In my understanding, there are two ways to go to compute the best BSFC line of an internal combustion engine. The first one is: fix the engine speed and find the torque associated with the minimum BSFC for this speed; change to another speed and do it again; finally we can get the best BSFC line. The second one is that: we can also fix the torque first and find the speed associated withe the minimum BSFC for this torque value; change to another torque and do it again. Looks like either way makes sense. But the lines that I got using this two methods were different, so I am very confused. Can anyone tell me where did I do wrongly?
Why not try to draw a 3D BSFC contour, which will clearly show you the best BSFC point or area in the whole engine map. As in this contour, you can use engine speed as X axis, and load (such as BMEP, torque, intake pressure, etc.) as Y axis, and BSFC as the Z axis. As show in the attached picture, or you could search BSFC map on Google Images for a sample. To draw this kind of diagram, you can use Matlab, or special data process software such as Uniplot and so on.
The topic areas indicate that this is for a gasoline internal combustion engine. Does the engine use an electronic engine management system (EMS)? If so, does the engine also use electronic throttle control (ETC)? How those two questions are answered have important implications regarding how to develop both torque and BSFC maps with the engine. EMS component protection modes have to be understood before the engine can be mapped accurately.
Once you are convinced that the map your are working with is an accurate reflection of engine operation, the speed-torque line for minimum BSFC can either be determined by visual inspection of a speed-torque-BSFC or speed-BMEP-BSFC contour map (assuming that the contours have sufficient resolution for your purposes) or it can be determined by a database or spreadsheet query of the raw data used to draw the contour map. There are sometimes more than one "island" of minimum BSFC depending on the engine design and it's control parameters. In the map example shown earlier, minimum BSFC was196 g/kW-hr at 17 bar BMEP, 2300 rpm and with an island of 200 g/kW-hr from about 14-21 bar BMEP and about 2000-2700 rpm.
Hi, Joesph, Thanks a lot for the answer. I believe this engine has both EMS and ETC. Actually I already had the speed-torque-BSFC map data. My question is regarding how to plot the minimum BSFC speed-torque line according to these data. I don't have multiple "islands" of minimum BSFC in my case. I want to get an accurate speed-torque line for the minium BSFC, meaning the exact number. So i think visual inspection is not going to work out. What do you mean by database query?
Thanks again.
BSFC is shown as contours of iso-BSFC, which forms shapes of an island on the speed-torque-BSFC map. Dense collection of data will result in finer contours. Inner-most island on your map showing zone of minimum BSFC. The line which you want is the border line of the BSFC contour island.
I need to understand your question again.
Are you looking for min. BSFC at each RPM? What you want to achieve?
Yes, I am looking for min BSFC at each RPM. If I link together all the min BSFC points at each RPM, I will get a minimum BSFC line. Right? On the other hand, if try to find min BSFC at each torque, and link together all the min BSFC points at each torque value, I will get another minimum BSFC. Why these two BSFC lines are different and which one is correct?
I suggest you to look the universal BSFC contour again in my earlier answer.
The minimum BSFC points at each speed and minimum BSFC points at each torque are totally differently located. So obviously you can not get two equal lines by connecting the two set of points. The deep reason is that, for each speed, the engine gets best BSFC at middle-to-high load (for example in gasoline engine, at low load pumping loss is higher, at WOT the spark timing has to be retarded due to knock); while for each torque level, the engine gets best BSFC at middle speed (at low speed the combustion is slow and leakage loss is higher, at high speed friction loss is higher). So your proposed two lines, one is longitudinal, while another is horizontal, no way to be the same. None of your proposed two lines is wrong. Just they are talking about different aspect of engine performance. This is why we have speed characteristic curves, load characteristic curves, and still universal maps to define engine performance. Finally the universal BSFC map (contour) is the best way to understand the whole and the best fuel consumption of an engine. This is why it is recommended by ISO and SAE Standards.
It is more correct to seek the minima at constant brake power. I suggest to plot iso-bsfc lines on a power vs engine-speed diagram. Your least bsfc line is the locus of the points on the iso-bsfc that have a horizontal tangent
A fairly easy method is to plot lines of constant power on the BSFC map, then search along each iso power line for the lowest BSFC, If you do this in steps of a few kW you can then join the dots and will then end up with an operating line that delivers the lowest BSFC for any power. See a paper I did a long time back for a simple procedure that works pretty well.
Article An operating point optimizer for the design and calibration ...
Unless there is unacceptable measurement uncertainty or a problem with the laboratory equipment or its calibration, there is no reason why fixing torque and sweeping speed or fixing speed and sweeping torque shouldn't yield the same results. Another issue might be how you are fitting points among a sparse data set. We often generate iso-bar contour plots of constant BSFC relative to BMEP and engine speed to find speed/load "islands" of best BSFC, but it usually takes about 200 or more separate speed-load points for us to have sufficient data for a reasonably good fit . Is it possible that your two mapping methods are not taking a sufficient number of data points to provide a reasonable fit of the data? How many data points are you using? What are you using to fit the BSFC data?
I have also seen problems with the turn-down ratio of some fuel flow meters causing unacceptable uncertainty in the lower range of their flow calibration. Do you have a means to cross-check your fuel flow measurement to see if it is a source of error in your measurements? Is your meter calibrated vs a traceable standard and properly selected for your encountered range of flow rates? Sometimes we use a mass balance method to cross-check fuel flow measurements. If you have a well-calibrated air-flow measurement (LFE or hot-wire anemometer) and a well calibrated UEGO lambda sensor (not a factory calibration - actual lean/rich calibration with gas standards), you can use air-flow and air/fuel measurements to calculate fuel mass flow at different engine operational points to check against the fuel meter measurements.
You didn't mention the engine type. There are engines that I have tested that require care taken to reliably reach certain operating points. For example, electronically controlled spark ignition engines operated near points of knock-limited operation or operated near points where component protection from high temperatures may be needed can have significant changes in fueling due to enrichment and/or changes to spark timing and electronic throttle position. This causes changes in measured fuel consumption that are real but that can be difficult to identify unless engine control parameters (spark timing, manifold pressure, throttle position, cam phasing, etc.) are monitored either via CAN signals or independent measurements. In some cases, particularly high-load or knock-limited operation, you may need to establish stability criteria (e.g., COV) so that you are not taking fuel flow data while A/F, spark timing, and (if so equipped) cam-phasing are all dynamically changing at a speed/load point in response to knock-sensor input or in response to the EMS providing component protection. This is particularly true of electronically controlled, boosted engines.
Hi Joseph, all good points. The method I used works on a previously obtained data set, In this case I was using neural networks to do the interpolation and estimate BSFC and emissions at each point whilst accounting for engine condition and actuator settings such as coolant temperature, injection timing, EGR ratio etc.
The data set was generated from a fairly extensive test campaign where all of these factors were perturbed on the test cell using a design of experiments approach to ensure adequate coverage as far as possible. There were several hundred test points and ideally I would have had many more.
In order to escape these limitations we have more recently been working on dynamic design of experiments techniques where all of the actuators are driven through a sine wave based excitation sequence designed to expose all of the dynamics of the system as well as visit all of the design space. Clearly this is demanding in terms of actuator and instrumentation bandwidth but approaches like this are increasingly important as we move to a more real-world calibration environment. There is simply not time to map the entire engine across the whole envelope in the traditional way.
Some of these dynamics are still very difficult to characterize. We have encountered TGDI engine calibrations with more than one WOT torque curve depending on knock sensor input and time. In some cases, this has required us to dynamically increase load up to the WOT threshold for brief periods in order to characterize these operational points.
Heavy-duty diesel engines (e.g., U.S. Class 8 truck applications) often have more than one full-load torque calibration that is switched between or iterated between depending on sensor feedback to the EMS. Dynamic electronic control of engines has certainly made performance mapping more interesting over the last decade. It also means that for engine dynamometer measurements to provide meaningful information, we need to make sure that they adequately represent a full range of conditions encountered during in-use operation.
Absolutely, a real challenge. We need full access to engine control variables and preferably the strategy to get a full understanding of what is going on. In simulation we aim to recreate the parts of the strategy that govern behaviour we are interested in, otherwise the simulation will be unrepresentative. In engine test we have to have much more capable cells than we did in the past. Exciting times.
https://x-engineer.org/automotive-engineering/internal-combustion-engines/performance/brake-specific-fuel-consumption-bsfc/
see this