Predictive Maintenance with MATLAB
This two-day course focuses on data analytics, signal processing, and machine learning techniques needed for predictive maintenance and condition monitoring workflows. Attendees will learn how to use MATLAB to import data, extract features, and estimate the condition and remaining useful life of equipment.
- Importing and organizing data
- Unsupervised anomaly detection
- Creating supervised fault classification models
- Preprocessing to improve data quality
- Extracting time and frequency domain features
- Estimating Remaining Useful Life (RUL)
- Interactive workflows with apps
Day 1 of 2
Importing Data and Processing Data
Objective: Bring data into MATLAB and organize it for analysis, including handling missing values. Process raw imported data by extracting and manipulating portions of data.
- Store data using MATLAB data types
- Import with datastores
- Process data with missing elements
- Process big data with tall arrays
Finding Natural Patterns in Data
Objective: Use unsupervised learning techniques to group observations based on a set of condition indicators and discover natural patterns in a data set.
- Find natural clusters within data
- Perform dimensionality reduction
- Evaluate and interpret clusters within data
Building Classification Models
Objective: Use supervised learning techniques to perform predictive modelling for classification problems. Evaluate the accuracy of a predictive model.
- Classify with the Classification Learner app
- Train classification models from labeled data
- Validate trained classification models
- Improve performance with hyperparameter optimization
Day 2 of 2
Exploring and Analyzing Signals
Objective: Interactively explore and visualize signal processing features in data.
- Import, visualize, and browse signals to gain insights
- Make measurements on signals
- Compare multiple signals in the time and frequency domains
- Perform interactive spectral analysis
- Extract regions of interest
- Generate MATLAB scripts for automation
Preprocessing Signals to Improve Data Set Quality and Generate Features
Objective: Learn techniques to clean signal sets with operations such as resampling, removing outliers, and filling gaps. Interactively generate and rank features.
- Use resampling to handle nonuniformly sampled signals
- Fill gaps in uniformly sampled signals
- Perform resampling to ensure common time base across signals
- Use the Signal Analyzer app to design and apply filters
- Use File Ensemble Datastore to import data
- Use the Diagnostic Feature Designer app to automatically generate and rank features
- Perform machinery diagnosis using envelope spectrum
- Locate outliers and replace with acceptable samples
- Detect changepoints and perform automatic signal segmentation
Estimating Time to Failure
Objective: Explore data to identify features and train decision models to predict remaining useful life.
- Select condition indicators
- Use lifespan data to estimate remaining useful life using survival models
- Use run-to-threshold data to estimate remaining useful life using degradation models
- Use run-to-failure data to estimate remaining useful life using similarity models