Bremer Institut für Produktion und Logistik GmbH

The nature of manufacturing systems faces ever more complex, dynamic and at times even chaotic behaviors. In order to being able to satisfy the demand for high-quality products in an efficient manner, it is essential to utilize all means available. One area, which saw fast pace developments in terms of not only promising results but also usability, is machine learning. Promising an answer to many of the old and new challenges of manufacturing, machine learning is widely discussed by researchers and practitioners alike. However, the field is very broad and even confusing which presents a challenge and a barrier hindering wide application. Here, this paper contributes in presenting an overview of available machine learning techniques and structuring this rather complicated area. A special focus is laid on the potential benefit, and examples of successful applications in a manufacturing environment.

Over the recent years Convolutional Neural Networks (CNN) have become the primary choice for many image-processing problems. Regarding industrial applications, they are hence especially interesting for automated optical quality inspection. However, with well-optimized processes is it often not possible to obtain a sufficiently large set of defective samples for CNN-based classification and the training objective shifts from defect classification to anomaly detection. Here we approach this problem with deep metric learning using triplet networks. Our evaluation shows promising results that even translate to novel surface/defect classes, which were not part of the training data.

An operative and versatile household energy management system is proposed to develop and implement demand response (DR) projects. These are under the hybrid generation of the energy storage system (ESS), photovoltaic (PV), and electric vehicles (EVs) in the smart grid (SG). Existing household energy management systems cannot offer its users a choice to ensure user comfort (UC) and not provide a sustainable solution in terms of reduced carbon emission. To tackle these problems, this research work proposes a heuristic-based programmable energy management controller (HPEMC) to manage the energy consumption in residential buildings to minimize electricity bills, reduce carbon emissions, maximize UC and reduce the peak-to-average ratio (PAR). We used our proposed hybrid genetic particle swarm optimization (HGPO) algorithm and existing algorithms like a genetic algorithm (GA), binary particle swarm optimization algorithm (BPSO), ant colony optimization (ACO), wind-driven optimization algorithm (WDO), bacterial foraging algorithm (BFA) to schedule smart appliances optimally to attain our desired objectives. In the proposed model, consumers use solar panels to produce their energy from microgrids. We also perform MATLAB simulations to validate our proposed HGPO-HPEMC (HHPEMC), and results confirm the efficiency and productivity of our proposed HPEMC based strategy. The proposed algorithm reduced the electricity cost by 25.55%, PAR by 36.98%, and carbon emission by 24.02% as compared to the case of without scheduling.

Environmental contours are an established method in probabilistic engineering design, especially in ocean engineering. The contours help engineers to select the environmental states which are appropriate for structural design calculations. Defining an environmental contour means enclosing a region in the variable space which corresponds to a certain return period. However, there are multiple definitions of environmental contours for a given return period as well as different methods to compute a contour. Here, we analyze the established approaches and present a new concept which we call highest density contour (HDC). We define this environmental contour to enclose the highest density region (HDR) of a given probability density. This region occupies the smallest possible volume in the variable space among all regions with the same included probability, which is advantageous for engineering design. We perform the calculations using a numerical grid to discretize the original variable space into a finite number of grid cells. Each cell's probability is estimated and used for numerical integration. The proposed method can be applied to any number of dimensions, i.e. number of different variables in the joint probability model. To put the highest density contour method in context, we compare it to the established inverse first-order reliability method (IFORM) and show that for common probability distributions the two methods yield similarly shaped contours. In multimodal probability distributions, however, where IFORM leads to contours which are difficult to interpret, the presented method still generates clearly defined contours.

Maritime vessels are complex systems that generate and require the utilization of large amounts of data for maximum efficiency. The successful utilization of sensors and IoT in the industry requires a forward-thinking approach to leverage the benefits of Industry 4.0 in a more comprehensive manner. While processes and manufacturing processes can be improved and advanced through such efforts, in order the industry to be able to benefit from data generation, integrated approaches are necessary. In order to develop truly value-added vessels, we introduce a descriptive approach for understanding Maritime 4.0.

With the improvement in transportation infrastructure and in-vehicle technology in addition to a meteoric increase in the total number of commercial and non-commercial vehicles on the road, traffic accidents may occur, which usually cause a high death toll. More than half of these deaths occur due to a delayed response by medical care providers and rescue authorities. The chances of survival of an accident victim could increase drastically if immediate medical assistance is provided at an accident location. This work proposes a low-cost accident detection and notification system, which utilizes a multi-tier IoT-based vehicular environment; principally, it uses V2X Communication and Edge/Cloud computing. In this work, vehicles are equipped with an On-Board Unit (OBU) in addition to mechanical sensors (accelerometer, gyroscope) for reliable accident detection along with a Global Positioning System (GPS) module for identification of accident location. In addition to this, a camera module is implanted on the vehicle to capture the moment when an accident takes place. In order to facilitate inter-vehicle communication (IVC), OBU in each vehicle incorporates a wireless networking interface. Once an accident occurs, a vehicle detects it and generates an alert message. It then sends the message along with the accident location to an intermediate device, placed at the edge of the vehicular network, and therefore called an edge device. Upon receiving the notification, this edge device finds the nearest hospital and makes a request for an ambulance to be dispatched immediately. It also performs some preprocessing of data and effectively acts as a bridge between the sensors installed inside the vehicle and the distant server deployed in the cloud. A significant issue that the traffic authorities are currently facing is the real-time visualization of data obtained through such environments. Wireless interfaces are usually capable of forwarding real-time sensor data; however, this feature is not yet commercially available in the OBU of the vehicle; therefore, practical implementation is carried out using the Internet of things (IoT) in order to create a network among the vehicles, the edge node, and the central server. By performing analysis on the adequate acquired data of road accidents, the constructive plans of action can be devised that may limit the death toll. In order to assist the relevant authorities in performing wholesome analysis of refined and reliable data, a dynamic front-end visualization is proposed, which is hosted in the cloud. The generated charts and graphs help the personnel at relevant organizations to make appropriate decisions based on the conclusive analysis of processed and stored data.

With the smart grid development, the modern electricity market is reformatted, where residential consumers can actively participate in the demand response (DR) program to balance demand with generation. However, lack of user knowledge is a challenging issue in responding to DR incentive signals. Thus, an Energy Management Controller (EMC) emerged that automatically respond to DR signal and solve energy management problem. On this note, in this work, a hybrid algorithm of Enhanced Differential Evolution (EDE) and Genetic Algorithm (GA) is developed, namely EDGE. The EMC is programmed based with EDGE algorithm to automatically respond to DR signals to solve energy management problems via scheduling three types of household load: interruptible, non-interruptible, and hybrid. The EDGE algorithm has critical features of both algorithms (GA and EDE), enabling the EMC to generate an optimal schedule of household load to reduce energy expense, carbon emission, Peak to Average Ratio (PAR), and user discomfort. To validate the proposed EDGE algorithm, simulations are conducted compared to the existing algorithms like Binary Particle Swarm Optimization (BPSO), GA, Wind Driven Optimization (WDO), and EDE. Results illustrate that the proposed EDGE algorithm outperforms benchmark algorithms in energy expense minimization, carbon emission minimization, PAR alleviation, and user discomfort maximization.

Currently, there is an ongoing transformation of classical products and machinery towards cyber-physical systems. Main features of these systems are the real-time data exchange between various technical and computational elements enabled by communication technologies and data processing ability provided by embedded systems. In the area of manufacturing, this trend boosts the development of cyber-physical production systems (CPPS). They enable the optimization of control processes, for example by autonomous decision-making, computational assistant systems for workers, or an extended human-machine collaboration. Subsequently, this increased computerization and automation provokes changes for human work in manufacturing. Following leading experts, the factories of the future will provide less easy and repetitive but more advanced and complex tasks. This trend changes the way how human factors or human-machine interaction influence the design of manufacturing systems. In order to achieve the promised productivity gains created by CPPS, these human-related topics have to be considered and included into the technical and organizational development of CPPS. As a starting point, a detailed view on remaining and newly added human tasks in CPPS is necessary. In this paper, we provide a listing of human task areas in existing and future CPPS. In this regard, we provide a trend estimation on the decline, rise, or further change of these tasks. The results can be used to facilitate the integration of human factors in the design of CPPS. We carry out our work by firstly deriving a standard listing of tasks for a generalized manufacturing system. Secondly, we combine the findings with expert judgments regarding CPPS trends and recent employment data from the German job market.

Integration of sensors into various kinds of products and machines provides access to in-depth usage information as basis for product optimization. Presently, this large potential for more user-friendly and efficient products is not being realized because (a) sensor integration and thus usage information is not available on a large scale and (b) product optimization requires considerable efforts in terms of manpower and adaptation of production equipment. However, with the advent of cloud-based services and highly flexible additive manufacturing techniques, these obstacles are currently crumbling away at rapid pace. The present study explores the state of the art in gathering and evaluating product usage and life cycle data, additive manufacturing and sensor integration, automated design and cloud-based services in manufacturing. By joining and extrapolating development trends in these areas, it delimits the foundations of a manufacturing concept that will allow continuous and economically viable product optimization on a general, user group or individual user level. This projection is checked against three different application scenarios, each of which stresses different aspects of the underlying holistic concept. The following discussion identifies critical issues and research needs by adopting the relevant stakeholder perspectives.

The computer-based representation of "things" in the real world is at the heart of today's virtual engineering practices. Digital Twin (DT) is a term that receives significant attention in academia and business within this domain. Despite its appealing metaphorical strength, people use it to describe quite different applications with specific conceptual backgrounds, goals and approaches. This paper aims to provide a first systematic classification about DT applications to support follow-up research. The first part of this paper focuses on three application cases described in the academic literature. It analyzes their conceptual background, the targeted problem and the implemented use case. The result of this analysis are seven dimensions that categorize the presented DT applications. They include distinctions of goals, focused users, life cycle phases, system levels, data sources, authenticity and data exchange levels.

The paper identifies the need for human robot collaboration for conventional light weight and heavy payload robots in future manufacturing environment. An overview of state of the art for these types of robots shows that there exists no solution for human robot collaboration. Here, we consider cyber physical systems, which are based on human worker participation as an integrated role in addition to its basic components. First, the paper identifies the collaborative schemes and a formal grading system is formulated based on four performance indicators. A detailed sensor catalog is established for one of the collaboration schemes, and performance indices are computed with various sensors. This study reveals an assessment of best and worst possible ranges of performance indices that are useful in the categorization of collaboration levels. To illustrate a possible solution, a hypothetical industrial scenario is discussed in a production environment. Generalizing this approach, a design methodology is developed for such human robot collaborative environments for various industrial scenarios to enable solution implementation.

Recent advances in natural language processing enable more intelligent ways to support knowledge sharing in factories. In manufacturing, operating production lines has become increasingly knowledge-intensive, putting strain on a factory's capacity to train and support new operators. This paper introduces a Large Language Model (LLM)-based system designed to retrieve information from the extensive knowledge contained in factory documentation and knowledge shared by expert operators. The system aims to efficiently answer queries from operators and facilitate the sharing of new knowledge. We conducted a user study at a factory to assess its potential impact and adoption, eliciting several perceived benefits, namely, enabling quicker information retrieval and more efficient resolution of issues. However, the study also highlighted a preference for learning from a human expert when such an option is available. Furthermore, we benchmarked several commercial and open-sourced LLMs for this system. The current state-of-the-art model, GPT-4, consistently outperformed its counterparts, with open-source models trailing closely, presenting an attractive option given their data privacy and customization benefits. In summary, this work offers preliminary insights and a system design for factories considering using LLM tools for knowledge management.

In cold regions like Northern Europe or Northern America, formation of ice on wind turbines not only reduces the power output of the system, but also leads to wind turbine breakdowns and poses a risk to nearby vehicles and people. Economical solutions for anti-icing systems such as preventive heating are required. Purely meteorological icing predictions are not sufficient to accurately predict icing on wind turbines. Therefore, this paper proposes an icing prediction approach that uses historical weather data and data from a supervisory control and data acquisition (SCADA) system plus methods from supervised machine learning to predict the risk of icing. The first results of the prediction model show its capability to predict most of the icing events. The discussion shows that the model should consider more meteorological data that describe the icing risk such as humidity or liquid water content to achieve better prediction results. The final aim is a prediction system that will use real-time data to determine icing risks hours beforehand, allowing the operation of existing anti-icing systems more efficiently.

Dispatching rules play an important role especially in semiconductor manufacturing scheduling, because these fabrication facilities are characterized by high complexity and dynamics. The process of developing and adapting dispatching rules is currently a tedious, largely manual task. Coupling Genetic Programming (GP), a global optimization meta-heuristic from the family of Evolutionary Algorithms, with a stochastic discrete event simulation of a complex manufacturing system we are able to automatically generate dispatching rules for a scenario from semiconductor manufacturing. Evolved dispatching rules clearly outperform manually developed rules from literature.

This paper addresses the implications of combining learning analytics and serious games for improving game quality, monitoring and assessment of player behavior, gaming performance, game progression, learning goals achievement, and user's appreciation. We introduce two modes of serious games analytics: in-game (real time) analytics, and post-game (off-line) analytics. We also explain the GLEANER framework for in-game analytics and describe a practical example for off-line analytics. We conclude with a brief outlook on future work, highlighting opportunities and challenges towards a solid uptake of SGs in authentic educational and training settings.

This paper reports on the evaluation methods and findings from serious games for teaching engineering and manufacturing. Two serious games are considered: Cosiga, a new product development simulation game and Beware, a risk management simulation game. These two games cover the front and middle parts of the engineering process – from design to manufacture to sale. For the Cosiga simulation evaluations of the communication and cognitive change were performed. For the Beware game evaluation of communication, risk awareness and improvement of risk management skills were performed The findings from the evaluations showed that serious games deliver learning outcomes. However, there are drawbacks to their use that need to be taken into account. Principally the high cost of development and the need for expert facilitators for running game sessions.

This paper arises from work ongoing in the GALA (Games and Learning Alliance — Network of Excellence for Serious Games). As part of GALA, a comprehensive state of the art analysis of existing serious games for the business and industry domain (loosely defined) was undertaken. A categorisation of the identified serious games was developed in order to analyse the characteristics of the serious games — the aspects they covered and those they do not cover. Of primary importance were the level, topic and skills mediated by the identified serious games. The simulation means the level or amount of the world that is simulated in the or serious game. This is a hierarchy starting with the World/ God/ Universe — in which level whole worlds are simulated, for example, in games such as Civilization. The hierarchy then proceeds downwards from nation, industry, inter-organisational, business/ organisation, intra-organisational/ processes, group/ team, discipline, techniques to games addressing the individual. Second the skills to be transferred by the serious game were also analysed. From this an analysis of the gaps in coverage of serious games was carried out, leading to identifying opportunities for, and recommendations of, serious games to be developed for the business and industry domain.

The ongoing introduction of cyber-physical systems in many areas of manufacturing will create profound changes in work design. Examples are new computerized tools or changed tasks due to a new allocation of work tasks between humans and machines. Hence the possible utilization of the potential enabled by cyber-physical production systems highly depends to what extent they are designed for humans. Therefore, an integrated system design is required which includes the human factors at an early stage. This work serves as a starting point for the development of a design-to-human factors for cyber-physical production systems. On the basis of the present state-of-the-art of relevant scientific research a hypothetical model is developed, which shows the interdependencies between human-oriented work design and the resulting job performance in regard of cyber-physical production systems. We use an interdisciplinary approach consisting of research and findings from human factors, ergonomics, human-machine-interaction and work psychology in connection with engineering goals. The resulting model provides exemplarywork design actions for cyber-physical production systems.

Product and process quality was and still is a key factor of success for manufacturing companies in the competitive global business environment. The stage gate model represents a well-established method for quality management in the product development domain. This paper discusses the application of the stage gate model in the domain of production. The two domains differ in certain areas, which has to be reflected by the adapted stage gate model. The preliminary findings of the two case studies, covering manufacturing and assembly processes, indicate that an adapted stage gate model may provide valuable support for product and process quality improvement. However, the success is strongly dependent of the right adaptation, taking the individual requirements, limitations and boundaries into consideration.

This article discusses the scientifically and industrially important problem of automating the process of unloading goods from standard shipping containers. We outline some of the challenges barring further adoption of robotic solutions to this problem, ranging from handling a vast variety of shapes, sizes, weights, appearances, and packing arrangements of the goods, through hard demands on unloading speed and reliability, to ensuring that fragile goods are not damaged. We propose a modular and reconfigurable software framework in an attempt to efficiently address some of these challenges. We also outline the general framework design and the basic functionality of the core modules developed. We present two instantiations of the software system on two different fully integrated demonstrators: (1) coping with an industrial scenario, i.e., the automated unloading of coffee sacks with an already economically interesting performance; and (2) a scenario used to demonstrate the capabilities of our scientific and technological developments in the context of medium- to long-term prospects of automation in logistics. We performed evaluations that allowed us to summarize several important lessons learned and to identify future directions of research on autonomous robots for the handling of goods in logistics applications.