A Spatio-Temporal Ensemble Deep Learning Architecture for Real-Time Defect Detection during Laser Welding on Low Power Embedded Computing Boards.

In modern production environments, advanced and intelligent process monitoring strategies are required to enable an unambiguous diagnosis of the process situation and thus of the final component quality. In addition, the ability to recognize the current state of product quality in real-time is an important prerequisite for autonomous and self-improving manufacturing systems. To address these needs, this study investigates a novel ensemble deep learning architecture based on convolutional neural networks (CNN), gated recurrent units (GRU) combined with high-performance classification algorithms such as k-nearest neighbors (kNN) and support vector machines (SVM). The architecture uses spatio-temporal features extracted from infrared image sequences to locate critical welding defects including lack of fusion (false friends), sagging, lack of penetration, and geometric deviations of the weld seam. In order to evaluate the proposed architecture, this study investigates a comprehensive scheme based on classical machine learning methods using manual feature extraction and state-of-the-art deep learning algorithms. Optimal hyperparameters for each algorithm are determined by an extensive grid search. Additional work is conducted to investigate the significance of various geometrical, statistical and spatio-temporal features extracted from the keyhole and weld pool regions. The proposed method is finally validated on previously unknown welding trials, achieving the highest detection rates and the most robust weld defect recognition among all classification methods investigated in this work. Ultimately, the ensemble deep neural network is implemented and optimized to operate on low-power embedded computing devices with low latency (1.1 ms), demonstrating sufficient performance for real-time applications.

Enhancing Adhesion of Fibrin-Based Hydrogel to Polythioether Polymer Surfaces.

The development of stable (bio)hybrid constructs composed of scaffolds and (bio)matrices is a major challenge in the field of tissue engineering. In the present work, the adhesion of fibrin-based hydrogels to the surface of polythioether-based polymers relevant to the 3D printing of polymer scaffolds produced by thiol-ene click chemistry was investigated. Adhesion properties were characterized by single-lap tensile shear testing. Both the sample preparation and the test method were optimized for the analysis of fibrin gel bonding to the polythioether surface. Our experimental results show that even without further modification, an adhesion between the fibrin hydrogel and polythioether is substantial, with an adhesion strength of 4.9 ± 1.0 kPa. To further improve the bonding, linear functional poly(N-vinylpyrrolidone-co-glycidyl methacrylate) (PVP-co-GMA) copolymers were used that are known for covalently binding to fibrin. The maximum adhesion strength in our study was found to be 18.4 ± 3.4 kPa. The pure PVP-co-GMA copolymers also demonstrate covalent binding to the thiol-ene-based polymers with a maximum adhesion strength of 32.2 ± 2.7 kPa. Therefore, compared to pure fibrin, the presence of copolymer coating both on the polythioether surface and in the fibrin gel led to a significant increase of the adhesion strength by a factor of 1.6.

LIFTOSCOPE: development of an automated AI-based module for time-effective and contactless analysis and isolation of cells in microtiter plates.

BACKGROUND: The cultivation, analysis, and isolation of single cells or cell cultures are fundamental to modern biological and medical processes. The novel LIFTOSCOPE technology aims to integrate analysis and isolation into one versatile, fully automated device. METHODS: LIFTOSCOPE's three core technologies are high-speed microscopy for rapid full-surface imaging of cell culture vessels, AI-based semantic segmentation of microscope images for localization and evaluation of cells, and laser-induced forward transfer (LIFT) for contact-free isolation of cells and cell clusters. LIFT transfers cells from a standard microtiter plate (MTP) across an air gap to a receiver plate, from where they can be further cultivated. The LIFT laser is integrated into the optical path of an inverse microscope, allowing to switch quickly between microscopic observation and cell transfer. RESULTS: Tests of the individual process steps prove the feasibility of the concept. A prototype setup shows the compatibility of the microscope stage with the LIFT laser. A specifically designed MTP adapter to hold a receiver plate has been designed and successfully used for material transfers. A suitable AI algorithm has been found for cell selection. CONCLUSION: LIFTOSCOPE speeds up cell cultivation and analysis with a target process time of 10 minutes, which can be achieved if the cell transfer is sped up using a more efficient path-finding algorithm. Some challenges remain, like finding a suitable cell transfer medium. SIGNIFICANCE: The LIFTOSCOPE system can be used to extend existing cell cultivation systems and microscopes for fully automated biotechnological applications.

YAP1 is essential for self-organized differentiation of pluripotent stem cells.

Induced pluripotent stem cells (iPSCs) form aggregates that recapitulate aspects of the self-organization in early embryogenesis. Within few days, cells undergo a transition from epithelial-like structures to organized three-dimensional embryoid bodies (EBs) with upregulation of germ layer-specific genes. However, it is largely unclear, which signaling cascades regulate self-organized differentiation. The Yes-associated protein 1 (YAP1) is a downstream effector of the Hippo pathway and essential mechanotransducer. YAP1 has been suggested to play a crucial role for early embryo development, but the relevance for early germ layer commitment of human iPSCs remains to be elucidated. To gain insights into the function of YAP1 in early cell-fate decisions, we generated YAP1 knockout (YAP-/-) iPSC lines with CRISPR/Cas9 technology and analyzed transcriptomic and epigenetic modifications. YAP-/- iPSCs showed increased expression of several YAP1 targets and of NODAL, an important regulator of cell differentiation. Furthermore, YAP1 deficiency evoked global DNA methylation changes. Directed differentiation of adherent iPSC colonies towards endoderm, mesoderm, and ectoderm could be induced, albeit endodermal and ectodermal differentiation showed transcriptomic and epigenetic changes in YAP-/- lines. Notably, in undirected self-organized YAP-/- EBs germ layer specification was clearly impaired. This phenotype was rescued via lentiviral overexpression of YAP1 and also by NODAL inhibitors. Our results demonstrate that YAP1 plays an important role during early germ layer specification of iPSCs, particularly for the undirected self-organization of EBs, and this is at least partly attributed to activation of the NODAL signaling.

Photochemistry with laser radiation in condensed phase using miniaturized photoreactors.

Miniaturized microreactors enable photochemistry with laser irradiation in flow mode to convert azidobiphenyl into carbazole with high efficiency.

Light driving force for surface patterning on azobenzene-containing polymers.

In this paper, we investigated the effect of light driving force induced surface deformation on azobenzene-containing polymers. The surface deformation is attributed to light-induced mass migration inside the polymers. Circular cap arrays are firstly fabricated by high power laser ablation via polarization controlled three-beam interference. The circular caps are subsequently exposed to polarization controlled two-beam interfering field. The results illuminate that when the interfering laser beams are both set to P polarization, the circular caps are deformed. While the laser beams are of other interfering modes like (S, S) and (+45° , -45°), the caps are seldom deformed. The circular caps are also exposed to single intensity-homogeneous linearly polarized laser beam. The deformation of the caps keeps the same direction as the irradiating polarization. A model based on the focusing effect of the circular caps is addressed to explain the origin of the light driving force for mass migration in azopolymers. The all-optical approach for the production of deformed caps can be used to generate aspherical lens, which may be applied to many domains.

Model-Based Estimation of the Strength of Laser-Based Plastic-Metal Joints Using Finite Element Microstructure Models and Regression Models.

Plastic-metal joints with a laser-structured metal surface have a high potential to reduce cost and weight compared to conventional joining technologies. However, their application is currently inhibited due to the absence of simulation methods and models for mechanical design. Thus, this paper presents a model-based approach for the strength estimation of laser-based plastic-metal joints. The approach aims to provide a methodology for the efficient creation of surrogate models, which can capture the influence of the microstructure parameters on the joint strength. A parametrization rule for the shape of the microstructure is developed using microsection analysis. Then, a parameterized finite element (FE) model of the joining zone on micro level is developed. Different statistical plans and model fits are tested, and the predicted strength of the FE model and the surrogate models are compared against experiments for different microstructure geometries. The joint strength is predicted by the FE model with a 3.7% error. Surrogate modelling using half-factorial experimental design and linear regression shows the best accuracy (6.2% error). This surrogate model can be efficiently created as only 16 samples are required. Furthermore, the surrogate model is provided as an equation, offering the designer a convenient tool to estimate parameter sensitivities.

All-optical fabrication of ellipsoidal caps on azobenzene functional polymers.

We have fabricated an azobenzene (azo) polymer microellipsoidal cap array of hexagonal symmetry with high-power laser ablation and subsequent single-beam-induced mass migration. High-power interference with polarization-controlled triple beams is utilized to inscribe a circular bump array directly on the surface of the azo polymer film. The produced circular cap array is exposed to the linearly polarized beam, and the caps are stretched along the polarization direction of the irradiating beam. A model of gradient force emerged by the interaction of the polarized beam and azo polymer is constructed to explain the mechanism of such polarization-induced microscale shape deformation.

Towards automation in biologics production via Raman micro-spectroscopy, laser-induced forward cell transfer and surface-enhanced Raman spectroscopy.

Mammalian cells have become the predominant expression system for the production of biopharmaceuticals due to their capabilities in posttranslational modifications. In recent years, the efficacy of these production processes has increased significantly through technical improvements. However, the state of the art in the development of producer cell lines includes many manual steps and is as such very time and cost consuming. In this study we developed a process combination of Raman micro-spectroscopy, laser-induced forward transfer (LIFT) and surface-enhanced Raman spectroscopy (SERS) as an automated machine system for the identification, separation and characterization of single cell-clones for biopharmaceutical production. Raman spectra showed clear differences between individual antibody-producing and non-producing chinese hamster ovary (CHO) cells after their stable transfection with a plasmid coding for an immunoglobulin G (IgG) antibody. Spectra of producing CHO cells exhibited Raman signals characteristic for human IgG. Individual producing CHO cells were successfully separated and transferred into a multiwell plate via LIFT. Besides, changes in concentration of human IgG in solution were detected via SERS. SERS spectra showed the same peak patterns but differed in their peak intensity. Overall, our results show that identification of individual antibody-producing CHO cells via Raman micro-spectroscopy, cell separation via LIFT and determination of changes in concentrations of overexpressed protein via SERS are suitable and versatile tools for assembling a fully automated system for biopharmaceuticals manufacturing.

Surface Topography Guides Morphology and Spatial Patterning of Induced Pluripotent Stem Cell Colonies.

The relevance of topographic cues for commitment of induced pluripotent stem cells (iPSCs) is largely unknown. In this study, we demonstrate that groove-ridge structures with a periodicity in the submicrometer range induce elongation of iPSC colonies, guide the orientation of apical actin fibers, and direct the polarity of cell division. Elongation of iPSC colonies impacts also on their intrinsic molecular patterning, which seems to be orchestrated from the rim of the colonies. BMP4-induced differentiation is enhanced in elongated colonies, and the submicron grooves impact on the spatial modulation of YAP activity upon induction with this morphogen. Interestingly, TAZ, a YAP paralog, shows distinct cytoskeletal localization in iPSCs. These findings demonstrate that topography can guide orientation and organization of iPSC colonies, which may affect the interaction between mechanosensors and mechanotransducers in iPSCs.

Surface topography enhances differentiation of mesenchymal stem cells towards osteogenic and adipogenic lineages.

Surface topography impacts on cell growth and differentiation, but it is not trivial to generate defined surface structures and to assess the relevance of specific topographic parameters. In this study, we have systematically compared in vitro differentiation of mesenchymal stem cells (MSCs) on a variety of groove/ridge structures. Micro- and nano-patterns were generated in polyimide using reactive ion etching or multi beam laser interference, respectively. These structures affected cell spreading and orientation of human MSCs, which was also reflected in focal adhesions morphology and size. Time-lapse demonstrated directed migration parallel to the nano-patterns. Overall, surface patterns clearly enhanced differentiation of MSCs towards specific lineages: 15 µm ridges increased adipogenic differentiation whereas 2 µm ridges enhanced osteogenic differentiation. Notably, nano-patterns with a periodicity of 650 nm increased differentiation towards both osteogenic and adipogenic lineages. However, in absence of differentiation media surface structures did neither induce differentiation, nor lineage-specific gene expression changes. Furthermore, nanostructures did not affect the YAP/TAZ complex, which is activated by substrate stiffness. Our results provide further insight into how structuring of tailored biomaterials and implant interfaces - e.g. by multi beam laser interference in sub-micrometer scale - do not induce differentiation of MSCs per se, but support their directed differentiation.

Local ultraviolet laser irradiation for gradients on biocompatible polymer surfaces.

Generation of supporting structures, which guide cell growth, is a challenging task in the field of tissue engineering. Cell guidance properties of a scaffold are important in the field of neuronal regeneration. Those guiding structures can provide guidance just by mechanical stimulus or by chemical stimuli like cell signaling molecules. For an enhanced guidance, chemical gradients are under investigation. With this study, we show that ultraviolet laser irradiation is a useful tool to activate polymer surfaces with a high temporal and spatial resolution. We demonstrated that poly(methyl methacrylate) (PMMA) and poly-ε-caprolactone (PCL) can be locally activated and functionalized with amine groups that can be used for immobilization of arginine-glycine-aspartic acid (RGD) peptide. The immobilized RGD was detected by neuronal B35 cells. By defined pulse accumulation functionalization density on the surface can be varied for the generation of gradients. We demonstrated that PMMA and PCL have different process windows for functionalization. Although PMMA has a very small process window for activation, PCL allows the generation of stepwise functionalization. The presented technology can help to develop assays for the analysis of cell migration and neuronal regeneration due to flexible patterning easily realized by changing the irradiation parameters.

Fabrication of 2D protein microstructures and 3D polymer-protein hybrid microstructures by two-photon polymerization.

Two-photon polymerization (TPP) offers the possibility of creating artificial cell scaffolds composed of micro- and nanostructures with spatial resolutions of less than 1 µm. For use in tissue engineering, the identification of a TPP-processable polymer that provides biocompatibility, biofunctionality and appropriate mechanical properties is a difficult task. ECM proteins such as collagen or fibronectin, which could mimic native tissues best, often lack the mechanical stability. Hence, by generating polymer-protein hybrid structures, the beneficial properties of proteins can be combined with the advantageous characteristics of polymers, such as sufficient mechanical stability. This study describes three steps toward facilitated application of TPP for biomaterial generation. (1) The efficiency of a low-cost ps-laser source is compared to a fs-laser source by testing several materials. A novel photoinitiator for polymerization with a ps-laser source is synthesized and proved to enable increased fabrication throughput. (2) The fabrication of 3D-microstructures with both systems and the fabrication of polymer-protein hybrid structures are demonstrated. (3) The tissue engineering capabilities of TPP are demonstrated by creating cross-linked gelatin microstructures, which clearly forced porcine chondrocytes to adapt their cell morphology.