Identification of nonlinearities in dynamic process systems
Process modelling is an essential element in the development of advanced (model-based) process control systems, accounting for up to 80% of the cost of development. Often, models based on historic process data are the only viable option when dealing with processes that cannot be modeled cost-effectively from first principles. Despite major advances that have recently been made in the field of nonlinear process modeling, some basic idea of the dynamic behaviour of a process system is important prior to system identification. While many process systems seem unlikely to be linear, the possible nonlinear aspects of their dynamics are not necessarily supported by the data. This may have significant implications for subsequent system identification or interpretation of the data. For example, it is quite possible that an underlying nonlinear deterministic signal may be obscured by a high degree of measurement or process noise, to the extent that it may be critical in the choice of model to fit to the data. The method of surrogate data analysis has recently been proposed as a means to classify the underlying dynamics of complex systems, and in this paper the use of Fourier transform surrogates to classify process systems is explored.