Ting-Hsueh Chuang

Dr. Ting-Hsueh Chuang

orcid

Research Focus & Metrics

Academic Career Length

5Years (2022-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
16 March 2022 in Applied Sciences
Verified
Journal Article
Open access
PDF file
YHYi-Cheng Huang
Yi-Cheng Huang
KHKuo-Chun Hung
Kuo-Chun Hung
CLChun-Chang Liu
Chun-Chang Liu
TCTing-Hsueh Chuang
Ting-Hsueh Chuang
SCSHEAN-JUINN CHIOU
SHEAN-JUINN CHIOU
Metal workpieces are an indispensable and important part of the manufacturing industry. Surface flaws not only affect the appearance, but also affect the efficiency of the workpiece and reduce the safety of the product. Therefore, the appearance of the product needs to be inspected to determine if there are surface defects, such as scratches, dirt, chipped objects, etc., after production is completed. The traditional manual comparison inspection method is not only time-consuming and labor-intensive, but human error is also unavoidable when inspecting thousands or tens of thousands of products. Therefore, Automated Optical Inspection (AOI) is often used today. The traditional AOI algorithm does not fully meet the subtle detection requirements and needs to import a Convolutional Neural Network (CNN), but the common deep residual networks are too large, such as ResNet-101, ResNet-152, DarkNet-19, and DarkNet-53. Therefore, this research proposes an improved customized convolutional neural network. We used a self-built convolutional neural network model to detect the defects on the metal’s surface. Grad–CAM was used to display the result of the last layer of convolution as the basis for judging whether it was OK or NG. The self-designed CNN network architecture could be customized and adjusted without using a large network model. The customized network model designed in this study was compared with LeNet, VGG-19, ResNet-34, DarkNet-19, and DarkNet-53 after training five times each. The experimental results show that the self-built customized deep learning model avoiding the use of pooling and fully connected layers can effectively improve the recognition rate of defective samples and unqualified samples, and reduce the training cost. Our custom-designed models have great advantages over other models. The results of this paper contribute to the development of new diagnostic technologies for smart manufacturing.

26 August 2026 in The International Journal of Advanced Manufacturing Technology
Verified
Journal Article
Closed access
PDF file
TCTing-Hsueh Chuang
Ting-Hsueh Chuang
FLFwu-Hsing Liu
Fwu-Hsing Liu
SKShou-Chih Ko
Shou-Chih Ko
CSChan-Chen Su
Chan-Chen Su
GCGuo-Ying Chen
Guo-Ying Chen
CHChun-Yao Hsu
Chun-Yao Hsu

15 October 2026 in Actuators
Verified
Journal Article
Open access
PDF file
CCChen Chen
Chen Chen
BWBo-Xuan Wu
Bo-Xuan Wu
SCSin-Wei Chen
Sin-Wei Chen
CTChung-chen Tsao
Chung-chen Tsao
TCTing-Hsueh Chuang
Ting-Hsueh Chuang
Deep learning has achieved high accuracy in machining condition monitoring, yet accuracy alone cannot reveal whether a model relies on physically meaningful evidence or on dataset shortcuts. The difficulty is that common tasks such as tool-wear and surface-roughness estimation lack a known ground-truth basis for discrimination, so attribution outputs cannot be validated. This study therefore constructs a benchmark task—three-class cutting-speed classification (Vc = 60, 80, 100 m/min)—in which, for a fixed tool diameter, the spindle-speed-dependent spectral content is an a priori known correct basis, enabling the reliability of explainable AI (XAI) methods to be examined under controlled conditions; the task is deliberately simple and serves as a validation framework for attribution methods rather than as a demanding monitoring problem. Cutting force signals were acquired from 3 × 3 full-factorial milling experiments on 17-4PH stainless steel using a Kistler 9170A rotating dynamometer; the torque Mz and resultant force Fa were segmented by sliding windows into 2970 single-channel samples (15 records × 99 windows × 2 channels), each represented by 50 time segments × 6 amplitude-independent statistics. A 1D-CNN, an LSTM, and a fully connected network (NN) were evaluated under five-fold StratifiedGroupKFold cross-validation. The 1D-CNN achieved 93.97% accuracy with roughly one-seventh of the parameters (NN 90.24%, LSTM 89.12%), the smallest fold-to-fold standard deviation, and 84.5% on the most difficult fold; as a single random seed was used, performance conclusions are descriptive only. Holding out whole cutting conditions reduced the mean accuracy to 83–86% (one feed level: 54–57%), so the accuracies describe generalization across records of the studied conditions, not to unseen conditions. Fold-wise application of LRP, LIME, and SHAP showed that the 1D-CNN and NN distributed importance across the entire signal (normalized entropy ≥ 0.97), whereas for the LSTM the gradient × input attribution used as the LRP implementation placed 9.4 times the uniform share on the final time segment, consistently across all five folds, a concentration that SHAP (expected gradients referenced to background training samples) and the perturbation-based LIME did not reproduce. Gradient-free occlusion and resampling reproduced it, and retraining showed that it follows the last input position while removing the final segments did not reduce accuracy; it therefore reflects positional sensitivity rather than reliance on the sequence end. In this run, cross-fold Spearman correlations of LIME rankings lay within −0.05 to 0.05, versus 0.20–0.70 for LRP and SHAP; because exploratory recalculations with alternative aggregation choices gave markedly different LIME correlations, this is reported as a run-specific observation rather than as a property of LIME. All three architectures relied predominantly on frequency-domain features; because the tooth-passing frequencies (159.2–265.3 Hz) exceed the 100 Hz Nyquist frequency of the 200 Hz sampling rate yet lie below the 300 Hz low-pass cutoff of the signal conditioner, they are aliased into the analyzed band, and the dominant-frequency feature reflects the spindle rotation frequency and aliased tooth-passing components, both proportional to spindle speed and hence consistent with the physical definition of cutting speed. The contributions are: a controlled validation framework that cross-validates attribution credibility through importance concentration, inter-method agreement, and cross-fold reproducibility; and the observation that the end-of-sequence concentration of the LSTM, revealed by gradient × input but not by SHAP or LIME, reflects positional sensitivity rather than the basis of classification, highlighting the need to verify attribution concentration with independent analyses rather than relying on a single attribution method.