Time Series Anomaly Detection Techniques for Predictive Maintenance

 Fault data is critical when designing predictive maintenance algorithms but is often difficult to obtain and organize. Many organizations are faced with a growing sea of time series sensor data, most of which represents normal operation. How can engineers analyze this data and design anomaly detection algorithms to identify potential problems in industrial equipment? Using real-world examples, this webinar will introduce you to a variety of statistical and AI-based anomaly detection techniques for time series data. 

Learn about:

Organizing, analyzing, and preprocessing time series sensor data

Feature engineering using Diagnostic Feature Designer

Distance-based approaches for exploring anomalies in historical data

One-class machine learning and deep learning approaches for algorithm development

Comparing and testing algorithm performance

Deploying anomaly detection algorithms in a streaming environment 

Predictive Maintenance Toolbox Examples: https://bit.ly/41g4aKi

About the Presenter:

James Wiken is a Senior Application Engineer at MathWorks, where he helps people with all things MATLAB, with a particular emphasis on Test & Measurement, Application Development, and Software Development Workflows. James also holds an S.B. and S.M. degree in Aerospace Engineering from MIT, where he specialized in controls and autonomous flight. 

Chapters:

00:00 Introduction to Anomaly Detection 

01:03 Predictive Maintenance Basics

03:12 Types of Time Series Anomalies

04:20 Time Series Anomaly Detection Techniques

06:39 Data Exploration using Distance-Based Pattern Matching in MATLAB

13:37 AI Algorithm Development Workflow

15:03 Developing Anomaly Detection Algorithms in MATLAB

17:15 Feature Engineering with the Diagnostic Feature Designer

24:29 Training AI Models for Anomaly Detection

25:27 AI Models for Anomaly Detection: One-Class SVM 

27:55 AI Models for Anomaly Detection: Isolation Forest

28:47 AI Models for Anomaly Detection: LSTM Autoencoder

34:44 Deploying Anomaly Detection Models

35:45 Further Resources