Research On Operation Mode Identification Of Offshore Platform Based On Fuzzy Clustering

Operating mode recognition was one of the key technologies of safety monitoring for offshore platform healthy structures.This technology can effectively eliminate the interference of operation state change on structural damage recognition and improve the accuracy of damage recognition.On the other hand,it was the technical basis for improving the operation and maintenance efficiency,ensuring safe production and realizing the digital twin of the offshore platform that understanding the operating mode of the platform.One kind of operating mode recognition method based on fuzzy clustering principle which combined with convolution neural network method was established to analyze and process the vibration response signals.Through the analysis of spatio-temporal difference of platform monitoring data,the characteristic parameters which can describe the vibration characteristics effectively were selected.The primary clustering analysis on these characteristic parameters was executed by fuzzy clustering algorithm based on density peak.For solving the unsatisfactory primary clustering results,a hierarchical clustering method was proposed,which divided the monitoring data into normal time series and special time series.Different characteristic parameters were extracted from these two kinds of data for secondary clustering analysis.According to the secondary clustering results,the vibration levels of each measuring point were divided.For normal time series and special time series,two primary convolutional neural network models were established to classify the vibration levels.Therefore,a panoramic space-time map reflecting the overall vibration of the platform was obtained.On this basis,the third-order tensor expression of the global information of the platform was established.The three-dimensional convolutional neural network model of operating mode was built for the panoramic space-time graph.The secondary convolutional neural network model was used to identify the operating mode for the target platform from February 2019 to January 2020.After verifying the representative operation conditions of the platform,the actual operation state of the target platform can be identified effectively.