Metrological assurance of digital twins of chemical production process lines at development, commissioning, and maintenance stages

Metrological assurance is considered for digital twins of chemical process lines used for operational prediction of equipment condition, deviation detection, operating-mode selection, and support of process-safety decisions. The reliability of calculation results obtained using a digital twin depends not only on the adequacy of the computational model of the technological process but also on the metrological condition of measurement channels, data synchronization, and traceability of the data and procedures used to determine the parameters of this model. Existing approaches do not integrate these determinants of calculation-result reliability with validation of the computational model of the technological process and operational control. A methodology is developed for the metrological assurance of digital twins of chemical production process lines at the development, commissioning, and maintenance stages. The methodology includes uncertainty evaluation for calculated output quantities, metrological traceability of input-quantity measurement results, time-synchronization control, validation of the computational kinetic model of the reaction unit against control data, and data-quality assessment. Its effectiveness is demonstrated using a model-based example of a continuous isothermal stirred-tank reactor with a first-order reaction. An uncertainty budget is developed for the calculated concentration of the target component, and estimates obtained using the linear law of uncertainty propagation and Monte Carlo simulation are compared. Scenario experiments demonstrate the effect of the metrological condition of input data on digital-twin suitability. Transition from operation without metrological assurance to metrologically assured operation reduces the expanded uncertainty of the calculated concentration from 0.042 to 0.016 kmol/m 3 . The standard uncertainty associated with the activation-energy estimate accounts for the largest share of the output variance, 42.7 %. The effect of a mismatch between the time stamps of input-quantity measurement results and the corresponding states of the reactor computational kinetic model is investigated for steady-state and transient reactor operation. In the transient mode, a 60 s mismatch increases the expanded uncertainty to approximately 0.026 kmol/m 3 . The structure of a digital metrological passport and suitability criteria for monitoring, diagnostics, and adaptive correction of the computational model of the technological process are proposed. The methodology can be applied at the design, commissioning, and maintenance stages of process-line digital twins by process engineers, metrologists, specialists in the digitalization of chemical production, and modeling-software developers.