Gearbox fault detection using vibration sensors with over 99.9% accuracy
Ensure gearbox reliability with accurate falult detection using AI-accelerated chip
Application overview
For complex industrial systems like gearboxes, it is critical to prevent unexpected equipment failures and optimize maintenance schedules. By monitoring vibration patterns through sensors and machine learning, anomalies can be detected before they become costly failures, moving from reactive or scheduled maintenance to a more efficient predictive approach that reduces downtime and extends equipment lifespan.
Edge AI enables real-time anomaly detection by processing vibration data locally at the machine level. This eliminates cloud dependency and latency while continuously learning normal patterns and identifying deviations. It transforms maintenance from fixed schedules to condition-based interventions, improving response times and enabling scalable deployment across multiple industrial machines.
This use case is powered by TI's proprietary anomaly detection model that has been co-optimized for TI's low-power MCU architecutre. The TI's MSP family of microcontrollers is powered by TinyEngine? NPU and the anomaly detection model are optimized to reach over 99.9% accuracy.
Starting evaluation
Data collection
The gearbox fault detection dataset includes the vibration dataset recorded by SpectraQuest’s Gearbox Fault Diagnostics Simulator. Dataset has been recorded using 4 vibration sensors placed in four different direction, and under variation of load from '0' to '90' percent.Two different scenario are included: healthy condition and broken tooth condition.
There are 20 files in total, 10 for healthy gearbox and 10 from broken one. Each file corresponds to a given load from 0% to 90% in steps of 10%. This open-source dataset is available on Kaggle.
Data quality assessment
Ensure your AI model is build on data that truly matters. Here is the visualization of testing dataset using LDA (Linear Discriminant Analysis) and PCA (Principal component analysis) conponents.?
Build and train your model
Accelerate development with advanced optimization and searching algorithms for design, training, and depolying AI models with superior performance.?
Find the right model for your needs
For this use case, we implement 8 different models. Please explore these models in MSPM0-SDK?and use the pareto front plot below for reference to see the performace of designed models with different hardware footprints.
Deploying your model
CCStudio? Edge AI Studio?gives a start to finish workflow for deploying trained models to embedded targets.For developers seeking deeper customization and control, the MSPM0-SDK?offers a comprehensive framework for building and integrating edge AI functionality into your own embeded applications.
Choosing the Right device for you
The MSP-series MCUs deliver scable performance for executing and accelerating gearbox fault detection models, along with key SoC features critical to your application.?
| Product number | Processing core | NPU available | Clock frequency (MHz) | ? ? ? ?Gearbox fault detection benchmarking metrics? | ||
|---|---|---|---|---|---|---|
| Flash (kB) | SRAM (kB) | |||||
| MSPM0G5187 | Arm? Cortex?-M0+ Core | Yes | 80 | 8947 | 2244 | |
| MSPM0G3507 | Arm? Cortex?-M0+ Core | No | 80 | 10169 | 1920 | |
| MSPM0G3519 | Arm? Cortex?-M0+ Core | No | 80 | 10169 | 1920 | |
| MSPM33C321A | Arm?Cortex?- M33 | No | 160 | 18515 | 1920 | |
All the hardware, software and resources you’ll need to get started
Hardware
LP-MSPM0G5187
MSPM0G5187 LaunchPad? development kit evaluation module for evaluating and testing the gearbox fault cases.
LP-MSPM0G3507
MSPM0G3507 LaunchPad? development kit evaluation module for evaluating and testing the gearbox fault cases.
LP-MSPM0G3519
MSPM0G3519 LaunchPad? development kit evaluation module for evaluating and testing the gearbox fault cases.
LP-MSPM33C321A
MSPM33C321A LaunchPad? development kit evaluation module for evaluating and testing the gearbox fault cases.
Software & development tools
CCStudio? Edge AI Studio
A fully integrated no-code solution for training and compiling PIR Motion detection models, to deploy onto TI embedded microcontroller devices.?
CLI tools
Use this end-to-end model development tool that contains dataset handling, model training and compilation?
MSPM0-SDK
The MSPM0 SDK provides the ultimate collection of software, tools and documentation to accelerate the development of applications for the MSPM0 MCU platform?under a single software package.?