— August 19th, 2021
Within an industrial setting, being able to determine if and/or when a machine malfunctions is vital to maintaining safety and uptime. This challenge is what prompted a maker who goes by javagoza on element14 to enter into their Design for a Cause 2021 contest with his device, which he calls the VenTTracker.
At its heart, the VenTTracker uses an Arduino Nano 33 IoT mounted onto a small protoboard that is attached to a sliding surface, such as a window or vent. Under normal operation, the device does nothing, but once an anomaly is detected, including an obstacle or breakdown, the onboard OLED screen shows an alert message.
Because this project uses machine learning to differentiate between normal operation and an anomaly, javagoza collected a large dataset of motions from an accelerometer and then uploaded it to Edge Impulse’s Studio. From there, he added a time series processing block and flattening block to generate the features that fed into the Keras neural network for training and validation. Once deployed back to the Arduino, the model performed very well at telling the difference between the window opening normally and something being in the way.
He even included Arduino Cloud functionality to display if the window is open and any anomalies that have been detected so far. There was an additional module constructed for environmental monitoring, which consists of a Nano 33 IoT and a BME680 sensor that sends CO2, temperature, and humidity data to another Cloud dashboard to let users know when to open the window.
To read about the VenTTracker in more detail and see its code, you can visit javagoza’s write-up on element14.
Website: LINK
