Scholarlink

<p>This paper addresses the integrated problem of production, distribution, and maintenance planning based on a data-driven approach, within the supply chain context, involving multiple production sites and multi-demand. Traditionally, these activities have been treated as separate topics in the literature. We aim to predict an optimal production plan for each site to meet its specific demand. Additionally, we aim to establish a collaborative distribution plan among the sites, bridging the gap between production and demand for needy sites, while minimizing the production-distribution costs and adhering to a service level for each production site. Given that demand forecasting presents a significant challenge, we analyze historical demand data for each site. This analysis aims to accurately forecast demand for each manufacturing site using various deep learning models. We also aim to predict preventive maintenance actions for each site, taking into account the dependency of the failure rate and the production cadence. We propose a data-driven approach that employs deep learning methods, namely the long short-term memory model for forecasting multi-demand and the NeuroEvolution of augmenting topologies model for predicting the three joint plans. The resulting integrated method is assessed using reference datasets. The proposed framework is evaluated using reference datasets, with results being compared across various approaches to highlight the advantages of the proposed framework.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Deep learning framework for multi-demand forecasting and joint prediction of production, distribution, and maintenance across multiple manufacturing sites

  • 期刊论文

  • 发表日期：

    2025年1月7日

查看全文
  • Mohamed Ali Kammoun,
  • Zied Hajej,
  • Sami Bennour,
  • Nader Salem,
  • Oumayma El Mabrouk,
  • Amir Baccar

摘要

This paper addresses the integrated problem of production, distribution, and maintenance planning based on a data-driven approach, within the supply chain context, involving multiple production sites and multi-demand. Traditionally, these activities have been treated as separate topics in the literature. We aim to predict an optimal production plan for each site to meet its specific demand. Additionally, we aim to establish a collaborative distribution plan among the sites, bridging the gap between production and demand for needy sites, while minimizing the production-distribution costs and adhering to a service level for each production site. Given that demand forecasting presents a significant challenge, we analyze historical demand data for each site. This analysis aims to accurately forecast demand for each manufacturing site using various deep learning models. We also aim to predict preventive maintenance actions for each site, taking into account the dependency of the failure rate and the production cadence. We propose a data-driven approach that employs deep learning methods, namely the long short-term memory model for forecasting multi-demand and the NeuroEvolution of augmenting topologies model for predicting the three joint plans. The resulting integrated method is assessed using reference datasets. The proposed framework is evaluated using reference datasets, with results being compared across various approaches to highlight the advantages of the proposed framework.