Multitask scheduling on cloud additive manufacturing using NSGA-II

Abstract

Purpose: Cloud manufacturing (CM) represents a new manufacturing paradigm that integrates distributed resources to provide on-demand services. The high consumer demand from various locations, coupled with the customizability and complexity of manufacturing, complicates task scheduling. In this context, 3D printers are crucial as innovative manufacturing technologies with significant potential in producing complex and custom products. Scheduling in CM falls under the non-deterministic polynomial time-hard category, where tasks must be scheduled and distributed rapidly. Considerations of distance, minimization of delays, and makespan become critical variables that must be considered. This research aims to schedule and distribute tasks in CM using the non-dominated sorting genetic algorithm II (NSGA-II) to minimize delays, reduce makespan, and decrease costs.

Methodology: NSGA-II is employed to tackle the complexities of scheduling in CM. The strength of NSGA-II lies in its ability to determine optimal and efficient solutions for multiobjective problems. Tasks originating from requests at various locations are adjusted based on material parameters and dimensions and then distributed to providers while considering aspects such as makespan, delay minimization, and cost.

Findings: The optimization results using NSGA-II demonstrate effective and efficient task distribution to providers. Across the four tested task distribution scenarios, the average computational time required was 5.59 seconds. Pareto analysis indicates a trade-off between various objective functions. Solutions with short distances tend to have increased maximum time and delays.

Originality/value: NSGA-II is effective for task distribution with multiobjective considerations. Not all three objective functions can be optimized simultaneously, given the trade-offs between distance, maximum time, and lateness. The priority of the objective functions should be determined to achieve optimal results. If minimizing lateness is most important, the focus should be on points with low lateness values. Further development can be done by modifying the Pareto front to make data-driven decisions that consider these trade-offs.