TY - JOUR A1 - Schmitt, Jan A1 - Seitz, Philipp A1 - Scherdel, Christian A1 - Reichenauer, Gudrun T1 - Machine Learning in the development of Si-based anodes using Small-Angle X-ray Scattering for structural property analysis JF - Computational Materials Science N2 - Material development processes are highly iterative and driven by the experience and intuition of the researcher. This can lead to time consuming procedures. Data-driven approaches such as Machine Learning can support decision processes with trained and validated models to predict certain output parameter. In a multifaceted process chain of material synthesis of electrochemical materials and their characterization, Machine Learning has a huge potential to shorten development processes. Based on this, the contribution presents a novel approach to utilize data derived from Small-Angle X-ray Scattering (SAXS) of SiO_2 matrix materials for battery anodes with Neural Networks. Here, we use SAXS as an intermediate, high-throughput method to characterize sol–gel based porous materials. A multi-step-method is presented where a Feed Forward Net is connected to a pretrained autoencoder to reliably map parameters of the material synthesis to the SAXS curve of the resulting material. In addition, a direct comparison shows that the prediction error of Neural Networks can be greatly reduced by training each output variable with a separate independent Neural Network. KW - machine learning KW - neural network KW - autoencoder Y1 - 2023 UR - https://doi.org/10.1016/j.commatsci.2022.111984 SN - 1879-0801 N1 - Link zum Datensatz: https://gitlab.vlab.fm.fhws.de/philipp.seitz/machinelearningandsaxs VL - 218 ER - TY - CHAP A1 - Rosilius, Maximilian A1 - Wilhelm, Markus A1 - Seitz, Philipp A1 - Von Eitzen, Ingo A1 - Wirsing, Benedikt A1 - Rabenstein, Markus A1 - Decker, Steffen A1 - Bräutigam, Volker T1 - Equalization of the Visibility Loss between AR and Real Stimuli Sizes T2 - 2022 IEEE International Symposium on Mixed and Augmented Reality Adjunct (ISMAR-Adjunct) N2 - The results of the empirical study based on an AR Landolt C vision were examined to compare real and virtual visibility as a function of stimulus size on an HMD (e.g. MS HoloLens 2). A counterbalanced within-subject-design study with n = 32 probands was conducted. The results show on one hand a visibility loss between real and immersive systems by the same size of representation. On the other hand, a transfer function can be derived to ensure comparable visibility between the systems. This transfer function can be used to compensate for the visibility loss as a design guideline for immersive applications. In approximation, the Weber-Fechner's relation can be applied on AR visibility versus size of stimuli. Y1 - 2022 UR - https://ieeexplore.ieee.org/iel7/9973799/9974160/09974190.pdf U6 - https://doi.org/https://doi.org/10.1109/ISMAR-Adjunct57072.2022.00178 SP - 821 EP - 826 PB - IEEE ER - TY - JOUR A1 - Seitz, Philipp A1 - Schmitt, Jan T1 - Alternating Transfer Functions to Prevent Overfitting in Non-Linear Regression with Neural Networks JF - Journal of Experimental & Theoretical Artificial Intelligence N2 - In nonlinear regression with machine learning methods, neural networks (NNs) are ideally suited due to their universal approximation property, which states that arbitrary nonlinear functions can thereby be approximated arbitrarily well. Unfortunately, this property also poses the problem that data points with measurement errors can be approximated too well and unknown parameter subspaces in the estimation can deviate far from the actual value (so-called overfitting). Various developed methods aim to reduce overfitting through modifications in several areas of the training. In this work, we pursue the question of how an NN behaves in training with respect to overfitting when linear and nonlinear transfer functions (TF) are alternated in different hidden layers (HL). The presented approach is applied to a generated dataset and contrasted to established methods from the literature, both individually and in combination. Comparable results are obtained, whereby the common use of purely nonlinear transfer functions proves to be not recommended generally. KW - Machine learning; nonlinear regression; function approximation; overfitting; transfer function Y1 - 2023 U6 - http://nbn-resolving.de/urn/resolver.pl?urn:nbn:de:bvb:863-opus-49199 UR - https://doi.org/10.1080/0952813X.2023.2270995 ER - TY - JOUR A1 - Seitz, Philipp A1 - Scherdel, Christian A1 - Reichenauer, Gudrun A1 - Schmitt, Jan T1 - Machine Learning in the development of Si-based anodes using Small-Angle X-ray Scattering for structural property analysis JF - Computational Materials Science Y1 - 2023 VL - 218 SP - 111984 EP - 111984 ER - TY - CHAP A1 - Seitz, Philipp A1 - Schmitt, Jan A1 - Engelmann, Bastian T1 - Evaluation of proceedings for SMEs to conduct I4.0 projects T2 - Procedia Cirp Y1 - 2019 VL - 86 SP - 257 EP - 263 ER - TY - JOUR A1 - Höfflin, Dennis A1 - Hartmann, Jürgen A1 - Rosilius, Maximilian A1 - Seitz, Philipp A1 - Schiffler, Andreas T1 - Opto-Thermal Investigation of Additively Manufactured Steel Samples as a Function of the Hatch Distance JF - Sensors N2 - Nowadays, additive manufacturing processes are becoming more and more appealing due to their production-oriented design guidelines, especially with regard to topology optimisation and minimal downstream production depth in contrast to conventional technologies. However, a scientific path in the areas of quality assurance, material and microstructural properties, intrinsic thermal permeability and dependent stress parameters inhibits enthusiasm for the potential degrees of freedom of the direct metal laser melting process (DMLS). Especially in quality assurance, post-processing destructive measuring methods are still predominantly necessary in order to evaluate the components adequately. The overall objective of these investigations is to gain process knowledge make reliable in situ statements about component quality and material properties based on the process parameters used and emission values measured. The knowledge will then be used to develop non-destructive tools for the quality management of additively manufactured components. To assess the effectiveness of the research design in relation to the objectives for further investigations, this pre-study evaluates the dependencies between the process parameters, process emission during manufacturing and resulting thermal diffusivity and the relative density of samples fabricated by DMLS. Therefore, the approach deals with additively built metal samples made on an EOS M290 apparatus with varying hatch distances while simultaneously detecting the process emission. Afterwards, the relative density of the samples is determined optically, and thermal diffusivity is measured using the laser flash method. As a result of this pre-study, all interactions of the within factors are presented. The process variable hatch distance indicates a strong influence on the resulting material properties, as an increase in the hatch distance from 0.11 mm to 1 mm leads to a drop in relative density of 57.4%. The associated thermal diffusivity also reveals a sharp decrease from 5.3 mm2/s to 1.3 mm2/s with growing hatch distances. The variability of the material properties can also be observed in the measured process emissions. However, as various factors overlap in the thermal radiation signal, no clear assignment is possible within the scope of this work. KW - additive manufacturing processes KW - material Y1 - 2021 U6 - https://doi.org/10.3390/s22010046 SN - 1424-8220 VL - 22 IS - 1 PB - MDPI ER -