Towards real time monitoring of reacting species and pH coupling electrical resistance tomography and machine learning methodologies

The development of smart sensors capable to analyse data gathered on the process line and to give a real time feedback has been undergoing extensive research in the last years due to its potential benefits on the process optimisation and products improvement. In this paper, a novel approach to detect and monitor pH and conductivity using 2D electrical resistance tomography (ERT) is proposed for the first time in a reacting system. As a study case, the reaction between phosphoric acid and potassium hydroxide in mediums of both water and sodium carboxymethylcellulose (CMC) aqueous solution was assessed. The information gathered using the ERT have been used to determine local and overall mixing time for a sequence of injections of base and acid to understand the overall performance of the system. In addition, the same information have been used to extrapolate live data about the variation of the pH coupling the ERT data and machine learning techniques. Three different approaches have been investigated to achieve the aforementioned objective all integrating ML to the data processing. The first two approaches did not provide satisfying results showing the limitation of a completely blind approach (pure statistical approaches). However, the last approach, which combined ML technique and physical/chemical knowledge, showed very successful results for the real time monitoring of the pH in a reacting system.