Search

Recurrent neural networks and exceedingly Long short-term memory (LSTM) have been investigated intensively in recent years due to their ability to model and predict nonlinear time-variant system dynamics. The present paper delivers a comprehensive overview of existing LSTM cell derivatives and network architectures for time series prediction. A categorization in LSTM with optimized cell state representations and LSTM with interacting cell states is proposed. The investigated approaches are evaluated against defined requirements being relevant for an accurate time series prediction. These include short-term and long-term memory behavior, the ability for multimodal and multi-step ahead predictions and the according error propagation. Sequence-to-sequence networks with partially conditioning outperform the other approaches, such as bidirectional or associative networks, and are best suited to fulfill the requirements.

Data-science-driven development projects are increasingly gaining the attention of small and medium sized enterprises. Since SME are often lacking the necessary competencies in data science, cooperation with other companies or universities is required. The efficient handling of data is one of the main challenges in joint cross-enterprise development projects. Actual cost driver is the development of data by labeling and classifying the data by domain experts, which is very time-consuming and labor-intensive with large amounts of data. Furthermore, clearance processes also have a high potential to cause delays before data can be shared with project partners. Moreover, before the actual work can begin, it is often necessary to clean up and repair incomplete or noisy data. The concept of Data Administration Shell presented in this paper addresses the challenge of structured information sharing and information management in joint cross-enterprise engineering. The Data Administration Shell links data sets to information regarding data origin and already performed analyses including their results and program scripts. Adding relations and documentation facilitates the reuse of data sets for subsequent projects. For this purpose, the Data Administration Shell adapts the concepts serving the information sharing in the research field of manufacturing and Digital Twin. The evaluation of the Data Administration Shell was based on time-series measurement data from a production process optimization scenario. Here, the Data Administration Shell manages the data sets of time series data and facilitates the joint cross-enterprise engineering of data-driven solutions.

The system development process is becoming increasingly complex as the systems themselves become more and more complex. At the same time, the various disciplines such as mechanical engineering, electrical engineering and software engineering are becoming increasingly intermixed, so that companies in one discipline are experiencing leaps and bounds in the complexity of their systems and their development. For this reason, this publication presents a concept of a virtual assistant that leads through a development phase. It shows that the software supporting the development needs an information model to store the data of the developed system and to connect it with the existing knowledge. This knowledge can either be available internally or on the web. The development process should therefore support cooperation so that the assistance software and engineers interact with each other.

The frequency of changes in production requirements is continuously increasing due to economic volatility, shorter innovation cycles and product life cycles. Therefore, a prediction of all goals of a production system at development time is impossible, which is leading to an increased reconfiguration demand during operation. Currently, there are still some weaknesses concerning the reconfiguration of production systems, which are highlighted in this article. The future of industrial automation will be dominated by Cyber-Physical Production Systems, which offer many promising potentials. Hence, this contribution discusses the Cyber-Physical Production Systems and some of their potentials regarding the reconfiguration issues. Corresponding concepts are required to utilize these theoretical potentials. Therefore, this research provides a basic concept for self-organized reconfiguration management.

The use of the Digital Twin as a promising technology during the reconfiguration of automated systems enables to meet the challenges of increasing product diversity and shortening product life cycles in today's industry. The use of Digital Twins supports engineers in all phases of the life cycle of a production line. Over time, however, production facilities are often modified and improved, while at the same time the created models during the engineering process of the system remain unchanged and no longer correspond to the real facility. Manual updating of the positions in the digital layout is very time-consuming and therefore expensive. This paper presents a concept for the automatic update of the layout of a production line. In this concept, the positions of the robots and their active and passive peripheral devices are automatically positioned in a digital plant model using information from the current configuration of the robots in a production line. Therefore, engineers can use the resulting synchronized digital plant model, Digital Twin of the system, to further optimize or expand the production plant.

Shorter product life cycles and increasing individualization of production leads to an increased reconfiguration demand in the domain of industrial automation systems, which will be dominated by cyber-physical production systems in the future. In constantly changing systems, however, not all configuration alternatives of the almost infinite state space are fully understood. Thus, certain configurations can lead to process instability, a reduction in quality or machine failures. Therefore, this paper presents an approach that enhances an intelligent Digital Twin with a self-organized reconfiguration management based on adaptive process models in order to find optimized configurations more comprehensively., Comment: 6 pages, 2 figures. Submitted to 54th CIRP Conference on Manufacturing Systems 2021

A multitude of high quality, high-resolution data is a cornerstone of the digital services associated with Industry 4.0. However, a great fraction of industrial machinery in use today features only a bare minimum of sensors and retrofitting new ones is expensive if possible at all. Instead, already existing sensors’ data streams could be utilized to virtually ‘measure’ new parameters. In this paper, a deep learning based virtual sensor for estimating a combustion parameter on a large gas engine using only the rotational speed as input is developed and evaluated. The evaluation focusses on the influence of data preprocessing compared to network type and structure regarding the estimation quality.

The capabilities of a cyber-physical production system are determined by their physical and virtual elements and their interaction. The interaction between virtual and physical entities must be precisely defined, and it is therefore necessary to develop capability models and requirements that enable self-reconfiguration of different elements and their relationships. Furthermore, an ultra-flexible production system requires communication and analysis of huge amounts of data as well as flexible, near-real-time communication between elements. This paper discusses and describes the functional and technical requirements of an ultra-flexible factory and presents an approach to the semantic structuring of elements and capabilities.

The heterogeneity and dynamics of IoT systems pose new challenges for their simulation, especially for hardware-in-the-loop simulations, which can be met by modular co-simulation. Therefore, several existing co-simulation approaches are presented and their suitability for hardware-in-the-loop co-simulations is evaluated. A new concept for dynamic hardware-in-the-loop co-simulation of IoT systems with a multi-agent system is presented, where each IoT component is simulated in its own simulation tool. Each individual simulation is represented by an agent and is therefore able to dynamically enter a co-simulation at run-time, which enables a "plug-and-simulate" behavior. The presented concept is evaluated by a prototypical implementation.

The early and robust detection of anomalies occurring in discrete manufacturing processes allows operators to prevent harm, e.g. defects in production machinery or products. While current approaches for data-driven anomaly detection provide good results on the exact processes they were trained on, they often lack the ability to flexibly adapt to changes, e.g. in products. Continual learning promises such flexibility, allowing for an automatic adaption of previously learnt knowledge to new tasks. Therefore, this article discusses different continual learning approaches from the group of regularization strategies, which are implemented, evaluated and compared based on a real industrial metal forming dataset., Comment: 6 pages, 5 figures, 3 tables, submitted to the CIRP Conference on Manufacturing Systems 2021

Modern production systems often form Internet-of-Things-systems (IoT). Because of the flexibility those systems are dynamic, meaning the entering of components during runtime, and heterogeneous. For the simulation of such systems, those challenges of dynamic and heterogeneity have to be met by a dynamic co-simulation. An important aspect of a co-simulation is the synchronization of the used simulations. In this contribution, challenges of synchronizing a co-simulation of IoT-systems are introduced and existing co-simulation synchronization concepts examined with regard to their usability for simulating IoT-systems. Afterwards a synchronization concept is presented, which can be used in the presented agent-based co-simulation concept.

Process anomalies and unexpected failures of manufacturing systems are problems that cause a decreased quality of process and product. Current data analytics approaches show decent results concerning the optimization of single processes but lack in extensibility to plants with high-dimensional data spaces. This paper presents and compares two data-driven self-learning approaches that are used to detect anomalies within large amounts of machine and process data. Models of the machine behavior are generated to capture complex interdependencies and to extract features that represent anomalies. The approaches are tested and evaluated on the basis of real industrial data from metal forming processes.

The increasing change of production leads to differences between the current shop floor and the state of planning. This difference causes significant challenges for production planners while integrating new products into existing production systems. To tackle this issue, this paper presents a concept for the automated creation of a digital twin of a body-in-white production system based on current resources, products as well as process information from the cyber-physical system. The paper focuses on the different data sources and information in cyber-physical systems necessary for integration planning. Furthermore, major parts of the concept are evaluated in a real body-in-white production system. The resulting digital twin enables faster product integration and Industry 4.0 concepts.

The heterogeneity and dynamic of IoT-systems pose new challenges for their simulation, which can be met by a modular co-simulation. Therefore several existing co-simulation approaches are presented and evaluated. A new concept for a dynamic co-simulation of IoT-systems utilizing a multi-agent-system is presented, wherein each IoT-component is simulated in a separate simulation tool. Each separate simulation is represented by an agent, and therefore able to enter a running co-simulation dynamically during runtime, which allows for a “Plug-and-Simulate” behavior. The connection between agents and simulation tools is realized by an interface concept. The presented concept is evaluated by a prototypical implementation.

Challenges on safeguarding distributed automation systems arise due to their increasing complexity and changeability. Functional changes in automation systems are mainly conducted by software modifications. Especially in distributed automation systems, the impacts of software modifications are difficult to estimate. Mainly, this will challenge plant operators who have to safeguard their automation systems after functionality changes were executed. If behaviour models of the automation systems are available, model-based techniques are suitable to estimate the impacts of software modifications on other system components. In fact, behaviour models of distributed automation systems are seldom available or maintained, due to the high complexity of the overall system and the changing structure caused by reconfigurations or software modifications. This often prevents the application of model-based techniques. This contribution presents a model-based approach with which the impacts of software modifications can be recognized and affected subsystems can be safeguarded efficiently by model-based verification methods. To achieve this an impact analysis is performed, identifying requirements which are affected by software modifications. As the behaviour models that are necessary to verify the identified requirements are seldom available, the necessary models are generated automatically. The approach is evaluated with modification to a large scale automation system.

A Digital Twin, an always in sync digital model of existing manufacturing cells, can be used to reduce time and risk of reconfiguration by early detection of design or process sequence flaws of the system in virtual commissioning and simulation. In this paper the need of Digital Twins in future production plants as well as the structure of the Digital Twin is presented. The engineering process of production systems is a cross-domain challenge between mechanics, electrics and software, but a lack of collaboration and universal information transfer between the domains leads to a high investment volume by synchronization the digital model from the time of commissioning. To synchronize cross-domain mechatronic data models of mechatronic components in the digital world during the life cycle of existing production systems this paper presents the anchor point method to firstly detect variances of cross-domain mechatronic data structure between the digital model and the real system in the specific domains electrics, mechanics and software and update the virtual models to have a consistent data model of the Digital Twin.

Increasing product variety and the shortening of product lifecycles require a fast and inexpensive reconfiguration of existing manufacturing automation systems. To face this challenge one solution is a Digital Twin, which can be used to reduce the complexity and time of reconfiguration by early detection of design or process sequence errors of the system with a cross-domain simulation. For engineering the Digital Twin and systemically synchronizing the data of mechatronic components in the interdisciplinary engineering models of a Digital Twin during the life cycle of manufacturing automation systems, this paper presents a concept for the engineering of a Digital Twin based on model integration in a PLM IT-Platform and an Anchor-Point method to systematically detect variances of the mechatronic data structure between the digital models and the physical system. The data of a mechatronic component from interdisciplinary domains, developed by the corresponding engineering tools are referred to as anchor points. This paper analyses domain-specific challenges in automation software-code to develop an assistance system for rule-based consistency check and for synchronizing the engineering models of the Digital Twin of the manufacturing automation system based on the Anchor-Point method.

In the age of the Industrial Internet of Things, distributed subsystems are established in many domains, which bring added value through their networking. In the field of industrial automation this means a distribution of the control system. As the distribution involves an increase in complexity, new approaches are needed that address this problem. This paper presents a model-driven development approach that involves a Service Oriented Architecture.

A requirement of future industrial automation systems is the application of intelligence in the context of their optimization, adaptation and reconfiguration. This paper begins with an introduction of the definition of (artificial) intelligence to derive a framework for artificial intelligence enhanced industrial automation systems: An artificial intelligence component is connected with the industrial automation system's control unit and other entities through a series of standardized interfaces for data and information exchange. This framework is then put into context of the intelligent Digital Twin architecture, highlight the latter as a possible implementation of such systems. Concluding, a prototypical implementation on the basis of a modular cyber-physical production system is described. The intelligent Digital Twin realized this way provides the four fundamental sub-processes of intelligence, namely observation, analysis, reasoning and action. A detailed description of all technologies used is given.

Digital Twins have been described as beneficial in many areas, such as virtual commissioning, fault prediction or reconfiguration planning. Equipping Digital Twins with artificial intelligence functionalities can greatly expand those beneficial applications or open up altogether new areas of application, among them cross-phase industrial transfer learning. In the context of machine learning, transfer learning represents a set of approaches that enhance learning new tasks based upon previously acquired knowledge. Here, knowledge is transferred from one lifecycle phase to another in order to reduce the amount of data or time needed to train a machine learning algorithm. Looking at common challenges in developing and deploying industrial machinery with deep learning functionalities, embracing this concept would offer several advantages: Using an intelligent Digital Twin, learning algorithms can be designed, configured and tested in the design phase before the physical system exists and real data can be collected. Once real data becomes available, the algorithms must merely be fine-tuned, significantly speeding up commissioning and reducing the probability of costly modifications. Furthermore, using the Digital Twin's simulation capabilities virtually injecting rare faults in order to train an algorithm's response or using reinforcement learning, e.g. to teach a robot, become practically feasible. This article presents several cross-phase industrial transfer learning use cases utilizing intelligent Digital Twins. A real cyber physical production system consisting of an automated welding machine and an automated guided vehicle equipped with a robot arm is used to illustrate the respective benefits., 6 pages, 7 figures, submitted to the CIRP Design Conference 2021

Increasing product variety and shortening product lifecycles require a fast and inexpensive reconfiguration of existing manufacturing systems. A synchronized Digital Twin of the manufacturing system is one solution to face these challenges. In order to use the Digital Twin for reconfiguration, a major challenge is to keep the developed Digital Twin of a manufacturing system, which was created during the engineering process, synchronized with the real system after commissioning. To automatically synchronize the cross-domain models of a Digital Twin after the commissioning of a manufacturing system, the authors introduced the Anchor-Point-Method in their previous papers. In this paper, the realization of the Anchor-Point-Method based on an assistance system is presented and the functionality is evaluated. This assistance system enables having an up-to-date Digital Twin of a manufacturing system available during the entire life-cycle of a system. Finally, a qualitative and quantitative evaluation of the advantages of a synchronized Digital Twin for the reconfiguration of a manufacturing system is presented. For this purpose, an automated system has been digitally and physically designed and built. On this system, a reconfiguration using the synchronized Digital Twin was performed and compared with another reconfiguration without Digital Twin collected through a survey. The results show that the Digital Twin can reduce the time of the reconfiguration process by up to 58 percent.

Intelligent industrial automation devices based on cyber-physical systems can establish ad-hoc networks of manufacturing systems to produce individual products. To realize those, all network-participants, with their offered services and properties need to be described in an explicit semantic. A cloud-based concept is conceived in this paper which includes a semantic to describe the manufacturing systems, their characterization as well as the specified user requirements. A cloud-based concept called “e-Production” maps the abilities of the manufacturing systems with the customer requirements and identifies the best collaborative units for production regarding to energy, costs and other quality characteristics.

Under the pressure of cost reduction and productivity improvement, this paper presents a new methodology which provides a fast inspection of defective objects and generates a real time motion trajectory for processing objects being conveyed with high speed in an industrial large-scale production. The image data obtained by a multispectral imaging system is analyzed within image processing algorithms using classification methods based on support vector machine. These data provide a basis for a path planning algorithm which considers location, orientation and arrangement of defects on the conveyed objects. Selective processing tool guided by the planed path is motion controlled.