Integration of systems and concurrent engineering using artificial intelligence in improving passenger ship preliminari design

Abstract

Designing ships having large and complex systems involves prescribed design parameters development and is typically executed exhaustively through iterations. The processes become challenging as the design complexity increases. This is due to the simplistic and sequential approach of the conventional ship design spiral model. To mitigate this, a data-centric and integrated design approach with artificial intelligence (AI) was proposed as objective in this thesis. It was applied to simulate the passenger ship’s preliminary design developments based on the identified design goals, requirements, and data. The methodology was carried out in deriving a ship design methodological framework, ship design processes, knowledge graph, predictive models, and computer-based design tools. An extension to the integrated Quality Function Deployment and Axiomatic Design (QFD-AD) method was proposed to establish and analyse the design functional requirements, parameters, data, and tasks concurrencies. It was further explored using graph theory to represent the ship design, data, and their relationships. Finally, AI and deep learning (DL) methods were explored to develop and deploy prediction models at the graph nodes to determine the ship preliminary design principal parameters. These steps led to the development of a computer-based design tool to simulate and evaluate the ship design. The method was then applied to investigate and evaluate a generic passenger ship design model. The results from the design modelling, prediction model and empirical approximation were compared, evaluated, and discussed. While the stepwise empirical model algorithm was ten times faster, it was restricted by the set of hard rules that are based on assumptions. Though, the speed was highly influenced by the algorithm complexity and number of iterations till convergence. This phenomenon was observed in the brake power (P) prediction where the data-centric approach outperforms the Bailey’s rule-based model by four times with a nearly accurate result. This work showed significant impact in terms of simplifying the existing ship design model, evoking design information, and producing fast and nearly optimal solution. It lightened the effort in initiating the design in terms of data collection, requirement analysis and planning in the conceptual and preliminary designs phases. Importantly, it shows the potential for a broad range of applications, scales, and design automation.