Abstract:Nickel-based single crystal superalloys, as engine blade materials, are prone to fatigue damage due to repeated startups and shutdowns. Therefore, monitoring and quantitatively estimating fatigue cracks are essential for engineering structures to ensure safety. In this study, we proposed a method for fatigue crack segmentation and damage prediction based on deep learning and in-situ high-temperature scanning electron microscopy (SEM). Sequential SEM images describing the crack initiation and propagation under near-service conditions were obtained by conducting in-situ high-temperature fatigue experiments. A fatigue crack dataset with high-quality was thus constructed for further dynamic and real-time crack segmentation and damage assessment. Deep learning-based models were used to segment cracks and predict damage behavior (i.e., crack area, length, width, and stress intensity factors) at future based on prior damage information. The short-term and long-term damage prediction capability were validated by comparing model performance when predicting damage at different future time points. Additionally, we compared the model performance when predicting damage at specific time point based on varying lengths of input sequence. Results demonstrated that the model could segment cracks and scales with different sizes accurately. The model performed well in short-term damage prediction. The long-term predictive performance showed decrease than that of short-term, which could be improved by feeding long length of input sequence. The proposed approach demonstrates the feasibility and effectiveness of deep learning-based crack segmentation and damage prediction, which facilitates the move toward real-time analysis and rapid diagnosis of material damage in the future.

Simplified ResNet approach for data driven prediction of microstructure-fatigue relationship

Prediction of Fatigue Crack Damage Using In-Situ Scanning Electron Microscopy and Machine Learning

A physics‐informed neural network framework based on fatigue indicator parameters for very high cycle fatigue life prediction of an additively manufactured titanium alloy

A defect-based physics-informed machine learning framework for fatigue finite life prediction in additive manufacturing

Fatigue life prediction at mesoscopic scale of samples containing casting defects: A novel energy based non-local model

A new data-driven probabilistic fatigue life prediction framework informed by experiments and multiscale simulation

Fatigue life prediction driven by mesoscopic defect data

Remaining fatigue life prediction of additively manufactured Inconel 718 alloy based on in-situ SEM and deep learning

On the Use of Data-Driven Machine Learning for Probabilistic Fatigue Life Prediction of Metallic Materials Based on Mesoscopic Defect Analysis

A frequency-domain enhanced multi-view network for metal fatigue life prediction

Materials fatigue prediction using graph neural networks on microstructure representations

A frequency domain enhanced multi-view neural network approach to multiaxial fatigue life prediction for various metal materials

A data-physics integrated approach to life prediction in very high cycle fatigue regime

Multi-physics information-integrated neural network for fatigue life prediction of additively manufactured Hastelloy X superalloy

Deep Learning-Based Fatigue Strength Prediction for Ferrous Alloy

Deep learning prediction of stress fields in additively manufactured metals with intricate defect networks

Image-driven prediction of fatigue crack growth in metal materials via spatiotemporal neural network

An artificial neural network modeling approach for short and long fatigue crack propagation

Defect driven physics-informed neural network framework for fatigue life prediction of additively manufactured materials

A Hybrid Mechanism and Data‐Driven Approach for Predicting Fatigue Life of MEMS Devices by Physics‐Informed Neural Networks

On the use of transfer modeling to design new steels with excellent rotating bending fatigue resistance even in the case of very small calibration datasets