<p>Total knee arthroplasty (THA) is currently one of the most common surgeries in the United States, with more than 600,000 operations performed each year. As life expectancy increases and the American population continues to age, TKA is likely to increase in popularity. While the surgery is generally highly successful – enjoying a 10-year revision rate of only 5% – revision due to pain and failure is costly and can have adverse impacts on patient outcomes. The possibility exists to reduce rates of catastrophic failure by early detection of damage in total knee replacements (TKR). To this end, previous work has been done to establish the ability of a structural health monitoring (SHM) technique known as the electromechanical impedance (EMI) method to detect certain types of damage prevalent in TKRs. This method utilizes the electromechanical coupling inherent in piezoelectric materials to identify and classify damage done to a structure using changes in its mechanical impedance. In the aforementioned work, 19 simulated TKRs were constructed and artificially damaged, impedance spectrum measurements were taken in healthy and various damaged states, and the data was compared via RMSD and p-value statistics to determine if significant differences between the healthy and damaged impedance responses exist. The work presented here expands upon the previous work by exploring the use of classification machine learning (ML) algorithms for translating any differences in impedance responses into discrete damage classes. On top of their predictive power, the applications of ML algorithms are innumerable, ranging from neural networks for facial recognition to spam filters to cancer detection. Recent work has been done to make use of various ML algorithms, such as Support Vector Machines (SVM), Convolutional Neural Networks (CNN), as well as unsupervised learning techniques, to process EMI-based SHM data to predict existence, location, and severity of damage in generic aluminum beams, vehicle brake discs, and concrete structures, among others. The goal of this work is to determine the ideal classification technique(s) for identifying the presence, type, and severity of damage within the simulated TKR systems assembled and tested previously. To this end, several classification algorithms will each be trained on the previously recorded impedance data in a comparable manner, and the results of a k-fold cross-validation scheme will be compared for accuracy, among other common ML performance metrics. Algorithms to be tested include various forms of SVMs, k-nearest neighbor, decision trees, etc., as well as certain popular Ensemble classification techniques. A 1-dimensional CNN will also be trained and tested using the same cross-validation technique, and compared to the more classical ML methods with the same metrics.</p>

Comparison of Classification Machine Learning Algorithms for Damage Detection in Simulated Total Knee Replacements

Paper Proceedings: https://asmedigitalcollection.asme.org/SMASIS/SMASIS2021/volume/85499

ASME 2021 Smart Materials, Adaptive Structures and Intelligent Systems (SMASIS2021)

Supporting the need among researchers worldwide for additional online and multimedia content, ASME now offers the ASME Conference Video Collection. This platform will collect and provide long-term access to conference presentation videos, encompassing all technology areas within ASME's technical conference and events program. About 30,000-40,000 engineers across all mechanical engineering and related disciplines participate in ASME conferences each year. This includes authors of approximately 6,000 to 8,000 technical papers published by ASME, the majority of whom also submit video presentations for our virtual conferences. The video collection will be available as an add-on to existing product packages or as a separate subscription.