BearingAnamolyDetector

Fan bearing anomaly detector.

The Problem

Many fans, pumps and motors in apartments and industrial centers run until complete failure after which they have to be thrown away. Not many people notice the early warning signs like imbalance or loosening until failure which also causes downtime and wasted energy. People in the industry solve this by using handheld analyzers but they’re expensive and need training for the analysts. Attaching sensors to all machines is also too costly for small industries with many small machines.

My Solution

A pen-shaped instrument that anyone can press against a running machine to test its health in a few seconds. I will use an MPU-6050 accelerometer that will be at the tip of the instrument, with a magnet for a steady contact, connected to the nRF54LM20 DK. The results of that will be shown on a phone over Bluetooth LE.

The instrument will have two modes -

• Learn/Check Mode – The user (while installing the new machine) presses the instrument against the new running machine to save its vibration fingerprint under a name. Later, the user can compare it against the on-device anomaly detection model, so as to catch the changes early.
• Diagnose Mode – For machines without any saved baseline, the instrument will estimate the RPM from the vibration spectrum and use order analysis (1x, 2x, harmonics) to flag like the cause for example unbalance, looseness etc and also a severity rating based on ISO 20816 vibration zones.

How I will use the nRF54LM20 DK -

I will run the anomaly on-device model, trained with the Nordic Edge AI lab (Neuron model), and will explore the faults classification using the Axon NPU.

• I will run the anomaly on-device model, trained with the Nordic Edge AI lab (Neuron model), and will explore the faults classification using the Axon NPU.

The Bluetooth LE will be used for sending alerts to a phone and also for managing the saved machine fingerprints. over a simple browser Bluetooth page.

• The Bluetooth LE will be used for sending alerts to a phone and also for managing the saved machine fingerprints. over a simple browser Bluetooth page.

I will use the low power operation so that the instrument can run on battery and can sleep between measurements.

• I will use the low power operation so that the instrument can run on battery and can sleep between measurements.