Data reconciliation combined with data evaluation by sensor networks

We developed a data evaluation technique that allows us to evaluate dynamically a kind of reliability of data from sensors by analytic redundancy among process variables. In this paper, we extend the data evaluation, combined with data reconciliation technique to handle the situation when both measurement error and gross error may exist. When gross error exists, data evaluation is used to filter out inadequate constraints used in data reconciliation. On the other hand, when measurement error exists, data reconciliation is used to filter out noise and hence provides data evaluation with clean relation among data. It is demonstrated by simulations that the combined method enhances diagnostic performance compared to that when both methods are applied independently