In safety-critical applications such as automotive or railway technology, prediction of the condition and remaining life of electronic components is a major challenge. To determine the condition, a comprehensive digital fingerprint of the components is needed, which already starts during the production phase. Currently, inspection data such as SPI, AOI and AXI data are used to group the components into two groups ("OK" and "not OK"). In the future, however, this inspection data will also be used to predict the service life and current condition of the assembly. This article deals with the investigation and processing of the imaging inspection data, as well as the geometry parameters determined from it. First, a dimensional reduction of the input data is carried out with the help of mathematical methods (PCA, t-SNE and Autoencoder). This reduction extracts the relevant information from the input data from the point of view of the AI. So far, clustering methods such as K-Means, DBSCAN and Hierarchical Clustering have been used to generate groups. In the future, these groupings can be used to predict the service life together with the loads on the components and thus make a statement about the current condition of the components

Machine learning implementations in an industrial setting poses various challenges due to the heterogeneous nature of the data sources. A classical machine learning algorithm cannot adapt to dynamic changes in the environment, such as the addition, removal, or failure of a data source. However, to handle heterogeneous data and the challenges coming with this, it is a mandatory capability to build robust and adaptive machine learning models for industrial applications. In this work, a novel architecture for robust and adaptive machine learning is proposed to address these challenges. For this, an architecture consisting of different modular layers is developed, where different models can be easily plugged in. The architecture can handle heterogeneous data with different fusion techniques, which are discussed and evaluated in this paper. The proposed architecture is then evaluated on two public datasets for condition monitoring of automation systems to prove its robustness and adaptiveness. The architecture is compared with baseline models and shows more robust performance in case of failing/removed data sources. In addition, new data sources can easily be added without the need to retrain the whole model. Furthermore, the architecture can detect and locate faulty data sources.

Advancements in the Internet of Things (IoT) applied by automation systems lead to an increasing and flexible interconnection of various devices and the generation of voluminous data. In the domain of home automation for instance, having a multitude of interconnected data leveraged to knowledge could enhance user-centeredness and energy efficiency, among other use cases. However, the data as well as data sources are dynamic and heterogeneous, posing a great challenge to unfold their full capabilities. Thus, a unifying and correspondingly dynamic architecture is needed for supporting the flexibility and extensibility of such systems and data sources. In this paper, we present the concept and implementation of a context-aware INTElligent LIVing (inteLiv) architecture. Consisting of a data layer, a context layer and a service layer, the architecture abstracts the heterogeneous data of the linked devices via a middleware. The resulting system enables the collection of data across networked heterogeneous devices and models the collected data by using a unified dynamic context model. A hybrid approach was chosen for the designed context model: A combination of an ontology-based and property graph-based model represent a core aspect of this work as opposed to common static ontological representations and pre-defined context-based use cases. In the following contribution, the inteLiv ontology is first defined to describe the concepts of the intelligent living system. Based on the modeled context, reasoning algorithms also enable the derivation of new context information. The context is stored as a knowledge graph and can be used for context-aware applications, which can be designed during runtime. Furthermore, a user interaction with the system is possible via a web interface.

Failure analysis is essential for improving the reliability and manufacturability of electronic devices. With the time-domain reflectometry method, failures can be analyzed non-destructively. The method enables the detection, location, and characterization of hard interconnection failures (open or shorts) as well as of soft interconnection failures, which can give an outlook on imminent hard failures. Generating measurement data from real failed devices is costly since failed devices need to be selected and the measurements need to be performed and prepared. In contrast, simulation models are often available where all possible kinds of failures can be created. Therefore, we propose simulation to real transfer learning for the failure analysis on time-domain reflectometry data. A deep learning model shall first be trained on time-domain reflectometry simulation data and then be transferred to measurement data of a power transistor. We investigate different possibilities of transfer and evaluate the performance of a SiC power transistor.

Current industrial automation systems are facing increasing dynamics. Thus, acquiring and managing heterogeneous data within the Digital Twin to enable decision making is necessary although challenging. Knowledge Graphs unify and relate data, enabling the derivation of new insights. In this contribution, an approach for context-enriched modeling of cyber-physical production systems is proposed, in order to realize a Knowledge Graph enhanced intelligent Digital Twin further considering the context. Therefore, the modeling approach of the Knowledge Graph considers context and serves as a base for the graph embeddings to gain further knowledge about the production system. This knowledge is used within the architecture of the intelligent Digital Twin. The resulting benefits, i.e. diverse manifestations of an improved decision making, are highlighted by the use case of self-organized reconfiguration management.