7.7.18. Gearbox Fault Detection

Classify gearbox operating conditions from vibration sensor data.

7.7.18.1. Overview

This example demonstrates real-time gearbox fault classification using 4-channel vibration data on the MSPM0G5187 MCU with integrated NPU. The model detects healthy operation vs. broken tooth faults, enabling predictive maintenance on resource-constrained microcontrollers without cloud connectivity.

Application: Predictive maintenance, industrial machinery, gearbox health monitoring

Task Type: Time Series Classification

Data Type: Multivariate (4 vibration sensors)

7.7.18.2. Device Support

Device	Description	Configuration File
`MSPM0G5187`	MSPM0 with NPU (primary)	`config_MSPM0.yaml`

7.7.18.3. System Components

Hardware

MSPM0G5187 with integrated NPU

Software

Code Composer Studio 12.x or later
MSPM0 SDK 2.10.00 or later
TI Edge AI Studio

7.7.18.4. Dataset

The example uses the SpectraQuest Gearbox Fault Diagnostics Simulator dataset:

Source: Gearbox Fault Diagnosis — Kaggle
Classes: 2 (Healthy, Broken Tooth)
Operating conditions: 10 load levels (0–90%)

Each CSV file contains 4 accelerometer columns (a1, a2, a3, a4) with ~88,000 rows per file. File naming convention: {b|h}30hz{load}.csv where b = broken tooth, h = healthy.

7.7.18.5. Feature Extraction

Sensor Input — 4-channel accelerometer data
Windowing — 256-sample fixed-length windows
Single frame passed directly to model

7.7.18.6. Models

Model	Parameters	Flash (bytes)	Description
`GearboxFault_model_1.2k`	~1,174	~8,233	4-layer network, progressive channel reduction (12→12→8→8)
`GearboxFault_model_1.5k`	~1,914	~8,878	3-layer network, constant channels (16→16→16)

Both models use INT8 quantization with 4-channel, 256-sample input.

Expected Accuracy: 97–100% depending on model selection.

7.7.18.7. Configuration

common:
  target_module: 'timeseries'
  task_type: 'generic_timeseries_classification'
  target_device: 'MSPM0G5187'

dataset:
  dataset_name: 'gearbox_fault_detection'

training:
  model_name: 'GearboxFault_model_1.2k'
  training_epochs: 30
  batch_size: 32
  quantization: 2
  quantization_method: 'QAT'
  quantization_weight_bitwidth: 8
  quantization_activation_bitwidth: 8

testing: {}
compilation: {}

7.7.18.8. Running the Example

cd tinyml-modelzoo
./run_tinyml_modelzoo.sh examples/gearbox_fault_detection/config_MSPM0.yaml

cd tinyml-modelzoo
run_tinyml_modelzoo.bat examples\gearbox_fault_detection\config_MSPM0.yaml

7.7.18.9. See Also

Motor Bearing Fault — Bearing fault classification from vibration data
Fan Blade Fault Classification — Fan blade fault classification
Example on GitHub
SpectraQuest Dataset on Kaggle
TI Neural Network Compiler User Guide