Reliability Estimation for Sensor Networks in Chemical Plants using Monte Carlo Methods

Abstract The aim of this work is to analyze and determine an efficient and flexible resolution methodology to address the design of a minimum cost instrument network subject to restrictions over a set of key variables. In this sense, a Monte Carlo Simulation method is proposes to evaluate the reliability of the network defined as the probability of continuing to observe the keys variables when the instruments fails according to a given failure model. This is a powerful technique to model this stochastic behavior of systems and components. The optimization engine chosen is a heuristic, Simulated Annealing, which has shown to have a good performance for this kind of problems. Industrial examples of increasing complexity are provided to show the efficiency of the algorithms.