top of page
  • Writer's pictureHiren Shethna

Digital Twins in Process Industry - their ease of use?

As we work through building and supporting digital twins with our clients, the complexity has become clear. But before we make the point of the complexity, let us understand the purpose and the meaning of a digital twin.

A digital twin is essentially as good a replica of the real plant as it can be. Okay so that is simple. For civil and construction engineers might think of this as a 3-D model of the plant that is there, with the intricate details of the piping and vessels and foundations, and can be used to understand whether the plant can withstand stresses during high winds or earthquakes, etc, for example.

Okay but I am a process engineer, and my world is process engineering. In my world, the digital twin is a mathematical model of the process which mimics the plant behavior (and misbehavior) - i.e. how plant responds to changes in feed quantity and quality, how does the plant respond to changes in operating conditions and ambient conditions, how can we manage or prevent an upset, and how far can we "push" the plant safely but get more value out of it. All these questions and more can be answered safely from the comfort of your own (home) office without touching the real plant if you have a good digital twin (mathematical model).

A key requirement for a good digital twin model needs the ability to “self-learn” from known plant behavior – i.e. adjust its internal constants, just like the weights and biases in an Artificial Nueral Network. A well-learned digital twin model can then be pushed against the plant limits (usually the Advanced Process Control limits) for maximizing certain objectives, be it profit or energy or some other thing that the process or control engineer deems fit. After all, the twin is a tool to assist the engineer. It is here that the complexity question arises. If the digital twin is not easy to use, will it be helpful to the engineer. Our realization in this important question is that, if the plant that you are operating has a complex set of interactions, a good digital twin will also carry with it those interactions and there is no escaping that complexity. And for that the years of experience with the plant will come to one’s assistance. But for the software OEM’s it would be good if the work processes in terms of using the digital twins are made as intuitive as using the DCS control room screen which is the common denominator for all plant engineers. For example, can I view and navigate the unit operation graphics in an identical way as the DCS, or can I search the variables in my digital twin with the same name such as “10FI203.PV” for example. These are just some of the many workflows that can make the adoption of digital twin easier.

But to reiterate, the paramount requirements that top the list is the ability to “self-learn” the plant conditions and to duplicate the complexity of interactions of the real plant.

104 views0 comments

Recent Posts

See All

Heat Exchanger Analytics and its economic impact

Our paper on 'Heat Exchanger Fouling Analytics' has been published in the Q3 Issue 2020 of the Petroleum Technology Quarterly magazine. Please spare some time to go through it especially if you're a C


bottom of page