When optimizing the micro geometry in the development process of a gear set, the deflection behavior under load situations is important to predict the realistic component life span and acoustic behavior. Most common experimental deflection tests are extraordinarily complex and need special preparation of the whole drive unit to be able to measure the deflection precisely. On the other hand, purely simulation approaches need to be validated because the models are getting extensive through many boundary conditions between the different components of the drive unit. To address these challenges, this paper presents a method for measuring the deflection of hypoid gear sets by an evaluation of thermographic contact pattern. This method is based on measuring the contact pattern via a thermographic camera and transferring the optical information of the images into objective parameters to describe the properties of the contact pattern such as position, size and inner temperature. Through experimental study the influence of different load and rotation speed combination on the contact pattern are investigated. With increasing load, the size and the inner temperature of the contact pattern rise and a deflection to the heel is detected for drive and coast flank. Using a linear approach, the load and contact pattern parameters are showing a high correlation (R2 > 0.9) coefficient. Although a higher rotational speed did not show any influence on the positioning of the contact pattern, an increased temperature leads to an enlargement contact pattern size. This is considered as disruptive effect which can be compensated by optimizing segmentation threshold in the image processing. As part of the publication, the experimental results are compared to a loaded tooth contact analysis (LTCA) conducted with the software BECAL. After optimizing the deflection for LTCA, the results in the different loading points are showing a high agreement in a direct comparison.

Quality predictions have great potential in production. However, establishing predictive quality models requires to understand and therefore explain the predictions generated by these models based on production data. Model-independent interpretation methods claim to deliver such insights. As to date there is no approach for assessing the suitability of these methods in production context, we provide a methodology to evaluate them against general requirements as well as requirements specific for the context of process- and product-centered quality prediction. The results show that the proposed methodology is capable to effectively evaluate the interpretation methods based on selected criteria. (c) 2024 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0)

In production environments, failure management and process optimization are essential elements for preventing interruptions and fulfilling customer expectations in terms of product quality. In this context, not only the establishment of standardized failure management and process optimization processes are of central importance, but also the use of modern analysis methods for the efficient identification and correction of product and process failures. Advances in the field of machine learning enable the prediction of quality characteristics during production, supporting in the identification of potential defects. However, merely predicting defects proves insufficient for supporting the decision-making process toward specific measures for process optimization.This paper examines the information demand for decision support through predictive quality models in production by means of a targeted information requirements analysis. For this purpose, established methods for failure analysis and process optimization, such as fault tree analysis, 8D-reports, and design of experiments, were evaluated. Thereupon, ten hypotheses with two additional research questions regarding the information demand were derived. Based on the derived hypotheses and questions, a cross-industry semi-structured expert survey (n=33) was conducted, followed by an evaluation of the results with regards to the information requirements for process-oriented decision support.The survey results (Cronbach’s alpha=0.77) confirm that predicting possible deviations and defects alone is insufficient for supporting decision-making process. The results indicate that additional information is required to support decision-making through predictive quality models with the objective of optimising the production process. The identification of important characteristics and interactions as well as the associated effect size is of paramount importance. Regarding the requirements for the provision of information via predictive quality models, the study reveals that the correctness of the determined causes, as well as the repeatability, real-time capability, and scalability of the analysis, along with independence from the underlying predictive quality models, are required.

Additive manufacturing (AM), especially the extrusion-based process, has many process parameters which influence the resulting part properties. Those parameters have complex interdependencies and are therefore difficult if not impossible to model analytically. Machine learning (ML) is a promising approach to find suitable combinations of process parameters for manufacturing a part with desired properties without having to analytically model the process in its entirety. However, ML-based approaches are typically black box models. Therefore, it is difficult to verify their output and to derive process knowledge from such approaches. This study uses interpretable machine learning methods to derive process knowledge from interpreted data sets by analyzing the model’s feature importance. Using fused layer modeling (FLM) as an exemplary manufacturing technology, it is shown that the process can be characterized entirely. Therefore, sweet spots for process parameters can be determined objectively. Additionally, interactions between parameters are discovered, and the basis for further investigations is established.

In the context of predictive quality in production, there is a need to explain and understand the predictive models used, as well as the dependencies present in the underlying data. For this purpose, we develop a framework for model-independent interpretation of predictive models to enable data-driven, process- and product-oriented decisions. This framework combines different model-agnostic interpretation methods and structures them into two modules, one for a process-oriented view and the other for a product-oriented view. In addition, an implementation concept for the two modules in a machine learning pipeline is also provided.

Failure management is essential in production environments to prevent disruptions and satisfy customers' expectations of product quality. Of central importance is not only the establishment of standardized failure management processes but also the use of modern analysis methods for the efficient correction of product and process failures. Due to the high complexity of production environments and failure management at the same time, the human factor still plays a central role in detecting, analyzing, and sustainable correcting product and process failures. However, according to a previous empirical study, around 50 % of all information collected in manufacturing companies remains unused. The study also confirms that around 70 % of the companies consulted are not aware of the information required for targeted analyses.

Due to the trend of electrification in the automotive industry, the economic production of electric drives with high acoustic quality requirements is a crucial factor to stay competitive in the global market. Low noise levels in the interior are an important criterion for the perceived quality of electric vehicles. Consequently, the noise generated by mounted gear components within integrated electric drive topologies must be minimized. Gears with unavoidable manufacturing deviations are usually randomly assembled, leading to random non-defined gear-related acoustic properties of the assembled electric drive. Furthermore, parameters of the gear manufacturing machines do not dynamically adapt to unknown changes in the production system leading to non-ideal quality output. To address these challenges, this paper presents a self-optimization concept in gear manufacturing and assembly in the production of electric drives by cognition enhanced control. A digital twin is developed which estimates the transmission error based on inline measurements. Through optimization, an optimal selection of gear pairs is achieved. Based on quality predictions, adaptive control of the gear manufacturing process can be implemented, leading towards a closed-loop self-optimization of the production system. The concept is developed and validated using an exemplary use case from the commercial vehicle industry.

Causal inference is a fundamental research topic for discovering the cause–effect relationships in many disciplines. Inferring causality means identifying asymmetric relations between two variables. In real-world systems, e.g., finance, healthcare, and industrial processes, time series data from sensors and other data sources offer an especially good basis to infer causal relationships. Therefore, many different time series causal inference algorithms have been proposed in recent years. However, not all algorithms are equally well-suited for a given dataset. For instance, some approaches may only be able to identify linear relationships, while others are applicable for non-linearities. Algorithms further vary in their sensitivity to noise and their ability to infer causal information from coupled vs. non-coupled time series. As a consequence, different algorithms often generate different causal relationships for the same input. In order to achieve a more robust causal inference result, this publication proposes a novel data-driven two-phase multi-split causal ensemble model to combine the strengths of different causality base algorithms. In comparison to existing approaches, the proposed ensemble method reduces the influence of noise through a data partitioning scheme in a first phase. To achieve this, the data are initially divided into several partitions and the base causal inference algorithms are applied to each partition. Subsequently, Gaussian mixture models are used to identify the causal relationships derived from the different partitions that are likely to be valid. In the second phase, the identified relationships from each base algorithm are then merged based on three combination rules. The proposed ensemble approach is evaluated using multiple metrics, among them a newly developed evaluation index for causal ensemble approaches. We perform experiments using three synthetic datasets with different volumes and complexity, which have been specifically designed to test causality detection methods under different circumstances while knowing the ground truth causal relationships. In these experiments, our causality ensemble outperforms each of its base algorithms. In practical applications, the use of the proposed method could hence lead to more robust and reliable causality results.

This paper examines and benchmarks different approaches in their ability to detect and predict faults in manufacturing processes, based on real-world use cases and with respect to their differing dataset properties. Knowing about the occurrence of faults becomes more and more important in manufacturing due to increasing quality demands and legal guidelines. In addition, the complexity of manufacturing processes is constantly increasing. This stems from a higher product variance resulting from individual and customized products as well as additional external influences such as human errors, environmental factors and tool wear. As a result, today’s process data is often no longer normal distributed. Furthermore, data volume steadily increases, thereby opening new opportunities for data-driven analytics approaches. Frequently applied control charts for statistical process control (SPC) often lack the ability to deal with multiple variables and non-normal distributed data at the same time, since multivariate and nonparametric control charts are underrepresented in past research. Consequently, there is a need for new process control methods in manufacturing that are suitable for large amounts of data and cover diverse and dynamic distribution models. Therefore, machine learning models have been recognized as feasible approaches to meet these requirements. For comparison a Hotelling’s T2 control chart, a K-Chart, an Isolation Forest, an ARIMAX model and a Neural Network have been implemented. We evaluate each method by missed detection rate (MDR), false alarm rate (FAR) and whether signals occurred before or after the faults. Real-world data sets of a commercial vehicle manufacturer serve as benchmarking basis.

AbstractIn short-term production management of the Internet of Production (IoP) the vision of a Production Control Center is pursued, in which interlinked decision-support applications contribute to increasing decision-making quality and speed. The applications developed focus in particular on use cases near the shop floor with an emphasis on the key topics of production planning and control, production system configuration, and quality control loops.Within the Predictive Quality application, predictive models are used to derive insights from production data and subsequently improve the process- and product-related quality as well as enable automated Root Cause Analysis. The Parameter Prediction application uses invertible neural networks to predict process parameters that can be used to produce components with desired quality properties. The application Production Scheduling investigates the feasibility of applying reinforcement learning to common scheduling tasks in production and compares the performance of trained reinforcement learning agents to traditional methods. In the two applications Deviation Detection and Process Analyzer, the potentials of process mining in the context of production management are investigated. While the Deviation Detection application is designed toidentify and mitigate performance and compliance deviations in production systems, the Process Analyzer concept enables the semi-automated detection of weaknesses in business and production processes utilizing event logs.With regard to the overall vision of the IoP, the developed applications contribute significantly to the intended interdisciplinary of production and information technology. For example, application-specific digital shadows are drafted based on the ongoing research work, and the applications are prototypically embedded in the IoP.

Abstract Den Kern des Exzellenzclusters Internet of Production bildet die domänen- und disziplinübergreifende Forschung in der Produktionstechnik. Der Fokus der Gruppe Short-Term Production Management liegt dabei insbesondere auf der Erhöhung von Entscheidungsqualität und -geschwindigkeit im Produktionsumfeld durch die datenbasierte Unterstützung der Anwender:innen. Dazu werden geeignete, kontextspezifische Daten aus Entwicklung, Produktion und Anwendung in Echtzeit und mit angemessener Granularität bereitgestellt, zusammengeführt und analysiert.

Data acquired during production processes often contain redundant and irrelevant features. Thus, to predict quality characteristics from process data, precise feature extraction is essential to sustain a low prediction error and to limit the computational complexity of the deployed machine learning models.

The quality of production planning and control in production companies largely depends on the accuracy of the database regarding the status quo of the production process, as well as the time required for each production step. Especially in single and small batch production, it is difficult to create a reliable database due to the large product spectrum and the variation in the production process. For this reason, this paper examines the potential of Bluetooth Low Energy (BLE) Beacons to create a low-cost real-time locating system (RTLS). In comparison to other existing RTLS solutions like RFID, BLE stands out through very low hardware costs and a high compatibility to almost every mobile device. Through a signal strength based approach in combination with probabilistic neural networks, the RTLS has been able to achieve a positioning accuracy with a mean absolute error of up to 0,74 m in an industry-like environment.

Die Umsetzung von Industrie 4.0 prägt den Wettbewerb produzierender Unternehmen auf globalen Märkten. Wer in diesem Wettbewerb dauerhaft eine Vorreiterrolle einnehmen will, ist gefordert, die Potenziale der Digitalisierung größtmöglich zu realisieren. Die konsequente Nutzung des impliziten Wissens, welches in den exponentiell ansteigenden unternehmerischen Datenmengen steckt, führt über eine starke Hebelwirkung zur kontinuierlichen Verbesserung der produkt- und prozessbezogenen Qualität. Gleichzeit adressiert eine qualitätsgetriebene Optimierung des Ressourceneinsatzes die stetig zunehmenden Nachhaltigkeitsforderungen aus Bevölkerung und Politik. Der rasante Anstieg der Datenverfügbarkeit resultiert zum einen aus einer wachsenden Vernetzung von Lieferanten, Produzenten und Kunden und zum anderen aus der Nutzung einer steigenden Anzahl unterschiedlicher Informationskanäle, die von integrierter Sensorik bis zu Online-Produktreviews reicht.

In this paper, we present a new framework for process optimizations, the Data Analytics Production Line Optimization Model (DAPLOM). Due to increasing efforts in the digitalization of production systems, an extensive amount of production data is available for analytics. This data can be used for the optimization of production lines and the prediction of their performance (e.g. drift of parameters or component quality) in order to achieve economic and technical improvements. The demand for systematical usage of data-driven methods involving technologies like Data Analytics and Machine Learning and the combination of engineering approaches is growing continuously. DAPLOM guides the implementation process of IT supported problem-solving solutions in production environments. It combines classical process- with data-driven approaches. Specific focus lies on achieving a holistic perspective with a macro- as well as a microscopic view on the given conditions. Here the macroscopic view covers the general material flow, whereas microscopic view considers process details. Additionally, DAPLOM provides useful methods in a step-by-step procedure structured in seven phases. The framework is validated in an industrial use case of an automated wire bending process. Thus, the effectiveness of the framework is demonstrated and further development potentials are identified.

This paper presents a data-driven approach for improving the process quality of production systems. Therefore, the product quality is detected during the production process. The worker is provided with reasonable parameter recommendations about the production process as decision support to improve the process quality. To achieve this, a cross-process data analysis of the process and quality data is carried out using decision trees. The results are visualized in a comprehensible form for the worker. Based on a case study from mass production, the approach is evaluated and its performance is demonstrated in comparison to classical statistical methods.

Purpose: This paper provides a domain specific concept to assess data suitability of various data sources along the production chain for defect prediction. Methodology/Approach: A seven-phase methodology is developed in which the data suitability for defect prediction in interlinked production steps is assessed. For this purpose, the manufacturing process is mapped and potential influencing variables on the origin of defects are identified. The available data is evaluated and quantified with regard to the criteria relevancy, completeness, appropriate amount of data, accessibility and interpretability. The individual assessments are then visualized in an overview, gaps in data acquisition are identified and needs for action are derived. Findings: The research shows a seven-phase methodology to systematically assess data suitability for defect prediction and identify data gaps in interlinked production steps. Research Limitation/implication: This research is limited to the analysis of contextual data quality for the use case of defect prediction. Other data analytics applications or processes outside of manufacturing are not included. Originality/Value of paper: The paper provides a new approach to identify gaps in data acquisition by systematically assessing data suitability for defect prediction and deducting needs for action. The accuracy of predictive defect models is then to be improved by the subsequent optimization of the data basis.