Experimental Investigation of Crack Jumps during Initiation and Growth in IN718

TL;DR

This study examines the statistical behavior of crack jumps during fatigue in Inconel 718 using in-situ measurements, revealing non-random crack length changes that can inform machine learning models for predicting crack growth.

Contribution

It provides the first detailed statistical analysis of crack jumps in Inconel 718 during fatigue, highlighting non-random patterns useful for predictive modeling.

Findings

Crack length change is not a random process.

Time history features can aid machine learning predictions.

Results applicable across different heat treatments.

Abstract

The following study investigates the statistical nature of crack jumps during fatigue of Inconel 718. In-situ measurement in atmospheric air of the crack length at several loading and temperature conditions was conducted using a direct current potential drop (DCPD) method. Both annealed and heat-treated Inconel 718 samples were investigated. For a single sample, the normalized change of crack length was confirmed to not be a random process. This finding is significant in highlighting the time history features, which could be used in training machine learning models for the fundamental understanding of oxidation and the prediction of crack initiation and growth.