Machine Learning-Optimized Dual-Band LoRa Elliptical Patch Antenna in LoRa Communication System for Waterborne Microplastic Detection

Abstract

This research proposes a dual-band LoRa elliptical patch antenna for the LoRa communication system to detect waterborne microplastics. The proposed LoRa communication system comprises a LoRa sensor node board and an IoT-LoRa gateway board. The LoRa sensor node board is used to capture microplastic images using a digital camera and collect analog signal data from an 8×8 photodiode array which detects the reflected light from microplastic fragments. The data is transmitted using a LoRa elliptical patch antenna in the sensor node board, operating at 0.915 GHz for long-range data transfer. The IoT-LoRa gateway board is used to forward data received from the LoRa sensor node board to a cloud server via the internet, and the stored data is accessible and viewable via a smartphone. In this research, the antenna design is optimized by using machine learning (ML) algorithms, unlike conventional antenna design methods which rely on the manual and iterative process. The ML-optimized dual-band LoRa elliptical patch antenna covers the LoRa, UHF RFID, and ZigBee frequency bands, with an omnidirectional radiation pattern. The measured impedance bandwidths are 8.93% (0.868–0.949 GHz) and 12.69% (2.36–2.68 GHz) for the lower and upper frequency bands, respectively, with the corresponding impedance matching (|<italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">S</i><sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">11</sub>|) of -23.02 dB at 0.907 GHz and -27.27 dB at 2.52 GHz. Two ML-optimized LoRa elliptical patch antennas are subsequently integrated into the LoRa communication system, i.e., one on the LoRa sensor node board and other on the IoT-LoRa gateway board. Furthermore, prior to indoor and outdoor experiments, the ML-driven waterborne microplastic detection scheme with the LoRa communication system is trained and tested using camera-captured images and analog signal-converted images from the photodiode array. The ML-driven microplastic detection scheme can classify different types of microplastics in water, achieving an accuracy of 100% for all types of microplastics. The detection scheme is also capable of identifying the presence of microplastics in water, achieving an overall accuracy of 98.5%. The originality of this work lies in the use of ML algorithm to optimize the antenna design and to streamline identification and detection of microplastics in water.