نگهداری ثابت دینامیکی تنظیم WIP در شبکه های تولید با سیستم های کار مستقل

In this paper, a method is presented for information sharing in production networks with large numbers of autonomous work systems for the purpose of maintaining constant dynamic properties when the structure of physical order flows between the work systems is omni-directional and variable. It is shown that information sharing is necessary if undesirable behaviors such as oscillation or slow response are to be avoided. A method for designing the dynamic properties of such networks is presented along with a method for distributed computation and communication of information needed to locally compensate for the expected order flows from other work systems.

The ability to establish and maintain desirable dynamic behavior is essential in production networks. This can be a particularly significant challenge when the individual work systems in a network have high levels of local autonomy, and cooperation and information sharing are used to ensure effective operation, rather than centralized control. Production networks are known to exhibit unfavorable dynamic behavior; for example, inventory levels can oscillate in supply chains as organizations respond individually to variations in orders [1]. Decentralized planning and control methods are an increasingly important alternative to centralized control of production networks; however, achieving effective cooperation and choosing the appropriate level of autonomy are significant challenges in design of these autonomous logistic systems [2], [3] and [4]. Due to the complexity of interactions between decision-making entities in production networks, modeling their behavior also is a challenge [5] and [6]. Two-level models have been developed that combine control of Work In Progress (WIP) with control of backlog [7] and final inventory [8]. Application of control theory to the production inventory problem has been reviewed [9], and control-theoretic approaches have been used to model supply chain management including the use of differential equations to study the stability of adjustments in inventories and production rates [10]. Autonomous work systems require coupling structures that create the information-based interactions necessary to ensure that local actions are globally effective [11], and the control laws and heuristic rules chosen need to create well-behaved network dynamics including desired responsiveness, absence of oscillatory behavior, and robustness in the presence of uncertainties. There is a need to limit the propagation of disturbances in a production network and to ensure that the dynamic behavior of the network remains as designed and does not change unpredictably or unfavorably with time. It is shown in this paper, through dynamic system analysis, that when the structure of order flows between the work systems is omni-directional and variable, there can be variations in the fundamental dynamic behavior of the work systems and the production network. It is also shown that information coupling created by sharing of order-flow structure information can produce desired and consistent dynamic behavior when the order-flow information is accurate. A method for designing the dynamic properties of a network is presented along with a method for distributed computation and communication of information needed to locally compensate for the expected order flows from other work systems.

It has been shown through control-theoretic dynamic analysis that a WIP regulation topology for autonomous work systems that includes order-flow information sharing can lead to more favorable and more consistent fundamental dynamic behavior. The goal of this dynamic consistency is to allow the responsiveness of regulation to be chosen (designed) and relied upon in network operation. There is a high level of autonomy because only local information is shared between work systems. Information is gathered and shared when it is needed rather than being archived either within the work systems or in a centralized network database. It has been shown in this paper that variations in the coupling between work systems created by omni-directional physical order flows can be result in variation in the fundamental dynamic behavior of the work systems and the production network. It is also shown that information coupling created by sharing of order-flow structure information can produce desired and consistent dynamic behavior when the order-flow information is accurate. As illustrated in Fig. 4, order-flow information sharing curtails propagation of turbulence to downstream work systems. On the other hand, as illustrated in Fig. 5, and inaccurate compensation for order-flow structure can result in dynamic behavior that deviates from that desired, potentially becoming oscillatory or requiring longer time periods to react to disturbances than designed. A method was presented for choosing a value of control parameter kc that improves responsiveness without producing oscillation, and an algorithm has been presented for distributed computation and communication of information needed to locally compensate for the expected order flows from other work systems. Delay in capacity adjustment has been included to represent the inability to make instantaneous adjustments, but non-linear variations in cost, delay and feasibility with capacity adjustment magnitude have not been modeled. Other logistic issues have not been addressed including the effects of capacity limits, buffer capacities, setup times, transportation times, starvation of work systems when WIP is low, and modeling the work content of orders.