Data fusion algorithms for assessing sensors’ accuracy in an oil production well : a Bayesian approach

Master's thesis in Information technologyOil industry faces an underutilization problem of the captured data during the extracting process. This issue is a consequence of the lack of information regarding sensors’ accuracy. One effect can be a serious obstacle in the development of computer assisted decision systems. In a production well, it can be experienced the inexistence of sensor redundancy and enough information to assess credible probabilities. In this situation, we have to strongly depend of the experts’ ability to provide alternatives based on their understanding. These skills can be a critical limitation and turns particularly difficult the establishment of a prediction model. With this work we propose a Bayesian Network approach as a promissory data fusion technique for surveillance of sensors accuracy. We proved the usefulness of this method when it seems there isn’t enough feasible data to construct a model. In presence of certain data constrains we suggest an inversion of the causal relationship. This approach can be a possible solution to help the expert in accessing conditional probabilities