Use of historical trending in the validation of hazardous waste site monitoring data

摘要

The Savannah River Site is a large nuclear weapons facility with over 2500 monitoring wells with an annual load of over 300,000 samples. Data are verified and validated using a program that performs routine data format checks as well as a statistical check on whether results fall within the expected range based on a trend of previous values. For analytical data, a flinear fit of the previous eight sampling events is calculated along with a predicted value and confidence interval. If the result is outside the confidence interval for a linear fit, then a quadratic fit is attempted and any value outside of the quadratic fit confidence interval is flagged as an anomaly. The algorithm takes into account 90th percentile detection limits and dilution factors. When there is not enough data on a well to predict a trend, the algorithm uses data from all wells that are in same well group. Use of this historical trending algorithm has provided tremendous improvement over the prveious method of performing the istorical check. False anomaly indications have been reduced by roughly 60percent.