Projects
AdjUST – Automatisierung in der Konfiguration von Unternehmensinformationssystemen
Contact Person: Matthias Olscher
Duration: September 2020 – December 2022
In the course of digitalization and globalization, companies in the textile industry are faced with the challenge of professionalizing their processes. This concerns the core processes in administration and production as well as, for example, the processes in sales. As smaller companies in particular only have small budgets, individual developments are usually ruled out, and so is the adaption of expensive standard software. To meet these special requirements, innovative techniques of process analysis and artificial intelligence are used in the described project to enable companies to use software solutions for process handling in the most resource-efficient way possible. The execution data generated from system use is used by the system operator to successively develop inductive reference models for the textile industry, from which other companies then can benefit. At the same time, AI technologies are used to enable automated, intelligent customization of the system for new customers. The goal is to provide companies with an ERP system that is easy to use. By making (anonymized) usage data available, best practices can be disseminated within the industry and thus processes that do not primarily add value can be digitized cost-effectively.
APPaM – Automated Process Planning and Mining
Contact Person: Peter Fettke
: August 2020 – July 2023
The correct execution of business processes is nowadays already being examined in companies with the help of process mining tools. Based on the execution logs of the IT systems, these tools develop human-readable process models that reflect the current state. For the readability of these models, the real executions are simplified, and atomic events are combined into generic alternatives. As a result, these process models are no longer suitable for intelligent planning or automation. Besides this, forecasting models which predict future events already exist, but the conversion into a concrete recommendation for action is still left to the employees. For effective support, a system is needed that automatically recognizes business processes from the system logs and converts them directly into cost-optimized recommendations for action to support employees in their planning.
ErgoBest – Erhebung ergonomischer Best Practices in industriellen Arbeitsprozessen mittels Internet-of-Things und Mixed Reality
Contact Person: Andreas Emrich
: June 2021 – May 2023
The ErgoBest project is developing an intelligent assistance system that combines process information and real-time data from a sensor network. The sensor network consists of smart wearables as well as sensor technology integrated into workstations. Based on this, the system recognizes ergonomically critical situations for employees in physically demanding jobs and optimizes them by providing immediate feedback with mixed reality glasses that offer tips, for example, on how to lift objects correctly. The results, supported by AI-based algorithms, will be used to identify best practices for intelligent employee relief. The development of such a networked infrastructure of sensors is carried out by an interdisciplinary consortium of research and industry.
ExPro – KOOPERATIVE EXPLORATIONSUMGEBUNG VON MACHINE-LEARNING-PROGNOSEN AM BEISPIEL DER PRODUKTIONSPLANUNG
Contact Person: Nijat Mehdiyev
: April 2021 – September 2023
Taking equal account of XAI methods and profound methods of human-computer interaction (HCI), the ExPro project pursues the goal of developing an ML module for predicting production-relevant key performance indicators (KPI) based on historical company data and production plans in a way that is comprehensible to the end user; the comprehensibility of the forecasts is ensured with local explanatory approaches. In order to implement this, ML procedures and XAI analytics must be adapted within the project for application to structured data. The resulting XAI concept will not be limited to highlighting influencing factors in visualized data, but based on this, it will focus on interactive exploration by users as an innovative core element of the comprehensibility of ML predictions. In order to make the use of these methods on the user interface manageable for end-users, methods and experiences from the areas of end-user development (e.g., visual programming, adaptability interfaces, collaborative tailoring) and appropriation support (e.g., group and usage histories, usage discourse environments, extensible help systems) need to be adapted to the specifics of ML architectures, especially to cope with the differences of heuristic computation methods compared to traditional deterministic program structures.
KEA-Mod – Kompetenzorientiertes E-Assessment für die grafische Modellierung
Contact Person: Peter Fettke
: November 2019 – October 2022
Graphic modulation is an established part of information systems higher education and numerous related study programs. The joint project KEA-Mod aims to develop a digital subject concept that improves the teaching quality of graphical modeling. To this end, tools that have so far been isolated from one another and used locally by the collaborative partners, such as task generators, feedback, and assessment systems, will be combined in a uniform overall system and further developed in an application-oriented manner. The result is an “e-assessment platform” that covers various teaching-learning scenarios such as lectures, exercises, and exams. A central feature of the platform is its transferability: It should be possible to use it at different universities in Germany. The use of the platform is accompanied by qualitative and qualitative methods of evaluation.
KIMonoS – KI-gestützte Mobility-On-Demand-Plattform im Saarland
Contact Person: Oliver Gutermuth
: September 2021 – February 2024
The objective of the KIMonoS project is to develop a mobility-on-demand platform solution based on artificial intelligence (AI) and to test it as a prototype at the municipal level in Saarland. This is intended to serve the needs of passengers in rural areas better and at the same time increase the economic efficiency and sustainability of transport operations in this environment. To achieve this, parts of the local public transport system (LPT) should be able to respond flexibly to demand situations. The concept involves decoupling public transport capacities from fixed-route services and managing them with an AI-supported mobility platform. This platform provides passengers with information about possible connections, handles bookings, and allows new connections to be integrated into a dynamic timetable at short notice. In the process, connections that have already been committed, the capacities of available vehicles, and the economic interest of the transport company must be taken into account. Data on demand and traffic volumes can be collected and used for forecasting. Accordingly, the processing of numerous information and a weighing of alternatives for the creation of flexible mobility offers is required that can be realized by recent technical developments of machine learning and AI.
Kosmox – Entwicklung einer neuartigen lokalen kontrafaktischen Erklärungsmethode und – Schnittstelle unter Berücksichtigung kognitiver Modellierungsansätze
Contact Person: Nijat Mehdiyev; Lea Götz
Duration: January 2020 – June 2022
Due to the enormous progress in the field of artificial intelligence (AI) in recent years, more and more of these AI technologies are being used – especially in the field of machine learning (ML), they can serve as decision support for users. For the analysis of the data and thus the use of the technology, a downstream verification, validation and interpretation of the results by the user is necessary – however, due to the black box character, there is currently a lack of transparency and explainability. A robust Explainable Artificial Intelligence (XAI) system should support the decision maker in understanding the decisions made by the system and identify potential improvements to achieve a desired outcome in the future based on the ML model used. To guarantee these functions, a local post-hoc explanation system is to be developed in the project. This is characterised by a semantic integration of two complementary explanatory approaches, local rule-based and simulation-based causal. Within the framework of the KOSMOX project, an explanatory interface is to be developed by incorporating findings from cognitive, IS, organisational sciences, HCI and behavioural economics. This should enable interactive communication between the end user and the used AI techniques or their explanations. The user should be supported in understanding the ML results for the decision-making process.
MDZ – Mittelstand-Digital Zentrum Kaiserslautern
Contact Person: Nina Obreschkova
: 08/01/2021 – 07/31/2024
Das Mittelstand-Digital Zentrum Kaiserslautern bietet kleinen und mittleren Unternehmen aus Rheinland-Pfalz und darüber hinaus kostenfreie Unterstützung in Form von Informationen und Anregungen bis hin zur Unterstützung bei der Umsetzung ihrer Projekte. Das Mittelstand-Digital Zentrum Kaiserslautern gehört zu Mittelstand-Digital. Mit dem Mittelstand-Digital Netzwerk unterstützt das Bundesministerium für Wirtschaft und Energie die Digitalisierung in kleinen und mittleren Unternehmen und dem Handwerk.
RACKET – Rare Class Learning and Unknown Events Detection for Flexible Production
Contact Person: Peter Fettke
Duration: September 2020 – August 2024
The RACKET project addresses the problem of detecting rare and unknown faults by combining model-based and machine learning methods. The approach is based on the assumption that a physical or procedural model of a manufacturing plant is available, which is not fully specified and has uncertainties in structure, parameters and variables. Gaps and errors in this model are detected by machine learning and corrected, resulting in a more realistic process model (nominal model). This model can be used to simulate system behavior and estimate the future characteristics of a product. Actual product defects can thus be attributed to anomalies in (Nominalmodell) the output signal and to inconsistencies in the process variables, without the need for a known failure event or an accurate failure model. Errors have a wide range, i.e., geometric errors such as scratches, out-of-tolerance dimensional variables, or dynamic errors such as deviations between estimated and actual product position on a conveyor belt, process steps or incorrect path assignment in the production flow, etc., and can occur at the product and process level.
SensoBike – Modell zur Erkennung von Fehlstellungen des Knies
Contact Person: Andreas Emrich; Michael Frey
: March 2021 – February 2023
A training and mobility system for orthopaedic prevention integrated into everyday life, using the example of knee arthrosis. The system is based on a pedelec ergometer that, to begin with, controls the user’s load depending on the knee position. The orthopaedic pedelec ergometer enables users to train in a controlled and safe way without fear of overload thanks to real-time feedback, to stay motivated because training sessions can be varied and carried out outdoors and the training progress can be measured, to train sustainably because training sessions can be easily integrated into everyday life, e.g. on the way to work, during leisure time, etc., to maintain or increase their mobility in everyday life, to train in groups with different levels of fitness, and to provide doctors and therapists with the necessary information: Plan, monitor, correct and document training sessions more flexibly.
SmartVigilance – Regulatorische Compliance durch KI-basierte Umfeldüberwachung in der Medizintechnik
Contact Person:
Medical device manufacturers are subjected to strict regulations regarding the safety of their products. Regulations affect both product approval and the post-market phase of approved medical devices. Companies are obliged to monitor the use and application of their products on the market and to take appropriate measures to eliminate defects or minimize risks. The aim of the project is the prototypical development of technologies and automated procedures for the regulatory required marked monitoring and risk assessment in medical technology. An internet based “SmartViligance” platform will automatically capture, analyze, and report to manufacturers dangerous incidents and product defects that are reported by users to regulators and are publicly available. The project uses artificial intelligence methods and technologies – natural language processing (NLP), machine learning (ML) and data analytics. The platform is used to monito the environment (“vigilance”); it is intended to make market monitoring (“post-market surveillance”) more reliable and to support and relieve companies in the medical technology sector in the process.
TRIPLEADAPT – Adaptive, interoperationale Weiterbildungsplattformen durch einen Digitalen Drilling
Contact Person: Peter Fettke; Sarah Rübel
: May 2021 – April 2024
The overall objective of the project is the participatory development of a digital triplet involving all relevant operational stakeholders and taking their interests into account. Here, the concept of the digital twin is expanded to include a digital learning environment in which employees are supported in real problems and obstacles in their work processes. The learning concepts arise from the analysis and modelling of process data, which are oriented towards developed competence profiles and match these anonymously with the actual competence profiles.
VisEP – Visualisierung von Entscheidungen in Prozessen
Contact Person: Alexander Beuther
: March 2021 – August 2022
The aim of the VisEP project (Visualization of Decisions in Processes) is to make current research results from decision mining and process management usable for knowledge-intensive processes. The project aims at the development of a prototype for the extraction of data from application systems of knowledge-intensive services of, for example, tax consultants, in order to represent decisions as well as their effects on the processes. In particular, the temporal, sequential and logical relationships between activities should be used to determine their impact on the further course of the process. In this way, the implicitly contained decisions are to be uncovered and made usable for the identification of anomalies or used for future process predictions. To solve the problem described above, the project “VisEP” will use process mining and artificial intelligence methods to first identify the process-influencing points in the process and divide them into relevant groups or clusters of decisions and, based on this, extract the underlying decision logics. Since decisions made by artificial intelligence are often not transparent, a visualization component is being created that is intended to display decision parameters using explainable AI, among other things. This should result in improved process visualization and process knowledge for knowledge-intensive processes.