A real time methodology of cluster-system theory-based reliability estimation using k-means clustering

With rapidly increase in the design and application of complex system engineering, life and reliability analysis methods for these engineering have received much attention. A novel real time reliability analysis methodology is proposed based on the cluster-system theory and k-means clustering. Firstly, the system consisting of three or more identical or similar sub-systems facing the same load or task is defined as cluster-system. By analyzing the key performance parameters, sub-systems with similar performance are divided into a family-system, so that each sub-system can be used as reference sample for other sub-systems in the identical family-system. The cubic spline interpolation method under first-order boundary conditions is used to fit the average variation of key performance parameters of family-systems. The residual life of family-system can be obtained by determining the threshold of key performance parameters of sub-system through experiments or experience. Then reliability of the whole system is estimated by the contribution of each sub-system in cluster-system to solve the problem that there is no fault data and reference samples in the reliability analysis of complex system. Application in the Five-hundred-meter Aperture Spherical Telescope (FAST) demonstrates the effectiveness of the proposed method, compared to traditional reliability analysis methods based on experimental statistics.