Conversion of Fieldbusses regarding Industrial Internet of Things

Transcription

1 Conversion of Fieldbusses regarding Industrial Internet of Things Ludwig Leurs Ethernet Convergence Bosch Rexroth AG Presented at the ODVA 2015 Industry Conference & 17th Annual Meeting October 13-15, 2015 Frisco, Texas, USA Abstract This paper shows the development of fieldbusses in factory automation and examines how new requirements from Industrial Internet of Things will impact communication technology. It further explores what future developments can be forecasted given that the Industrial Internet of Things defines that machines automatically coordinate and exchange information leading to increased real time requirements. Finally, this paper will examine the possibility of whether Time Sensitive Networking technology ( TSN ) could be the unified hardware architecture for EtherNet/IP. Keywords Fieldbus, Industrial Ethernet, EtherNet/IP, Sercos, OPC UA, Industrie 4.0, Vertical Integration, Time Sensitive Networks Introduction A lot of companies are claiming to support Connected Industry (often called Industrie 4.0 in Europe) or the Industrial Internet of Things (IIoT). This is also true for fieldbus or Industrial Ethernet organizations. But what are the features that need to be supported? Do these systems need to change? Another arising communication technology is Time Sensitive Networks (TSN). TSN allows maintaining determinism with the confidence of being able to satisfy the requirements of less demanding traffic sharing the medium, which means successful convergence of critical control, non-critical control, and data streams on a single network [1]. But TSN also will have different effects and benefits to these systems. To understand this, we need to step back in history and look at the original applications where these systems come from. Properties of fieldbus applications fields and their resulting solutions Fieldbusses were introduced in the late 1980 th to replace hard wired cabling between controllers and devices. The application areas were spread from Simple I/O direct wiring replacement Data exchange between automation controllers Distributed closed loop control for synchronized motion Replacing direct I/O wiring For larger machines it has always been a challenge to directly wire the signals from the PLC to the I/O. In the 1980s several companies developed I/O rack extension system. These could be placed at 2015 ODVA Industry Conference ODVA, Inc.

2 distances of more than 100m apart and were connected to the central PLC via a serial point to point connection. Data ranges were in the range of 10 kbit/s. As customers wanted to be more flexible in choosing their I/O hardware supplier, at the beginning of the 90s a couple of serial busses (fieldbusses) specialized for I/O data transfer turned up: Interbus-S was one of the first. It used a data rate of 500kbits/s and a high efficient communication system for small data packages using a common frame for all devices. The second important I/O system was Profibus DP which was derived from Profibus FMS mentioned in the next section. Profibus DP was developed in several steps: first improving performance by optimizing the protocol and then driving the data rate up to 12 Mbit/s in Version 0. Version 1 added acyclic data access in a rack based I/O station data model. Version 2 tried to extend to motion control, but never succeeded because of compatibility needs with the prevous versions. ODVA started with DeviceNet based on CAN in As everybody knows CAN is the International Standard ISO and widely used in vehicles. DeviceNet was designed for a wide variety of I/O applications: Master/Slave and peer to peer, cyclic and change of state, unicast and multicast communication. The outstanding property is that it has been constructed using a fully object oriented data model. That enabled ODVA to support a whole family of networks with the Common Industrial Protocol in a standardized way. Data transport for I/O messages is efficiently compressed via assembly objects. But information exchange also included. Objects also include behavior and device profiles form an object collection to facilitate interoperable devices from various vendors. Controller to controller communication (C2C) Automation was and still is highly driven by the automotive industry. In the 80s central control was the concept and vertical integration was intended to support this target. The Manufacturing Automation Protocol (MAP) and the Manufacturing Message Specification (MMS) were driven by General Motors and pushed a big hype in the industry. This was a giant project and influenced many architectural items in the fieldbus area. So the first Profibus Implementation was called FMS for Fieldbus Message Specification. It is a subset of MMS (ISO9506) and was intended for communication between automation controllers. The services are grouped into variable access, program invocation, domain management and Virtual Manufacturing Device (VMD) support. As the demand to simplify I/O wiring increased and the C2C communication required a lot of expensive software and processor resources, this communication technology went out of focus for fieldbus standardized solutions. But nevertheless this is the area where standard Ethernet was first introduced. The reasons are: HMI are often built on MS-Windows based PCs and offer an Ethernet interface on board. TCP/IP can be used to build customized interfaces to all systems using Ethernet. This can be useful for communication between machines. CNC, RC and Motion Control applications Coordinated motion using electronic controllers was first introduced in machine tools industries and robotics. As CNCs need very high precision and fast processing time the first implementations used analog interfaces and analog control loops. Figure 1 shows the closed loop control structure of a servo drive controller. The position command values are given by an interpolation task of the CNC in equidistant time steps to the position control. The position control loop gets current position input from the feedback sensor of the motor or directly from the machine axis. The position difference leads to an increase/decrease in speed command value. The difference to the current speed value measured by the motor feedback leads to an increase/decrease of the momentum/current command value to the momentum/current control loop. As speed of the control loops is essential to the performance of the drive controller, the fastest control loops can be implemented locally in the drive controller, because no communication delays are present. In chained / non Cartesian kinematics systems cross influences might limit this ODVA Industry Conference ODVA, Inc.

3 Figure 1: Servo Drive Control Loops Traditionally the position controller resides in the CNC and the inner control loops in the drive controller. The interface used to be the +- 10V analog interface to the drive and the incremental encoder interface from the motor to the controller. It is very important to understand that these signals must match exact points in time for excellent closed loop control, that precision is dependent on fast cycle time and on exact synchronization of the participating axis in the coordinated motion. It can be calculated, that 1µm spatial precision correlates to 1µs synchronization time precision at a processing speed of 1m/s. Coordinated motion requirements of are much higher and also different from the requirements of the I/O wiring replacement. This leads to specialized solutions in this sector where the first one (Sercos 1 ) was shown in Concerning a motion control only application some requirements from the I/O area are not that important as e.g. operating with missing devices and independent start up and shut down of devices. A motion control application normally can t work, if one of the axes is missing. For an I/O based application it is a regular use case to work with some I/O missing. A system that does not cover this case could only be used in few applications. Another difference is configuration. While I/O configuration is done offline in the PLC project, drive configuration is historically done online due to the number of concerned parameters to be individually adjusted (see Table 1). This might be changing as drive systems behavior can be simulated offline. Summary of requirements of the considered application areas I/O C2C Motion Offline configuration X Online configuration X Config at connection X X Cycle time 5-10ms 5-100ms 0.5-4ms Synchronization < ±5% - <1µs Browse Table 1: Summary of Requirements X 1 Other ones are Profidrive, CANopen DS402 and CIP Motion 2015 ODVA Industry Conference ODVA, Inc.

4 Several systems focused on one of these application areas and developed specialized solutions. CIP technology included Information exchange right from the beginning. Its object oriented design includes standardization of the information model, behavior and application profile. This enables CIP networks to cover several physical layers and to operate across network boundary as described in the Bridging and Routing Chapter of the Specification. This also allows to integrate other networks and to extend onto future physical layers and lower end sensor devices (IO-Link, DoT, new wireless technologies). Migration to Ethernet Two requirements lead to introduction of Ethernet to the field device area: Performance: higher data rate needs for increasing the number of devices and/or decreasing the cycle time Convergence: o Elimination of a separate commissioning interface or improvement of the data rate for that purpose, if commissioning was already done via the fieldbus. Serial interfaces were dropped in standard PCs and Ethernet was becoming attractive when replacing the fieldbus and the commissioning interface in the device. o Enable IT integration into devices Web server for diagnostics and maintenance SNMP FTP for firmware update The only problem to consider was real time behavior as Ethernet was said to be non-deterministic. This has been solved in several ways Introducing a Master/Slave MAC layer protocol (PowerLink) Using switched Ethernet and QoS Adding IEEE1588 support for synchronized motion Enabling ultra low latency by using TDMA scheduling and multi-device packets (e.g. Sercos, Profinet IRT) ODVA was one of the early adopters of Ethernet for industrial networks. EtherNet/IP was built solely on publicly available standards and Commercial of the Shelf (COTS) products. This strategy had already been very successfully incorporated into the DeviceNet standard. There might be a disadvantage by not being able to serve every niche application, but the advantage is to gain maximum benefits from the continuous development of the Ethernet standard and not having to solve migration problems. Connected Industries, market requirements and technical opportunities Recent developments show a trend in market development and in technical development. The market trend is the requirements for more individual service for the customer. The technical trend is the vast improvement of communication technology. Mass production of individual items While Henry Ford made cars affordable by everyone through standardization and automation of mass production in an economy of scale, the Japanese focused on flexibility to follow changing markets and invented the Toyota Production System (TPS) which allows making a larger variation of products by optimizing the material flow incorporating human flexibility [2]. This has been adopted by many companies and seems to be common in industry today. The automotive industry reached the point where hardly two cars are exactly the same. This allows higher prices because individuality is a value to the customer. The necessary infrastructure (e.g. product configuration system) is becoming available in the internet also for other products. Direct integration down to the production is not easy and still needs a lot of efforts in defining standardized data structures and flexible manufacturing methods. ODVA SIG Optimization of Machine Integration is seeking to facilitate the informational integration of machines into supervisory systems and also improve M2M communication by defining standardized and extensible data structures. The SmartFactory e.v. in Kaiserslautern Germany has been pushing Industrie 4.0 (Connected Industries) for quite some years and has built an industrial demo line showing flexible production. The line is built by independently operated individual modules connected by a transport 2015 ODVA Industry Conference ODVA, Inc.

5 system that allows flexible arrangement of the modules. Information exchange with the supervisory systems is done via OPC UA. ODVA Machinery SIG has defined data models for easy integration of machinery into supervisory systems. Common members of ODVA and SmartFactory are in the process of testing these data structures in the demo production aimed for Hannover Fair Ubiquitous availability of information The Industrial Internet of Things seeks to speed up all manufacturing and business processes in industry. Besides formal description and management of business rules the key is the ubiquitous availability of information. For the area of fieldbusses this means that all information needs to be accessible. For CIP based devices this is already true. Objects and services allow getting and setting the available data and even allow state changes through the appropriate services. Bridging and Routing is an intrinsic part of the CIP specification and allowed communication between DeviceNet networks at the time other fieldbusses didn t think about the requirement. Today this enables integration of legacy devices into the IIoT world. Additional work is being done in three more areas o The ODVA SIG DeviceNet of Things (DoT) is working on an easy connectable DeviceNet extension for simple, low cost in cabinet devices. o The proposal by Rockwell Automation Extending EtherNet/IP to Resource-Constrained Industrial Things brings EtherNet/IP down to applications that could not be reached by Ethernet based technologies. o Remote diagnostics and control wants to dig deeper down to each bit in the sensor domain. Proximity sensors may need complex configuration but only deliver a single bit as resulting data. The IO-Link device manufacturers have developed their fieldbus independent and fully backwards compatible system for integration into the networked world. The ODVA SIG IO-Link integration is developing a seamless integration. Browsing But this also implies that the information is described and can be found. Today this would be implemented by something like a browse service. Communication protocols like OPC have defined and used this already for quite some time. Some systems like Sercos contain a mandatory directory in each device. This can be used by a controller to detect all available data in the device including access rights, data type, limits, units and names and offer the browse service to the cloud. CIP based networks will have to define this service. The necessary data is already there, but not accessible from the device. As EDS files are the place to contain that information, the device itself or the controller could either store that file or just an URL to the file (see proposal in [3]). That would be a natural solution in the connected world. It can be supposed that all current IT systems support XML data definitions. That means that vertical integration is easy using XML. So either a gateway (in most cases the controller) must translate the EDS defined data to XML or new approaches to CIP device descriptions [4] could be implemented. Turning the wheel back in history we detect that directory services already have been part of the MAP/MMS specification. In some ways Connected Industries is the successor of MAP/MMS. Wireless technologies If the information is needed at arbitrary places, wireless technology is the natural choice. In general e.g. due to local regulations limiting medium access methods in the frequencies available for industrial use, wireless does not yet allow hard real time applications. For less time critical applications as HMI WLAN (802.11) is available as an established standard and in use for EtherNet/IP [5] and OPC protocols. Extending EtherNet/IP to IPv6 in order to connect to worldwide distributed devices has been started some time ago, but there doesn t seem to be a demand at the present time ODVA Industry Conference ODVA, Inc.

6 Future trends New development arising from Audio Video Broadcasting and supported by the automotive industry are allowing hard real time features in the presence of mixed traffic in the next generation of standard unmodified Ethernet. This is called Time Sensitive Network (TSN). The authors explanation for this name is that at frequencies of 1Gbit/s and above time and space are no more to be considered independently and all network elements need to be time sensitive. TSN consist of a set of specifications (see Figure 2), where some are independent and some form a set of related things. The set has been introduced in the previous presentation Time Sensitive Networks (TSN) Protocols and use in Industrial Automation [1]. At this point we want to consider the effect this suite of specifications could have on CIP and related networks. AB Qca br Qbu Qch TSN Q AS Rev Qav Qbv AS Qcc Qci CB Figure 2: Set of IEEE 802 Specifications relating to TSN Qat Legend: TSN AVB Ethernet Transport Generally speaking it can be expected, that networks relying on unmodified Ethernet standards can improve most with only minor addition to the specifications. Now let s look at the new IEEE standards and their effect on Industrial Ethernet Systems. Time Synchronization (IEEE1588, IEEE802.1ASrev) Time synchronization by the Precision Time Protocol (PTP) has been incorporated in nearly all synchronized motion networks with the exception of Sercos III. The idea of these networks is that latency is not a major problem, because it can be reduced by selecting a topology which only needs little number of hops. The synchronization required for synchronous motion is nominated to be in the range of less than 1µs. This can be achieved by already by IEEE1588v2 used in CIP Sync. The necessary hardware support is already available in chips supporting AVB (like Intel i210) and will be continued in TSN chips. Sercos III does not use PTP but hardware synchronization by the Sercos telegram and only works in line or ring topologies. So the number of hops was the reason not to use PTP. Reducing latency (IEEE802.1Qbu, IEEE802.3br, IEEE802Qbv) The problem of standard Ethernet has always been that long frames that have been started to send cannot be interrupted and cause significant delays, e.g. up to 125µs at 100Mbit/s in each switch in the delivery chain. This can be prevented by two ways. Interrupting long frames: this is done in the MAC layer (IEEE802.1Qbu) according to the rules of standard Ethernet and in the physical layer (IEEE802.3br) for improving protocol efficiency. This could bring the maximum hop delay by long packets down to 11.2µs at 100Mbit/s. That can be done with the new TSN chips and without changes in the network specification of EtherNet/IP ODVA Industry Conference ODVA, Inc.

7 Scheduling time slots for transmission of the critical frames. This has been done e.g. by Sercos, Profinet IRT and PowerLink. Interrupting long frames is not necessary, but would improve media availability for best effort traffic. TSN supports this by IEEE802.1Qbv. Systems not using scheduling today e.g. EtherNet/IP could align latency to the above mentioned systems, but would introduce the complexity of scheduling and need quite some specification update. Seamless Redundancy (IEEE802.1CB) As most industrial networks have already introduced methods to satisfy redundancy requirements, the questions are: which advantages would be offered, what is the cost for the change and what does the migration path look like? ODVA has introduced Device Level Ring (DLR, standardized in IEC ) to enable media redundancy in EtherNet/IP ring topologies for fast recovery in the lower ms range. Sercos has built in redundancy by the (optional) ring topology and this is seamless as the output data is concerned. For input data the maximum loss of packets can be guaranteed to 1. As EtherNet/IP and Sercos have already redundancy appropriate to their application use cases the IEEE802.1CB standard would not offer a big advantage, especially if considering the cost of migration. 100Mbit/s versus 1Gbit/s Current Industrial Ethernet systems are built on Fast Ethernet (100Mbit/s) technology. Investigations in 2007 by Jasperneite et al [6] and Danielis et al [7] show that not all Industrial Ethernet systems gain by 1Gbit/s in the same way. This is mainly caused by topological constraints in these systems. For a deeper understanding we have to look at the behavior of delay and transmission time at the relevant data rates. The first fact to note is that transmission time and node delay behave different when going to a higher data rate. In order to limit the maximum used frequency the coding scheme of the data has to be changed in implementations of higher data rates. This leads to increased complexity in the encoding and decoding process. The result is that cut-through delay is about the same at 100 Mbit/s and 1 Gbit/s while transmission time decreases by a factor of 10. µs M-64B 1000M-64B 100M-1518B 1000M-1518B Number of Hops Figure 3: Cut-through packet delay versus number of hops 2015 ODVA Industry Conference ODVA, Inc.

8 Figure 3 shows the packet delay depending on data rate and the number of hops for cut-through forwarding. The considered packet size is 64 bytes for the smallest real time packet (short) and 1518 bytes for a maximum size Ethernet packet (big). At 100 Mbit/s there is a substantial difference between small and big packets. At 1 Gbit/s the node delay is dominating the total delay. 14,000.0 µs 12, , , , , M64B 1000M64B 100M1518B 1000M1518B 2, Number of Hops Figure 4: Store & forward packet delay versus number of hops If we look at the same diagram for store and forward delay shown in Figure 4, we see a similar effect. At 100 Mbit/s store and forward delay is a big problem for hard real time applications like coordinated motion. Being able to split the big packets into chunks of 64 to 128 bytes using TSN technology would really help. But large networks are usually not structured in line topology. So we should look at the range of 1 to 10 hops magnified in Figure 5. This shows, that store&forward switched networks with low number of hops can be acceptable in real time application using 1Gbit/s data rate ODVA Industry Conference ODVA, Inc.

9 µs M64B 1000M64B 100M1518B 1000M1518B Number of Hops Figure 5: Store & forward packet delay (scaled) Example: if we consider a network structured by switches, 100 devices could be connected in star topology by cascaded switches leading to two hops only. The other scenario could be to use one 11port switch in store and forward technology and 10 lines containing 10 devices with internal switches using cut-through technology. To retain maximum real time performance these switches should be based on TSN. This shows that a combination of standard TSN based switches and today s Industrial Ethernet line topology could be an excellent combination. Summary Automation technology developed in several application fields in factory automation, each with its specific requirements which lead to tailored solutions. The demand for more complex applications made Ethernet technology attractive to automation due to performance, price and IT integration. ODVA has always supported the use of COTS technology to enable long term sustainability. Also bridging and routing has always been part of the CIP specification to allow communication across network boundaries. Connected Industries need deeper vertical integration than conventional requirements from Computer Integrated Manufacturing demanded. CIP technology is ready for this. New requirements as discovery services can be resolved by gateway functions based on device descriptions or added as functionality in the devices itself. Increasing demand for M2M communication will need more Real Time support for the machine network. This will be a field to introduce Time Sensitive Networks. But TSN can do more. It can serve CIP Motion to minimize latency and enable faster control loops. It can be used as unified transport media for most Industrial Real Time networks and therefore enable new topologies using distributed controls. TSN will help to converge the formerly separated application areas of fieldbusses. References [1] G. A. Ditzel und P. Didier, Time Sensitive Network (TSN) Protocols and use in EtherNet/IP Systems, ODVA Industry Conference, Frisco, TX, ODVA Industry Conference ODVA, Inc.

10 [2] T. Ohno, Toyota Production System: Beyond Large-Scale Production, Portland, OR: Productivity, Inc., [3] S. Zuponcic, L. Leurs und R. Beudert, Machinery Information Base data Structure, ODVA Industry Conference, Frisco, TX, [4] R. Blair, A Modern Approach to CIP Device Descriptions, ODVA Industry Conference, Frisco, TX, [5] P. Brooks und P. Didier, n - Wireless Performance for Control?, ODVA Industry Conference, Phoenix, AZ, [6] J. Jasperneite, K. Weber und M. Schumacher, Limits of Increasing the Performance of Industrial Ethernet Protocols, in s Emerging Technologies and Factory Automation, ETFA. IEEE Conference on, Patras, Greece, [7] P. Danielis, J. Skodzik, V. Altmann, E. B. Schweissguth, F. Golatowski, D. Timmermann und J. Schacht, Survey on Real-Time Communication Via Ethernet in Industrial Automation Environments, in s Emerging Technologies and Factory Automation, IEEE Conference on, Barcelona, Spain, ****************************************************************************************************************************************************** The ideas, opinions, and recommendations expressed herein are intended to describe concepts of the author(s) for the possible use of ODVA technologies and do not reflect the ideas, opinions, and recommendation of ODVA per se. Because ODVA technologies may be applied in many diverse situations and in conjunction with products and systems from multiple vendors, the reader and those responsible for specifying ODVA networks must determine for themselves the suitability and the suitability of ideas, opinions, and recommendations expressed herein for intended use. Copyright 2015 ODVA, Inc. All rights reserved. For permission to reproduce excerpts of this material, with appropriate attribution to the author(s), please contact ODVA on: TEL FAX odva@odva.org WEB CIP, Common Industrial Protocol, CIP Energy, CIP Motion, CIP Safety, CIP Sync, CompoNet, ControlNet, DeviceNet, and EtherNet/IP are trademarks of ODVA, Inc. All other trademarks are property of their respective owners ODVA Industry Conference ODVA, Inc.