Application of fuzzy inference system for analysis of oil field data to optimize combustion engine maintenance

摘要

The condition of a technical system has been subject to intense scrutiny in recent years. Monitoring the technical condition of a system may be performed by applying different approaches. The main intention of the monitoring is to get the information about the instant system condition, and to estimate and predict reliability measures. In the article, the authors suggest possible ways to process diagnostic measures which have the potential to determine the system condition and to predict its future development. The diagnostic measures are in this case indirect and they are introduced in the form of oil data. The diagnostic data are obtained from the tribodiagnostic system which is composed of kinematic pairs and oil. The analysed oil samples come from the combustion engine of a heavy ground vehicle. The authors focus on the output values in the form of wear particles, iron and lead, and additive particles. The concentration of these particles in the oil is influenced by operating time and calendar time. However, the particles include inherent and natural levels of uncertainty and fuzziness. Therefore, the authors apply and present the models imitating the development of the particles which are based on a fuzzy inference system. Highly valuable and extensive data set records enabled the authors to perform two-dimensional data modelling based both on operation time and calendar time. The obtained results enable us to predict the remaining useful life of the system. Moreover, the results could also be beneficial when modifying hard time scheduled preventive maintenance intervals (e.g. when to change the oil). The major contribution of this paper is the fact that all analysed diagnostic data are not artificial but real; moreover, they were collected for more than 10 years and therefore contain hundreds of records.