Can you trust your numeric simulations—How to verify your code and validate your model

Abstract

In many fields where numeric simulations are utilized, there are standard practices on how to make sure that the simulation results are of high quality. In room acoustic simulations this seems to be quite rare although there are several factors that affect the accuracy of the simulations. First of all, there should be guarantees that the mathematical model, typically partial differential equations, actually model the physical phenomena under investigation. This can be made sure by validation and it can take place, for example, by comparing the simulation results to some reference solution or to measurement data. This is typically not a concern for the linearized wave equation although adding realistic boundary conditions makes it more challenging. Another essential factor affecting the correctness and reproducibility of the results is the quality of the implementation. Code verification is a procedure that aims to guarantee that an implementation is free of errors. In this paper, we review some common practices of code verification. As a practical example, we show verification studies conducted with our finite-difference time-domain solver. In addition, we show convergence rates obtained with the same solver to demonstrate the order of the accuracy of the underlying models.