I know that symmetry in mathematics is a deep and very well studied area. However, I'm looking specifically to define a rule by which we might define symmetry of graphs of functions in 2D/3D Cartesian coordinates and in (2D) polar coordinates.
In this sense, symmetry implies some form of periodicity of the graph about some line or set of lines, but I'd like something more clear and formal than that. Are there any such useful definitions?
A graph has a symmetry when you can apply a transformation to it, and end up with the same graph again. For example, a graph is symmetric about the line $x=2$ if you can apply the transformation $x\mapsto 4-x$ to its equation, and have it simplify to the same thing:
$\begin{align} y=x^2-4x+12 \mapsto\,\,\, &y = (4-x)^2 - 4(4-x) + 12\\ &y = 16 - 8x + x^2 - 16 + 4x + 12\\ &y = x^2 - 4x + 12 \end{align}$
Similarly, a graph is symmetric about the line $y=x$ if you can apply the transformation $x\mapsto y, y\mapsto x$, and have the equation come out the same:
$xy=2\mapsto\,\,\, yx=2\\$
A graph like $y=\sin x$ has translational symmetry, because you can apply the transformation $x\mapsto x+2\pi$, and end up with the same graph:
$\begin{align} y=\sin x \mapsto\,\,\, &y = \sin(x+2\pi)\\ &y = \sin x\cos 2\pi + \cos x\sin 2\pi\\ &y = \sin x \end{align}$
Does this help?