Predictive Maintenance Toolbox

Design and test condition monitoring and predictive maintenance algorithms

Predictive Maintenance Toolbox provides functions and apps for designing condition monitoring and predictive maintenance algorithms for motors, gearboxes, bearings, batteries, and other applications. The toolbox lets you design condition indicators, detect faults and anomalies, and estimate remaining useful life (RUL).

With the Diagnostic Feature Designer app, you can interactively extract time, frequency, time-frequency, and physics-based features. You can rank and export the features to develop application-specific algorithms for fault and anomaly detection. To estimate RUL, you can use survival, similarity, and trend-based models.

The toolbox helps you organize and analyze sensor data imported from local files, cloud storage, and distributed file systems. You can generate simulated failure data from Simulink and Simscape models.

To operationalize your algorithms, you can generate C/C++ code for edge deployment or create production applications for cloud deployment. The toolbox includes application-specific reference examples that you can reuse for developing and deploying custom predictive maintenance algorithms.

Screenshot of the Time Series Anomaly Detection dashboard.

Anomaly and Fault Detection

Train statistical, machine learning, and deep learning algorithms to detect anomalies and faults in time series data. Track changes in your system, detect anomalies, and identify faults.

Time Series Anomaly Detection for MATLAB

Documentation | Examples

Remaining Useful Life (RUL) Estimation

Train RUL estimator models on historical data to predict time-to-failure. Use the Health Indicator Designer app to interactively transform features into a composite health indicator for RUL model training.

Three Ways to Estimate Remaining Useful Life

Documentation | Examples

The Diagnostic Feature Designer app shows signal data in four panes: signal traces, power spectra, a table of features ranked by one-way ANOVA, and a bar chart sorting the features by importance.

Feature Engineering

Use the Diagnostic Feature Designer app to automatically extract and rank features to train statistical and AI models.

Feature Design for Predictive Maintenance (3:03)

Documentation | Examples

Illustration of a rotating wheel and battery icon with a green lightning bolt.

Component-Specific Predictive Maintenance

Apply component-specific predictive maintenance tools to rotating machinery and batteries. Classify bearing faults, detect leaks in pumps, track changes in motor performance, identify faults in gearboxes, detect anomalies in lithium-ion cells and battery packs, and estimate remaining battery cycle life. Get started quickly with a library of reference examples.

Documentation | Examples

MATLAB code that shows how to create a fileEnsembleDatastore from a set of vibration data files stored locally. The output shows the ensemble represented as a tall table.

Data Management and Preprocessing

Access sensor data stored locally or remotely. Prepare data for algorithm development by removing outliers, filtering, and applying various time, frequency, and time-frequency preprocessing techniques.

Documentation | Examples

A Simscape model showing a pump housing, three plungers, and a crankshaft connected together.

Synthetic Data Generation

Simulate system behavior, faults, and degradations using physics-based models built in Simulink and Simscape, or inject synthetic anomalies directly into time series data. Create digital twins to monitor performance and predict future behavior.

Modeling Methods to Create Digital Twins (8:28)

Documentation | Examples

A MATLAB Coder Report shows MATLAB code for a Remaining Useful Life prediction function on the left and corresponding C++ code on the right. A colorful region maps a single line of MATLAB code to many lines of C++ code.

Embedded Deployment

Use MATLAB Coder to generate C/C++ code directly from feature computation functions, condition monitoring algorithms, and predictive algorithms for real-time embedded processing.

Documentation | Examples

Deploy predictive algorithms within your enterprise ecosystem using MATLAB Production Server.

Cloud Deployment

Use MATLAB Compiler and MATLAB Compiler SDK to scale algorithms in the cloud as shared libraries, packages, web apps, Docker containers, and more. Deploy to MATLAB Production Server on Microsoft^® Azure^® or AWS^® without recoding.

Documentation | Examples

Product Resources:

Documentation Examples Videos Technical articles Functions Requirements Release notes

Predictive Maintenance Video Series

Watch the videos in this series to learn about predictive maintenance.

Watch videos (4 videos)

Predictive Maintenance Toolbox FAQs

Predictive Maintenance Toolbox provides functions and apps for designing condition monitoring and predictive maintenance algorithms for motors, gearboxes, bearings, batteries, and other applications, enabling you to design condition indicators, detect faults and anomalies, and estimate remaining useful life (RUL).

The Diagnostic Feature Designer app lets you interactively extract time, frequency, time-frequency, and physics-based features from sensor data, rank them for effectiveness, and export them to develop application-specific algorithms for fault and anomaly detection.

The toolbox contains survival, similarity, and degradation models that can be trained on historical data to predict time-to-failure.

Yes, you can generate C/C++ code with MATLAB Coder for embedded deployment or create production applications for cloud deployment using MATLAB Compiler, MATLAB Compiler SDK, or MATLAB Production Server.

You can organize and analyze multi-channel, multi-member time series sensor data imported from local files, cloud storage, and distributed file systems. You can also generate simulated failure data from Simulink and Simscape models.

While the toolbox can be used for any predictive maintenance application with time series sensor data, it also contains component-specific tools and reference examples for rotating machinery and batteries. This includes classifying bearing faults, detecting pump leaks, tracking motor performance changes, identifying gearbox faults, detecting anomalies in lithium-ion cells and battery packs, and estimating remaining battery cycle life.

<a href="/content/mathworks/www/en/products/time-series-anomaly-detection.html">Time Series Anomaly Detection for MATLAB</a> is a support package for Predictive Maintenance Toolbox. This support package contains functions and an app for characterizing normal system behavior and detecting anomalies in time series sensor data using ready-to-train statistical, machine learning, and deep learning detectors.