Prediction of flammable conditions of a methanol storage tank using Neural Networks.

Methanol is one of the most important commodities in industrial chemistry and can be stored in fixed roof tanks provided with an inertization system. In this process it is necessary to continuously monitoring the oxygen concentration inside the tank, the measurement can be done by means of an in-situ sensor. As an alternative to continuous oxygen measurement, some inferential methods can be used. Recently, neural network-based models are used to solve several different problems such as pattern recognition, measuring system back-up, what-if analysis, real-time prediction for plant control, sensor validation and fault diagnosis strategies. In this work, a soft sensor for estimating the concentration of vapors inside a fixed roof methanol storage tank equipped with a nitrogen inerting system is presented. This soft sensor is built using a neural network model and trained using the data obtained through the phenomenological model of the fuel tank. Both the neural network and the tank model were developed using Matlab