COMBINING HARD AND SOFT MODELLING TO ESTIMATE KINETIC RATE CONSTANTS IN CURING REACTIONS MONITORED WITH NIR

M. GARRIDO; M. S. LARRECHI; F. X. RIUS

Barcelona

Jornada; XI Jornadas de Análisis Instrumental; 2005

Asociación Española de Química Analitica

Current research on epoxy resins focuses on improving their characteristics in order to respond to the requirements for new and advanced applications. To understand the curing process and the relationships between structure, property and processing for manufacture and use of epoxy thermosets, we need to determine the reaction mechanism and quantify the kinetic parameters. An important issue in epoxy-amine kinetics is the change in reactivity of a secondary amine with respect to that of a primary amine, the so-called substitution effect (¦Ñ). This effect is normally calculated from the ratio of the reaction rate constant of an epoxide reacting with a secondary amine to that of an epoxide reacting with a primary amine. Knowledge of this value is crucial to adequately describing of the kinetics of network formation and therefore to better understanding the morphology of the cured structure. One of the most suitable technique for monitoring the curing process is near-infrared (NIR) spectroscopy since it is non-destructive and, if the appropriate accessories are available, it enables the cure reactions to be monitored in situ. Spectroscopic data from the monitoring of kinetic processes can be analysed using several techniques. Hard modelling involves fitting kinetic parameters of a well-known chemical model, which describes the mechanism of the reaction. Soft-modelling enables us to describe the process without explicitly using the associate chemical model and only requires a series of constraints that are inherent to the chemical knowledge of the system. The interaction between hard and soft modelling involves fitting the concentration profiles obtained by soft modelling, such as Multivariate Curve Resolution-Alternating Least Square (MCR-ALS), to a previously established kinetic model. In this study we compared two strategies. First, the concentration profiles obtained by MCR-ALS (the pure soft-modelling approach) are fitted to a kinetic model in a separate step (the hard-modelling approach). The other strategy incorporates the kinetic model as a constraint at each iteration of the MCR-ALS optimization.  We applied these strategies to NIR data from the monitoring of two model reactions (phenyl glycidyl ether (PGE)/aniline and Glycyciloxydimetylphenyl silane (GDMPS)/aniline) and discuss the results.