Evening Workshop Wednesday, February 27, 2008: Day Three

5:30pm - 8:30pm Conducting Statistical Analysis for Proactive Maintenance

Dinner will be served

The challenge in feeding back information from the field to the enterprise is to separate real issues with device operation from the “normal” variation that devices experience from minute-to-minute, day-to-day, or season-to-season. The cost of “getting it wrong” equals the expense of spinning up the service organization to troubleshoot problems that don’t really exist plus the expense of mistakenly letting a customer’s device run long after it should have been serviced (to the point where it needs overhauling or wholesale replacement). In this course we examine statistical methods that can be used in M2M software solutions to reduce this cost.

Participants will learn how to:

• Understand univariate statistical methods to detect shifts and drifts in readings from sensors
• Understand multivariate statistical methods to detect shifts and drifts
• Understand the challenge of correlative effects like autoregression, autocorrelation, and seasonality

How you will benefit from this workshop:

• Reduce the rate of false alarms
• Increase accuracy and responsiveness in detecting real issues
• Gain a deeper understanding of the issues in designing algorithms to detect real problems that need real intervention as opposed to expected variation in the asset’s environment

Workshop Leader:

Matthew Lehr
Senior Consultant
EMRT

[ Register Now] · [ Next: Post-Conference Masterclass Thursday, February 28, 2008: Day Five ]