Computation Details for the Curious and Unsuspecting
Unless you are really getting excited about doing your own calculations or may be refining those presented here, you may wish to skip this section. We are going to wallow in some of the nitty-gritty details needed to solve the equations. It also will assume the reader is familiar with physics and computer programing.
drag coefficient calculation-non spinning
case
The drag coefficient on a non spinning ball was calculated using the
data of Achenbach That is displayed in the section on the Magnus Effect.
To obtain the proper value of the drag coefficient at every time interval,
the curve was fit to an arbitrary function of a type known as a rational
equation. The points to fit were manually transcribed off the published graph.
Rational equations have the general form:
There are many software programs which have regression routines to
fit data to specific functions. SigmaPlot (ver 4.0), a scientific graphing
program which has regression capabilities, was used to do a least squares
minimization to find the constants a-g. The points shown in the Achenbach
figure were the data from the experimental plot. The line represents the
actual fitted data.
drag and lift coefficient-spinning case
In a similar manner, the data for the drag and lift coefficients for the
spinning paintball were calculated using a rational equation. Two sets of data were
merged for this, the data from Davies and the high V/U data from Maccoll. This was necessary
because paintballs will achieve V/U values of around 3 near the end of their flight. The
Davies data only goes to 0.56. The fit to
the Davies/Maccoll data is shown at the right. The points are the data taken from
Davies and Maccoll figures, and the line represents the fit. However, there is a complication
in the calculation of the drag coefficient data for a spinning paintball.
The data by Davies was generated at somewhere around 105 ft/s, and the
data replotted by Bearman and Harvey is in terms of V/U (spin velocity/linear
velocity). The Reynold's number for Davies data was ~9 x 105 and
shows a value of 0.42 at zero spin velocity. Achenbach's data shows a value
of 0.50 in this same Reynolds number region. To get
adequate correspondance between spin and non spin data, it was necessary
to scale the Davies drag data according to the Achenbach data. To do this,
a first order correction was applied: The Achenbach data was used
to find the CD for the non spinning case, and this value was
substituted in the "a" value of the rational equation for Davies spinning
ball rational fitting equation. (The first constant "a" represents the
zero spin case, because V/U = 0.) This assumes that the relationship between
CD and V/U of Davies spinning sphere data are the same at all
relevant Reynold numbers. This assumption has not been proved, but I believe
it is valid enough for the calculations here. This generally brings up
a point that CL data would be better plotted not as rotational
velocity/linear velocity (V/U), but rotational velocity/Reynolds number
(V/Re). The constants for all the different rational equations that were
fitted to the data can be obtained from the source program.
Also, I considered that the spin velocity was constant throughout the entire flight. This assumption is not realistic. The air resistance around the ball would likely cause the spin rate to decrease. The question is how fast. Some people on the rec.sport.paintball group think that it will be damped very fast,within a short distance from the gun. Others like me are not so sure. I believe a rough answer to this question is already available somewhere in the fluid dynamics literature. I just have not had time to dig it out. Until I or someone else does, I will assume the the drag effect on spin is small and therefore negligible. Realize that the answer to this could make a real change in my estimations of spin effects.
equations of motion integration
The equations of motion were solved using the Euler-Richardson numerical
integration method. The code was based on an example in the book by Gould
and Tobochnik (see references); this was modified for the 3D case here.
I found this method particularly easy for following the programing logic.
A better integration method probably is a Runge-Kutta strategy. This may
someday make its way into the Java version of the program, but required
more work intially then I had time for. The time interval was fixed at
0.001 seconds. Most trajectories were on the order of 3- 5 seconds, so
that 3k-5K calculation cycles were needed. On my PII/400 the time for most
simple trajectory calculations was under ten seconds, even in the worst
cases. The original coding for all the results produced here were written
in Excel 97 brand of Visual Basic so that all the calculated values could
be dumped to a spreadsheet. All the graph information displayed here is
the result of some variant of the Visual Basic code.
The original VBA coding turned out to be straightforward. Converting
to the proper units produced the most frustration. The program uses three
measuring systems: The trajectory calculations use the mks system, while
the calculation of the CL and CD constants used the cgs system. There are
two functions to obtain these numbers. Since many of the readers are still
mired in the old English system, the metric units were then converted to
English units at the end for output.
Preamble
Introduction
Trajectory
Equations Section
Results
Section
Relevance
Section
Discussion
Section
Explorations
Section
Conclusions
References
Back to Paintball Main Page