Diseño de sensores virtuales con corrección de retardo fisiológico para el monitoreo de glucosa en pacientes con Diabetes Mellitus Tipo I

The design and implementation of virtual sensors (VS) is a task often necessary for the reconstruction of variables that are difficult to measure online. However, all the problems associated with the available measurements (noise, uncertainty, delays, among others) negatively affect the performance of VS. This paper addresses the problem of measurement delays for the design of virtual sensors. The main contribution consists of a virtual sensor design methodology for a class of nonlinear systems, over which measurements are obtained with an unknown but bounded delay. The proposed methodology integrates the techniques of maximum correlation analysis, the Unscented Kalman Filter, and the Smith Predictor to estimate and correct delay times. As a result, a new cascade estimation scheme was obtained, which allows for obtaining better performance than a conventional estimator, which does not consider the delay in the measurement. The proposed methodology was validated by simulation with insilico-generated data, for the case of glucose monitoring in patients with Type I Diabetes Mellitus.