Estudo do comportamento à fadiga de compósitos utilizando rede neural artificial modular

Artigo publicado no periódico International Journal of Fatigue, por Prof. Dr. Bruno da Cunha Diniz e Prof. Dr. Raimundo Carlos Silvério Freire Júnior

Título da pesquisa: Estudo do comportamento à fadiga de compósitos utilizando rede neural artificial modular com a incorporação de probabilidade de falha à posteriori

Título original: Study of the fatigue behavior of composites using modular ANN with the incorporation of a posteriori failure probability

Palavras chave: Composites, Fatigue, Weibull distribution, Modular ANN

Disponível em: https://www.sciencedirect.com/science/article/pii/S014211231930461X

Resumo:

"Resumo: To study the fatigue behavior of composites, a large number of experimental assays for both deterministic and probabilistic analysis are needed. Recent studies have demonstrated mathematical and methodological models that are applied to determine the deterministic fatigue of composite materials, but to date, no reliable model has been developed to analyze probabilistic fatigue behavior using a small amount of experimental data and considering failure probability in analysis. As such, this study aimed to develop an artificial neural network (ANN) with modular architecture in order to model probabilistic fatigue behavior, using only three S-N curves in training and applying a posteriori failure probability (after ANN training). To that end, two methodologies were used to obtain Weibull distribution parameters and this result was incorporated into deterministic ANN architecture The advantage of this strategy is that training and using a deterministic ANN has demonstrated repeatability (robustness) in the training and generalization of the final result, while attempts to train an ANN with probabilistic data do not always obtain satisfactory results after training.".

Fig.1: Rede neural de arquitetura modular com entradas: Número médio de Ciclos e Tensão média e com saída: Tensão alternada

Fig.2: Método 1: Obtenção da rede neural probabilística utilizando equacionamento

Fig.3: Método 2: obtenção da rede neural probabilística utilizando uma RNA determinística na análise da probabilidade

Fig.4: Erro Médio Quadrático (RMS) obtido com a comparação entre os dois métodos considerando três valores de probabilidade de falha

Fig.5: Diagrama de vida constante para o compósito laminado de fibra de carbono DD16 para probabilidade de falha de 10% obtida pelo método 1.

Fig.6: Diagrama de vida constante para o compósito laminado de fibra de carbono DD16 para probabilidade de falha de 10% obtida pelo método 2.

The results demonstrate that the methodologies used in this article make it possible to reliably obtain probabilistic fatigue behavior at probabilities greater than or equal to 5%. Method 2 (which uses the deterministic ANN to obtain parameters α and β) is more conservative than method 1 and should be used when greater design safety is required. It can be concluded that the methods presented here are repeatable and reliable, even when a small amount of experimental data was used, and can be applied with other analytical models instead of an ANN.