ارزیابی راحت و کاربر پسند از مدل کنترل نظری از مقررات WIP در شبکه های سیستم کار مستقل

One means of adapting to variation in demand is making capacity flexible so that work in progress (WIP) can be regulated; however, this can significantly influence the dynamic behavior of production networks in which there is high local autonomy. Control-theoretic models are a convenient means for investigating and designing the dynamics of such networks, but the fidelity of these models is not well understood. In this paper, results obtained using discrete-event simulations are used to assess the control-theoretic approach, providing evidence that fidelity varies depending upon factors such as WIP level and the magnitude of capacity adjustments.

Industries are being continually challenged by variability due to product variety, shortened lead times, rush orders, etc. Industries have attempted to address these issues using concepts such as agile manufacturing, quick response manufacturing and lean manufacturing. Centralized control and information sharing has been recommended to reduce effects of variability in demand in production networks and unfavorable dynamic behavior such as the ‘bullwhip effect’ [1]. However, required information sharing and complexities due to growing numbers of echelons and relationships in networks complicates centralized control, and autonomous decentralized control has been recommended to better respond to changing markets [2] and [3]. As production networks expand and increase in complexity, it is important to ensure that local decision making and disturbances do not adversely affect their dynamic behavior [4]. Techniques and tools of control theory can be used to understand dynamics of production networks. A review of research in this area suggested that control theory can be used to reduce inventory variations, demand amplifications and optimize order release rules in the networks [5]. Industries often struggle with maintaining optimal work in progress (WIP) to satisfy conflicting objectives of short lead times and high utilization when there is high variability in demand [6]. Non-linear operation rules have been developed to adjust WIP in complex networks [7]. Control-theoretic models have been used to analyze stability of production networks [8], and recommended for regulating lead times and improving customer service [9]. Control-theoretic approaches have been proposed for WIP regulation to improve operating performance, and dynamic models have been developed for WIP regulation in networks of autonomous work systems [10], [11] and [12]. Control-theoretic models of production networks can be developed more quickly than discrete-event simulation models, and produce estimates of fundamental dynamic properties such as time constants and damping ratios that characterize how rapidly production networks respond to turbulence and whether responses are oscillatory. However, the fidelity with which control-theoretic models predict the fundamental dynamic behavior of production networks is not well understood and has not been assessed. Many assumptions are required to make control-theoretic models tractable, and many details are ignored regarding the logistics of production. Such an assessment is described in this paper. A specific industrial scenario was used to assess the fidelity of a control-theoretic model of a network of autonomous work systems with local WIP regulation. A discrete-event simulation was developed for the same scenario and used as the benchmark for comparison. In the following sections, these models are described and areas of agreement and deviation are identified. The benefits of WIP regulation using a control-theoretic approach are also discussed.

Discrete-event simulation models and an industrial dataset were used in this work to assess the fidelity of control-theoretic models. Except at low levels of WIP or when response was extremely oscillatory, the response of WIP regulation to work disturbances such as rush orders predicted by the control-theoretic models was nearly identical to that predicted using discrete-event simulation. Results of discrete event simulation of WIP regulation at low WIP showed lower utilization, higher capacities and lower settling times than predicted by the control-theoretic models because the latter do not represent individual orders and machines. Also, there was more variation in WIP with discrete-event simulation than with control-theoretic simulation because the latter neglected daily changes in work flow structure. It was concluded that the fidelity of the control-theoretic model decreased at extreme conditions such as low WIP and large capacity adjustments at very high gain Kc, but predictions of fundamental dynamic behavior using transfer functions were otherwise good. Control-theoretic simulations of operation at extreme operating conditions were of significantly lesser fidelity than discrete-event simulations; thus, further research is required to improve their fidelity. Research already conducted on order-flow information sharing between autonomous work systems [12] may serve as a starting point in this regard. Comparison of discrete-event simulation results with and without WIP regulation indicated significant reductions in variation of WIP from planned levels. More extensive simulation studies using distributions for arrival rates, service times, etc. are needed to confirm these results in a broader range of scenarios.