Providing timely actuation guarantees with heterogeneous SAN for industrial process control.doc

为工业过程控制提供及时异构SAN驱动保证Abstract When wireless sensor networks are introduced in industrial settings, they will be part of a much larger heterogeneous sensor and actuation network that includes those sub-networks together with Ethernet cabled Programmable Logic Controllers (PLCs) and control stations. Performance issues arise in such systems. Particularly, actuation latencies are considered crucial in those scenarios. In this paper, we present an approach for planning and evaluating a heterogeneous supervisory control and data acquisition(SCADA) system with wireless sensor networks. We propose closed-loop schemas and mechanisms for measuring and verifying the latency bounds in the whole system. As a case study, we deploy and evaluate our approach on an oil refinery, where several heterogeneous networks are used to monitor and control a water treatment process. Results demonstrated the effectiveness of our method in planning, debugging and providing feedback about the network (in)ability to provide timely actuation guarantees.摘要：当无限传感器网络被引入工业环境中，他们将成为更大的异构传感器和驱动网的一部分，包括那些子网络与以太网电缆可编程逻辑控制器（PLC）网络和控制站。在这样的系统中性能问题大大出现了。特别是在这些情况下启动延迟是至关重要的。在本文中，我们提出了一种方法，用于评估和规划一个异构的数据采集与监控系统（SCADA）与无线传感器网络系统。我们提出了闭环模式和机制，用于测量和验证整个系统的延迟界限。作为一个案例研究，我们部署和评估我们的方法在一个炼油厂，有几个异构网络是用于监测和控制水处理工艺。结果表明，我们的方法在有效规划，调试和提供网络反馈（中）提供及时的激励保障能力。关键词：SAN，异质性，时间界限，闭环控制1、 引言With the evolution and increased adoption of wireless sensor technology and networks, and their easier and much cheaper deployment, there is a current trend to partially replace or complement the existing infrastructure. When deployed in industrial settings, it will be integrated on a larger heterogeneous sensor and actuation network composed by cable-based networks, wireless sensor networks (WSN), Programmable Logic Controllers (PLCs) and control stations.随着时代的演变，无线传感器技术和网络的使用也在增加以及他们更容易和便宜得多的部署，目前的趋势是有部分取代或补充现有的基础设施目前的趋势。当部署在工业环境中，它将被集成在一个更大的异构传感器和驱动网络由有线网络，无线传感器网络（WSN），可编程逻辑控制器（PLC）和控制站。In that context, sensor and actuator networks require a high level of reliability concerning network latency, delay and message losses, to ensure that monitoring and control loop actions can be made within required time bounds.在这样的背景下，传感器和执行器网络需要高水平的可靠性有关的网络延迟，延迟和消息损失，确保监测和控制回路的动作可以在所需的时间界限。Time-base guarantees can be provided by deploying WSN networks with real-time specific algorithms that include completely pre-planned synchronous time-division mechanisms. But, sensor and actuation infrastructures typically are heterogeneous systems, composed with specific software parts, some possibly offering real-time, while others do not. For instance, in a typical setting, a Time Division Multiple Access (TDMA) protocol can be applied to wireless sensor networks, to deliver some degree of in-network time predictability. However, the monitoring and control systems frequently are heterogeneous and composed by off-the-shelf networking and computerized systems without real-time operating services.时基担保可以由部署的WSN网络的实时的具体算法，包括完全预先计划好的同步时分机制来提供。但是，传感器和驱动的基础设施通常是异构系统的组成部分，与特定的软件，有可能提供实时的，而其他的没有。例如，在一个典型的设置中，时分多址（TDMA）协议，可用于无线传感器网络，提供一定程度的网络时间的可预测性。然而，监测和控制系统通常是异构的，由现成的网络和没有实时操作服务计算机系统组成。Performance control requirements (e.g. latencies or losses) introduce some constraints into the design and deployment of overall SAN architecture. Instead of devising a single wireless sensor network with thousands of nodes in a large monitoring and control deployment, there are a multitude of small networks with tens of nodes with planned functionalities connected with several gateways and cabled networks.性能控制要求（如延迟或损失）提出一些约束条件纳入总体SAN架构的设计和部署。与制定一个单一与成千上万个节点的大型监控部署的无线传感器网络不同，有数十节点与多个网关和有线网络连接功能规划众多的小型网络。A SCADA structure includes resource-constrained sensor nodes, such as TelosB nodes, and other more powerful nodes, such as PLCs, PC or servers, interconnect with cable or wireless links, as shown in figure 1.SCADA结构包括资源受限的传感器节点，如TelosB节点，和其他更强大的节点，如PLC，PC或服务器，以有线或无线链路互连，如图1所示。图1. 一般的SCADA系统Due to the nature of control, the data transmitted from sensors is only valid for a short time. If the data is delivered too late it is of limited use, as in most real-time systems.由于控制的本质，由传感器传输的数据只有在短时间内才有效。在大多数实时系统中，如果数据交付太迟只能是有限的使用。In industrial processes, some functions are safety-critical and thus are dependent on the system ability to operate within specific time boundaries. The prevention of accidents is extremely important. As an example, if a set point sent by the control system to a control valve cannot be delivered, the valve should fall back into a safe state (normally fully closed or fully open) after a timeout. The timeout depends on how long the process can tolerate a “malfunction” of the actuator before entering into a possibly dangerous situation, typically ranging from milliseconds to seconds. Additionally, the control system should also detect and alert control engineers when time boundaries are exceeded or communication losses occur. This is crucial to timely detect, mitigate and avoid error propagation to the rest of the control network.在工业生产过程中，有些功能是安全的关键，因此是依赖于系统的能力，在特定的时间范围内运行。事故的预防是非常重要的。例如，如果一个被控制系统发送到一个控制阀集合点不能交付，阀门在超时后应该回到安全状态（正常全关或全开）。超时多久取决于过程可以容忍一个“失灵”的驱动器进入一个可能有危险的情况之前，通常从毫秒到秒。此外，该控制系统还应检测和报警控制工程师的时间界限是超过或通信损失发生。这是及时发现减轻和避免错误传播到剩下的控制网络的是至关重要。In this work we propose an approach to plan closed-loop tasks with restricted time boundaries in such a heterogeneous system. We assume TDMA wireless sensor networks and consider measures and constraints from all parts of the system, including non-real-time operating components.在这项工作中，我们提出了一个计划闭环任务在这样的异构系统的限制时间边界的方法。我们假设TDMA的无线传感器网络和系统的所有部分考虑采取措施，限制，包括非实时操作系统组件。The rest of the paper is organized as follows: section II reviews related work; section III describes how to dimension a SCADA system with time guarantees, and discusses the time related guarantees that each component of the architecture can provide to the system. Section IV defines two closed-loop alternatives that can provide time requirements. In section V we present the experimental setup and the oil refinery requirements. Section VI shows the experimental results. Lastly, section VII concludes the paper.本文的其余部分安排如下：第二节回顾相关工作；第三节介绍了如何维与时间保证SCADA系统，并讨论了该体系结构的各组件可以提供给系统的时间相关的保证。第四节定义了两个闭环方案，可以提供时间要求。在第五节，我们目前的实验装置和炼油厂的要求。第六部分：试验结果。最后，第七章总结全文。II. RELATED WORKII. 相关的工作In this section we first review related work on schedule based planning and monitoring tools. One important key issue in WSNs that influences whether the deployed system will be able to meet time requirements is the MAC protocol and its configurations. The next related works concern implementing real-time strategies in single WSNs.在这一部分中我们首先回顾相关的以规划和监测工具为基础的工作表上的工作。在无线传感器中的一个重要的问题是影响是否已部署的系统将能够满足时间要求是MAC协议及其配置。下一个相关工作与在单传感器实施实时策略相关。In RT-Link protocol, time-slot assignment is accomplished in a centralized way at the gateway node, based on the global topology in the form of neighbor lists provided by the WSN nodes.在RT链路协议，时隙分配完成在网关节点的一个集中的方式，基于邻居列表的无线传感器网络节点提供的全局拓扑形式。WirelessHART is designed to support industrial process and automation applications. In addition, WirelessHART uses at its core a synchronous MAC protocol called TSMP , which combines TDMA and Frequency Division Multiple Access (FDMA).WirelessHART是专为支持工业过程自动化中的应用。此外，WirelessHART使用为核心的同步MAC协议称为结合TDMA和频分多址接入（FDMA）的TSMP 。GinMAC is a TDMA protocol that incorporates topology control mechanisms to ensure timely data delivery and reliability control mechanisms to deal with inherently fluctuating wireless links. The authors show that, under high traffic load, the protocol delivers 100% of data in time using a maximum node duty cycle as little as 2.48%.proposed protocol is also an energy efficient solution for time-critical data delivery with neglected losses.GinMAC是结合拓扑控制机制TDMA协议，以确保提供及时的数据和可靠的控制机制，解决固有的波动无线连接。作者表明，在高流量负载，该协议提供100%使用节点的最大占空比2.48%少时间数据。该协议也是一种被忽视的损失的时间关键数据提供高效节能的解决方案。PEDAMACS is another TDMA scheme including topology control and routing mechanisms. The sink centrally calculates a transmission schedule for each node, taking interference patterns into account and, thus, an upper bound for the message transfer delay can be determined. PEDAMACS is restricted by the requirement of a high-power sink to reach all nodes in the field in a single hop. PEDAMACS is analyzed using simulations, but a real-world implementation and corresponding measurements are not reported.PEDAMACS是另一个TDMA方案包括拓扑控制和路由机制。汇集中计算每个节点的发送时间表，以干扰模式的考虑，因此，一个上限的消息传递延迟可以确定。PEDAMACS由一个高功率的下沉到所有节点在该领域的一个单跳的要求限制。进行了模拟PEDAMACS，但实际的实现和相应的测量是不报道。Our approach is also related to monitoring tools that can be used to evaluate network performance. It requires information of latencies and other simple metrics that provide enough information about the network health to avoid safety control activation or accidents.我们的方法也与可以用来评估网络性能监测工具的相关。它要求延迟和其他简单的指标，提供足够的信息，以避免网络健康安全控制的激活或事故信息。There already exist some tools to monitor wireless sensor networks. These include Sympathy , Sensor Network Management System (SNMS) , Sensor Network Inspection Framework (SNIF) and Distributed Node Monitoring in Wireless Sensor Networks (DiMo) .已经有一些工具来监测无线传感器网络。这些包括支持，传感器网络管理系统（SNMS），传感器网络检测框架（SNIF）和无线传感器网络中的分布式节点监控（DIMO）。We have reviewed existing tools to monitor network health, and designed our own simplified tool adapted to our planning objectives. The existing tools are focused on specific goals, are complex to implement and work only inside a WSN.我们回顾了现有的工具来监视网络的健康，和我们自己的设计简化工具适合我们的规划目标。现有的工具都集中在特定的目标，实现起来很复杂，只有在一个无线传感器网络的工作。III. CLOSED-LOOP DIMENSIONING FOR HETEROGENEOUS NETWORKIII. 为异构网络闭环的尺寸Most performance-critical applications can be found in the domain of industrial monitoring and control. In these scenarios control loops are important and can involve any node and any part of the SCADA system. For instance, the closed loop actuation value can be determined inside any WSN subnetwork, any PLC or control station.大多数的性能关键的应用程序可以在工业监控领域的发现。在这些情况下，控制回路是重要的和可涉及的任何节点和SCADA系统的任何部分。例如，闭环驱动值可确定在无线传感器网络的子网，任何PLC或控制站。The SCADA system shown in Figure 1 assumes an industrial network with multiple WSN sub-networks. Given computational, energy and performance considerations, closed loop paths may be entirely within a single sub-network, with decision logic resident in the sink node (e.g. taking few ms) or outside the WSN (e.g. for applying more computational complex supervision controller), or it may span more than one WSN, with supervision control logic residing in one of the distributed PLC outside the WSNs (middleware servers).如图1所示的SCADA系统与多个无线传感器网络的子网络是一个工业网络。给定的计算的，能源和性能考虑，闭环路径可能是完全一个单独的子网内，在汇聚节点的决策逻辑的点（例如以几毫秒）或外部的无线传感器网络（例如使用更多的计算复杂的监督控制器），或者在一个在无线传感器网络的分布式PLC（中间件服务器）在监督控制逻辑点下它可能跨越多个无线传感器网络。One important factor in control is the closed-loop latency. The closed loop latency is the time taken from sensing node to the actuator node, passing through closed-loop manager. It will be the time taken since the value (event) happens at sensing node to the instant when the action is performed at actuator node. Since the value must cross several parts of the system, this latency can be decomposed into latencies for each part of the path. The latency can be divided into three main parts: upstream part (from sensing node to the closed-loop manager), processing latency (it is dependable of the closed-loop algorithm) and downstream part (from closed-loop manager to the actuator).在控制的一个重要因素是闭环延迟。闭环延迟是通过闭环管理从传感节点到执行器节点的时间。这将是自值时间（事件）发生在传感节点的瞬间作用于执行器节点进行。由于值必须跨系统的几个部分，这样的延迟可以被分解成的潜伏期为每个路径的一部分。潜伏期可划分为三个主要部分：上游（从传感节点到闭环管理），处理时间（这是可靠的闭环算法）和下游部分（从闭环管理器）。The position of closed-loop manager may depend on timing restrictions and data needed to compute decisions. For instance, if minimal latency was required and a single sub-network is considered, the closed-loop manager must be deployed at sink node, but the closed-loop decision computation inside embedded devices (sink node) is very limited. In this case, the upstream latency corresponds to the time for transmission between leaf node and sink node; typically, the processing time is small, because simple closed-loop techniques are used. The downstream latency is the time to transmit a command from sink node to the actuator.闭环管理点可能取决于计算所需的时间限制和数据的决定。例如，如果最小的延迟是必需的，被认为是一个单一的子网络，闭环管理必须部署在汇聚节点，但闭环决策计算在嵌入式设备（sink节点）是非常有限的。在这种情况下，上游延迟对应于叶节点和汇聚节点之间的传输时间；通常情况下，处理时间是很小的，因为使用的是简单的闭环控制技术。下游延迟时间发送命令从汇聚节点的执行器。Figure 2 shows a scenario example of closed-loop system where the closed-loop decision is done by the sink node. The computation capability for taking closed-loop decisions inside embedded devices is very limited. It allows, for example, the selection of which nodes will participate in the sensing and decision, which threshold and conditions will be used to trigger the actuator and the actuation value.图2显示了一个例子的情况的闭环系统，闭环的决定是由汇聚节点完成。以闭环决策在嵌入式设备的计算能力是很有限的。它允许，例如，选择的节点参与感知和决策，阈值和条件将被用来触发执行器和驱动值。图2.在汇聚节点控制决策At the sink node, when a data message from sensing nodes participating in the decision arrives, or at defined time periods, the condition and thresholds are analyzed, and the actuator is triggered if one of the defined conditions is matched.在汇聚节点，当数据消息从传感节点到参与决策，或在规定的时间周期，和阈值的条件进行了分析，如果一个定义的条件相匹配执行器触发。The other alternative for closed-loop control with more powerful resources is to deploy the supervision control logic in one of the distributed PLC outside the WSNs. This alternative is also shown in Figure 3. In this case it is possible to read data from several WSN, to compute a decision based in more complex algorithms, and to actuate over the whole industrial network. The closed-loop algorithm will receive data coming from sensors and will produce actuation commands for the actuator(s).更强大的资源用于闭环控制的另一种方法是在一个分布式PLC在无线传感器网络部署的监督控制逻辑。这个方案也如图3所示。在这种情况下，可以从几个传感器读取数据，计算出一个基于更复杂的算法的决策，并驱动整个产业网络。闭环控制算法将接收来自传感器的数据并将为执行器产生驱动的命令。In this case the control loop may traverse multiple, most probably non-real-time hardware and software systems, nevertheless the control loop will still need to be under expected time bounds.在这种情况下，控制回路可能跨越多个，最有可能的非实时的硬件和软件系统的控制回路，但仍需在预期的时间界限。图3.在整个网络的控制回路In this scenario, the first part of latency (upstream latency) can be sub divided into - WSN latency (time for transmission between leaf node and sink node); latency for sink-gateway (the time taken for the message to go from the sink to the gateway, plus gateway processing time); latency for cabled network (e.g transmission between PLC-gateway and Control Station); control station latency; and end-to-end latency (leaf node to Control Station).在这种情况下，延迟的第一部分（上游延迟）可以分为：无线传感器网络的延迟（叶节点和汇聚节点之间的传输时间）；为Sink网关的延迟（时间为消息走到网关，网关处理时间加上有线网络延迟（）；PLC网关和控制站之间传输）；如控制站和终端到终端的延迟（延迟；叶节点控制站）。The second part (processing time) depends of the closedloop computation to apply, and it can take several seconds. Lastly, the third part (downstream) corresponds to the path used by a command to reach an actuator. This part can be subdivided into several subparts, such as: control station latency, gateway latency, gateway to WSN interface latency and WSN downstream latency.第二部分（处理时间）取决于闭环计算应用，它可能需要几秒。最后，第三部分（下游）对应的命令到执行器的路径。这一部分可分为几个部分，如：控制站的无线传感器网络网关的延迟，延迟，延迟和延迟的无线传感器网络网关接口的下游。In the next sub-sections we discuss how schedule-dictated “time guarantees” are achieved in WSN sub-networks and the issues related with time guarantees limitations and solutions for the global SCADA system.在下一小节讨论进度决定“保证”是无线传感器网络的子网络实现的问题，时间保证的局限性和解决方案相关的全球SCADA系统。A. Schedule based protocol for WSN sub-networksA. 基于无线传感器网络的子网络协议的时间表The medium access control protocols for wireless sensor networks can be classified into two major categories: Contention based and Schedule based. Contention based protocols can easily adapt to topology changes after being deployed, as new nodes join and others leave. These protocols are based on Carrier Sense Multiple Access (CSMA) technique and have higher costs related to message collisions, overhearing and idle listening. In contrast, schedule based protocols (TDMA) avoid collisions, overhearing and idle listening by scheduling the transmission and listening periods to specific