Optimization validation of online system for COD variation measurement in wastewater at levels of parallel matrixes designed through bias correction by Deming regression under AD hypothesis test

Abstract

Abstract Chemical oxygen demand (COD), with its unique socio-economic structure and conditions in regional wastewater of Dalian, China, can bring about the long-term potential harm impacts to ecosystems and human health. As the key indicators of total sewage discharge control and the must-measure item for routine testing in China, COD test is basically completed by the online system (as X-method). However, X-method belongs to the domains of non-standard, and the validity of its reported results deserves attention. The background effect treatment has been debated when dealing with the accuracy of COD test, and the usual practice is referred to the matrixes spikes at multiple levels. However, we hold the opinion that COD of the actual extraction efficiency (naturally existing matrix matched), is out of the question while measuring for unstable samples greatly fluctuated in wastewater, with its instantaneous effluent and temporal and spatial characteristics, and cannot be repeated and predicted. Instead, we decided to use regression to solve this problem. We may assess the methodology of leverage (hi) and Cook’s distance (Di) to identify influential observation prior to model transformation for ordinary least squares (OLS) fitting, in which, the random parallel samples designed, within each level consistent differing by ≤ 10%, is deployed to establish the bias correction reasoning between X-method and the arbitration, referee or primary system (as Y-method). As OLS fitting is, yielding biased and inefficient estimates, not suitable for uncertainties existed in both regressed variables, instead, we use a knowledge-based Deming regression (DR) to optimally monitor and validate the validity of X-method system. In this paper, we give the detailed fitting process of DR technique with its weighted iteration accounted for measurement error in both methods. To ensure the residuals, in chronological order, deduced from the bias correction function, under independence identical distribution (i.i.d) condition, we strongly advocate a more robust Anderson Darling (AD) hypothesis test for validation of X-method. If the AD null hypothesis is held, we further hold the opinion that X-method is, under the site precision (sR'), reliable for its COD determination in wastewater. Meanwhile, the sR' is more appropriate to maximally incorporate all cumulative effects, even with the annoying interaction, into the data quality objective (DQO) of the COD system over an entire range levels range, and minimize the intractable problem caused by matrix effect. Our motive is to compare the COD variation measurement from real-time online system, across all matrices, with the results obtained historically or subsequently, to meet client’s needs in a way that allows the operation of the COD system from X-method in consistency, in impartial, in competency, and to assist management decision making.