具有时滞特征的动态供应链建模与稳定性分析

Dynamic Supply Chains with Time Delay Characteristic Modeling and Stability Analysis

动态供应链的平稳运行是供应链管理的重要基础,针对具有时滞特征的动态供应链建模波动性的难题,在归纳总结正交函数、正交函数集性质的基础上,将正交神经网络应用到动态供应链模型的稳定性分析中,建立了具有时滞特征的供应链单层和多层模型,研究了时滞特征供应链系统的稳定性,设计了基于正交神经网络的时滞特征的动态供应链网络模型结构的权值调整的求解算法,并以四层时滞供应链系统进行仿真研究,仿真结果显示:对于多层时滞供应链模型,使用该方法,提高了收敛速度和增强了最优解的逼近能力,同时增加了时滞供应链的稳定性。

Because of the recent push of information technology, supply chain management has been making a positive impact on all parties, including suppliers, manufacturers, wholesalers, retailers, and customers, involved in the supply chain operation. As customers demand more personalized and diversified products and services, the product life cycle is shortened, and the market uncertainty of competition environments are increased. Rapid market changes require that companies adjust their management structure. It is in this context that the dynamic supply chain is being widely implemented in order to obtain faster and more flexible market strain capacity. The dynamic supply chain has drawn extensive attention in different fields, including management science, mathematics, physics, petrochemical industry, living services, and military logistics. Real-time operation and research of dynamic supply chains have become current focuses. Having a dynamic supply chain run smoothly is an important foundation of supply chain management. In order to solve difficulties in the modeling volatility of dynamic supply chains with time delay characteristics, Orthogonal Neural Network is used to discuss single layer and multilayer supply chain structures and their stability. Orthogonal neural network is a global convergence of the neural network. In the orthogonal neural network, each component unit（similar to neurons with the component of the neural network） of dynamic supply chain systems with time delay is able to avoid choosing initial value and numbers of the hidden layers, and confirming each layer neurons in order to speed up the training speed and realize system identification easily. This method that does not affect optimal solution can improve the speed of convergence and enhance the optimal solution of approximation ability. Compared with the existing research literature, this article focuses on the application of recursive orthogonal neural network algorithm for dynamic supply chain modeling with time delay characteristics, and further discusses the stability of dynamic supply chains with its characteristics in order to improve convergence speed and approximation capability.In the first part, on the basis of qualitative analysis of the orthogonal functions and their properties, dynamic supply chain system models with time-delay characteristics are established. These include two models： a monolayer model and a multilayer model. This research shows that the orthogonal neural network is an artificial neural network, similar to the neurons of neural network. This paper also shows that when the connection weight of dynamic supply chain system is made up of orthogonal function extension and the neural networks have linear excitation function, the supply chain system of units will have orthogonal relations. Forming the relations can help avoid the problems that the initial value is set, network layer is selected, and each layer of neurons number is determined. At the same time, the convergence speed is accelerated so as to implement system identification. In the second part, the dynamic supply chain system is discussed. The weight adjustment algorithm is designed based on dynamic supply chain network model structure with time delay characteristics. The orthogonal neural network is used to calculate the weight adjustment formula of monolayer and multilayer supply chain network structure. In the third part, the research question has carried on the simulation experiment with the existing literature results contrastive analysis. Finally, it is verified that the orthogonal neural network method can improve the convergence speed and enhance the stability of the supply chain system with time delay characteristics. In summary, based on the stability of the supply chain model and the limitations of traditional methods, this study used the orthogonal neural network algorithm, and dynamic supply chain with time delay to explore the stability of the studied modeling methods. This is a crossover study. This paper＇s main results and contributions are as follows： firstly, this paper summarized the properties of orthogonal function and set of orthogonal function, and applied to the stability analysis of dynamic supply chain model. Secondly, this paper established a monolayer and multilayer model of supply chain with time delay. Finally, this study used orthogonal neural network algorithm in order to study the stability analysis of dynamic supply chain model with time delay, and simulated supply chain system with 4 layer time delay. The simulation results show that the method could improve convergence speed, enhance optimal solution approximation ability, and increase stability of the supply chain with time delay for multilayer supply chain models.