An algorithmic process used within data reconciliation to identify and isolate non-random errors, such as sensor biases, physical leaks, or instrument failures. Detecting these anomalies prevents corrupt data from skewing plant optimization models and performance assessments.