{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,2,11]],"date-time":"2026-02-11T13:14:52Z","timestamp":1770815692565,"version":"3.50.1"},"reference-count":26,"publisher":"MDPI AG","issue":"4","license":[{"start":{"date-parts":[[2026,2,10]],"date-time":"2026-02-10T00:00:00Z","timestamp":1770681600000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["Processes"],"abstract":"<jats:p>Automated Fiber Placement (AFP) is a critical process in composite manufacturing, where precise fiber tow placement is essential for achieving high-quality and high-performance engineering components. However, deviations in process variables frequently lead to defects such as gaps and overlaps, which can compromise structural integrity. While various monitoring techniques exist, accurately predicting and understanding the formation of these defects from complex sensor data remains challenging. This work introduces a novel application of a Transformer-based deep learning architecture to enhance the estimation of gap widths in AFP. Leveraging a publicly available industrial AFP dataset, our methodology incorporates a customized positional encoding scheme to effectively integrate the critical spatial context of the tow layup process. The model\u2019s predictive performance was evaluated, achieving a Mean Absolute Percentage Error (MAPE) of 1.04% and an R-squared (R2) value of 0.9143, demonstrating its capability for accurate gap width estimation. Furthermore, SHapley Additive exPlanations (SHAP) analysis was employed to assess the complex interplay between sources of manufacturing process variation. This study establishes the Transformer architecture as a promising and interpretable data-driven tool for AFP process monitoring. The results serve as a proof of concept for attention-based virtual metrology, offering a pathway towards deeper process understanding and defect mitigation.<\/jats:p>","DOI":"10.3390\/pr14040609","type":"journal-article","created":{"date-parts":[[2026,2,11]],"date-time":"2026-02-11T09:16:08Z","timestamp":1770801368000},"page":"609","update-policy":"https:\/\/doi.org\/10.3390\/mdpi_crossmark_policy","source":"Crossref","is-referenced-by-count":0,"title":["Automated Fiber Placement Gap Width Prediction Using a Transformer-Based Deep Learning Approach"],"prefix":"10.3390","volume":"14","author":[{"ORCID":"https:\/\/orcid.org\/0000-0003-1737-5821","authenticated-orcid":false,"given":"Diogo","family":"Cardoso","sequence":"first","affiliation":[{"name":"INEGI\u2014Institute of Science and Innovation in Mechanical and Industrial Engineering, 4200-465 Porto, Portugal"},{"name":"FEUP\u2014Faculty of Engineering, University of Porto, 4200-465 Porto, Portugal"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-4146-6224","authenticated-orcid":false,"given":"Ant\u00f3nio Ramos","family":"Silva","sequence":"additional","affiliation":[{"name":"INEGI\u2014Institute of Science and Innovation in Mechanical and Industrial Engineering, 4200-465 Porto, Portugal"},{"name":"FEUP\u2014Faculty of Engineering, University of Porto, 4200-465 Porto, Portugal"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-6486-3954","authenticated-orcid":false,"given":"Nuno","family":"Correia","sequence":"additional","affiliation":[{"name":"INEGI\u2014Institute of Science and Innovation in Mechanical and Industrial Engineering, 4200-465 Porto, Portugal"},{"name":"LAETA\u2014Associated Laboratory of Energy, Transports and Aerospace, 4200-465 Porto, Portugal"}]}],"member":"1968","published-online":{"date-parts":[[2026,2,10]]},"reference":[{"key":"ref_1","doi-asserted-by":"crossref","first-page":"100182","DOI":"10.1016\/j.jcomc.2021.100182","article-title":"Automated fiber placement: A review of history, current technologies, and future paths forward","volume":"6","author":"Brasington","year":"2021","journal-title":"Compos. Part C Open Access"},{"key":"ref_2","doi-asserted-by":"crossref","first-page":"2099","DOI":"10.1007\/s10845-021-01774-3","article-title":"Review of image segmentation techniques for layup defect detection in the Automated Fiber Placement process: A comprehensive study to improve AFP inspection","volume":"32","author":"Meister","year":"2021","journal-title":"J. Intell. Manuf."},{"key":"ref_3","doi-asserted-by":"crossref","first-page":"317","DOI":"10.4271\/2014-01-2272","article-title":"Improving AFP cell performance","volume":"7","author":"Rudberg","year":"2014","journal-title":"SAE Int. J. Aerosp."},{"key":"ref_4","doi-asserted-by":"crossref","first-page":"106610","DOI":"10.1016\/j.compositesa.2021.106610","article-title":"Experimental investigation of the effect of half gap\/half overlap defects on the strength of composite structures fabricated using automated fibre placement (AFP)","volume":"150","author":"Cadran","year":"2021","journal-title":"Compos. Part A Appl. Sci. Manuf."},{"key":"ref_5","doi-asserted-by":"crossref","first-page":"107654","DOI":"10.1016\/j.compositesa.2023.107654","article-title":"Review of in-process defect monitoring for automated tape laying","volume":"173","author":"Yadav","year":"2023","journal-title":"Compos. Part A Appl. Sci. Manuf."},{"key":"ref_6","doi-asserted-by":"crossref","first-page":"109160","DOI":"10.1016\/j.compositesb.2021.109160","article-title":"Investigations on Explainable Artificial Intelligence methods for the deep learning classification of fibre layup defect in the automated composite manufacturing","volume":"224","author":"Meister","year":"2021","journal-title":"Compos. Part Eng."},{"key":"ref_7","doi-asserted-by":"crossref","unstructured":"Stojkovic, M., and Butt, J. (2022). Industry 4.0 Implementation Framework for the Composite Manufacturing Industry. J. Compos. Sci., 6.","DOI":"10.3390\/jcs6090258"},{"key":"ref_8","doi-asserted-by":"crossref","unstructured":"Lachekhab, F., Benzaoui, M., Tadjer, S.A., Bensmaine, A., and Hamma, H. (2024). LSTM-Autoencoder Deep Learning Model for Anomaly Detection in Electric Motor. Energies, 17.","DOI":"10.3390\/en17102340"},{"key":"ref_9","doi-asserted-by":"crossref","first-page":"111891","DOI":"10.1016\/j.compositesb.2024.111891","article-title":"Predicting gaps and overlaps in Automated Fiber Placement composites by measuring sources of manufacturing process variations","volume":"291","author":"Pantoji","year":"2025","journal-title":"Compos. Part B Eng."},{"key":"ref_10","unstructured":"Druiff, P. (2023). Data-Driven Manufacturing in the Automated Lay-Up of Composites. [Ph.D. Thesis, University of Bristol]."},{"key":"ref_11","doi-asserted-by":"crossref","unstructured":"Schmitt, R., Mersmann, C., and Damm, B. (2010). In-Process 3D Laser Measurement to Control the Fiber Tape-Laying for Composite Production, SPIE.","DOI":"10.1117\/12.853882"},{"key":"ref_12","doi-asserted-by":"crossref","first-page":"505","DOI":"10.1016\/j.promfg.2020.10.071","article-title":"Inline quality control for thermoplastic automated fibre placement","volume":"51","author":"Schuster","year":"2020","journal-title":"Procedia Manuf."},{"key":"ref_13","doi-asserted-by":"crossref","first-page":"112514","DOI":"10.1016\/j.compstruct.2020.112514","article-title":"Machine learning in composites manufacturing: A case study of Automated Fiber Placement inspection","volume":"250","author":"Sacco","year":"2020","journal-title":"Compos. Struct."},{"key":"ref_14","doi-asserted-by":"crossref","first-page":"116051","DOI":"10.1016\/j.compstruct.2022.116051","article-title":"An in-process inspection method integrating deep learning and classical algorithm for automated fiber placement","volume":"300","author":"Tang","year":"2022","journal-title":"Compos. Struct."},{"key":"ref_15","doi-asserted-by":"crossref","unstructured":"Zambal, S., Heindl, C., Eitzinger, C., and Scharinger, J. (2019). End-to-End Defect Detection in Automated Fiber Placement Based on Artificially Generated Data, SPIE.","DOI":"10.1117\/12.2521739"},{"key":"ref_16","doi-asserted-by":"crossref","first-page":"109432","DOI":"10.1016\/j.compositesb.2021.109432","article-title":"On the effect of manual rework in AFP quality control for a doubly-curved part","volume":"227","author":"Sacco","year":"2021","journal-title":"Compos. Part B Eng."},{"key":"ref_17","doi-asserted-by":"crossref","first-page":"100174","DOI":"10.1016\/j.dajour.2023.100174","article-title":"A predictive maintenance model using long short-term memory neural networks and Bayesian inference","volume":"6","author":"Pagano","year":"2023","journal-title":"Decis. Anal. J."},{"key":"ref_18","doi-asserted-by":"crossref","first-page":"101916","DOI":"10.1016\/j.aei.2023.101916","article-title":"Cluster-based industrial KPIs forecasting considering the periodicity and holiday effect using LSTM network and MSVR","volume":"56","author":"Wang","year":"2023","journal-title":"Adv. Eng. Inform."},{"key":"ref_19","doi-asserted-by":"crossref","unstructured":"Chabane, A.N., Sahnoun, M., and Bettayeb, B. (2021, January 26\u201328). Forecasting KPIs of Production Systems Using LSTM Networks. Proceedings of the 2021 1st International Conference on Cyber Management and Engineering (CyMaEn), Virtual.","DOI":"10.1109\/CyMaEn50288.2021.9497278"},{"key":"ref_20","doi-asserted-by":"crossref","first-page":"1539","DOI":"10.1109\/TIE.2017.2733438","article-title":"Machine health monitoring using local feature-based gated recurrent unit networks","volume":"65","author":"Zhao","year":"2017","journal-title":"IEEE Trans. Ind. Electron."},{"key":"ref_21","doi-asserted-by":"crossref","unstructured":"Xu, D., Zhang, Z., and Shi, J. (2022, January 23\u201325). Failure prediction using gated recurrent unit and autoencoder in complex manufacturing process. Proceedings of the 2022 4th World Symposium on Artificial Intelligence (WSAI), Jilin, China.","DOI":"10.1109\/WSAI55384.2022.9836412"},{"key":"ref_22","unstructured":"Vaswani, A. (2017, January 4\u20139). Attention is all you need. Proceedings of the 31st Conference on Neural Information Processing Systems (NIPS 2017), Long Beach, CA, USA."},{"key":"ref_23","doi-asserted-by":"crossref","unstructured":"Wen, Q., Zhou, T., Zhang, C., Chen, W., Ma, Z., Yan, J., and Sun, L. (2022). Transformers in time series: A survey. arXiv.","DOI":"10.24963\/ijcai.2023\/759"},{"key":"ref_24","first-page":"32053","article-title":"Towards learning universal hyperparameter optimizers with transformers","volume":"35","author":"Chen","year":"2022","journal-title":"Adv. Neural Inf. Process. Syst."},{"key":"ref_25","unstructured":"Lundberg, S.M., and Lee, S. (2017). A unified approach to interpreting model predictions. arXiv."},{"key":"ref_26","doi-asserted-by":"crossref","first-page":"e70056","DOI":"10.1111\/cts.70056","article-title":"Practical guide to SHAP analysis: Explaining supervised machine learning model predictions in drug development","volume":"17","author":"Schmitt","year":"2024","journal-title":"Clin. Transl. Sci."}],"container-title":["Processes"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/www.mdpi.com\/2227-9717\/14\/4\/609\/pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,2,11]],"date-time":"2026-02-11T10:01:10Z","timestamp":1770804070000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.mdpi.com\/2227-9717\/14\/4\/609"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,2,10]]},"references-count":26,"journal-issue":{"issue":"4","published-online":{"date-parts":[[2026,2]]}},"alternative-id":["pr14040609"],"URL":"https:\/\/doi.org\/10.3390\/pr14040609","relation":{},"ISSN":["2227-9717"],"issn-type":[{"value":"2227-9717","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,2,10]]}}}