Interval-Based Diagnosis of Biological Systems - a Powerful Tool for Highly Uncertain Anaerobic Digestion Processes

Abstract

Anaerobic digestion (AD) is a highly nonlinear time-varying process commonly used for biological wastewater treatment, which is subject to large disturbances of both influent concentrations, and flow rates that may lead the process to a breakdown. In order to compensate the effect of these disturbances, the dynamics of the main state variables – including biomass – must be closely monitored and used to improve the process performance. However, AD processes still suffer from a lack of reliable and cheap sensors of key process variables to insure the right process operation. This has led to the development of estimation schemes, which infer the information of such key variables from the available measurements. Nevertheless, reliable measurements are not always possible to get because these readings may be corrupted by noise or erroneous due to sensor failures and as a consequence, they may lead to deteriorated control efforts and the eventual crash of the AD process. In this article, we propose an integrated system for the detection, isolation, and analysis of faults in AD processes by using interval observers (IO). The proposed approach was experimentally implemented on a 1-m3 pilot scale anaerobic digester. Based on the comparison between the measured outputs and their corresponding estimates, results show that this approach was able to detect sensor failures as well as faults in the basic hypotheses made during the design step.