Model based control

Wrong automation focus

My observation is that automation focus often is on control hardware and tools selection, rather than control system performance. It’s like an author focusing on computer brand for writing a novel instead of focusing on novel content.

I want this site to contribute to focusing automation content; not so much the equipment and the tools used.

An example illustrates this: A chlor-alkali plant owner experienced plant trips almost once a week. That is expensive in terms of profit loss and other negative things. The problem was the brine mass-balance control. My company Prediktor made a study. After some Matlab simulations we came up with a solution. It took us 5 minutes to implement it in the Honeyewll DCS system that controlled the plant. The problem disappeared. The profit increased by more than $ 20 000 per week.

About me and my background

This site is made and maintained by me: Steinar Saelid. I have been working in the control- and automation field for more than 40 years, starting with my masters degree in Sensitivity Analysis in Optimal Systems in the 1970ties and a PhD in the field of Cement kiln modelling, estimation and control. 

cement-production

 

The background for the latter was some nasty oscillations in a kiln in a real cement plant and the quest for better control. My conclusion after extensive modelling and simulations was that the oscillations were difficult to avoid by control.

Steinar SaelidThe reasons for the oscillations were exessive CO2 gas development in the first part of the rotary kiln causing a significant batch of material to flow into the heat zone of the kiln and starting off the kiln oscillating behaviour.

I learnt a lot from this, even if I did not find a control based solution. This was a problen caused by design.

Later I have worked as a professor (full and adunct) at NTNU (the Norwegian University for Science and Technology) in Trondheim, Norway. I teached and researched during these years in Modelling for Control, System Identification and Estimation, Model Predictive Control and Stochastic Systems.

Dynamic positioning - model based control

Dynamic positioning Kalman filter - model based control

I moved to the industry and worked with delelopment of Dynamic Positioning systems (voted as the greatest engineering achievment in Norway, ever), development of a DCS system and various control system realizations at Kongsberg Maritime.
At Hydro (later split into Yara, Hydro and one part joining Statoil) I worked with model based operator systems for fertilizer production, model based control in the aluminum production sector and also in the oil and gas sector.

From 1995 I started the company Prediktor in Norway together with Helge Mordt and Rune Storkås. I have worked with modelling and control in Prediktor in all these years, and I am now a CTO of that company.

On control theory and education

A lot of books on theory for engineering students and for researchers in the field of estimation and model based control exists. The focus of this website is on the application of mathematical models for real time estimation and control and how to select appropriate models for such use. The purpose is also to show real applications or potential applications of models in this context by maintaining a blog containin posts on various themes related to this.

During all the years working in the control and automation field, I have been aware of the difficulty of making suitable models for use in estimation and control. When teaching topics at the university it is easy to think that a model for a system is a fixed entity, in the same way as the heat diffusion model is universally valid.

This is a very nonproductive attitude. The task of finding/developing a model for use for control purposes do not have one definite answer. A real system might be modelled in very different ways. The art is to find a SUITABLE model for a purpose. Not too complicated, but still capturing the important aspects important for control purposes.

This site is about selecting such models and how to apply these for real world control, analysis and estimation problems, and also in general on automation, production execution (MES and MOM) and plant operation optimization.