Physics-informed machine learning is particularly relevant in the fields of artificial intelligence, industry and Industry 4.0, as well as Big Data and Smart Data. Here, traditional physics knowledge comes together with modern algorithms. The goal: to solve problems more efficiently and accurately.
In machine learning, computers use data to recognise patterns and learn from them. The physics-informed approach additionally incorporates knowledge from physics. This means that existing natural laws – such as gravity or fluid dynamics – support the computer's learning processes. The result is that the systems' predictions are more realistic and adapted to the real world.
For example: In industry, one wants to predict how a certain material will behave under high stress. Pure machine learning requires vast amounts of measurement data for this. With physics-informed machine learning, much less data is often sufficient, because the system already knows the fundamental laws of material physics.
Businesses benefit from more accurate analyses, can save costs, and develop innovative solutions more quickly. Physics-informed machine learning thus combines the strengths of science and modern data analysis for tangible everyday benefits.
Deutsch 
English (UK)