Twin-AI: Intelligent Barrier Eddy Current Separator with Digital Twin and AI Integration
Abstract
1. Introduction
- The primary contributions of this paper are as follows:
- The development of an intelligent optimization model for a conveyor belt, vibration feeder, and magnetic drum angle in a BECS system using real industrial data.
- The integration of a machine vision module with YOLOv11 and a line-scan camera to classify materials by shape (sharp or round).
- Implementation of a thermal monitoring system using an infrared camera to detect fire hazards from ferromagnetic materials.
- A fully automated conveyor belt misalignment correction mechanism based on edge detection and motor control.
- End-to-end integration with a Digital Twin and PLC infrastructure, enabling real-time monitoring, prediction, and self-adjustment.
1.1. Background
1.2. Issues
1.3. Existing Body of Knowledge and State of the Art
1.4. Gap Detection
1.5. Research Questions
2. Supplementary Literature and Related Work
3. System Overview and System Architecture
3.1. System Architecture
- WebPortal (React):
- Each ui component manages its data and directly invokes Use-case modules, ensuring clear functional boundaries and high testability. Cross-component dependencies are handled by lightweight state-slices that broadcast mutation events without storing domain data, thereby avoiding the complexity of a global store while guaranteeing coherence.
- ManagementBackend (.net 7):
- A minimalist controller/hub layer accepts rest and WebSocket traffic; a thin service layer encapsulates integrations with Azure Digital Twins and Event Grid; and ef Core’s DbContext offers transactional access to the structured content database. This layered approach decouples business logic from transport concerns and allows horizontal scaling behind an Azure App Service during peak throughput.
- MachineBackend (Python/Flask):
- Deployed at the edge, it augments raw sensor data with ai-based inference (YOLOv11 object detection, anomaly scoring) before relaying only high-value information northbound. A built-in scheduler executes periodic acquisition and model-evaluation tasks locally, reducing cloud round-trips and guaranteeing the deterministic response times required by the BECS.
- DigitalTwin and Event Grid:
- The graph model provides a semantically rich, versioned representation of conveyors, drums, and feeders, facilitating what-if simulations and complex-event detection. Event Grid’s serverless pub/sub pattern decouples producers and consumers, so new analytic modules can be attached without modifying the existing code.
- IdentityServer:
- Outsourcing authentication to an oidc-compliant Duende service eliminates the duplicated security code across micro-services and enables seamless single-sign-on for the web portal, mobile clients, and future api consumers.
3.2. Digital Twin
- Domain model:
- The logical structure of the plant is expressed in Digital Twins Definition Language (DTDL). Location aggregates any number of Machine twins via the rel_has_machines relationship; each Machine twin exposes operational properties such as SpeedOfBelt or SortingQuality and maintains bi-directional links to its Sensor set. Specialized sensors—e.g., a line-scan Camera—inherit from the generic sensor interface to avoid duplication. This inheritance pattern allows us to extend the model (e.g., add a hyperspectral camera) without refactoring existing code.
- Service integration:
- An ADT instance is provisioned in the same Azure region as the Manage Backend to minimize latency. The ManageBackend exposes a thin TwinService that wraps the Azure.DigitalTwins.Client SDK: the service updates and inserts property patches received from the MachineBackend and establishes relationships on first contact. All twin modifications are configured to trigger an EventGrid topic, which disseminates messages to a function that stores snapshots in cold storage for offline analytics, as well as to the WebSocket hub that feeds the operator dashboard.
- Outcome:
- By adopting ADTs we achieved a single source of truth for both the static topology and the dynamic state of the BECS. The graph query language has simplified cross-asset correlations (e.g., which sensors are attached to machines in location A?), while the built-in eventing mechanism reduced round-trip latency between PLC state changes and browser visualization.
3.3. PLC System Integration
4. Implementation of Core Modules
4.1. Color-Based Material Detection on Moving Components
4.2. Preventing Undesired Events in BECS Operation
4.2.1. Detection and Correction of Conveyor Belt Misalignment
4.2.2. Early Fire Detection
- Local alert: A high-decibel buzzer activates for immediate on-site attention.
4.2.3. Material Edge Detection
4.2.4. Custom Dataset Preparation
4.2.5. YOLOv11-Based Edge Detection
- The input image is passed through a series of convolutional layers to extract spatial features.
- The image is divided into a grid; each grid cell predicts bounding boxes and class labels.
- The model outputs the predicted location (x, y, w, h), class label (e.g., “sharp” or “smooth”), and a confidence score.
- In the segmentation version, a pixel-wise mask is also generated for each detected object.
- Bounding box coordinates.
- Class label (e.g., sharp or smooth).
- Segmentation mask.
- Confidence score (e.g., 0.93).
- where
- BoxLoss: Penalizes inaccuracies in bounding box predictions.
- ClassLoss: Measures classification error using cross-entropy.
- DFLLoss: Distribution Focal Loss for better localization precision.
- SegmentationLoss: Assesses pixel-level segmentation accuracy.
4.3. Smart Separation
4.3.1. Constructing a Reliable Dataset for Machine Learning-Based BECS Optimization
- Copper particles mistakenly sorted into the plastic stream.
- Plastic particles incorrectly deposited in the metal bin.
- Brass fragments misclassified as plastic.
- Aluminum components misdirected into the plastic fraction.
4.3.2. Dataset and Data Preparation
- Input Features (X):
- Vibration speed (low, medium, high);
- Conveyor belt speed (Hz);
- Separation errors for aluminum, copper, brass, and plastic (grams);
- Drum angle (degrees).
- Output Variables (y):
- Optimized vibration speed;
- Optimized conveyor belt speed;
- Optimized drum angle.
4.3.3. Machine Learning-Based Approach
- is the vector of input features, consisting of
- –
- : vibration speed (low, medium, high);
- –
- : conveyor belt speed (Hz);
- –
- : separation error for aluminum in plastics (grams);
- –
- : separation error for copper in plastics (grams);
- –
- : separation error for brass in plastics (grams);
- –
- : separation error for plastics in metals (grams);
- –
- : drum angle (degrees).
- represents the predicted optimized operational settings, including
- –
- : optimized vibration speed (low, medium, high);
- –
- : optimized conveyor belt speed (Hz);
- –
- : optimized drum angle (degrees).
- is a nonlinear function approximated by the random forest model, trained to find the optimal settings for the BECS system.
- represents the model error, accounting for the discrepancy between the predicted and actual optimal operational settings.
4.4. Integration of AI and Internet of Things with PLC
- A thermal image is captured, annotated with the temperature value, and stored.
- A remote warning is sent via a Telegram bot, including the annotated image, to notify the technician.
- A shutdown command is issued to the Siemens PLC using the OPC UA protocol.
- As ns = 2, s = OPC_Daten.Anlage_ausschalten
- A Boolean value of true is written to this node, deactivating the entire system.
5. System Testing and Evaluation
- Aluminum: 96.2% accuracy (19 g error)
- Copper: 98.2% accuracy (9 g error)
- Brass: 97.0% accuracy (15 g error)
- Plastic: 94.8% accuracy (26 g error)
- mAP@50: 0.994, mask recall: 0.979
- Class-wise mAP: aluminum 0.993, copper 0.991, plastic 0.995
- Inference speed: ∼6 ms per image
5.1. Discussion
- Part 1—Training Loss Curves:
- train/box_loss shows a smooth and steady decline from around 1.2 to below 0.7, reflecting better localization accuracy of bounding boxes as training progresses.
- train/seg_loss also steadily decreases, indicating the model’s improved ability to perform precise segmentation of objects.
- train/cls_loss drops sharply in the early epochs and stabilizes, which indicates the model quickly learns to classify objects correctly.
- train/dfl_loss (Distribution focal loss) steadily decreases, improving localization precision in predicted object positions.
- Part 2—Evaluation Metrics for Class B and Class M:
- metrics/precision(B) and metrics/recall(B), as well as their counterparts for class M, show consistent improvement and stabilize around 0.8 to 0.9, indicating strong classification performance for both datasets.
- metrics/mAP50(B) and metrics/mAP50(M) reach close to 0.9, indicating high detection accuracy at a relaxed IoU threshold.
- metrics/mAP50-95(B) and metrics/mAP50-95(M) stabilize around 0.6, which is very good for strict IoU thresholds from 0.5 to 0.95.
- Part 3—Validation Loss Curves:
- val/box_loss, val/seg_loss, val/cls_loss, and val/dfl_loss show significant fluctuation in early epochs but gradually settle, a typical behavior when adapting to unseen data.
- The initial spikes, especially in val/cls_loss, suggest either a higher variability in validation data or class imbalance, but the trend improves over time.
5.2. Limitations
- While the YOLOv11 and YOLOv11n-seg models demonstrated high accuracy in material and shape classification, deploying them on low-powered edge devices, such as the Raspberry Pi or Jetson Nano, still presents computational and memory limitations. Near-real-time inference under harsh industrial conditions remains partially dependent on offloading to more powerful local servers or optimizing the model architecture. To explore lighter alternatives, YOLOv5 and YOLOv8 models were also tested. YOLOv5 showed significantly lower detection performance and is not recommended for industrial or real-time scenarios. However, YOLOv8 provided reasonably good results with moderate accuracy and better efficiency, making it suitable for users with limited computational resources or tighter budgets. All training datasets and source codes for YOLOv8 are available in our public GitHub repository. For applications where lower cost outweighs maximum accuracy, YOLOv8 is a practical choice. Conversely, for projects requiring high detection precision and near real-time performance, applying the YOLOv11 model remains the most effective option.
- The current classification modules primarily focus on four material types: aluminum, copper, brass, and plastics. Although sufficient for many recycling applications, the system’s performance on composite, coated, or contaminated materials has not been directly evaluated on the conveyor belt. However, this issue was addressed in our previous project, PiVisionSort, which was implemented on a vibration feeder system, shown in Figure 36. In that project, a color-based classification approach was combined with a K-means clustering algorithm to detect material types initially. For composite or coated samples with mixed or ambiguous color patterns, a secondary algorithm—decision tree—was introduced. The decision tree received the K-means output and applied a threshold-based rule (e.g., if 80% of the pixels were classified as orange, the material was categorized as copper). This hybrid method significantly improved classification accuracy to over 97%. Tested composite and coated material samples are illustrated in Figure 31. The entire experiment was performed on the vibration feeder setup and published in the PiVisionSort paper [28]. In future work, we plan to investigate this issue further using hyperspectral imaging technology to assess whether spectral analysis can yield more accurate and robust classification results for composite and coated materials.
- During all experimental scenarios, the magnetic drum speed was held constant to ensure mechanical safety. While this approach simplifies optimization, it restricts the ability of the model to explore dynamic drum speed adjustments, which may provide additional gains in separation accuracy under varying input conditions.
- The shape detection model uses a monochrome line-scan camera to distinguish between sharp and smooth materials. While effective for edge recognition, the lack of color limits the system’s ability to perform material identification or distinguish similarities between the materials.
6. Conclusions and Future Work
Author Contributions
Funding
Institutional Review Board Statement
Informed Consent Statement
Data Availability Statement
Acknowledgments
Conflicts of Interest
References
- Forti, V.; Baldé, C.P.; Kuehr, R.; Bel, G. The Global E-Waste Monitor 2020; United Nations University (UNU): Bonn, Germany; International Telecommunication Union (ITU): Geneva, Switerland; International Solid Waste Association (ISWA): Rotterdam, The Netherlands, 2020; Volume 120. [Google Scholar]
- Cucchiella, F.; D’Adamo, I.; Koh, S.; Rosa, P. Recycling of WEEEs: An economic assessment of present and future e-waste streams. Renew. Sustain. Energy Rev. 2015, 51, 263–272. [Google Scholar] [CrossRef]
- Ramanayaka, S.; Santhirasekaram, K.; Vithanage, M. Urban mining of E-waste: Treasure hunting for precious nanometals. In Handbook of Electronic Waste Management; Elsevier: Amsterdam, The Netherlands, 2019; pp. 19–54. [Google Scholar] [CrossRef]
- Aniley, A.A.; Simegn, A.A. A Comprehensive Review on Challenges and Opportunities of e-waste Management Practices in Ethiopia. ASRIC J. Soc. Sci. Humanit. 2023, 4, 99–109. [Google Scholar]
- Luther, L. Managing Electronic Waste: Issues with Exporting E-Waste; DIANE Publishing: Collingdale, PA, USA, 2010. [Google Scholar]
- Smith, Y.R.; Nagel, J.R.; Rajamani, R.K. Eddy current separation for recovery of non-ferrous metallic particles: A comprehensive review. Miner. Eng. 2019, 133, 149–159. [Google Scholar] [CrossRef]
- Rem, P.; Leest, P.; Van den Akker, A. A model for eddy current separation. Int. J. Miner. Process. 1997, 49, 193–200. [Google Scholar] [CrossRef]
- Jujun, R.; Yiming, Q.; Zhenming, X. Environment-friendly technology for recovering nonferrous metals from e-waste: Eddy current separation. Resour. Conserv. Recycl. 2014, 87, 109–116. [Google Scholar] [CrossRef]
- Aschenbrenner, D.; Gros, J.; Fangerow, N.; Werner, T.; Colloseus, C.; Taha, I. Recyclebot–using robots for sustainable plastic recycling. Procedia CIRP 2023, 116, 275–280. [Google Scholar] [CrossRef]
- Lubongo, C.; Bin Daej, M.A.; Alexandridis, P. Recent developments in technology for sorting plastic for recycling: The emergence of artificial intelligence and the rise of the robots. Recycling 2024, 9, 59. [Google Scholar] [CrossRef]
- Wilts, H.; Garcia, B.R.; Garlito, R.G.; Gómez, L.S.; Prieto, E.G. Artificial intelligence in the sorting of municipal waste as an enabler of the circular economy. Resources 2021, 10, 28. [Google Scholar] [CrossRef]
- Cheng, T.; Kojima, D.; Hu, H.; Onoda, H.; Pandyaswargo, A.H. Optimizing Waste Sorting for Sustainability: An AI-Powered Robotic Solution for Beverage Container Recycling. Sustainability 2024, 16, 10155. [Google Scholar] [CrossRef]
- Lakhouit, A. Revolutionizing Urban Solid Waste Management with AI and IoT: A review of smart solutions for waste collection, sorting, and recycling. Results Eng. 2025, 25, 104018. [Google Scholar] [CrossRef]
- Bin, C.; Yi, Y.; Zhicheng, S.; Qiang, W.; Abdelkader, A.; Kamali, A.R.; Montalvao, D. Effects of particle size on the separation efficiency in a rotary-drum eddy current separator. Powder Technol. 2022, 410, 117870. [Google Scholar] [CrossRef]
- Shan, Z.; Yuan, Y.; Cao, B.; Miao, S.; Li, G.; Wang, Q. The effect of interaction between particles on eddy current separation. Sep. Purif. Technol. 2024, 346, 127382. [Google Scholar] [CrossRef]
- Fuller, A.; Fan, Z.; Day, C.; Barlow, C. Digital twin: Enabling technologies, challenges and open research. IEEE Access 2020, 8, 108952–108971. [Google Scholar] [CrossRef]
- Botín-Sanabria, D.M.; Mihaita, A.S.; Peimbert-García, R.E.; Ramírez-Moreno, M.A.; Ramírez-Mendoza, R.A.; Lozoya-Santos, J.d.J. Digital twin technology challenges and applications: A comprehensive review. Remote Sens. 2022, 14, 1335. [Google Scholar] [CrossRef]
- Yu, W.; Patros, P.; Young, B.; Klinac, E.; Walmsley, T.G. Energy digital twin technology for industrial energy management: Classification, challenges and future. Renew. Sustain. Energy Rev. 2022, 161, 112407. [Google Scholar] [CrossRef]
- Zhang, D.; Gao, X. A digital twin dosing system for iron reverse flotation. J. Manuf. Syst. 2022, 63, 238–249. [Google Scholar] [CrossRef]
- Kender, R.; Rößler, F.; Wunderlich, B.; Pottmann, M.; Golubev, D.; Rehfeldt, S.; Klein, H. Development of control strategies for an air separation unit with a divided wall column using a pressure-driven digital twin. Chem. Eng. Process.-Process Intensif. 2022, 176, 108893. [Google Scholar] [CrossRef]
- Guiqiang, W.; Junbao, C.; Chengzhang, L.; Shuo, L. Edge-YOLO: Lightweight Multi-Scale Feature Extraction for Industrial Surface Inspection. IEEE Access 2025, 13, 48188–48201. [Google Scholar] [CrossRef]
- Hao, X.l.; Liang, H. A multi-class support vector machine real-time detection system for surface damage of conveyor belts based on visual saliency. Measurement 2019, 146, 125–132. [Google Scholar] [CrossRef]
- Rahman, M.A.; Bakker, M. Sensor-based control in eddy current separation of incinerator bottom ash. Waste Manag. 2013, 33, 1418–1424. [Google Scholar] [CrossRef]
- Brooks, L.; Gaustad, G.; Gesing, A.; Mortvedt, T.; Freire, F. Ferrous and non-ferrous recycling: Challenges and potential technology solutions. Waste Manag. 2019, 85, 519–528. [Google Scholar] [CrossRef] [PubMed]
- Chamorro, J.; Vallejo, L.; Maynard, C.; Guevara, S.; Solorio, J.A.; Soto, N.; Singh, K.V.; Bhate, U.; Ravi Kumar, G.V.V.; García, J.; et al. Health monitoring of a conveyor belt system using machine vision and real-time sensor data. CIRP J. Manuf. Sci. Technol. 2022, 38, 38–50. [Google Scholar] [CrossRef]
- Ramly, R.; Sajak, A.A.B.; Rashid, M. IoT recycle management system to support green city initiatives. Indones. J. Electr. Eng. Comput. Sci. 2019, 15, 1037–1045. [Google Scholar] [CrossRef]
- Tabaghchi Milan, S.; Darbandi, M.; Jafari Navimipour, N.; Yalcın, S. An energy-aware load balancing method for IoT-based smart recycling machines using an artificial chemical reaction optimization algorithm. Algorithms 2023, 16, 115. [Google Scholar] [CrossRef]
- Kia, S.; Leiding, B. PiVisionSort: Integrating Image Processing and Machine Learning for Material Recognition on Conveyor Belts. In Data, Information and Computing Science; IOS Press: Amsterdam, The Netherlands, 2025; pp. 26–40. [Google Scholar]
- Waheed, F.; Omar, M.; Ibrahim, S.Z.; Chugtai, R.; Aejaz, S.H. Application of industrial IoT in developing a sustainable and automatic liquid filling plant. Pak. J. Sci. Res. (PJOSR) 2023, 3, 130–140. [Google Scholar] [CrossRef]
- Bargal, N.; Deshpande, A.; Kulkarni, R.; Moghe, R. PLC based object sorting automation. Int. Res. J. Eng. Technol. (IRJET) 2016, 3, 103–108. [Google Scholar]
- Almtireen, N.; Reddy, V.; Sutton, M.; Nedvidek, A.; Karn, C.; Ryalat, M.; Elmoaqet, H.; Rawashdeh, N. PLC-Controlled Intelligent Conveyor System with AI-Enhanced Vision for Efficient Waste Sorting. Appl. Sci. 2025, 15, 1550. [Google Scholar] [CrossRef]
- Mustafa, M.N. Classification of maintenance techniques and diagnosing failures methods. In Proceedings of the Journal of Physics: Conference Series; IOP Publishing: Bristol, UK, 2021; Volume 2060. [Google Scholar]
- Schröder, M.; Falk, B.; Schmitt, R. Failure classification and analysis for technical products. Procedia CIRP 2016, 51, 116–121. [Google Scholar] [CrossRef]
- Lee, W.Y. PLC Control of Scrap Battery and Lead Reclamation Process. Bachelor Thesis, Universiti Teknologj PETRONAS, Seri Iskandar, Perak, Malaysia, June 2004. [Google Scholar]
- Kasiemkhan, A. Sensor Based Optimisation of Eddy Current Separation in Bottom Ash Recycling. Master’s Thesis, TU Delft, Delft, The Netherlands, 2013. [Google Scholar]
- Aquib, K.; Fansupkar, Y.; Nath, S.; Gunde, O. Automated Precision: PLC-Based Object Sorting System. In Maharashtra State Board of Technical Education Sponsored Technical Paper Presentation Competition 2023–2024, Electronics and Telecommunication Engg. Group, Mumbai Region; MSBTE: Pune, India, 2024; p. 62. [Google Scholar]
- Moray, R.; Pabalkar, V. Creation of Smart Cities—Perception and Strategies towards Liveable Futures. Int. J. Innov. Creat. Change 2020, 12, 124–126. [Google Scholar]
- Nagel, J.R. An analytic model for eddy current separation. Miner. Eng. 2018, 127, 277–285. [Google Scholar] [CrossRef]
- Zhang, S.; Forssberg, E.; Arvidson, B.; Moss, W. Separation mechanisms and criteria of a rotating eddy-current separator operation. Resour. Conserv. Recycl. 1999, 25, 215–232. [Google Scholar] [CrossRef]
- Yang, H.; Kuang, Z.; Zhu, J.; Xu, F.; Nie, F.; Sun, S. Digital twin key technology on rare earth process. Sci. Rep. 2022, 12, 14727. [Google Scholar] [CrossRef]
- Yang, Y.; Miao, C.; Li, X.; Mei, X. On-line conveyor belts inspection based on machine vision. Optik 2014, 125, 5803–5807. [Google Scholar] [CrossRef]
- Xianguo, L.; Lifang, S.; Zixu, M.; Can, Z.; Hangqi, J. Laser-based on-line machine vision detection for longitudinal rip of conveyor belt. Optik 2018, 168, 360–369. [Google Scholar] [CrossRef]
- Zghaibeh, M. A Blockchain-Based, Smart Contract and IoT-Enabled Recycling System. J. Br. Blockchain Assoc. 2023, 7. [Google Scholar] [CrossRef]
- Li, M.; Chen, F.; Zhou, W. Digital-twin-based system for foam cleaning robots in spent fuel pools. Appl. Sci. 2024, 14, 2020. [Google Scholar] [CrossRef]
- Ruiz, C.; Dashti, A. Mixed Models for Product Design Selection Based on Accelerated Degradation Testing. In Proceedings of the 2025 Annual Reliability and Maintainability Symposium (RAMS), Destin, FL, USA, 27–30 January 2025; IEEE: Piscataway, NJ, USA, 2025; pp. 1–6. [Google Scholar]
- Li, L.; Aslam, S.; Wileman, A.; Perinpanayagam, S. Digital twin in aerospace industry: A gentle introduction. IEEE Access 2021, 10, 9543–9562. [Google Scholar] [CrossRef]
- Kritzinger, W.; Karner, M.; Traar, G.; Henjes, J.; Sihn, W. Digital Twin in manufacturing: A categorical literature review and classification. Ifac-PapersOnline 2018, 51, 1016–1022. [Google Scholar] [CrossRef]
- Sujatanagarjuna, A.; Kia, S.; Briechle, D.F.; Leiding, B. MushR: A smart, automated, and scalable indoor harvesting system for gourmet mushrooms. Agriculture 2023, 13, 1533. [Google Scholar] [CrossRef]
- Piromalis, D.; Kantaros, A. Digital twins in the automotive industry: The road toward physical-digital convergence. Appl. Syst. Innov. 2022, 5, 65. [Google Scholar] [CrossRef]
- Vachálek, J.; Bartalskỳ, L.; Rovnỳ, O.; Šišmišová, D.; Morháč, M.; Lokšík, M. The digital twin of an industrial production line within the industry 4.0 concept. In Proceedings of the 2017 21st International Conference on Process Control (PC), Strbske Pleso, Slovakia, 6–9 June 2017; IEEE: Piscataway, NJ, USA, 2017; pp. 258–262. [Google Scholar]
- Chandan, G.; Jain, A.; Jain, H.; Mohana. Real time object detection and tracking using Deep Learning and OpenCV. In Proceedings of the 2018 International Conference on Inventive Research in Computing Applications (ICIRCA), Coimbatore, India, 11–12 July 2018; IEEE: Piscataway, NJ, USA, 2018; pp. 1305–1308. [Google Scholar]
- Bradski, G.; Kaehler, A. Learning OpenCV: Computer Vision with the OpenCV Library; O’Reilly Media, Inc.: Sebastopol, CA, USA, 2008. [Google Scholar]
- Boyko, N.; Basystiuk, O.; Shakhovska, N. Performance evaluation and comparison of software for face recognition, based on dlib and opencv library. In Proceedings of the 2018 IEEE Second International Conference on Data Stream Mining & Processing (DSMP), Lviv, Ukraine, 21–25 August 2018; IEEE: Piscataway, NJ, USA, 2018; pp. 478–482. [Google Scholar]
- Sharma, A.; Pathak, J.; Prakash, M.; Singh, J. Object detection using OpenCV and python. In Proceedings of the 2021 3rd International Conference on Advances in Computing, Communication Control and Networking (ICAC3N), Greater Noida, India, 17–18 December 2021; IEEE: Piscataway, NJ, USA, 2021; pp. 501–505. [Google Scholar]
- Flusser, J.; Farokhi, S.; Höschl, C.; Suk, T.; Zitova, B.; Pedone, M. Recognition of images degraded by Gaussian blur. IEEE Trans. Image Process. 2015, 25, 790–806. [Google Scholar] [CrossRef]
- Tsomko, E.; Kim, H.; Izquierdo, E. Linear Gaussian blur evolution for detection of blurry images. IET Image Process. 2010, 4, 302–312. [Google Scholar] [CrossRef]
- Chityala, R.N.; Hoffmann, K.R.; Bednarek, D.R.; Rudin, S. Region of interest (ROI) computed tomography. In Proceedings of the Medical Imaging 2004: Physics of Medical Imaging, San Diego, CA, USA, 14–19 February 2004; SPIE: Bellingham, WA, USA, 2004; Volume 5368, pp. 534–541. [Google Scholar]
- Khan, I.H.; Javaid, M. Role of Internet of Things (IoT) in adoption of Industry 4.0. J. Ind. Integr. Manag. 2022, 7, 515–533. [Google Scholar] [CrossRef]
- Rath, K.C.; Khang, A.; Roy, D. The role of Internet of Things (IoT) technology in Industry 4.0 economy. In Advanced IoT Technologies and Applications in the Industry 4.0 Digital Economy; CRC Press: Boca Raton, FL, USA, 2024; pp. 1–28. [Google Scholar]
- Lampropoulos, G.; Siakas, K.; Anastasiadis, T. Internet of things in the context of industry 4.0: An overview. Int. J. Entrep. Knowl. 2019, 7, 4–19. [Google Scholar] [CrossRef]
- García, G.B.; Suarez, O.D.; Aranda, J.L.E.; Tercero, J.S.; Gracia, I.S.; Enano, N.V. Learning Image Processing with OpenCV; Packt Publishing: Birmingham, UK, 2015. [Google Scholar]
- Culjak, I.; Abram, D.; Pribanic, T.; Dzapo, H.; Cifrek, M. A brief introduction to OpenCV. In Proceedings of the 2012 35th International Convention MIPRO, Opatija, Croatia, 21–25 May 2012; IEEE: Piscataway, NJ, USA, 2012; pp. 1725–1730. [Google Scholar]
- Howse, J. OpenCV Computer Vision with Python; Packt Publishing: Birmingham, UK, 2013; Volume 27. [Google Scholar]
- Modrzyk, N. Building Telegram Bots: Develop Bots in 12 Programming Languages Using the Telegram Bot API; Apress: New York, NY, USA, 2018. [Google Scholar]
- Idhom, M.; Fauzi, A.; Alit, R.; Wahanani, H.E. Implementation system telegram bot for monitoring Linux server. In Proceedings of the International Conference on Science and Technology (ICST 2018), Bali, Indonesia, 18–19 October 2018; Atlantis Press: Dordrecht, The Netherlands, 2018; pp. 1089–1093. [Google Scholar]
- Jorge, V.L.; Bendaoud, I.; Soulié, F.; Bordreuil, C. Deep learning-based YOLO for semantic segmentation and classification of weld pool thermal images. Int. J. Adv. Manuf. Technol. 2025, 137, 3573–3585. [Google Scholar] [CrossRef]
- Toapaxi, C.C.; Eduardo, C. Assessment Performance and Emittance Measurements Tests of Basler Digital Camera vs. the Standard BTV System at CLEAR; Report CERN-STUDENTS-Note-2019-045; CERN: Geneva, Switzerland, 2019. [Google Scholar]
- Bodenstorfer, E.; Hasani, Y.; Fürtler, J.; Brodersen, J.; Mayer, K.J. High-speed line-scan camera with multi-line CMOS color sensor. In Proceedings of the 2012 IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops, Providence, RI, USA, 16–21 June 2012; IEEE: Piscataway, NJ, USA, 2012; pp. 9–14. [Google Scholar]
- Ciaglia, F.; Zuppichini, F.S.; Guerrie, P.; McQuade, M.; Solawetz, J. Roboflow 100: A rich, multi-domain object detection benchmark. arXiv 2022, arXiv:2211.13523. [Google Scholar]
- Liang, S.; Wu, H.; Zhen, L.; Hua, Q.; Garg, S.; Kaddoum, G.; Hassan, M.M.; Yu, K. Edge YOLO: Real-time intelligent object detection system based on edge-cloud cooperation in autonomous vehicles. IEEE Trans. Intell. Transp. Syst. 2022, 23, 25345–25360. [Google Scholar] [CrossRef]
- Han, B.G.; Lee, J.G.; Lim, K.T.; Choi, D.H. Design of a scalable and fast YOLO for edge-computing devices. Sensors 2020, 20, 6779. [Google Scholar] [CrossRef]
- Bhavana, N.; Kodabagi, M.M.; Kumar, B.M.; Ajay, P.; Muthukumaran, N.; Ahilan, A. POT-YOLO: Real-Time Road Potholes Detection using Edge Segmentation based Yolo V8 Network. IEEE Sensors J. 2024, 24, 24802–24809. [Google Scholar] [CrossRef]
- Aboyomi, D.D.; Daniel, C. A comparative analysis of modern object detection algorithms: YOLO vs. SSD vs. faster R-CNN. ITEJ (Inf. Technol. Eng. J.) 2023, 8, 96–106. [Google Scholar] [CrossRef]
- Olorunshola, O.; Jemitola, P.; Ademuwagun, A. Comparative study of some deep learning object detection algorithms: R-CNN, fast R-CNN, faster R-CNN, SSD, and YOLO. Nile J. Eng. Appl. Sci 2023, 1, 70–80. [Google Scholar] [CrossRef]
- Redmon, J.; Farhadi, A. Yolov3: An incremental improvement. arXiv 2018, arXiv:1804.02767. [Google Scholar]
- Zhao, Z.Q.; Zheng, P.; Xu, S.t.; Wu, X. Object detection with deep learning: A review. IEEE Trans. Neural Networks Learn. Syst. 2019, 30, 3212–3232. [Google Scholar] [CrossRef] [PubMed]
- Wang, C.Y.; Bochkovskiy, A.; Liao, H.Y.M. YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, Vancouver, BC, Canada, 17–24 June 2023; pp. 7464–7475. [Google Scholar]
- Vlachos, A.; Bargiota, E.; Krinidis, S.; Papadimitriou, K.; Manglis, A.; Fourkiotou, A.; Tzovaras, D. iblueCulture: Data Streaming and Object Detection in a Real-Time Video Streaming Underwater System. Remote Sens. 2024, 16, 2254. [Google Scholar] [CrossRef]
- Gomathi, N.; Sridevi, I. Recovery of noble metal from E-waste using leaching, electro deposition and electro generative process. Der Pharma Chem. 2015, 7, 219–224. [Google Scholar]
- Peng, T.; Sellami, S.; Boucelma, O.; Chbeir, R. Multi-output regression for imbalanced data stream. Expert Syst. 2023, 40, e13417. [Google Scholar] [CrossRef]
- Garreta, R.; Moncecchi, G. Learning Scikit-Learn: Machine Learning in Python; Packt Publishing: Birmingham, UK, 2013; Volume 2013. [Google Scholar]
- Rodriguez-Galiano, V.; Sanchez-Castillo, M.; Chica-Olmo, M.; Chica-Rivas, M. Machine learning predictive models for mineral prospectivity: An evaluation of neural networks, random forest, regression trees and support vector machines. Ore Geol. Rev. 2015, 71, 804–818. [Google Scholar] [CrossRef]
ID | Scenario | Average Temp (°C) | Max Temp (°C) | Response Time (s) |
---|---|---|---|---|
0 | No Iron Present | 19.0 | 21.7 | 1.8 |
1 | Iron Present (Low Speed) | 29.5 | 42.0 | 2.1 |
2 | Iron Present (High Speed) | 40.1 | 45.3 | 2.3 |
3 | Iron and Dust Present | 38.5 | 43.8 | 2.5 |
4 | Mixed Materials (No Iron) | 20.5 | 22.0 | 1.7 |
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. |
© 2025 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
Share and Cite
Kia, S.; Mayer, J.B.; Westphal, E.; Leiding, B. Twin-AI: Intelligent Barrier Eddy Current Separator with Digital Twin and AI Integration. Sensors 2025, 25, 4731. https://doi.org/10.3390/s25154731
Kia S, Mayer JB, Westphal E, Leiding B. Twin-AI: Intelligent Barrier Eddy Current Separator with Digital Twin and AI Integration. Sensors. 2025; 25(15):4731. https://doi.org/10.3390/s25154731
Chicago/Turabian StyleKia, Shohreh, Johannes B. Mayer, Erik Westphal, and Benjamin Leiding. 2025. "Twin-AI: Intelligent Barrier Eddy Current Separator with Digital Twin and AI Integration" Sensors 25, no. 15: 4731. https://doi.org/10.3390/s25154731
APA StyleKia, S., Mayer, J. B., Westphal, E., & Leiding, B. (2025). Twin-AI: Intelligent Barrier Eddy Current Separator with Digital Twin and AI Integration. Sensors, 25(15), 4731. https://doi.org/10.3390/s25154731