Study of construction of optimal production and distributio systems based on JIT philosophy

Just-In-Time生产系统的特点如下：1、适用于多品种小量生产的顺序提货后补生产系统，简称提货系统。2 .自主分散生产系统；3 .持续改进彻底消除浪费的机制及改进活动；多功能工人和u型生产线。在pull系统中，在生产和提取过程中使用看板或e-看板进行沟通和控制，并用电子信息代替看板。然而，在这些日子里，信息技术是先进的，包括电子看板在内的看板系统减少了浪费。在本研究中，我们试图构建一个不确定环境下的最优生产和分配系统，该系统利用了上述特征，除了第一个特征作为JIT哲学。为构建基于JIT理念的最优生产与分配系统，明确了以下三点：通过将马尔可夫决策过程与强化学习和模拟技术相结合，开发了一种不使用看板的多项目、多阶段生产和分配系统中最优生产和退出策略的神经DP算法。2 .将排队理论与仿真有机结合，开发适应客户需求多样化和产品生命周期短的各类单元生产系统的性能评价算法。发展实用的元启发式方法来解决物流中的基本问题，例如三维包装问题和车辆常规问题，以获得基于JIT哲学的最优配送系统。

The features of the Just-In-Time production system are as follows :1. the sequenced withdrawal and later-replenishment production system, which is called shortly the pull system and is adapted to multi-item small-sized production,2. the autonomous decentralized production system,3. the mechanism of thorough elimination of waste by continuous improvement and the improvement activity,4. multi-function workers and a U-shaped production line.In the pull system, Kanbans or e-Kanbans replacing then by electronic information are used for communication and control in production and withdrawal. In these days, however, IT is advanced and the Kanban system that includes e-Kanbans is reduced to waste. In this study, we attempt to construct an optimal production and distribution system under uncertain environments that utilizes the above features except the first as the JIT philosophy. The following three points are clarified for construction of the optimal production and distribution systems based on the JIT philosophy :1. Development of a Neuro DP algorithm that determines an optimal production and withdrawal policy in a multi-item, multi-stage production and distribution system without using Kanbans by combining the Markov decision process with reinforce learning and simulation technique.2. Development of performance evaluation algorithms of carious types of cell production systems, which are adapted to diversification of customer demand and a short product life cycle, by organically combining queueing theory with simulation.3. Development of practical meta-heuristics for basic problems in logistics, e.g. three-dimensional packing problems and vehicle routine problems, to attain optimal distribution systems based on the JIT philosophy.