A Method for Knowledge Mining of Satellite State Association

This paper presents a new scheme for data mining in spacecraft state association and abnormal detection. A method which includes state association knowledge mining (SAKM) and Similar Density Merge Clustering (SDMC) is developed. Data from the satellite are the most critical thing for analyzing satellite state and abnormal detection, therefore correct analyze of the regulation is quintessential for the detection of satellite abnormal detection. The associations in any of the subsystems are known when satellite was designed; however, the associations between any two of the subsystems are to be found. In such cases a SAKM algorithm becomes essential for obtaining the correct regulation. Clustering algorithm such as K-means is important in the SAKM algorithm, although it has its own limitations; however in this paper a novel method has been proposed for conditioning of K- means method. So that it can be effectively used for regulation discovery of data and satellite abnormal detection. The SDMC approach involves determination of classes' centers from the classes established by K-means method. In this way the points in one class are close enough. The high similarity points fall into the same class while the low similarity points fall into the different class based on the SDMC algorithm, the SAKM algorithm use Apriori algorithm mining frequent item sets and association rules of parameter feature characters. The experimental results show that as compared to the K-means algorithm the SDMC method can effectively cluster the data, the SAKM algorithm can correctly mining the satellite association knowledge. Spacecraft in orbit return a large number of data to the ground control center every day. Day and months multiplying, data and information are quite large. The analysis and feature extraction of satellite data is also difficult. The abnormal determine and fault diagnose, usually use human experience related to the possible data analysis to determine the change of satellite. In such cases research on the state association knowledge mining method and discovery the regulation among data will play the most significant part in correctly positioning the abnormal in spacecraft. Current approaches to satellite association knowledge mining are based on the human experiences and require long- time testing with a great amount of data. Fault detection capability of the current strategies is also fairly weak. Motivated by these considerations, a new method based on K- means algorithm is proposed in for clustering the high similarity points to one class. The low similarity points will typically lead too far away. In this paper an improved K-means method (SDMC) has been proposed the aim of which is to overcome the inherent limitations of K-means and then correctly discover the state change regulations of satellite. For this purpose, a large amount of data has been used for clustering. As in the entire in SDMC algorithm the data points are mapped into high dimension space using a kernel function, where it then adopts the Apriori algorithm method for mining the association knowledge and frequent pattern. the exact class centers are determined using the classes which is established before using history data in satellite.