Vis Vibro –Diagnosing machine failures

Created: Applied AI Course - June 25, 2017

Vibration Analysis is one of the most critical and widely used methods in predictive condition monitoring of mechanical devices. Most of this analysis is performed using industry grade accelerometers as transducers along with custom-built hardware and software to process the accelerometer data to predict the extent of damage in machinery. These vibration analyzers are costly (typically $30,000) and do not perform very well very the motors are rotating at slow speeds (under 300 RPM).

We are leveraging ubiquitous smartphone cameras to measure mechanical vibrations and perform predictive condition monitoring resulting in a massive (1/10th) cost-reduction. We leverage advances in computer-vision, rapidly improving smartphone cameras and Adreno GPUs to solve this problem very economically. Our solution works very well at low RPM speeds where current accelerometer solutions do not fare well. Currently, early prototypes are been tested in the field in various factories and plants in India. We hope to test our solution extensively and roll out to our early customers by end of 2018.

Reference: Acceleration Velocity and Displacement time-series (bi-axial) along with their spectrum from one of our deployments. Notice the clear peaks at low RPM/CPM values.


DeepTrader –
Stock Trading Using DeepLearning

Applied AI Course - June 25, 2017
READ MORE

KHOJ-
Simplifying Product Research

Applied AI Course - June 25, 2017
READ MORE