Deep Learning-Based Quality Classification of Cold Spray Coatings with Explainable Feature Analysis

Abstract:

This study investigates the applicability of deep learning models for automated quality classification of cold spray coatings, focusing on three deposition categories: good, degraded, and poor deposition. Three state-of-the-art convolutional architectures, ResNet-50, EfficientNet-B0, and ConvNeXt-Tiny, were evaluated across two training phases designed to assess the impact of dataset balancing, data augmentation, and higher input resolution. In the first phase, models were trained on an imbalanced dataset using only class weighting; EfficientNet-B0 achieved the best performance (ACC 80%, F1 77%), while ResNet-50 showed notable instability (ACC 60%, F1 56%). In the refined second phase, oversampling, advanced augmentation, 380×380 resolution, and early stopping led to substantial performance gains for all models. ConvNeXt-Tiny achieved the most robust and balanced results (ACC 93.3%, F1 90.3%), outperforming EfficientNet-B0 and ResNet-50 particularly in sensitivity and specificity for minority classes. Grad-CAM analysis provided qualitative insights into the decision-making process: poor samples elicited strong, spatially extended activations corresponding to defective regions, degraded samples produced more localized responses aligned with mid-scale irregularities, and good samples yielded diffuse, low-intensity activation patterns associated with surface uniformity. These interpretable attention maps validated the physical relevance of the learned features and confirmed the suitability of ConvNeXt-Tiny for reliable and explainable cold spray quality assessment.

• Periodical: Materials Science Forum (Volume 1187)
• Pages: 17-27
• DOI: https://doi.org/10.4028/p-4P6a64
• Online since: April 2026
• Authors: Alessia Serena Perna, Domenico Rossi, Alessia Auriemma Citarella, Dario de Fazio, Fabiola De Marco*, Luigi Di Biasi, Massimo Durante, Genoveffa Tortora, Antonio Viscusi
• Keywords: Cold Spray, Deep Learning, Explainable Artificial Intelligence

References

[1] Falco, R., & Bagherifard, S. (2025). Cold spray additive manufacturing: A review of shape control challenges and solutions. Journal of Thermal Spray Technology, 1-19.

DOI: 10.1007/s11666-025-01970-0

[2] Rezzoug, A., Gholenji, R. B., & Yandouzi, M. (2025). Innovative Thermal Spray Deposition Techniques for Polymer and Polymeric Matrix Composite Substrates: Methodologies, Characteristics, and Real-World Applications. Journal of Thermal Spray Technology, 1-44.

DOI: 10.1007/s11666-025-02023-2

[3] Zhang, C., Molla, T., Brandl, C., Watts, J., Mccully, R., Tang, C., & Schaffer, G. (2025). Critical velocity and deposition efficiency in cold spray: A reduced-order model and experimental validation. Journal of Manufacturing Processes, 134, 547-557.

DOI: 10.1016/j.jmapro.2024.12.077

[4] Li, W. Y., Zhang, C., Li, C. J., & Liao, H. (2009). Modeling aspects of high velocity impact of particles in cold spraying by explicit finite element analysis. Journal of Thermal Spray Technology, 18(5), 921-933.

DOI: 10.1007/s11666-009-9325-2

[5] Rahmati, S., Mostaghimi, J., Coyle, T., & Dolatabadi, A. (2024). Jetting phenomenon in cold spray: a critical review on finite element simulations. Journal of Thermal Spray Technology, 33(5), 1233-1250.

DOI: 10.1007/s11666-024-01766-8

[6] Eberle, M., Pinches, S., King, H., Guzman, P., Qin, K., & Ang, A. (2025). Predicting Deposition Efficiency Across Diverse Cold Spray Process Parameters Using Machine Learning. Journal of Thermal Spray Technology, 1-24.

DOI: 10.1007/s11666-025-01983-9

[7] Citarella, A. A., Carrino, L., De Marco, F., Di Biasi, L., Perna, A. S., Viscusi, A., & Tortora, G. (2025). AI Data-Driven Optimization of Cold Spray Coating Manufacturing. IEEE Transactions on Industrial Informatics.

DOI: 10.1109/tii.2025.3582358

[8] Savangouder, R. V., Patra, J. C., & Palanisamy, S. (2024). A machine learning technique for prediction of cold spray additive manufacturing input process parameters to achieve a desired spray deposit profile. IEEE Transactions on Industrial Informatics, 20(10), 12275-12283.

DOI: 10.1109/tii.2024.3417300

[9] He, F., Liu, T., & Tao, D. (2020). Why resnet works? residuals generalize. IEEE transactions on neural networks and learning systems, 31(12), 5349-5362.

DOI: 10.1109/tnnls.2020.2966319

[10] Banerjee, C., Mukherjee, T., & Pasiliao Jr, E. (2019, April). An empirical study on generalizations of the ReLU activation function. In Proceedings of the 2019 ACM southeast conference (pp.164-167).

DOI: 10.1145/3299815.3314450

[11] Mascarenhas, S., & Agarwal, M. (2021, November). A comparison between VGG16, VGG19 and ResNet50 architecture frameworks for Image Classification. In 2021 International conference on disruptive technologies for multi-disciplinary research and applications (CENTCON) (Vol. 1, pp.96-99). IEEE.

DOI: 10.1109/centcon52345.2021.9687944

[12] Koonce, B. (2021). EfficientNet. In Convolutional neural networks with swift for Tensorflow: image recognition and dataset categorization (pp.109-123). Berkeley, CA: Apress.

DOI: 10.1007/978-1-4842-6168-2_10

[13] Mistry, J., Koshti, N., & Gautam, G. K. (2024, December). EfficientNetB0 for AI-Generated and Real Image Classification. In International Conference on Information and Communication Technology for Competitive Strategies (pp.307-316). Singapore: Springer Nature Singapore.

DOI: 10.1007/978-981-96-5751-3_26

[14] Mortaz, E. (2020). Imbalance accuracy metric for model selection in multi-class imbalance classification problems. Knowledge-Based Systems, 210, 106490.

DOI: 10.1016/j.knosys.2020.106490

[15] Woo, S., Debnath, S., Hu, R., Chen, X., Liu, Z., Kweon, I. S., & Xie, S. (2023). Convnext v2: Co-designing and scaling convnets with masked autoencoders. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition (pp.16133-16142).

DOI: 10.1109/cvpr52729.2023.01548

[16] Saini, M., & Susan, S. (2023). Tackling class imbalance in computer vision: a contemporary review. Artificial Intelligence Review, 56(Suppl 1), 1279-1335.

DOI: 10.1007/s10462-023-10557-6

[17] Zhu, T., Liu, X., & Zhu, E. (2022). Oversampling with reliably expanding minority class regions for imbalanced data learning. IEEE Transactions on Knowledge and Data Engineering, 35(6), 6167-6181.

DOI: 10.1109/tkde.2022.3171706

[18] Wang, Z., Wang, P., Liu, K., Wang, P., Fu, Y., Lu, C. T., ... & Zhou, Y. (2025). A comprehensive survey on data augmentation. IEEE Transactions on Knowledge and Data Engineering.

[19] Fayyaz, A. M., Abdulkadir, S. J., Talpur, N., Al-Selwi, S. M., Hassan, S. U., & Sumiea, E. H. (2025). Grad-CAM (Gradient-weighted Class Activation Mapping): A systematic literature review. Computers in Biology and Medicine, 198, 111200.

DOI: 10.1016/j.compbiomed.2025.111200