Date of Award


Document Type

Master Thesis

Degree Name

Master of Engineering (Research)


Electronic Engineering

First Advisor

Dr. Tom O'Mahony

Second Advisor

Dr. Barry O'Connor


The primary focus of this research is to identify a linear model for a direct steam injection (DSI) pasteurisation process which is a component of a large-scale pilot evaporation process widely used in the dairy industry. The pasteurisation process has been identified as a key component of the evaporation process. The author believes that no previously published work has focused on identifying models for a DSI pasteurisation process which makes this work an original contribution to the area. The pasteurisation process can be modelled as a single-input, single-output (SISO) process with a measurable disturbance. Thus two transfer functions were required to be identified, GD(s)relating the dynamics of the disturbance to the output and, Gp(s) relating the dynamics of the input to the output. Initially, the focus of this work was on the use of operating data to estimate models, however a thorough analysis of the results raised suspicions regarding model accuracy and resulted in low user confidence. Thus specially designed system identification tests were applied and the resulting model has been proven to be accurate over a 65°C to 75°C range (which is the operating range most frequently used). It is believed that substantial energy savings can be made from a re-tuning of the PID settings using the identified models in this thesis.

Because the sampling rate of the data was high, compared with the fundamental dynamics of the process, there were concerns regarding the integrity of the discrete-time results. Alternative methods of determining continuous-time models were investigated which identified direct continuous-time methods as a viable alternative. Therefore, a second contribution of this thesis is the evaluation of direct continuous-time versus indirect continuous-time system identification methodologies. Specifically, the Box Jenkins (indirect method) model and the Simplified Refined Instrumental Variable (SRIVC) (direct method) model were used for analysis purposes. The results indicated that models with smaller standard deviations were estimated using the SRIVC method - implying that these models are more accurate. Furthermore, the low standard deviations resulted in high user confidence in the estimated models. To date only one previous comparison between these approaches has been performed and therefore the research reported here forms a significant contribution to this research area.


Submitted to Cork Institute of Technology in partial fulfilment of the requirements for Degree of Master of Engineering

Access Level