Werfen: predictive maintenance for freezers

01 - 10 - 2021
We are very glad to announce that our IT team has completed another Data Science project. We are currently collaborating with Werfen, a well-known world leader in specialized diagnostics.

We spoke with Werfen Original Equipment Manufacturing Technology Center, and they had a very specific need: prevent biological samples from spoiling. To store these samples, they use ultra-low freezers (ULT). These types of freezers preserve the biological samples at temperatures between -40⁰C and -86⁰C.

Werfen’s challenge to PROCON SYSTEMS: predict when these freezers will malfunction and prevent the biological sample from spoiling. They provided us data from the freezer environment that could affect its operation, such as freezer temperature, ambient temperature, and the freezer brand and model.

Using a recurrent neural network to create an autoencoder we used an unsupervised model for anomaly detection to improve the current alarm system that they have.

Yet another example of how to apply the latest Machine Learning techniques in the Industry 4.0 sector!