The exploration and production of oil and gas lead to many logistics challenges. In the case of the Brazilian offshore production, the operation size and the fact that the exploration occurs up to 300 km apart from the coast make these challenges even greater. There are many thousands of different types of materials, hundreds of suppliers, as well as hundreds of different destinations served by distinct routes. In short, this is a complex system, in which it is necessary to deal with an intrinsically combinatorial problem of sharing resources, such as warehouses, ports and means of transportation. Logistics processes need to deliver the materials in time, making sure that the production is not affected.
In addition, operational costs and immobilized capital must be minimized. It is crucial to evaluate distribution probabilities for lead-times, identify bottlenecks and predict the effect of adopting specific policies for supply, picking and transportation. In order to address these issues, it is necessary to answer complex what-if queries that take into account the temporal relations, either specified or dictated by the process dynamics, between the events.
Since all operations are logged, a substantial amount of historical data is generated. However, these data do not necessarily cover all situations, simply because certain feasible and relevant combinations of events may not have occurred during the period that data is logged. This motivates the use of simulations that generate huge amounts of data augmenting the (logged) historical data, and making big data analytics necessary. In this paper, we briefly describe the process of model building based on historical data as well as the construction of a simulation engine that permits efficient large scale simulations. The simulation results, together with the logged historical data are subjected to big data analytics in order to create global prediction models.
The proposed methodology aims to answer complex what-if queries about the logistics processes with a high degree of efficiency and prediction accuracy. The software tool, based on this methodology, is designed so that a decision maker can interactively detect critical situations and study the global effect of changes in policies. Some examples of queries that can be supported by this research are: (i) estimation of distribution probabilities for lead-times under varied circumstances; and (ii) probability of critical materials shortage during periods of high demand.