Predictive Maintenance Platform

IoT

Background

The client is a start-up that aims to deliver a predictive maintenance system for the industrial IoT, monitoring plants, maritime shipping, and so on. The client needed an entire platform, including:

  • integration with multiple sensor interfaces and gateways (i.e., Modbus, MQTT);
  • data streams collection;
  • data streams processing and classification;
  • design of classification models;
  • building data models with historical data in batch processing;
  • time series visualization.

The platform should be fully reconfigurable and should support any data model.

Our work

We delivered an entire system and deployed it on the IBM cloud. The system can handle any number of sensors attached to any number of devices, since its throughput scales in a linear way. The system configuration, e.g., changing the installation topology, machine learning models and so on, can be reconfigured online, without any system downtime. The applied algorithms enable anomaly detection and downtime prediction in real-time. Moreover, the algorithms may be tuned with new data.

The outcome

A state-of-the-art predictive maintenance platform (PDM) that can be extended to different sensor interfaces and industries.

Technology used
  • Scala
  • Python
  • Akka
  • Spark
  • Kafka
  • Cassandra
  • InfluxDB
  • Grafana
  • IBM Bluemix
  • Docker