


Power output formulas
I've seen engine output calculators like desktop dyno and the like flying around, and the sometimes unrealistic figures people get from them, and it made me wonder, What is the program actually doing?
I figured it could work one of two ways: it has a library of known values for such things as flow, lift, compression, I/E events, spark timing, and frictional losses in a given motor with given parts and calculates power output based on how other similar motors have performed. Or it takes these values and applies them to some calculus based formula for power output. My question is this: Are there better formulas for calculating power output than those used by these programs, and if so, what are they, and how do they work? I would appreciate any information on this. Thank you. Travis 
Sponsored Links  
Advertisement 



There are a lot of calculations that are performed by the good programs, instead of listing the types of figures and amount of information that is processed have a look at this web page. There is a good screen shot of the text dump that is created through the algorithm.
http://users.erols.com/srweiss/winscren.htm
__________________
Outlawed tunes from outlawed pipes 


Thanks chuck, many more variables involved in those calculations than I was aware of. Most people just can't bear to give a no BS assessment of what their motor really has in it. Had a buddy with a warmed over 350 in his 62 pickup and thought he was pushing 350400 horses. Just had to laugh to myself. He ended up putting a rod through the pan thinking his 4 bolt mains would keep it together at 6000. Yeah deuce, GIGO, eh? Thanks for the help. By the way, would you recommend any particular program over the others? Gracias, muchachos.
Travis 


Quote:
__________________
Outlawed tunes from outlawed pipes 


i agree with chuck, those programs leave a lot of room for error. I think a better way of calculating the results from those programs would be to have actually tuned and dynotested severalhundred combos and tell you results of two or three that are similar to yours (ie, your motor has the same heads, a little less compression and a little less cam than this combo that makes its peak 435 horse at 5800 RPM, motor #2 has less cam and less compression than yours and makes a peak of 4400RPM and 385 horse. Logically yours should fall in the 400 range.) I actually have all the cam and flow data for cams and heads specific to manufactures in my dyno2000 program, and i don't think it has any means of descriminating that a head that flows such a cfm may mix and burn better than another head that flows the same cfm, so I've "dynoed" a few combos I found on the web and have been way off. Other times its pretty close, just too much error.
When it all comes down to it, numbers don't mean a thing. We all just end up building the best motor we can for our money. K 


Not all computer programs are bad. All major racing teams use them BUT they are in the thousands of dollars with phone books for instructions and a instructor the comes out the teach you how to use it.
There is alot of things you can not come up with in a normal program for example the way the heads are ported. You can have such small nuances in porting that will change a power output dramatically. I personally don't use them for power output numbers. Specially since most are very incomplete (i.e. changes in using an electric water pump or windage tray). But if you do use one I would say aim low on the program and you'll be more pleased in the end with the result of the engine. 


unless you are doing a 3d scan of you intake, heads, carb, and have very accurate physics models determining turbulence, swirl, and other fluid properties of the intake charge, i don't see how you would get an accurate answer, unless it's by chance. Then you have the temperature of the intake charge as well, and how and when it warms up, then you have how well the exhaust scavenges, spark plug settings, ignition timing, weight of oil, types of bearings, etc.
simply too many variables for a program to take input from a keyboard and mouse and give a correct answer 


Quote:
The goal of any simulation is to determine the most pertinent physical parameters to the process being modeled and develop a mathematical relationship that will yield a result that accurately predicts an outcome. There are two basic but very broad classes of models; "Analytical" and "Finite Element or Numerical". Analytical models typically consist of a few root equations that give you the 80% answer. They can be applied easily with paper and pencil or at worst on a hand held programmable or simple PC program. I have written and published several such analytical models for oil reservoir processes and they in fact yield the desired 80% result. Would I spend $50,000,000 based on their output? Not in this lifetime!! But they are great for pointing me in the right direction. They take virtually no time to apply and give very useful "ball park" answers that lead to forgetting the project or induce committing $$ for further detailed study. A great example of a simple analytical model is the equation thrown around on this board for determining the length of the primary pipe in an exhaust header, L = 120V/rpm For L = pipe length, less port length in head, in inches and V = velocity of sound in hot gasses. Values of 1300ft/sec to 1700 ft/sec are common. Using V = 1700ft/sec the equations simplifies to L = 204,000/rpm. A lot of experimental engineering went into that simple equation but all that work determined that of the admitted nearly infinite number of variables in an exhaust system, gas velocity is the one that is most determinant in the final optimal length. From this result, an engineer is way down the road in coming up with the final design. From here he can go to a finite element analysis or cut and try physical model (build some and see what they do!). Numerical or finite element models are orders of magnitude more comprehensive and expen$ive to run. Software for good models easily cost 6 figures to purchase, 5 figures in annual maintenance fees and who knows how much in computer power to run them. The oil reservoir process simulators we use take several months to just configure the model in the computer then several days in computer time to make a single run. But, they are vastly more precise and accurate IF we put in the correct parameters in the first place. The only way to trust one of these is to 'history match' an existing project. If we can twist knobs in the computer to get it to predict performance that closely matches historical data, then we have confidence that it will correctly predict the future performance of our target project. These things divide the widget being studied into very small blocks or "finite elements". A finite element model for an Aarm in an auto suspension may have several thousand blocks. The physical characteristics of each little block (metallurgy, size and shape, etc.) is input for the computer model. Several equations are written for every conceivable impact each block has on its neighbor. There are typically 9 neighbors for each block so all equations must be calculated for each neighbor for each block for each time step. I haven't mentioned time steps yet but these models are predicting the result of a dynamic process  what happens over time?  so not only is the model busted up into small pieces, so is time. A change is made to the input, ie., a force is applied to our Aarm, then the effect of the force on each finite element and it's neighbors is calculated. That may require solving multiple millions of equations for one time step. Then the computer model looks at the result and determines if that one result makes sense based on limits specified by the engineer. For example, if the time step chosen for our Aarm was too large and the model calculation predicted a ridiculous 1" deflection in a pivot bushing, the time step was too large so the computer throws out that calculation, chooses a smaller time step, and makes another run. It is not uncommon to make many 10s of iterations on a single time step to find one that yields an acceptable result so the computer can continue to the next time step. An Aarm deflection happens over tenths of a second but it may take time steps measured in 100,000ths of a second to properly model the dynamic so a single computer run may take hours or days to do the required billions of calculations. An oil field reservoir process lasts decades but some parts of the simulation must be broken into time steps measured in days so, again, computer time escalates to massive numbers. The point of this dissertation (yes there is a point!) is that the DeskTop Dynos out there are 80% analytical models. They do a great job of giving quick, generally accurate and RELATIVELY correct answers to otherwise complex problems. I have one and am very happy with the results, realizing that it is the result of an analytical model. It does exactly what it was intended to do and shouldn't be rejected or held to a standard it was not designed to meet. None of us on this board could afford the finite element software and necessary support hardware to achieve the 95% answer. That is still generally the purview of the Fortune 500 company with a stable full of PHDs. Actually there have been huge inroads in the past couple of years in software and hardware development that is putting these tools on the desks of application engineers. Still an order of magnitude out of our reach though! Last edited by willys36@aol.com; 04302004 at 11:45 AM. 


POWER OUTPUT FORMULAS
Reading this thread,which is very interesting,no doubt.There is one thing I noticed with your formulas,unless I missed it,(which is possible 'cause I'm doing 30 things @ once)
I noticed nothing was mentioned in the way of....ummm...call it external factors? or ambient factors maybe,,anyway,ambient air temp.,density,barometric pressure,and such are significant,aren't they? Did I miss something? 


Re: POWER OUTPUT FORMULAS
Quote:



good post Willys....
That's definitely one of the better descriptions of what FEA does. I test FEA software for a living and I must say that it's impressive to see just how accurate this stuff is getting. But even if we're charging nearly 6 figures for a solver (for one license for one year ), it all comes down to good input. If the user doesn't know what he's doing, he's going to get garbage. To give you guys an idea of how intricate these models are....I have worked with models of engine blocks that are over 8 gigabytes in size. A while back, I was looking at gas tank designs because there has been a push in the past few years to quiet the sloshing of fuel. Currently, a manufacturer will form a clear acrylic model of a tank and fill it part way with dyed water to test how the fuel is going to move inside the tank. To perform the analysis on a computer, it can take a month or so to prepare the model and then as much as 45 days computing the results. We're not talking about slow machines either. We usually run large jobs like that on a Sun machine with 9 Gb of RAM. These machines are typically over $50k each. As far as the external factors, the engineer preparing the analysis must decide which of the external factors are going to have a reasonable impact on the results. For example, in willys36 case of an Aarm being deflected, gravity is probably not going to have enough of an impact on the results to make the extra effort to add it to the analysis. 


Quote:
The equivalent equation for intake, bandied about at Chrysler in the fifties and sixties, as they developed their "ram" manifolds, was: L = 84,000/rpm The "L," in this case, was measured from the intake valve head to the plenum. A few years ago, I was asked to sit in on some dyno tests being run, by one of the magazines, on a BBC with FI stacks. I noticed that the Chrysler equation was "missing" the torque curve "bumps" produced by the engine. A young man on the dyno crew suggested that this might be due to a difference in cam duration. I immediately dismissed this possibility since the equation, at Chrysler, was considered something "handed down from the gods" and was never questioned. But, upon further consideration, I realized he was absolutely correct and then offered my apologies for my hasty rejection of his suggestion. (At least, I hope he got them, for I had already gone home and had to send them through a third party.) With this adjustment to the Chrysler equation, I was able to "explain" each bump on the curve. I present the above to point out that some of the modeling used by engineers is not analytically derived, but is obtained solely from observation (empirical)(as Willys points out: "A lot of experimental engineering...."), with little or no understanding of the underlying principles. In this case, the equation was based on the locations of the primary "bumps" on the torque curve as an engine was operated with manifolds having different "tuned" lengths. I seriously doubt if those early researchers understood that they were observing that which is commonly called, in liquid flow, water hammer. If they had, they could have easily modified the equation to predict the other "harmonics" and the effect of intake valve duration. With FEA, engine intake and exhaust flow has been modeled with a high degree of precision in recent years, but, judging from what I've been told and that which I've read since I retired, I still have serious doubts as to the modeling methodology. But, if the results are sufficiently accurate (and, in the case of modern engine modeling, I think they're more than adequte), full understanding is unnecessary. A saying I like: In any successful manufacturing venture, there comes a time when you must shoot the engineer and ship the product. 

Recent General Rodding Tech posts with photos 
Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)  
Thread Tools  

