A non-linear dynamic model identification challenge problem

A non-linear dynamic model identification challenge problem is proposed. The problem is to identify a dynamic model from single input single output data from a laboratory surge tank system that was built expressly for generating non-linear input output data. The problem is divided into six separate tests that challenge the ability of the modeling method to produce a model that interpolates well, degrades gracefully when extrapolating and is robust to noise in the identification data. Instructions are given for the development and testing of the models and for reporting the results. It is hoped that this problem will lead to a more constructive comparison of non-linear modeling methods than has been seen to date.