The process is conducted by first choosing the domain in which data is presented.
The third task is to formulate a function which has the parameters that are expected to be design variables, and which represents the distance between the measured data and the finite element model predicted data.
The fourth step is to implement the optimization method to identify parameters that minimize this function.
For nonlinear analysis, more specific methods like response surface modeling, particle swarm optimization, Monte Carlo optimization, and genetic algorithms can be used.
Recently, finite element model updating has been conducted using Bayesian statistics which gives a probabilistic interpretation of model updating.