previous day
all days

View: session overviewtalk overview

09:00-10:00 Session 9: Keynote Lecture
In search of emotional traces: detection of depression through social media

ABSTRACT. Every day the Internet witnesses a great amount of people activity in many different social media networks. They allow us to share –personal– information very easily, and with it, they offer us innumerable new opportunities for business intelligence, personalization, and health monitoring. This talk aims to resume some of the work we have developed in the previous two years towards the detection of social media users suffering from depression. Particularly, it presents three machine learning methods based on the analysis of emotions, which are supported on idea that depressed and non-depressed users must show different emotion distributions and trends in their posts. The talk will conclude with a brief overview of our ongoing work in this task, considering the application of deep learning techniques and the development of cross-language solutions.

10:00-11:40 Session 10A: Power Converters
Quadratic Buck/Boost Converter in Series Connection for Photovoltaic Applications

ABSTRACT. This paper proposes a quadratic buck-boost converter, which is based on a non-cascade connection of two basic switching cells. The proposed converter is analyzed into a photovoltaic architecture in series connection. The voltage conversion ratio, steady state operating condition, converter operating modes, and semiconductor stress are derived. The characteristics of the proposed converter are verified through simulation results, where the layout of the photovoltaic system uses two quadratic converters, each with an MPPT P&O strategy.

Battery Types and Electrical Models: A Review

ABSTRACT. Batteries performance is an important issue for those systems with an implicated energy storage system where it is important to known three fundamental internal parameters, state of charge (SoC), state of health (SoH) and state of operation (SoF). In order to know these internal states some techniques as adaptive observers are use together with a battery model. In this paper, the main characteristics of the most common and commercial batteries, as well as the most cited batteries models in the literature are studied. Then a comparative analysis making emphasis in its qualities and applications is performed. The main idea of the paper is to provide a wide and fast view of the main characteristics together with the advantages and drawbacks of the batteries and battery models. In future works, this information can be a good reference to design algorithms for parameter identification or internal states estimation.

DC-DC Buck Converter with an LC Filter for Battery Parameter Identification

ABSTRACT. In this paper, the identification of the parameters of a battery model is addressed by measuring the current and voltage at the terminals of the battery and by using a moving window least square method in an online application. The battery is considered to supply a DC voltage reference to a dedicated load through a buck converter with an LC filter at the input. The converter output voltage is regulated through a conventional controller that ensures stability, sufficient damping and zero steady-state error at constant voltage references. The results of numerical simulations show that the presented scheme is capable of identifying battery parameters such as internal resistance, battery capacity and the time constant of an RC network when referring to an equivalent electric circuit model that considers the battery relaxation effect.

Evaluation of a Thermoelectric Generation system based on Differential-Power Processing architecture under non-uniform temperature conditions
PRESENTER: Victor M. Bravo

ABSTRACT. This paper presents a Differential Power Processing (DPP) architecture applied to series-connected thermoelectric generators (TEG). Currently, thermoelectric technology is being considered as a promising power generation technology that can be used to recover waste heat energy. Thus, a thermoelectric generation system is studied under non-uniform temperature conditions in multiple TEG devices. The main objective is to allow each thermoelectric sub-module to reach its maximum power point more quickly. The purpose has been to improve the maximum power point tracking (MPPT) in each sub-module, thus it is possible to increase the efficiency with respect to the traditional method based on a global MPPT. Differential Power converters have been implemented in each TEG sub-module to provide an effective solution and mitigate the impact of the mismatch in the power obtained. The DPP architecture consists of a small micro-converter, at the submodular level, applied to the thermoelectric cell. The control algorithm is oriented to polarize each TEG device at its optimal point, which allows us an active balancing among the different TEG sub-modules regardless of the operating temperature. Matlab-Simulink has been the software used to develop the TEG module-array.

Off-grid Wind Energy Conversion Systems based on Multilevel Cascaded H‐bridge Topology in Real Time Simulation

ABSTRACT. In this paper, the modeling and real time simulation of an off-grid wind energy conversion systems (WECS) based on isolated multilevel cascaded H‐bridge topology is presented. The system is tested under a synthetic wind speed signal generated by Gaussian Process (GP), even with wind variations. The off-grid WECS structure is made through an AC/DC power converter formed by a VSC in rectifier mode, which is designed through the Permanent Magnet Synchronous Generator (PMSG) characteristics; and a DC/AC power converter based on isolated multilevel cascaded H‐bridge of 5-levels. The system validation is achieved in MATLAB-Simulink and with the laboratory tests using the concept of Model In the Loop (MIL) and the real-time simulator Opal-RT Technologies. The obtained results show a robust and effective methodology generating a total output power of 30 kW off-grid.

10:00-11:40 Session 10B: Biomedical Applications
Blind Deconvolution Estimation by an Exponentials Library

ABSTRACT. The deconvolution process allows to extract the impulse response of a sample by collecting the input/output response. In the blind deconvolution estimation (BDE), this process is implemented without the input signal information. In particular, this work is focused on fluorescence lifetime imaging microscopy (FLIM) datasets, where the fluorescence impulse responses are extracted by assuming an exponential library model and a common instrument response (input signal) to all the measurements. Due to the nonlinear interaction of the free variables, an alternated least-squares methodology is adopted, which is based on constrained quadratic optimizations. The new BDE algorithm is validated with synthetic FLIM datasets by comparing the standard deconvolution methodology with an exponential library under different model orders, and types and levels of noise, which shows the applicability and robustness of the proposal.

Preprocessing and Labeling Tool for Lateral Skull X-Ray Images Applied to Cephalometric Analysis

ABSTRACT. Cephalometric analysis is a tool to study the craniofacial relationships, commonly used by dentists and orthodontists for skull growth analysis, diagnosis, planning, and treatment. Automatic cephalometric analysis systems rely on the precise labeling of lateral skull radiographs. This task demands from the health expert, an excellent visual acuity, and a significant amount of time in the medical image's manual labeling. This paper proposes a computational tool to support the process of semi-automatic labeling of Lateral Cephalometric Radiographs (LCR). This tool receives as input a set of images corresponding to LCR and performs automatic tasks such as filtering and digital signal processing to assist the health expert on the labeling task, using computational visual human perception models. The tool output is a set of N cephalometric landmarks (x, y) for each LCR. This tool has a set of image preprocessing techniques that clarify the manual labeling getting the accuracy required by the experts. It supports the experts in the cephalometric analysis. The second application is the creation of a new database for the development of automatic cephalometric landmark identification systems.

Analysis and design of a photoplethysmograph for glucose measurement using near-infrared spectroscopy analyzing human body physical and functional variables

ABSTRACT. Diabetes mellitus is a chronic disease that increases blood glucose levels. Glucose measurement devices, called glucometers, normally use invasive techniques. These devices are used by the patients themselves. They are small size and require a lancing device to extract a drop of blood and pinpoint the glucose concentration present, This technique is uncomfortable and painful for the patient when continuous blood glucose monitoring is required throughout the day. To improve glucose measurement conditions, non-invasive blood glucose meters have been proposed using near-infrared spectroscopy. The problem that occurs when using the near-infrared spectroscopy method is the physical and functional characteristics of the human body, which interact with the light emitted making it difficult to measure glucose levels, These characteristics are, for example, the fat, pressure, humidity, temperature, among others.

Brain Tumor Segmentation using an Encoder-Decoder Network with a Multiscale Feature Module

ABSTRACT. The correct identification of brain tumors in multi-sequence images is a very challenging task and of high relevance in the correct treatment of patients affected by this illness. In this paper, we present a segmentation neural network approach named FPM-BrainSeg. FPM-BrainSeg incorporates important features on its architecture such as residual connections, anisotropic convolutions, instance normalization, and a Feature Pooling Module (FPM) that increases the field of view of the extracted features. We compare the performance of FPM-BrainSeg with three state-of-the-art segmentation networks trained over a common Graphic Processing Unit (GPU) of the GTX family, by using the BraTS 2018 dataset. The performance obtained with FPM-BrainSeg surpassed the other neural models. The features extracted by the FPM increase the ability of the network to segment the different tumor regions as demonstrated in the visualization of the activation layers. Furthermore, a second comparison of our approach with state-of-the-art models (as presented in their original papers), demonstrate that our architecture attained competitive performance.

Mechatronic Design and Robust Control of an Artificial Ventilator in Response to the COVID-19 Pandemic

ABSTRACT. The crisis resulting from the COVID-19 pandemic has generated an adverse situation in which hundreds of people die due to the lack of artificial ventilation devices. In this sense, this work presents a proposal for the robust mechatronic design and control of a low-cost non-invasive ventilator, for which rapid prototyping manufacture strategies such as 3D printing and product design are used. In order to guarantee the reliability of the system operation, in this work, a robust control scheme based on super-twisting sliding modes is proposed, which guarantees the trajectory tracking control corresponding to the breathing profiles required by the patients. Experimental and simulation results validate the effectiveness of the proposed prototype design.

10:00-11:40 Session 10C: Data Science
A new approach for Pneumonia diagnosis using Convolutional Neural Networks

ABSTRACT. Convolutional Neural Networks (CNN) can be used as an efficient tool for detecting diseases between different types of medical imaging in a fast and reliable way, so that the article focuses on pneumonia disease, as yearly about 2.56 million people die in consequence of this illness. This paper illustrate the importance of data preprocessing as an effective approach for producing better results since raw data repercute in the training time, in consequence, it require more computational time to complete a determined machine learning problem. One of the main focal point is to introduce a novel and effective method to work with large amount of data and how it can be preprocessed for getting almost ideal results with a minimal lost of information due preprocessing. As the main result, the solution mentioned can help radiology and medical personnel to diagnose X-Ray images. Regarding the dataset, its name is Chest X-Ray Image Dataset, it’s a public dataset of Kaggle and contains 5856 JPEG images organized in three directories. As future work, this model can be used to work with other types of Medical Images due to its adatpability.

Analysis model of the most important factors in Covid-19 through data mining, descriptive statistics and random forest

ABSTRACT. The Covid19 pandemic has had a great impact worldwide, it has become a major problem due to the demand for care in hospitals and clinics despite the low level of mortality. This is because the disease has spread rapidly as the spread between people is accelerated. So in this document we propose using a classification-oriented machine learning method, we do a classic data science process so that we can perform noise cleaning and data processing to do descriptive statistical analysis in such a way that the most important variables or factors are identified through unsupervised learning. And with this it is appreciated that the most important variables for the risk of infection and mortality that Covid-19 disease can have are diseases that affect the immune system, such as diabetes, heart disease, hypertension and also kidney disease. They can cause serious kidney problems. And the evaluation of our method will be carried out through quality measures. Finally, this work opens the door to other investigations with the aim of conducting centralized investigations on each variable related to Covid19, in order to find relevant information that can promote an improvement in the current situation.

Roundness Estimation of Sedimentary Rocks Using Eliptic Fourier and Deep Neural Networks

ABSTRACT. —Sedimentary rocks analysis is useful in geological science, economic sector, and risk evaluation. Roundness is a morphological parameter that provide information to characterize and classify sedimentary material. Roundness degrees is estimated from the contour of the particle. Waddell (1932) proposed a remarkable method based on the measurement of particle’s curvature. This method is accurate; nevertheless, it is not invariant to scale and rotation. This problem can be solved by mapping the contour to the frequencydomain, however, spectral analysis is a difficult task. Based on these two approaches, we propose to use a deep neural network whose input is the elliptical Fourier spectrum and target is roundness proposed by Wadell. The training database consists of 623 realrocks images from some geological phenomena. We have found the neural networks perform very well on the 94% of rocks.

Subgroup classification model identifying the most influential factors in the mortality of a patients with COVID-19 using data analysis

ABSTRACT. This research assesses the health conditions of the people in the study and determines the reason why a person dies after being infected with COVID-19. In this study, 538 sample groups that provided medical data from people in different locations were analyzed. The biggest challenge in this study was to carry out 2 different criteria within the same data set to conclude that the mortality of the persons inside a group depends more than anything on the age of the person at risk and the presence of one or more other health disorders of the primary disease, which in this case is COVID-19.

For this study, the public data set "COVID analytics" was used, which provided all the necessary medical information and the classification of the groups, which are then interpreted as useful labels to better deduce the degree of mortality of the affected person. After completing the data analysis, it is determined that the factors that aggravate the condition of a patient with COVID-19 are: hypertension, advanced age and any other disease.

Multi-step forecasting of waiting time on emergency department overcrowding using multilayer perceptron neural network algorithm

ABSTRACT. In the next work, a multilayer perceptron artificial neural network (MLP-ANN) was implemented to perform a seven days multi-step prediction of waiting times in the emergency department of a public hospital. A dataset of more than two years was used to training the MLP-ANN. The imputation technique was used to interpolate the data. The dataset was distributed in training and testing with 80 and 20%, respectively. The results of the MLP-ANN were compared with the Persistence and ARIMA models, obtaining much better results than the other two methods, especially on weekends.

12:00-13:20 Session 11A: Power
A New Iterative Power Flow Method for AC Distribution Grids with Radial and Mesh Topologies

ABSTRACT. This brief discusses the classical problem of power flow analysis in alternating current (ac) distribution networks through Taylor series expansion. The main advantage of this approach is that it can work with radial and mesh topologies without modifications in its formulation. This method can deal with the hyperbolic relation between voltages and currents at $k$ node, i.e., $\mathbb{I}_{k} = \frac{\mathbb{S}^{\star}_k}{\mathbb{V}_k^{\star}}$, by transforming this into a linear approximation. To minimize the error in this linear transformation, an iterative procedure is implemented by updating the linearizing point, which allows reaching the same solution of the classical power flow methods for distribution systems in less processing time. Numerical results confirm the effectiveness of the proposed approach when compared to classical Gauss-Seidel, Newton-Raphson, and Backward/forward methods that can work with radial and mesh distribution network structures. All the numerical validations are conducted in MATLAB software.

Methodology for Measurement-based Load Modeling considering Integration of Dynamic Load Models

ABSTRACT. A methodology for measurement-based load modeling based on the integration of dynamic load models is presented in this paper. A composite load model (CLM) and an exponential recovery load (ERL) model are combined and used as a model structure for load representation coupled with a heuristic criterion. The parameter estimation problem is solved using an improved particle swarm optimization (PSO) algorithm, considering a speed update modification to enhance the global search capabilities of the technique. The validation of the load model is performed in two test systems, using measurements of load response after small and severe disturbances, considering noise measurement to simulate monitoring devices. The results show that the proposed load model has a better performance than the individual load models, and also PSO algorithm is adequate for parameter identification for measurement-based load modeling.

Remedial Action Scheme based on Automatic Load Shedding for Power Oscillation Damping

ABSTRACT. This paper presents an algorithm for automatic load shedding as a Remedial Action Scheme (RAS) for power oscillation damping. The algorithm computes the exact amount of power for the load shedding in all buses and then detects the critical areas in the system. Therefore, the trip actions are only in the areas where are necessary. In addition to this, the RAS is equipped with an oscillation monitoring stage since the system presents oscillatory transitions. The proposed RAS methodology is based on the online computation of the PV curves and the voltage stability margin for calculating the amount of power to trip and used a load shedding strategy based on modal participation factors to identify the most critical areas. On the other hand, the algorithm identifies ringdown data for the detection of oscillations, and Prony’s analysis monitors the damping of oscillations to determine online if the load disconnection has the desired effect. Performance evaluation of the proposed scheme has been made on a 12-bus network and, the results show excellent damping of oscillations for critical contingencies and an increase in the stability margin of the whole power system.


ABSTRACT. En este documento se describen algunos aspectos técnico necesarios para extraer los parámetros de dispersión y coeficientes de reflexión utilizados para calibrar instrumentos de medición en radiofrecuencia y microondas tales como el analizador vectorial de redes. Se describe brevemente una metodología para corregir los errores que se presentan al momento de realizar las mediciones con este instrumento. Finalmente, se presenta un ejemplo de medición de un transformador monofásico con relación de trasformación 124/24 volts de 150 VA de potencia.

12:00-13:20 Session 11B: Biomedical Applications
Linear controller proposal applied to the virtual servomechanism from a open source mechanical ventilator system

ABSTRACT. Nowadays a new infectious disease has been spread world-wide with a virus called Covid-19, this disease mainly affects the respiratory system and forces the use of mechanical ventilators unto those who suffer from it; unfortunately, the worldwide availability of such artificial ventilators is scarce. Therefore, some international efforts have been developed to get singled and cheaper mechanical ventilators, but the main challenge from this technological development has is the air flow control, because of the dynamical complexity that links the servomechanism and the bio-hydraulic system. In this context, as a first stage in the development of robust control algorithms with the described intention, this paper presents the design and analysis of a feedback vector linear controller implemented in a virtual servomechanism from an open source mechanical ventilator. The numerical validation of this feedback vector controller is analyzed through two gains tuning scenarios, and the results shows a good implementation by considering the strong nonlinearity that the system has.

Linear Sintonization of Two Cascade Vectorial Controllers from the Numerical Prediction of a Mechanical Ventilator System

ABSTRACT. With the increasing number of COVID-19 cases, the demand of ventilators has increased to such an extent that there is not enough supply of equipment available. For this reason, we have based our investigation on an open source project presented by the Massachusetts Institute of Technology; this project involves the conditioning of an Artificial Manual Breathing Unit (AMBU bag), which acts as an aid for patients to maintain a constant respiratory cycle. Currently the project is not viable, as it does not meet the requirements for its use. This document presents the calculation of the control matrix for the system so its behavior becomes similar to commercial ventilators therefore it would be available as an emergence ventilator. The calculation of the control matrix for both mechatronic and the bio pneumatic systems are detailed and simulated with MATLAB to show results.

Linear Parametric Identification from the Preliminary Setup for a Servomechanism in a Mechanical Ventilator System

ABSTRACT. Covid-19 has caused a health and economic crisis worldwide. Unfortunately, the lack of hospital infrastructure and equipment, such as ventilators, represent a significant obstacle for the medical community to treat patients suffering from this virus, therefore we decided to create an algorithm that can correctly estimate linear parameters for an auxiliary mechanical ventilation system. This work was developed in the facilities of Universidad Anahuac Querétaro and it addresses the implementation of a parameter identification algorithm on a signal produced by a system emulated in MATLAB. The project consists of parametric identification for an auxiliary mechanical ventilator system where the angular position, angular speed and torque are estimated. To ensure the lowest percentage of error possible, proposed values were chosen based on their proximity to the original given matrix. After several tests, an accuracy of 98% was achieved with the A matrix values, this allowed us to correctly estimate linear parameters.

12:00-13:20 Session 11C: Vision and Language
Photovoltaic module segmentation and thermalanalysis tool from thermal images

ABSTRACT. The growing interest in the use of clean energy has led to the construction of increasingly large photovoltaic systems. Consequently, monitoring the proper functioning of these systems has become a highly relevant issue. In this paper, a segmentation methodology of photovoltaic modules from thermal images, based on digital image processing techniques, as well as a statistical thermal analysis are developed. Besides, a graphical user interface has been designed as a potential tool that provides relevant information of the photovoltaic modules.

Detection of foliar diseases with convolutional neural networks implemented on Raspberry Pi

ABSTRACT. In this paper, an approach is described to develop deep learning models that recognize plant diseases of commercial interest through the classification of leaf images. A Raspberry Pi 4 microcomputer was used for hardware implementation. Some models employed from a subset of artificial intelligence, called Deep Learning, were used for pest detection by finetuning Transfer Learning to obtain high accuracy rates based on the PlantVillage dataset containing thirty-eight different classes, including diseased and healthy leaves. The study’s main objective is to classify various images of plants and leaves with high precision using convolutional neural networks with transfer learning implemented in a hardware device. The models were evaluated through an analysis based on precision, recall, F-score, and accuracy. The results present significant values obtained by the VGG16 technique, with 90% sensitivity and 90% accuracy. It is possible to conclude that the VGG16 trained model can be a useful tool for farmers to help and protect plants from the diseases mentioned.

Identification of agricultural parcels using optical and synthetic aperture radar data

ABSTRACT. Data fusion methodologies have been implemented in agricultural applications with different types of sensors. One of the problems in delineating cultivation areas is the mixture of spectral signatures due to the transitions between the types of cultivation, built areas, and other natural covers. In order to improve discrimination and identification of crop types, structure data fusion techniques were evaluated. This article aims at showing the potential of using satellite data from the European Space Agency, both optical and SAR, in order to improve land cover classification of agricultural land located in Mexico. To achieve this, an analysis of the spectral, spatial and textural data was performed. Specifically, two classification algorithms were used and compared. The first is based on vector support machines and the second one on Random Forests. The methodology was applied for the study of 4 types of crops in 2017 in the municipality of Villa de Arriaga located in the state of San Luis Potosí. As final results, maps were obtained with the areas with a kappa greater than 0.80.

Optimized Transmission of Encrypted Images over Wireless Channels

ABSTRACT. The transmission of images over wireless channels is increasing with advances in wireless communication technologies. However, ensuring secure and efficient transmission is a major challenge, because intrusive receivers may be present, wireless channels have high bit error rates and limited bandwidth. This paper proposes a wireless transmission scheme for encrypted images to establish reliable and robust communications. The proposed scheme comprises, on the transmitter side, compression, channel coding and modulation, and their inverse processes in the receiver. The Discrete Transformed Wavelet and Huffman coding are employed for image compression, whereas convolutional coding is used as a channel coding technique for error correction. The system is implemented with the USRP NI-2920 modules and Labview Communications 2.0 software, allowing the transmission of encrypted images in a real and simulated environment. The results show that the proposed system is robust and efficient against channel degradation.

14:00-15:00 Session 12: Keynote Lecture
Vision and Challenges of CENACE Faced with the Pandemic

ABSTRACT. On August 28, 2014, the agreement to create the National Center for Energy Control (CENACE) was published, as a decentralized public body of the Federal Public Administration, sectorized to the Ministry of Energy, with its own legal personality and assets, residing in Mexico City. The tasks assigned to CENACE were “to exercise the Operational Control of the National Electric System; the operation of the Wholesale Electricity Market and guarantee open and non-unduly discriminatory access to the National Transmission Network and the General Distribution Networks, and propose the expansion and modernization of the National Transmission Network and the elements of the General Distribution Networks that correspond to the Wholesale Electricity Market. The challenge is enormous given that electricity is the engine of economic and social development of any country, and it is the function of CENACE to execute the control in an effective and efficient way, but in a transparent and impartial way for all the actors of the National Electric System. Hence the importance of the vision that CENACE has and must have, and the challenges it must face. inthis talk, a general overview is presented about the challenges CENACE faces every day, in its daily function, in harmonizing the different types of electric power generation technologies to satisfy the country’s demand for electric power, with adequate margins of safety and reliability, sustainability, sufficiency, quality and economy. This requires and implies many fronts to fulfill its basic tasks, among which we can mention, the application of technological developments, interaction and feedback to technology developers, strategies and synergies that must be done with participants in the electricity sector, as well as the interaction with research centers and the academic sector, both national and international.