AI-Reparability-scores-analysis

• The project provides a supervised and unsupervised learning framework to evaluate the product repairability scores for academic research. An example of a smartphone dataset is used to demonstrate how the frameworks work. The models are built on pytorch and trained by GPU.

Install Dependencies

• Python3
• matplotlib
• pandas
• opencv-python
• torch
• torchvision
• openpyxl
• pytorch

Supervised learning framwork

• The framework is shown in the below picture. The input is a teardown image, and the output is the repairability scores.

• Repairability scores evaluation by ResNet50 in a 3-class scale in the testing phase for (left) Samsung Galaxy S6 Edge and (right) Samsung Galaxy Note Fan Edition; Both are in the same cluster based on similarity assessment.

• Repairability scores evaluation by ResNet50 in a 3-class scale in the testing phase for Samsung Galaxy Note 20: (left) teardown image and (right) X-ray image.

Unsupervised learning framwork

• The unsupervised learning framework used ORB to extract features from teardown images before applying K-means to cluster the group. This framework is useful when the repairability scores are unknown.

• The ORB keypoints matching results of Huawei Mate 10 Pro (left) and LG G6 (right) with 79 matching keypoints are in the same cluster.

Citation

If you use the packages, please cite the paper by the following BibTex:

@article{liao2024automated,
  title={Automated Evaluation and Rating of Product Repairability Using Artificial Intelligence-Based Approaches},
  author={Liao, Hao-Yu and Esmaeilian, Behzad and Behdad, Sara},
  journal={Journal of Manufacturing Science and Engineering},
  volume={146},
  number={2},
  year={2024},
  publisher={American Society of Mechanical Engineers Digital Collection}
}