The integration of IoT technologies and data-driven Machine Learning (ML) algorithms is revolutionizing the industrial sector, mainly through predictive maintenance and Industry 4.0. However, most of the solutions used today are based on fully-supervised ML approaches, necessitating large datasets and detailed knowledge of the system under monitoring. This work proposes an innovative, fully unsupervised, fault detection and predictive maintenance system for industrial electric motors based on an self-sustainable IoT sensor node and an online clustering algorithm. The system acquires real-time sparse data through an ultra-low power wireless sensor node, using a DBSCAN-based approach to identify operational conditions and anomalies without the need for pre-training or dataset collection. Experimental field tests on different three-phase motors, based on a 6-month continuous trial, have demonstrated the method’s effectiveness, highlighting the algorithm’s ability to dynamically adapt to new operating conditions correctly, discriminating among 7 classes. Our system reached a 97% accuracy in correctly discriminating anomalies and 74% in automatically identifying operative conditions on a previously unknown electric motor. The results confirm the potential of the proposed approach to improve industrial maintenance reliability and efficiency, reducing costs and unexpected downtime.
Leaf wetness duration (LWD) is key in predicting and preventing plant diseases. However, accurate detection of leaf wetness is often challenged by the presence of agricultural contaminants that can interfere with sensor readings. This study presents a highly selective reduced graphene oxide (rGO)-based capacitive leaf wetness sensor (LWS) fabricated on a flexible polyimide substrate, emphasizing its performance under synthetic conditions mimicking real agricultural environments. The sensor is fabricated by drop-casting 0.01 mg of rGO dispersed in 0.5 ml of deionized water. It demonstrates a remarkable 25,000% capacitive response to water exposure, with response and recovery times of 125 s and 37 s, respectively. Crucially, the sensor exhibits excellent selectivity to water, showing negligible response to common agricultural contaminants such as NaCl, CaCl2, CdCl2, urea, humic acid, NPK fertilizer, vermicompost, and decomposed compost, even at high concentrations. This performance is attributed to the hydrophilic surface characteristics of rGO and the capacitance-based sensing mechanism, which provide robust immunity to non-aqueous interference. The proposed LWS represents a significant advancement for precision agriculture by enabling reliable, real-time LWD monitoring in chemically complex environments.
Real-world deployments of Internet of Things (IoT) sensing systems equipped with artificial intelligence (AI) models generally experience a reduction in accuracy over time due to concept drift and data shift. To overcome this issue, it is suggested to periodically retrain the embedded AI model using new incoming data. However, this requires powerful processing hardware and labelled data, which are not generally available on IoT edge-based devices deployed in the real-world. In this study, we propose a method that benefits from a Deep Learning (DL) model for feature extraction and K-Means clustering for classification. The K-Means clustering technique can use the new incoming unlabeled data to update the model, thus helping the system to adapt to any changes that might occur in the new data. The proposed method is evaluated on two image datasets: the first is a public dataset with artificially added concept drift, and the second is a real-world dataset that suffers from concept drift issues. This is a lightweight model with only 610 KB of size and 608 KB of peak memory, which needs less than 0.6 s to perform and lower than 0.7 J to analyse each sample and update the model. Therefore, the method could easily be stored and executed on resource-constrained microcontroller-based (MCU-based) devices to deal with concept drift.
This paper considers the problem of precise geolocation of an emitter of electromagnetic waves on the Earth performed by a set of satellites in Low Earth Orbit (LEO). The sensing platforms are supposed to be composed by an array of antennas able to perform Angle of Arrival (AOA) measurements.The study deals with the optimization problem on the distribution of the satellites in order to reduce the Position Dilution of Precision (PDOP), considering the specific kinematics of the satellite orbits, the measurement error of the sensing platforms, and the geometrical constraint related to the Line of Sight (LOS).The main advancement of the paper is on the evaluation of the effect of the Signal-to-Noise (SNR) on the optimal distribution of the satellites compared to previous approximate investigations that considered a constant model for the standard deviation of the measurement error.Main metrics of evaluation are on the angles of the right ascension of the ascending node, true anomaly, and spacing angle between the satellites with respect to the emitter position.Among the results, even if restricted to the case of a system composed by two satellites, there is the confirmation of emerging properties of symmetry and anti-symmetry of such angles and the presentation of quantitative trends on the differences, that suggests that the effects of SNR on these geolocation problems can have an important impact on the optimal distribution of the satellites.
This paper describes an experimental characterization of a planar double-spiral resistive temperature microsensor. The proposed microsensor has been fabricated by means of a thermal evaporation process by employing a 1.2 mm-thick borofloat glass substrate and by depositing a 100 nmthick aluminum (Al) layer stacked between two 20 nm-thick chromium (Cr) layers. The fabricated microsensor has a track width of 50 mu m with an inter-track spacing of approximately 300 mu m. The experimental characterization has been performed by means of a tailored PID controlled temperature chamber including two Peltier cells, two fans and a Pt100 resistive temperature detector (RTD) for temperature control and monitoring. A sensitivity of 1.2 Omega/degrees C in the considered temperature range of 16 - 66 degrees C has been achieved showing a maximum nonlinearity error of 0.41 % referred to the span. The obtained experimental results confirm the possibility to exploit the proposed microsensor for temperature measurements.
