you can see, the predicted flow numbers correlate reasonably well with the measured values. There is around 7% difference in the region of interest, which considering the CAD model has been cobbled together from photos and limited dimensional data available on the web, I think is none too shabby. I will only be using the model to compare different options rather than looking at absolute values, so I believe this is a reasonable starting point.