DETECTION AND FAULT DIAGNOSIS OF BLOOD GLUCOSE LEVEL THROUGH NON INVASIVE TECHNIQUE

Abstract

Several works have proposed in the past alternative control strategies for regulating the blood glucose levels inside a healthy range for patients who suffer Type I Diabetes Mellitus (T1DM). All of them considered the availability of accurate measurements of glycaemia, generally recorded from invasive methods. However, a common  problem of this methodology, since the control point of view, is the variable delay depending of the type of sample extraction, that is subcutaneous (SC) or intravenous (IV). In the last decade, intensive research has been focused on developing and testing non-invasive sensors of blood glucose levels. However, it has been reported some faulty or inaccurate measurements with them. Up to now, nobody has considered the use of tolerant control for this kind of systems.  Hence, in this paper a fault diagnosis and identification (FDI) system based on the Discrete Wavelet Transform (DWT) is used to detect faulty behaviour from those non-invasive sensors. The FDI is integrated to the Predictive functional controller (PFC), turning it as fault tolerant control which supports the decision about the correct insulin dosage. Several simulation results are presented by employing a rigorous endocrine model which can emulate the interaction among glucose-insulin-glucagon.