Sistema Regional de Información
en línea para Revistas Científicas de América Latina,
el Caribe, España y Portugal

Context: This article contains an analysis of the applications of different Deep Learning and Machine Learning techniques used in a wide range of industries to ensure quality control in finished products through the identification of surface defects.Method: A systematic review of the trends and applications of Deep Learning in quality processes carried out. After consulting several databases, the articles were filtered and classified by industry and specific work technique applied to later analyze their usefulness and performance.Results: The results show by means of success cases the adaptability and potential applicability of this artificial intelligence technique to almost any process stage of any product, due to the handling of complementary techniques that adjust to the different particularities of the data, production processes, and quality requirements.Conclusions: Deep Learning, complemented with techniques such as Machine Learning or Transfer Learning, generates automated, accurate, and reliable tools to control the quality of production in all industries.

Context: Infrastructure enables the satisfaction of the population’s needs and contributes significantly to the economic development of countries and regions. However, Flyvbjerg points out that the success rate of construction projects is estimated at only 25 % and, particularly in megaprojects, it is 8 successful projects per 1000. On the other hand, several studies point out that complexity has negative effects on project performance, so it is of interest to evaluate such complexity and to sensibilize project managers to anticipate its negative effects. Method: Trough a literature review, four relevant complexity models were identified. Using a heuristic analysis technique, they were analyzed in three aspects: 1) factors contributing to project complexity, 2) types of projects and their specific complexity factors and, 3) techniques and tools used in the models to study project complexity. Results: The most comprehensive model is Lessard, Sakhrani & Miller’s HoPC. By considering the project’s life cycle, on Bosch-Rekveldt’s TOE framework, seven complementary complexity aspects were identified: project architecture, financial complexity, governance, the validation process of project stages, project management maturity, cultural aspects, and the regulatory framework. Conclusions: Recent studies highlight that environment and externalities are increasingly relevant in assessing the complexity of infrastructure construction projects. Projects exhibit aspects of complexity depending on their internal components and on the specific context in which they are undertaken, so the development of subject-specific models is recommended. Project complexity has been addressed mainly from Project Management and Systems Dynamics approaches, however, for the study of the diversity, interdependence, and dynamics among the complexity factors future research from the Complex Systems approach is needed.

Context: Simulation and wear joint mechanisms have been studied and applied in knee biomechanical systems for more than 30 years. However, these have not been widely reported with regard to their control and/or automation strategies. This work aims to present the advances made in the technological development of the different platforms and models of knee simulators, based mainly on the Oxford and Stanmore platforms.Method: An exhaustive review of commercial equipment patents and scientific papers was conducted. The approach considered the kinematics and dynamics of the platforms and the control models actuators, interface, and tuning method used, as well as the tests conducted and the system error. Results: Biomechanical knee systems have not been widely reported, as far as their control and/or automation strategies are concerned, because many of them are commercial and patented. Some platforms are certified under certain standards but depend only on the controlled variable. In addition, a detailed comparison of the different types of existing platforms is presented, highlighting the hydraulic models with PID controllers. Conclusions: There is an area of opportunity to propose new design alternatives and/or control strategies for knee simulators. This, in turn, opens the possibility of developing new designs for other joints, as well as improved models of existing ones.

Editorial Vol.27 No.1

Editorial Vol.26 No.3

Context: The accelerated growth of cities and rural areas requires the adequate expansion of electrical distribution systems in order to meet the electrical energy requirements with efficiency, reliability, and safety for commercial, residential, and industrial users. To serve the different users of the electrical network, a typical methodology used by network operators corresponds to the optimal assignment of the calibers of the conductors associated with the distribution routes. This selection is made while considering its cost of investment and operation for a determined planning horizon. Method: To solve the problem regarding optimal selection in three-phase distribution networks, the application of an optimization algorithm of the family of combinatorial techniques known as Newton’s metaheuristic algorithm (NMA) is proposed. The main advantage of the NMA is that it uses evolution rules based on the first and second derivatives of the objective function, which are applied to each individual in the population. In addition, the evolution rules of the NMA cause this algorithm to have a proper balance between the exploration and exploitation of the solution space as the iterative process advances. Results: Numerical validations in two three-phase distribution systems composed of 8 and 27 nodes with balanced and unbalanced operation scenarios show that the NMA reaches the optimal solution reported in the literature for the 8-node system and improves the scientific reports for the 27-node test system in both test scenarios. Conclusions: The results obtained through the application of the NMA to the problem regarding the optimal selection of conductors in distribution systems demonstrate the effectiveness of the proposed solution methodology for mixed integer nonlinear programming problems in electrical engineering with exponential sizes of the solution space. Furthermore, the reported results for the 8-node and 27-node test systems constitute a benchmark for further research.

Context:  In the education sector, variables have been identified which considerably affect students’ academic performance. In the last decade, research has been carried out from various fields such as psychology, statistics, and data analytics in order to predict academic performance. Method: Data analytics, especially through Machine Learning tools, allows predicting academic performance using supervised learning algorithms based on academic, demographic, and sociodemographic variables. In this work, the most influential variables in the course of students’ academic life are selected through wrapping, embedded, filter, and assembler methods, as well as the most important characteristics semester by semester using Machine Learning algorithms (Decision Trees, KNN, SVC, Naive Bayes, LDA), which were implemented using the Python language. Results: The results of the study show that the KNN is the model that best predicts academic performance for each of the semesters, followed by Decision Trees, with precision values that oscillate around 80 and 78,5% in some semesters. Conclusions: Regarding the variables, it cannot be said that a student’s per-semester academic average necessarily influences the prediction of academic performance for the next semester. The analysis of these results indicates that the prediction of academic performance using Machine Learning tools is a promising approach that can help improve students’ academic life allow institutions and teachers to take actions that contribute to the teaching-learning process.

Context: The power flow is a classical problem for analyzing and operating power distribution networks. It is a challenging problem due to a large number of nodes, the high $r/x$ ratio -typical in low voltage networks- and the unbalanced nature of the load. Method: This paper review four methods for power flow analysis, namely: the conventional Newton's method, Newton's method in a complex domain, the fixed-point algorithm using $Y_\text{bus}$ representation, and the backward-forward sweep algorithm. It is well-known that Newton's method has quadratic convergence, whereas the backward-forward sweep algorithm has linear convergence. However, the formal analysis of this convergence rate is less known in the engineering literature. Thus, the convergence of these methods is presented in theory and practice. Results: A set of simulations in the IEEE 900 node test system is presented. This system is large enough to demonstrate the performance of each algorithm. In addition, a Matlab toolbox is presented for making numerical simulations both for the static case and for quasi-dynamic simulations. Conclusions: Fixed point algorithms were faster than Newton's methods. However, the latter required less number of iterations.

Nowadays, optimization methodologies based on combinatorial strategies (i.e., metaheuristic methods) and exact methods can be easily found through-out the scientific literature in all areas of engineering, including electrical, mechanical, chemical, computational, and food engineering, among others. The common denominator in these areas of research corresponds to the complexity of the optimization models, as well as to the large dimensions of the solution space where these models are defined [1]. In addition, most of these models combine binary (also integer) decision variables with continuous ones into nonlinear non-convex formulations, which complicates the application of exact solution methods.

A Possible Classification for Metaheuristic Optimization Algorithms in Engineering and Science