.png){kind=icon}
.png){kind=icon}
Return
Diffusion model-based generative optimisation for resilient service composition in cloud manufacturing
.png){kind=icon}
PRE
AI
DOI：10.1080/00207543.2026.2625247.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Recent disruptions in supply networks highlight the need for resilient, data-driven decision support. We propose a novel multi-stage optimisation framework that integrates deep generative models with operations planning to address dynamic resource allocation under disruptions. Specifically, a variational autoencoder (VAE) learns a latent manifold of feasible allocation plans, and a diffusion-based generative model produces realistic refined disruption-aware allocations. These components combine in a stage-wise optimisation process. The approach yields flexible allocation strategies that adapt in real time to evolving uncertainties, improving overall network resilience. We demonstrate practical relevance of our method through a case study with disruptions, where our generative-optimisation framework outperforms traditional optimisation by providing faster adaptation and better performance under disturbance. This work highlights the novelty of embedding generative AI into production decision support. By leveraging learned latent spaces and diffusion processes, it offers a new direction for operations research tools that blend generative modelling with optimisation, enhancing resilient and informed decision-making. The computational approach scales to realistic planning problems and provides decision-makers with actionable insights. Overall, our results suggest that generative optimisation can serve as a practical decision-support tool for managing network disruptions.
Keywords:
Generative AI
diffusion models
resource allocation
resilience
cloud manufacturing
Journal
IF:
7.3
Papers:
1.1W
Citations:
3.7W