The ASNT Annual Conference, held in Nashville, TN, from October 31 – November 3, 2022, is where the NDT world comes together! FDH team members Joshua Scott, NDT Research Engineer III; Dave Edeal, Product Manager; and Nadia Sa’d Mulaire, Vice President of Business Development & Partnerships, will be exhibiting at booth 1215.
In addition, Joshua Scott was invited to present his research on the topic, Machine Learning-Based Approach to Measuring Bolt Tension Without Reference Measurements, on Thursday, November 3rd, from 9:15-10 am.
The presentation will examine an ultrasonic bi-wave approach to flange bolt tension measurement using signal processing and machine learning techniques via an easy-to-use handheld tool. Current testing methods require heavy equipment and knowledge of unstressed bolt lengths. Considering the large number of bolts installed on critical structures like wind turbines, existing processes are inefficient, and results can be widely inaccurate.
In the proposed approach, signal processing techniques indicate whether collected data is of sufficient quality, eliminating field technicians’ need for manual analysis. Then a three-stage machine learning approach identifies the bolt coating, whether the data should be considered an “outlier,” and uses a regression model to predict the bolt tension.
Machine learning performance results indicate that, for grade 10.9 M42 bolts, the approach has a 95% confidence error interval of 7.6%, where the percent error is relative to bolt yield stress. Verification of the results was performed on previously untested bolts of the same material properties and total lengths between 260mm and 425mm. The average predictive error was approximately 2.9%.
For more information on FDH’s innovative solutions for critical infrastructure or to inquire about participating in field testing, visit FDH’s products page.
Learn more about the event here.