>>12810364As I'm sure you aware, since the advent of cheap computers, we have moved away from simple linearized models and towards more accurate non-linear coupled modelling that does not permit anatical solutions. Solving these often require modern numerical methods; optimization algorithms for parameterisation, PDE discretisation and time integrators etc. A lot of modern applied mathematics and computer science has gone into improving these algorithms, especially since new problems/models we encounter often require new methods. Now we all publish these improvements in an abstract problem statement, so that they may be used by other scientists, engineers and (applied) mathematicians who have a similar system.
Phycists are not ignorant of this, they are usually taught numerical analysis and programming in undergrad. The problem is that they are (on average) _bad_ at it, they reason they are bad at it is because of the way upper level physics is still taught with an emphasis on those old perturbed/linearized models. That is how physicists are trained and how they try to solve new modelling and simulation problems. A tell of this is that Phycisists are comfortable using commercial simulation tools, but they are not comfortable developing new numerical methods to solve new problems (which is something researchers can only do by using open source code that they can modify; this community contains researchers from math, engineering, computer scientists, but virtually no physicists). So instead of developing a new method to solve a complex model, physicists instead to reduce complex models to simpler systems that are easier to solve/simulate using methods they are comfortable with.
In particular in multidisciplinary areas such as nanofluids, MRI, superconductors, thermophysics etc. I always see worthless papers written by Phycisits that fuck around with baby perturbation algebraic manipulations that end up producing worse results than even older numerical methods.
