摘要
•A human-centric decision-making approach assesses in situ production resources and preferences based on job delivery risks.•The robust scheduling approach accounts for fluctuating process time in System-in-Package Assembly (SiP).•A scheduling framework for SiP with surface-mount technology (SMT) phase and dual-inline package (DIP) phase.•We conduct an empirical study focusing on the stochastic scheduling problem in SiP.
System-in-package (SiP) assembly integrates substrate-level interconnect fabrication, which includes patterning redistribution layers on the substrate surface and forming through-substrate vias via photolithography, sputter deposition, and electroplating processes to enable high-density lateral and vertical electrical interconnections. Production involves both surface-mount technology (SMT) and dual in-line package (DIP) processes. Given that the SMT processing time is short and predictable, analyses typically focus on the makespan. However, as the DIP processing time is random, this creates a multi-objective, stochastic flow shop scheduling problem for SiP assembly production that requires balancing the makespan and total tardiness. We address this challenge by proposing a simulation-based scheduling method using a genetic algorithm to integrate the SMT and DIP stages while evaluating delivery risk. A mixed-integer linear programming model with a chance constraint is developed to minimize the makespan and total expected tardiness under uncertain processing times. A discrete event simulation system evaluates various dispatching rules in the DIP stage, identifies Pareto optimal solutions, and defines a cumulative processing time distribution to assess delay risk. This framework enables the reallocation of production resources based on real-time conditions. Empirical validation at a factory demonstrates that our approach improves both the makespan and job tardiness simultaneously, showcasing its practicality. This study presents a human-centric decision-making procedure for scheduling plants that is guided by a dynamic balance between the risk level, time buffer, and resource adjustment to maintain on-time delivery and production stability in an uncertain environment. The proposed approach leverages practical the technique for order of preference by similarity to ideal solution (TOPSIS) applications combined with managerial preferences to simplify implementation through weighted objectives. In addition, it provides ranked scheduling recommendations, enabling decision-makers to assess in-situ production resources and risk preferences based on job delivery risks.