A novel method for total chlorine detection using machine learning with electrode arrays

Multiple Statistical Model Ensemble Predictions of Residual Chlorine in Drinking Water: Applications of Various Deep Learning and Machine Learning Algorithms

Applicability of statistical models in predicting chlorine decay remains minimally explored. This study predicted residual chlorine using six deep learning and nine machine learning techniques. Suitability of multimodel ensembles (MMEs) including arithmetic mean of all the models (Ens1), average of the best three performing models (Ens2), and weighted mean of outputs from all the 15 models was investigated. A total of nine “goodness-of-fit” measures (such as distance correlation (r_d) and Taylor skill score) were used to rank the models. The two best deep learning methods were the nonlinear autoregressive model with exogenous input (NARX) (r_d = 0.51) and feedforward backpropagation (FFB) (r_d = 0.61). The two best machine learning algorithms included random forests (RF) (r_d = 0.64) and Gaussian process regression (GPR) (r_d = 0.59). Eventually, Ens2 was obtained using RF, FFB, and GPR. Ens2 performed better than Ens1 and Ens3. The amount of variance explained by individual models and MMEs was over the ranges of 13–66% and 51–74%, respectively. Ens2 explained 74% of the total variance in observed residual chlorine. Remarkably, the appropriateness of the MMEs depends on the approach for combining model outputs, and the number of models considered. This study demonstrated the acceptability of statistical MMEs in predicting chlorine residual concentration in drinking water.

A new insight for real-time wastewater quality prediction using hybridized kernel-based extreme learning machines with advanced optimization algorithms

Accurate prediction of inlet chemical oxygen demand (COD) is vital for better planning and management of wastewater treatment plants. The COD values at the inlet follow a complex nonstationary pattern, making its prediction challenging. This study compared the performance of several novel machine learning models developed through hybridizing kernel-based extreme learning machines (KELMs) with intelligent optimization algorithms for the reliable prediction of real-time COD values. The combined time-series learning method and consumer behaviours, estimated from water-use data (hour/day), were used as the supplementary inputs of the hybrid KELM models. Comparison of model performances for different input combinations revealed the best performance using up to 2-day lag values of COD with the other wastewater properties. The results also showed the best performance of the KELM-salp swarm algorithm (SSA) model among all the hybrid models with a minimum root mean square error of 0.058 and mean absolute error of 0.044.

Continuous monitoring of propofol in human serum with fouling compensation by support vector classifier

We present a new method for electrochemical sensing, which compensates the fouling effect of propofol through machine learning (ML) model. Direct and continuous monitoring of propofol is crucial in the development of automatic systems for control of drug infusion in anaesthesiology. The fouling effect on electrodes discourages the possibility of continuous online monitoring of propofol since polymerization of the surface produces sensor drift. Several approaches have been proposed to limit the phenomenon at the biochemical interface; instead, here, we present a novel ML-based calibration procedure. In this paper, we analyse a dataset of 600 samples acquired through staircase cyclic voltammetry (SCV), resembling the scenario of continuous monitoring of propofol, both in PBS and in undiluted human serum, to demonstrate that ML-based model solves electrode fouling of anaesthetics. The proposed calibration approach is based on Gaussian radial basis function support vector classifier (RBF-SVC) that achieves classification accuracy of 98.9% in PBS, and 100% in undiluted human serum. The results prove the ability of the ML-based model to correctly classify propofol concentration in the therapeutic range between 1μM and 60μM with levels of 10μM, continuously up to ten minutes, with one sample every 30s.