Abstracts:The primary concern of this article is to stabilize the rotating speed of the permanent magnet DC (PMDC) motor driven by a DC-DC sepic converter under mismatched disturbances via higher order PID sliding surface (PIDSS) controller. This controller offers numerous benefits, including robustness, enhanced control performance, flexibility, simple implementation, and low cost. An algorithm for the above-said control is developed for the load torques such as: no-load, constant, frictional, and propeller types. Further, the features of PIDSS are compared with classical sliding surface, sliding mode control (SMC) and proportional integral controller (PIC) by taking into consideration of peak overshoot, steady-state error and settling time. Simulation and experimental results are obtained satisfactorily.

Abstracts:Distribution systems are increasingly experiencing the penetration of photovoltaic (PV) systems. Although PV penetration is beneficial up to a point, beyond that point, it begins to generate issues related to voltage levels and grid stability. In modern distribution system planning, it is essential to identify an optimal operational point where the integration of PV supports the voltage profile rather than causing any adverse effects. The purpose of this paper is to explore and evaluate strategies to enhance Hosting Capacity and reduce Power Losses in distribution systems through an optimization algorithm that iteratively uses power-flow simulations and a Multi-Objective Genetic Algorithm. Different strategies taking advantage of conventional distribution system assets are formulated to avoid new system reinforcement. The strategies include Network Reconfiguration, Capacitor Switching, On-Load Tap Changer Switching, Volt-VAR Control Settings and the Combination of all strategies. To evaluate the efficiency of each approach, a comprehensive simulation study is conducted on the IEEE 123 bus distribution system modeled in OpenDSS, with an algorithm created in Python to control the optimization process.

Abstracts:The amorphous drug structure represents an important feature to be reached in the pharmaceutical field due to its possibility of increasing drug solubility, considering that at least 40% of commercially available crystalline drugs are poorly soluble in water. However, it is known that the amorphous local structure can vary depending on the amorphization technique used. Therefore, recognizing such variations related to a specific amorphization technique through the pair distribution function (PDF) method, for example, is an important tool for drug characterization concerns. This work presents a method to classify amorphous drugs according to their amorphization techniques and related to the local structure variations using machine learning. We used experimental PDF patterns obtained from low-energy X-rays scattering data to extract information and expanded the data through the Monte Carlo method to create a synthetic dataset. Then, we proposed the evaluation of such a technique using a Deep Neural Network. Based on the results obtained, it is suggested that the proposed technique is suitable for the amorphization technique and local structure recognition task.

Abstracts:Given the pressing demand for energy conservation, the petrochemical sector faces increasingly stringent energy-saving mandates. Heat exchangers, essential to this sector, suffer efficiency losses and increased energy consumption due to fouling. To ensure optimal operation of heat exchange systems, regular assessment of solid deposits and the implementation of cleaning schedules are imperative. However, the multitude of influencing factors renders traditional estimation methods unreliable. Consequently, we developed a stacking model to predict the fouling factor of heat exchangers. Specifically, we first constructed fouling factor prediction models using various machine learning techniques, then selected the best-performing models random forest, extreme gradient boosting , and light gradient boosting machine for integration. Finally, the predictions from these three models were fed into a linear regression layer to form the final stacking model. The results indicate that the constructed stacking model significantly enhances the accuracy of fouling factor prediction. This model not only surpasses traditional multilayer perceptron neural network methods but also outperforms the well-performing gaussian process regression. This achievement not only validates the effectiveness of our model but also provides robust support for future research and applications in related fields.

Abstracts:Long non-coding RNAs (lncRNAs) represent a fundamental category of epigenetic modulators. Recent research has revealed that lncRNAs play critical roles in gene regulatory mechanisms, substantially influencing the pathogenesis of various human diseases. In this study, a multilayer heterogeneous network was created and we introduced the fast random walk with restart (FRWR) for predicting connections between lncRNAs and diseases. By combining the similarity network of lncRNA, similarity network of disease, and association network of existing lncRNA-disease, a multilayer heterogeneous network was constructed, and the fast random walk with restart method (FRWR) was applied on this network to predict additional potential lncRNA-disease associations. The AUROC value of 0.9034, achieved through leave-one-out cross-validation, underscored the predictive precision of the FRWR technique. Furthermore, a case study of three different diseases provided further validation of the reliability of prediction results. Overall, the multilayer network FRWR method proposed in this work could effectively forecasting the connections between lncRNAs and diseases, offering valuable insights into comprehending the functions of lncRNAs in the context of human health and disease. The source code for the FRWR method can be accessed at: https://github.com/TianTianTian14/FRWR.

Abstracts:It is common to use sequential testing strategies to help reduce the time of automated detection of an auditory steady-state response (ASSR). However, the application of repeated tests leads to an increase of false positive rate. Monte Carlo-based strategies are used to overcome this obstacle. Despite several paper could be found describing such strategies, no comprehensive comparison was found in the literature. The chosen strategies are based on Monte Carlo simulations to calculate critical values and were faithfully replicated for comparison purposes, and then the test application parameters were varied to suggest an optimization. The detection rate and/or the detection speed improved with each implemented strategy, except for the one related to the year 2013, which increased the false positive rate to 15.3%. The other strategies kept the false positive rate under control. The Pareto curves compared the optimizations of the strategies and revealed that the modified 2015 strategy had the performance achieving 5.6% higher than the original parameters. The automated detection of ASSR improved with each implemented strategy, but not all of them kept a controlled false positive rate (2013 and 2015). The 2015 modified strategy had the highest detection rate in the shortest time.

Abstracts:Vehicle counting systems detect, classify, and count vehicles with sensors or image processing, providing valuable information for road management. Image processing systems provide detailed information on vehicle flow with adequate lighting conditions and a higher computational cost compared to sensor systems. The image processing systems with higher accuracy require higher computational cost. This feature limits the number of application cases in cities with low technology level. This research analyzes urban vehicle counting using an automatic image processing system using YOLOv5 in the vehicle detection-classification stage and the SORT algorithm in the tracking stage. The study used videos recorded from a pedestrian bridge in Popayan, Colombia, for an exploratory study of the influence of preprocessing operations on the performance of a low-tech vehicle counting system. The study performed a comparative statistical analysis to determine the impact of different settings on system performance. An ANOVA analysis evaluates the incidence of frame cut and reshape on YOLO processing. The results indicate that a 30% cut of the image area prior to YOLO processing produces the lowest weighted average error. In addition, the frame reshape only increases the processing time. The study proposes improvements in the performance of an offline automatic vehicle counting system from the video preprocessing stage.

Abstracts:Obesity is a complex chronic disease resulting from the interaction of multiple behavioral factors. This paper presentsthe application of Machine Learning to identify the primary groups of behaviors contributing to the development of obesity.Supervised machine learning emphasizes decision trees and deep artificial neural networks from datasets. The study also references related work that utilizes predictive methods to estimate obesity levels based on physical activity and dietary habits. Furthermore,it compares the performance of classification algorithms such as J48, Naive Bayes, Multiclass Classification, Multilayer Perceptron, KNN, and decision trees when predicting diabetes cases. The objective is to analyze different tools in the assessment based on physical activity and dietary habits, contributing to the improvement of obesity risk diagnosis. In addition, MLP and J48 demonstrated strong performance among all the algorithms, but BPTT achieved the highest overall performance.