yusha-fias / hkg

Hierarchical Knowledge Guided Fault Intensity Diagnosis of Complex Industrial Systems

Requirements

• Python 3.8.11
• torch 1.9.1
• torchvision 0.10.1

Note: our model is trained on NVIDIA GPU (A100).

Code execution

• train.py is the entry point to the code.
• main.py is the main function of our model.
• models/xxx.py is the network structure of our method (e.g. resnet_add_gcn.py, mobilenet_v2_add_gcn.py, vit_add_gcn.py and so on).
• opts.py is all the necessary parameters for our method (e.g. comprehensive output factor, learning rate and data loading path and so on).
• engine.py contains the construction of the different correlation matrices (e.g. SCM, HKCM, Binary HEKCM and Re-weighted HEKCM).
• gcn_layers.py is the network structure of GCN.
• train/test_data_loader.py represents the loading of training and test datasets.
• generate_adj_file.py indicates the generation of the adjacency matrix.
• generate_word_embedding.py is the generation of word embeddings for the target classes (e.g. GloVe, GoogleNews, FastText and so on).
• Execute train.py

Test dataset

Download datasets from here and place test signals in the subdirectories of ./Data/Test/

Updates

[08.02.2024] Our manuscript was submitted to the 30TH ACM SIGKDD International Conference on Knowledge Discovery & Data Mining (ACM SIGKDD2024).

[18.02.2024] For the time being, some codes are being made available and the full codes will be released when the manuscript is accepted.