(103a) Using Measurement Uncertainty Information for Effective Model Development

Obtaining and handling uncertainty information is still a challenge in the Chemical Processing Industries (CPI). Typically, among the variables most affected by uncertainty, one typically finds the process outputs, comprising concentrations (main product and sub-products, reactants, etc.), measurements of quality properties (mechanical, chemical, etc.) or other relevant information about the end use of the product. With the increasing flexibility of processing units, these quantities can easily span different orders of magnitude and present rather different uncertainties associated with their measurements. This means that heteroscedasticity in the process outputs (Y’s) is a rather prevalent feature in CPI, which must be properly managed and integrated in all the tasks that make use of process data. In this presentation we address this critical issue, by considering two sub-problems in which it can be decomposed: i) estimation of measurement uncertainty in heteroscedastic contexts; ii) integration of uncertainty in the outputs in model development. We present modelling approaches that describe the heteroscedastic behaviour of the Y-uncertainties in the operational domains of the measurements of interest and explore their interpolation and extrapolation capabilities and limitations. We also present the results of a Monte Carlo study where the added value of integrating different levels of information regarding the uncertainty affecting the output measurements is comparatively assessed. As a result of this study, valuable information is gathered on the benefits of collecting and using uncertainty information, which can be considered in decision making as a trade-off for the additional work the use of uncertainty information may encompass.