Listagem por temas:

Escolhendo um título você terá acesso ao arquivo original em Post-Script.

COB432 AVALIAÇÃO DE DANOS ESTRUTURAIS UTILIZANDO O ALGORITMO GENÉTICO / STRUCTURAL DAMAGE ASSESSMENT USING THE GENETIC ALGORITHM

Mario Antonio Bavaresco, Francisco José da Cunha Pires Soeiro & Cicero Vianna de Abreu

Departamento de Engenharia Mecânica e de Materiais - IME/RJ, CEP 22290-270 Rio de Janeiro, Brasil

e-mail: cvabreu@taurus.ime.eb.br

Departamento de Engenharia Mecânica - UERJ, CEP 20550-013 Rio de Janeiro, Brasil - e-mail: soeiro@uerj.br

The present paper describes an approach for structural damage assessment that has its basis in methods of system identification. Response of a damaged structure differs from that obtained from the original and healthy structure. A refined analytical model that predicts correctly the response of the structure is available. The output error approach of system identification is employed to determine changes in the analytical model necessary to minimize differences between the measured and predicted response. Structural damage are represented by damage variables in which are lumped changes in the stiffness matrices resulting from variations in geometry or material properties of the structure during damage. The design space of the resulting optimization problem is highly nonconvex and has several local optima. In some cases methods of mathematical programming using gradients are not able to detect damage. This work presents a departure from that approach using one stochastic method of global optimization which is the genetic algorithm. Measurements of static deflections are used in the identification procedure. Principal shortcomings in the proposed approach and methods to circumvent these problems are also discussed.

Keywords: Structural Damage Assessment, Genetic Algorithms, System Identification, Structural Optimization

Avaliação de Danos Estruturais, Algoritmos Genéticos, Identificação de Sistemas, Otimização de Estruturas

COB450 SUBSTITUIÇÃO DAS CHAVETAS DA COROA POLAR DO ROTOR DA UG-07 DA UHE ILHA SOLTEIRA / REPLACEMENT OF THE MAGNETIC WHEEL’S SPLINES IN THE GENERATOR # 7 OF ILHA SOLTEIRA POWER PLANT

Orlando Marcondes Machado Filho

Setor de Geração de Ilha Solteira -CESP - Companhia Energética de São Paulo, CEP 15385-000, Ilha Solteira, Brasil.

In present times, under reduction of the financial investiments resources and increase of the demands for electrical power, the improvement of the maintenance actions and increase of the available time in big generators units are a must. Allowing for the maintenance know how and experience of the ABB and CESP, it was possible a new and dared solution for the rehabilitation of a 480 ton’s weight generator rotor by the replacement of its magnetic wheel’s splines. This solution meant a quality level profit and kept the maintenance time restricted to the original preventive maintenance program of generator unit.

Keywords: Manutenção - Máquinas Rotativas - Gerador - Chavetas - UHE Ilha Solteira

Maintenance - Rotating Generator - Hydroelectric Generator - Splines - Ilha Solteira Power Plant

COB458 ANÁLISE DE FALHA EM REDUTORES UTILIZANDO TRANSFORMADA DE WAVELET/ GEARBOX FAULT ANALYSIS USING WAVELET TRANSFORM

Mamede, W. F. & Teodoro, E. B.

DEEME - Universidade Federal de Uberlândia - Campus Santa Mônica

Caixa Postal 593 - Rua Duque de Caxias, 285, e-mail teodoro@ufu.br, 38400-902 - Uberlândia/MG

The Wavelet Transform - WT is a powerful tool used for analysis of non-stationary signals, but it has not been frequently used in this area. A brief mathematical formulation of WT is presented along with various wavelet functions used in this paper. The main feature of WT is to present at the same graphic information of the signal in the time and frequency domain. Therefore it is possible to have different resolutions and simultaneous information of the signal in both domains. A series of simulated signals are generated by computer and processed with the proposed transformation in order to understand how it works. "Doing so we can learn more about the WT concepts such as: scale, level, signal reconstruction and translation". Experimental results from signals measured in a gearbox are presented, when processed with the WT. The analysis of these results showed that WT is more appropriate to locate transients in non-stationary signals than the well known and used Fourier Transform. The WT technique also proved to be possible the detection of transients in the signal, generated by local faults in the gear through time versus scale plots.

Keywords: Transformada de Wavelet - Redutores - Identificação - Engrenagens - Processamento de Sinais.

Wavelet Transform - Gearbox - Identification - Gear - Signal Processing.

COB480 PREVISÃO DO COMPORTAMENTO VIBRATÓRIO DE SISTEMAS MECÂNICOS / MECHANICAL SYSTEMS VIBRATION BEHAVIOUR PREDICTION

Rivaldo Araújo Vilela & Robson Pederiva

Departamento de Projeto Mecânico, FEM - UNICAMP, Caixa Postal 6051- Cidade Universitária, CEP 13083-970 - Campinas/SP, Brasil E-mail: robson@fem.unicamp.br

An important information about mechanical systems is the fact that vibration phenomena is the chief cause and also the consequence of the main wear processes. This information is used to set up a feedback ARMAX forecasting model for machines which have on wear its main source of faults. For evaluating purpose, the ARMAX forecasting model is used for predicting the trend of some simulated cases. Satisfactory results have been obtained.

Keywords: Machinery prognostics, failure prediction, trend analysis, ARMAX models, predictive maintenance, prognóstico em máquinas, previsão de falhas, análise de tendência, modelos ARMAX, manutenção preditiva.

COB481 DIAGNÓSTICO DE DEFEITOS EM UM CONJUNTOMOTO-BOMBA BASEADO EM CRITÉRIOS ESTATÍSTICOS / FAULT DIAGNOSTIC IN A CENTRIFUGAL PUMP BASED ON STATISTICAL PATTERN RECOGNITION

Marcus Vinícius Portari & Robson Pederiva

Departamento de Projeto Mecânico, Faculdade de Engenharia Mecânica - UNICAMP - CEP 13083-970 Campinas, Brasil

E-mail: robson@dpm.fem.unicamp.br

This work analyzes two fault diagnostic methods one based on Euclidean distance and another on Mahalanobis distance. These methods don’t require high efforts in learning phase, are easily implemented and present satisfactory results in machinery fault diagnostic. An experimental setup involving a centrifugal pump was designed to analyze these methods. Some faults like unbalance, suction and discharge piping obstructions, wear on rotor blades were studied. Data from healthy and fault conditions were used to evaluate the sensitivity (ST) of the two methods.

Keywords: Diagnostic, fault, monitoring, pattern recognition.

COB492 DETECÇÃO E LOCALIZAÇÃO DE FALHAS EM SISTEMAS MECÂNICOS ATRAVÉS DA ANÁLISE DE CORRELAÇÕES / FAULT DETECTION AND ISOLATION IN MECHANICAL SYSTEMS THROUGH CORRELATIONS ANALYSIS

André Garcia Chiarello, Robson Pederiva

Departamento de Projeto Mecânico, Faculdade de Engenharia Mecânica, UNICAMP

Cidade Universitária, CP 6051, CEP 13083-970, Campinas, SP, Brasil, E-mail: garcia@fem.unicamp.br

In the context of model based fault detection and diagnosis, there is a residual generation which can be obtained through redundancy equations. The redundancy equations ( analytical redundancy ) are based on the use of mathematical models of the process being monitored for faults. This paper presents a new formulation involving the generation of redundancy equations for linear, time invariant, stationary systems, based on correlations functions of the system. The proposed approach is demonstrated using a simulated model of mechanical system

Keywords: Detecção de falhas, diagnose de falhas, equações redundantes, sistemas mecânicos, funções de correlação / Fault detection, failure detection, fault diagnosis, parity equations, redundancy equations, correlations functions, model based, stationary systems.

COB983 DIAGNOSE DE FALHAS ATRAVÉS DE OBSERVADORES DE ESTADO EM SISTEMAS COM PARÂMETROS DESCONHECIDOS / FAULTS DETECTION THROUGH STATE OBSERVERS IN SYSTEMS WITH UNKNOWN PARAMETERS

Gilberto Pechoto De Melo* & Robson Pederiva**

*DEM - FEIS - UNESP- Ilha Solteira,C. Postal 31 - Cep 15385-000 Ilha Solteira S.P., Email:gilberto@feis.unesp.br

**DPM - FEM - UNICAMP - Campinas,C. Postal 6051 - Cep 13081-970 Campinas S.P, Email:robson@dpm.fem.unicamp.br

The application of the deterministic state observer techniques for machine monitoring can be severely restricted because of limitations in accuracy reliability and robustness. The observers can produce large transient or steady-state errors, which can cause frequent false alarms and these limitations in transient and steady-state performance can be described by ill-conditioning factors.This paper describes a Model-based monitoring system on state observer theory with unknown parameters. A time-domain technique based on Fourier series is used for the parameters identification. The method is based on the orthogonality property of the Fourier series which enables the integration of the equations of motion. These equations are then converted to a linear algebraic model that is solved to obtain the unknown parameters. The goal is monitoring a mechanical system with an automatic system for detection and isolation of the fault using robust observation.

Keywords: Faults Detection and Isolation, Parameters Identification, State Observers/Observadores de Estado, Falhas.