Mechatronic systems are monitored while its operation parameters are subjected to several influences affecting the efficiency, functionality, and safety. It is of major interest to infer the actual State-of-Health of critical components from acquired data. The degree of wear and the quality of mechatronic systems or cost-sensitive machine components are significant for the system reliability. The overall reliability of technical systems must be ensured in order to reduce the risks and costs of a system failure. Automated monitoring of wear and the classification of the machine state are necessary for State-of-Health evaluation. These techniques were used in the currently developed approach to judge both fault probability and system reliability. Optimization techniques are used to improve developed approaches concerning the reliability of State-of-Health evaluation. Core of this contribution is the comparison of two algorithms which can be easily used, applied, and handled. These algorithms were specially developed facing industrial data or measurements from technical systems. As example in this contribution a hydraulically driven machine part sliding over another one is used. A connection between measured hydraulic data to the degree of wear of the lubricated surface is used to calculate information about the state of a sliding surface between the two machine parts. The pressure time behavior is taken and filtered for better evaluation using arithmetic mean value and sliding window technique [1]. For further generation of suitably defined characteristics, the data has to be post processed and analyzed. Using the filtered data, the wear state is classified using two different methods. The first method uses thresholds to distinguish the surface condition into three states of wear. For the second method, further filtering and calculation of the trend lead to a classification of the wear state. In extension to previous work of the authors [1], here the thresholds of method one and the window size of method two are optimized to minimize the difference of the classification results for the same experiment. The main idea of this contribution is to make the results similar by means of optimization in order to assure the plausibility of the employed approaches.
doi: 10.12783/SHM2015/80