Thesis for the Degree of Master of Science


Year 2005
Tommi Rönkönharju

Design and implementation of an operators’ support system for managing disturbances in industrial processes

Operators are still actively involved in the controlling of processes, even though the level of automation has been increasing steadily at industrial sites. Operator interventions are especially common during process disturbances. There would accordingly be demand for a system, that could memorize the best operator decisions for re-use when similar problem situations arise in the future. The utilization of such a system would enable better and more reliable operator actions during process disturbances. This assumption is well founded, since the system evaluates the relational quality of operator actions automatically and all recommendations provided by the system are based on accepted operator actions of the past. The aim of this thesis was to develop a dynamic operators’ support system that could be used for these purposes. The literature survey part of this study discusses theoretical aspects of decision support systems and by what means and methods they have been developed. The main emphasis is on the process history-based methods (case-based reasoning cycle, ART-Kohonen neural networks, principal component analysis etc.). Based on the some of the studied methods, a software application was created, and its functioning was tested on simulated paper machine runs. The chosen disturbance situation was a grade change, since in grade change situations operators commonly intervene with the process controls to speed up the change to the new desired grade. The test runs were carried out with simulator software (APROS Paper).The efficiency and usability of the implemented software was examined in the light of special parameters, which were derived from the program’s operation during the test runs. The evaluation of the results mainly consisted of demonstrating the evolution of the application’s operations in the long run, since any occasional result in this case is greatly dependent on the arbitrary nature of the simulation runs. It could be shown, that the program functioned as desired. A great emphasis was placed on the computational efficiencies of the presented algorithms, but they were not analyzed here, since they were beyond the scope of this study. Nonetheless, the benefit of using the presented search/store-algorithms is evident if this kind of operator support system is to be generated.

This info last modified 16 Jul 2018 by Jerri Kämpe-Hellenius