Blower Imbalance 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

Blower Imbalance detection is a classification problem to understand the correct working of fan blowers. The problem is to identify whether the fans of a running motor have any imbalance in rotation or not. This becomes important in real-world scenarios such as dust accumulation on fan blades, cooling systems, elevators, and anywhere there is a running motor. This example uses current readings instead of additional sensors to do the classification.

This project demonstrates implementation of an AI-based blower imbalance detection system on AM26x microcontrollers, helping prevent hazards through early detection of faults.

Supported Combinations

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

Dataset and Model Details

Dataset

TI has created a specialized blower imbalance dataset containing AC current measurements (3-phase).

Parameter	Value
Sensor	Current Meter
Channels	3 (AC Current - 3 phases)
Sampling Rates	20 Hz, 30 Hz, 50 Hz, 60 Hz (variable)
Samples per File	Variable: 36,000 - 90,000 samples
Total Files	123 files (69 normal, 54 fault)

Output Classes:

• Class 0: Normal Operation
• Class 1: Blower Imbalance

Model Architecture

This lightweight classification model CLS_1k_NPU contains approximately 1,000 parameters and follows a streamlined architecture consisting of four convolutional layers (each enhanced with BatchNorm and ReLU activation functions) followed by a single linear layer.

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 3-phase AC current
• H (128): samples of timeseries AC current which is 128 in this example
• W (1): width of samples is restricted to 1 for timeseries applications

Output

This model produces a 1D output representing the two possible classes - normal operation or blower imbalance condition.

Feature Extraction Configuration

The feature extraction pipeline uses FFT-based frequency analysis:

• FE_FFT: Performs Fast Fourier Transform (256-point)
• FE_DC_REM: Removes DC component
• FE_BIN: Groups frequencies into 16 bins (bin size 8)
• FE_BIN_NORMALIZE: Normalizes bin values
• FE_LOG: Applies logarithmic scaling
• FE_CONCAT: Combines 8 frames for temporal context

The pipeline processes 256-sample frames per channel, producing 16 features per frame, concatenated across 8 frames for 128 total features per channel.

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 will classify current readings as normal or imbalanced

See Also

AI Examples

Sample Output

Blower Imbalance Detection Example Started ...
Feature extraction mismatches 0
All tests have passed!!
Golden vectors matched: 2 not matched: 0