As the customization of new energy vehicles accelerates, the quality requirements for die-cast components are becoming increasingly stringent. Accurate and robust defect prediction has thus become critical for enhancing the competitiveness of manufacturing enterprises. In response to challenges such as small sample sizes, severe label imbalance, and strong inter-label correlations in die-casting quality prediction, this paper proposes CastMoE, a Casting-specific Mixture of Experts Framework for multi-label classification. CastMoE is designed to integrate multiple heterogeneous models with structurally diverse architectures through a Gated Mixture-of-Experts Fusion mechanism, which dynamically assigns label-wise softmax weights to adaptively aggregate predictions from different sub-models based on their performance. The framework incorporates a combination-based structure search strategy to explore and evaluate various model subsets, and selects the optimal fusion configuration based on cross-validated metrics. To ensure temporal consistency among variable-length time-series samples from the die-casting process, the Fast Dynamic Time Warping (FastDTW) algorithm is utilized. All sub-models and the fusion network are trained with Focal Loss to address class imbalance and improve the sensitivity to minority labels. During inference, a label-wise adaptive thresholding strategy is applied to convert probabilistic outputs into final binary decisions, with thresholds optimized using validation F1 scores to enhance recall and classification quality. Through the CastMoE framework, the optimal model combination identified by structure search consists of BiLSTM and CNN-LSTM-Attention. This configuration achieved a Micro-F1 score of 96.50% and a Micro-Recall of 95.83% under five-fold cross-validation, demonstrating the framework’s robustness, effectiveness, and industrial applicability for multi-label defect prediction in die-casting production.
PsarommatisFMayG. A systematic analysis for mapping product-oriented and process-oriented zero-defect manufacturing (ZDM) in the industry 4.0 era. Sustainability2023; 15(16): 12251.
2.
SutharJPersisJGuptaR. Critical parameters influencing the quality of metal castings: a systematic literature review. Int J Qual Reliab Manag2023; 40(1): 53–82.
3.
LinJSChenKH. A novel decision support system based on computational intelligence and machine learning: towards zero-defect manufacturing in injection molding. J Ind Inf Integr2024; 40: 100621.
4.
AndriosopoulouGMastakourisAMasourosD, et al. Defect recognition in high-pressure die-casting parts using neural networks and transfer learning. Metals2023; 13(6): 1104.
5.
LiuDDuYChaiW, et al. Digital twin and data-driven quality prediction of complex die-casting manufacturing. IEEE Trans Ind Inform2022; 18(11): 8119–8128.
6.
TangWVianCMTangZ, et al. Anomaly detection of core failures in die casting X-ray inspection images using a convolutional autoencoder. Mach Vis Appl2021; 32(4): 102.
7.
WangXZhangYBaiN, et al. Class-imbalanced time series anomaly detection method based on cost-sensitive hybrid network. Expert Syst Appl2024; 238: 122192.
8.
FrumosuFDKhanARSchiølerH, et al. Cost-sensitive learning classification strategy for predicting product failures. Expert Syst Appl2020; 161: 113653.
9.
WangTWangXMaR, et al. Random forest-bayesian optimization for product quality prediction with large-scale dimensions in process industrial cyber–physical systems. IEEE Internet Things J2020; 7(9): 8641–8653.
10.
LiuFWangYWangK. Quality sensitive feature extraction for end product quality prediction in injection molding processes. In: CCF conference on big data, 8–10 January 2022, pp.182–193. Singapore: Springer.
11.
TercanHMeisenT. Machine learning and deep learning based predictive quality in manufacturing: a systematic review. J Intell Manuf2022; 33(7): 1879–1905.
12.
KhaireUMDhanalakshmiR. Stability of feature selection algorithm: a review. J King Saud Univ Comput Inf Sci2022; 34(4): 1060–1073.
13.
DinevaAMosaviAGyimesiM, et al. Fault diagnosis of rotating electrical machines using multi-label classification. Appl Sci2019; 9(23): 5086.
14.
LiuYYuanYLiuJ. Deep learning model for imbalanced multi-label surface defect classification. Meas Sci Technol2021; 33(3): 035601.
15.
TangWDaiRYanW, et al. Identifying multi-functional bioactive peptide functions using multi-label deep learning. Brief Bioinform2022; 23(1): bbab414.
16.
BoutellMRLuoJShenX, et al. Learning multi-label scene classification. Pattern Recognit2004; 37(9): 1757–1771.
17.
ReadJPfahringerBHolmesG, et al. Classifier chains for multi-label classification. Mach Learn2011; 85: 333–359.
18.
TsoumakasGVlahavasI. Random k-labelsets: an ensemble method for multilabel classification. In: European conference on machine learning, 2007, pp.406–417. Berlin/Heidelberg: Springer.
19.
WangSTianJLiangP, et al. Single and simultaneous fault diagnosis of gearbox via wavelet transform and improved deep residual network under imbalanced data. Eng Appl Artif Intell2024; 133: 108146.
20.
JoeHKimHG. Multi-label classification with XGBoost for metabolic pathway prediction. BMC Bioinform2024; 25(1): 52.
21.
YanZLiuFLiuY. Euk-mPSL: a deep learning framework for multi-label eukaryotic protein subcellular localization prediction with imbalanced datasets. In: Proceedings of the 2023 4th international symposium on artificial intelligence for medicine science, Chengdu, China, 20–22 October 2023, New York, NY: Association for Computing Machinery, Inc.; 2023, pp.1141–1145.
22.
WangYPanZYuanX, et al. A novel deep learning based fault diagnosis approach for chemical process with extended deep belief network. ISA Trans2020; 96: 457–467.
23.
RamliAIMa’arofHRamliMS. A review on surface defect detection using vision-based approach. J Adv Manuf Technol2019; 13(2): 69–84.
24.
SzeliskiR. Computer vision: algorithms and applications. New York, NY: Springer Nature, 2022.
25.
WuYLinYChenZ, et al. Feature learning based on deep model for classifying imbalanced data. Pattern Recognit2018; 81: 205–217.
26.
ParlakIEEmelE. Deep learning-based detection of aluminum casting defects and their types. Eng Appl Artif Intell2023; 118: 105636.
27.
LinCHHuGHHoCW, et al. Press casting quality prediction and analysis based on machine learning. Electronics2022; 11(14): 2204.
28.
TarekegnANGiacobiniMMichalakK. A review of methods for imbalanced multi-label classification. Pattern Recognit2021; 118: 107965.
HeHBaiYGarciaEA, et al. ADASYN: adaptive synthetic sampling approach for imbalanced learning. In: 2008 IEEE international joint conference on neural networks (IEEE World Congress on Computational Intelligence), Hong Kong, China, Piscataway, NJ: IEEE; 2008, pp.1322–1328.
31.
FernándezAGarcíaSGalarM, et al. Learning from imbalanced data sets. Cham: Springer, 2018.
32.
ElujideIFashotoSGFashotoB, et al. Application of deep and machine learning techniques for multi-label classification performance on psychotic disorder diseases. Inform Med Unlocked2021; 23: 100545.
DuanJGuYYuH, et al. ECC++: an algorithm family based on ensemble of classifier chains for classifying imbalanced multi-label data. Expert Syst Appl2024; 236: 121366.
35.
AlotaibiRFlachP. Multi-label thresholding for cost-sensitive classification. Neurocomputing2021; 436: 232–247.
36.
HuKChengYWuJ, et al. Cross-domain intelligent diagnostics for rotating machinery using domain adaptive and adversarial networks. J Ind Inf Integr2024; 42: 100722.
37.
ZhangXWangCZhouW, et al. Trustworthy diagnostics with out-of-distribution detection: a novel max-consistency and min-similarity guided deep ensembles for uncertainty estimation. IEEE Internet Things J2024; 11(13): 23055–23067.
38.
HaoCLiWMaoX, et al. An intelligent prediction paradigm for milling tool parameters design based on multi-task tabular data deep transfer learning integrating physical knowledge. J Manuf Process2025; 134: 998–1020.
39.
DietterichTG.Ensemble methods in machine learning. In: International workshop on multiple classifier systems, 2000, pp.1–15. Berlin/Heidelberg: Springer.
XiaYChenKYangY. Multi-label classification with weighted classifier selection and stacked ensemble. Inf Sci2021; 557: 421–442.
44.
LiJZhuXWangJ. AdaBoost. C2: boosting classifiers chains for multi-label classification. In: Proceedings of the AAAI conference on artificial intelligence, Vancouver, BC, Canada, Palo Alto, CA: AAAI Press; 2023, vol. 37. no. 7, pp.8580–8587.
45.
AbdillahAFPurnama PutraCBApriantoniA, et al. Ensemble-based methods for multi-label classification on biomedical question-answer data. J Inf Syst Eng Bus Intell2022; 8(1): 42–50.
DingCSunSZhaoJ. MST-GAT: a multimodal spatial–temporal graph attention network for time series anomaly detection. Inf Fusion2023; 89: 527–536.
48.
WuHXuJWangJ, et al. Autoformer: decomposition transformers with auto-correlation for long-term series forecasting. Adv Neural Inf Process Syst2021; 34: 22419–22430.
49.
LeeWJXiaKDentonNL, et al. Development of a speed invariant deep learning model with application to condition monitoring of rotating machinery. J Intell Manuf2021; 32(2): 393–406.
50.
NieYNguyenNHSinthongP, et al. A time series is worth 64 words: long-term forecasting with transformers. arXiv preprint arXiv:2211.14730, 2022.
51.
ChenXQiuPZhuW, et al. TimeMIL: advancing multivariate time series classification via a time-aware multiple instance learning. arXiv preprint arXiv:2405.03140, 2024, https://arxiv.org/abs/2405.03140 (cited 19 June 2025).
