Evaluation of biological wastewater treatment process using Mahalanobis distances in original and principal component space: a case study

Petr Praus

doi:10.1007/s13201-018-0794-7

Abstract

The evaluation of wastewaters treated by biological wastewater treatment plant was performed using multivariate analysis. The samples taken during 1 year were characterized by the Mahalanobis distances (MDs) calculated from 11 original parameters and 4 principal components extracted by principal component analysis. The principal components were interpreted using Ward’s hierarchical clustering analysis and factor analysis. Statistical processing of the samples by means of the MDs calculated in the original and PC space was found to be complementary. Since MDs were not normally distributed, the statistical analysis of their log-transformed values (logMDs) was preferred to common Hotelling’s or Chi-squared statistics of MD2 ones. The outliers were confirmed by Ward’s method and by inspection of their chemical composition. In contrast to complexity and different magnitudes of the original wastewater parameters, the logMD charts provided a simple and effective tool for the evaluation of biological wastewater treatment process.

Highlights

Quality monitoring of processes has a long tradition in many technical and economic fields (Montgomery 1980, 1996)
The aim of this paper is to demonstrate the utilization of the Mahalanobis distance for the characterization of treated wastewater composition
The treated wastewater samples taken at the outlet of BWWTP during a year were evaluated by multivariate statistical analysis

Summary

Introduction

Quality monitoring of processes has a long tradition in many technical and economic fields (Montgomery 1980, 1996). Some applications of statistical control in non-industrial processes and in environment monitoring were reviewed by Cobert and Pan (2002). Shewhart’s control charts (Shewhart 1939) of selected individual variables were used for the evaluation of sewage treatment stations (Berthoex et al 1978; Orssatto et al 2014) and river water quality (Iglesias et al 2016). Univariate control charts of individual parameters can lead to erroneous conclusions. They may incorrectly identify out-of-control situations (Montgomery 2009). The application of Hotelling’s control chart in the monitoring of BWWT process was referred by Capilla (2009)

Objectives

Methods

Results

Conclusion