The Predictive Maintenance Toolbox™ provides capabilities and reference examples for designing and testing condition monitoring and predictive maintenance algorithms for ball bearings, pumps, batteries, and other machines.

Use the Diagnostic Feature Designer to extract features from sensor data without writing any MATLAB^® code. Filter and preprocess sensor data signals and extract time domain features such as mean and standard deviation. You can also estimate a signal’s power and order spectra and extract frequency domain features such as spectral peak values. After you have computed your features, you can plot and rank them to determine which features are best suited for your fault classification and remaining useful life algorithms, and export them.

You can estimate the time to failure of your machine or its remaining useful life using similarity methods which require run-to-failure data, survival methods—which require lifetime data related to events such as part replacement and part failure—and trend-based methods, which require a known failure threshold.

As you can see, the methods also provide confidence intervals for the predictions made. 

Every algorithm needs data, and you can import yours from the cloud, HDFS, and local files before organizing it in MATLAB. If you don’t have any failure data, you can generate simulation data from Simulink^® models of your machine that incorporate fault conditions.

The documentation and examples help you get started by stepping you through the workflow of the algorithm development process.

For more information on the Predictive Maintenance Toolbox, please return to the product page.