Sensor networks design for Observability of non-linear chemical processes

RODRIGUEZ, LEANDRO; GUSTAVO SCAGLIA; SANCHEZ, MABEL CRISTINA

Mendoza

Congreso; Actas del Congreso Latinoamericano de Ingenierías y Ciencias Aplicadas - CLICAP 2022; 2022

The performance of any modern monitoring strategy depends strongly on the structure of the sensor network installed in the plant. One of the properties that most influences the process monitoring is the Observability, which is defined as the ability of a system to reconstruct the state variables given the outputs. In nonlinear dynamic systems, the computation of this property is normally carried out using a method based on Lie derivatives, which is not computationally tractable for complex systems. This drawback was solved in the bibliography using the Empirical Observability Grammian, however the effect of the loss of measurements on the Observability was not addressed. In this work, two robust sensor network design strategies are presented, which contemplate the incidental loss of sensors and its effect on this property. These design problems are solved using a level traversal search with cutting criteria algorithm. A copolymerization process is used to demonstrate the performance of the proposed techniques.