Investigation of stress corrosion cracking in CMSX-4 turbine blade alloys using Deep Learning assisted X-ray microscopy and using 3D images for finite element modelling

Session

Microscopy to Modelling

Authors

Mr Andy Holwell (1), Dr Hrishikesh Bale (2), Mr Mustafa Elsherkisi (3), Dr Maadhav Kothari (1)

Affiliations

1. Carl Zeiss Microscopy Ltd
2. Carl Zeiss X-ray Microscopy LLC
3. Cranfield University

Keywords

Superalloy, CMSX-4, XRM, deep learning, hot stress corrosion cracking, C ring, FEA, FEM, finite element modelling

Abstract text

Single crystal nickel superalloys are typically are used in power generation and aviation applications due to their unique properties. Recently, incidents of failure due to increased temperature has caused Type II hot corrosion leading to cracking in blade roots resulting in catastrophic failure. Understanding the failure mechanism and crack characterization is vital in solving this issue.

 

After exposing a salted C-ring specimen to 500°C air for 92 hours, we demonstrate a novel high resolution X-ray microscopy (XRM) workflow using deep-learning based algorithms for data reconstruction and segmentation, combined with scanning electron microscopy in order to study cracks, crack tips and crack arrest points developed during stress corrosion cracking.

 

By extracting the fracture tip, both crystal plasticity and crystal deformity can be studied in detail resulting in orientation tomography of the corroded region. Using this correlative workflow we are able to identify structural defects and fracture mechanisms not visible using typical microscopy techniques.

 

Furthermore, deep learning reconstruction datasets have enabled the integration of XRM with finite element models (FEM) to enable mapping of real-life cracks that are translated to realistic model meshes. Computational modelling can complement experimental efforts by providing estimations of attributes (e.g., stress) concurrent with the material characterisation offered by deep-learning enhanced XRM We will demonstrate that the approach has been instrumental to uni-vocally discover the damage mechanism involved in stress corrosion cracking. We will further present an autonomous integration approach between XRM and FEM that can be implemented concurrently with the material characterisation. The seamless XRM-FEM exchange has the potential to control experiments based on magnitudes that are not measured but modelled.