Using statistical models to detect leaks in underground storage tanks

Abstract

In modern service stations, intermediate grades of gasoline are formed by blending different volumes of unleaded and super unleaded gasolines stored in underground storage tanks. The volumes displaced from these coupled tanks can be modeled in a two-dimensional model as a function of the volumes dispensed through the flow meters located at the pumps (or positive flow displacement meters). The intercepts in these models represent the sizes of leaks in the respective underground storage tanks. Inferences on the vector of intercepts are discussed based on a regression model, a measurement-error model and a sequential model. In the last model, bivariate control charts are used to monitor the leak status of the tanks. Copyright © 2000 John Wiley & Sons, Ltd.