Development and use of statistical process control (SPC) techniques for real-time interpretation of wastewater treatment plant data
Abstract
The two Indianapolis Advanced Wastewater Treatment (AWT) plants are complex, computer controlled facilities designed for tertiary oxidation of carbonaceous BOD and ammonium-nitrogen (NH$\sb4$-N). Since 1982, distributed process control computers have assisted the operations staff in maintaining compliance with stringent effluent discharge permits. Along with providing clear benefits, current computer technology (such as employed in Indianapolis) has created problems characterized largely by the overproduction of real-time process data. This "data deluge" hinders an operator's effective use of this computer system. The ability to qualitatively diagnose process parameter shifts or trends that indicate a subsequent overload or upset condition in the nitrification systems was beyond the capability of the computer. This research effort addressed this problem by developing and implementing a diagnostic algorithm which evaluated key nitrification parameter trends in real-time. The study involved the extensive use of both daily average data and real-time data. Descriptive and correlative statistical analyses were performed on representative daily data sets to develop a foundation for the subsequent formulation of diagnostic rules. A Statistical Process Control (SPC) approach leads to the use of single and multiple modified Shewhart x control charts to detect corresponding parameter shift patterns. Control chart limits for each nitrification parameter chart were based on the standard deviation obtained from the results of ARIMA time-series modelling and a smoothed 24-hour moving average. An evaluation of the multiple control chart technique using episodal data was then presented and discussed. A retrospective approach to discerning contributing factors to prior nitrification upsets was accomplished with Pareto analysis. This technique characterized 64% to 79% of the identified upsets. Low reactor temperature, high flow, high specific NH$\sb4$-N loading, and inhibition were among the leading nitrification upset factors. Overall, this research effort involved the development, testing, and implementation of an innovative computer-based diagnostic inferencing technique. These diagnostics were both visually and verbally generated, representing a unique approach to optimization of the human/hardware interface.
Degree
Ph.D.
Advisors
Alleman, Purdue University.
Subject Area
Civil engineering|Sanitation|Operations research
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.