Paulo Eduardo Maciel de Almeida

A comprehensive review for industrial applicability of artificial neural networks

This paper presents a comprehensive review of the industrial applications of artificial neural networks (ANNs), in the last 12 years. Common questions that arise to practitioners and control engineers while deciding how to use NNs for specific industrial tasks are answered. Workable issues regarding implementation details, training and performance evaluation of such algorithms are also discussed, based on a judiciously chronological organization of topologies and training methods effectively used in the past years. The most popular ANN topologies and training methods are listed and briefly discussed, as a reference to the application engineer. Finally, ANN industrial applications are grouped and tabulated by their main functions and what they actually performed on the referenced papers. The authors prepared this paper bearing in mind that an organized and normalized review would be suitable to help industrial managing and operational personnel decide which kind of ANN topology and training method would be adequate for their specific problems.

Neural optimal control of PEM-fuel cells with parametric CMAC networks

This work shows an application of the parametric CMAC (P-CMAC) network, a neural structure derived from Albus CMAC algorithm and Takagi-Sugeno-Kang parametric fuzzy inference systems. It resembles the original CMAC proposed by James Albus in the sense that it is a local network, i.e., for a given input vector, only a few of the networks nodes (or neurons) will be active and will effectively contribute to the corresponding network output. The internal mapping structure is built in such a way that it implements, for each CMAC memory location, one linear parametric equation of the network input strengths. First, a new approach to design neural optimal control (NOC) systems is proposed. Then, P-CMAC is used to control output voltage of a proton exchange membrane-fuel cell (PEM-FC), by means of NOC. The proposed control system allows the definition of an arbitrary performance/cost criterion to be maximized/minimized, resulting in an approximated optimal control strategy. Practical results of PEM-FC voltage behavior at different load conditions are shown, to demonstrate effectiveness of the NOC algorithm.

Parametric CMAC networks: fundamentals and applications of a fast convergence neural structure

This work shows fundamentals and applications of the parametric CMAC (P-CMAC) network, a neural structure derived from Albus CMAC algorithm and Takagi-Sugeno-Kang parametric fuzzy inference systems. It resembles the original CMAC proposed by James Albus in the sense that it is a local network, i.e., for a given input vector, only a few of the networks nodes (or neurons) will be active and will effectively contribute to the corresponding network output. The internal mapping structure is built in such a way that it implements, for each CMAC memory location, one linear parametric equation of the network input strengths. This mapping can be thought of as the corresponding of a hidden layer in a multi-layer perceptron (MLP) structure. The output of the active equations are then weighted and averaged to generate the actual outputs to the network. A practical comparison between the proposed network and other structures is accomplished. P-CMAC, MLP and CMAC networks are applied to approximate a nonlinear function. Results show advantages of the proposed algorithm, based on the computational efforts needed by each network to perform nonlinear function approximation. Also, P-CMAC is used to solve a practical problem at mobile telephony, approximating a RF mapping at a given region to help operational people while maintaining service quality.

Applications of Celluar Neural Networks for Shape from Shading Problem

The Cellular Neural Networks (CNN) model consist of many parallel analog processors computing in real time. CNN is nowadays a paradigm of cellular analog programmable multidimensional processor array with distributed local logic and memory. One desirable feature is that these processors are arranged in a two dimensional grid and have only local connections. This structure can be easily translated into a VLSI implementation, where the connections between the processors are determined by a cloning template. This template describes the strength of nearest-neighbour interconnections in the network. The focus of this paper is to present one new methodology to solve Shape from Shading problem using CNN. Some practical results are presented and briefly discussed, demonstrating the successful operation of the proposed algorithm.

A novel movable partitions approach with neural networks and evolutionary algorithms for solving the hydroelectric unit commitment problem

This paper presents a method based on Neural Networks and Evolutionary Algorithms to solve the Hydroelectric Unit Commitment Problem. A Neural Network is used to model the production function and a novel approach based on movable partitions is proposed, which makes it easier to model the desired power output equality constraint in the optimization modeling. Three evolutionary algorithms are tested in order to find optimized operation points: differential evolution DE/best/1/bin, a balanced version of DE and Particle Swarm Optimization algorithm (PSO). The results show that the proposed method is effective in terms of water consumption, reaching in some cases more than 1% of economy whether compared to the traditional commitment strategy.

A novel mathematical modeling approach to the electric dispatch problem: Case study using Differential Evolution algorithms

Nowadays, the population growth and economic development causes the need for electricity power to increase every year. An unit dispatch problem is defined as the attribution of operational values to each generation unit inside a power plant, given some criteria to be obeyed like the total power to be generated, operational bounds of these units etc. In this context, an optimal dispatch programming for hydroelectric units in energy plants provides a bigger production of electricity to be generated with a minimal water amount. This paper presents an optimization solution for hydroelectric generating system of a plant, using Differential Evolution algorithms. The novel mathematical model proposed and validation of the obtained algorithms will be performed with practical simulation experiments. Throughout the text, the equations and models for the system simulation will be fully described, and the experiments and results will be objectively analysed through statistical inference. Simulation results indicate savings of 6.5 million litres of water for each month of operation using the proposed solution.

Mobile telephony RF mapping using fuzzy-CMAC neural networks

In this work, an engineering solution to a very common problem in mobile telephony is proposed using the artificial neural networks theory. The radio frequency (RF) level mapping problem is described and some considerations about its solution are taken. The structures and operation details of CMAC and fuzzy-CMAC neural networks are also discussed and a solution to the above problem is derived. The steps needed to reach accurate and reliable results are depicted. Results obtained with real field data acquired by Telemig Cellular Company from Brazil are shown and a final analysis is made considering the performance achieved with the described approach. Finally, some improvements to the presented solution are proposed, as suggestions to continuing the current research.

Correcting geometric deviations of CNC Machine-Tools

An experimental methodology of Design for Manufacturing (DFM) is used for survey and analysis of geometric deviations of a CNC Machine-Tools.Artificial Neural Networks (ANN) with back propagation algorithm (BPNN) has been applied to predict the fabrication parameters.The performance of the trained neural network has been tested for compensation of geometric deviations of a CNC Machine-Tools. This paper presents an experimental methodology of Design for Manufacturing (DFM) used for survey and analysis of geometric deviations of CNC Machine-Tools, through their final product. These deviations generate direct costs that can be avoided through the use of Intelligent Manufacturing Systems (IMS), by the application of Artificial Neural Networks (ANNs) to predict the fabrication parameters. Finally, after the experiments, it was possible to evaluate the experimental methodology used, the equations, the variables of data adjustment and thus enable the validation of the methodology used as a tool for DFM with high potential return on product quality, development time and reliability of the process with wide application in various CNC Machines.

Simulação da movimentação de pessoas em situações de emergência: aspectos ergonômicos e computacionais com autômatos Fuzzy e sua aplicação ao projeto arquitetônico

Este artigo apresenta o programa computacional Fuga, desenvolvido para simular a movimentacao de pessoas em ambientes construidos durante uma situacao de abandono. Esse programa se baseia na modelagem celular e possui como principais paradigmas a utilizacao de aspectos ergonomicos associados a movimentacao humana e a utilizacao da Logica Fuzzy como ferramenta de inteligencia computacional para emulacao do processo de tomada de decisao humana. Os aspectos ergonomicos e o processo de tomada de decisoes sao apresentados. E realizada uma validacao do modelo, assim como varias simulacoes, ilustrando como pode ser utilizado na concepcao de ambientes mais seguros, de uma forma que dificilmente seria obtida pela simples aplicacao das legislacoes vigentes.

Simulation of People Flow by a New Fuzzy Discrete Automata Model and an Ergonomic Approach

This work presents the computer program FUGA v. 1.0, developed to simulate the movement of people in constructed environments in normal situations and also during an evacuation in emergency situations. FUGA is based on a discrete automata model using pre-defined rules. This program uses an ergonomic approach associated with human movement and fuzzy logic as a computer intelligence tool to emulate the human decision-making process. The model incorporates mechanical and mental aspects, as well as their quantitative and qualitative nature. This work shows how selected ergonomic quantities are incorporated into a human decision-making process emulated by a fuzzy logic system. FUGA simulates environments with any internal or external geometry; with one or more floors; with or without staircases or ramps and with uni or multi directional flows. Some simulations are performed showing how the software FUGA can be used in the design of safer environments, in a way that could hardly be achieved by simply applying the existing regulations.