Following the development of the prototype in the summer semester of 2023, the Pyramid app has now been officially published in the app stores. The app modernises communication between students and companies at Augsburg University of Applied Sciences' Pyramid Fair by supplementing existing sources of information such as the website, print and PDF catalogues. →
A machine learning application predicts patient volumes and thus facilitates nursing staff scheduling in order to efficiently manage staff shortages.
In cooperation with XITASO GmbH, an application was developed that uses machine learning to predict and visualise patient volumes in order to improve nursing staff scheduling. Based on historical data, the time series forecasting method is used, whereby models such as Random Forest, SARIMA and Holts Winters seasonal method are utilised. The model quality is evaluated by RMSE, MAPE and MAE. The predictions are presented in a user-friendly web interface, which enables care facilities to optimise staff planning in line with demand. →
Employees at swa Netze lack an overview of the stock of usable spare parts and emergency materials. A tool for cross-departmental inventory management would be desirable, with which material stocks can be consolidated and sustainably reused in order to save money and resources. →
In the KI-HEAP4.0 project, the production system of an automotive component manufacturer is represented as a digital twin.
This digital twin serves as a simulation environment to support decision-making in detailed production scheduling, using a hybrid approach that combines mathematical optimization with reinforcement learning. →
This paper presents a conceptual approach and a successful implementation guide for a digital twin of a modular manufacturing system. →
The increasing demand for components with tailored properties and functionally graded materials featuring adjustable interfaces is raising the requirements in gravity sand and die casting processes of aluminium. In response to these challenges, a novel controlled process based on injector casting is being developed. →
In collaboration with XITASO, we have developed a trade fair showcase that presents object recognition and segmentation using neural networks. By using RGB and infrared cameras, road users such as cyclists and pedestrians are recognised under different environmental conditions. Pre-trained models such as YOLOv8 and Mask R-CNN are used, which have been subsequently fine-tuned to infrared data to enable reliable detection even in poor lighting conditions. →
The German Academic Exchange Service (DAAD) supports the program "Study@HSA - Support Initiative German Schools Abroad and Partner Schools (BIDS) 2023-25" with funds from the Federal Foreign Office (AA). The program, which is integrated into the AA's Partner School Initiative (PASCH), aims to work at the interface between PASCH schools and German universities. With the help of BIDS, German universities build networks with PASCH schools abroad and support their graduates in their transition to studying in Germany. This includes intensive information and support at school, preparatory measures, advice during the application phase and intensive support, especially in the first two semesters. →
Plan graphic 2.0 →
