To getting accurate results from numerical analysis, the results must be approaches to experimental tests result. What methods that can be used to reduce the difference between the experimental and numerical analysis results?
The quality of input determines the quality of output. My experimental input consists of 20 pairs of values stress-strain. The theoretic material models in Ansys Mechanical usually use one stress value (strength, peak) and the slope before the peak (young's modulus) - another stress value.
Article Project CC Modelling and quasistatic simulation of the stiff...
In mechanical labs, experiments are typically designed to be as simple as possible to focus on a certain parameter of interested. Take an example of loading on a beam, a point load and distributed loading will give different moment and shear distribution across the beam. Another example is on the boundary condition, compare simply supported beam at two points, to a beam with two fixed supports.
Modelling the accurate experimental representation numerically is the first step to modelling. If a wrong loading and/or boundary condition is specified, the numerical model is describing a different problem from the one being investigated and will not give a accurate result.
Of course there will be uncertainties due to parameters not typically investigated, such as air density in mechanical tests. These parameters will definitely affect the results, but typically negligible difference. It is therefore important to understand the problem of interest, and apply the appropriate fundamentals to solve them.