The digital twin begins life as a prototype model, intended to match the process, but possessing ideal attributes that need to be adapted to the actual process. First, one must adjust model coefficients and constituent relations to best match the model to actual process operation. Second, over time as the process is reconfigured or substantially altered, reflect those process changes in the model. This may also require adjustment of model coefficients. Third, as the process operates, there will be continual changes in features such as ambient heat losses, heat exchanger fouling, catalyst reactivity, friction losses in piping, raw material properties, etc. Incrementally adapt the model to match the local reality.
This is part 2 of a three-part series on the digital twin. Part 1, in the October issue, defined a digital twin as a model of the process that's frequently adapted to match data from the process. This keeps it useful for its intended purpose. Here, part 2 discusses how to adapt the model. And part 3, in the December issue, will discuss tempering adaptation in response to noise.
Do not attempt to code the entire composite model at once. Make the model modular. Construct each elementary object as its own model, coded as its own subroutine or function block. Use its output as the input for another sequential object module. Build the process digital twin as a collection of interconnected subroutines. This permits the individual models to be easily adapted with limited data, to be edited, and to be understood.
