How can I find a parabolic function that mimics a hyperbolic one? How would I find the parabolic function for the hyperbolic function $y=5\cosh(\frac x5)$?
-
4That depends on what you mean by "mimic". – Gerry Myerson Feb 21 '13 at 00:32
-
Probably by the Taylor series? – Berci Feb 21 '13 at 01:09
4 Answers
For the hyperbolic cosine the Taylor series is $\cosh(x)=1+x^2/2!+x^4/4!+\cdots$. Hence $5\cosh x/5 = 5(1+(x/5)^2/2!+\cdots)$ and a close approximation to your function, for values close to $0$ is $y=5+{x^2 \over 10}$.
A hanging cable or chain appears in the shape of a parabola but that is only an approximation. The ideal shape is hyperbolic or catenary.
- 3,702
It sounds like you want a parabola that traces a path similar to $\cosh$. These are very different functions. Far from the origin, parabolas can only rise as $kx^2$, while $\cosh$ is exponential, which rises much faster. Over a particular range, you can find the best parabola that fits $\cosh(x)$ but it is a deception to claim that all even functions that go to $+\infty$ as $x \to \infty$ are similar in any other way.
- 374,822
For simplicity, let's take parabola opening up with value $y=1$ at $x=0$.
Take the cosh function as $y = b \cosh(x/b)$ and the parabola as $y = a x^2+1$.
To get them as close as possible, we will try to match the second derivative at $x = 0$, since both have zero slope there.
For the parabola, $y''(0) = 2a$.
For the cosh, $y' = \sinh(x/b)$ and $y'' = (1/b)\cosh(x/b)$, so $y''(0) = 1/b$.
To have the parabola match this, we must have $2a = 1/b$ or $a = 1/(2b)$.
Therefore the parabola that most closely matches this hyperbola is $y = x^2/(2b) + 1$. In your case, $b = 5$ so the parabola is $x^2/10+1$.
- 107,799
-
The question seems to be about the hyperbolic cosine, not about any hyperbolas. – Gerry Myerson Feb 21 '13 at 01:33
The circular trig functions relate to the complex numbers. The hyperbolic ones to the split-complex numbers. Presumably the best analogy for parabolic functions would be the dual numbers.
These are boring. $\exp(\epsilon x) = 1 + \epsilon x$, so we would define
- $\mathop{\text{cosp}}(x) = 1$
- $\mathop{\text{sinp}}(x) = x$
-
1Wow. You don't seriously think this is what the question is about, do you? – Gerry Myerson Feb 21 '13 at 02:25