Antônio Morais da Silveira

Case study: Comparing the use of nonlinear discriminating analysis and Artificial Neural Networks in the classification of three fish species: acaras (Geophagus brasiliensis), tilapias (Tilapia rendalli) and mullets (Mugil liza)

This study used the Discriminant Analysis statistical technique and Artificial Neural Networks, multilayer perceptron, in the classification of three fish species sampled in the state of Rio de Janeiro, Brazil: Geophagus brasiliensis (acaras), Tilapia rendall (tilapias) and Mugil liza (mullets). These fish were sexed when possible, weighed, measured, and had their Gonadosomatic and Hepatosomatic Indices calculated, as well as their Condition Factor. The use of an Artificial Neural Network (ANN) presented satisfactory results, even though the groups were composed of very diverse-sized animals. Without the need for non-violation assumptions and other considerations, the Artificial Neural Network was found to be the excellent alternative to classification problems of unbalanced data, such as the one presented in this study.

Metal and physico-chemical variations at a hydroelectric reservoir analyzed by Multivariate Analyses and Artificial Neural Networks: Environmental management and policy/decision-making tools

This paper compared and evaluated seasonal variations in physico-chemical parameters and metals at a hydroelectric power station reservoir by applying Multivariate Analyses and Artificial Neural Networks (ANN) statistical techniques. A Factor Analysis was used to reduce the number of variables: the first factor was composed of elements Ca, K, Mg and Na, and the second by Chemical Oxygen Demand. The ANN showed 100% correct classifications in training and validation samples. Physico-chemical analyses showed that water pH values were not statistically different between the dry and rainy seasons, while temperature, conductivity, alkalinity, ammonia and DO were higher in the dry period. TSS, hardness and COD, on the other hand, were higher during the rainy season. The statistical analyses showed that Ca, K, Mg and Na are directly connected to the Chemical Oxygen Demand, which indicates a possibility of their input into the reservoir system by domestic sewage and agricultural run-offs. These statistical applications, thus, are also relevant in cases of environmental management and policy decision-making processes, to identify which factors should be further studied and/or modified to recover degraded or contaminated water bodies.

Categorization of the trophic status of a hydroelectric power plant reservoir in the Brazilian Amazon by statistical analyses and fuzzy approaches.

The Amazon area has been increasingly suffering from anthropogenic impacts, especially due to the construction of hydroelectric power plant reservoirs. The analysis and categorization of the trophic status of these reservoirs are of interest to indicate man-made changes in the environment. In this context, the present study aimed to categorize the trophic status of a hydroelectric power plant reservoir located in the Brazilian Amazon by constructing a novel Water Quality Index (WQI) and Trophic State Index (TSI) for the reservoir using major ion concentrations and physico-chemical water parameters determined in the area and taking into account the sampling locations and the local hydrological regimes. After applying statistical analyses (factor analysis and cluster analysis) and establishing a rule base of a fuzzy system to these indicators, the results obtained by the proposed method were then compared to the generally applied Carlson and a modified Lamparelli trophic state index (TSI), specific for trophic regions. The categorization of the trophic status by the proposed fuzzy method was shown to be more reliable, since it takes into account the specificities of the study area, while the Carlson and Lamparelli TSI do not, and, thus, tend to over or underestimate the trophic status of these ecosystems. The statistical techniques proposed and applied in the present study, are, therefore, relevant in cases of environmental management and policy decision-making processes, aiding in the identification of the ecological status of water bodies. With this, it is possible to identify which factors should be further investigated and/or adjusted in order to attempt the recovery of degraded water bodies.

Logistic regression and fuzzy logic as a classification method for feral fish sampling sites.

This study presents a classification method combining logistic regression and fuzzy logic in the determination of sampling sites for feral fish, Nile Tilapia (Tilapia rendalli). This method statistically analyzes the variable domains involved in the problem, by using a logistic regression model. This in turn generates the knowledge necessary to construct the rule base and fuzzy clusters of the fuzzy inference system (FIS) variables. The proposed hybrid method was validated using three fish stress indices; the Fulton Condition Factor (FCF) and the gonadossomatic and hepatossomatic indices (GSI and HSI, respectively), from fish sampled at 3 different locations in the Rio de Janeiro State. A multinomial logistic regression allowed for the FIS construction of the proposed method and both statistical approaches, when combined, complemented each other satisfactorily, allowing for the construction of an efficient classification method regarding feral fish sampling sites that, in turn, has great value regarding fish captures and fishery resource management.

Generalized minimum variance control under long-range prediction horizon setups

This paper presents the design and evaluation of a minimal order Generalized Minimum Variance controller with long-range prediction horizon and how it affects the controller and plant output variances. This study investigates how the increased prediction horizon can contribute to mitigate stochastic disturbances and attenuate oscillations. In order to design high order prediction minimum variance filters, a design procedure independent of the Diophantine Equation solution is used. The evaluation is conducted through simulations and practical essays with two different plants: a first order water flow rate problem and a second order under-damped electronic circuit. Both problems are assessed under an incremental control scheme and based on identified stochastic models. Also, two optimal tuning procedures for the algorithm are proposed.

Partial discharge analysis and inspection alert generation in high power transformers: a case study of an autotransformer bank at Eletrobrás-ELETRONORTE vila do conde station

This paper presents an exploratory study using statistical and IA techniques in the partial discharge database located in Vila do Conde substation, Barcarena, Para state, Brazil, ELETRONORTE property. Through ambiental and system variables analysis, it was possible to identify that the 230kV reactive power and period of day have a strong relation to partial discharge measures. With the obtained knowlegde and specialists knowlegde, a initial fuzzy system is proposed for partial discharge classification in diferents operational situations of alert, contributing to the operational status diagnosis of power transformers and amplifying the knowledge about the theme.

On the design of stochastic RST controllers based on the generalized minimum variance

This paper presents a general framework based on the RST structure for performing fair comparisons between deterministic and stochastic digital linear controllers. Due to the chosen RST formulation, it can be shown that any linear Single-Input Single-Output controller may be designed as a Generalized Minimum Variance Controller, i.e. as a stochastic controller. This in particular reduces the variance of the control signal and in turn leads to a better output characteristic. The method is applied to the controllers of an exemplary Single-Machine Infinite Bus system, namely the AVR and governor, where it is shown that the voltage and power regulation for the non-stochastic and the GMVC are similar whereas the control signal given by the latter is much smother than for the former, even using the same set of gains for both.

On the unrestricted horizon predictive control — A fully stochastic model-based predictive approach

This paper presents the design of the Unrestricted Horizon Predictive Controller, short UHPC, evaluated from the extended horizon of a minimal order Generalized Minimum Variance controller investigated in a free Diophantine equation form based on ARMAX models. In contrast to common MPC methods, UHPC does not employ the receding horizon approach for calculation of its future control steps, but a state space solution where a 1-step ahead Kalman Filter and the inherent calculation of two Diophantine equations - one for the control signal and other for the noise - are utilized. Simulation results show that UHPC performs well compared to the most common approaches (Minimum Variance and Generalized Predictive Control) for several different plant conditions, such as regulation (noise rejection) and reference tracking.

An Intelligent Method for Scheduling Inspections in Transformers Based on Analysis of Combustible Gases and Physico-Chemical Parameters of the Insulating Oil

This paper presents a proposed solution using fuzzy approach, to assist the process of scheduling inspections in high power transformers (500KV), through the analysis of dissolved gases in the insulating oil, obtained using the technique of gas chromatography, and physico-chemical analysis of the oil. The development of the method was based on the criteria of gas analysis referenced standards, the statistical analysis of data and the tacit knowledge of experts. Sought to produce a solution which combines the results of traditional methods already established in the technical literature with additional situations arising from the physical-chemical analysis of the oil and the knowledge of experts, in order to increase the efficiency of such equipment inspection procedure. Keywords-component; Fuzzy Logic; Gas Chromatography; Physico-chemical; Factorial.

Computational intelligence applied to competitiveness evaluation of supply chains: an adaptive neuro-fuzzy model

Technological advances and economic turmoil are some of the factors leading to increased competition in a global scale, which led companies, industries and countries to be concerned to maintain a leadership position in the market for competitive advantage. One way to achieve this goal is to make use of computer technologies to facilitate and accelerate decision-making in these environments of uncertainty. This work aims to show the use of a hybrid approach of Artificial Neural Networks and Fuzzy Logic, an Adaptive Neuro-Fuzzy system, to supply chain competitiveness evaluation. To validate the method is used a case of study based on the supply chain of broilers in Brazil. The results were satisfactory considering the low errors obtained in the validation tests.