Instrumental Odour Monitoring System Classification Performance Optimization by Analysis of Different Pattern-Recognition and Feature Extraction Techniques

Abstract

Instrumental odour monitoring systems (IOMS) are intelligent electronic sensing tools for which the primary application is the generation of odour metrics that are indicators of odour as perceived by human observers. The quality of the odour sensor signal, the mathematical treatment of the acquired data, and the validation of the correlation of the odour metric are key topics to control in order to ensure a robust and reliable measurement. The research presents and discusses the use of different pattern recognition and feature extraction techniques in the elaboration and effectiveness of the odour classification monitoring model (OCMM). The effect of the rise, intermediate, and peak period from the original response curve, in collaboration with Linear Discriminant Analysis (LDA) and Artificial Neural Networks (ANN) as a pattern recognition algorithm, were investigated. Laboratory analyses were performed with real odour samples collected in a complex industrial plant, using an advanced smart IOMS. The results demonstrate the influence of the choice of method on the quality of the OCMM produced. The peak period in combination with the Artificial Neural Network (ANN) highlighted the best combination on the basis of high classification rates. The paper provides information to develop a solution to optimize the performance of IOMS.

Application of Machine Learning for Fenceline Monitoring of Odor Classes and Concentrations at a Wastewater Treatment Plant

“…Consequently, the measurement cycle consisted of 1 min of ZA, 2 min sample draw-in by IOMS (acquisition phase), and 3 min for baseline recovery; for each sample, the measurement cycle was repeated four times. The duration of each phase was determined on the basis of the literature [ 13 ] and observing the response of the sensors in terms of the time required to reach a steady state value during the feeding of the samples, which was less than 90 s (see Section 2.3 ). The sampling frequency of the instrument was set to 0.1 Hz, so that 12 data points were available for each sample replicate.…”

“…Consequently, for the sensor i , the fractional difference x i in the signal relative to the peak values was used [ 12 , 13 ], by considering the associated piecemeal signal within the 2-min acquisition phase: where R i is the resistance peak value after feeding and is the baseline resistance, as detected during the zero-air drawn-in cycle, for the sensor i . At present, we do not have sufficient information to confirm whether the response of the sensors that gave a positive response (nanocomposite-based sensors) varies with the variation in chemical species and, therefore, whether this information can be used in ML algorithms.…”

“…Then, the response curves, representing the dynamic variation in the electrical resistance of the sensors [39], were analyzed to extract piecemeal signal features. Although several features are extractable from the response curves (steady-state response, rising time, falling time) [12], recent studies suggest that the peak values of the signals, representing the maximum degree of change in the sensors responding to odor, are more effective for odor classification [13]. Consequently, for the sensor i, the fractional difference x i in the signal relative to the peak values was used [12,13], by considering the associated piecemeal signal within the 2-min acquisition phase:…”

“…Furthermore, the rapid development of low-cost sensors, which allows for air pollution monitoring at a lower cost and with a higher spatial density than reference measurement methods, the work undertaken by the Technical Committee CEN/TC 264 “Air quality” and the relative working group (WG41) for defining technical specifications for the performance evaluation of air quality sensors, and the development of the high-speed communications network (5G) are the most significant recent developments for the field of IOMS hardware. The software developments in recent years are remarkable both in the field of data pretreatment and feature extraction [ 12 , 13 ], and algorithms for odor classification [ 14 , 15 ] and quantification [ 16 ]. Several machine learning (ML) algorithms were proposed for pattern recognitions of odor, such as PCA, LDA, ANN, PLS, and SVM, and these were employed as regression techniques in order to obtain odor concentrations [ 17 , 18 , 19 ].…”

Classification Method of Uniform Circular Array Radar Ground Clutter Data Based on Chaotic Genetic Algorithm

“…Particle Swarm Optimization (PSO) and Genetic Algorithm (GA) belong to Swarm Intelligence Optimization (SIO), so they can be used for reference in optimization and algorithm application. Applying genetic algorithm to the pattern recognition and classification of actual overvoltage data can realize effective data classification and recognition [ 32 , 33 , 34 ]. In addition, semi-supervised clustering based on the genetic algorithm can be used to effectively classify hyperspectral images [ 35 ].…”