MODELO PROBABILISTA DE DAÑO POR CORROSIÓN EN DUCTOS TERRESTRES DE TRANSPORTE DE HIDROCARBUROS

Abstract

In this work is presented a probabilistic model of cumulative damage, it is based on Markov chains theory to model the depth pitting inside corrosion propagation in a hydrocarbon transport pipeline that have an inspection report. The cumulative damage model is a stationary Markov process, of discrete time and finite states. The damage cumulating mechanism is unit-jump type and depends of the state. Is used the idea of the shock model, based on Bernoulli trials and probabilities to stay in a state or to go to the next. To obtain the transition probability matrix was necessary to analyze the inspection data, to fit the distribution function and to obtain probabilities to stay in the different discrete damage states measured to inspection time. The data was fit to Lognormal distribution, and it was checked with a Kolmogorov-Smirnov test. It was developed the equation that result from multiply the initial state vector by the matrix transition and was result the equation system to find everyone to transition probabilities. To known the inside corrosion propagation after to the inspection was proposed an equation and was fit to the tendencies known at that moment. So it was possible to calculate the transition probabilities until the failure and to know the mean of the time spent in every state using the Bogdanoff and Kozin theory of cumulative damage. To know the mean of the time spent in every state is possible to know the mean of the time of failure. The time spent in every state was compared with Monte Carlo simulations and was find similar values, and it is possible to say the model is consistent. Every step of time was fit to the real time with the data mean and the known time of pipeline operation.