Technical Papers & Journals

Advanced MBE methods that inform Bidirectional Communication between Manufacturing and Quality Disciplines

Tom Groff

Manufacturing and Quality disciplines consume data and publish data to the digital thread.  What data should be included for optimizing the various interactions between Manufacturing and Quality?  Learn how Manufacturing can benefit from Quality data, and how Quality can benefit from Manufacturing data.  When you know Manufacturing operation details from CAM and inspection details and results from MBE,   the inputs and feedback to machines and software can be more comprehensive and correct.  This yields higher throughput, better quality and optimized costs.  We will discuss the following tools and processes for maximizing on the Manufacturing and Quality digital thread data that is available through Model Based Enterprise (MBE) practices.

• Simulating Manufacturing Data in the Design Phase to inform proper  GD&T
• Using Manufacturing Datum and Setup Information from CAM in inspection
• When to use STEP vs QIF for Digital Thread Standardization
• Determining Feedback from Part Inspection to Manufacturing Tools to Control Processes
  • Traditional Tolerancing vs GD&T
  • Single Part  vs MultiPart SPC
• Iterating on  Manufacturing Setup Operational possibilities  to yield optimized functional requirements and GD&T characteristic results.
• Quality input differences between Subtractive and Additive Manufacturing
• Using  Quality data in Casting operations.

Stability Assessment of Motion Capture Systems

Jorian Khan

Motion capture technology is extensively used across biomechanics, robotics, and the arts and entertainment industries, that rely on the systems' ability to track the positions of multiple targets with low latency in large volumes. These abilities suggest that motion capture is a promising tool for non-contact, high-speed, in-situ measurements in manufacturing environments. However, the utilization of motion capture in support of manufacturing necessitates the development of meaningful performance metrics to evaluate the system's measurement accuracy and repeatability. Moreover, the growing variety of these systems creates a need for standardized system evaluation to allow the comparison of performance across different motion capture systems and other measuring technologies. This investigation describes the development of specific tests aimed at quantifying the stability of motion capture systems utilizing data obtained during typical motion capture calibration procedures. The proposed testing methodology yields an aggregate performance metric derived from a geometric interpretation of extrinsic and intrinsic camera parameters. The analysis of this metric includes an assessment of its sensitivity to variability in camera parameters, based on both simulations and experimental data. Results demonstrate that the proposed metric provides insights into the variability of target indications, offers a practical means to assess system stability, and supports the diagnosis of instability. These findings support the promise of the proposed methodology serving as a practical tool for evaluating motion capture systems in manufacturing environments. Ultimately, use of this methodology may serve as a basis to allow industrial users to determine the "fitness for use" of a motion capture system for a particular dimensional metrology task. 

Metrology Applications of Industrial X-Ray Computed Tomography (CT) in Medical Device Manufacturing

Krishnakumar Gopal, Tony Calduch

This paper presents practical examples of innovative applications of X-ray computed tomography (CT) for dimensional inspection and product acceptance in medical device manufacturing. CT technology significantly enhances throughput and reduces inspection time. CT inspection eliminates operator variation, as demonstrated by the measurement system analysis. Unlike traditional CMM inspection, which requires the operator to correctly position the part on a fixture and teach the CMM its location, CT inspection removes the need for manual alignment and rigid fixturing. The paper highlights various practical examples of how CT scanner measurement technology was utilized to inspect manufactured components, thereby reducing inspection time and increasing throughput. Currently, parametric coded measurement (PCM) plans are employed in Zeiss Calypso CMM programs. Python scripting language enabled PCM programming capabilities for CT scanner programming. The examples include using CT metrology to check internal dimensions that traditional coordinate measuring machines cannot measure and combining dimensional and material quality control in a single measurement inspection run. The measurement cycle time was significantly reduced due to the ability to inspect multiple parts simultaneously, unlike traditional CMM measurements, without adding additional measurement time.

User-centered training for optical measurements in digitalized manufacturing processes

Wiktor Harmatys

The TOPMEAS project aims to develop curricula and educational materials that focus on student learning outcomes while addressing skills mismatches and promoting entrepreneurship. By ensuring alignment with labor market and societal needs, the project introduces flexible learning paths and modular course schemes in optical metrology for digital production management. It integrates transdisciplinary approaches and innovative pedagogical methods, such as online international collaborative learning and blended curricula, fostering transferable skills in digital technologies and entrepreneurship. Additionally, the project supports digital transformation in higher and adult education by enhancing digital competence in optical metrology and equipping educators with expertise in digital tools, accessibility technologies, and digital content creation.

A core element of the project is the development of innovative learning materials for coordinate metrology, with a particular focus on optical measurement techniques. These materials will incorporate blended learning methodologies, combining e-learning, face-to-face lessons, and workshops to enhance accessibility and engagement. Designed to be student-centered, the content will be both digestible and engaging, ensuring that learners gain practical knowledge applicable in modern industrial settings. Additionally, the project fosters strategic and structural cooperation among consortium members, primarily universities, by developing and testing various collaboration models. This includes virtual networks, the use of digital tools and platforms, and mechanisms to enhance student mobility through the recognition of qualifications and learning outcomes.

The project stands out due to its holistic approach to training, addressing the needs of diverse industrial stakeholders. Unlike previous initiatives that focused primarily on metrologists in quality assurance, this project expands its scope to engineering students at the bachelor’s and master’s levels, equipping them with skills in modern product development and supervision. These students will be trained in the latest industry standards, ensuring they are prepared for careers in advanced manufacturing. The project integrates concepts such as Product Manufacturing Information (PMI), which conveys non-geometric attributes in 3D computer-aided design (CAD), as well as Model-Based Definition (MBD) and digital twins, which play a key role in ensuring quality assurance at different stages of product development and production.

In this paper, we present the idea behind this project, examples of learning material, multimedia and interactive content developed within the project. Paper explains how this content may be used both by students/trainees and tutors. Integration of the content with virtual reality is also described.

By enhancing communication among stakeholders involved in production, the project ensures that modern optical measurement techniques are better understood and integrated into industrial workflows. It covers the entire product supervision cycle, from design to final production, using a combination of digital learning methods and hands-on workshops. By targeting a broad audience and meeting the industry’s demand for well-trained professionals, the project contributes to strengthening Europe’s digital manufacturing capabilities. The innovative training solutions and new approaches developed within TOPMEAS will not only increase the effectiveness of technical education but also improve the adaptability of future engineers to the rapidly evolving industrial landscape.

MACHINE LEARNING APPROACH FOR IMPROVING DATA ACCURACY OF AT-LINE AUTOMOTIVE MEASUREMENTS

Matt Cornelius

At-line metrology has revolutionized dimensional quality control within the automotive industry and enables applications in real-time process monitoring, closed-loop control, and digital twins.  High data accuracy is foundational to this revolution; however, traditional manual data review methods are infeasible at the volume of at-line measurement, therefore an improved data review method is needed.  Each day, the BMW Plant Spartanburg Body Shop measurement process generates approximately 1.2 million reported characteristics from a strategic collection of measurement features.  Under normal operating conditions, approximately 0.1% of these measured features are anomalies, caused either by a real dimensional condition (60% of anomalies) or a misrepresentative measurement extraction (40% of anomalies).  These features and characteristics are utilized by a variety of critical users and automated processes to drive the dimensional quality of the vehicle, making data accuracy a core component of the measurement process.  Traditional methods for achieving high data accuracy include proactively reviewing full measurement programs or out-of-tolerance features prior to data storage or reactively reviewing data points after an end user identifies suspicious data.  The traditional proactive methods require significant manual effort, such as having to unnecessarily review many representative measurement extractions and are not feasibly achievable with the measurement volume of at-line metrology systems.  The reactive review methods require less manual effort but decrease initial data accuracy and degrade the trust and usefulness of the measurement system. 

In this presentation and paper, a novel machine learning data review system is presented to detect and classify measurement anomalies while minimizing unnecessary manual data review.  Firstly, the system uses an unsupervised clustering and point assignment method for improving the likelihood of correctly extracting features from complex mating surfaces, such as gap and flush profiles in full Body-in-White measurements.  This method autonomously subdivides point clouds into part point clouds for more robust feature extraction.  In a study of Body-in-White measurement anomalies, this method contributed to a 28% reduction in misrepresentative extractions.  Additionally, an unsupervised machine learning method is used for dynamic anomaly detection, which approximates the instantaneous probability distribution of measurement features.  This method quickly and reliably filters complete measurement projects into suspicious points for further manual or automated review.  Lastly, a convolutional neural network (CNN), trained on composite images of multiple extraction views, classifies misrepresentative measurement extractions.  While testing the CNN on various types of circle features, 96% of misrepresentative extractions and 75% of representative extractions were successfully identified.  Applying these percentages to historical measurements estimates a 50% reduction in manual data review as compared to using the dynamic anomaly detection method alone, while limiting reported misrepresentative measurement extractions to 0.002% of measured features.  Ultimately, this holistic machine learning approach to data review improves the efficiency, reliability, and usefulness of at-line metrology systems.

An Alternative Approach to Align Spacecraft to RF Test System Reference Axis

Dr. Değer Akın, Ahmet Akgül

Alignment measurement process has a crucial role in the assembly, integration and test (AIT) activities of a spacecraft. In general approach, laser trackers and theodolites are used. Activities such as determining the reference axis of the spacecraft, aligning critical equipment onto the spacecraft, and aligning the spacecraft with test systems can be accomplished through theodolite measurements in the content of alignment measurement of spacecraft subjected to this article. This article focuses on aligning a spacecraft undergoing the AIT process with the reference axis of an RF test system by using theodolite measurements. Due to the absence of alignment cubes on the spacecraft and the lack of measurement infrastructure, the measurements were conducted by changing the orientation of spacecraft with respect to the test facility reference axis. Spacecraft was positioned onto the RF test system by means of a specific L-shaped adaptor. Due to the absence of alignment cubes on the spacecraft, an alignment cube was placed on the L-shaped adapter back side where it is assumed as coincide with Launch interface adaptor center point. The alignment cube positioned on L-shaped adaptor was measured by using theodolites. The alignment cube positioned on the L-shaped adaptor was rotated by 180 degrees around its own axis, and another measurement was taken in this configuration. Therefore, the axes obtained from these two measurements were determined. The angle bisector of these two axes was determined as the nominal axis and accepted as the representative reference axis of the spacecraft. Subsequently, the reference axis of the RF test system was measured by using the alignment cube located on the test facility. The measurement results revealed the orientation of the spacecraft relative to the reference axis of the RF test system, with the assumed axis as the spacecraft’s reference axis. Therefore, the spacecraft was adjusted and positioned relative to the reference axis of the test facility. The conducted RF measurements showed that the orientation of the spacecraft aligned using the approach, that is the subject of this study, was in agreement with the reference axis of the test facility. In this alignment campaign, where constraints such as the absence of alignment cube on the spacecraft, placement of a new alignment cube on the L-shaped adaptor, and the limitation measurement infrastructure were present, the spacecraft was aligned with the RF test facility using an alternative approach.

Laser Tracker Direct Scanning Vacant SMR Nests

Steve Seiler

A network of fixed target holders that repeatably accept 1.5” Spherically Mounted Retroflectors (SMRs) can serve as a valuable reference to critical features if they are included in the laser tracker survey that characterizes a rigid part.  These “nests” are especially important when the meaningful features of the part are subsequently obscured by components added in the fabrication process.  What happens when these fiducial nests are no longer safely accessible to accept an SMR?  This is the scenario of some Naval ships’ radar arrays – a laser tracker on the deck has line of sight to each fiducial nest, but it is impractical and unsafe to hold an SMR in each non-magnetic nest for the survey due to the array’s height. 

This paper explores the feasibility of directly scanning the nests with a laser tracker.  SMR nests were 3D printed using the model of those used on the radar arrays.  The tests were performed with a Leica ATS600 and a rotatable rigid plate with affixed printed nests and standard magnetic nests.  The “true” values of the plate’s targets were established by a multi-instrument network of redundant observations of SMRs in all nests.  The magnetic nests were used to locate the instrument to the plate in various configurations and the non-magnetic nests were scanned in each configuration so that the impact of distance and approach angle variables on extracted center point accuracy could be isolated and quantified.  Automated analysis with metrology software was used to programmatically extract the virtual SMR center from the raw point cloud of the nest surface. 

The study looked at differences in measurement methods and instrument settings for the scans.  The center point extraction algorithm is described, as well as its modified form to include corrections to compensate for errors caused by approach angle and/or distance if these are found to be systematic.  The evaluation of accuracy for a given distance and approach angle is presented in graphical format so that the accuracy can be considered against improved safety and measurement ease for the naval radar array survey and similar applications.

CMM Calibration

Rich Dittman, Brian Wallin

In an effort to standardize the calibration process for Coordinate Measuring Machines (CMM) across GE Aerospace facilities, Brian Wallin and I conducted a comprehensive review of existing calibration practices. Our investigation revealed a lack of uniformity in the calibration standards being applied, with variations in the use of ASME B89 and ISO 10360 standards across different facilities and suppliers.

Upon detailed analysis of these standards, we identified ISO 10360-2009 as the most suitable calibration standard for recommendation due to its comprehensive and consistent approach. The ISO 10360-2009 standard provides clear guidelines for the performance verification of CMMs, ensuring that measurements are accurate and reliable. This standard was chosen for its ability to address the diverse needs of GE Aerospace facilities and suppliers, providing a common framework for calibration.

To address the inconsistency in calibration requests to OEMs and third-party calibration sources, we developed a standardized Statement of Work (SOW) for CMM calibration. This SOW was formulated in collaboration with industry leaders such as Renishaw, Hexagon Manufacturing Intelligence, and East Coast Metrology. The SOW outlines specific requirements for calibration procedures, including the use of certified artifacts, environmental conditions, and detailed reporting of calibration results. By establishing these guidelines, we aim to ensure that calibration results are consistent and reliable regardless of the service provider.

The implementation of this standardized calibration process is crucial for maintaining the health of CMM equipment. Regular monitoring and adherence to these calibration standards help minimize downtime and ensure accurate measurement results. This proactive approach to equipment calibration supports the overall goals of GE Aerospace in terms of Safety, Quality, Delivery, and Cost (SQDC), with safety being the most critical factor.

By adopting the ISO 10360-2009 standard and the standardized SOW for CMM calibration, GE Aerospace facilities can achieve higher efficiency and reliability in their measurement processes. This initiative not only enhances the performance of CMM equipment but also contributes to the overall operational excellence of GE Aerospace.

Performance Evaluation of an Automated Laser Tracker System for the Vera C. Rubin Observatory

Guillem Homar

The Vera C. Rubin Observatory’s Simonyi Survey Telescope (Simonyi Telescope) is an 8.4m telescope now in the final testing phase, on Cerro Pachon in Chile. To fulfill the Rubin Observatory scientific objectives of conducting a decade-long time domain survey of the optical sky, the Legacy Survey of Space and Time (LSST), the telescope requires delivering a consistent exquisite image quality over its 3.5 degrees field of view (FoV). This is accomplished using a sophisticated active optics system (AOS) and perfect alignment. 

Building on the established use of laser tracker technology during telescope integration, we now report the first alignment results for the Symoni telescope optical system; the LSSTCam, the largest digital camera ever built, and the telescope's secondary mirror (M2) relative to its primary/tertiary mirror. Using the Leica AT930 laser tracker system, we characterize the alignment of LSST Cam and M2 with respect to the telescope’s optical axis and assess the system’s ability to maintain alignment within stringent tolerances.

This work also presents a detailed analysis of observed drifts in the alignment measurements over time. These findings confirm the metrology system’s capability to support precise alignment during final integration and highlight important considerations for maintaining optical performance during operations.

Data Fusion of Point Clouds from Multiple Measuring Instruments

Rilyn Fox

Effective defect detection is critical to the qualification of manufactured products. Common practices involve the inspection of dimensional point clouds to identify discrepancies such as scratches or dents. Color image data is frequently analyzed to classify discolorations and other surface irregularities. Both methods have shortcomings that can be solved by incorporating the other. The first half of this presentation will discuss efforts to combine more dimensionally accurate point cloud data with colorized data to effectively detect defects, while the second half of the presentation will describe current work on using variable, non-rigid transformations. Typical data fusion/registration processes use rigid transformations to manipulate point clouds. However, this leaves small incompatibilities where dimensional and color data may not align correctly. This is highlighted in Figure 1, which depicts a small patch of two larger point clouds that have been registered using ICP. The colorized point cloud is produced using stitched iPhone photos and is therefore relatively dimensionally inaccurate. The lighter circle next to each fastener shows the location of the fastener in what is referred to as the “dimensional point cloud,” taken from a more dimensionally accurate device (structured light scanner). The misalignment of these fasteners in the two clouds highlight the need for a new, better method of registration between point clouds. To overcome the constraints of rigid transformations such as ICP, non-rigid transformations will be applied on patches of point cloud data to locally stretch and shrink the less-accurate color cloud to match features identified on the more-accurate cloud. The work will eventually expand beyond simple planar cases such as the one shown in the figure to include closed cylindrical shapes as well as freeforms.