Geometric dimensioning, tolerancing and inspection

Silicon Carbide (SiC) power devices are highly sought in modern power electronics for their superior performance; however, yield is severely impacted by crystal-growth defects often termed killer defects. Fast and accurate post-growth inspection of SiC wafers is therefore critical to identify such defects and improve manufacturing outcomes. We propose an improved YOLO-based one-stage detector that integrates a Non-Local Block + Gradient Reversal Layer (GRL) module into the YOLO backbone between the Spatial Pyramid Pooling (SPP) and BottleneckCSP layers. The Non-Local blocks enable the CNN to leverage global context in the microscopic images, improving its ability to recognize subtle defect patterns across the wafer. Meanwhile, the GRL-based domain adaptation allows the model to learn domain-invariant features, addressing distribution shifts between different datasets or imaging conditions. The proposed architecture, comprising these components, is evaluated on a dataset of SiC wafer optical micrographs containing various defect types to demonstrate its effectiveness (e.g., dislocations and stacking faults). The proposed method achieves mAP@0.5=95% and mAP@0.5:0.95=94.5%. These improvements demonstrate the efficacy of combining attention and adversarial domain adaptation in the YOLO architecture. The enhanced model detects SiC defects with higher accuracy and robustness to varying imaging conditions, marking a significant step forward in automated wafer defect inspection.

Keywords: Deep Learning , Defects analysis , Non-local block , Silicon-Carbide

Deploying deep neural networks on embedded systems is challenging due to stringent constraints on memory, compute, and power. DISTILLO (aDvanced Interactive ST knowledge distILLation) is a knowledge distillation framework tailored for classification tasks in resource-limited environments. It features two strategies: DISTILLO-Classical, using dynamic loss weighting with Kullback-Leibler or Jensen-Shannon divergence, and DISTILLO-Ensemble, which combines multiple student models distilled at different temperatures. Jensen-Shannon divergence is adopted to provide symmetric, bounded, and stable gradients, improving the robustness of the training. Experiments using Vision Transformers as teachers and lightweight CNNs as students show that DISTILLO-Classical achieves up to +5.37% accuracy gain on CIFAR-100 and +4.42% on Imagenette. DISTILLO-Ensemble further improves performance, reaching +13.97% and +8.32% gains, respectively, with minimal memory and latency overhead. DISTILLO enables the effective deployment of deep models on embedded platforms such as STM32 microcontrollers.

Keywords: embedded systems , knowledge distillation , low-resources

Automotive seat comfort is a critical factor in enhancing driver satisfaction, especially in sports cars, where design must balance comfort features and performance-oriented features like lateral containment and anti-submarining. This study adopts an empirical-analytical approach for assessing and modelling perceived comfort in sports car seats using both objective and subjective data. A total of 64 participants (50 males, 14 females) evaluated two types of sports car seats—a road model (SEAT A) and a racing model (SEAT B)—during 15-min driving simulations using a dynamic simulator equipped with a full-body pressure mat (XSENSOR X3 PRO). Comfort was assessed through a postural comfort questionnaire using 10-point Likert scales. Statistical analysis revealed significant correlations between anthropometry, pressure distribution, and perceived comfort. In light of the correlation analysis, regression models were developed for four anthropometric percentile clusters (0–25th, 25–50th, 50–75th, 75–100th). Models were validated (accuracy > 75%) and one of them (named Model III) achieved accuracies of 95%, 96%, 90%, and 97% for its percentile clusters. The proposed models offer actionable insights for tailoring sports car seats to different user percentiles, enabling more personalized and effective seat designs that enhance both performance and comfort.

Keywords: comfort assessment method , human seat interaction , pressure distribution , sports car seats

Real-time rendering is increasingly used in augmented and virtual reality (AR/VR), interactive design, and product visualisation, where materials must prioritise efficiency and consistency rather than the extreme accuracy required in offline rendering. In parallel, the growing demand for personalised and customised products has created a need for digital materials that can be generated in-house without relying on expensive commercial systems. To address these requirements, this paper presents a low-cost digitisation workflow based on photometric stereo. The system integrates a custom-built scanner with cross-polarised illumination, automated multi-light image acquisition, a dual-stage colour calibration process, and a node-based reconstruction pipeline that produces albedo and normal maps. A reproducible evaluation methodology is also introduced, combining perceptual colour-difference analysis using the CIEDE2000 ((Formula presented.)) metric with angular-error assessment of normal maps on known-geometry samples. By openly providing the workflow, bill of materials, and implementation details, this work delivers a practical and replicable solution for reliable material capture in real-time rendering and product customisation scenarios.

Keywords: digitalisation workflow , material scanner , photometric stereo , real-time rendering , texture evaluation

This study introduces a numerical methodology for computing the statistical assembly shift in patterns of fits, addressing scenarios with variable numbers of elements and scrap caused by assembly failure. Using Monte Carlo simulations, the methodology estimates rejection rates, determines gap distributions, and calculates assembly shifts while considering both intrinsic and external datum systems. The findings indicate that adding more elements to a pattern reduces assembly shifts exponentially, presenting a design opportunity to control alignment and optimize component performance. A case study involving engine block and cylinder head alignment demonstrates the methodology's applicability to real-world mechanical design. Three approaches for tolerance stack-up are evaluated: the standard Root Sum Square (RSS) method, where assembly shift is treated as a worst-case scenario; a proposed RSS method that models the assembly shift as a Gaussian distribution with a standard deviation derived from Monte Carlo simulations; and the Monte Carlo approach, which considers the full shape of the assembly shift distribution. By comparing these approaches, the study underscores the effectiveness of the proposed methodology in capturing the statistical behavior of assembly shifts. This work contributes a robust tool for tolerance analysis, advancing the precision and accuracy of pattern fit modeling and assembly shift evaluation in mechanical design.

Keywords: Assembly shift , Boundary condition , Clearance fits , Monte carlo simulation , Rejection rate estimation , Tolerance analysis , Transition fits

Metal binder jetting (MBJ) is an additive manufacturing technology of increasing interest due to its potential competitiveness in medium- and large-scale production, especially from a sustainability perspective. However, challenges in controlling the product accuracy and precision significantly limit the widespread adoption of this technology. This work investigates the achievable accuracy, precision, and spatial repeatability of parts produced using the MBJ process. Additionally, the paper aims to identify the causes of inaccuracy and suggest countermeasures to improve the product quality. The study was conducted experimentally by designing a benchmark geometry with various basic features. This geometry was scaled to three sizes—10–20 mm (small), 20–30 mm (intermediate), and 30–50 mm (large)—and produced using two different stainless-steel powders: AISI 316L and 17-4PH. In the green state, the dimensional tolerances ranged from IT8 to IT12 for features parallel to the build direction (heights) and from IT9 to IT13 for features parallel to the build plane (lengths). In the sintered state, the tolerances ranged from IT10 to IT16. This study reveals the challenges in scaling geometries to compensate for accuracy loss originating from the printing and sintering stages. In the green state, accuracy issues are likely due to non-uniform binder application and drying operations. In the sintered state, the accuracy loss is related to variable shrinkage based on the feature size, anisotropic shrinkage depending on the print direction, and differing densification mechanisms influenced by the material type. This study offers novel insights for improving MBJ process precision, supporting wider adoption in the manufacturing industry.

Keywords: additive manufacturing , dimensional accuracy and precision , metal binder jetting , shrinkage on sintering

Metal Binder Jetting (MBJ) is an additive manufacturing (AM) technology increasingly gaining interest due to its capability of producing complex products in medium-large batches, cost-effectively compared to other AM technologies. Nevertheless, dimensional accuracy is still challenging, also considering the significantly anisotropic shrinkage on sintering. The influence of printing parameters on the anisotropic shrinkage of two different geometries is studied in this work. Using fractional factorial Design of Experiments (DOE), four factors at three levels were considered (printhead speed, binder saturation, rake speed, and shell thickness). 30 cubes with squared through-hole, and 30 cylinders with circular through-hole were produced for each experiment using 316L; the shrinkage along different printing directions was derived from the dimensions measured at green and sintered state by Coordinate Measuring Machine. The optimal combination of printing parameters for the different geometries was identified using ANOVA. The effect of printing parameters on the two geometries was highlighted and discussed.

Background: – We aimed to describe a protocol and related surgical treatment for mentoplasty using virtual surgical planning (VSP) and piezosurgery. Methods: – Twelve patients requiring mentoplasty underwent VSP based on preoperative computed tomography. Virtual osteotomy was defined and the bone fragment was repositioned. Cutting and repositioning guides were exported as STL files and created using a 3-dimensional (3D) printer with biocompatible material. Plates were preblended on the printed model. Postoperatively, morphometric analysis was performed by comparing the planned and postoperative 3D models: distance between the apex of the cusp of 43 and 33 and lower edge of the chin; distance between the lower edge of the right and left mental foramen and osteotomy; and distance between the mandibular lingual fossa and pogonion. Results: – Six cutting guides were made composite and six were monobloc. The repositioning guides were all made monobloc. No adverse intraoperative events were observed. Postoperative transient hypoesthesia was reported in 1 case. The average distance between the planned and postoperative 3D models of the chin was recorded in translation (1.91 mm), rotation (right, 1.38 mm; left, 1.30 mm), and cranial-caudal repositioning (right, 1.39 mm; left, 1.23 mm). VSP and 3D printing have improved surgical precision and reproducibility; they help the surgeon decide the extent of chin movement and the most suitable type of surgical guide in advance. Conclusions: – Mentoplasty using 3D-printed guides and piezosurgery is simple, accurate, and safe. The costs of guide production are reduced using 3D printers by approximately €400–€800 per case compared with industrial printers.

The accurate motion of roller-bearing-supported rings is critically influenced by shape and positional tolerances, which are often underestimated in conventional modeling approaches. The aim of this study is to develop and validate a multibody dynamic framework capable of quantifying the impact of roundness and positional errors on the motion accuracy of roller-bearing-supported rings. Shape errors are modeled using Fourier series and incorporated into a high-fidelity multibody simulation environment. Experimental validation using laser triangulation reveals a maximum runout error of 72.9 μm, compared to a numerically predicted value of 88.6 μm, resulting in a quantified numerical overestimation of 21.5%. Parametric studies investigated the effects of harmonic order, error amplitude, and combined error scenarios on key performance metrics, including trajectory runout and initial offset displacement. Results reveal that the trajectory errors range between 0.29 mm and 0.63 mm for shape error orders and can escalate to 2.84 mm for high amplitude errors, demonstrating the critical role of error order and amplitude. Furthermore, combined simulations show that bearing position errors exert a more pronounced effect on radial accuracy than shape deviations alone. The proposed approach enables precision design evaluation and tolerance optimization in high-accuracy applications, including robotics, aerospace mechanisms, and optical alignment systems.

Keywords: Fourier series , manipulation , motion accuracy , multibody dynamics , positional error , robotics , roller bearings , roundness error , runout error , tolerance analysis

This study presents a comparative analysis of two portable 3D scanning technologies, structured light and LiDAR, applied to the reverse engineering of fiberglass boat molds. Key aspects such as acquisition speed, data completeness, resolution, feature detectability, and operational usability were evaluated. Due to the absence of a CAD reference, a relative geometric deviation analysis between the overlapping regions of the scans was performed. Results indicate that while the LiDAR-based scanner offers rapid and comprehensive coverage suitable for large-scale geometry capture, the structured light scanner provides higher resolution in localized areas but requires more careful setup and markers for effective scanning. The complementary nature of these technologies suggests potential advantages in hybrid scanning strategies for marine manufacturing applications.

Keywords: 3D scans , LiDAR scanning , quality control , reverse engineering , structured light scanning

Industrial Robots (IR) are currently employed in several production areas as they enable flexible automation and high productivity on a wide range of operations. The IR low positioning performance, however, has limited their use in high precision applications, namely where positioning errors assume importance for the process and directly affect the quality of the final products. Common approaches to increase the IR accuracy rely on empirical relations which are valid for a single IR model. Also, existing works show no uniformity regarding the experimental procedures followed during the IR performance assessment and identification phases. With the aim to overcome these restrictions and further extend the IR usability, this paper presents a general method for the evaluation of IR pose and path accuracy, primarily focusing on instrumentation and testing procedures. After a detailed description of the experimental campaign carried out on a KUKA KR210 R2700 Prime robot under different operating conditions (speed, payload and temperature state), a novel online compensation approach is presented and validated. The position corrections are processed with an industrial PC by means of a purposely developed application which receives as input the position feedback from a laser tracker. Experiments conducted on straight paths confirmed the validity of the proposed approach, which allows remarkable reductions (in the order of 90%) of the orthogonal deviations and in-line errors during the robot movements.

Keywords: Error compensation , Experimental approaches , Industrial robots , Laser tracker , Path accuracy , Pose accuracy

At Legnaro National Laboratories of the Italian National Institute for Nuclear Physics, Design for Assembly principles were applied to ease and improve the assembly procedure of the Forced Electron Beam Induced Arc Discharge ion source for the SPES Isotope Separation On-Line facility. Such device is a key component for the whole system, as its correct functioning is fundamental to provide the radioactive ion beam to the experimental users; furthermore, its reliability and performance stability are directly affected by the construction and placement accuracy of the parts that compose the assembly. A deep critical analysis of the current assembly procedure was performed to identify the main issues that affect the mounting accuracy of the ion source components. Consequently, a new assembly procedure was developed to fix the identified issues without applying relevant modification to the ion source parts; specifically, a custom assembly tool was designed not only for an easier coupling of the components, but also to improve the accuracy of the functional surfaces positioning to increase the device performance stability during operation. To achieve this, a custom assembly tool was designed to reduce the number of the elements of the tolerance stackup and, therefore, the uncertainty propagation. The proposed assembly procedure results significatively less time consuming than the current one, moreover it improves the accuracy of the FEBIAD ion source components position once assembled, as confirmed by Coordinate Measuring Machine measures.

Keywords: assembly tool , Design for assembly , Geometrical Product Specifications , Tolerance analysis

The Geometrical Product Specification (GPS) and the Geometrical Dimensioning and Tolerancing (GD&T) are communication languages to code the tolerable morphology of manufactured parts and assemblies. Both languages should be unambiguous tools to communicate such information between designers, process engineers, and Coordinate Measuring Machines operators (CMM). GPS is the one developed in the ISO (International Organization for Standardization) environment. GPS language is a complex code of 143 standard with further 15 under development. Moreover, as in each complex body of standards, most of them recalls other standards in a very intricate manner. So, the need to have a flexible tool to search and navigate through the standards is great, as is the need to optimize the work of the designer and to minimize the design, production, and control costs. The basic effort in building such a tool has been the development of the database and the structure for the search engine, called “GPS Navigator”. In the following, the requirements for the coding phase have also been issued, to realize a powerful, efficient, fast, robust, and rigorous tool to navigate through the GPS standards. The final step of the “GPS-Navigator project” is the delivery of a software tool able to help and guide the designer to quickly consult the appropriate standard or set of standards.

Keywords: Advanced search criteria , Database , Design methods , Geometrical product specifications (GPS)

Compliant mechanisms (CMs), along with soft robotics devices formed therewith, may be defined as engineering systems achieving force and motion transmission via the deflection of flexible members. CMs have increasingly gained a strong foothold in the scientific arena owing to their hinge-less nature, shock resistance, potential single-piece manufacturability, safety in human–machine interaction, minimal maintenance requirements, and adaptability to work in unstructured environments. In parallel, current advances in the production of inherently compliant sensory-motor apparatus, as well as progresses in the development of robust control methods, are paving the way to practical CM adoption in a large variety of engineering fields, here including healthcare, manufacturing, inspection/maintenance, and agrifood.

Keywords: Flexible manufacturing systems , Human computer interaction , Manufacturing processes , Mechatronics , Soft robotics , Special issues and sections

The Laser Powder Bed Fusion (LPBF) is an Additive Manufacturing (AM) technology suitable to produce almost free-form metallic components. At Legnaro National Laboratories (LNL) of the Italian National Institute for Nuclear Physics (INFN), the LPBF process was recently used to produce parts of the Forced Electron Beam Induced Arc Discharge (FEBIAD) ion source for the SPES Isotope Separation On-Line (ISOL) facility. In this work are presented the feasibility assessment and production steps of tantalum cathodes produced via AM; in addition, the results concerning both the dimensional–geometrical measurements and the preliminary high-temperature test are reported.

The Fourth Industrial Revolution conceptualizes the rapid change of industries resulting from the convergence of technologies such as artificial intelligence, genetic editing, and advanced robotics that enable increasing interconnectivity and machines that can analyze and diagnosing problems without human intervention, through intelligent automation. In this scenario, the use of augmented reality technologies is of great interest. The paper aims to explore the use of augmented reality in support of traditional inspections for assisting textile experts in fabric defect detection. The contribution of this study consists of three main phases, necessary for the future development of the system: (1) the analysis of possible automatic defect detection techniques; (2) the analysis of hardware solutions for the realization of a system based on important criteria such as operator comfort, system footprint, and so on; (3) the proposal of a possible comprehensive solution. Considering these aspects this paper identifies and investigate the best scenario for the introduction of artificial intelligence and augmented reality technologies to help the operator in the detection of textile defects.

Keywords: Artificial intelligence , augmented reality , machine vision , textile defect detection

The objective of this study was to develop a Computational Fluid Dynamics (CFD) 3D model for an industrial application, to investigate how surface roughness affected pressure drop and thermal performance in the ACEPACKTM DRIVE, a commercial SiC-based power module used in traction inverters, that features a pin-finned baseplate with cylindrical pin-fins. The surface roughness of the pin-fins was modelled by using a sand-grain model available in Ansys Fluent, which represents a novelty in the literature and can pave the way for the development of accurate and computationally efficient simulations. The CFD model exhibits an error from experimental data within the range 10%–22% when the flow rate varies from 2 l/min to 12 l/min (maximum error at 2 l/min), while the validation of the thermal behaviour was obtained with a maximum error of 2.5%. Subsequently, a parametric analysis was carried out, varying the flow rate of the coolant, in order to investigate the influence of pin-fins surface roughness on performance, by considering two possible design option featuring respectively a surface roughness equal to 1 μm and 6 μm. The results showed a reduction in pressure drop from 4% to 11% and a degradation of the thermal performance, less than the 2%, when the pin-fins roughness increases.

Keywords: CFD , Pin-fins , Power module , Surface roughness

Additive Manufacturing is widely applied in aerospace, automotive and marine engineering. Indeed, large-scale components are often required in these applications, such as for non-structural parts of aircraft, spare parts or small lots of cars or marine components. Fused Deposition Modelling is one of the Additive Manufacturing processes used to affordably convert digital models into mockups, prototypes, and functional parts: a slicing software converts the object's digital model into a list of instructions for the machine. However, commercial slicing software packages often fail to accurately estimate the time required to produce models, especially when their size is significant: the errors could be up to several hours, which cannot be adequate in a real-life industrial context where production must be scheduled in a precise way. This manuscript compares the build time estimation of several commercial slicing software considering a real-life part. Furthermore, the evaluation of the manufacturing setting mainly affects the error in estimating the build time achieved through a Design of Experiment approach. The more time-impacting printing parameters have been detected, allowing fine helpful tuning to increase the accuracy of the build time in commercial slicing software. A case study included in the manuscript supports the analyses. Proper setting of the commercial slicing software can significantly improve the accuracy of the printing time.

Keywords: Additive Manufacturing , Design of Experiment , Fused Deposition Modelling , Slicing , Time Estimation

Virtual assembly has become a popular trend in recent years and is used for various purposes, including selective assembly and adaptive tooling. Monte Carlo approaches based on Finite Element Method (FEM) simulations are commonly used for production applications. However, during the design phase, when testing different configurations and design options, a variational method is more suitable. This paper aims to test different implementations of the Method of System Moments applied to the second-order tolerance analysis method when actual distributions, which are non-centered and non-normal, are used as input for the simulation. The study reveals that the simulation results can significantly vary depending on the simulation settings in some cases. As a result, a series of best practices are highlighted to improve the accuracy and reliability of the simulation outcomes.

Keywords: Assembly simulation , Computer-aided tolerancing , Tolerance stack-up , Virtual assembly

In tolerancing activities focusing on the allocation of geometrical tolerances, many critical issues originate from the non-optimal assignment of responsibilities among the organization units involved. This paper aims to depict relations between different tolerancing activities and relevant specifications, assigning them to the proper actor and, therefore, expanding the ISO 8015:2011 “responsibility principle”. A classification among tolerancing activities, specifications, and media is proposed; a horizontal hierarchical framework among functional, manufacturing, and verification specifications and a vertical hierarchical framework along the supply chain are discussed. Examples of both hierarchical structures are presented.

Keywords: functional specification , functional tolerancing , geometrical product specification , ISO/TS 21619:2018 , manufacturing specification , manufacturing tolerancing , tolerance specification , verification specification , verification tolerancing