Analyse et caractérisation multidimensionnelles pour l’optimisation du fini de surface de pièces en composites

Content Provider	Semantic Scholar
Author	Rioux, Michel
Copyright Year	2011
Abstract	RESUME Mots-cles : Fini de surface, RTM, caracterisation multidimensionnelle, rugosite Dans le domaine des transports, le procede RTM presente beaucoup de potentiel. Celui-ci necessite cependant encore plusieurs ameliorations, notamment en ce qui a trait a la qualite du fini de surface. Etant donne qu’il n’existe pas encore de theorie permettant de modeliser ce phenomene, l’approche empirique devient necessaire. La theorie des plans d’experience est toute designee, mais elle necessite d’avoir une reponse quantitative representant bien le phenomene. Les lacunes de l’indice de rugosite usuel (Ra) comme parametre de caracterisation sont demontrees. Toutefois, en utilisant les profils bruts mesures par un rugosimetre, plusieurs statistiques ont ete calculees en s’inspirant des techniques de traitement du signal, de la statistique classique ou des series chronologiques. Une methode de redressement des profils a ete mise au point. Ensuite, a l’aide d’une analyse multidimensionnelle, quatre variables de caracterisation (minimum, maximum, asymetrie et aplatissement) ont ete retenues. Bien que d’allure elementaire, ces quatre variables sont non seulement tres complementaires mais egalement beaucoup plus riches d’information que des Ra, meme avec differents seuils de coupure frequentielle. Ces quatre variables de caracterisation comblent le besoin limitant l’emploi des plans d’experiences dans ce domaine. Elles peuvent meme fournir de l’information sur les types de defauts predominants (min./cavites, max./asperites, asymetrie/cretes tronquees et aplatissement/denivellations). Elles peuvent egalement etre agregees. Parce qu’elles contiennent de l’information redondante, l’usage de la distance de Mahalanobis s’est demontree utile. La presente recherche demontre l’interet des analyses statistiques dans le domaine des composites.----------ABSTRACT Keywords: Surface finish, RTM, multidimensional characterization, roughness In the field of transportation, the RTM process offers a lot of potential. It however requires several improvements, notably regarding the quality of surface finish. Since no theory currently allows to model the process, the empirical approach is necessary. The experimental design theory is an obvious choice, but it requires a quantitative response to represent this phenomenon. The shortcomings of the usual roughness index (Ra) as a characterization parameter are demonstrated. However, by using the raw profiles measured by a profilometer, several statistics were calculated based on signal processing techniques, statistics and time series. A profile straightening method was developed. Four characterization variables (i.e. minimum, maximum, skewness, and kurtosis) were subsequently selected by resorting to a multivariate analysis. Although elementary-looking, these four variables are not only highly complementary but they also provide much richer information than Ras, even with different cutoff frequency thresholds. These four characterization variables fill the need restricting the use of design of experiments in this field. They can even provide information on the types of predominant defects (min/cavities, max/peeks, asymmetry/truncated crests, and kurtosis/depressions. They can also be aggregated. Because they contain redundant information, the use of the Mahalanobis distance has proven useful. This research shows the relevance of statistical analysis in the field of composites.
File Format	PDF HTM / HTML
Alternate Webpage(s)	https://publications.polymtl.ca/563/2/2011_MichelRioux.pdf
Language	English
Access Restriction	Open
Content Type	Text
Resource Type	Article