The growing adoption of data-driven techniques to analyze sensor data in production engineering reinforces the significance of time series in research and development. In this context, time series management is on the verge of being reconceptualized. Modern data mining pipelines depend on seamless data flows and reproducible datasets. However, data acquisition and analysis are often treated as separate processes, creating a disconnect between tools. To create a unified time series management system for acquisition and analysis, two challenges must be addressed: contrary storage requirements and rapidly evolving information models.This paper proposes a data management solution for time series and related metadata for mechanical engineers seeking to aggregate and analyze data from machine prototypes, test beds, or small pilot lines. The solution integrates the previously disparate activities of data acquisition and analysis. This includes incorporating new sensors, as well as creating, augmenting, and analyzing labeled datasets. The solution is flexible by design and can be used with information models from OPC UA companion specifications, Asset Administration Shells, or as a standalone. A detailed description of the setup, tools, and validation on a machine tool is provided.

In event-based control systems, actuator operations are typically triggered by physical sensor signals to ensure process safety. However, this can lead to systematic delays, especially in pneumatic systems, due to system-related reaction times. This paper presents a method to compensate such delays by statistically advancing control signals without the need for additional sensors. A bidirectionally coupled digital twin, based on virtual commissioning models, continuously evaluates live data and derives optimized advance signals during operation. The method integrates event-based control with time-based triggering and employs a hybrid approach combining empirical quantiles with analytical confidence intervals to maintain safe operation. Validated in a testbed under realistic industrial conditions, the approach achieves an average time saving of 315 ms per compensated transition. The system architecture decouples real-time PLC execution from non-real-time analysis, enabling scalable integration with diverse control and simulation platforms. The results demonstrate the potential of digital twins for active, data-driven optimization in discrete manufacturing systems.

This work introduces a novel method for the generation of process-aligned robotic pathways specifically designed for surface processing applications. The proposed approach integrates the interpretation of sensor data, computer vision algorithms, and process knowledge modeling to address the complexities inherent in robotic programming. To mitigate programming challenges, the method incorporates intuitive interaction techniques, including hand gestures and human-computer interaction (HCI), thereby facilitating the efficient generation of robotic paths. Additionally, it augments the user’s teaching experience by enabling the seamless deployment of the methodology in serial production settings, accommodating the variability of both workpieces and environmental conditions. The proposed framework ensures smooth integration of robotic systems into complex workflows by aligning robotic paths with the unique requirements of surface processing tasks.

313

Industrial anomaly detection models proved to be a viable solution for automating the task of visual inspection. Despite their effectiveness, the deployment of current deep learning models for anomaly detection for high-resolution images involves large memory requirements, which are unattainable in industrial deployments on edge devices. In this paper we introduce a tiling approach for a memory bank based feature embedding model to reduce the memory requirements during training and inference. Additionally, we analyze the correlations between image size, dataset size and required memory. To evaluate the detection performance of different tiling schemes, we create and apply a real industrial training and validation dataset collected in the electric motor housing production of the BMW Group. Our experiments show that depending on the tiling scheme the overall required memory can be decreased by a factor of 12 while leaving the detection performance unchanged and even increase the detection performance when the memory requirement is only halved. This venture underscores the viability of our tiling approach in real-world industrial applications but also marks a significant leap towards resolving the issue of resolution limitations in deploying advanced anomaly detection models.

Automation of inspection for quality control is needed to overcome the errors and delays inherent in manual processes. Machine learning methods have the potential to greatly improve automated inspection. However, machine learning techniques require training data that are precisely labeled and reflect the distribution of defects to be detected. Physically collecting suitable training data requires significant time and prolongs the overall time for system development. To address this challenge, a new study is presented that explores synthetic data generation for a state-of-the-art anomaly detection (AD) model in the electric motor housing (EMH) surface inspection. The study successfully demonstrates using synthetic data for anomaly detection and presents a comparison of detection performance by models trained solely on synthetic data and models trained on both synthetic and real data. The study shows that real data combined with synthetic data can increase overall model performance. The study also addresses current challenges in using synthetic data and proposes directions for future work.

Industrial robot programming necessitates specialized expertise and significant time commitment, particularly for small-batch productions. In response to the escalating demand for production agility, novel approaches have emerged in intuitive robot programming. These inventive systems, rooted in diverse conceptual frameworks, are designed to expedite the deployment of robot systems. A prominent innovation in this domain is adopting no-code robot programming through finger-based gestures. A robot program can be generated by capturing and tracking non-expert users’ finger movements and gestures, converting 3D coordinates into an executable robot programming language. However, accurately determining finger positions for 3D coordinates and precise geometrical features presents an ongoing challenge. In pursuit of heightened trajectory precision and reducing more significant effort for the users, we propose a hybrid methodology that amalgamates finger-gesture programming with point cloud data. This synergistic integration demonstrates promising outcomes, substantiating its potential to facilitate the precise and adaptive generation of robot paths within robot applications.

The automated production of individualized products in a cyber-physical production system (CPPS) requires the combined automation of software and machine components. While this leads to increased productivity, the complexity of the CPPS may result in long unplanned downtimes when faults occur, and no system model is available to guide the maintenance team. Knowledge-driven, data-driven or hybrid modeling are available approaches in the literature to obtaining a system model. While expert-driven and data-driven modeling face limited applicability to CPPS, hybrid models, combining both approaches can offer a solution. This paper proposes a causality-driven hybrid model for fault diagnosis in complex CPPS, represented in a causal knowledge graph (CKG). The CKG serves as a transparent system model for collaborative human-machine fault diagnosis. We provide a concept for the continuous hybrid learning of the CKG, a maturity model to classify the resulting CKG’s fault diagnosis capabilities, and the industrial setting inspiring the approach.

The planning phases of industrial surface defect detection systems for machined, automotive workpieces are time-consuming and require project teams comprised of experts from different disciplines in order to achieve optimal systems regarding the prioritized objectives. To master this challenge current research is aiming for flexible inspection system concepts that are not application-specific and thus can be used for a range of industry-dependent specifications without the necessity to complete each planning phase from the beginning. This paper presents a thorough analysis of the influencing factors at the beginning of the inspection planning process of an industrial surface defect detection system for machined, automotive workpieces. Based on this analysis, design requirements for a modular framework for industrial surface defect detection system designs in this sector are derived and contextualized with available system components. Finally, an outlook for future research is presented with the goal to further increase the flexibility of industrial surface defect detection systems.

The programming of industrial robots necessitates specialized expertise and significant time and effort, particularly for small batch sizes. However, with the increasing demand for agility in production, the solutions used for robot programming have evolved significantly. Intuitive robot programming systems based on diverse concepts have been introduced to facilitate rapid deployment of robot systems. One such approved concept is no-code robot programming with finger-based gesture. In this concept, non-expert users draw a robot path via finger movement, which is subsequently translated into robot programming language to facilitate the corresponding movement. A significant challenge associated with this method is the valid replication of the corresponding robot Tool Center Point (TCP) orientation. Reachability issues, non-compliant hand contortions, and sensor occlusions make it difficult to directly derive the robot’s TCP orientation from the finger’s orientation. This work presents two novel approaches for estimating robot orientation using numerical analysis and point cloud information for finger-based robot programming without requiring prior knowledge. The first approach utilizes numerical analysis to estimate the relative robot orientation based on the geometry of the trajectory. In contrast, the second approach uses point-cloud to derive the robot orientation based on the object contour. Input shaping algorithms are employed and evaluated to reduce the divergence in the orientation estimations. Experiments demonstrate the effectiveness of the proposed approach utilizing a low-cost camera as a cost-efficient alternative to existing no-code programming strategies, potentially accelerating the real-world deployment of robotic applications in industrial environments.

Factory Planning is a multidisciplinary and complex process, which is often characterized by a large set of influencing factors and interrelated decisions. These may be supported by simulation and optimization tools for a more quantitative and validated outcome. Within the different planning phases, the required and available data base has decisive impact on the quality of results and may range from estimated, static average data to granular dynamic time series data as well as information on stochastic uncertainty. Furthermore, the different data requirements lead to the necessity of a method based on a structured process. By using Building Information Modeling (BIM) for a whole factory planning project, a more standardized design and realization process may be established. This leads to the working hypothesis that BIM could provide the methodological and technical environment in modern data-intense factory planning processes. Therefore, a framework of required input data, a so-called Information Delivery Manual (IDM) is derived as well as closing the loop by defining output information, e.g., out of simulation models that may be fed back into a common BIM-based data environment. Existing structuring approaches in factory planning are consolidated with the formalized data structuring options within the BIM-based workflows. As a result, a derived concept may serve as a generalized template for specifying IDMs in future planning projects.

Circular economy and agile manufacturing require a safe and efficient industrial robot system working in close human proximity. Although, close proximity local sensing enables safe collaboration with small cobots. However, they cannot ensure safety at high velocities with a heavy-duty industrial robot. Stereo-camera and 3D LiDAR-based touch-less global sensing methods exist, but they do not address the safety standards. This work proposes a novel method for estimating the safety parameters in speed and separation monitoring mode for 3D vision sensors. Accurate estimation of these parameters ensures compact sensing zones. Thus, enabling efficient human-robot collaboration for permanent operator presence. The method requires less effort in setup. The developed software with a graphical user interface enables workers from wide technical expertise to perform safety measurements at a precision of ±15ms in reaction time estimation. The experiments are repeatable and capture the statistical data for low error in estimation. The estimated parameters for two exemplary 3D sensors enable a human to work in close proximity of 20 cm while enabling safety from a collision.

Accurate detection of the human body and its limbs in real-time is one of the challenges toward human-robot collaboration (HRC) technology in the factory of the future. In this work, a new algorithm has been developed to fuse thermal, depth, and color information. Starting with the calibration of the sensors to each other, proceeding with data fusion and detection of human body parts, and ending with pose estimation. The proposed approach has been tested in various HRC scenarios. It has shown a significant decrease in errors and noise in industrial environments. Furthermore, it produces not only a higher positioning accuracy of the head and hands compared to the state of art algorithms (e.g., OpenPose), but it is also faster. Hence, such an algorithm could be joined with digital twins for safe and efficient human-robot collaboration.

In this paper, the authors deal with the questions of use case-oriented process identification. Based on a real world example, implementation problems such as experiment planning under real production conditions, data preparation and pre-processing, selection and application of an appropriate analysis technique as well as its result interpretation are discussed. A specialized software tool supports the key steps of the process identification and controller design. The advantages of a holistic software-supported approach of guided modeling and subsequent controller design to simplify the entire process are demonstrated.

The development of digital process twins goes hand in hand with the digitalization of manufacturing's production chain, especially in the machining of parts. The essential basis of these digital process twins is the information model, which describes the properties and relationships of all relevant data and information required to realize the processing task and a digital representation. This paper describes such an information model and presents the benefits and application potentials of digital process twins in the machining of parts.

In order to guarantee the quality and the productivity of a production system in a competitive marketplace, it is important to be able anticipate the changes in specifications of products and systems. The time limits in running productions, the complexity of manufacturing systems, and the diversification of components, are the challenges that human experts cannot handle without cognitive systems. Capabilities of cognitive Systems in observing, learning, and predicting the behavior and the evolution of the manufacturing systems make them special candidates for solving these problems. This mapping study provides an insight into the application of cognitive systems in the domain of production. We categorize different approaches and estimate their progress. We also discuss the optimizations and persisting problems and barriers. These representations can help in recognizing the concrete problems of the field. According to the results of our mapping study, Human-Machine Interaction and Knowledge Gaining/Sharing represents the largest categories of the domain. A gain in efficiency and maximized effectiveness can be achieved as optimization. The most common problem is the missing or only difficult generalization of the presented concepts.