Analysis Method Of Ocean Structure Anomaly Based On Monitoring Database

Today,with the rapid development of the global economy,various types of energy play avital role in social development.At the same time,the ocean is rich in resources,and the ocean platform is a large civil engineering facility used for the exploitation of offshore resources,so the structural safety of the ocean platform is the foundation of the development of the ocean economy.At the same time,if there is a serious accident on the ocean platform,the crude oil mined will cause pollution to the marine environment,so the structural safety of the offshore platform has received widespread attention from all walks of life.This article organically combines the two methods of modal analysis and machine learning,and applies it to the research of offshore platform structural anomaly detection.Specifically,this article first uses the offshore platform monitoring database as the data base,which includes long-term in-situ monitoring data for multiple FPSOs in the South China Sea during the 11 th,12th,and 13 th Five-Year Plans;then,the offshore platform monitoring data is considered,it has the characteristics of non-continuous,non-linear and non-stationary.It uses data cleaning,data first-order difference equation smoothing,data segmentation data pre-processing operations,and then extracts the offshore platform structure freedom from the data pre-processed response data.The response characteristics are used as algorithm training data;then the four commonly used anomaly detection algorithms are experimentally compared,and finally OCSVM is selected as the offshore platform anomaly detection algorithm;then because the offshore platform will have a variety of operating conditions during long-term service,the design is based on OCSVM,it is a core long-term online learning architecture based on single-class samples,and horizontally compares the accuracy and space complexity of the three methods of offline learning,online boundary incremental learning and online re-learning,and analyzes the advantages of long-term online boundary incremental learning.Finally,finally apply long-term online boundary incremental learning to Real data on ocean platforms,analyzes the false positive rate and accuracy of detection of abnormalities two areas of particular interest to structure demonstrates the applicability of the boundary-line incremental learning platform in ocean engineering practice.It can be concluded that the long-term online boundary incremental learning method is suitable for offshore platform structural anomaly detection,which can solve the engineeringdifficulty in obtaining abnormal data samples and long-term online while ensuring low false alarm rate and high accuracy.It can be considered that this analysis method has certain engineering significance for the problem of learning the complexity of the architecture space.