RESULTS
Scientific Publications
A robot-based inspecting system for 3D measurement
Author: Marche Polytechnic University
Zero Defect Manufacturing aims to minimize the number of defects within a process through proper measurement and control that make possible defect prediction and prevention. This procedure should ideally be performed in a non-destructive approach. Hence, this paper presents two novel non-destructive measurement systems for the geometric control of metal bars concerning a plant producing steel parts. The aforementioned systems are designed to be integrated into the actual process with minimum intervention supporting the online quality control of the manufactured product.The two measurement systems exploit an industrial robotic manipulator and an optical sensor mounted on the robot’s end effector. They differ in the strategy of motion of the laser line triangulation sensor relative to the steel part to be measured. The proposed systems have been implemented as prototypes and deployed at the premises of a steel manufacturer to test and validate their performance, with the preliminary findings being provided and discussed in this work.
Beam Straightness Measurement with Laser Triangulation System: a steel industry use case
Author: Marche Polytechnic University
With reference to a steel bars manufacturing process, there is a variety of factors which can contribute to defect generation such as geometrical non-conformity. Nondestructive Inspection systems (NDIs) are a key element for the early detection of defects in a production line. This paper considers a particular steel industry use case focusing on the design and development of an NDI system to measure the straightness of steel bars in line, by a non-intrusive approach. This NDI is based on the laser line triangulation technique, interacts with a robot and is connected to a software platform where additional services may be exposed. The paper presents a parametric study of the laser line triangulation system to be developed, highlighting the influence of design parameters over system resolution and measurement range. Considering the use case this paper focuses on, the straightness deviation can be correctly estimated as the resolution value can be extremely fine: 0.01 mm if using subpixel accuracy. The steel beam is ideally modelled as a parallelepiped, however in reality its shape can deviate: this causes uncertainty due to the model of the measurand. The paper then discusses this uncertainty and the one related to the misalignment of the laser plane with respect to the beam axis. Results show erroneous estimation of the straightness deviation up to 1.8 mm in some cases of beam distortions analysed. The simulation presented in the paper is therefore of primary importance to evaluate the factors which may influence the overall uncertainty of the straightness measurement process.
Integration of Non-Destructive Inspection (NDI) systems for Zero-Defect Manufacturing in the Industry 4.0 era
Author: Marche Polytechnic University
Industry 4.0 paradigm and its enabling technologies, such as Internet of Things (IoT), have increased the potential of industrial automation with the growing implementation of Cyber-Physical Production Systems (CPPS). They improve the production processes thanks to an optimal use of data. Based on the Reference Architecture Model Industry 4.0 (RAMI 4.0) and the Asset Administration Shells (AAS), according to standards and communication protocols, the European project OpenZDM is realizing an open platform to perform the Zero-Defect Manufacturing paradigm. The platform will be developed and demonstrated in five representative production lines combining ICT solutions and innovative Non-Destructive Inspection systems (NDI). NDI systems will be used as an Industry 4.0 enabling technology to collect and analyze data from the physical assets and convert them to value-added information, as an IoT technology. A relevant example of an NDI system is presented in this paper, based on the acquisitions of an infrared camera, to focus on the business logic unit and the OT/IT convergence and integration.
Repeatability and Reproducibility method of hand-handle laser triangulation profilometer
Author: Marche Polytechnic University with U-Sense.it
This paper explores the use of a hand-held laser triangulation-based measurement system for gap and flush measurements in the automotive industry, focusing on operator variability. In modern manufacturing, especially with Industry 5.0, precise measurements are essential for quality and efficiency. While laser triangulation technology offers high accuracy, when the device is portable human factors influence performance. The study evaluates the U-Sense.IT srl G3F laser profilometer in two modes: manual and with a magnetic holder. Using Measurement System Analysis (MSA), the research assesses repeatability (Equipment Variation, EV) and reproducibility (OperatorVariation, AV), determining the Gauge Repeatability and Reproducibility (GRR). Data were collected from three operators measuring three cars, with repeated measurements for both gap and flush. Results show that the holder technique has lower variability in both repeatability and reproducibility compared to the manual method. The GRR for the holder was 2,22% for flush and 2,32% for gap, while the manual technique showed 3,13% for flush and 3,66% for gap. Both systems had GRR values below the acceptable 10% threshold, with the magnetic holder system demonstrating greater precision due to better sensor positioning. This research highlights the effectiveness of support systems in reducing human error, even if it reduces flexibility, showing that while both methods are valid, the magnetic holder system aligns better with Industry 5.0 principles by improving collaboration between humans and intelligent systems.
Laser Line Triangulation Sensor With Wide Measurement Range: A Steel Industry Use Case
Author: Marche Polytechnic University
Laser line triangulation is a key measurement technology that can be applied in quality control processes to inspect product geometry without contact. This paper presents the design of a Laser Line Triangulation (LLT) sensor developed in the context of the Horizon Europe (HE) project openZDM and tested inline in a steel industry use case.The LLT sensor has been developed focusing on two main target objectives: wide measurement range along transversal (X) and axial (Z) directions and small resolution along the Z axis. These have been achieved thanks to the optimization of the design parameters of the system which has been implemented with a triangulation angle of 45° reaching over a stand-off distance of 1000 mm and a 1000 mm camera-laser distance. The overall measurement range results in a transversal Field-OfView (X-FOV) from 800 mm to 2000 mm and a minimum Z distance of 630 mm in the near field up to 1580 mm in the far field.The resolution along the Z axis has been enhanced thanks to the introduction of subpixel accuracy reaching 0.026 mm at the stand-off distance which is 1000 mm. In this paper, these values are compared to other Laser Line Triangulation sensors available on the market showing that the system developed by UNIVPM is innovative.The Laser Line Triangulation sensor has been also tested inline and this paper presents some results of the laser profiles acquired on a steel bar held by a robot. These laser profiles are then processed in order to detect geometric defects, in particular the non-straightness of the bar, using an algorithm properly developed for the use case. From the results of these feasibility tests, the application has proven effective and opens the path to a full-inline installation, in order to measure non-straightness on 100% of the products.
Laser line triangulation measurement on incandescent steel objects:
methodologies to improve optical signal to noise ratio
Author: Marche Polytechnic University
Integrating optical sensors into harsh industrial environments poses challenges and limits to the application. This paper discusses the implementation of a LLT (Laser Line Triangulation) sensor in a steel industry use case, designed to measure the straightness of steel bars at very high temperatures, up to 1000-1200 °C. Due to the bar incandescence it is challenging to use the optical instrument, as Signal to Noise Ratio (SNR) of the laser line projected onto the inspected object is lower than 2 dB. Both hardware and software solutions are investigated, in particular the choice of a very narrowband filter is selected to remove quite all the other spectral content coming from the bar irradiance. Even though this solution enhances SNR, the narrow band filter blocks the laser irradiance at the edges of the field of view: a reference ruler measured at ambient temperature has shown that 40.3 % of the length of the bar is not detected with respect to the measurement without the narrow band filter. To cope with this issue, a software solution is implemented which consists in the preprocessing of the image with a Region of Interest (ROI) around the laser line. This solution imposes the constraint of repeatable product positioning which has to be assured by the robot which picks up the product for the measurement. Applying the ROI, the Laser Line Extraction (LLE) algorithm detects clearly the laser line, with outliers reduction of 97.5 % with respect to the case without ROI introduction.
2D Temperature Distribution Reconstruction of Steel Bars under Thermal Transient from Sequences of Occluded Infrared Images
Author: Marche Polytechnic University
In-line thermography evaluates surface temperature distributions for quality control. This paper addresses the issue when the thermal imaging camera cannot fully capture a moving component undergoing successive stages of processing with transient thermal behavior. Reconstruction of the 2D temperature distribution requires stitching sequential partial images. When the object is moving and undergoing a thermal transient, simple stitching of sequential images leads to discontinuities and erroneous temperature distribution because the images are framed at different times. The correction of such artifacts is demonstrated using steel bars coming out of an induction furnace as an example. Two strategies are compared: temporal alignment and spatial alignment. Spatial alignment considers cooling relative to the distance from the furnace, requiring knowledge of the transient thermal pattern of the bar. The performance of the method is discussed in terms of effectiveness, uncertainty, and practical implementation.
Assessing the Effectiveness of Sharpness Metrics to Determine the Presence of Contamination on Thermographic Cameras in Harsh Environments
Author: Marche Polytechnic University
This article assesses the effectiveness of sharpness metrics for monitoring the cleanliness of in-line thermographic systems and enabling self-diagnosis, in order to prevent degradation of metrologic performance and increase of measurement uncertainty. When optical measurement systems are installed in harsh industrial environments, external contaminants may compromise their operation conditions. Dust settling on the lens or protective window deteriorates the signal quality, reducing the sharpness of thermal images and making it challenging to extract object features accurately. This study investigates whether monitoring image sharpness can serve as an indicator of contamination presence. Contamination was simulated in controlled laboratory experiments by incrementally adding 0.4 g of dust mixture to the lens of a thermal camera observing high-temperature steel objects. Among the metrics evaluated, Histogram Entropy and the Brenner gradient showed monotonic trends and high sensitivity to small amounts of contamination, with slopes greater than 0.25. The uncertainty of these metrics is less than 0.3. The combined metric, derived through multiple linear regression, improved accuracy, with an R² of 0.96, up from 0.93 for Histogram Entropy and 0.95 for Brenner. Validated with real industrial data, the combined metric proved effective for real-time inline contamination diagnostics in manufacturing environments.
A methodology to assess circular economy strategies for sustainable manufacturing using process eco-efficiency
Author: Laboratory for Manufacturing Systems & Automation (LMS)
This study discusses an extended eco-efficiency indicator targeted at a manufacturing process level, facilitating decision support on the selection of circular economy strategies that a company may adopt to improve its environmental impact. The impact of non-zero-defect manufacturing processes is coupled with the process’s environmental impact and associated cost to derive a new indicator that can support decision-making at a process level regarding the adoption of more sustainable strategies.
A hybrid digital twin approach for proactive quality control in manufacturing
Author: Laboratory for Manufacturing Systems & Automation (LMS)
Quality control is a critical aspect in today’s fast-paced and competitive business landscape. The increasing digital transformation of manufacturing companies allows for the implementation of even proactive quality control strategies. This, however, requires the proper integration and analysis of shop floor data, regarding monitoring, diagnosis, and prognostics. This is to support defect recognition and recuperation, along with potential system reconfiguration based on knowledge extraction and human experience integration. Digital twins, being virtual replicas of physical assets, support real-time monitoring, analysis, and optimization. However, quality-related aspects may not be related to monitored parameters, thus solely data-driven models may not be accurate enough for proactive quality control. In this work, a hybrid digital twin is proposed, where data-driven models are used to finetune the behaviour of the digital twin based on its physics model. A use case concerning an industrial asset and the heat transfer to a steel bar is investigated with the results presented and commented on.
A Zero-Shot Learning Approach for Task Allocation Optimization in Cyber-Physical Systems
Author: Laboratory for Manufacturing Systems & Automation (LMS)
The design and reorganization of Cyber-Physical Systems (CPSs) faces challenges due to the growing number of interconnected devices. To effectively handle disruptions and improve performance, rapid CPS design and development is crucial. The Task Resources Estimator and Allocation Optimizer (TREAO) addresses these challenges, by simulating and optimizing the tasks assignment to the CPS machines, recommending suitable software layouts for the CPS characteristics. It employs Zero-Shot Learning (ZSL) to predict task requirements in heterogeneous devices, enabling the characterization of software pipeline execution in distributed systems. The Genetic Algorithm (GA) component then optimizes the task assignment across available machines. Through experiments, the tool is evaluated for task characterization, CPS modeling and optimization performance. TREAO, when compared with similar tools, allows the simulation of more resource usage metrics (CPU, RAM, processing time and network delay) and increases flexibility in heterogeneous CPSs by predicting the task execution behavior and optimizing the task assignment.
Quality control in manufacturing through temperature profile analysis of metal bars: A steel parts use case
Author: Laboratory for Manufacturing Systems & Automation (LMS)
Non-uniform heating during metal bar hot forming may impact its straightness. In this study, an infrared non-destructive inspection system is proposed to acquire steel temperature profiles in runtime which should correlate to straightness deviations. Additionally, a machine learning algorithm detects outliers to identify oxides on the metal, which in turn is correlated to process parameters. This allows for proactive temperature adjustment to mitigate the risk based on historical profiles. The proposed approach has been tested in a use case coming from the steel industry.
Data Analytics and AI for Quality Assurance in Manufacturing: Challenges and Opportunities
Author: Laboratory for Manufacturing Systems & Automation (LMS)
Data analytics and Artificial Intelligence (AI) have emerged as essential tools in manufacturing over recent years, providing better insight into production systems. Their importance can be highlighted by the way it can transform quality control, from prescriptive to proactive. Data analytics combined with AI can identify abnormal trends and patterns in huge amounts of data, that could uncover potential defects and allow pre-emptive action to minimize or even prevent these from happening. A direct effect of this is the contribution to waste reduction, as well as saving time and resources. While data in a digital factory is ample and the resources for developing artificial intelligence applications are accessible, the implementation of accurate, robust, standard, and economically viable quality monitoring and assessment approaches is not straightforward. This is also strengthened by the scarce skillset in today’s manufacturing companies in this area. In this study, the capabilities and potential of data analytics combined with AI are reviewed with a focus on manufacturing. The implementation challenges posed for a practitioner, as well as the benefits of implementing a solution for a manufacturer using data analytics and AI for quality assessment are discussed, based on real-world experiences from existing production environments. Lastly, a learning approach utilizing a high-fidelity digital twin at its core is presented which a practitioner can utilize to create, test and continuously improve a predictive analytics model.
3D point cloud analysis for surface quality inspection: A steel parts use case
Author: Laboratory for Manufacturing Systems & Automation (LMS)
A manufacturing process includes inspecting the product to verify it meets its quality standards. Such steps, however, are time-consuming and, depending on the means, prone to errors. If not identified in time, defects occurring at an early step of a manufacturing process may result in significant waste, especially if the product is not easy to re-work. Today, however, the combination of AI with computer vision technologies can enable manufacturers to transform quality inspection by automating the detection of defects. This study discusses the use of products’ 3D shape for inline surface defect detection, facilitating the adoption of proactive control strategies facilitating the reduction of waste. The product’s 3D shape, represented by a point cloud is acquired by two fixed laser triangulation sensors orthogonally arranged. The K-means method is adopted for the point cloud data analysis, while Voxel Grid filters are used for downsampling to reduce computational time. The proposed approach has been evaluated in a use case related to the production of steel parts, with the findings supporting that an in-line implementation can facilitate the detection of surface or geometry defects, which, in turn, may facilitate the reduction of waste, by avoiding further processing of the defective product.
A process-level LCA for evaluating the contribution of digitalization in the greening of a manufacturing system
Author: Laboratory for Manufacturing Systems & Automation (LMS)
Artificial Intelligence (AI) can significantly support manufacturing companies in their pursuit of operational excellence, by maintaining efficiency while minimizing defects. However, the complexity of AI solutions often creates a barrier to their practical application. Transparency and user-friendliness should be prioritized to ensure that the insights generated by AI can be effectively applied in real-time decision-making. To bridge this gap and foster a collaborative environment where AI and human expertise collectively drive operational excellence, this paper suggests an AI approach that targets identifying defects in production while providing understandable insights. A semi-supervised convolutional neural network (CNNs) with attention mechanisms and Layer-wise Relevance Propagation (LRP) for explainable active learning is discussed. Predictions but also feedback from human experts are used to dynamically adjust the learning focus, ensuring a continuous improvement cycle in defect detection capabilities. The proposed approach has been tested in a use case related to the manufacturing of batteries. Preliminary results demonstrate substantial improvements in prediction accuracy and operational efficiency, offering a scalable solution for industrial applications aiming at zero defects.
Adapting Vision Transformers for Cross-Product Defect Detection in Manufacturing
Author: Laboratory for Manufacturing Systems & Automation (LMS)
Advanced defect detection solutions that can easily adapt to different products and defect types are of high value for modern manufacturing companies. A significant challenge in developing and deploying such AI models is ensuring they generalize efficiently across diverse visual domains. This challenge is driven by limited data availability of high quality and the substantial effort required for labeling such datasets. This paper explores the adaptation of a Vision Transformer (ViT), originally trained to identify aesthetic defects in battery modules, for application in moulded plastic parts. By using transfer learning and generative AI techniques, this study evaluates fine-tuning and synthetic data augmentation strategies. The proposed approaches are assessed for their potential to enhance model adaptability and reduce dependency on extensive labelled datasets. A case study involving a battery manufacturing company with real-world data serves as the basis for this evaluation. Our preliminary findings suggest promising directions for enhancing the flexibility and efficacy of AI-driven defect detection systems in diverse manufacturing environments.
An explainable active learning approach for enhanced defect detection in manufacturing
Author: Laboratory for Manufacturing Systems & Automation (LMS)
Life Cycle Assessment (LCA) methodology is usually applied to products’ lifecycle and value chains to identify their environmental performance and identify circular economy strategies and options. As a result, different product configurations and strategies may be evaluated and compared, ensuring a positive environmental balance. However, LCA is not usually focused on a manufacturing process level. In this study, LCA is performed at a process level to assess the correlation between digitalization and green manufacturing. The GaBi professional database was used for assessing the manufacturing gate-to-gate processes.
Human-Centric Proactive Quality Control in Industry5.0: The critical role of explainable AI
Author: Laboratory for Manufacturing Systems & Automation (LMS)
The integration of human knowledge and experience with artificial intelligence, especially in the context of Industry5.0, holds the promise of advanced capabilities for manufacturing that may facilitate reduced waste and increased efficiency. However, there is a gap between the two. This work discusses the critical role of Explainable AI (XAI) within this paradigm, fostering a collaborative environment where human operators can leverage AI-driven insights. A framework for data-driven proactive quality control is coupled with XAI and human-centric approaches to enable a path towards zero-defect manufacturing processes, improved operational efficiency, and enhanced workforce empowerment. Furthermore, practical implications, the impact of XAI and recommendations for upskilling and reskilling the manufacturing personnel are discussed with a focus on small and medium-sized enterprises.
Inline-acquired product point clouds for non-destructive testing: A case study of a steel parts manufacturer
Author: Laboratory for Manufacturing Systems & Automation (LMS)
Modern vision-based inspection systems are inherently limited by their two-dimensional 13 nature, particularly when inspecting complex product geometries. These systems are often unable 14 to capture critical depth information, leading to challenges in accurately measuring features such as 15 holes, edges, and surfaces with irregular curvature. To address these shortcomings, this study in-16 troduces an approach that leverages computer-aided design oriented three-dimensional point 17 clouds, captured via a laser-line triangulation sensor mounted on a motorised linear guide. This 18 setup facilitates precise surface scanning, extracting complex geometrical features, which are sub-19 sequently processed through an AI-based analytical component. Dimensional properties, such as 20 radii and inter-feature distances, are computed using a combination of K-nearest neighbours and 21 least-squares circle fitting algorithms. The approach is validated in the context of steel parts manu-22 facturing, where traditional 2D vision-based systems often struggle due to material reflectivity and 23 complex geometries. The system achieves an average accuracy of 95.78% across three different prod-24 uct types, demonstrating robustness and adaptability to varying geometrical configurations. Uncer-25 tainty analysis confirms that measurement deviations remain within acceptable limits, supporting 26 the system’s potential for improving quality control in industrial environments. Thus, the proposed 27 approach may offer a reliable, non-destructive in-line testing solution, with the potential for enhanc-28 ing manufacturing efficiency.
A Zero-Shot Learning Approach for Task Allocation Optimization in Cyber-Physical Systems
Author: University of Porto
A manufacturing process includes inspecting the product to verify it meets its quality standards. Such steps, however, are time-consuming and, depending on the means, prone to errors. If not identified in time, defects occurring at an early step of a manufacturing process may result in significant waste, especially if the product is not easy to re-work. Today, however, the combination of AI with computer vision technologies can enable manufacturers to transform quality inspection by automating the detection of defects. This study discusses the use of products’ 3D shape for inline surface defect detection, facilitating the adoption of proactive control strategies facilitating the reduction of waste. The product’s 3D shape, represented by a point cloud is acquired by two fixed laser triangulation sensors orthogonally arranged. The K-means method is adopted for the point cloud data analysis, while Voxel Grid filters are used for downsampling to reduce computational time. The proposed approach has been evaluated in a use case related to the production of steel parts, with the findings supporting that an in-line implementation can facilitate the detection of surface or geometry defects, which, in turn, may facilitate the reduction of waste, by avoiding further processing of the defective product.
0-DMF - A Decision-Support Framework for Zero Defects Manufacturing
Author: University of Porto
Manufacturing companies are increasingly focused on minimising defects and optimising resource consumption to meet customer demands and sustainability goals. Zero Defect Manufacturing (ZDM) is a widely adopted strategy to systematically reduce defects. However, research on proactive defect-reducing measures remains limited compared to traditional defect detection approaches. This work presents the 0-DMF decision support framework, which employs data-driven techniques for defect reduction through (1) defect prediction, (2) process parameter adjustments to prevent predicted defects, and (3) clarifying prediction factors, providing contextual information about the manufacturing process. For defect prediction, Machine Learning (ML) algorithms, including XGBoost, CatBoost, and Random Forest, were evaluated. For process parameter adjustments, optimisation algorithms such as Powell and Dual Annealing were implemented. To enhance transparency, Explainable Artificial Intelligence (XAI) method s, including SHAP and LIME, were incorporated. Tailored for the melamine-surfaced panels process, the methods showed promising results. The defect prediction model achieved a recall value of 0.97. The optimisation method reduced the average defect probability by 28 percentage points. The integration of XAI enhanced the framework’s reliability. Combined into a unified tool, all tasks delivered fast results, meeting industrial time constraints. These outcomes signify advancements in predictive quality through data-driven approaches for defect prediction and prevention.
HyPredictor - Hybrid Failure Prognosis Approach combining Data-Driven and Knowledge-Based Methods
Author: University of Porto
In modern manufacturing, marked by an unprecedented surge in data generation, utilising this wealth of information to enhance company performance has become essential. Within the industrial landscape, one of the significant challenges is equipment failures, which can result in substantial financial losses and wasted time and resources. This work presents the HyPredictor framework, a comprehensive failure prediction and reporting system designed to enhance the reliability and efficiency of industrial operations by leveraging advanced machine learning techniques and domain knowledge. Six machine learning algorithms were evaluated for failure prediction. The predictions from the algorithms are then refined using rule-based adjustments derived from domain knowledge. Additionally, Explainable Artificial Intelligence (XAI) techniques were incorporated, as well as the capability of users to customise the system with their own rules and submit failure reports, prompting model retraining and continuous improvement. Integrating domain-specific rules improved the performance by up to 28 percentage points in the F1 Score metric in some prediction models, with the best hybrid approach achieving an F1 Score of 90% and a Recall of 92% in failure prediction. This adaptive, hybrid approach improves prediction accuracy and fosters proactive maintenance, significantly reducing downtime and operational costs.
Data-Driven Predictive Maintenance for Component Life-Cycle Extension
Author: University of Porto
In the era of Industry 4.0, predictive maintenance is crucial for optimizing operational efficiency and reducing downtime. Traditional maintenance strategies often cost more and are less reliable, making advanced predictive models appealing. This paper assesses the effectiveness of different survival analysis models, such as Cox Proportional Hazards, Random Survival Forests (RSF), Gradient Boosting Survival Analysis (GBSA), and Survival Support Vector Machines (FS-SVM), in predicting equipment failures. The models were tested on datasets from Gorenje and Microsoft Azure, achieving C-index values on test data such as 0.792 on the Cox Model, 0.601 using RSF, 0.579 using the GBSA model and 0.514 when using the FS-SVM model. These results demonstrate the models’ potential for accurate failure prediction, with FS-SVM showing significant improvement in test data compared to its training performance. This study provides a comprehensive evaluation of survival analysis methods in an industria l context and develops a user-friendly dashboard for real-time maintenance decision-making. Integrating survival analysis into Industry 4.0 frameworks can significantly enhance predictive maintenance strategies, paving the way for more efficient and reliable industrial operations.
Survival Analysis-Based System for Predictive Maintenance Optimization
Author: University of Porto
The adoption of predictive maintenance (PdM) in Industry 4.0 has become essential for optimizing operational efficiency and reducing downtime. Traditional maintenance approaches, such as reactive and preventive maintenance, often result in inefficiencies, highlighting the need for more proactive and data-driven strategies. This paper presents a modular PdM framework that uses survival analysis methods to estimate the remaining useful life (RUL) of industrial components and integrates a Simulated Annealing (SA)-based optimization algorithm to schedule maintenance interventions dynamically. Four survival models—Cox Proportional Hazards (Cox PH), Random Survival Forests (RSF), Gradient Boosting Survival Analysis (GBSA), and Survival Support Vector Machines (Survival SVM)—were evaluated to predict components’ failure, with GBSA emerging as the most robust model due to its resistance to overfitting and ability to capture non-linear degradation patterns. The maintenance scheduling optimization algorithm minimizes downtime by aligning maintenance windows with non-linear degradation/survival functions and production requirements, achieving a 25% reduction in downtime compared to non-optimized methods. The framework was validated on real-world welding electrode degradation data and a synthetic Microsoft Azure PdM dataset, demonstrating its adaptability to heterogeneous industrial environments. Unifying survival analysis with production-aware scheduling optimization enabled cost-effective, risk-informed decision-making for Industry 4.0 applications.
Design of an ISO 23247 Compliant Digital Twin for an Automotive Assembly Line
Author: Polytechnic Institute of Bragança
The integration of Industry 4.0 (I4.0) technologies is transforming various society sectors, and particularly the manufacturing sector, promoting a digital transformation that fosters smart and interconnected systems. The Digital Twin (DT) acts as a key component in the digital transformation landscape and its integration with 14.0 technologies paves the way for the implementation of Zero Defect Manufacturing strategies. A spotlight on the automotive industry underscores the power of DT applications for real-time defect detection and improvement of product quality and process efficiency. However, the proper DT implementation, contributing to their integration and interoperability, requires the compliance with standards and reference architectures. Having this in mind, this paper describes the design of a DT architecture compliant with the ISO 23247 standard and aligned with the RAMI 4.0 to cover the dimensional measurement process comprising inspection stations placed in the body shop area of an automotive assembly line. This implementation enables the real-time and early identification of defects along the assembly process, allowing to prevent their occurrence at a single stage and their propagation to downstream processes.
Positioning Cyber-Physical Systems and Digital Twins in Industry 4.0
Author: Polytechnic Institute of Bragança
Industry 4.0 has brought innovative concepts and technologies that have greatly improved the development of more intelligent, flexible and reconfigurable systems. Two of these concepts, Cyber-Physical Systems (CPSs) and Digital Twins (DTs), have gained significant attention from various stakeholders, e.g., researchers, industry practitioners, and governmental organizations. Both are vital to support the digitalisation of products, machines, and systems, and they focus on the integration of physical and cyber processes, where one affects the other through feedback loops. Having this in mind, this paper aims to better understand how CPS and DT are correlated, particularly exploring their similarities and differences, their positioning within the Industry 4.0 paradigm, and their convergence to develop Industry 4.0 solutions. Some research challenges to develop Industry 4.0 solutions by integrating these concepts are also discussed.
Exploring Digital Twin Dynamics: An Analysis of Structure Configurations
Author: Polytechnic Institute of Bragança
Digital twin (DT) is an important technology to support the realization of the digital transformation, connecting the physical asset to its virtual copy and fostering realtime monitoring, simulation, and decision-support. However, the benefits of DT depend on its intended purpose and application, which is impacted by the design structure configuration that is used for its implementation. This paper discusses the advantages and challenges of considering different organizational structures for the implementation of DTs, namely centralised, hierarchical and decentralised, complemented with a case study that was used to analyse the implementation of DT focusing on centralised and decentralised approaches. Additionally, the paper includes an analysis of the main aspects of DT structures and their design guidelines, the key enabling technologies and the main challenges of distributing DTs.
Digitalization of Industrial Inspection Assets through the Asset Administration Shell
Author: Polytechnic Institute of Bragança
Developments in industrial manufacturing systems, through the use of new digital technologies, are revolutionizing the traditional means by which companies operate in terms of productivity and competitiveness. The adoption of technologies related to Industry 4.0 (I4.0) has led to the possibility of digitizing industrial assets to achieve more responsive, reconfigurable and efficient production systems. The Asset Administration Shell (AAS) corresponds to a form of digital representation of a physical asset, such as, a machine or device, that describes its properties, functionalities, and behaviours, as well as guarantee interoperability between different systems. Implementing AAS in the industrial environment allows the creation of a standard for digitalising asset scenarios, which can be used for intelligent and non-intelligent products. This paper explores modelling geometric inspection assets on an automotive assembly line using AAS standards. Once modelled, the AAS of the assets are applied to a new approach for reactive AAS based on a REST API for real-time data transmission to a framework focused on Zero Defects Manufacturing. The implications associated with the asset modelling and the creation of the AAS server are presented and discussed, giving a critical opinion on the current status of the technologies associated with AAS in I4.0.
An Augmented Reality Intelligent Guide for the Automotive Industry
Author: Polytechnic Institute of Bragança
Throughout the 21st century, there has been a rise in interest of increasing inter-connectivity and smart automation in the realms of industrial production, often called the 4th Industrial Revolution. Thus, interest in areas such as virtual, mixed, and augmented reality has increased as new devices and technologies related to these areas are seen as a possible solution to increase Industrial efficiency, create safer work environments for employees, and more effective training. In this project, a HoloLens app was developed, capable of identifying and showing to the user various zones where a specific vehicle in a production line requires checking and in which users have full spatial perception. Each vehicle is composed of several zones and each of these zones is associated with specific stations in a sequential order, so the user will only be able to see the zones associated with the station that is being treated. Using a specific “gesture” users can change the zone’s status in the database to indicate that the zone has been checked. These changes will be visible and the vehicle, stations, and zones will be updated to reflect the modifications. Various scripts in C# and PHP were used to allow modifications to the behaviour of the objects and database in the augmented reality scene through the access to a RESTful service in Unity.
Relationship of Digital Product Passport and Digital Twin in Industry 4.0 Context
Author: Polytechnic Institute of Bragança
The advances brought by Industry 4.0 contribute to the digital transformation and green transition, accelerating advances toward a circular economy. In this context, some interconnected concepts share similarities and complementarities, that can be analyzed to understand the potential benefits of their associated use. This paper discusses the relationship between the Digital Product Passport (DPP) and the Digital Twin concepts, highlighting their alignment and exploring their similarities and differences, aiming to understand how they complement each other. Therefore, the typology of assets, data model, and functionalities were analyzed, as well as an architectural and component alignment based on the elements provided by ISO 23247 while considering that the DPP concept is still in development. A case study is also presented to illustrate the implementation of both concepts.
Real-Time Rule-based Monitoring Tool to Achieve Zero Defect Manufacturing
Author: Polytechnic Institute of Bragança
The demands of innovative production systems are shifting from mass production to the creation of smaller quantities with a focus on high quality. To achieve these evolving demands, Zero Defect Manufacturing has emerged as a key paradigm. This approach requires an innovative architectural monitoring tool where real-time data is continuously gathered and analysed to predict defects and assess their potential impacts. It also necessitates the seamless integration of diverse data sources, advanced processing algorithms, and Digital Twins to align with industrial requirements. In this paper we present a real-time, rule-based monitoring tool applied to a real-world car manufacturing use case. The tool successfully generated early alerts for quality deviations, enabling production engineers to shift from a reactive to a proactive approach by detecting potential quality issues early in the process.
Industrial Metaverse Digital Twin: ISO 23247 Compliant Architecture for AI-Driven Simulation
Author: Polytechnic Institute of Bragança
The Industrial Metaverse marks a new stage in Industry 4.0, raising the representation level of Digital Twins (DT) from discrete elements to an interconnected network of assets covering the entire production ecosystem. This paradigm changing reflects advances in enabling technologies such as Artificial Intelligence (AI) and complex what-if simulations. As complexity increases, adopting established industrial standards for implementing DT functionalities becomes imperative to specify guidelines for companies and researchers. This paper proposes a functional architecture for DT implementation in compliance with ISO 23247 standard, aiming to support the development of interoperable and standardized solutions combined within the Industrial Metaverse. The architecture was employed to develop a DT framework for an automotive assembly line, covering the quality inspection process and embedding AI-based mechanisms to leverage the what-if simulation of deviations in structural parameters of the vehicle’s body. Experimental results demonstrate the tool’s ability to accurately predict outputs for critical quality parameters according to hypothetical measurement scenarios, leveraging the production stakeholders’ understanding regarding the correlated impact of deviations at different structural points, and demonstrating the versatility and potential of combining AI strategies for what-if simulations.
Exploring Automotive Quality Correlations through Explainable Machine Learning What-If Simulation
Author: Polytechnic Institute of Bragança
High-dimensional variability in manufacturing processes presents significant challenges for quality control, demanding predictive strategies capable of capturing complex parameter dependencies. Machine learning (ML) offers robust mechanisms for this purpose, but reliance on black-box models often limits interpretability and hinders producing stakeholders’ identification of meaningful correlations for model optimization. This paper introduces an interactive what-if simulation platform designed to explore structural quality correlations in automotive assembly through explainable ML techniques, enhancing transparency and enabling uncertainty quantification. The platform is based on a modular Digital Twin (DT) architecture aligned with the ISO 23247 standard, guiding expert and non-expert users through correlation-driven feature selection, regression modelling and SHapley Additive exPlanations (SHAP) based post-hoc explanations. A case study using real inspection data from a vehicle assembly line demonstrates the tool’s capacity to support variable relevance assessment, dimensionality reduction, and model interpretability. Furthermore, an uncertainty-aware SHAP analysis enhances confidence in the model’s prediction stability, reinforcing the platform’s suitability for quality-driven decision support and integration into future DT ecosystems.
Microservices based Artificial Intelligence Diagnosis System for Zero Defect Manufacturing
Author: Polytechnic Institute of Bragança
As manufacturing evolves towards high-quality and high-output production, the demand for intelligent, defect-free systems becomes increasingly critical. Zero Defect Manufacturing (ZDM) emerges as a key paradigm, relying on real-time data and predictive analytics to proactively detect and mitigate quality deviations. In this work, a AI-based diagnosis system is proposed as part of an integrated approach to transform quality monitoring from a reactive to a proactive strategy by enabling the early detection of potential defects. This approach considers representing algorithms for predicting product and measurement quality based on live shop floor data, and optimization algorithms for fine-tuning the regression models aiming to enhance the prediction performance. The proposed approach was applied to an automotive assembly line, and the preliminary results shown benefits in improving the results of the diagnosis process.