Future of engine ECUs & control strategies

[SomptingGuy]

I'm looking for opinions.

I mentioned in another thread that I think future engine control strategies are likely to be based on cycle-by-cycle individual in-cylinder conditions (measured or predicted).

This will require cheap and robust sensors, and/or sophisticated hardware/software.

Any thoughts or comments?

- Steve

 

[JSteve2]

This is absolutely coming. The current legislative emissions cycle peaks in 2012 or so, and what you are talking about is the generation after that as we begin to fine-tune secondary emissions that are not being measured or legislated today, and also try to recover power and fuel efficiency that have been degrading due to meeting emissions requirements. Many of these technologies are well into the pipeline, and some modeling of individual cylinder conditions is already in use commercially.

 

[Bribyk]

I've seen some ignition systems that monitor cylinder conditions by the capacitance of the plugs. I don't recall if the spark plugs were special (read: expensive) or if the soft/hardware that was making the measurements was just uber-sophisticated. I think I saw this is Racecar Engineering a couple of years ago...

 

[patprimmer]

I expect (from intuition only) that will be an area where increased efficiency will soon be exploited.

I know one argument for plastic manifolds was that they could have more consistent cylinder to cylinder air flow and therefore reduce the need for individual cylinder fuel trim.

Regards

eng-tips, by professional engineers for professional engineers
Please see FAQ731-376 for tips on how to make the best use of Eng-Tips Fora.

 

[MikeHalloran]

A long time ago I read some SAE papers and patents covering 'The Optimizer', a device that adjusted distributor timing (mechanically; it was a long time ago) in response to a crank velocity signal pulled off the flywheel starting gear. The patents were assigned to Harmon Electronics of Grain Valley, Missouri, which last time I checked was still in business making railroad electronics.

What was surprising, then, was the amount of useful cycle by cycle information that could be gleaned from a single sensor.

I'm pretty sure the patents have run out.

I'd be very surprised if current ECUs >didn't< have a little code in them based on the core technology... which amounted to diddling something at random just a tiny bit, measuring the engine's response over the next few cycles, and biasing the controller's setpoint in whatever direction produced an improvement.



Mike Halloran
Pembroke Pines, FL, USA

 

[GregLocock]

I think Torsional Vibrations are the most likely approach, as all the instrumenation is already on the engine. Other cheap ways of looking at each cylinder would include microphones in the exhaust header.

However as an example of how difficult it is to analyse performance from that data look at this amateurish webpage

http://members.optusnet.com.au/greglocock/cylinderpressure.htm

I don't see why a high pressure pressure transducer in mass production necessarily has to be expensive, so true in-cylinder measurement may be feasible. However they are a bit tricky to package.





Cheers

Greg Locock

SIGlease see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[mattsooty]

I don’t see any real world gain from having a measurement/prediction of the in-cylinder conditions of an engine.

One problem with any cycle-by-cycle analysis of an IC engine (especially gasoline) is the combustion process itself, which is inherently chaotic and has very high variability. This variability is such that some form of averaging must be performed in order that a statistical meaningful value can be calculated, even during steady state running. Since most IC engines spend their time operating transiently the signal to noise ratio of the in cylinder measurements is going to be far too high to be useful.

I have also heard the notion of 'adaptive spark advance' (not knock control) whereby the spark is advanced until just on the borderline of knock (often, but not always, the most efficient setting). This sounds good but since there is no torque feedback it cannot be guaranteed that this is the most efficient setting. In addition to that the fact is NOx emissions are always going to be affected would make a compliant emissions calibration impossible to achieve.

Having said that, there do already exist systems, such as adaptive knock control & misfire detection, which operate on a cylinder specific level. These use knock sensors & the acceleration of the flywheel, respectively, with good results. Some EMS systems also have an idle speed stabilisation control but, once again, crank accelerations are also perfectly adequate for this. I suppose that in cylinder monitoring could be used in these functions, within a feedback system, but I am sceptical that there is a requirement. After all, surely, in using in cylinder measurements you are just swapping one feedback signal (i.e. crank accels.) for another (i.e. in cylinder pressure) with no gain?

In terms of emissions control, the main factor is the cold start, not individual cylinder conditions. Once catalyst light off has been achieved tail pipe emissions can be reduced to almost negligible levels. It is the period before light off in which most emissions tests are passed or failed. Especially when there is a prolonged period of open loop fuel control and high feed gas emissions.

After light off the next major emissions contributor tends to be transient fuelling control and catalyst O2 clearing. Both of which are greatly aided by GDI technologies, which greatly reduced catalyst breakthrough when properly calibrated.

I can’t see how any of the above would be greatly helped by in cylinder measurements.

MS

 

[JSteve2]

MS,
The spirit of much of what you are saying is true today, but it will not be true in just a few years. For example, "Once catalyst light off has been achieved tail pipe emissions can be reduced to almost negligible levels" is roughly true today - although probably only for SI engines - but it will not be true by the 2012 emissions cycle when the targets are so low that a bit of off-nominal operation in a cylinder for just a few minutes can cause an emissions failure regardless of the catalyst performance.

Also - when you say "After all, surely, in using in cylinder measurements you are just swapping one feedback signal (i.e. crank accels.) for another (i.e. in cylinder pressure) with no gain?" I'm with you until those last three words. There are things you cannot detect with the crankshaft - temperature & composition being two big ones - that can at least in theory be detected in-cylinder or near-cylinder. Further, finding things out from the crankshaft is more complicated (algorithmically, not environmentally of course) than an in-cylinder detection. You have multiple cylinder events occurring on a flexible crankshaft with harmonics bouncing through it, and further a transmission and a flywheel tugging on it. We can sort through all that to detect complete misfires, but any better resolution than that is suspect.

 

[BrianPetersen]

Upcoming diesel engines (example, 2009 VW Jetta TDI common-rail, and it's not the only one) will be using cylinder pressure sensors. It's the only practical way that the ECU can "measure" the ignition delay with any degree of accuracy. On these upcoming engines, they are using lower compression pressure for a variety of reasons (mostly emissions but partly for performance by allowing more boost pressure), and that reduces the "window of opportunity" for ignition.

On engines that propose to use some variation of HCCI during some portion of the speed/load regime (for emissions reasons), this type of feedback will probably be necessary to fine-tune injection timing for reliable ignition at the right time.

 

[Thecardoc]

Individual cylinder balancing is not new, especially when it comes to ignition timing. Chrysler accomplished this feat back in the eighties, and had spark timing maps for individual cylinder control that we could see in scan data. Just imagine if cylinder #3 was carboned up a little more than the rest, and tended to detonate more. The timing on just that one cylinder would be slightly retarded and correct for the pinging. Engineers also use timing control as the first step in RPM control. If the engine slows slightly, instead of opeing the IAC right away, a couple degrees of timing advance will usually speed the engine up 5-25rpm. A touch too fast now, and retarding the timing will slow it.

We haven't even disscussed high resolution crankshaft sensor's and faster PCM's which first need to sample the crankshaft waveform for a reference, and then once learned, can use small variances to that waveform in order to sense engine missfires. This system even provides feedback to be sure that variable displacement has occured correctly, and works with variable engine mount dampening, and active noise cancellation so that the driver doesn't notice the change in the way the engnie is running.

While pressure transduecers would be an interesting addition, you have the problem that by increasing the number of parts, you increase the potential number of failures. Plus you also have to deal with the added weight that has to be hauled around for the lifetime of the vehicle.

 

[patprimmer]

Even the humble VW Beetle had a 2 deg retarded cylinder to compensate for the said cylinder running hotter because of oil cooler placement. This was obtained by the indexing of the lobes on the breaker cam in the distributor.


Regards

eng-tips, by professional engineers for professional engineers
Please see FAQ731-376 for tips on how to make the best use of Eng-Tips Fora.

 

[Lou Scannon]

Another example of German thoroughness, I guess!

 

[dcasto]

From the web site of Caterpillar:

Caterpillar ADEM III control system provides electronic governing integrated with air/fuel ratio control and individual cylinder ignition timing control.

Custom control system software is available for non-standard ratings. Software is field programmable using flash memory.

 

[ivymike]

yeah, but that's for a much more expensive (and larger) engine than I thought we were talking about here, even though the original post didn't specify.

 

[SomptingGuy]

The OP (me) is interested in engines of all sizes and flavours. Have MVEMs had their day, or will NNs keep them alive?

- Steve

 

[TDIMeister]

The paradigm in engine controls is actually moving away from "sensor-based" to "model-based".

Instead of sensors acting in a phenomenologically-based feedback loop and a relatively "dumb" look-up method, future systems will have physics-based models that basically stores equations of physical processes instead of mapped values.

The objective here is to minimize the number of sensors, because each added component increases cost and the size of the BOM, as well as represent potential failure modes. Another reason is that, as explained above, they're rather dumb: sensors work on historical sensed information and don't do well to look-ahead and compensate for engine wear, variations, and sensor deviations/deterioration.

An example might be emissions control, which is key to the move toward model-based control because of regulations that require emission control systems to be functional and still meet stringent tailpipe limits to 120,000 miles. With a sensor-based system, the O2 sensor is critical to knowing the composition of the exhaust gas. Model-based control may not eliminate the O2 sensor in this case, but the ECU will have physics-based models that can basically calculate almost everything the ECU would need to know based on a more limited set of sensed inputs.

 

[catserveng]

CAT has been using charge density control on larger natural gas fueled engines for about 4 years now. Instead of an O2 sensor they are using quite a bit of modeling and three dimensional maps to obtain very good mechaincal efficiency and low emissions. Seems to work well, but some of the early problems getting all the mapping worked out was and continues to be quite a chore.

Some of the prechambered gas fueled engines are using combustion feedback, it looks at the amount of time the flame front travels across the cyliner starting the time from when the primary igniton pulse fires the coil.

Yes, lot's of cost and hardware, but in this size range a couple of points in effciency of a few ppm of NOx is a big deal for an engine running over 8000 hours a year.

We have also seen test injectors with on board pressure transducers, something some of the factory engineers indicate we may see as a part of high performance packages full time.

It has been a pretty interesting last few years with technology improvemetns on the larger engines, and expect with desire for lower emissions and fuel consumption, that the complexity and capability of engine controls in our market will continue to advance. And we see in our market that technology that was on the larger engines a few years ago is now moving to the smaller size units.

The challenges from a field perspective? Making the service tooling as smart as the engine controls. Having reasonable service life for the sensors being installed, and having the design engineers listen to the field folks when they report problems and come up with real world diagnostic procedures and solutions.

Just my two cents.