Abstract
IGUSA (Intelligent Global Ultimate Strength Analysis) is a tool developed by PETRONAS to predict the ultimate strength of a fixed offshore jacket platforms installed in Malaysian waters using machine learning techniques.
The ultimate strength, or more commonly represented by Reserve Strength Ratio (RSR), is a gauge of the robustness and redundancy inhibited in a fixed offshore structure. It is very useful in being an indicator for fitness-for-purpose of the platform and which is an integral part of Structural Integrity Management (SIM). However, a typical deterministic ultimate strength analysis for a fixed offshore structure is a time intensive process, using specialized software in the realm of plastic collapse analysis. As such, it is intended that machine learning techniques to be utilized to perform a prediction for the RSR, subsequently optimizing resources in SIM processes.
This paper will discuss the development of data-driven predictive model of IGUSA. Various machine learning techniques were experimented on PETRONAS' Global Ultimate Strength Analysis (GUSA) data. The objective is to obtain an accurate and reliable model to predict the RSR. Nonlinear regression using Artificial Neural Network (ANN) was found to provide the best model to predict the Base Shear Collapse, and hence the RSR for a typical jacket platform. The ANN model was incorporated into the IGUSA tool for deployment within PETRONAS. It is envisaged that IGUSA will be a valuable rapid screening tool for the typical platforms and the deterministic ultimate strength efforts can be focused on the more critical platforms.
Based on IGUSA development, the usage of machine learning techniques is proven to be useful in the structural engineering discipline. It is hoped that IGUSA will be able to assist PETRONAS and other Oil and Gas Operators in the region to optimize their resources in SIM processes.