Efficient oscillation detection for verification of mechatronic closed-loop systems using search-based testing

摘要

In mechatronic system design, the systematic early-stage verification in simulation environments is a crucial aspect. In particular, the development of highly sophisticated and well performing controllers requires complementary testing by extensive simulations. The assessment and evaluation of the control strategy and its tuning is ideally carried out by specialists and is usually very time consuming. Due to time constraints, often only the nominal configuration and a limited number of other configurations are considered in the simulation-based analysis of the closed-loop system. To cover a more comprehensive set of test cases and speed up the analysis, assistive and automated tools are essential. This paper proposes a method to systematically search for oscillations in closed-loop mechatronic systems, which, in most cases, indicate a poor control performance or even an unstable system behavior. Although oscillation detection methods are very well established in the process industry, these methods do not directly apply to mechatronic systems where frequent changes of the operating points and tracking of fast trajectories have to be taken into account. A recently proposed method which can cope with these challenges uses the Empirical Mode Decomposition (EMD) combined with an assessment of the decomposed modes. This is the basis for the construction of a normalized and continuous oscillation index, which provides a suitable measure for the oscillation strength. The performance of the selected EMD approach compared to other state-of-the-art concepts is illustrated by measurement data of a hydraulic valve. The introduced oscillation index serves as an objective function for a Search-based Testing (SbT) approach to systematically select those parameter combinations which yield a worst-case scenario for the closed-loop system. The high potential of the combination of SbT with an oscillation-based objective function for the automated verification of closed-loop mechatronic systems is demonstrated for the example of a goods lift.