As the other poster succinctly said, it’s about iterative solving problems. What might not be obvious is where you want to use this.
Let’s say you are holding onto a continuous, all-metal cooking pan’s handle and you turn on the burner; at what second will it be too hot to hold? If we assume the pan is starting at 20 C and it’ll be too hot for your leathery, lemmy-browsing hand at 100 C, then we at least have some basic bounds on the problem, but how do we think about the middle parts?
We could say that the burner is putting out about 5 kW of energy (or more simply, that the pan’s bottom has a fixed input of 5 kW (a so-called boundary condition). With a basic equation translating energy input into temperature (factoring in specific heat), we at least know the temperature of the bottom of the pan at a given moment, but what about the top of the pan and, most importantly, the handle? The temperature in these places isn’t climbing because they’re getting licked by flames, but due to thermal conductivity away from the bottom.
Thermal conductivity is actually one of the simplest questions in partial differential equations (PDEs), and is addressed with the aptly named heat equation. In extremely casual terms, the heat equation (du/dt = L u) is saying that the heat at a given position is determined by its past heat plus the past heats of positions around it. Our job as mathematicians is to express our problem using a so-called weak form of the heat equation that makes it possible for an iterative solution.
We can solve weak form equations with a framework like the Finite Differences method (which is a little dated but easier to understand than other methods, and also still quite effective). This very basically means we’ll create a fixed grid of points, then express the relationships between them with algebra.
Once we have our algebraic mesh and we are happy with all of our boundary conditions, we can start the simulation. We watch as the temperature immediately shoots up on the bottom of the pan (followed soon after by the top and edges), then as heat slowly transfers up the handle until it reaches 100 C at the end.
In doing these problems, we make as many simplifications as possible in the boundary conditions, because the equation’s solution space is unfathomably large. Everything we can do to constrain it (easy heat source, exact physical parameters of our materials, etc) shrink the solution space closer to one that matches our actual problem.
As a final thought, lots of very bright people have spent entire careers designing heatsinks using numerical methods, so don’t take my description of the heat equation as an easy problem to mean it’s not a great line of inquiry, but rather that the field is really fucking hard. Anyone who does this stuff can, in fact, hold a 90 C pan handle because their tolerance for pain is astronomical.
What’s numerical methods all about?
I mean I think I can intuit a little and could Google a lot, but hearing from someone who seemingly somewhat recently studied is a unique chance.
As the other poster succinctly said, it’s about iterative solving problems. What might not be obvious is where you want to use this.
Let’s say you are holding onto a continuous, all-metal cooking pan’s handle and you turn on the burner; at what second will it be too hot to hold? If we assume the pan is starting at 20 C and it’ll be too hot for your leathery, lemmy-browsing hand at 100 C, then we at least have some basic bounds on the problem, but how do we think about the middle parts?
We could say that the burner is putting out about 5 kW of energy (or more simply, that the pan’s bottom has a fixed input of 5 kW (a so-called boundary condition). With a basic equation translating energy input into temperature (factoring in specific heat), we at least know the temperature of the bottom of the pan at a given moment, but what about the top of the pan and, most importantly, the handle? The temperature in these places isn’t climbing because they’re getting licked by flames, but due to thermal conductivity away from the bottom.
Thermal conductivity is actually one of the simplest questions in partial differential equations (PDEs), and is addressed with the aptly named heat equation. In extremely casual terms, the heat equation (du/dt = L u) is saying that the heat at a given position is determined by its past heat plus the past heats of positions around it. Our job as mathematicians is to express our problem using a so-called weak form of the heat equation that makes it possible for an iterative solution.
We can solve weak form equations with a framework like the Finite Differences method (which is a little dated but easier to understand than other methods, and also still quite effective). This very basically means we’ll create a fixed grid of points, then express the relationships between them with algebra.
Once we have our algebraic mesh and we are happy with all of our boundary conditions, we can start the simulation. We watch as the temperature immediately shoots up on the bottom of the pan (followed soon after by the top and edges), then as heat slowly transfers up the handle until it reaches 100 C at the end.
In doing these problems, we make as many simplifications as possible in the boundary conditions, because the equation’s solution space is unfathomably large. Everything we can do to constrain it (easy heat source, exact physical parameters of our materials, etc) shrink the solution space closer to one that matches our actual problem.
As a final thought, lots of very bright people have spent entire careers designing heatsinks using numerical methods, so don’t take my description of the heat equation as an easy problem to mean it’s not a great line of inquiry, but rather that the field is really fucking hard. Anyone who does this stuff can, in fact, hold a 90 C pan handle because their tolerance for pain is astronomical.
It’s all about solving things iteratively. Usually applied to how computers solve for things.
Thank you for the summary