Application of Machine Learning Methods for an Analysis of E-Nose Multidimensional Signals in Wastewater Treatment

Abstract

The work represents a successful attempt to combine a gas sensors array with instrumentation (hardware), and machine learning methods as the basis for creating numerical codes (software), together constituting an electronic nose, to correct the classification of the various stages of the wastewater treatment process. To evaluate the multidimensional measurement derived from the gas sensors array, dimensionality reduction was performed using the t-SNE method, which (unlike the commonly used PCA method) preserves the local structure of the data by minimizing the Kullback-Leibler divergence between the two distributions with respect to the location of points on the map. The k-median method was used to evaluate the discretization potential of the collected multidimensional data. It showed that observations from different stages of the wastewater treatment process have varying chemical fingerprints. In the final stage of data analysis, a supervised machine learning method, in the form of a random forest, was used to classify observations based on the measurements from the sensors array. The quality of the resulting model was assessed based on several measures commonly used in classification tasks. All the measures used confirmed that the classification model perfectly assigned classes to the observations from the test set, which also confirmed the absence of model overfitting.

Environmental Engineering Applications of Electronic Nose Systems Based on MOX Gas Sensors

“…A system for the classification of stages of wastewater treatment processes and full-scale WWTP (particular devices) was designed and presented by Łagód and coworkers [ 103 , 104 ]. The mentioned system was based on a gas sensors matrix consisting of 17 sensors by Figaro.…”

“…The mentioned system was based on a gas sensors matrix consisting of 17 sensors by Figaro. For dimensionality reduction and preliminary data visualization, principal component analysis was applied [ 103 ] as well as the t-SNE method [ 104 ]. Basing on a distance matrix in multidimensional space, the k-means method (non-hierarchical cluster analysis) was used to find homogeneous clusters of data [ 103 ].…”

“…Basing on a distance matrix in multidimensional space, the k-means method (non-hierarchical cluster analysis) was used to find homogeneous clusters of data [ 103 ]. For evaluating the discretization potential of the collected multidimensional data enabling to seek general relations and the relationships between groups of data in multidimensional space, the k-median method was used [ 104 ]. As a simple predictive model with sufficient accuracy of about 98%, a decision tree model was used [ 103 ], while a more advanced random forest model allowed achieving an accuracy of 100% [ 104 ] basing on an analysis of 185 multidimensional datasets and readouts from 17 gas sensors.…”

Rapid Method of Wastewater Classification by Electronic Nose for Performance Evaluation of Bioreactors with Activated Sludge

“…Thus, the term "gas fingerprint" is frequently employed when considering different signal combinations. Appropriate statistical analyses of multidimensional data are conducted for this purpose, such as artificial neural networks (ANN) [11,12], decision trees (DT) and random forests (RF) [13], support vector machines (SVM) [14], t-distributed stochastic neighbor embedding (t-SNE) [15], cluster analysis (CA) methods, or principal component analysis (PCA) [16].…”

“…Unfortunately, due to the complexity of the relationships between gas sensor array readings, using deterministic models for the classification of objects is not sufficient. However, high classification capability can be achieved using an appropriate advanced machine learning model [15].…”