Kai Kreisköther

Optimized Factory Planning and Process Chain Formation Using Virtual Production Intelligence

The increasing complexity of products creates new challenges in production planning. Hence, the methodology of process development has to be designed valuable. An innovative approach to reach efficient planning consists in the virtualization of planning processes. The concept of the ”Digital Factory” enables a preliminary evaluation of the planning success. In the present work, a framework is presented, which allows for the integration of dedicated applications into an integrative data model to gain a holistic mapping of the production. Using Intelligence approaches, data can be analyzed to provide decision support and optimization potentials. The advantages involved are demonstrated by a production structure planning approach in connection with a process chain optimization.

Improving vehicle localization using semantic and pole-like landmarks

In this paper, we present a framework for vehicle self-localization in urban environments. It utilizes semantic and distinctive physical objects such as trees, traffic signs or street lamps as robust landmarks and deduces the global vehicle pose in conjunction with an offline map. Since it is independent from the availability of road markings and the knowledge of street courses, application in dense urban areas with high rates dynamic objects and road users is possible. This paper introduces novel methods for vehicular environment perception via LiDAR scanner and stereo camera, as well as models for their association with a high-precision digital map to estimate the vehicles position via Adaptive Monte-Carlo Localization. Evaluation in urban areas indicates the potential for global positioning accuracy below 0.30 m for LiDAR and below 0.50 m for stereo camera, as well as a corresponding heading error below 1°.

Virtual Production Intelligence (VPI)

The research area Virtual Production Intelligence (VPI) focuses on the integrated support of collaborative planning processes for production systems and products. The focus of the research is on processes for information processing in the design domains Factory and Machine. These processes provide the integration and interactive analysis of emerging, mostly heterogeneous planning information. The demonstrators (flapAssist, memoSlice und VPI platform) that are information systems serve for the validation of the scientific approaches and aim to realize a continuous and consistent information management in terms of the Digital Factory. Central challenges are the semantic information integration (e.g., by means of metamodeling), the subsequent evaluation as well as the visualization of planning information (e.g., by means of Visual Analytics and Virtual Reality). All scientific and technical work is done within an interdisciplinary team composed of engineers, computer scientists and physicists.

Return on engineering: Design to cost for electric engine production

Producing companies for electric traction engines used in electric cars are faced with new challenges as production volumes and market demands change frequently. Missing standards in the product and production development process itself increase the expenses for development and production of an electric engine. As 70% of the product costs are determined by the product development and still generated solutions do not meet the required project cost targets, the focus of electric engine companies needs to be shifted to the product and process development. Therefore, the Return on Engineering approach pursues the maxim of forcing the cost target particularly in the early stages of development. This paper presents an approach to structure the product and process development process by modularizing the production process following the product architecture of an electrical engine. Thereby, the producing companies have the opportunity to decrease product costs and shorten their development process. The approach is being clarified with the aid of an exemplary development process of an electric engine for small city-cars. Especially technology chains for component production as well as final assembly are considered.

Selection of transformable production technologies as a reaction on a varying demand of electric traction motors

Manufacturers of electric vehicles face the difficult challenge to adjust proper production systems, facilities and technologies. The market is developing (regulations, policies and new competitors) and todays small volumes could increase fast, so that todays production systems might no longer be reasonable. This paper illustrates the necessity of transformable production technologies regarding a varying and increasing demand of electric drivetrains and explains how to deal with these developments. The paper shows and categorizes the external and internal influences regarding the selection of the right technology for this trendsetting industry. The requirements and important aspects to select an optimal production technology for electric traction motors will be exposed. An efficient selection process supports the manufacturer to choose the best individual technology in a developing market environment. The selection method is applied to the production step of winding and compares different winding technologies.

Towards Multi-Object Detection and Tracking in Urban Scenario under Uncertainties.

Urban-oriented autonomous vehicles require a reliable perception technology to tackle the high amount of uncertainties. The recently introduced compact 3D LIDAR sensor offers a surround spatial information that can be exploited to enhance the vehicle perception. We present a real-time integrated framework of multi-target object detection and tracking using 3D LIDAR geared toward urban use. Our approach combines sensor occlusion-aware detection method with computationally efficient heuristics rule-based filtering and adaptive probabilistic tracking to handle uncertainties arising from sensing limitation of 3D LIDAR and complexity of the target object movement. The evaluation results using real-world pre-recorded 3D LIDAR data and comparison with state-of-the-art works shows that our framework is capable of achieving promising tracking performance in the urban situation.

Vorausschauende Instandhaltung durch Maschinelles Lernen in der Prozessindustrie

Kurzfassung Durch den Einfluss der Zuverlässigkeit der Instandhaltung auf die Produktionskosten und die Anlagenverfügbarkeit gewinnt die vorausschauende Instandhaltung als Evolution der präventiven Instandhaltung zunehmend an Bedeutung. In diesem Beitrag werden die Entwicklung, Funktionalität, Einsatz und Potenzial eines vorausschauenden Instandhaltungsmodells am Beispiel eines Großanlagenkompressors dargelegt. Das auf Machine Learning basierende Modell konnte erfolgreich in der Industrie validiert werden, indem kritische Veränderungen elektrischer Maschinen vor ihrem Eintreten vorhergesagt wurden.

Discrete event simulation approach considering scalable systems and non-expert users in the early phase of production planning for electric powertrains

Electric vehicles will gain a significant market share within the next decade. Therefore, the automotive industry faces challenges regarding increasing number of units as well as technology uncertainty. To address these challenges production systems have to be developed that provide volume flexibility and reduce cost. It is essential to provide companies with planning methods that have the ability to assess manufacturing systems in a short time to react quickly in this disruptive environment. In this paper, we argue that through the use of discrete event simulation in an early phase, planning time can be reduced and the output quality increased. Conventional approaches focus mostly on the further improvement of existing manufacturing lines for a low uncertainty in production volume. The simulation of scalable production systems in the early phase requires a fully variable control between different modules. These modules have to be parametrical because of new technologies in the environment of electric mobility and the associated unsecure input data. Due to the high amount of scenarios production planners without simulation expertise have to be able to use the method. In this paper a method is proposed that uses predefined models to configure the scope of an assessment as well as to determine the required data. Preprogrammed modules are used to analyze the scalable concept of a manufacturing line efficiently. An application shows that the required accuracy can be achieved in a shorter time than using conventional methods.

Anlauf disruptiver Produkte

Kurzfassung Die Serienproduktion disruptiver Produkte stellt bestehende Antriebsproduzenten in der Automobilindustrie vor vollkommen neue Herausforderungen. Um den Wandel von der Verbrennungsmotorenproduktion hin zur Produktion elektrischer Antriebssysteme erfolgreich zu realisieren, muss der Erfahrungsaufbau mit den neuen Technologien im operativen Bereich systematisiert und beschleunigt werden. Der Produktionsanlauf als Schlüsselphase soll hierzu genutzt werden.

An Interaction Framework for a Cooperation between Fully Automated Vehicles and External Users in Semi-Stationary Urban Scenarios

Automated vehicles are becoming gradually available in restricted environments and are planned to be available for more challenging situations in the near future. Fully automated vehicles (FAVs) will have no drivers and still need to cooperate and interact with other road users outside the vehicle. In this work we propose an interaction framework, which makes it possible for external users to interfere with the FAV guidance in an abstract level via communicating a desired maneuver. The external user can be assumed as a road participant, who shares drivable areas with the FAV, or an operating person such as delivery person, who wants to guide a delivery vehicle remotely. The application area of this framework is the low velocity range, which can be also assumed as semi-stationary environments. The proposed framework explores the percepted static environment and identifies all possible paths with respect to vehicle dynamics, safety and comfort parameters. These paths are processed in order to build a set of meaningful candidates for the further steps. For this goal we have proposed two different methods based on a modified RRT algorithm and a skeletonization of the freespace. In order to extract possible drivable maneuvers out of the current scene, the candidate paths are assigned to predefined maneuver classes and selected with respect to their length and reasonableness. The set of meaningful and drivable maneuvers will be communicated to the user in form of an abstract and simplified catalogue. With this framework we provide both the FAV and the external user with a mutual understanding about the scene and avoid the possible ambiguity in goal understanding. The proposed framework is validated with sensor data from real scenarios.