Preliminary results - horsepower

As I stated in one of my prior entries, I was going to run the data twice, once for horsepower and once for torque. Let's take a look at the preliminary results separately.

For the horsepower test, values ranged from 29 to 50 horsepower. This represents the highest number for each of the 150 datasets. Here is the distribution:

Number Of
horsepower Occurances
---------- ----------
29 1
30 1
31 1
32 2
33 3
34 19
35 3
36 1
37 3
38 6
39 17
40 6
41 4
42 16
43 15
44 19
45 19
46 4
47 6
48 2
49 1
50 1
---------
150 total

For the top three horsepower values, 48-49-50, the "path" to each was reached by changing only ONE variable. I'll get to which one in a minute. I have opted to not use these cases for two reasons. First, the number ofoccurances per value is minimal, 1-1-2. Second, if you look at a "distribution" of this test, these 3 values would probably be considered "outliers" and invalid for inclusion.

I think now would be a good time to mention that the ID implementation I use is capable of "extrapolation". Here's what that means. For the intake, advancing the cam one tooth or 7.1 degrees gives you intake opening at 41.1 degrees. If you advance the intake cam 2 teeth, that gives you an intake opening of 48.1 degrees. Now if the optimal intake opening is somewhere between stock, 34 degrees, and one or two teeth advanced, 41.1 or 48.1 degrees, the ID implementation I use can extrapolate, or rough out, the value.

If there is any intest, I can put the entire data file available for download on the Keller Racing website, http://KellerRacing.net. That way you may peruse all the data values used.

OK, I had promised that I would let you know what the single variable to change to achieve the maximum horsepower. An intake valve opening of LESS THAN 23.4 degrees gives you 48-49-50 horsepower, one occurance of each. An intake valve opening of GREATER THAN OR EQUAL TO 23.4 degrees can give you one occurance of 48 horsepower.

So if you discount those three figures, 48, 49 and 50, the number 47 appears in horsepower 6 times. The shortest route to get to that figure is an intake valve opening LESS THAN 37.55 degrees AND an exhaust valve opening of AROUND 65.6 degrees, preferably LESS THAN.

What is very interesting is the fact that intake valve closing appears NOWHERE in the "map"!
For my next entry, I will detail the results for torque which should be much more appropriate for dealing with the "intake valve closing" question.

Data

At this point, I have 25 data sets.
1. Base or stock configuration
2. Advance intake valve opening 7.1 degrees
3. Advance intake valve opening 14.2 degrees
4. Retard intake valve opening 7.1 degrees
5. Retard intake valve opening 14.2 degrees
6. Advance exhaust valve opening 7.1 degrees
7. Advance exhaust valve opening 14.2 degrees
8. Retard exhaust valve opening 7.1 degrees
9. Retard exhaust valve opening 14.2 degrees
10. Increase intake valve diameter +5%
11. Increase intake valve diameter +10%
12. Decrease intake valve diameter -5%
13. Decrease intake valve diameter -10%
14. Increase exhaust valve diameter +5%
15. Increase exhaust valve diameter +10%
16. Decrease exhaust valve diameter -5%
17. Decrease exhaust valve diameter -10%
18. Increase intake valve lift +5%
19. Increase intake valve lift +10%
20. Decrease intake valve lift -5%
21. Decrease intake valve lift -10%
22. Increase exhaust valve lift +5%
23. Increase exhaust valve lift +10%
24. Decrease exhaust valve lift -5%
25. Decrease exhaust valve lift -10%
I also restricted the output to look at to between 4500 and 7000 RPM in 500 RPM increments. That totals 150 datasets.Due to the basic engine design, I knew that peak torque and horsepower would occur in that RPM range. Also, I separated torque and horsepower as two distinctly separate result goals. Why? Well, usually to achieve the maximum of either requires different configurations. Adding valve diameter and lift also makes the test more valid as it definitely complicates that data and adds more possible contributory factors.
Time to run the data. I'll post the results next time.

Tools of the trade

OK, we have decided on a problem, identify primary discriminatory factors among valve timing events. Now we have to assemble our tools and develop our data and methodology.
There is an artificial intelligence method called an "iterative dichotimizer"(ID) that is used to identify the shortest path from data to a specific result.For example, it we have the four valve events, which one is the quickest way to maximum torque or maximum horsepower. We will use this method to determine which valve event is most important. For this tool, I will use an old DOS program that I've used for years. Why a DOS tool? Well, it's cheap (I already have it). I'm familiar with it and have used it for years.And last, it is a very good implementation of the ID algorithm and I know it works.
Next, because I don't have the time or money to build a bunch of motors and dyno them to get results, I'll use an engine simulator. There are many of these available and I would prefer not to recommend one over the other because I've only used a few. The ones I haven't used, I do not have sufficient knowledge to judge them.
However, I will make a generalization. Try not to use a simulation program that uses a table lookup for values for Volumetric Efficiency. Try to find one that actually computes VE for the entire 720 degrees of 4 stroke rotation.
I will also be using Microsoft's Excel spreadsheet for some data analysis and a few progams I have written myself.
In regards to the data, I will "simulate" engines similar to the one I use for racing. Now here is where a Triumph engine shines. The cams (separate intake and exhaust) on a Triumph are gear driven. There is 51 teeth on these gears. This means that altering the timing by one tooth is a 7.0589 degree advance or retard (360 degrees divided by 51 teeth). Additionally, there are three keyways on a Triumph cam gear and they are 5 degrees apart. Soooo, by a one tooth and one keyway change, I can alter timing by a minimum of 2.06(rounded off) degrees. That's cutting it fairly fine!
I think I will also add valve diameter and lift to the mix. When using an ID algorithm, it's best to have "complicated" data.
Now off to the computer to generate data, which I'll present and detail in the next entry.

Description of problem/question

First a little background. I like performance and racing. I've even been known to watch lawnmower races on TV. I really love racing. I really,really love drag racing. And I really, really, REALLY love motorcycle drag racing! I've been riding bikes since '62 but it's only been relatively recently (last 10-12 years) that I've been involved with motorcycle dragracing.
During that time, I've read many, many books on racing and performance. One common thread that ran through all of them was that of the four valve events(intake opening/intake closing, exhaust opening/exhaust closing), intake closing was the "most important". I understood the rationale behind this statement. You keep the intake valve open after BDC (bottom dead center)and rely on the inertia of the incoming charge to continue to fill the cylinder beyond a VE of 100%.
However, I was curious if the statement would hold up under close scrutiny.I then set about to devise a method of testing it (short of building several dozen motors to test all the changes).
In my next post, I will detail the data I used to confirm/deny the hypothesis.
Ahh, what the heck! I think I'll also use this blog to keep my many fans updated on the changes to the drag bike for the 2010 season (for pictures of the bike, please check out the Keller Racing website).

While Keller Racing does have a website at www.KellerRacing.net, which has a page on using Artificial Intelligence to enhance your racing experience, it is to much work ;-)) to constantly update a web page. First you have to design the page, then enter the content, finally FTP the revised file up to the website.

In view of some of the results I have been getting, I thought it may be easier to just "blog" the data.Now this blog won't be about daily entries, or multiple entries per day. It will probably only see entries about once a week, maybe a bit less. But that is the nature of the beast. Much of what I'm attempting requires a lot of research and pre- and post- processing of data, both endeavors require time.

So bear with me and hopefully this will be beneficial to both of us.

Also, I'm attempting to get the "e-mail to blog" function working and not having much success.