Verbeterde Kalman-filtering voor systemen met cyclische input en parameter variaties met nadruk op mechanische transmissiesystemen

In modern mechatronic systems, control strategies and condition monitoring are essential for optimizing performance and ensuring safety during operation. The effectiveness of these techniques often relies heavily on the availability of measured data. However, direct measurement of essential variables such as forces and torques is often impractical, costly or even impossible. Moreover, when monitoring the health of a mechanical system, the critical parameters influencing its condition may not be directly measurable. The absence of direct measurements can be mitigated by constructing a system model that relates the quantities of interest with the available measurements. However, the accuracy of such models and their descriptive parameters are often inadequate for deterministic methods. Instead, stochastic state estimators, which account for uncertainties in both the model and measurements, are more appropriate. Moreover, state estimator schemes have been extended to jointly estimate uncertain parameters or unknown disturbances along with the system states. For the unknown variable, a zeroth order hold or random walk model is typically utilised. A low uncertainty is assumed when the unknown is a constant or slowly varying parameter, while a high uncertainty is required for tracking an unknown dynamic input force. The advantage is the generic estimation of unknown inputs and parameters when the trajectory is fully unknown. However, for many mechanical systems, the trajectory is not fully unknown but there are some recurrent or cyclic patterns in the input or parameter variations. Cyclic input or parameter variations are common in rotating machinery, drivetrains, and linkages, collectively referred to as mechanical transmission systems. The goal of this research is to improve existing estimators for these systems by leveraging the cyclic variations. Given that real-time availability of the estimates is often required in industrial applications, the research focuses solely on single-step estimator schemes to minimize computation time and maximize industrial applicability. The simple but efficient and effective Kalman filter serves as the core for the developed estimators. In order to allow tailored estimation solutions with a direct industrial applicability, the methods are developed case per case instead of targeting one general strategy. Several industrially relevant estimation problems for mechanical transmission systems are selected.