7.7.9. Fan Blade Fault Classification

Detect faults in BLDC fans from accelerometer data.

7.7.9.1. Overview

This example demonstrates fault detection in brushless DC (BLDC) fans using vibration data from accelerometers. It can identify various fault conditions including blade damage, bearing wear, and imbalance.

Application: Cooling systems, industrial fans, computer hardware

Task Type: Time Series Classification

Data Type: Multivariate (accelerometer X, Y, Z axes)

7.7.9.2. Demo Setup

Fan Blade Fault Demo Setup

Hardware setup for fan blade fault classification demo

7.7.9.3. Fault Types

The model can identify various fault conditions:

Normal Operation

Normal Operation - Fan running without any faults
Blade Damage

Blade Damage - Physical damage to fan blades
Blade Imbalance

Blade Imbalance - Uneven weight distribution causing vibration
Blade Obstruction

Blade Obstruction - Foreign object interfering with fan operation

7.7.9.4. Configuration

common:
  target_module: 'timeseries'
  task_type: 'generic_timeseries_classification'
  target_device: 'F28P55'

dataset:
  dataset_name: 'fan_blade_fault_classification'

training:
  model_name: 'CLS_4k_NPU'
  training_epochs: 50
  batch_size: 32

testing: {}
compilation: {}

7.7.9.5. Running the Example

cd tinyml-modelzoo
./run_tinyml_modelzoo.sh examples/fan_blade_fault_classification/config.yaml

7.7.9.6. Dataset Details

Input Variables:

  • Accelerometer X-axis

  • Accelerometer Y-axis

  • Accelerometer Z-axis

Classes:

  • Normal operation

  • Blade fault

  • Bearing fault

  • Imbalance

7.7.9.7. Results and Analysis

ROC Curves:

ROC Curves

One-vs-Rest Multi-class ROC curves showing excellent classification performance

Class Score Histogram:

Class Score Histogram

Distribution of class score differences

Feature Extraction Quality:

PCA on Training Data

PCA visualization showing class separation in feature space

7.7.9.8. Anomaly Detection Variant

This example also supports anomaly detection mode:

./run_tinyml_modelzoo.sh examples/fan_blade_fault_classification/config_anomaly_detection.yaml

Reconstruction Error Analysis:

Reconstruction Error

Reconstruction error distribution for anomaly detection

7.7.9.9. See Also