Predictive Maintenance with LSTM Siamese Network

Detect Failures from Sensor Time Series

Marco Cerliani
Towards Data Science
6 min readNov 13, 2019
Photo by sol on Unsplash

Predictive Maintenance nowadays is an active field where Artificial Intelligence solutions are introduced to improve the productivity of every kind of manufacturing system. Typical requests involve developing a solution that produces warnings when particular parts of observed components are closed to failure. As can be guessed, these are classic ‘situations of unbalance’ where we have tons of data coming from every kind of sensor but a lack of positive labeled samples, where a positive sample is the presence of a failure.

In a previous post, I’ve introduced a solution that classifies the degradation status of a particular component in a hydraulic pipeline system; the results were cool! Now, I try to develop a pipeline that detects if a storage system is close to failure. We want to classify continuous time series coming from sensors making use of Neural Networks. In particular, I avoid the problem of lack in failure samples using Siamese Networks adapted to fit time series with an LSTM architecture.

THE DATASET

As always, I prefer to work with real data. I found a fantastic dataset on Kaggle: High Storage System Data for Energy Optimization, released by Smartfactory. The dataset stores the signals of 4 short…

Create an account to read the full story.

The author made this story available to Medium members only.
If you’re new to Medium, create a new account to read this story on us.

Or, continue in mobile web

Already have an account? Sign in

Towards Data Science
Towards Data Science

Published in Towards Data Science

Your home for data science and AI. The world’s leading publication for data science, data analytics, data engineering, machine learning, and artificial intelligence professionals.

Marco Cerliani
Marco Cerliani

Written by Marco Cerliani

Statistician Hacker & Data Scientist