Fan Blade Anomaly Detection

Table of Contents

• Introduction
• Supported Combinations
• Dataset and Model Details
  • Dataset
  • Model Architecture
  • Input Features
  • Output
• Feature Extraction Configuration
• Steps to Run the Example
• See Also
• Sample Output

Introduction

Fan blade faults such as imbalance, damage, and obstruction can lead to reduced efficiency, increased energy consumption, and potential system failure. Early detection of these anomalies through vibration analysis enables predictive maintenance, preventing costly downtime and extending equipment lifespan.

This project demonstrates implementation of an AI-based fan blade anomaly detection system on AM26x microcontrollers. It uses an autoencoder model trained only on normal vibration data to detect deviations that indicate potential faults - including fault types never seen during training.

Supported Combinations

Parameter	Value
CPU + OS	r5fss0-0 nortos
Toolchain	ti-arm-clang
Board	am261x-lp
Example folder	examples/ai/fan_blade_anomalydetection/

Dataset and Model Details

Dataset

TI has created a fan blade vibration dataset collected using a 3-axis ADXL355 accelerometer. The dataset captures both normal operation and various fault conditions.

Parameter	Value
Sensor	ADXL355 3-axis Accelerometer
Sampling Rate	4 kHz
Channels	3 (Vibration X, Y, Z axes)
Samples per File	~20,000 samples (~5 seconds of data)
Total Files	287 files (100 Normal, 187 Anomaly)

Anomaly Types:

• Blade imbalance
• Blade damage
• Blade obstruction

Important: For anomaly detection, the model is trained only on normal data. All anomaly samples are used exclusively for testing.

Model Architecture

This autoencoder model AD_17k contains approximately 17,000 parameters. The encoder compresses the input vibration features into a compact representation, and the decoder reconstructs the features. When presented with anomalous vibration data, the reconstruction error is significantly higher than for normal data.

Input Features

The model takes 4D input (N,C,H,W):

• N (1): batch size which is restricted to 1
• C (3): channels which is 3 for vibration X, Y, Z axes
• H (64): features per channel (16 frequency bins x 4 concatenated frames)
• W (1): width of samples is restricted to 1 for timeseries applications

Output

The model produces a reconstructed feature vector of the same shape as the input (1, 3, 64, 1). The reconstruction error (MSE between input and output) is compared against a threshold:

if reconstruction_error > threshold:
    -> ANOMALY
else:
    -> NORMAL

Feature Extraction Configuration

The feature extraction pipeline uses FFT-based frequency analysis on 3-axis vibration data:

• FE_FFT: Performs Fast Fourier Transform (256-point)
• FE_DC_REM: Removes DC component
• FE_BIN: Groups frequencies into 16 bins
• FE_BIN_NORMALIZE: Normalizes bin values
• FE_LOG: Applies logarithmic scaling
• FE_CONCAT: Combines 4 frames for temporal context

The pipeline produces 16 features per frame x 4 frames = 64 features per channel. With 3 channels (X, Y, Z), the total input is 192 features.

Steps to Run the Example

• When using CCS projects to build, import the CCS project for the required combination and build it using the CCS project menu (see Using SDK with CCS Projects).
• When using makefiles to build, note the required combination and build using make command (see Using SDK with Makefiles)
• Launch a CCS debug session and run the executable, see CCS Launch, Load and Run
• The application computes reconstruction error and classifies as Normal or Anomaly

See Also

AI Examples

Sample Output

Fan Blade Anomaly Detection Example Started ...
Feature extraction mismatches 0
All tests have passed!!
Golden vectors matched: 2 not matched: 0
Reconstruction error is 0.002345, Threshold is 0.010000
The test sample is a Normal sample