In the context of Industry 4.0, constantly evolving production systems generate the need for a highly adaptive and autonomous automation system with lean maintenance, minimum downtime, maximum reliability, and resilience. Due to the high complexity of production environments future manufacturing execution systems (MES) will have to be distributed physically and logically to cope with the arisen complexity and dynamic enough to adapt to the changes in the production processes execution. However, such systems lack a global overview and the possibility to centrally intervene to assure the quality of the product and/or the process.
The goal of the project is to develop new methodologies for quality assurance in versatile heterogeneous production systems with the focus on predictive analysis, autonomous health monitoring, and management of the operation processes to reduce system reaction time and engineering efforts by adapting following production steps to compensate for the deviations. To this end, we develop a simulation framework, which enables digital production process planning and validation. In particular, we address the following aspects:
Dynamic clustering to the relevant environment discovery: Key elements in this area are new dynamic cluster building concepts and methods allowing effective negotiation and propagation of mitigation measures as well as representation of the overall production environment, beyond its locally accessible information. Autonomous knowledge extraction to increase adaptivity and reduce engineering effort: Semantic enhancement using minimum a priori knowledge requires intelligent data analysis, such as various integration and reduction techniques. Self- and context-awareness, data abstraction and scattered-data fusion, and machine learning techniques are the main methods considered for this task bringing added value by enhancing adaptivity and minimizing engineering-effort-heavy tasks of defining suitable abstraction and knowledge extraction methods. Fault diagnosis and prognosis for quality assurance and resilience enhancement: In a distributed system no single (sub)system knows the complete state of the overall system. Therefore, additional measures need to be taken to detect, analyze, predict, and/or mitigate errors, faults, and failures. Proper communication infrastructure, smart data fusion, and learning are key elements of this task. Cognitive decision-making for enhanced intelligence: In complex manufacturing systems (CPPS) thorough and precise modeling of the system and its environment is challenging. Cognitive systems are adept at taking decisions efficiently despite the lack of a complete or precise model. Cognitive decision-making improves the efficiency of the system in using extracted knowledge about each agent, neighbors in the cluster, the overall system, and possible courses of action, to re- or pro-actively change the CPPS. Especially by mitigation of failures, degraded health, or anomalies, it ensures the quality of the product and process in adaptive heterogeneous manufacturing systems.
We will test SAVE methodologies in an IEC 61499 based simulation environment and use an adaptive automobile motor production use-case to evaluate them regarding their functionality and efficiency.
|Duration||01/04/2018 - 31/03/2021|
|Program||Produktion der Zukunft|
|Principle investigator for the project (Danube University Krems)||Priv.-Doz.Dipl.-Ing.Dr. Thilo Sauter|
Comparison of Preprocessors for Machine Learning in the Predictive Maintenance Domain
2020 IEEE 29th International Symposium on Industrial Electronics (ISIE), 17/06/2020
Artificial Intelligence for Process Monitoring
21st International Workshop on Computer Science and Information Technologies, TU Wien, Wien, Österreich, 02/11/2019
Self-monitoring based process adaptation in flexible manufacturing
Austrian -Finnish competence & experience exchange in RDI, Tampere, Finnland, 05/06/2019