Skip to main content
Disciplines
Automotive Business IT + Informatics Construction + Real Estate Electrical Engineering + Electronics Energy + Sustainability Insurance + Risk Finance + Banking Management + Leadership Marketing + Sales Mechanical Engineering + Materials
Events
DE
EN
Books
Journals
Topic Page
Marketing
Start single access now
Access for companies
EXTENDED SEARCH
Log in
Top

2016 | Book

Development of a Cyber-Physical System based on selective Gaussian naïve Bayes model for a self-predict laser surface heat treatment process control Read chapter Read first chapter

Machine Learning for Cyber Physical Systems

Selected papers from the International Conference ML4CPS 2015

Editors: Oliver Niggemann, Jürgen Beyerer

Publisher: Springer Berlin Heidelberg

Book Series : Technologien für die intelligente Automation

Part of: Springer Professional "Wirtschaft+Technik" , Springer Professional "Technik" , Springer Professional "Wirtschaft"

Login to get access
insite
SEARCH

About this book

The work presents new approaches to Machine Learning for Cyber Physical Systems, experiences and visions. It contains some selected papers from the international Conference ML4CPS – Machine Learning for Cyber Physical Systems, which was held in Lemgo, October 1-2, 2015.

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.

Table of Contents

Frontmatter
Development of a Cyber-Physical System based on selective Gaussian naïve Bayes model for a self-predict laser surface heat treatment process control
Abstract
Cyber-Physical Systems (CPS) seen from the Industrie 4.0 paradigm are key enablers to give smart capabilities to production machines. However, close loop control strategies based on raw process data need large amounts of computing power, which is expensive and difficult to manage in small electronic devices. Complex production processes, like laser surface heat treatment, are data intensive, therefore, the CPS development for these type of processes is challenging. As a result, the work described in this paper uses machine learning techniques like naïve Bayes classifiers and feature selection optimization, in order to evaluate its performance during surface roughness detection. Additionally, the feature selection techniques will define optimal measuring zones to reduce generated data. The models are the first step towards its future embedding into a laser process machine CPS and bring self-predict capabilities to it.
Javier Diaz, Concha Bielza, Jose L. Ocaña, Pedro Larrañaga
Evidence Grid Based Information Fusion for Semantic Classifiers in Dynamic Sensor Networks
Abstract
We propose an anytime fusion setup of anonymous distributed information sources with spatial affiliation. For this approach we use the evidence grid mapping algorithm which allows to fuse sensor information by their inverse sensor model. Furthermore, we apply an online Mixture of Experts training such that faulty voters are detected and suppressed during runtime by a gating function.
Timo Korthals, Thilo Krause, Ulrich Rückert
Forecasting Cellular Connectivity for Cyber-Physical Systems: A Machine Learning Approach
Abstract
Many applications in the context of Cyber-Physical Systems (CPS) can be served by cellular communication systems. The additional data traffic has to be transmitted very efficiently to minimize the interdependence with Human-to-Human (H2H) communication. In this paper, a forecasting approach for cellular connectivity based machine learning methods to enable a resource-efficient communication for CPS is presented. The results based on massive measurement data show that the cellular connectivity can be predicted with a probability of up to 78 %. Regarding a mobile communication system, a predictive channel-aware transmission based on machine learning methods enables a gain of 33 % concerning the spectral resource utilization of an Long Term Evolution (LTE) system.
Christoph Ide, Michael Nick, Dennis Kaulbars, Christian Wietfeld
Towards Optimized Machine Operations by Cloud Integrated Condition Estimation
Abstract
The requirements concerning the Overall Equipment Effectiveness (OEE) – especially machine availability – increase constantly in production nowadays. Unplanned down-times have to be prevented by efficient methods. Predictive, condition based maintenance represents a valuable approach for fulfilling these demands. Existing concepts lack of information, training data or interconnectedness. The objective of this paper is to present a novel approach in the context of Industrie 4.0 by using machine models with integrated uncertainties in the beginning, resolving these by methods of machine learning during operation and integrating both into a cloud-based service architecture.
C. Brecher, M Obdenbusch, W. Herfs
Prognostics Health Management System based on Hybrid Model to Predict Failures of a Planetary Gear Transmission
Abstract
Health condition monitoring has developed over several years. However, in the area of health assessment algorithms, most of the research has focused on data-driven approaches that do not rely on the knowledge of the physics of the system, while physics-based model (PbM) approaches which rely on the understanding of the system and the degradation mechanisms, are more limited and have the potential to provide more robust predictions due to the understanding of the failure mode phenomena. This paper proposes a Physics-based Model approach to detect incipient metal-metal contact and fatigue degradation of a planetary transmission of an aircraft. Both models are integrated in a realtime Prognostics Health Management (PHM) system that calculates the Remaining Useful Life (RUL) of the component. This tool also incorporates the decision-making process that is performed in the aircraft to connect/disconnect the transmission. A theoretical hybrid model that fuses a machine learning approach with the Physics-based approach to obtain a more robust prediction is also proposed.
Adrian Cubillo, Suresh Perinpanayagam, Marcos Rodriguez, Ignacio Collantes, Jeroen Vermeulen
Evaluation of Model-Based Condition Monitoring Systems in Industrial Application Cases
Abstract
In this paper, model-based condition monitoring methods are investigated. Reliable process monitoring allows costs and risks to be reduced by the early detection of faults and problems in the process behavior and the prevention of component failures or in extreme cases a production stop of the complete plant. The principal of model-based condition monitoring consists of comparing the actual process behavior with the behavior as predicted from process models. For this purpose, a Hidden Markov Model and a method based on principal component analysis are applied. Both methods are evaluated in industrial application cases. In doing so, F-measures of 88.25% and 98.84% are achieved for a wind power station and a glue production plant, respectively.
S. Windmann, J. Eickmeyer, F. Jungbluth, J. Badinger, O. Niggemann
Towards a novel learning assistant for networked automation systems
Abstract
Due to increasing requirements on functionality (e.g. self-diagnosis, self-optimization) or flexibility (e.g. self-configuration), future automation systems are demanded to be more and more intelligent. Therefore the systems are desired to learn new knowledge from other systems or its environment. The purpose of this work is to propose a prospective concept of learning assistant for networked automation systems. With the help of the assistant, an automation system can obtain new knowledge by collaborating with other systems to improve its prior knowledge. So that the system user is liberated from continuously providing new knowledge to an automation system.
Yongheng Wang, Michael Weyrich
Efficient Image Processing System for an Industrial Machine Learning Task
Abstract
We present the concept of a perceptive motor in terms of a cyber-physical system (CPS). A model application monitoring a knitting process was developed, where the take-off of the produced fabric is controlled by an electric motor. The idea is to equip a synchronous motor with a smart camera and appropriate image processing hard- and software components. Subsequently, the characteristics of knitted fabric are analysed by machine-learning (ML) methods. Our concept includes motor-current analysis and image processing. The aim is to implement an assistance system for the industrial large circular knitting process. An assistance system will help to shorten the retrofitting process. The concept is based on a low cost hardware approach for a smart camera, and stems from the recent development of image processing applications for mobile devices [1–4].
Kristijan Vukovic, Kristina Simonis, Helene Dörksen, Volker Lohweg
Efficient engineering in special purpose machinery through automated control code synthesis based on a functional categorisation
Abstract
Individual customer demands in special purpose machinery shift the focus towards efficient engineering. However, automated engineering approaches fail due to lack of an appropriate description language for the system modelling. Therefore, a categorisation of capabilities and functions is proposed, that can serve as basis for the description of machine parts in model-based approaches. As an example an approach that synthesises control code from a plant model, based on the description language developed in this paper, is presented.
Tobias Helbig, Steffen Henning, Johannes Hoos
Kognitive Architektur zum Konzeptlernen in technischen Systemen
Abstract
Durch die Komplexität technischer Systeme und die benötigte Flexibilität für Systemänderungen werden innovative Ansätze benötigt die schnell an neue Situationen angepasst werden können. In dieser Arbeit wird die Implementierung der Architektur CATS (Cognitive Architecture for Technical Systems) beschrieben. Im Vergleich mit bestehen Architekturen existieren einige Unterschiede: Die Kommunikation zwischen System und Bediener wird über natürliche Sprache realisiert, wobei die Sprachverarbeitung auf dem automatischen Lernen von Konzepten besteht. Da technische Systeme ihr Verhalten über die Zeit ändern, kann CATS nicht auf statische Textdokumente als Wissensbasis zurückgreifen. Stattdessen müssen die nötigen Informationen aus Echtzeitdaten extrahiert werden.
Um die Funktion von CATS zu validieren, wurde die Architektur auf einem Demonstrator implementiert. Das Stanford CoreNLP Framework wurde darin zur Sprachverarbeitung benutzt. Die Wissensmodellierung wird durch die Web Ontology Language (OWL) realisiert und die Vorverarbeitung der Maschinendaten mit verschiedenen maschinellen Lernalgorithmen implementiert. Der Demonstrator ist in der Lage 84% der Eingabetexte richtig zu beantworten.
Durch die Benutzung von CATS in technischen Systemen kann der Installations- und Adaptionsaufwand reduziert werden, da Ansätze für maschinelles Lernen Informationen über das System automatisch gewinnen. Weiterhin wird der Informationsaustausch zwischen Bediener und Maschine durch natürliche Sprache und Konzepte vereinfacht.
Alexander Diedrich, Andreas Bunte, Alexander Maier, Oliver Niggemann
Implementation and Comparison of Cluster-Based PSO Extensions in Hybrid Settings with Efficient Approximation
Abstract
This contribution presents a comparison between two extensions of the particle swarm optimization algorithm in a hybrid setting where the evaluation of the objective function requires a high computational effort. A first approach using simulation-based optimization via particle swarm optimization was developed in order to reach an improved setup optimization support of the workpiece position and orientation in a CNC tooling machine. For that, a 1:1 interface between the machine simulation model and the simulation-based optimization approach produced a high number of simulation runs. The idea arose that the extension of the PSO algorithm as well as the usage of an NC interpreter operating as a pre-processing component could support the setup process of the tooling machines. The extension of the PSO algorithm deals with the segmentation of the parameter search space taking collisions and lower computational effort into consideration. A significant reduction of simulation runs has been achieved.
André Mueß, Jens Weber, Raphael-Elias Reisch, Benjamin Jurke
Machine-specific Approach for Automatic Classification of Cutting Process Efficiency
Abstract
The identification of an inefficient cutting process e.g. in selfpropelled harvesters is a great challenge for automatic analysis. Machinespecific parameters of the process have to be examined to estimate the efficiency of the cutting process. As a contribution to that problem a simple method for indirect measurement of the efficiency is presented and described in this article.
To establish a general algorithm, the vibration data of a harvesting machine were extracted. The data from two sensors were recorded while gathering whole crop silage and while standing still in operation mode. For every data stream, a spectral analysis and a feature extraction was performed.
For the development of the algorithm, exploration techniques of Machine Learning were implemented. Artificial Neural Networks were optimized using subsets of the recorded data and then applied to the independent validation data to compute the efficiency of the cutting process. The established algorithm is able to identify the process efficiency without using additional machine-specific parameters.
The validation results are presented as confusion matrices for each data set and the case-specific population of the generated Artificial Neural Networks. The described algorithm is able to automatically determine an inefficient and machine-specific cutting process as an additional information using vibration data only.
Christian Walther, Frank Beneke, Luise Merbach, Hubertus Siebald, Oliver Hensel, Jochen Huster
Meta-analysis of Maintenance Knowledge Assets Towards Predictive Cost Controlling of Cyber Physical Production Systems
Abstract
Successful transition to Industry 4.0 requires cross domain and interdisciplinary research to develop new models for enhancing data and predictive analytics. Predictive models in particular should be applied to real time and remotely maintenance cost planning, monitoring and controlling of cyber physical production systems (CPPS). This paper presents a knowledge-based model, Costprove, discusses its mathematical meta-analysis approach for evidence extraction, and studies its application in the state-of-the-art industry towards its prospective in causality detection and predictive maintenance cost controlling of CPPS.
Fazel Ansari, Madjid Fathi
Towards Autonomously Navigating and Cooperating Vehicles in Cyber-Physical Production Systems
Abstract
This paper presents a (ROS-based) framework for the development and assessment of (decentralized) multi-robot coordination strategies for Cyber-Physical Production Systems (CPPS) taking into account practical issues like network delays, localization inaccuracies, and availability of embedded computational power. It constitutes the base for (a) investigating the beneficial level of (de-) centrality within Automated Guided Vehicle-based CPPS, and (b) finding adequate concepts for navigation and collision handling by means of behavior-, negotiationand rule-based strategies for resolving or proactively avoiding multi-robot path planning conflicts. Applying these concepts in industrial production is assumed to increase flexibility and fault-tolerance, e. g., with respect to machine failures or delivery delays at the shopfloor level.
Adrian Böckenkamp, Frank Weichert, Jonas Stenzel, Dennis Lünsch
Metadata
Title
Machine Learning for Cyber Physical Systems
Editors
Oliver Niggemann
Jürgen Beyerer
Copyright Year
2016
Publisher
Springer Berlin Heidelberg
Electronic ISBN
978-3-662-48838-6
Print ISBN
978-3-662-48836-2
DOI
https://doi.org/10.1007/978-3-662-48838-6

Premium Partner

    Image Credits