Show simple item record

dc.contributor.authorFernandez Martinez, Roberto
dc.contributor.authorJimbert, Pello
dc.contributor.authorBarbero, Jose Ignacio
dc.contributor.authorCallejo, Lorena M.
dc.contributor.authorSomocueto, Igor
dc.date.accessioned2022-10-18T09:13:05Z
dc.date.available2022-10-18T09:13:05Z
dc.date.issued2019-10-22
dc.identifier.citationFernandez Martinez, Roberto, Pello Jimbert, Jose Ignacio Barbero, Lorena M. Callejo, and Igor Somocueto. “Material Fracture Life Prediction Using Linear Regression Techniques Under High Temperature Creep Conditions.” Lecture Notes in Computer Science, 2019, 535–44. https://doi.org/10.1007/978-3-030-33904-3_50.en
dc.identifier.issn0302-9743en
dc.identifier.urihttp://hdl.handle.net/11556/1421
dc.description.abstract9–12% Cr martensitic steels are widely used for critical components of new, high-efficiency, ultra-supercritical power plants because of their high creep and oxidation resistances. Due to the time consuming effort of obtaining creep properties for new alloys under high temperature creep conditions, in both short-term and long-term testing, it is often dealt with simplified models to assess and predict the future behavior of some materials. In this work, the total time to produce the material fracture is predicted according to models obtained using several linear techniques, since this property is really relevant in power plants elements. These models are obtained based on 344 creep tests performed on modified P92 steels. A multivariate analysis and a feature selection were applied to analyze the influence of each feature in the problem, to reduce the number of features simplifying the model and to improve the accuracy of the model. Later, a training-testing validation methodology was performed to obtain more useful results based on a better generalization to cover every scenario of the problem. Following this method, linear regression algorithms, simple and generalized, with and without enhanced by gradient boosting techniques, were applied to build several linear models, achieving low errors of approximately 6.75%. And finally, among them the most accurate model was selected, in this case the one based on the generalized linear regression technique.en
dc.description.sponsorshipThe authors wish to thanks to the Basque Government through the KK-2018/00074 METALCRO.en
dc.language.isoengen
dc.publisherSpringer Natureen
dc.titleMaterial Fracture Life Prediction Using Linear Regression Techniques Under High Temperature Creep Conditionsen
dc.typeconference outputen
dc.identifier.doi10.1007/978-3-030-33904-3_50en
dc.rights.accessRightsembargoed accessen
dc.subject.keywordsLinear regressionen
dc.subject.keywordsGeneralized linear regressionen
dc.subject.keywordsEnhanced linear regressionen
dc.identifier.essn1611-3349en
dc.journal.titleLecture Notes in Computer Scienceen
dc.page.final544en
dc.page.initial535en
dc.volume.number11896en
dc.conference.titleCIARP 2019: Progress in Pattern Recognition, Image Analysis, Computer Vision, and Applicationsen


Files in this item

Thumbnail

    Show simple item record