Sensor Network Reallocation and Upgrade for Efficient Fault Diagnosis

Fault diagnosis is a pre-requisite for ensuring safe, efficient and optimal operation of chemical process plants. The success of any diagnosis strategy depends critically on the sensors measuring the process variables. With potentially many sensor locations, sensor placement can be optimized based on criteria like cost, reliability etc. We present formulations to perform sensor reallocation and upgrade of an existing sensor network to ensure comprehensive fault diagnosis. These formulations are based on minimizing cost required to achieve the desired reliability, and maximizing reliability for a given cost. A combined formulation which maximizes reliability and, among the various solutions, selects the one with minimum cost is also presented. The cause-effect information for these formulations is obtained from a semi-quantitative representation of the process which might be useful for performing sensor location for industrial processes. The utility of the proposed approach is demonstrated through a retrofit analysis of the Tennessee Eastman case study.