Creep-rupture prediction by naive bayes classifiers
Résumé
The purpose of this study was to predict the failure of composite materials by developing and evaluating an artificial learning algorithm that could predict their life time. This will be done by predicting whether a specimen will break within 30 seconds or not. Specimens were tested according to the creep test by the traction method. Naive Bayesian classifiers have been developed retrospectively in a group of 90 samples and tested prospectively in a group of 30 samples to evaluate and ensure the performance of this learning method. Each sample was characterized by a number of relevant parameters. During the five cross-validations, the learning machine achieved a mean sensitivity of 78% and a mean specificity of 82%. The mean area under the ROC curve (Receiver Operating Curves) reached 0.88. The study can be regarded as a very important step in the term of prediction of composite material time life remaining.
Origine : Fichiers éditeurs autorisés sur une archive ouverte
Loading...