[Technologien fÃ1⁄4r die intelligente Automation] Machine Learning for Cyber Physical Systems Volume 13 (Selected papers from the International Conference ML4CPS 2020) ||

This open access proceedings presents new approaches to Machine Learning for Cyber Physical Systems, experiences and visions. It contains selected papers from the fifth international Conference ML4CPS - Machine Learning for Cyber Physical Systems, which was held in Berlin, March 12-13, 2020. Cyber Physical Systems are characterized by their ability to adapt and to learn: They analyze their environment and, based on observations, they learn patterns, correlations and predictive models. Typical applications are condition monitoring, predictive maintenance, image processing and diagnosis. Machine Learning is the key technology for these developments. The Editors Prof. Dr.-Ing. Jürgen Beyerer is Professor at the Department for Interactive Real-Time Systems at the Karlsruhe Institute of Technology. In addition he manages the Fraunhofer Institute of Optronics, System Technologies and Image Exploitation IOSB. Dr. Alexander Maier is head of group Machine Learning at Fraunhofer IOSB-INA. His focus is on the development of algorithms for big data applications in Cyber-Physical Systems (diagnostics, optimization, predictive maintenance) and the transfer of research results to industry. Prof. Oliver Niggemann got his doctorate in 2001 at the University of Paderborn with the topic "Visual Data Mining of Graph-Based Data". He then worked for almost 8 years in leading positions in the industry. From 2008-2019 he held a professorship at the Institute for Industrial Information Technologies (inIT) in Lemgo/Germany. Until 2019 Prof. Niggemann was also deputy head of the Fraunhofer IOSB-INA, which works in industrial automation. On April 1, 2019 Prof. Niggemann took over the university professorship "Computer Science in Mechanical Engineering" at the Helmut-Schmidt-University in Hamburg / Germany. There he does research at the Institute for Automation Technology IfA in the field of artificial intelligence and machine learning for cyber-physical systems Preface 6 Table of Contents 7 1 Energy Profile Prediction of Milling Processes Using Machine Learning Techniques 9 1 Einleitung 9 2 Methode 11 3 Datenerhebung und -aufbereitung 12 3.1 Gewinnung der Zielwerte Energie- und Zeitbedarf 12 3.2 Gewinnung der Inputparameter für die Regressionsmodelle 12 3.3 Feature Engineering 13 4 Modellbildung 13 5 Ergebnisse und Validierung 15 6 Diskussion und Ausblick 17 References 18 2 Improvement of the prediction quality of electrical load profiles with artificial neural networks 20 1 Introduction 20 2 Analysis of the load profiles 21 2.1 Primary data preparation and plausibility check 21 2.2 Data analysis and creation load profile classes 21 2.3 Parameter estimation 21 2.4 Splitting the data sets 22 3 Artificial neural network as prediction model 22 3.1 Research studies 23 3.2 Basic specifications of the model 24 3.3 Investigation scenarios 25 4 Simulation and evaluation of the results 25 5 Conclusion and Outlook 27 References 27 3 Detection and localization of an underwater docking station in acoustic images using machine learning and generalized fuzzy hough transform 29 1 Introduction 29 2 Methodology 31 3 Experimental results 34 4 Conclusions and future work 35 5 Acknowledgements 36 References 36 4 Deployment architecture for the local delivery of ML-Models to the industrial shop floor 38 1 Introduction 38 2 Aim of the presented work 39 3 Related Work 39 4 Architecture 40 5 Data connectivity and collection 41 6 ML-Model Serving 42 7 Monitoring Strategies 43 8 Lifecycle Management 44 9 Discussion 45 10 Acknowledgement 46 References 46 5 Deep Learning in Resource and Data Constrained Edge Computing Systems 47 1 Introduction 47 2 Methods & Related Work 48 2.1 Variational Autoencoder 48 2.2 Federated Learning 48 3 Results 49 3.1 Clustering and Visualization of Wafermap Patterns 49 3.2 Anomaly Detection for Sensor Data of a Furnace 51 3.3 Predictive Maintenance using Federated Learning on Edge Devices 52 4 Conclusion 54 References 54 6 Prediction of Batch Processes Runtime Applying Dynamic Time Warping and Survival Analysis 56 1 Introduction 56 2 Dynamic Time Warping 58 3 Survival Analysis 59 4 Data 59 5 Proposed System 60 6 Results 61 7 Conclusion 62 References 62 7 Proposal for requirements on industrial AI solutions 65 1 Introduction 65 1.1 Usage of AI in Industrial Production 66 1.2 Industrial AI 66 2 Requirements on industrial AI 66 2.1 Adaption of Industrial AI systems 67 2.2 Engineering of Industrial AI systems 68 2.3 Embedding of Industrial AI system in existing production system landscape 69 2.4 Safety and Security of Industrial AI systems 69 2.5 Trust in functionality of Industrial AI systems 70 3 Discussion 71 4 Conclusion 71 Acknowledgements 71 References 71 8 Information modeling and knowledge extraction for machine learning applications in industrial production systems 75 1 Introduction 75 2 Information modeling 77 3 Tool chain for knowledge extraction 78 4 Conclusion 80 5 Acknowledgement 80 Appendix: Entities of the proposed information model 80 References 82 9 Explanation Framework for Intrusion Detection 84 1 Introduction 84 2 Explanations for Intrusion Detection 85 3 The Modular Phases of Explanations 86 4 Experiment 89 5 Summary 90 References 91 10 Automatic Generation of Improvement Suggestions for Legacy, PLC Controlled Manufacturing Equipment Utilizing Machine Learning 93 1 Introduction 93 2 Related Works 94 3 Hypothesis 95 4 Evaluation 99 5 Conclusion And Future Works 100 References 101 11 Hardening Deep Neural Networks in Condition Monitoring Systems against Adversarial Example Attacks 103 1 Introduction 103 2 Related work 105 3 Solution 106 4 Results 108 5 Conclusion 109 References 109 12 First Approaches to Automatically Diagnose and Reconfigure Hybrid Cyber-Physical Systems 112 1 Introduction 113 2 State of the Art 115 3 The multiple-tank model 116 4 Diagnosing Hybrid Systems 117 5 Reconfiguration after faults occurred 118 6 Conclusion and future work 119 7 Acknowledgement 120 References 120 13 Machine learning for reconstruction of highly porous structures from FIB-SEM nano-tomographic data 122 1 Introduction 122 2 Network architecture and and training the model 124 3 Results 125 4 Conclusion 125 References 127 This open access proceedings presents new approaches to Machine Learning for Cyber Physical Systems, experiences and visions. It contains selected papers from the fifth international Conference ML4CPS - Machine Learning for Cyber Physical Systems, which was held in Berlin, March 12-13, 2020. Cyber Physical Systems are characterized by their ability to adapt and to learn: They analyze their environment and, based on observations, they learn patterns, correlations and predictive models. Typical applications are condition monitoring, predictive maintenance, image processing and diagnosis. Machine Learning is the key technology for these developments. The Editors Prof. Dr.-Ing. Jürgen Beyerer is Professor at the Department for Interactive Real-Time Systems at the Karlsruhe Institute of Technology. In addition he manages the Fraunhofer Institute of Optronics, System Technologies and Image Exploitation IOSB. Dr. Alexander Maier is head of group Machine Learning at Fraunhofer IOSB-INA. His focus is on the development of algorithms for big data applications in Cyber-Physical Systems (diagnostics, optimization, predictive maintenance) and the transfer of research results to industry. Prof. Oliver Niggemann got his doctorate in 2001 at the University of Paderborn with the topic "Visual Data Mining of Graph-Based Data". He then worked for almost 8 years in leading positions in the industry. From 2008-2019 he held a professorship at the Institute for Industrial Information Technologies (inIT) in Lemgo/Germany. Until 2019 Prof. Niggemann was also deputy head of the Fraunhofer IOSB-INA, which works in industrial automation. On April 1, 2019 Prof. Niggemann took over the university professorship "Computer Science in Mechanical Engineering" at the Helmut-Schmidt-University in Hamburg / Germany. There he does research at the Institute for Automation Technology IfA in the field of artificial intelligence and machine learning for cyber-physical systems