Real-Time Simulation, Simulation-Based Monitoring and Predictive Maintenance

The real-time simulation of mechatronic systems in the automotive development phase is becoming even more important just as it is also in demand in vehicle operations where the use of virtual observers and assistance systems is becoming more common.

In all real-time applications, the interplay of modeling, model complexity, and numerics is of great significance. In addition to standard applications like the real-time simulation of HIL/SIL tests, we are also working on the problems encountered in real-time capable simulation processes. Our special focus is on the analysis and evaluation of co-simulation and real-time capable integration methods. Depending on the application, we use standard prototyping environments as well as real-time models implemented directly on micro controller boards as simulation platforms.

In addition to the numeric and process engineering analysis of real-time processes, real-time simulation is also applied in diverse application areas, to include in the vicinity of the Human-in-the-Loop driving simulator installed at ITWM. The demands placed on a vehicle are strongly influenced by the driver's behavior; the reactions of the driver, in turn, depend on the actual system behavior. To realistically implement this in its entirety in a simulator requires that vehicle dynamics and the interactions of the vehicle with its environment be simulated in real-time. Real-time capable physical tire models and efficient surface and interplay simulation are therefore also major research topics at our institute.

In simulation-based monitoring, the relevant but difficult to measure variables are simulated in real-time during operation using simple, standard system values recorded in the vehicle. The virtual measurements generated for the system variables in this way can then be used as a virtual observer in the driver assistance systems, for test bench control, or as virtual measurements as feedback in the development processes.  

A simulation based monitoring of operating stress was developed and published as part of a doctoral thesis at Fraunhofer ITWM. To guarantee high vehicle reliability, operational stability testing is carried out in the development and design phase. The stresses must reflect the real operating requirements to prove the predicted life expectancy in the test is equivalent to reality. This can only be approximated because of the high variety and diversity of use, in particular, in commercial vehicles and mobile work equipment.

Compute Simulations of Stresses

Simulation based monitoring allows the stresses on critical components during operation to be simulated and estimates of the degradation experienced by the component to be calculated. At the core of this development is real-time simulation of component stresses by means of flexible MBS models, which is performed on an embedded PC in the vehicle electronics. Suitable data processing and online classification algorithms (linear discriminant analysis) for the evaluation of the simulated stresses are also components of simulation-based monitoring.

Knowledge of the stresses arising during operation can enable maintenance intervals that are tuned to the actual component wear (»predictive maintenance«). The costs incurred for an unnecessarily early replacement of the part or by an unforeseen failure can be reduced. We methodically pursue and apply both physics-based models and data-driven approaches from the fields of Artificial Intelligence (AI) and Machine Learning (ML).