...
Since we have generalized solution in the first place...
First, we have the observations of some system made by a scientist. Careful measurements are taken and patterns begin to emerge.
The scientist then takes these observations and tries to model the system mathematically, usually in the form of an initial value problem (IVP) containing a differential equation, where all assumptions and boundaries are stated explicitly.
If the scientist is lucky, there is a solution for the differential equation. If not, then numerical techniques are used to approximate the solution, or the equation is linearized.
From the solution, predictions can be made to either verify the model, or discredit it. Once it is verified by experimentation over time, it may eventually become known as a law.
Because differential equations relate rates of change, and things change with time, position, or other variables, the resulting model will frequently be in the form of a differential equation.