How Digitalization and Machine Learning Can Advance the Vibration Analysis

Paulo Cipriano, Global CBM Services CoE Manager, SKF Australia Pty


This presentation provides a new approach in the field of vibration analysis that aims to improve efficiency and accuracy of data analysis to produce diagnostics and prognostics by introducing digitalisation and machine learning technologies. The principal goal on the utilisation of digitalisation and machine learning technologies associated with vibration data is to automate the identification of incipient failures and recommendation of maintenance actions to improve the usage and extend the life of the asset by predicting its future condition. In last past 20 years, several attempts have been made on trying to use alarms to automate the process of vibration analysis and diagnostics with the objective of improving efficiency, cost and performance but none has been successful. So, till today most vibration analysis data are manually reviewed. This manual evaluation of vibration data is inconsistent, time-consuming and an expensive task. The major part of the study was done by quantifying the performance of the proposed approach on the generation of “false positive alarm” (machine is good, but alarm is raised), “false negative alarm” (machine is defect, but no alarm is raised), “true positive alarm” (machine is defect and alarm is raised) and “true negative alarm” (machine is good and no alarm is raised). The proposed method was applied to real vibration and repair data from several asset types. One of the most important result achieved in this study was a reduction between 70 to 80% of time spent on vibration data analysis. New technologies such as digitalisation and machine learning have the potential to automate the complex process of vibration analysis entirely and at the same time use this data in correlation with others to predict the remaining life or reliability of assets in real-time and consequently cause a significant impact on the asset management program.
– New approach to vibration data analysis
– Introduction to machine learning
– Potential benefits of digitalization and machine learning


Joined SKF in 2000. Was responsible for the creation and set up of the first SKF’s Remote Diagnostic Center in 2002. Worked in different positions related to condition-based maintenance (CBM) services and integrated service solutions in South America, Asia and SKF Global. Over 20 years of experience in CBM services. Responsible for SKF’s Global CBM Service Standards. These standards outline the know-how for local capability set-up and delivery of CBM Services. The standardization helps SKF to maximize efficiency, safety, repeatability and quality while enabling SKF Digitalization program.