New Analytics Shows New Ways To Make Machines Run Better

Broad Frequency Spectrum analysis might hold the key to getting ahead with predictive maintenance.

For years data scientists have been grappling with the problem of predicting machine failures based on conventional process data, and have made significant advances: Today, vital insights can be told from increased oil consumption, slowing machine speeds, sluggish motor acceleration data. But the quest is open for more insights with more sensors and more data!

New methods based on frequency spectrum analysis give surprisingly accurate information about the inner workings of your machines. The technique is similar to what a human engineer might do: listen to the sounds of the mechanism to spot any unusual patterns. Except, when done electronically, a sensor can improve upon the human ear by listening to a much broader frequency spectrum, be placed inside the gut of the machine where an offending frequency would be much more pronounced, and be there 24/7 to spot even the smallest of changes to the normal operating environment.

Frequency Spectrum Analysis and Degradation over time in a category of rotational machines

Our analysis has shown that many mechanical issues are detectable through frequency spectrum analysis. We custom-designed a sensor to provide once-per-minute spectrum information for IoT-based analysis. The acoustic sensor can predict degrading mechanical performance in time to prevent unplanned downtime.

Fast Fourier Transforms (FFTs) is an example of a class of algorithm that can decompose a complex spectrum of frequencies into constituent parts. By continuously recording a band of frequencies that range far beyond what the ear can hear – starting with low-frequency vibrations through the audible spectrum to high-frequency noise, our FFT-based sensors are combined with machine learning techniques that adapt to usual operating noise and immediately lock onto any unusual patterns that can be indicative of machine problems.

This technique is particularly applicable to rotating machinery and has proven its value in real-world applications on applications on bearings, motors and drives, and pumps.

We call our system of sensors and analytics “Symphony”; because we are sure while it’s just noise to most people, to you the sound of your machines is pure music.