TY - CONF A1 - Schnur, C. A1 - Moll, J. A1 - Lugovtsova, Yevgeniya A1 - Schütze, A. A1 - Schneider, T. T1 - Explainable machine learning for damage detection - In carbon fiber composite plates under varying temperature conditions N2 - Understanding on how a machine learning model interprets data is a crucial step to verify its reliability and avoid overfitting. While the focus of the scientific community is nowadays orientated towards deep learning approaches, which are considered as black box approaches, this work presents a toolbox that is based on complementary methods of feature extraction and selection, where the classification decisions of the model are transparent and can be physically interpreted. On the example of guided wave benchmark data from the open guided waves platform, where delamination defects were simulated at multiple positions on a carbon fiber reinforced plastic plate under varying temperature conditions, the authors could identify suitable frequencies for further investigations and experiments. Furthermore, the authors presented a realistic validation scenario which ensures that the machine learning model learns global damage characteristics rather than position specific characteristics. T2 - 48th Annual Review of Progress in Quantitative Nondestructive Evaluation CY - Online meeting DA - 28.07.2021 KW - Explainable machine learning KW - Guided waves KW - Damage detection KW - Structural health monitoring KW - Composite structures PY - 2021 SN - 978-0-7918-8552-9 U6 - https://doi.org/10.1115/QNDE2021-75215 SP - 1 EP - 6 PB - American Society of Mechanical Engineers (ASME) CY - New York, NY AN - OPUS4-54219 LA - eng AD - Bundesanstalt fuer Materialforschung und -pruefung (BAM), Berlin, Germany ER - TY - CONF A1 - Schnur, C. A1 - Moll, J. A1 - Lugovtsova, Yevgeniya A1 - Schütze, A. A1 - Schneider, T. ED - Kundu, T. ED - Reis, H. ED - Ihn, J.-B. ED - Dzenis, Y. T1 - Explainable Machine Learning for Damage Detection: in Carbon Fiber Composite Plates Under Varying Temperature Conditions N2 - Understanding on how a machine learning model interprets data is a crucial step to verify its reliability and avoid overfitting. While the focus of the scientific community is nowadays orientated towards deep learning approaches, which are considered as black box approaches, this work presents a toolbox that is based on complementary methods of feature extraction and selection, where the classification decisions of the model are transparent and can be physically interpreted. On the example of guided wave benchmark data from the open guided waves platform, where delamination defects were simulated at multiple positions on a carbon fiber reinforced plastic plate under varying temperature conditions, the authors could identify suitable frequencies for further investigations and experiments. Furthermore, the authors presented a realistic validation scenario which ensures that the machine learning model learns global damage characteristics rather than position specific characteristics. T2 - 2021 48th Annual Review of Progress in Quantitative Nondestructive Evaluation CY - Online meeting DA - 28.07.2021 KW - Explainable machine learning KW - Guided waves KW - Damage detection KW - Structural health monitoring KW - Composite structures PY - 2021 SN - 978-0-7918-8552-9 U6 - https://doi.org/10.1115/QNDE2021-75215 VL - QNDE2021-75215 SP - 1 EP - 6 PB - The American Society of Mechanical Engineers (ASME) CY - New York, USA AN - OPUS4-56723 LA - eng AD - Bundesanstalt fuer Materialforschung und -pruefung (BAM), Berlin, Germany ER - TY - JOUR A1 - Schnur, C. A1 - Goodarzi, P. A1 - Lugovtsova, Yevgeniya A1 - Bulling, Jannis A1 - Prager, Jens A1 - Tschöke, K. A1 - Moll, J. A1 - Schütze, A. A1 - Schneider, T. T1 - Towards interpretable machine learning for automated damage detection based on ultrasonic guided waves N2 - Data-driven analysis for damage assessment has a large potential in structural health monitoring (SHM) systems, where sensors are permanently attached to the structure, enabling continuous and frequent measurements. In this contribution, we propose a machine learning (ML) approach for automated damage detection, based on an ML toolbox for industrial condition monitoring. The toolbox combines multiple complementary algorithms for feature extraction and selection and automatically chooses the best combination of methods for the dataset at hand. Here, this toolbox is applied to a guided wave-based SHM dataset for varying temperatures and damage locations, which is freely available on the Open Guided Waves platform. A classification rate of 96.2% is achieved, demonstrating reliable and automated damage detection. Moreover, the ability of the ML model to identify a damaged structure at untrained damage locations and temperatures is demonstrated. KW - Composite structures KW - Structural health monitoring KW - Carbon fibre-reinforced plastic KW - Interpretable machine learning KW - Automotive industry PY - 2022 U6 - http://nbn-resolving.de/urn/resolver.pl?urn:nbn:de:kobv:b43-542060 SN - 1424-8220 VL - 22 IS - 1 SP - 1 EP - 19 PB - MDPI CY - Basel AN - OPUS4-54206 LA - eng AD - Bundesanstalt fuer Materialforschung und -pruefung (BAM), Berlin, Germany ER - TY - JOUR A1 - Bastuck, M. A1 - Baur, T. A1 - Richter, Matthias A1 - Mull, B. A1 - Schütze, A. A1 - Sauerwald, T. T1 - Comparison of ppb-level gas measurements with a metal-oxide semiconductor gas sensor in two independent laboratories N2 - In this work, we use a gas sensor system consisting of a commercially available gas sensor in temperature cycled operation. It is trained with an extensive gas profile for detection and quantification of hazardous volatile organic compounds (VOC) in the ppb range independent of a varying background of other, less harmful VOCs and inorganic interfering gases like humidity or hydrogen. This training was then validated using a different gas mixture generation apparatus at an independent lab providing analytical methods as reference. While the varying background impedes selective detection of benzene and naphthalene at the low concentrations supplied, both formaldehyde and total VOC can well be quantified, after calibration transfer, by models trained with data from one system and evaluated with data from the other system. The lowest achievable root mean squared errors of prediction were 49 ppb for formaldehyde (in a concentration range of 20–200 ppb) and 150 μg/m³ (in a concentration range of 25–450 μg/m³) for total VOC. The latter uncertainty improves to 13 μg/m³ with a more confined model range of 220–320 μg/m³. The data from the second lab indicate an interfering gas which cannot be detected analytically but strongly influences the sensor signal. This demonstrates the need to take into account all sensor relevant gases, like, e.g., hydrogen and carbon monoxide, in analytical reference measurements. KW - Indoor air quality KW - Volatile organic compounds KW - Calibration transfer KW - Selective quantification KW - Inter-lab comparison PY - 2018 U6 - https://doi.org/10.1016/j.snb.2018.06.097 SN - 0925-4005 VL - 273 SP - 1037 EP - 1046 PB - Elsevier B.V. AN - OPUS4-45609 LA - eng AD - Bundesanstalt fuer Materialforschung und -pruefung (BAM), Berlin, Germany ER -