Interval-Valued Data Learning for Robustness of Energy Recovery Systems

Rachid Ouaret (1) , Pascal Floquet (1) , Jean-Pierre Belaud (1) , Stéphane Négny (1)

1 LGC - Laboratoire de Génie Chimique

Modeling of energy recovery systems, Heat Exchanger Networks (HEN) as an example, are complex processes due to many parameters involved and the lack of knowledge on the impact of operational variables on the response. These variables are often affected by different types of uncertainty as to measurement errors, computation errors, or imprecision related to the underlying method. The uncertainty in the data may be treated by considering, rather than a single value for each variable, the interval of values in which it may fall, histograms, or other multivalued data: symbolic data. This work aims to use, when possible, the symbolic data analysis to adapt the classical mathematical HEN models. It deals with the study of continuous interval data through suitable Principal Component Analyses and Regression for two purposes: clustering exchanges (i) classification of exchangers to detect those impacted by uncertainty factors and (ii) evaluation of the relationship between the different process parameters (inlet temperature, heat transfer coefficient, etc.) on interval data. The new method has been tested on a real data set and the numerical results are reported. The symbolic approach provides a simple way to study a great number of scenarios.

Mots clés: Machine Learning for Symbolic Data; Interval-valued data; Flexibility and Robustness; Heat Exchanger Networks (HEN)

Domaines: Statistiques [math.ST]; Probabilités [math.PR]; Applications [stat.AP]; Machine Learning [stat.ML]

Rachid Ouaret, Pascal Floquet, Jean-Pierre Belaud, Stéphane Négny. Interval-Valued Data Learning for Robustness of Energy Recovery Systems. 2020. ⟨hal-03045777⟩