Machine Learning for the Detection and Diagnosis of Anomalies in Applications Driven by Electric Motors.

Manufacturing systems are becoming increasingly flexible, necessitating the adoption of new technologies that allow adaptations to a turbulent and complex modern market. Consequently, modern concepts of production systems require horizontal and vertical integration, extending across value networks and within a factory or production shop. The integration of these environments enables the acquisition of a substantial amount of data containing information pertaining to production, processes, and equipment located on the shop floor. When these data and information are processed and analyzed, they have the potential to reveal valuable insights and knowledge about the manufacturing systems, offering interpretive outcomes for strategic decision making. One of the opportunities presented in this context includes the implementation of predictive maintenance (PdM). However, industrial adoption of PdM is still relatively low. In this paper, the aim is to propose a methodology for selecting the main attributes (variables) to be considered in the instrumentation setup of rotating machines driven by electric motors to decrease the associated costs and the time spent defining them. For this, the most well-known data science and machine learning algorithms are investigated to choose the one most adequate for this task. For the experiments, different testing scenarios were proposed to detect the different possible types of anomalies, such as uncoupled, overloaded, unbalanced, misaligned, and normal. The results obtained show how these algorithms can be effective in classifying the different types of anomalies and that the two models that presented the best accuracy values were k-nearest neighbor and multi-layer perceptron.

Cyber-Enabled Optimization of HVAC System Control in Open Space of Office Building.

Thermal comfort is crucial to well-being and work productivity. Human thermal comfort is mainly controlled by HVAC (heating, ventilation, air conditioning) systems in buildings. However, the control metrics and measurements of thermal comfort in HVAC systems are often oversimplified using limited parameters and fail to accurately control thermal comfort in indoor climates. Traditional comfort models also lack the ability to adapt to individual demands and sensations. This research developed a data-driven thermal comfort model to improve the overall thermal comfort of occupants in office buildings. An architecture based on cyber-physical system (CPS) is used to achieve these goals. A building simulation model is built to simulate multiple occupants' behaviors in an open-space office building. Results suggest that a hybrid model can accurately predict occupants' thermal comfort level with reasonable computing time. In addition, this model can improve occupants' thermal comfort by 43.41% to 69.93%, while energy consumption remains the same or is slightly reduced (1.01% to 3.63%). This strategy can potentially be implemented in real-world building automation systems with appropriate sensor placement in modern buildings.

Cyber-Physical System for Smart Traffic Light Control.

In recent years, researchers have proposed smart traffic light control systems to improve traffic flow at intersections, but there is less focus on reducing vehicle and pedestrian delays simultaneously. This research proposes a cyber-physical system for smart traffic light control utilizing traffic detection cameras, machine learning algorithms, and a ladder logic program. The proposed method employs a dynamic traffic interval technique that categorizes traffic into low, medium, high, and very high volumes. It adjusts traffic light intervals based on real-time traffic data, including pedestrian and vehicle information. Machine learning algorithms, including convolutional neural network (CNN), artificial neural network (ANN), and support vector machine (SVM), are demonstrated to predict traffic conditions and traffic light timings. To validate the proposed method, the Simulation of Urban Mobility (SUMO) platform was used to simulate the real-world intersection working. The simulation result indicates the dynamic traffic interval technique is more efficient and showcases a 12% to 27% reduction in the waiting time of vehicles and a 9% to 23% reduction in the waiting time of pedestrians at an intersection when compared to the fixed time and semi-dynamic traffic light control methods.

FEA and Machine Learning Techniques for Hidden Structure Analysis.

This study focuses on investigating and predicting two hidden structures: plant root system architecture and non-visible bubbles in plexiglass. Current approaches are damaging, expensive, or time-consuming. Infrared imaging was used to study the root structure and depth of small plants and to detect the diameter and depth of bubbles in plexiglass. A finite element analysis (FEA) model was built to simulate the infrared imaging process to realize the detection and prediction given the amount of heat flux required to obtain thermal images and data. For the root system, based on a tree structure thermal profile over time derived from the FEA model, a line scan method was developed to predict root structure. The main root branches can be viewed from the detection results. Polynomial regression, support vector machine (SVM), and artificial neural network (ANN) models were designed to predict root depth. For bubble defects, three ANN models were developed to predict bubble size using temperature data generated by the FEA model. Results indicated that these models provide valid predictions. Statistical tests were applied to evaluate and compare the above predictive models. Results suggest that infrared imaging and machine learning models can be used to provide information on both hidden structures.

The fabrication and thermal properties of bismuth-aluminum oxide nanothermometers.

Bismuth (Bi) nanowires, well controlled in length and diameter, were prepared by using an anodic aluminum oxide (AAO) template-assisted molding injection process with a high cooling rate. A high performance atomic layer deposition (ALD)-capped bismuth-aluminum oxide (Bi-Al2O3) nanothermometer is demonstrated that was fabricated via a facile, low-cost and low-temperature method, including AAO templated-assisted molding injection and low-temperature ALD-capped processes. The thermal behaviors of Bi nanowires and Bi-Al2O3 nanocables were studied by in situ heating transmission electron microscopy. Linear thermal expansion of liquid Bi within native bismuth oxide nanotubes and ALD-capped Bi-Al2O3 nanocables were evaluated from 275 °C to 700 °C and 300 °C to 1000 °C, respectively. The results showed that the ALD-capped Bi-Al2O3 nanocable possesses the highest working temperature, 1000 °C, and the broadest operation window, 300 °C-1000 °C, of a thermal-expanding type nanothermometer. Our innovative approach provides another way of fabricating core-shell nanocables and to further achieve sensing local temperature under an extreme high vacuum environment.

Thermography and machine learning techniques for tomato freshness prediction.

The United States and China are the world's leading tomato producers. Tomatoes account for over $2 billion annually in farm sales in the U.S. Tomatoes also rank as the world's 8th most valuable agricultural product, valued at $58 billion dollars annually, and quality is highly prized. Nondestructive technologies, such as optical inspection and near-infrared spectrum analysis, have been developed to estimate tomato freshness (also known as grades in USDA parlance). However, determining the freshness of tomatoes is still an open problem. This research (1) illustrates the principle of theory on why thermography might be able to reveal the internal state of the tomatoes and (2) investigates the application of machine learning techniques-artificial neural networks (ANNs) and support vector machines (SVMs)-in combination with transient step heating, and thermography for freshness prediction, which refers to how soon the tomatoes will decay. Infrared images were captured at a sampling frequency of 1 Hz during 40 s of heating followed by 160 s of cooling. The temperatures of the acquired images were plotted. Regions with higher temperature differences between fresh and less fresh (rotten within three days) tomatoes of approximately uniform size and shape were used as the input nodes for ANN and SVM models. The ANN model built using heating and cooling data was relatively optimal. The overall regression coefficient was 0.99. These results suggest that a combination of infrared thermal imaging and ANN modeling methods can be used to predict tomato freshness with higher accuracy than SVM models.

Exploiting maximal dependence decomposition to identify conserved motifs from a group of aligned signal sequences.

UNLABELLED: Bioinformatics research often requires conservative analyses of a group of sequences associated with a specific biological function (e.g. transcription factor binding sites, micro RNA target sites or protein post-translational modification sites). Due to the difficulty in exploring conserved motifs on a large-scale sequence data involved with various signals, a new method, MDDLogo, is developed. MDDLogo applies maximal dependence decomposition (MDD) to cluster a group of aligned signal sequences into subgroups containing statistically significant motifs. In order to extract motifs that contain a conserved biochemical property of amino acids in protein sequences, the set of 20 amino acids is further categorized according to their physicochemical properties, e.g. hydrophobicity, charge or molecular size. MDDLogo has been demonstrated to accurately identify the kinase-specific substrate motifs in 1221 human phosphorylation sites associated with seven well-known kinase families from Phospho.ELM. Moreover, in a set of plant phosphorylation data-lacking kinase information, MDDLogo has been applied to help in the investigation of substrate motifs of potential kinases and in the improvement of the identification of plant phosphorylation sites with various substrate specificities. In this study, MDDLogo is comparable with another well-known motif discover tool, Motif-X. CONTACT: francis@saturn.yzu.edu.tw