woot woot!

Ok.....

I'm trying to find a value within a function f(x)

I have two bounds (upper and lower)

and the function should be quadratic in nature

problem is that the quadratic might be either positive or negative

You have it in the form ax^2 + bx + c, or is the actual function hidden from you?

It's OK if any of those constants are negative, I think the math should work out the same

no I don't have the function known to me per se

well I guess I kinda do

this should help us a little

github.com/KronosKoderS/sie552/blob/dev/venus_example.ipynb

This is the specific file that I'm working with

go down to cell 23

the transfer_ellipse function

that's what I'm trying to get working correctly

ow.

Oof, this appears to be a good ways over my head

ya...

it's really simple honestly

the thing that I'm trying to do

the whole designing the transfer ellipse isn't so much

it's why it's a grad course :P

I'm trying to reduce it down to something simple.....

@BenN waaaaaay over mine

hm

Well, hm, how is it currently working? Does it work in some cases?

def find_val():
    upper = 100
    lower = 0
    found = False


    while not found:
        i = upper - lower / 2
        val = f(i)

        if desired_val - tolerance < val < desired_val + tolerance:
            found = True
        elif val > desired_val:
            lower = i
        else:
            upper = i

I only know some of these words from watching KSP videos

ok... this is basically what I'm trying to do

@KronoS So did you just need the pos result, or both, or ..?

BUT... val may be inverted and therefore the upper and lower would have to be inverted

I just don't know exactly how to do that

@BenN yes it does....

specifically in the example of the earth to venus case

ah

Probably silly question, but by inverted do you mean negative?

BUT I took a different apprach

@BenN not necessarily

think of a simple parabola

Oh, OK, I think I see

@KronoS and you don't have access to the quadratic formula?

if you did, you'd take the determinant and figure out whether it's inverted from there...

at least that's all I remember :(

well... its actually a cos wave

o.O

but I'm looking at a small section of it, which makes it more like a quadratic

t = (E - e sin E) / n

that's the specific formula

Oh, that can be handled I think, one moment

soooooooo... by "inverted" what you really mean is "decreasing"?

I'm altering values of E to find a specific t

more accurately, that the differential is, uh, negative?

If you take the derivative of that sine wave and find zeros in the range, that should do it?

whoops. derivative.

Keep a little list of the values at the critical points and choose the one that makes the biggest/bestest value, and don't forget the ends of the range

wolframalpha.com/input/?i=x^2

wolframalpha.com/input/?i=-x^2

dt/dE = (1 - e cos E) / n I think

wow... much link failure

Calculus I can do, physics not so much :)

@BenN if the bounds are already known and you only need to swap them, then it'd be easier to just check if it's pos or neg at that point, no?

@BenN I'm not trying to find a max/min though... just a value

Oh

this is bringing back horrible memories :P

looking for a specific value of t, by altering E

yea so, if the derivative is negative at that point then it looks like an "inverted" parabola part of the wave

perhaps what I can do... is divide up the range into sections

Or check whether the value gets "worse" after taking a small step in the direction the current code would suggest

examine the value at the edges of those sections and then refine it down that way

:/

Would it be feasible to just scan the entire range in steps of tolerance to see which try is closest?

Actually, since the derivative is pretty manageable, you could evaluate that to tell you which way to go