Design knowledge extraction framework and its application in multi-objective ship optimization

Abstract

Currently, hull form optimization performs design space searches through an optimization algorithm to find the optimal solution and discusses the optimal result only while ignoring a large amount of optimized data. In the optimized data, there are implicit relationships between variables and between the variables and objective function, which can contribute to a better understanding of optimization problems. For the above problems, this paper studies the optimized data in hull form optimization and proposes a set of design knowledge extraction methods suitable for multi-objective ship optimization; in addition, we study the optimization of ship form lines under various working conditions using a 7500-ton inland twin-skeg bulk carrier as the research object. On this basis, the design knowledge is extracted via sensitivity analysis and self-organizing mapping neural network in data mining, and the effectiveness of the rule extraction technology is verified through a comparative analysis of the results. The concluded rules play a significant role in guiding hull form optimization.