Search

In the woodworking industry, a huge amount of effort has to be invested into the initial quality assessment of the raw material. In this study we present an AI model to detect, quantify and localize defects on wooden logs. This model aims to both automate the quality control process and provide a more consistent and reliable quality assessment. For this purpose a dataset of 1424 sample images of wood logs is created. A total of 5 annotators possessing different levels of expertise is involved in dataset creation. An inter-annotator agreement analysis is conducted to analyze the impact of expertise on the annotation task and to highlight subjective differences in annotator judgement. We explore, train and fine-tune the state-of-the-art InternImage and ONE-PEACE architectures for semantic segmentation. The best model created achieves an average IoU of 0.71, and shows detection and quantification capabilities close to the human annotators., Comment: FZI Workshop - K\"unstliche Intelligenz im Mittelstand (KI-KMU 2024)

In industrial manufacturing of glass bottles, quality control of bottle prints is necessary as numerous factors can negatively affect the printing process. Even minor defects in the bottle prints must be detected despite reflections in the glass or manufacturing-related deviations. In cooperation with our medium-sized industrial partner, two ML-based approaches for quality control of these bottle prints were developed and evaluated, which can also be used in this challenging scenario. Our first approach utilized different filters to supress reflections (e.g. Sobel or Canny) and image quality metrics for image comparison (e.g. MSE or SSIM) as features for different supervised classification models (e.g. SVM or k-Neighbors), which resulted in an accuracy of 84%. The images were aligned based on the ORB algorithm, which allowed us to estimate the rotations of the prints, which may serve as an indicator for anomalies in the manufacturing process. In our second approach, we fine-tuned different pre-trained CNN models (e.g. ResNet or VGG) for binary classification, which resulted in an accuracy of 87%. Utilizing Grad-Cam on our fine-tuned ResNet-34, we were able to localize and visualize frequently defective bottle print regions. This method allowed us to provide insights that could be used to optimize the actual manufacturing process. This paper also describes our general approach and the challenges we encountered in practice with data collection during ongoing production, unsupervised preselection, and labeling., Comment: VISAPP 2024 Conference

In the food industry, reprocessing returned product is a vital step to increase resource efficiency. [SBB23] presented an AI application that automates the tracking of returned bread buns. We extend their work by creating an expanded dataset comprising 2432 images and a wider range of baked goods. To increase model robustness, we use generative models pix2pix and CycleGAN to create synthetic images. We train state-of-the-art object detection model YOLOv9 and YOLOv8 on our detection task. Our overall best-performing model achieved an average precision AP@0.5 of 90.3% on our test set., Comment: FZI Workshop - K\"unstliche Intelligenz im Mittelstand (KI-KMU 2024)

A major challenge in the development of Model Predictive Control (MPC)-based energy management systems (EMSs) for buildings is the availability of an accurate model. One approach to address this is to augment an existing gray-box model with data-driven residual estimators. The efficacy of such estimators, and hence the performance of the EMS, relies on the availability of sufficient and suitable training data. In this work, we evaluate how different data availability scenarios affect estimator and controller performance. To do this, we perform software-in-the-loop (SiL) simulation with a physics-based digital twin using real measurement data. Simulation results show that acceptable estimation and control performance can already be achieved with limited available data, and we confirm that leveraging historical data for pretraining boosts efficacy., Comment: 5 pages, 4 figures. To be published in 2024 IEEE PES Innovative Smart Grid Technologies Europe (ISGT EUROPE) proceedings

The Semmeldetector, is a machine learning application that utilizes object detection models to detect, classify and count baked goods in images. Our application allows commercial bakers to track unsold baked goods, which allows them to optimize production and increase resource efficiency. We compiled a dataset comprising 1151 images that distinguishes between 18 different types of baked goods to train our detection models. To facilitate model training, we used a Copy-Paste augmentation pipeline to expand our dataset. We trained the state-of-the-art object detection model YOLOv8 on our detection task. We tested the impact of different training data, model scale, and online image augmentation pipelines on model performance. Our overall best performing model, achieved an AP@0.5 of 89.1% on our test set. Based on our results, we conclude that machine learning can be a valuable tool even for unforeseen industries like bakeries, even with very limited datasets.

The replacement of fossil fuels in combination with an increasing share of renewable energy sources leads to an increased focus on decentralized microgrids. One option is the local production of green hydrogen in combination with fuel cell vehicles (FCVs). In this paper, we develop a control strategy based on Model Predictive Control (MPC) for an energy management system (EMS) of a hydrogen plant, which is currently under installation in Offenbach, Germany. The plant includes an electrolyzer, a compressor, a low pressure storage tank, and six medium pressure storage tanks with complex heuristic physical coupling during the filling and extraction of hydrogen. Since these heuristics are too complex to be incorporated into the optimal control problem (OCP) explicitly, we propose a novel approach to do so implicitly. First, the MPC is executed without considering them. Then, the so-called allocator uses a heuristic model (of arbitrary complexity) to verify whether the MPC's plan is valid. If not, it introduces additional constraints to the MPC's OCP to implicitly respect the tanks' pressure levels. The MPC is executed again and the new plan is applied to the plant. Simulation results with real-world measurement data of the facility's energy management and realistic fueling scenarios show its advantages over rule-based control., Comment: 8 pages, 3 figures. To be published in IEEE 3rd International Conference on Power Electronics, Smart Grid, and Renewable Energy (PESGRE 2023) proceedings

One of the major challenges in the development of energy management systems (EMSs) for complex buildings is accurate modeling. To address this, we propose an EMS, which combines a Model Predictive Control (MPC) approach with data-driven model error compensation. The hierarchical MPC approach consists of two layers: An aggregator controls the overall energy flows of the building in an aggregated perspective, while a distributor distributes heating and cooling powers to individual temperature zones. The controllers of both layers employ regression-based error estimation to predict and incorporate the model error. The proposed approach is evaluated in a software-in-the-loop simulation using a physics-based digital twin model. Simulation results show the efficacy and robustness of the proposed approach, Comment: 8 pages, 4 figures. To be published in 2023 IEEE Conference on Control Technology and Applications (CCTA) proceedings

This paper describes a machine learning approach to determine the abrasive belt wear of wide belt sanders used in industrial processes based on acoustic data, regardless of the sanding process-related parameters, Feed speed, Grit Size, and Type of material. Our approach utilizes Decision Tree, Random Forest, k-nearest Neighbors, and Neural network Classifiers to detect the belt wear from Spectrograms, Mel Spectrograms, MFCC, IMFCC, and LFCC, yielding an accuracy of up to 86.1% on five levels of belt wear. A 96% accuracy could be achieved with different Decision Tree Classifiers specialized in different sanding parameter configurations. The classifiers could also determine with an accuracy of 97% if the machine is currently sanding or is idle and with an accuracy of 98.4% and 98.8% detect the sanding parameters Feed speed and Grit Size. We can show that low-dimensional mappings of high-dimensional features can be used to visualize belt wear and sanding parameters meaningfully., Comment: Conference IEEE SENSORS 2022

This paper proposes a comparison of four popular interface capturing methods : the volume of fluid (VOF), the standard level set (SLS), the accurate conservative level set (ACLS) and the coupled level set and volume of fluid (CLSVOF). All methods are embedded into a unified low-Mach framework based on a Cartesian-grid finite-volume discretization. This framework includes a sharp transport of the interface, a wellbalanced surface tension discretization and a consistent mass and momentum transport which allows capillary-driven simulations with high density ratio. The comparison relies on shared metrics for geometrical accuracy, mass and momentum conservation which exposes the weakness and strengths of each method. Finally, the versatility and capabilities of the proposed solver are demonstrated on the simulation of a 3D head-on collision of two water droplets. Overall, all methods manage to retrieve reasonable results for all test cases presented. VOF, CLSVOF and ACLS tend to artificially create little structures while SLS suffers from conservation issues in the mesh resolution limit. This study leads us to the conclusion that CLSVOF is the most promising method for two-phase flow simulations in our specific framework because of its inherent conservation properties and topology accuracy.

A series of f-block chromates, CsM(CrO$_4$)$_2$ (M = La, Pr, Nd, Sm, Eu; Am), were prepared revealing notable differences between the AmIII derivatives and their lanthanide analogs. While all compounds form similar layered structures, the americium compound exhibits polymorphism and adopts both a structure isomorphous with the early lanthanides as well as one that possesses lower symmetry. Both polymorphs are dark red and possess band gaps that are smaller than the LnIII compounds. In order to probe the origin of these differences, the electronic structure of $\alpha$-CsSm(CrO$_4$)$_2$, $\alpha$-CsEu(CrO$_4$)$_2$ and $\alpha$-CsAm(CrO$_4$)$_2$ were studied using both a molecular cluster approach featuring hybrid density functional theory and QTAIM analysis, and by the periodic LDA+GA and LDA+DMFT methods. Notably, the covalent contributions to bonding by the f orbitals was found to be more than twice as large in the AmIII chromate than in the SmIII and EuIII compounds, and even larger in magnitude than the Am-5f spin-orbit splitting in this system. Our analysis indicates also that the Am-O covalency in $\alpha$-CsAm(CrO$_4$)$_2$ is driven by the degeneracy of the 5f and 2p orbitals, and not by orbital overlap.

Recent experiments have presented more accurate data on the $\Lambda\Lambda$-binding energies of a few $\Lambda\Lambda$- hypernuclei. This is important as the $\Lambda\Lambda$- bond energies ($\Delta B_{\Lambda\Lambda}$) of double-$\Lambda$ hypernuclei provide a measure of the in-medium strength of the $\Lambda\Lambda$ interaction. A mass formula, optimized with the newly available $\Lambda\Lambda$ binding energy data, is used to estimate the binding energy and bond energy over a wide range of hypernuclei. The $\Delta B_{\Lambda\Lambda}$ values calculated with this mass formula are in good agreement with the experimental data, predictions of the quark mean-field (QMF) model and the relativistic mean-field (RMF) model as well. The $\Lambda\Lambda$-bond energy is found to diminish with neutron numbers, approaching zero near the neutron-drip line. In this formalism, the calculated binding energy difference in mirror nuclei arises from the Coulomb contributions and can be utilized to extract the Coulomb-corrected charge symmetry breaking effect., Comment: 7 pages, 3 figures

Thin film solid state lithium-based batteries (TSSBs) are increasingly attractive for their intrinsic safety due to the use of a nonflammable solid electrolyte, cycling stability, and ability to be easily patterned in small form factors. However, existing methods for fabricating TSSBs are limited to planar geometries, which severely limits areal energy density when the electrodes are kept sufficiently thin to achieve high areal power. In order to circumvent this limitation, we report the first successful fabrication of fully conformal, 3D full cell TSSBs formed in micromachined silicon substrates with aspect ratios up to ~10 using atomic layer deposition (ALD) at low processing temperatures (at or below 250C) to deposit all active battery components. The cells utilize a prelithiated LiV$_2$O$_5$ cathode, a very thin (40 - 100 nm) LiPON-like lithium polyphosphazene (Li$_2$PO$_2$N) solid electrolyte, and a SnN$_x$ conversion anode, along with Ru and TiN current collectors. Planar all-ALD solid state cells deliver 37 {\mu}Ah/cm$^2${\mu}m normalized to the cathode thickness with only 0.02% per-cycle capacity loss for hundreds of cycles. Fabrication of full cells in 3D substrates increases the areal discharge capacity by up to a factor of 9.3x while simultaneously improving the rate performance, which corresponds well to trends identified by finite element simulations of the cathode film. This work shows that the exceptional conformality of ALD, combined with conventional semiconductor fabrication methods, provides an avenue for the successful realization of long-sought 3D TSSBs which provide power performance scaling in regimes inaccessible to planar form factor devices., Comment: 32 pages, 15 figures

Several active areas of research in novel energy storage technologies, including three-dimensional solid state batteries and passivation coatings for reactive battery electrode components, require conformal solid state electrolytes. We describe an atomic layer deposition (ALD) process for a member of the lithium phosphorus oxynitride (LiPON) family, which is employed as a thin film lithium-conducting solid electrolyte. The reaction between lithium tert-butoxide (LiO$^t$Bu) and diethyl phosphoramidate (DEPA) produces conformal, ionically conductive thin films with a stoichiometry close to Li$_2$PO$_2$N between 250 and 300$^\circ$C. The P/N ratio of the films is always 1, indicative of a particular polymorph of LiPON which closely resembles a polyphosphazene. Films grown at 300$^\circ$C have an ionic conductivity of $6.51\:(\pm0.36)\times10^{-7}$ S/cm at 35$^\circ$C, and are functionally electrochemically stable in the window from 0 to 5.3V vs. Li/Li$^+$. We demonstrate the viability of the ALD-grown electrolyte by integrating it into full solid state batteries, including thin film devices using LiCoO$_2$ as the cathode and Si as the anode operating at up to 1 mA/cm$^2$. The high quality of the ALD growth process allows pinhole-free deposition even on rough crystalline surfaces, and we demonstrate the fabrication and operation of thin film batteries with the thinnest (<100nm) solid state electrolytes yet reported. Finally, we show an additional application of the moderate-temperature ALD process by demonstrating a flexible solid state battery fabricated on a polymer substrate., Comment: 31 pages, 8 figures