How can I reduce false positives in vibration anomaly detection on an AVR128DA48?
I'm still on the project tho where I've installed a vibration sensor to monitor industrial machinery using an AVR128DA48 microcontroller. While my basic anomaly detection code is functioning,
it tends to produce too many false positives. Given the limitations of the microcontroller, I'm looking for a more refined algorithm to improve the accuracy of detecting anomalies in the vibration data. Any suggestions on how to enhance my approach?
5 Replies
Look into signal processing and reducing noise with it.
hey guys I was able to get a lead on this, I implemented a sliding window averaging technique, which smoothed out the vibration data over a short period, reducing the impact of sudden spikes that were causing false positives.
By averaging the readings before comparing them to the threshold, I was able to improve detection accuracy
I also introduced a derivative-based approach to monitor the rate of change in vibration, which helped me detect sharp increases rather than just focusing on absolute values.
Also, I adjusted the threshold based on normal operating conditions, ensuring the system was less sensitive to minor fluctuations and better at detecting true anomalies😤😤