Modelling of microstructural effects in the fatigue of austempered ductile iron

Predicted global amounts of bainite within austempered ductile iron correspond well to hardness, toughness and some aspects of damage tolerance. However, local microstructural fatigue events (e.g. fatigue initiation) do not correlate simply to overall predictions of phase distributions within the mi...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Materials science & engineering. A, Structural materials : properties, microstructure and processing Structural materials : properties, microstructure and processing, 2003-04, Vol.346 (1), p.273-286
Hauptverfasser: Reed, P.A.S, Thomson, R.C, James, J.S, Putman, D.C, Lee, K.K, Gunn, S.R
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Predicted global amounts of bainite within austempered ductile iron correspond well to hardness, toughness and some aspects of damage tolerance. However, local microstructural fatigue events (e.g. fatigue initiation) do not correlate simply to overall predictions of phase distributions within the microstructure, appearing to be dependent on a complex combination of size and morphology of microstructural features. Fatigue evaluation of high hardness austempered ductile iron (ADI) has established the role of graphite nodule clustering and size in crack initiation. Adaptive numerical modelling, the SUpport vector Parsimonious ANalysis Of VAriance (SUPANOVA) approach, has been applied to image analysis data to identify those features which distinguish between the graphite nodules which initiate fatigue and those that do not. Critical crack initiating features are identified to be a combination of graphite nodule size and local clustering within a mesoscopic region containing a lower volume fraction of graphite nodules. The interpretable features introduced by the SUPANOVA technique allow visualisation of these complicated relationships, whilst maintaining a 72% successful classification rate.
ISSN:0921-5093
1873-4936
DOI:10.1016/S0921-5093(02)00545-2