Software-Based Analysis of Microcracking in Turbine Blades 

Fatigue accounts for 80% or more of all mechanical service failures.  

High-cycle fatigue testing (HCF) is the most prevalent type of fatigue testing for failure analysis of turbine blades currently available. The testing material gets exposed to cyclic loads until the material fails or meets a set number of cycles. 

HCF testing helps determine and validate if turbine blades will meet design criteria and endurance limits to prevent turbine engine failure.  

ITS performs HCF testing on an electro-dynamic shaker, digital vibration controllers, sensors, and heating systems, all tuned to a particular frequency. The blade runs at the specific stress level until it reaches a five percent shift in frequency, which is ITS’ definition of failure, or 100 million cycles.  

During a recent post-testing analysis, ITS found a blade that never showed a 5% frequency shift, yet it exhibited microcracking.  

This finding prompted ITS to review test cycle data to discover evidence of microcracking and better understand microcracks that don’t trigger failures or halt the testing process. The goal was to create a postmortem testing tool that could flag parts for failure earlier in the process due to these microcracks that might otherwise go unnoticed.  

Further refinement of the current filtering method and comparison tool is still ongoing. The tool was initially developed to explore micro-cracks that did not trigger a failure or stop the vibration test. The idea was to find a postmortem tool to locate the initiation of these cracks in the hopes to be able to trigger or flag a part for failure earlier due to these micro cracks. 

A software-based algorithm that uses Python and MATLAB libraries to analyze vibration data was needed, which ITS custom created.  

Early use of the tool was successful, and analysis located important events based on frequency drops, events in voltage, and acceleration that could potentially indicate the initiation of circumstances that would lead to microcracks.  

However, the algorithm results did not quite align with the engineers’ assessments of the test data. Noise filtration and optimized methods used to search for events led to further iterations of the tool, and further refinement remains ongoing. 

From early versions that parsed the entire data set to more refined versions that use a stepwise sequence, the tool shows promise. Despite its potential for engine turbine blade testing, the algorithm has not yet been implemented as an evaluation method.  

The process of fine-tuning, filtering, and producing repeatable results continues, with possible additional testing methods such as infrared imaging or using eddy currents under consideration.