Unsupervised Learning to improve MFC accuracy

While the MFC nominal accuracy can be as low as 0.001”, data processing techniques currently employed by legacy software are simply unable to effectively correct for the effects of tool eccentricy at every depth sample. Incorrect centralization results can represent as much as 30 or 40% variation in reported penetration figures, with the end user being forced to accept a poor compromise between accuracy and practicality. In this webinar we discuss how unsupervised learning, advanced heuristics and anomaly detection can dramatically improve MFC survey accuracy and reliability.
John Guntoro is a Machine Learning Engineer with ANSA Data Analytics. He started working as a geophysicist in Indonesia upon graduating from the University of Southampton in 2014. Starting in 2021 he joined the University of Aberdeen as a research assistant before joining ANSA in 2023. He is now focused on delivering automated solutions for Multi-Finger Caliper processing at ANSA, using various optimisation and Machine Learning algorithms.