nach oben

2017 | Buch

Kapitel lesen Erstes Kapitel lesen

Machine Learning for Cyber Physical Systems

Selected papers from the International Conference ML4CPS 2016

herausgegeben von: Jürgen Beyerer, Oliver Niggemann, Christian Kühnert

Verlag: Springer Berlin Heidelberg

Buchreihe : Technologien für die intelligente Automation

Enthalten in: Springer Professional "Wirtschaft+Technik" , Springer Professional "Technik" , Springer Professional "Wirtschaft"

Einloggen, um Zugang zu erhalten

Über dieses Buch

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 Karlsruhe, September 29th, 2016.

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.

Inhaltsverzeichnis

Frontmatter

A Concept for the Application of Reinforcement Learning in the Optimization of CAM-Generated Tool Paths

Abstract

Cyber physical systems (CPS) are changing the way machine tools function and operate. As the CAD-CAM-CNC tool chain gains intelligence the boundaries of the elements of the tool chain become blurred and new features, based on advancements in artificial intelligence can be integrated. The main task of the CAD-CAM-CNC chain is to generate the cutter trajectories for the manufacturing operation. Driven by sustainability and the need for capacity, the need arises to optimize the paths through this tool chain. In this paper a concept for path optimization with reinforcement learning is proposed, with focus on the reward function, specific to tool path optimization via the channel method.

Caren Dripke, Sara Höhr, Akos Csiszar, Alexander Verl

Semantic Stream Processing in Dynamic Environments Using Dynamic Stream Selection

Abstract

Cyber-physical systems (CPS) require a new level of dynamics in information processing. Databases and query approaches need to be extended towards dynamic stream aggregation and analysis systems. In this paper, we designed ECQELS, a semantic stream processing engine, to support CPS applications by adding essential features like dynamic sensor selection. We present a feature complete first implementation and show competitive performance results.

Michael Jacoby, Till Riedel

Dynamic Bayesian Network-Based Anomaly Detection for In-Process Visual Inspection of Laser Surface Heat Treatment

Abstract

We present the application of a cyber-physical system for inprocess quality control based on the visual inspection of a laser surface heat treatment process. To do this, we propose a classification framework that detects anomalies in recorded video sequences that have been preprocessed using a clustering-based method for feature subset selection. One peculiarity of the classification task is that there are no examples with errors, since major irregularities seldom occur in efficient industrial processes. Additionally, the parts to be processed are expensive so the sample size is small. The proposed framework uses anomaly detection, cross-validation and sampling techniques to deal with these issues. Regarding anomaly detection, dynamic Bayesian networks (DBNs) are used to represent the temporal characteristics of the normal process. Experiments are conducted with two different types of DBN structure learning algorithms, and classification performance is assessed on both anomaly-free examples and sequences with anomalies simulated by experts.

Alberto Ogbechie, Javier Díaz-Rozo, Pedro Larrañaga, Concha Bielza

A Modular Architecture for Smart Data Analysis using AutomationML, OPC-UA and Data-driven Algorithms

Abstract

Today, heterogeneous tool landscape and different data suppliers in the production environment complicate a universal component for the processing of process and quality data. The development effort of a suitable system for process data analysis comprises a serious effort for the connection to the data sources, the comprehension of the recorded data, and the development of a feasible visualization. To avoid this, an integrated architecture based on existing industrial standards can be used. The present paper discusses such a modular architecture which makes the possibilities of process optimization and predictive maintenance transparent to the user. The architecture is based on standards für production plant modelling (AutomationML) and for the connection to the production process (OPC UA). It includes an example implementation of water quality monitoring using principal component analysis.

Christian Kühnert, Miriam Schleipen, Michael Okon, Robert Henßen, Tino Bischoff

Cloud-based event detection platform for water distribution networks using machine-learning algorithms

Abstract

Modern water distribution networks are equipped with a large amount of sensors to monitor the drinking water quality. To detect anomalies, usually each sensor contains its own threshold, but machine-learning algorithms become an alternative to reduce the parametrization effort. Still, one reason why they are not used in practice is the geographical restricted data access. Data is stored at the plant, but data scientists needed for the data analysis are situated elsewhere.

To overcome this challenge, this paper proposes a cloud-based event-detection and reporting platform, which provides a possibility to use machine learning algorithms. The plant’s measurements are cyclically transferred into a secure cloud service where they are downloaded and analyzed from the data scientist. Results are made available as reports.

Thomas Bernard, Marc Baruthio, Claude Steinmetz, Jean-Marc Weber

A Generic Data Fusion and Analysis Platform for Cyber-Physical Systems

Abstract

In the future, production systems and information technology will merge, providing new ways for data processing and analysis. Still, the current situation is that for different production environments, different IT infrastructures exist. This makes data gathering, fusion and analysis process an elaborate work or even unfeasible.

Hence, this paper presents a generic, extendable and adaptable data fusion and analysis platform. Within this platform it is possible to connect onto different production systems, collect and process their measurements in realtime and finally give feed-back to the user. To keep the platform generic, the architecture follows a plug-in based approach. It is possible to integrate data from new productions systems into the platform as well as tailor made algorithms for analysis. As a use case, the platform is used on an industry 4.0 testbed which is used to monitor and track the lifecycle of a load process.

Christian Kühnert, Idel Montalvo Arango

Agent Swarm Optimization: Exploding the search space

Abstract

Agent Swarm Optimization is a framework that combines the use of evolutionary algorithms, data mining, modeling and other techniques to find the best compromises among objectives in complex decision problems. It has been applied mainly in engineering cases where using classic optimization algorithms would require undesired simplifications of the problem or the use of simulators for evaluating the objective functions. The flexibility of evolutionary algorithms makes possible to use them in practically any case. Nevertheless, in this paper we are presenting a complex problem where using “pure” evolutionary algorithms was not resulting in good solutions. A different situation appeared after using rules for reducing the search space and moving the evolutionary process toward zones with a higher probability of containing good solutions. The results of using rules is also presented in this paper for the case studied. Additionally, the paper explores the capacity of the algorithms to discover additional rules that can improve the search process and the way the evolutionary algorithms behave in problems where the expert knowledge to generate search rules is limited.

Idel Montalvo Arango, Joaquín Izquierdo Sebastián

Anomaly Detection in Industrial Networks using Machine Learning: A Roadmap

Abstract

With the advent of 21st Century, we stepped into the fourth industrial revolution of cyber physical systems. There is the need of secured network systems and intrusion detection systems in order to detect network attacks. Use of machine learning for anomaly detection in industrial networks faces challenges which restricts its large-scale commercial deployment. ADIN Suite proposes a roadmap to overcome these challenges with multi-module solution. It solves the need for real world network traffic, an adaptive hybrid analysis to reduce error rates in diverse network traffic and alarm correlation for semantic description of detection results to the network operator.

Ankush Meshram, Christian Haas

Titel: Machine Learning for Cyber Physical Systems
herausgegeben von: Jürgen Beyerer
Oliver Niggemann
Christian Kühnert
Verlag: Springer Berlin Heidelberg
Electronic ISBN: 978-3-662-53806-7
Print ISBN: 978-3-662-53805-0
DOI: https://doi.org/10.1007/978-3-662-53806-7

Springer Professional

Über dieses Buch

Inhaltsverzeichnis

Frontmatter

A Concept for the Application of Reinforcement Learning in the Optimization of CAM-Generated Tool Paths

Semantic Stream Processing in Dynamic Environments Using Dynamic Stream Selection

Dynamic Bayesian Network-Based Anomaly Detection for In-Process Visual Inspection of Laser Surface Heat Treatment

A Modular Architecture for Smart Data Analysis using AutomationML, OPC-UA and Data-driven Algorithms

Cloud-based event detection platform for water distribution networks using machine-learning algorithms

A Generic Data Fusion and Analysis Platform for Cyber-Physical Systems

Agent Swarm Optimization: Exploding the search space

Anomaly Detection in Industrial Networks using Machine Learning: A Roadmap

Premium Partner