Times they are a-changin’ in predictive maintenance

Photo by Victor Garcia on Unsplash

Predictive maintenance, despite its measurable results and successes, has been suffering from lack of scalability. Models designed and trained for a given machine with a particular set of sensors often had to be retrained before being used with other similar machines. In addition, the quantity and variety of machines in a factory makes developing, training, and maintaining models for each machine a challenge. Therefore, it‘s natural that predictive maintenance became limited to the most valuable assets in a plant.

But that is quickly changing.

Traditionally, vibration analysis requires a vibration expert, detailed knowledge of the machines, and significant contribution from domain experts. This also contributes to the lack of scalability, especially in a situation in which all machines would have their own sensor.

Machine Learning becomes the solution to that, by allowing large-scale unsupervised analysis of the history of an asset or a population of assets. Additionally, our API makes it easy for asset health management, condition monitoring, and maintenance providers to build their own solution on top of it . Just connect data from vibration sensors or upload their own historic data and let our engine help with:

  • Mode identification: narrowing asset data investigation with the identification of operating or failure modes based on periods of similar behavior over time.
  • Black sheep using population-wide anomaly detection to enable the prioritization of assets that require attention by comparing the behavior of similar machines and finding the ones that behave differently.

And more features are in the pipeline ! We also provide a Python package for those who are not familiar with a REST API or would like to integrate it in their data analysis routine in Jupyter Notebook , for example.

Like the Bob Dylan song that inspired the title of this text says Don’t wait for the walls to start rattling, click to read more about MultiViz Vibration API and request your trial: https://multiviz.vikinganalytics.se/vibration-api

About The Author

Founded in 2017, Viking Analytics enables the digital transformation of industrial companies by bringing together people, data, insights, and value. With our software MultiViz, users can quickly prepare, analyze, and process large sensor data without being a data scientist. We also empower organizations to capture expert knowledge and build scalable and reliable smart manufacturing applications including optimization, predictive maintenance, anomaly detection, and digital AI operators.

AI-based predictive maintenance applications for industrial assets.