This thesis is concerned with simultaneous estimation of the time delay and system parameters of continuous-time transfer function models. Estimation of the delay is different from the estimation of the rest of the parameters due to the fact that the delay does not appear explicitly in the model equation. The initiative undertaken in this research is based on the idea of bringing the delay term within the param- eter vector. The idea was facilitated by a specific formulation of a linear filter for continuous-time identification. A new filter structure has been proposed and the en- suing iterative method has been developed in a novel way to estimate the delay plus other system parameters.

Over the last few decades, a significant number of new methods and techniques for system identification have been developed. More theoretical aspects of the identifica- tion problem such as the convergence of the parameter estimates have been addressed in much details. However, many practical problems in real world applications of dif- ferent identification techniques remain unsolved. In this work, aspects of system identification with respect to practical implementation are considered.

Lack of availability of uniformly sampled data is a common yet often overlooked prob- lem in real industrial data. A simple algorithm for identification from non-uniformly sampled output data has been proposed based on the idea of model based prediction. Techniques based on step test are commonly used in process industries for identifica- tion. Novel identification methods based on open loop and closed loop step test data have been proposed in this thesis that use raw industrial data without preprocessing. The methods are applicable even if the step input is applied when the process is not at steady state.

A multiple input multiple output (MIMO) model identification method is introduced that involves transformation of MIMO data into its single input single output (SISO) equivalents and uses SISO model identification algorithms for the purpose of identification. Model validation is a complementary step to the identification exercise. A validation scheme for SISO and MIMO continuous-time models is also presented in this thesis. The proposed method is applicable for models with delays.