Country : United States
Client : Aiden Livingston, ThermoAI
This is a classification machine learning model. We recorded vibrations with an accelerometer at two locations as well as a step-down ameter and two microphones recording the sound of the pump.
We ran 6 different types of pumps (size and capacity) and recorded all the data. The pumps were run normally and with 18 different types of anomalies such as gavitation in the pump, outlet valve partially or fully closed, bearing problems, etc.
We then trained a machine learning model with this data to be able to classify the operation of the pumps in realtime.
The trained model was able to infer the pumps and surrounding equipment condition with a 95% accuracy.
We were able to also combine the full training for all types of motors into one model, reducing the cost and man hours required for deployment & future updates at the edge.
“Richard is a total pro! His communication is top notch and he really goes above and beyond to exceed all expectations. We hired him just for ML modeling and he decided a dashboard would be the best way to showcase the results and mocked one up on his own time. And he was totally right, it was way better! Highly recommend!”
★★★★★ 5/5