Dynamic Vision-Enabled Machine Condition Monitoring: A Point Cloud-Based Diagnostic Methodology

📄 Paper: Dynamic Vision-Enabled Machine Condition Monitoring: A Point Cloud-Based Diagnostic Methodology

This article presents one of the pioneering attempts in the current literature on using a 3D point cloud data algorithm to process event data for machine fault diagnosis. A geometric data structure is proposed to represent event information and a deep learning method on point cloud data is employed for classification to perform diagnosis tasks. The article conducts experiments on diagnosing rotating machine faults using event-based cameras.

Figure 1: The experimental setup used to collect event-based data for fault diagnosis

Setup Instructions

Prerequisites

Ensure the following software is installed:

• Python 3.x

Installation Steps

Clone the repository:

git clone https://github.com/Tamphie/Dynamic_vision.git
cd Dynamic_vision/faultNet

Repository Structure

Main Scripts

• train_classification.py: Script for training the fault classification model.
• test_classification.py: Script for evaluating the trained classification model.
• train_partseg.py: Script for part segmentation tasks (if applicable).
• train_semseg.py: Script for semantic segmentation tasks (if applicable).

Figure 2: The network architecture designed for classifying fault types

Data Preprocessing

• Generate_txt_*.py: Scripts for generating point cloud data files from event data for specific fault types (e.g., ball, healthy, inner race, outer race).
• dat_to_txt.py: Converts .dat files (raw event data) into text-based formats for further processing.
• load_data_to_txt.py: Handles data loading and transformation into required formats.

Event Data Visualization and Analysis

• dataset_visualization.py: Visualizes datasets for better understanding and debugging.
• event_count.py: Counts events for data statistics.
• fft.py: Computes Fast Fourier Transform for frequency-domain analysis.
• random_roi_fig.py: Creates figures of random regions of interest for visualization purposes.

Figure 3: Visualization of randomly selected regions of interest (ROIs) from the dataset. These ROIs demonstrate the spatial distribution of event data points captured by the dynamic vision sensor.

Utility Scripts during Experiments

• obtain_file_names.py: Helper script for managing file names and paths.
• loadResult.py: Loads and manages experiment results.
• read_total_time.py: Processes timing information for experiments.

Contributions

Feel free to contribute by submitting pull requests or issues. For major changes, please open an issue to discuss proposed modifications.