A Holistic Evaluation of Multivariate Statistical Process Monitoring in a Biological and Membrane Treatment System

Abstract

Unsupervised process monitoring for fault detection and data cleaning is underdeveloped for municipal wastewater treatment plants (WWTPs) due to the complexity and volume of data produced by sensors, equipment, and control systems. The goal of this work is to extensively test and tune an unsupervised process monitoring method that can promptly identify faults in a full-scale decentralized WWTP prior to significant system changes. Adaptive dynamic principal component analysis (AD-PCA) is a dimension reduction method modified to address autocorrelation and nonstationarity in multivariate processes and is evaluated in this work for its ability to continuously detect drift, shift, and spike faults. For spike faults, univariate drift faults, and multivariate shift faults, implementing AD-PCA on data that are subset by treatment processes and operating states with significant differences in covariates and whose model parameters use week-long training windows, moderate cumulative variance, and a high threshold for detection was found to detect faults prior to existing operational thresholds. To improve the consistency with which the AD-PCA method detects out-of-control conditions in real time, additional work is needed to remove outliers prior to model fitting and to detect multivariate drift faults in which the covariates change slowly.

Long-Term Statistical Process Monitoring of an Ultrafiltration Water Treatment Process

“…MSPM PCA-based methods used without any adjustments are referred to as static PCA, but many adjustments to PCA have been proposed to improve fault detection performance . These include dynamic PCA, , adaptive PCA, , and adaptive-dynamic PCA (AD-PCA). ,, Dynamic PCA accounts for autocorrelation in the data by including lags of detrended monitoring variables, adaptive PCA uses a rolling window approach to account for nonstationarity, and AD-PCA incorporates both adjustments, making it suited for monitoring multivariate autocorrelated data over long periods of time.…”

“… 37 These include dynamic PCA, 19 , 38 adaptive PCA, 39 , 40 and adaptive-dynamic PCA (AD-PCA). 33 , 41 , 42 Dynamic PCA accounts for autocorrelation in the data by including lags of detrended monitoring variables, adaptive PCA uses a rolling window approach to account for nonstationarity, and AD-PCA incorporates both adjustments, making it suited for monitoring multivariate autocorrelated data over long periods of time.…”

Applications of Artificial Intelligence, Machine Learning, and Data Analytics in Water Environments

“…Other contributions introduce novel methodologies that combine ML with statistical approaches to address various environmental problems, such as sewer overflow pollution abatement, fault detection in water and wastewater treatment, , an assay for source apportionment of per- and polyfluorinated substances (PFAS), detection of freshwater algae, and influent water data . Beyond water quality, ML has been applied to model water quantity, exploring dominant factors influencing urban industrial wastewater discharges, model energy consumption of wastewater treatment, identify endocrine-active pollutants in the organic Unregulated Contaminant Monitoring Rule (UCMR 1–4) and their toxic potentials, and employ quantitative biodescriptors to predict in vivo toxicity …”