The development within an SFG entails specific challenges. The heterogeneity of the MMUs, the flexibility of the work processes and the diversity of variants result in new operating concepts for machines and components. A precise forecast of the system states requires an in-depth understanding of the future load profiles. However, these loads are highly context-dependent, which makes them difficult to predict. The associated uncertainties lead to a limited accuracy of the forecasts.

Initial approaches to forecasting future load profiles are based on historical and current data and the consideration of different operating points. Despite a variety of model-based, data-based and hybrid methods, prediction under time-varying operating conditions remains a key challenge. In addition, limited data quality and availability make forecasting more difficult, especially in the case of incomplete or censored processes. To date, there is no comprehensive methodology to overcome these problems.

The chosen research approach aims to better understand and precisely predict the relationships between load and degradation. The focus is on developing a methodology for predicting future load profiles. A key focus is on considering the different types of time-varying load profiles, taking into account both known and unknown as well as discrete and continuous profiles. The integration of these profiles into the methodology is a key focus.

Furthermore, a focus is placed on the challenges of incomplete, censored and qualitatively limited data, which play a central role especially for data-driven models. Research into innovative approaches such as data augmentation and data generation aims to expand the database and increase the robustness of the models. The research approach thus strives for a holistic solution that addresses the complexity of variable load profiles and the quality of the available data.