Adaptive convolutional neural network-based principal component analysis algorithm for the detection of manufacturing data
Herein, an adaptive convolutional neural network (CNN)-based principal component analysis (PCA) algorithm for the detection of manufacturing data is proposed. The mentioned algorithm adaptively selects a suitable classification scheme (a CNN-based scheme or PCA-based support vector machine scheme) o...
Saved in:
| Main Author: | |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
SAGE Publishing
2025-04-01
|
| Series: | Advances in Mechanical Engineering |
| Online Access: | https://doi.org/10.1177/16878132251325420 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Herein, an adaptive convolutional neural network (CNN)-based principal component analysis (PCA) algorithm for the detection of manufacturing data is proposed. The mentioned algorithm adaptively selects a suitable classification scheme (a CNN-based scheme or PCA-based support vector machine scheme) on the basis of various types of inputs to detect manufacturing data. Insufficient image signals usually lead to poor performance or failure in visual inspection tasks. For example, feature extraction might fail in vision-based detection methods when no image signals are detected in a manufacturing process. Vision-based inspection is challenging in manufacturing processes. The proposed algorithm can be employed to detect image signals and to obtain manufacturing data when image features cannot be extracted. In an experimental investigation, this algorithm successfully classified manufacturing data on the basis of different inputs and outperformed existing methods for detecting and classifying such data. In summary, the proposed algorithm utilizes a CNN to recognize image signals and employs PCA-based support vector machine scheme for processing measurement data inputs. This adaptive algorithm is capable of learning from limited image signals or features, enhancing data interpretability and increasing the amount of feature information for detecting manufacturing data. |
|---|---|
| ISSN: | 1687-8140 |