Validation Of Numerical Models: Empiricism Vs. The Laws Of Physics

Reliable computer prediction of the test-bench performance of new Stirling machines remains elusive due to inaccuracies arising both from the conceptualization of the machine to be modelled and the numerical scheme employed. Attempts to validate a code by balancing off numerical errors with compensating errors of other kinds is risky at best. Numerical inaccuracies that arise from the finite number of space and time steps in a numerical model are presumably unrelated to thermodynamic losses in the machine that is modelled. More reliable results can be obtained with a numerically accurate model that minimizes empirical inputs. The price is extensive use of computer time to carry out the necessary context. calculations for a large number of time and space steps and abandonment of the attempt to predict brake performance directly.