To maintain the competitiveness of shipyards in the current, difficult situation, further improvements to technology are necessary. Recently, various production technologies have been developed to advance the shipyard production environment under the influence of the Industry 4.0 toward automation, smart factories, and intelligent planning systems. To contribute to such efforts, we introduce a research case aimed at a simulation-based shipbuilding planning system. Shipbuilding planning processes are reorganized using an integrated planning and scheduling system, and a process-centric discrete event system simulation is used to enhance planning quality. Moreover, the proposed simulation-based planning system is applied to an actual shipbuilding process, proving that it could enhance the quality of production planning through several productivity evaluation indices.

Multi-agent scheduling algorithm is a useful method for the flexible job shop scheduling problem (FJSP). Also, the variability of the target system has to be considered in the scheduling problem that includes the machine failure, the setup change, etc. This study proposes the scheduling method that combines the independent learners with the implicit quantile network by modeling of the FJSP with high variability to the form of the multi-agent. The proposed method demonstrates superior performance compared to the several known heuristic dispatching rules. In addition, the trained model exhibits superior performance compared to the reinforcement learning algorithms such as proximal policy optimization and deep Q-network.

Currently, product technologies, information technologies, and market requirements vary rapidly to realise high productivity and reduce cost, creating many problems in the manufacturing industries. In particular, shipbuilding industries face serious scheduling problems owing to their complexity. The existing simulation-based methodology for shipbuilding is limited in that no systematic analysis method exists and complex shipbuilding processes cannot be simulated. This paper proposes a modelling and simulation methodology that is expected to overcome this limitation. The proposed methodology can create shipbuilding simulation models through a systematic analysis and modelling procedure with a specific technical module, whereas the existing methodology only concentrates on the definition of the modelling and simulation procedure. The methodology is confirmed as an efficient tool for the evaluation of shipbuilding scheduling information through the development of simulation modelling for the shipbuilding fabrication process. Furthermore, a web-based simulation system that could be used in an actual planning department is developed with a server and client architecture.

In recent years, shipyards have conducted various studies to accurately investigate their production capacities and improve the accuracy of their production plans. Nevertheless, frequent delays in schedules inevitably occur at production sites. Simulation-based techniques are used in various fields to solve similar problems, especially those related to predicting loads on facilities and upgrading production planning by considering various constraints. Process-centric simulation modeling techniques are used in shipyards because the shipbuilding process has different characteristics from the general manufacturing process. Conventional resource-oriented modeling techniques are inadequate to construct a simulation model to analyze the logistical behavior of shipyards. In this study, a process-oriented simulation modeling method is applied and improved through the identification and development of additional modules needed to simulate the behavior of shipyard logistics. The core of the simulation is improved using a logistic token that can analyze the physical movement effected by each process. Moreover, such process modeling modules as the geographic information system, the route search module, and the spatial arrangement module are developed and integrated into one solution. 1. Introduction: The shipbuilding industry in South Korea was affected by the recent global financial crisis. It has since suffered from a decline in global freight volume that has led to massive losses in all areas of commercial shipping and offshore plants since 2010. The shipbuilding industry in South Korea, especially large shipbuilders, received a large number of orders by competing in the market while reorganizing its business structure around the growth of offshore plants. This move was prompted by the flooding by Chinese shipbuilders of the local market with general merchant vessels (Fig. 1), as well as by prospects for the growing offshore plants' market due to crude oil depletion and high oil prices. However, the industry has suffered astronomical deficits in recent years due to unexpected losses, and the three major shipbuilders in South Korea are experiencing high operating loss and debt-to-equity ratio (Fig. 2), leading to fears of insolvency. Additional losses are expected due to large backlogs in orders for offshore plants.

During the shipbuilding process, a block assembly line suffers a bottleneck when the largest amount of material is processed. Therefore, scheduling optimization is important for the productivity. Currently, sequence of inbound products is controlled by determining the input sequence using a heuristic or metaheuristic approach. However, the metaheuristic algorithm has limitations in that the computation time increases exponentially as the number of input objects increases, and separate optimization calculations are required for every problem. Also, the heuristic such as dispatching algorithm has the limitation of the exploring the problem domain. Therefore, this study tries a reinforcement learning algorithm based on a pointer network to overcome these limitations. Reinforcement learning with pointer network is found to be suitable for permutation flowshop problem, including input-order optimization. A trained neural network is applied without re-learning, even if the number of inputs is changed. The trained model shows the meaningful results compared with the heuristic and metaheuristic algorithms in makespan and computation time. The trained model outperforms the heuristic and metaheuristic algorithms within a limited range of permutation flowshop problem.