Chemicals
Industrial production has objective challenges
Uncertainty and fluctuations
Physical models describe only general dependencies
Expertise of operators is not easily scaled
Process control is limited
  • Reactive in its nature

  • Operation within allowed "ranges"

  • Static rules instead of dynamic adjustments
With AI, operational decisions become "personalized"
  • Automatic predictions and recommendations

  • Generated individually for each iteration
Example case: Virtual measurement of incoming gas composition in gas fractionation unit
Client: large petrochemical company
2-6%
Average error in defining the incoming concentration of individual components (compared to the lab data)
Problem
Solution
Data Used
What is the current composition of the gas feed? What will it be in 15 min?
Sample model output
  • To efficiently manage the fractionation process we need to know gas composition

  • Laboratory measurements are available only twice per day. Chromatographs estimate only some components, with a delay and limited accuracy.
  • Create virtual sensors that estimate the chemical composition of the gas stream in real-time

  • Further use this data for timely adjustment of unit parameters
Data for 2 years:
  • Process parameters
  • Available gas measurements
  • Characteristics of the resulting product
  • Reference data
Extension plans
Additional models can be developed for gas fractionation unit based on virtual sensing

  • Estimation of current column load

  • Estimation of the current load of heat exchanger

At the next stage, a real-time recommender system can be built

  • Suggestion on optimal feed supply parameters, plate temperature change, reflux flow rate etc.

Similar applications can be developed for adjacent or downstream processes

  • Crude oil refining (atmospheric and vacuum distillation)

  • Pyrolysis during production of monomers (ethylene, propylene etc.)
Want to learn more?