Page 44 - MSDN Magazine, May 2019

MACHINE LEARNING
Using Survival Analysis for
Predictive Maintenance
Zvi Topol
Some years ago, I introduced the basics of survival analysis and described how to implement a non-parametric algorithm called Kaplan-Meier in C# (msdn.com/magazine/dn630650). Now, I’m going to take another look at survival analysis, in particular at two more advanced methodologies that are readily available on two popular machine learning platforms, Spark Machine Learning Library (MLLib) and h2o.ai, which are both supported by Azure HDInsight. I’ll use a predictive maintenance use case as the ongoing example.
Predictive Maintenance for the
Industrial Internet of Things
The main idea behind the Industrial Internet of Things (IIoT) is to connect computers, devices, sensors, and industrial equipment and applications within an organization and to continually collect data, such as system errors and machine telemetry, from all of
these with the aim of analyzing and acting on this data in order to optimize operational efficiencies.
The goal of predictive maintenance is to accurately predict when a machine or any of its components will fail. If you can do this, you can perform maintenance just before such failure is predicted to occur. This is more efficient than not performing any maintenance until a failure occurs, in which case the machine or component will be unavailable until the failure is fixed, if indeed it’s reparable. Such unplanned downtime is likely to be very costly.
The goal of predictive maintenance is to accurately predict when a machine or any of its components will fail.
Predictive maintenance is also more effective than performing preventive maintenance at frequent intervals, which could also be costlier because unnecessary maintenance may be applied.
The example and the data I’ll use are an adapted version of the example at bit.ly/2J4WnbN. The example includes 100 manufacturing machines, with no interdependencies among the machines. Each machine is one of four possible models.
This article discusses:
• Predictive maintenance
• Industrial Internet of Things
• Survival analysis
• Cox proportional hazards regression
• The Weibull Accelerated Failure Time Regression model Technologies discussed:
R, Spark MLLib, h2o.ai, Azure HDInsight
38 msdn magazine