Condition Monitoring, Anomaly Detection, and Predictive Maintenance

[MERSI]

Hello everyone,

I'm currently working on a project involving condition monitoring, anomaly detection, and predictive maintenance using sensors that only provide a voltage measurement (e.g., a speed sensor which outputs a proportional voltage). I wanted to ask if anyone has any experience or has worked on similar projects involving such sensors, particularly when it comes to using them in these areas of predictive maintenance and anomaly detection.

If you have any advice, examples, or ideas on how to handle this type of sensor data effectively, I would be very grateful. I'm especially interested in learning about any techniques or algorithms that can be used to detect anomalies or predict failures based on voltage signals alone.

Any tips or insights on the topic would be highly appreciated!

Thank you in advance for your help!

Best regards

[Julian E. · Accepted solution]

Hello @MERSI ,

For anomaly detection / predictive maintenance, I would suggest you take a look at NanoEdge AI Studio.

It is an auto machine learning tool that will train thousands of models (and apply preprocessing to your data) to hopefully find something that works for your use case.

The idea behind NanoEdge is that it uses machine learning model (quite small compared to neural networks) so it is more adapted to sensing data (current, vibration etc). I would not advise to do image classification with it.

With NanoEdge you only need datasets, and you get the C code libraries with the preprocessing. And one of the available kinds of project is Anomaly detection that can even be retrained directly on microcontroller.

The main advantage of retraining onboard is to be able to for example train a model for a kind of motor and deploy the same model on every motor on a factory, but you can retrain it specifically on each motor to be more adapted to the final environment (giving a better accuracy).

You could also try tools with ST working with the ST Edge AI Core. 

The ST Edge AI Core is the heart of many tools that uses Neural Networks, it converts them to C code for STM32, MPUs, MEMs, NPUs etc. 

The tools that use the ST Edge AI Core are CubeAI, a GUI in CubeMX that can will generate your C code Neural Network and template of STM32CubeIDE application

The ST Developer Cloud, where you can upload a Neural Network directly to make quantization, optimization and benchmarks. You can then get the C code neural network or a CubeMX or CubeIDE project.

You also have the ST Model Zoo, where you can find scripts and models for different uses cases, but Anomaly detection in not one of them.

My point here is that you could also look in the literatures for neural network that are used for anomaly detection using current and convert them for microcontrollers. 

With NanoEdge, it is easier and more straightforward. You collect the data and get the model.

With Neural Networks and our tools, you need a trained model, and you get the C code to make it run on a microcontroller.

I suggest you look at this tutorial for anomaly detection on a motor using vibration. It is not current, but the idea is the same. You will face the same pain points are the one cited on this tutorial and in terms of usage, using current or vibration is the same thing for NanoEdge.

AI:How to Build an Anomaly Detection Project for Predictive Maintenance with NanoEdge AI Studio - stm32mcu

Have a good day,

Julian