A Novel High-Throughput Fatigue Testing Method for Accelerated Material Characterization
The prediction of fatigue failures remains a challenging endeavor despite the critical importance of mechanical reliability in our modern economy. The advances in experimental and computational modeling techniques have continued to push mechanistic understanding and predictive capabilities forward. However, the ability to generate fatigue data has not experienced these gains. Traditional fatigue testing methods are not only slow and costly but also plagued by high variability, making it difficult to calibrate predictive models for real-world utilization. The limited progression toward creating large datasets constrains the application of modern statistics and computational techniques, including machine learning and artificial intelligence, for predicting low-probability fatigue failures. To bridge this gap, this project aimed to develop a testing methodology that conforms to popular fatigue testing standards, maintains existing test costs, and improves throughput by orders of magnitude. This dissertation explores various concepts to improve the throughput of uniaxial fatigue testing and presents a thorough mechanical analysis of the most promising approach. This approach, the stacked cylinder concept, involves a stack of compliant cylinders, each surrounded by a ring-shaped test specimen. The stacked cylinders are then compressed by an actuator, causing each ring-shaped specimen to counteract the radial expansion. These interactions simultaneously induce a nearly uniaxial stress state in rings throughout the stack. A prototype was developed and demonstrated with 39 aluminum 6061-T6511 test specimens subjected to 2 million loading cycles. The performance of the prototype was assessed against the popular force-controlled fatigue testing standards via numerous strain gauge measurements throughout the test, comparing the failure distribution to a traditional Metallic Materials Properties and Development Standardization (MMPDS) fatigue dataset, and a fractographic analysis using a scanning electron microscope. The fatigue data produced from the prototyped high-throughput design aligns well with the traditionally collected MMPDS equivalent life dataset, but failed to meet requirements for both fatigue testing standards. After identifying the shortcomings, pathways were discussed to improve future performance and obtain efficient, standardized, and economic fatigue lifetime data. To close, potential extensions and applications to the method were presented.