Predictive Maintenance Modelling Guide in the Cortana Intelligence Gallery
This post is authored by Fidan Boylu Uz, Senior Data Scientist at Microsoft.
Predictive Maintenance is among the most sought-after advanced analytics applications, and is increasingly viewed as a life-saver in asset-heavy industries such as manufacturing and aerospace because of its potential to deliver significant cost reductions through the avoidance of downtime and delays caused by mechanical problems. Many businesses are interested to predict such problems in advance and proactively take actions to prevent issues before they occur.
Microsoft has been an active contributor in this domain, having published several resources to help businesses understand the data science process behind predictive maintenance and build end-to-end solutions. We are now pleased to announce the Predictive Maintenance Modelling Guide, available as a collection in Cortana Intelligence Gallery.
What’s the Guide About?
The guide provides the steps needed to implement a predictive model for a scenario that’s based on a synthesis of multiple real-world business problems, bringing together the common elements observed among many Predictive Maintenance use cases. The specific business case in the guide is about predicting problems caused by component failures such that the question “What is the probability that a machine will fail in the near future due to a failure of a certain component” can be answered. The problem is formatted as a multi-class classification problem and a machine learning algorithm is used to create the predictive model that learns from simulated historical data that includes telemetry, error logs, maintenance records, machine properties and failure records.
The Predictive Maintenance Modelling Guide collection includes the relevant data sets, an R notebook and experiment:
Predictive Maintenance Modelling Guide Data Sets: The experiment that contains the data sets used in the collection.
Predictive Maintenance Modelling Guide R Notebook: The R notebook that explains the steps of implementing the solution.
Predictive Maintenance Modelling Guide Experiment: The experiment that demonstrates the training and evaluation of the predictive model.
How Can I Use this Guide?
This guide is intended to provide a walk-through of the data science process for Predictive Maintenance. It explains feature engineering, label creation, and the training and evaluation methods commonly used to model Predictive Maintenance problems. It will help you understand the underlying data science process in more detail and modify it to suit your needs.
How is the Guide Related to Other Resources?
The newly announced Predictive Maintenance Modelling Guide complements existing resources by looking at a larger scenario with more data sources. It focuses on the data science of Predictive Maintenance, following the ideas from the Playbook for Predictive Maintenance to demonstrate and explain how these methods can be applied to a common business scenario. You can also refer to the Predictive Maintenance Template for a starter failure prediction experiment in Azure ML and the Predictive Maintenance for Aerospace Solution Template for an end-to-end data pipeline.
Where Do I Start?
Get started with the Predictive Maintenance Modelling Guide and make yourself familiar with the data sets, R notebook and the experiment. You can then create your own models following the steps in the guide. And do share your feedback and comments as you work your way through the guide, we would love to hear from you.