Impedance-Assisted Neural Network-Based Model for Real-Time Prediction of H₂S Using Pd Anchored CuCrO₂ Sensor

The presence of highly toxic hydrogen sulfide (H₂S) in the atmosphere can have adverse effects on human health. Therefore, it is crucial to monitor this gas for gas leak alarms and security purposes. Considerable efforts have been focused on creating and improving gas sensors to enhance their performance in detecting H₂S. Creating a simple method to manufacture H₂S sensors with both exceptional performance and prolonged stability presents a notable challenge. To address this challenge, the integration of the Internet of Things (IoT) and Machine Learning (ML) in sensor technology is crucial for advancing gas sensing capabilities. In this context, we introduced an ML-based H₂S gas sensor utilizing Pd-anchored CuCrO₂, designed for ultra-low concentrations and operating at 150 °C for accurate concentration prediction. Most MOS-based sensors typically operate in chemiresistive mode, offering a univariate output of DC resistance at a specific moment. However, due to inherent issues with MOS-based sensors, a univariate output is inadequate for accurately estimating concentration. Various strategies have been employed to enhance sensor performance and achieve precision. These strategies include creating sensor arrays, applying temperature modulation to boost sensor responses, and utilizing broad-range impedance spectroscopy. Notably, these techniques yield multivariate outputs from single or multiple sensors, allowing real-time comparison of multiple variables for accurate environmental assessment. Sensor array platforms and temperature modulation approaches are progressing towards imminent field implementation. Ongoing research in the field is dedicated to overcoming these challenges and further improving sensor capabilities. Our study focuses on developing a composite of CuCrO₂-based material to prevent sulfur poisoning during continuous sensor operation. Additionally, the material is adorned with Pd to enhance selectivity towards H₂S. Furthermore, the CuCrO₂-based MOS sensors are integrated with an impedance-based multivariate analysis technique. This involves considering multiple impedance-related variables, facilitating more sophisticated data processing. The use of a neural network-based multi-layer perceptron (MLP) allows the system to analyze a combination of impedance-based variables at multiple frequencies. This approach enables the system to better discern genuine changes in H₂S concentration from external factors or drift, contributing to improved accuracy and reliability. These intelligent systems, capable of real-time monitoring and adaptive responses, aim to offer more dependable and efficient gas detection solutions across various industries.<br/><br/>Keywords: chemiresistive gas sensors, multi-layer perceptron (MLP), concentration prediction, Impedance measurement.