A just-in-time learning-based compensation prediction method for hull plate welding

Abstract

ABSTRACT The accuracy of plate welding substantially impacts the quality of the final hull and the hull construction efficiency. To control the accuracy of plate welding, one effective method is to estimate the extent of plate welding compensation for planning a reasonable accuracy management plan. This paper proposes a method that utilizes a Gaussian mixture model (GMM) and just-in-time learning (JITL) for this purpose. To develop the method, the raw data collected by the shipyard were used to perform accuracy judgment and missing value processing. Then, correlation analysis of various parameters and feature extraction methods were implemented to extract features that can be used to predict the compensation of plate welding. The plate welding process can be divided into different stages by using GMM to cluster the different data distributions. To track the welding process, the local models should be established based on the JITL principle. By comparing the performance of different models, the support vector regression model optimized by the particle swarm optimization algorithm was applied as the local model of the JITL model. The results demonstrated that the proposed model yields excellent compensation prediction performance.