Data Size Increment for Fault Detection on Rotating Machinery

In recent years, the use of data-driven modeling techniques for the diagnosis of failures in rotating machinery has increased. These techniques require large amounts of data that cannot always be obtained because they generate high costs and excessive time, which are difficult to solve from the economic and technical point of view. The present work focuses on the pre-processing of vibration signals and proposes a method to increase the number of informative time series of a rotating machine without increasing the time and costs in the signal acquisition stage. As a result, an increase from 315 signals in the data acquisition phase to 429000 after the application of the method has been obtained; an adequate amount for the construction of data-based models, including deep learning for the detection of faults in rotating machinery.