Thomas Schromm

Mr. Thomas Schromm

Technical University of Munich, Munich, Bavaria, Germany
orcid

Research Focus & Metrics

Academic Career Length

8Years (2019-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
31 July 2022 in Journal of Nondestructive Evaluation
Verified
Journal Article
Open access
PDF file
FBFabian Bauer
Fabian Bauer
DFDaniel Forndran
Daniel Forndran
TSThomas Schromm
Thomas Schromm
CGChristian Große
Christian Große
Robot-guided computed tomography enables the inspection of parts that are too large for conventional systems and allows, for instance, the non-destructive and volumetric evaluation of mechanical joining components within already assembled cars in the automotive industry. However, the typical scan time required by such setups is still significant and represents a major barrier for its industrial large-scale application. As an approach to mitigate the necessary time demand, we propose a part-specific adjustment of the acquisition trajectory. Common circular standard trajectories are inherently inefficient, since they are applied independently of the considered inspection task, while the use of acquisition orbits tailored particularly to the investigated object effectively allows a reduction of the required number of projections, which in turn has the potential to directly decrease the scan time significantly. In contrast to former simulation-guided approaches, this work is considered to be the first successful task-specific trajectory optimization being performed on a robot-based industrial CT platform and aims towards providing a first proof of concept that such methods can be practically applied in a shop floor environment. Based on representative results, a reduction of the number of required projections by approx. 55 % or an image quality improvement according to the root-mean squared error by approx. 40 % compared to the conventionally applied planar acquisition trajectory was achieved.

20 March 2022 in Materials
Verified
Journal Article
Open access
PDF file
NTNevine Tagscherer
Nevine Tagscherer
TSThomas Schromm
Thomas Schromm
KDKlaus Drechsler
Klaus Drechsler
The great potential of Extrusion Additive Manufacturing (EAM) for structural prototyping in the automotive industry is severely limited by the directional bias in the build direction. The layerwise fabrication leads to reduced mechanical properties at the layer-to-layer interface compared to the bulk of the strand. Especially for the often-used semi-crystalline thermoplastics, the mechanical properties strongly depend on the processing parameters, even more so if short fibers are used as fillers. Therefore, ideal processing windows in which the mechanical strength and modulus in the z-direction reach their maximum can be identified for these parameters, resulting in a reduced directional bias. The influence of the EAM processing parameters on mechanical strength has already been investigated, correlating strength with thermal conditions during printing. However, these considerations did not distinguish between the thermal effect on the polymer properties, the formation of voids and pores at the layer interface, and the resulting fiber orientation for different strand proportions. Therefore, in this study, the effect of different processing temperatures and layer heights on the pore size and distribution, as well as the fiber orientation in the different regions of the mesostructure was investigated using X-ray Computed Tomography (XCT).

19 June 2024 in Discover Applied Sciences
Verified
Journal Article
Open access
PDF file
TSThomas Schromm
Thomas Schromm
FBFelix Beckmann
Felix Beckmann
JMJulian Moosmann
Julian Moosmann
DBDaniel Berthe
Daniel Berthe
FPFranz Pfeiffer
Franz Pfeiffer
CGChristian Grosse
Christian Grosse
Despite recent advancements, contemporary laboratory-based industrial computed tomography (CT) technology has not yet achieved the status of an efficient and economically viable in situ non-destructive method for examining riveted joints that are commonly used in the automotive industry. 2D X-ray or 3D CT systems can only reliably assess a rivet joint’s condition in rare, and favorable circumstances. Moreover, a comprehensive analysis to identify the obstacles preventing CT from efficiently capturing high-resolution, and contrast-rich scans of riveted joints in general is lacking. This work delves into the challenges and limitations of the technology that prevent operators from reaching the desired precision required for rigorous quality control. The study specifically addresses questions concerning in situ reachability and positioning of the sample structure, radiodensity and beam hardening, the relationship between spot size and power, and economic feasibility—all within the context of analyzing riveted joints in the automotive industry. Additionally, we provide an abstracted overview of the current industrial X-ray tube market, which informs our discussion on these specific aspects. Through a combination of experimental findings, analytical insights, and deductions, we shed light on the persistent difficulties in achieving resolutions comparable to macro-sectioning and microscopy, which are currently considered the gold standard in examining riveted joints. Our investigation further explores the potential of a photon-counting detector paired with a conventional, laboratory-based X-ray source. Furthermore, we investigate the use of a synchrotron beamline as an X-ray source in conjunction with a custom-designed detector optimized for such beam geometries. This detector employs scintillators coupled with optical magnification, functioning similarly to an X-ray microscope. Our findings and data collection conclusively reveal that commercially available CT systems still fail to meet the requirements for efficient testing, underscoring the urgent need for further research and development to enable widespread implementation.

1 March 2019 in e-Journal of Nondestructive Testing
Verified
Journal Article
Open access
PDF file
BABmw Ag
Bmw Ag
TSThomas Schromm
Thomas Schromm
T(Technical University of Munich (TUM)
Technical University of Munich (TUM)
FDFabian Diewald
Fabian Diewald
CGChristian U. Grosse
Christian U. Grosse
Industries, which produce hundreds of terabyte of CT data per year, demand automated evaluation approaches. This work provides a first glance of an attempt to automatically detect and characterize possible defects and/or anomalies which formed during common joining processes. We investigated a standard riveting process with respect to the resulting final head height of steel selfpiercing half-hollow rivets. The methods include conventional image processing algorithms, like edge-detection, thresholding and principle component analysis (PCA) which were used to pre-process the CT data. In order to automatically evaluate the reconstructed volumes, which contained several of the aforementioned rivets, we compared the performance of different, publicly available, convolutional neural network (CNN) architectures. Furthermore, we investigated the impact of data augmentation and showed by means of a k-fold cross-validation that the training data causes no overfitting of the network. The obtained results suggest that an automated evaluation of the generated computed tomography scans, with regard to a rivet’s final head height, is feasible. However, in order to increase the network’s reliability and accuracy, the amount of training data needs to be further enlarged and diversified.

9 October 2021 in SN Applied Sciences
Verified
Journal Article
Open access
PDF file
TSThomas Schromm
Thomas Schromm
CGChristian Große
Christian Große
In this work, we present a methodology for shifting from a conventionally destructive, manual quality analysis for repetitive processes towards a non-destructive and largely automated process. The objects subjected to the quality analysis are mechanical joining elements like rivets or flow-drilling screws. We propose an algorithm that can automatically find and extract such joining elements from a computed tomography (CT) scan, rotate these elements to an upright orientation and eventually generate radial cross sections parallel to the elements’ longitudinal axis. The proposed algorithm was tested on five grayscale-based computed tomography volumes, with one synthetically generated volume. We will discuss both, cases in which the duo of CT and our proposed algorithm produces satisfying results, as well as cases in which it fails. Limitations of both the scan acquisition process and the proposed algorithm will be elaborated on and potential improvements will be mentioned.

1 February 2020 in e-Journal of Nondestructive Testing
Verified
Journal Article
Open access
PDF file
BABmw Ag
Bmw Ag
TSThomas Schromm
Thomas Schromm
JHJonas Holtmann
Jonas Holtmann
FLFraunhofer Institute for Structural Durability and System Reliability LBF
Fraunhofer Institute for Structural Durability and System Reliability LBF
MKMichael Koch
Michael Koch
CGChristian U. Grosse
Christian U. Grosse
Industrial tasks that produce hundreds of terabytes of data per year require efficient evaluation tools for the purpose of saving valuable resources such as time or highly trained personnel. Therefore, processes with high automation potential need to be identified and put into practice. Before deploying a new tool, however, thorough investigation is required. It needs to be tested whether the achievable degree of automation of such a new tool yields results which are on one hand reliable and on the other hand of sufficient quality. This work evaluates the applicability of artificial neural networks (ANNs) to detecting micrometer-cracks and delaminations in reconstructed computed tomography (CT) volumes of previously stressed carbon fibre reinforced polymer (CFRP) pressure rods. Common network architectures, varying network parameters and different training sets have been investigated and compared in order to determine the combination that performs best. The constellation (network, hyperparameters, training data) that performed best reached an average precision (AP) of 0.87. Based on the rather small data set of approx. 1.6E3 images and the unstructured nature and great diversity of the investigated features, this result can be regarded as very good. The detected feature sizes ranged from approx. 100 micrometer to a centimeter in length and from tens of microns to a few hundred microns in width. The results suggest that artificial neural networks have the potential to be used reliably for the automatic detection of micrometer-cracks and delaminations in CT volumes of CFRP pressure rods, provided the data set used for training is large and diverse enough and the network is being updated when new data is available.