Technologies For Managing Field-Ubiquitous Computing

NOGUCHI Akira¹ OKABE Nobuo¹ YAMAGUCHI Kenji² YAMAMOTO Shuji¹ OHNO Takeshi¹ OHTANI Tetsuya³

We have long been developing field computing technologies for industrial systems. The new paradigm of ubiquitous computing is also making inroads into the domain of field computing. In this paper, we describe our vision for field-ubiquitous computing from three aspects: 1) establishment of a network infrastructure based on open standards, 2) computing architecture for more flexible system expandability and common computing platforms for field facilities and 3) operation support technologies using online plant models.

1. Ubiquitous Field Computing Research Center, Corporate R&D
2. Administration & Technical Information Dept., Corporate R&D
3. Instrument & Control Research Center, Corporate R&D

I ENABLING FIELD-UBIQUITOUS COMPUTING

INTRODUCTION

Over recent years, in line with the new IT evolution trend, more and more network infrastructure is being installed as a key infrastructure for society. Thanks to this network environment, the concept of "ubiquitous" has emerged as one of the new paradigms in technological trends. The word "ubiquitous" means ever present or occurring everywhere. In other words, ubiquitous means utilizing information in a manner in which it is accessible from anywhere at anytime, without being aware of the actual accessing of that information, while conjuring up the lifelike images and sensations that are beyond time and space, from user's standpoint, and by meeting the changing requirements flexibly and effectively.

In the area of measurement and control systems that pertain to manufacturing systems, there exists a great deal of information and computing resources in the field, including sensors, controllers, and actuators. We believe that these resources in the field should be easily accessible and available from anywhere via networks and that field computing in the future will comprise a system environment that makes these resources available to more people and broader objectives effectively when necessary.

This paper describes our work on a new computing architecture that will facilitate such a field-ubiquitous computing environment.

21ST CENTURY APPLICATION FIELDS

Figure 1 Ever-Spreading Field

As Internet Protocol (IP) has gained widespread use in information and public systems, IP networking is also gradually becoming a core technology in network installations for field applications. The spread of IP networking has enabled field systems to exchange information without the constraints of physical distance or system layers, such as connections with remote locations over the Internet or direct data exchange with upper-layer information systems such as corporate servers. In addition, the emergence of autonomous controllers with Web server or mail functions has allowed users to communicate directly with equipment, enabling direct access to a wide variety of networked equipment. These direct access capabilities of such networked equipment allow users to directly select and utilize field information, broadening its potential usability.

At the same time, the requirements for manufacturing production systems have changed over time. These requirements are no longer limited to production operations, but are now coming to include safety, energy conservation, traceability, environmental protection, and an even broader range of issues (Figure 1). This trend will continue and require field systems to address various increasing problems in the future.

Therefore, it is necessary to consider a broader range of areas for field computing when developing system solutions which address these issues. These systems need to be designed to handle a variety of objectives under a common architecture operating through a networked environment, without limiting field computing to the operation of equipment-specific functions.

With an acceleration of the use of IP-based field networks, if individual field devices could be connected in a single network and those individual devices and systems could autonomously transmit information over the network, then a variety of field information could be accessed and used for field solutions over the network. Another important benefit of ubiquitous field computing in the future will be the flexibility to use any combination of components, including temporary sensors, for problem solving.

Field-ubiquitous computing in the 21st century will mark a shift away from solutions based on proprietary systems toward a field computing environment that can grow and be updated to handle the issues of numerous users, with an emphasis on flexibility and interconnectivity on the network.

CHALLENGES FOR ENABLING FIELD-UBIQUITOUS COMPUTING

Field-ubiquitous computing is premised upon the notion that a network environment based on open standards will pervade all areas of the field. Below we address the issues of ensuring that a network infrastructure can be established in the field networks, and creating an architecture that can be flexibly adapted to changes based on this network infrastructure.

• Establishing the field network infrastructure
  In order to ensure system interoperability, it is important to establish a network environment based on open-standards technologies as the field network infrastructure. Therefore, we will focus on IP and wireless technology as open-standards technologies.
  The challenge is to ensure system reliability and operability in the field by incorporating these technologies into equipment with limited computing resources that is installed in environments other than office environments, such as outdoors or in explosion-proof areas.
  Another anticipated technical field is wireless communications which enable the collection of information from high locations and other hazardous areas where it is difficult to run wires, as well as the installation of temporary sensors for diagnosing and analyzing equipment which is difficult to wire due to high costs. From this viewpoint, we are working on the implementation of wireless communications in the field, based on IEEE 802.15.4, a low-power wireless communications standard that enables fully wireless implementations (including power supply) by using batteries or the excess power of local equipment as a power supply.
• Flexible adaptation to a variety of problems
  When the entire field is connected in a single network, the conditions are in place to build a new system for problem solving, overlaid on existing systems. However, it is not enough to simply create a network in the field; the system must be designed so that devices distributed in the field can autonomously expand and increase in functionality. The key lies in establishing this autonomy in functionality as a common platform which is independent of equipment implementation. The aim of realizing this computing environment is to enable problem solving systems to be dynamically added and built. In such an environment, application programs for solving a variety of problems will be able to be easily added and updated on devices distributed throughout the network, and will have their operations dynamically linked to each other. In addition to this requirement, the system should provide a bird's eye view for managing the operations of these problem solving systems.
• Real-time utilization of plant models
  In the final form of a ubiquitous environment, it is essential to rearrange and provide raw information from the field according to the perspectives of each user. We are developing technology that the application of simulation technology in real time enables the inference of data that cannot be measured from existing sensor data. We are also developing technology that predicts future states from current states.
  In addition, we have created a prototype of a transparent operations based on essential plant information deduced using simulation technology. Under transparent operations, the entire factory or plant is modeled as a virtual environment independent of the physical configuration of devices and equipment. In response to operation commands, concrete processes corresponding to the physical configuration of devices and equipment are performed autonomously at each factory or plant. Plant-specific controls are locally adjusted, so it should be possible to reuse operations expertise in multiple plants.

CONCLUSION

Our work has focused on introducing network technology based on open-standards technology in the field, and achieving new, paradigmatic changes in a field environment. In the follow sections, we discuss our work in greater detail.

 

II PROSPECTS FOR IP TECHNOLOGY AND FIELD IMPLEMENTATIONS

INTRODUCTION

Internet Protocol (IP) and Ethernet which have been designed for Local Area Network (LAN) is deploying to both Wide Area Network (WAN) and inter-hardware communications, thereby enabling networks to be scalable and seamless from a technical point of view.

IP and Ethernet are also expected to be a part of the major network technologies in Process Automation (PA) and Factory Automation (FA), which are control systems supporting corporate infrastructure (referred to below as "field systems"). In this section, we discuss trends in IP and Ethernet, their importance to field systems and related issues, and the areas of research we are focusing on.

IP AND ETHERNET

Market Penetration

Ethernet won out against competing technologies, such as FDDI and ATM, for several reasons: 1) maintaining backward compatible with older versions of Ethernet; 2) selection of technology giving priority to cost performance; 3) assurance of growth potential due to simplicity. In addition, IP won out against competing technologies, such as NetWare and OSI, due to: 1) providing an open process for establishing specifications; 2) participation of numerous vendors; 3) assurance of growth potential due to simplicity. Because IP and Ethernet beat competing technologies, they secured a very large market and became the critical LAN technologies.

Trends

Beginning with the 10 Gps specification, Ethernet has entered the WAN arena as well. Although Ethernet specifications set 40 km as a maximum distance, in actual practice this can be extended to 100 km, and it is expected to gain market share due to better cost performance than competing technologies such as Synchronous Optical Network/Synchronous Digital Hierarchy (SONET/SDH).

IP and Ethernet are being applied in areas such as internal buses for connecting hardware devices together. Examples include Remote Direct Memory Access (RDMA), InfiniBand, and iSCSI. IP overhead (i.e., code size and processing speed), which has been considered problematic in embedded applications, is tried to be solved by a hardware-enabling technology called TCP/IP Offload Engine (TOE). Going forward, if TOE gains general acceptance, IP technology should become more popular for embedded applications in small devices.

From a field systems perspective, IPv6 has the following advantages over IPv4:

• NAT-independent network
  The address space in IPv4, 32-bit address, is not large enough, making Network Address Translation (NAT) necessary. NAT causes a loss of network symmetry, imposing various operational burdens on. Complexities of network configuration and operation due to NAT are always serious issues. A vast address space which IPv6 provides, 128-bit address, makes NAT unnecessary. Consequently, IPv6 can eliminate the abovementioned complexities.
• Unique local addresses
  Unique Local Unicast Addresses (ULA) in IPv6 allows global uniqueness to be provided even in local address space not connected to the Internet. This makes it possible to avoid the IP address renumbering and NAT, when connecting with private networks.
• Stateless address auto-configuration
  Stateless address auto-configuration in IPv6 assigns IP addresses to network interfaces autonomously without any specific infrastructure. This means that IPv6 node requires neither manual configuration nor Dynamic Host Configuration Protocol (DHCP) server for IP addresses, and can make network configurations simplified.
• Router discovery
  Router discovery in IPv6 maintains router information autonomously without any specific infrastructure. This can also eliminate manual configuration and DHCP server, then contributes to network simplicity.

IMPORTANCE FOR FIELD SYSTEMS

Two advantages to using IP and Ethernet in field systems are discussed below.

Network Configuration Flexibility

Flexibility in the configuration of field systems will gain importance because the critical factors in competition are continuously changing with the globalization of market competition. In light of the market penetration and advances in integrating technologies such as Enterprise Resource Planning (ERP) and Manufacture Execution System (MES), the connectivity of current production systems could be a focal point for vertical integration and horizontal integration. IP and Ethernet can achieve these types of integration.

Cost Performance

IP and Ethernet have enormous markets and provide superior cost performance and technological advances. Therefore, in the long term, IP and Ethernet will be very important in field systems as well. For example, hardware-enabling technologies like TOE and the two-wire metal cable technology specified by IEEE 802.3ah may be useful in field system from a viewpoint of application.

CHALLENGES IN FIELD SYSTEMS

1. Evaluation of Ethernet and study of alternate link layer technology
   • Topologies
   The latest Ethernet topologies are star and ring. In field systems, there is also a need for a bus topology for ease of wiring and other reasons; however, this requiresconsideration of constraints (e.g., reliability, maximum distance) particular to this topology.
   • Power consumption
   Ethernet has high power consumption because of its high- performance. In cases where there are constraints on the electric power consumed by the devices used in a field system, an alternate link layer technology is required. If the system uses IP as a network layer, then differences in link layer technologies are hidden, so applications will not be affected.
   • Intrinsic safety
   The issue of whether Ethernet can support intrinsic safety and the possible need for alternate link layer technology needs to be studied and evaluated.
2. Dependability
   If a field system which used to be a closed system becomes an open system, the system will be heterogeneous. Furthermore, if a system becomes heterogeneous, communication infrastructure based on system reliability will be more complicated. For example, security and Quality of Service (QoS) have to work under heterogeneous environment. We refer to "dependability" as being assured by multiple factors, in order to differ from conventional reliability. When considering an open system, factors which the system makes dependable (e.g., security and QoS) have to be fully interoperable.
3. Security
   Firewall which assume specific network topology is not always a complete solution for network security. It is challengeable to manage security of normative devices with firewall. In addition, Wireless technologies can expose network traffic behind a firewall easily. Therefore, end-to- end security mechanism which do not need to assume any specific network topology are necessary. Furthermore, the security mechanism has to be carefully chosen to be suited to small embedded devices commonly used in a field system because of their limited computational performance.
4. Real-time response for IP
   In embedded applications, the overhead related to IP processing is often a problem. As hardwired IP such as TOE is available , it could be one of the solutions for achieving a real-time responsivity (i.e., response on the order of 10 ms) required for small embedded devices commonly used in a field system.
5. Scalability
   The typical number of devices used in a field system is expected to continue increasing steadily. This increase makes engineering work more complex, resulting not only in higher costs but also increasing the potential for human errors.

RESEARCH AND DEVELOPMENT

Figure 1 Research Topics

Based on the foregoing analysis, we have conducted R&D on the technologies required for future field systems using IP, especially IPv6, and Ethernet as the basic network technologies (see Figure 1).

1. Dependability
   Referring to dependability, we are researching and developing as followings: 1) QoS mechanism which assumes that multiple systems share common networks, 2) end-to-end security mechanism which is suited to small embedded devices commonly used in a field system, and 3) network measurement which can detect Denial of service Attack (DoS) against a field system/networks.
2. Plug and play
   With respect to the issues discussed in ealier section, we are researching and developing a mechanism for automatically configuring devices. The mechanism also cares network security which is suited to small embedded devices.
3. Field-oriented Offload Engine (FOE)
   First, our efforts have been directed toward implementing software on silicon. It can be usefull for small embedded devices in a field system to use hardware-based technology not only for IP (i.e., TOE- like approach), but also for above the IP layer (i.e., implementing plug and play, security or applicationon a silicon). Second, the efforts has also been focused on implementation of software efficiently in size and speed.

CONCLUSION

Although IP and Ethernet are important network technologies, simply using them does not ensure that a system will be competitive. This paper has presented a number of topics to show the potential for proprietary developments. We are focused on researching and developing technologies that will help Yokogawa through greater product competitiveness, using network technologies as a base.

• Product names and other names appearing in this document are the trademarks or registered trademarks of their respective.

 

III DEVELOPMENT OF AN IPv6 CHIP

INTRODUCTION

We have developed a research prototype of a dedicated IPv6 chip to serve as a strategic device for expanding the IPv6 market. The IPv6 chip is a TCP/IP Offload Engine with IPv6 (not IPv4). The TCP/IP stack, which is implemented as software in conventional designs, consists of hardware logic on a chip. This approach makes it possible to eliminate communication processing from the host CPU and to incorporate IPv6 easily.

FEATURES

Figure 1 Example of Simple IPv6 Node Configuration (LAN/RS232C Converter)

Taking advantage of the parallel processing and high speed enabled by a hardware configuration, we added functions to the IPv6 chip that cannot be implemented through software. At the same time, it is possible to achieve extremely low-power consumption by reducing the operating speed. In addition, because of the growth of integration in today's integrated circuits, we were able to integrate peripheral functions for improved functionality and performance (Table 1). These features are described below.

The idea behind IPv6 is to network everything, therefore the transmission of small-size data should suffice for most of the IPv6 nodes expected to emerge in the future. The IPv6 chip is intended to be easily applicable to such tiny nodes with an 8-bit bus for interfacing with the host CPU, which must be a low-end MCU(Micro Controller Unit). Figure 1 shows an example of a tiny node configured using the IPv6 chip. Since the IPv6 chip contains the Ethernet Media Access Control (MAC) and Physical Layer (PHY) functions, it is possible to configure the LAN/ RS232C converter using only a few low-cost parts in addition to the IPv6 and MCU chips.

Table 1 IPv6 Chip Specifications

Item	Specification
Protocol	Multi-session TCP, UDP, ICMPv6,MLDv1,MIPv6 for correspondent nodes
	IPv6 core protocol, IPsec transport mode, Encryption accelerators (AES, 3DES, SHA-1)
	Ethernet MAC	IPv6 over PPP/AHDLC
Network interface	100BASE-TX	Telephone modem interface
Bus interfaces	8-bit data bus, 14-bit address bus
Power supply	3.3 V/2.5 V

In the ubiquitously networked world of the future, secure data communication will be essential even for tiny nodes. However, encryption of Security Architecture for Internet Protocol (IPsec) is beyond the capabilities of CPUs like MCU. The IPv6 chip has a built-in encryption process which takes advantage of the hardware-based implementation. Encapsulating Security Payload (ESP) is processed automatically and quickly on the chip, so the host CPU only needs to handle plain-text data. The IPv6 chip also includes a function which supports the tiny node's key negotiation protocol called Kerberized Internet Negotiation of Keys (KINK).

Figure 2 Field Sensor System Application

The way the host CPU controls the IPv6 chip is simple. Control is based on exchanging parameters, such as the IP address and port number, and payload data between them. The IPv6 chip has the flexibility to support both simple and high-level applications. This flexibility is characterized by the ease of modifying any existing product when adding IPv6 communications functions to it. This is because the IPv6 chip is less dependent on the product's CPU or operating system.

Yokogawa has a product group called field instruments which occupy a critical position in our solutions. In the past, most of these products transmitted signals using analog DC current values. However, in recent years there have been advances in data communication system such as FF (Foundation Fieldbus), and IP- based solutions named High Speed Ethernet (HSE), which are also gaining widespread attention. The IPv6 chip is positioned for the future trend toward IPv6-based HSE system, and incorporates field instruments functions such as the following (Figure 2).

• Low power consumption to satisfy intrinsic safety specifications
• 2-wire data communication system using IPv6 over AHDLC (Asynchronous High-level Data Link Control procedure)

CONCLUSION

We developed a TCP/IPv6 Offload Engine chip conforming to IPv6 Ready Phase 2 Core Protocol. This chip has a simple host interface and a integrated physical layer, making it easy to implement IPv6 communication functions in legacy products by reusing their resources. The chip uses IPsec for encryption and supports the KINK key negotiation protocol, making it possible to create a secure network. The chip also has an IPv6 over PPP/ AHDLC function and is designed for low power applications, so it can be used to implement IPv6 in future field instruments.

 

IV WIRELESS FIELD NETWORKS

INTRODUCTION

W ireless communication technology is used to provide very convenient usability in a wide variety of applications, such as office wireless LANs and cellphones. These communications include applications involving high transfer rates for transmitting voice and video communications. The wireless devices used in these applications are powered either by power lines or batteries for short time periods.

In field applications, the focus is on wireless sensor networks. The information used in such applications is not video or voice, but primarily data such as temperature values and process data that can be processed using extremely narrow bandwidth. It should also be noted that field environments are extremely difficult for wireless communications as they contain lots of electromagnetic noise and radiowave obstacles. Given these factors, the wireless devices used in such field applications should be capable of operating on battery power for a period from several months to several years.

We are conducting R&D on field wireless function elements and system technologies that can be used in such field systems.

IMPORTANT CHALLENGES

Figure 1 Comparison of Wireless Systems

Wireless field networks may have a variety of requirements depending on the application. Our research focuses primarily on providing optimal solutions to satisfy the following requirements:

1. Figure 2 I EEE 802.15.4 Function Layers

   Coverage area and point count
   The system needs to be compatible with a variety of coverage areas, ranging from small-scale sites with a coverage area of several tens of square meters, to large-scale sites several square kilometers in area, as well as pipelines that may be several hundreds of kilometers in length. Within the coverage area, there may be anywhere from several tens to several tens of thousands of points providing sensor data such as temperature, flow rate and other physical quantities, as well as points providing switch states or device operating states.
2. Transmitted information
   The information being transmitted is rather primitive, consisting of data such as contact states and analog data. The amount of information is small, on the order of several tens to several hundred bytes. The information is used for purposes such as system monitoring, and collecting and setting maintenance data and other supplementary information. Access is no more frequent than once every several seconds, and no less frequent than once every several tens of seconds to once every minute.
3. Communication quality
   Because the field is an open area, the system must be designed to prevent unauthorized access from prohibited locations and perationo. In addition, the system must be able to withstand data loss and interference in the data path due to noise and other factors in the field environment.
4. Installation and operation procedures
   The field environment may contain numerous metal obstacles such as pipes or tanks. These objects can block or reflect radiowaves, so it is important to properly determine installation positions, antenna directivity, and other factors. In addition, the operating durations of sensors, devices, and other equipment can vary widely, from extremely short durations on the order of several days to longer periods up to several years. For these durations, it is preferable that the equipment operate without battery changes. In cases where the power required for wireless communications is supplied by the field device, the maximum power is several tens of mW, so the design must enable communication under low power consumption. In addition, the installation of wireless functions should be simple and not require complex configuration of the communication system.

In order to meet these requirements, we selected IEEE 802.15.4 after consideration of many different wireless communication technologies (Figure 1). We made this selection due to: 1) the assumption that the equipment will operate under battery power for a long period of time; 2) the fact that up to approximately 60,000 nodes can depend on a single network; 3) IEEE 802.15.4 provides a multi-hop function for covering wide-area fields; 4) IEEE 802.15.4 is a global standard.

IEEE 802.15.4 FEATURES AND USAGE

Figure 3 Wireless Node

IEEE 802.15.4 consists of wireless communication hardware (physical layer) and functions for using this hardware (MAC layer) (Figure 2). The MAC layer controls communications between point to point, while the network layer uses the MAC function to control communication among multiple points. The network layer has various network topologies which are useful in field applications, such as tree and mesh topologies. An application platform and various applications can be set above the network layer. We use the features of IEEE 802.15.4 to provide the abovementioned characteristics needed in the field in the network layer and application platform.

 

 

FUNCTION ELEMENTS

Figure 4 Wireless Field Device

We created prototypes of a wireless node (Figure 3), a function element in a wireless field network, as well as the elements required for building a system. The wireless node is approximately 3 1.8 cm small enough to be embedded in a field device. Figure 4 shows an example of a wireless node attached to a field device.

1. Wireless node
   The wireless node consists of a low-power 16-bit MPU with 56 KB ROM and 4 KB RAM, and a wireless transmission chip with IEEE 802.15.4 functions. The software in MPU has a multi-hop function for communicating via multiple wireless nodes with locations not directly accessible by radiowaves, and a plug & play function that eliminates the need for network configuration and allows the wireless node to automatically link to the system and start operating. The plug & play function has an interface allowing detected node information to be transmitted to a higher-level application, so the application and wireless node can connect dynamically. In addition, the wireless node is designed to enable standalone acquisition of permissions required by the Radio Law.
2. System integration technology
   A wireless field network must connect with device server and control system networks for purposes such as facilities management. We have designed an IP network gateway protocol to serve for making this connection. This gateway makes interconnection between the wireless nodes and the IP applications. We also designed prototype tools allowing real- time monitoring of wireless communication packets for use in installing wireless nodes and monitoring operations.

WORKING TOWARDS COMMERCIAL APPLICATIONS

As described above, we designed prototypes of the basic functions required for building a wireless field network. Going forward, we will work on implementing new concepts in the field using these function elements. Below are basic descriptions of a temporal sensor and maintenance sub-line, which are priority projects. The basic system configuration is shown in Figure 5.

Figure 5 Basic System Configuration

1. Temporally sensors
   A temporal sensor (Figure 5-a) serves as short-term monitoring concerned points during periodic maintenances or when a plant abnormality occurs. This application requires a battery-powered function for installing the sensor wirelessly with a power supply, and a plug & play function for simplifying operations. In order to use a battery power supply over a long time period, we developed a method for intermittently operating the sensor and a method for synchronizing the operation timing in the system. The plug & play function is implemented by a method whereby the sensor is detected dynamically, allowing the monitoring system to monitor data simultaneously with the sensor installation.
2. Maintenance sub-line for analog field devices
   We have developed a wireless network function (Figure 5-b) for a maintenance sub-line, which adds remote diagnostics/ checking functions to sensors or actuators connected through a 4-20 mA analog signal. The power for the wireless node is supplied by the surplus power of the 4-20 mA signal of the device, so additional power line is required. This function facilitates installation of digital communication lines. As a result, even existing devices using a 4-20 mA signal can be incrementally upgraded with sophisticated self-diagnostic functions supporting FOUNDATION Fieldbus.

CONCLUSION

There are high expectations for the wireless field network market. Going forward, we will use the prototype function elements to identify constraints and issues for each application, and to provide solutions. In particular, we will focus on developing synergisms between wireless networks and F OUNDATION Fieldbus, or 4-20 mA signal wiring—the standard technology in the measurement and control field—. In addition, we will continue our efforts to gain broad acceptance for the standardization of wireless field networks through the ZigBee Alliance¹ which is based on IEEE 802.15.4, being one of the wireless networking standard, as well as employed by us.

REFERENCE

1. http://www.zigbee.org/

• Product names and other names appearing in this document are the trademarks or registered trademarks of their respective companies.

 

V A FRAMEWORK FOR EVOLVING SYSTEMS

INTRODUCTION

Figure 1 Systems with Complex Overlays

Over the years, managers of on-site production systems have striven to improve product quality, throughput, and efficiency. Improvements have been achieved by using networks to achieve vertical interconnection of on-site production systems and corporate management systems, horizontal interconnection of upstream to downstream processes, and utilization of information based on experience. If these improvements can be considered to be the first phase of networking-based improvements, we are now witnessing the start of the second phase of improvements, based on ubiquitous computing.

What is now needed is a multifaceted approach to a variety of problems surrounding conventional production practices, free of the constraints of the existing system framework (Figure 1). There are numerous challenges at production sites, including inventory which stagnates between processes, the need to improve the safety of production facilities, environmental issues, and energy problems. These problems grow in variety over the years. In many cases, those problems cannot be predicted prior to the actual operation and constantly crop up as new challenges. In some cases, solving such problems requires collecting and analyzing detailed field data and expanding functions to cover across the various existing systems, going beyond the conventional framework of the systems.

In the past, whenever a problem was addressed, the systems would be deactivated and have undergone the various modifications in existing functions or device configurations with respect to each device. However, in order to accommodate changes which take place frequently, it is necessary to develop a method for adding functions to each device, and a method for managing their operations through a network. Therefore, we have developed a pragmatic mechanism we call Field Overlay Architecture (FOA) which composed of two main functions: 1) functions as a common computing environment for add-on functions in devices, 2) overlays, on existing systems, a system for solving problems using this common platform. An overview of FOA is presented in this section.

FOA FEATURES

Figure 2 Mechanism for Implementing Add-on Functions

In the first phase of improvements, the platform for solving problems was connections between networked processes, and an establishment of process information management system for storing a variety of field data. Improvements to production processes themselves were made on this platform.

In the second phase of improvements, a computing environment that implements add-on functions when needed is the key to solving problems which occur frequently in relation to production processes.

FOA is used to build a computing environment for add-on functions in devices forming a system, so that an existing system can be expanded.

The features desired in this environment are as follows: 1) The environment should be a common platform independent of the individual architecture of the various devices found in a plant, so as to maximize the reusability of the add-on functions being implemented. 2) Implemented add-on functions should be able to communicate each other across various devices, and should be able to operate independently of each other as it is likely that various functions will be added as the system grows. 3) The environment must be designed so that the installation and execution of add-on functions does not adversely affect the real- time response and reliability required in the existing system.

FOA MECHANISM

This section discusses the FOA mechanism from two perspectives.

The first perspective focuses on the program execution environment of each individual device, and the mechanism for implementing add-on functions on devices.

The second perspective focuses on the operations of the problem solving system to be achieved across various devices incorporating the abovementioned mechanism. From this perspective, a mechanism we call "system overlay", whereby various problem solving systems are overlaid on a existing system, is described.

Mechanism for Implementing Add-on Functions

In order to incrementally implement add-on functions in addition to functions in existing devices, a program execution environment called the "FOA EmbeddedModule" is embedded to the devices. The structure of this environment is shown in Figure 2.

This FOA EmbeddedModule is provided on an interpreter as a common computing environment for heterogeneous devices. This makes it possible to implement add-on functions independent of a device's architecture (CPU, operating system, etc.). In addition, the functions can be still retained even when the devices are replaced.

Programs running on the FOA EmbeddedModule mechanism use the information and functions of these devices and multiple programs running in parallel are able to form add-on functions. In addition, the FOA EmbeddedModule has a function for downloading programs and their settings via the network, and a function that allows various operation commands (program start, stop, etc.) to be executed via the network.

Programs on the FOA EmbeddedModule are referred to as FOA Elements. Elements are installed as application components for forming add-on functions. They have communication link interfaces which are used to exchange data and event messages between the Elements. Elements communicate together with these communication link interfaces to form add-on functions.

The FOA EmbeddedModule provides an environment in which these add-on functions are executed independently of each other. This environment is called FOA ElementSpace. There are multiple ElementSpaces in a single FOA EmbeddedModule. These ElementSpaces can execute and control Elements without affecting each other.

System Overlay Mechanism

In order to form problem solving systems, Elements on devices exchange messages across multiple devices via the network. Problem solving systems differ in terms of the details of their particular problems, system device configurations, and the users in charge of them, so they must operate independently of problems and users in charge of which systems. Therefore, it is necessary to have an environment which provides a mechanism that controls operation access privileges for each system, and controls the assignment of computing resources and the functions already present in devices. This environment is called FOA SystemSpace and shown in Figure 3.

Figure 3 System Overlay Mechanism

SystemSpace is composed of multiple ElementSpaces generated in devices. Therefore, Elements operate under the ElementSpace environment, and communicate with other Elements in the same SystemSpace to form add-on functions.

Multiple SystemSpaces can be overlaid in multiple devices to operate multiple problem solving systems in parallel. This system operation configuration is called "system overlay," and the realization of this type of operation is a goal of FOA. The FOA ManagementSystem is a management function for realizing the system overlay mechanism.

The FOA ManagementSystem functions as a register for multiple SystemSpaces, ElementSpaces, and Elements which form problem solving systems. It contains a mechanism for downloading this configuration information and Element programs via the network to the FOA EmbeddedModule, which executes them. In addition, the FOA ManagementSystem has functions for knowing and monitoring system configuration and operation states.

FOA ManagementSystem functions are provided separately for the FOA administrator, who manages the overall FOA system, and the SystemSpace administrator, who is responsible for problem solving systems. Only the FOA administrator has a privilege that assigns SystemSpaces to SystemSpace administrators for defining operating conditions such as computing resources and access privileges for each SystemSpace. SystemSpace administrators have privileges for building and operating within their assigned SystemSpace, independently of other SystemSpaces.

EXAMPLE OF INDIVIDUAL MECHANISM IMPLEMENTATION AND OPERATIONS

We have developed a prototype of the basic FOA functions. Figure 4 shows the system configuration. An example of operations using the currently implemented mechanisms is described below.

Figure 4 Configuration of Current Implemented System

Implementation

We have developed a prototype of the FOA EmbeddedModule on an embedded experimental device running SH3/Linux. The prototype FOA EmbeddedModule was achieved by performing a system overlay through functional enhancements of Java Embedded Real-time Operating System (JEROS), an application execution environment which uses the Java interpreter. JEROS has a mail function and Web server, and provides flexibility for using Java programs to add add-on functions for transmitting information from embedded devices. JEROS is already embedded in Yokogawa products such as DUONUS¹ and STARDOM².

In addition, the FOA ManagementSystem was implemented on a PC. The FOA ManagementSystem consists of management consoles that allow administrators to control configuration data, a database for managing configuration separately for each SystemSpace, and a download server for downloading Elements and configuration information to the FOA EmbeddedModule.

Operation Example

In the example described below, FOA is used in a production system to overlay a failure trend diagnostic system as a problem solving system. First, the flow up to the point of system operation is described.

1. The FOA administrator uses the FOA management console to generate a new SystemSpace to be assigned to the diagnostic system, and ElementSpaces for SystemSpace are generated in each FOA EmbeddedModule. Next, the FOA administrator assigns SystemSpace administrator access privileges to the equipment maintenance system manager. This SystemSpace is used to install and operate a diagnostic system.
2. The equipment maintenance system manager, through the SystemSpace management console on a PC in his/her department, configures the deployment and link relations of the Elements forming the diagnostic system. These Elements are used to collect data, make judgments, and provide notifications. The defined configuration information is stored in a database on the FOA ManagementSystems.
3. After registering the configuration information and Element programs in the download server, the equipment maintenance system manager sends the FOA EmbeddedModule download commands and Element activation commands. Based on the downloaded configuration information, the FOA EmbeddedModule generates Elements in the ElementSpace generated in step 1. These Elements start communication with other Elements.
4. The diagnostic system begins to function. The operations after the diagnostic system begins functioning will now be described. The equipment maintenance system manager monitors the diagnostic system configuration and operating states by using the FOA ManagementSystem as necessary.

If a device is replaced during operations, it is only necessary to download the previously used configuration information and Elements to the FOA EmbeddedModule on the new device for resuming operations. Upgrading of Elements on devices performed in conjunction with application updates can be done within one SystemSpace at a time via the network while other systems are operating and without deactivating devices.

CONCLUSION

We devised an architecture called FOA for overlaying new problem solving systems on a current existing system and verified its effectiveness with a prototype of FOA. In the future, we plan to improve FOA functions in preparation for commercial implementation, and will work to address the portability of the FOA EmbeddedModule to a wide variety of devices.

In addition to arrangements where the FOA EmbeddedModule is embedded on individual devices, we also plan to work on arrangements in which a dedicated device containing only the FOA EmbeddedModule is externally connected to existing devices.

REFERENCES

1. Noguchi A., Ibaraki M., Ohno T., Iwamura T., "Industrial Network Computer 'DUONUS' ", Yokogawa Technical Report, Vol. 42, No. 4, 1998, pp. 173-178 (in Japanese)
2. STARDOM Feature Edition, Yokogawa Technical Report, No.33, 2002, pp.1-16

• 'DUONUS' and 'JEROS' are registered trademarks and 'STARDOM' is a trademark of Yokogawa Electric Corporation. Other product names and other names appearing in this document are the trademarks or registered trademarks of their respective companies.

VI REAL-TIME UTILIZATION OF PLANT MODELS

INTRODUCTION

Production sites become highly-automated in materials process industries, such as oil refining, petrochemicals, steel, and paper mills. However, in many cases people still have to handle error conditions of plants or make final decisions on production plans. These industries have been supported by skilled engineers over the years in high-growth economic period. But, there is a problem called the "year 2007 problem" that those many skilled engineers will retire in 2007. As those workers start to retire beginning in 2007, a new approach to plant operations should be required. At the same time, interests in environmental and safety issues have been increasing, and technology for ensuring safety is strongly expected.

 

Figure 1 Batch Polymerization Reactor

Due to the recent advancements in networking technologies, integration of field information has been advancing to an extent. This trend will be a prospective solution against the year 2007 problem and for safety and reliable operation. Moreover, it will become increasingly important for users to rearrange and modify actual plant information and provide it in a form that is easy to use based on users' perspectives. Therefore, in terms of technology supporting human-based decision-making and operations, we propose real-time utilization of plant models to make it possible to understand more essential plant conditions.

Figure 2 Distillation Column Model and Internal State

Figure 3 Tracking Simulator

It seems likely that ubiquitous production, whereby the required products are produced as they are needed, will become more common to reduce distribution costs and environmental costs, and hedge the risks of large-scale accidents. When functionally-similar types of plants are distributed around the world, or small-scale plants are located close to consuming regions, they will be needed to be maintained, monitored, and run from a remote location. If there are large differences between the individual installations, then the plants will have to be dealt with as individual ones at the remote control site, and thereby we cannot make efficient use of remote and concentrated operations. Therefore, it is desired to establish a mechanism that allows numerous similar plants to be handled in the same manner by common planning, operations, monitoring, and maintenance, as well as the differences in installations to be dealt with at the local plant level. This also means it is necessary to understand essential conditions common to all plants by calculating key performance indexes utilizing plant models in real time, while also enabling operations based on common control variables.

In section 2 we will discuss our work on modeling and simulation technology, and the importance of setting initial values and identifying parameters in dynamic simulations. In section 3 we present a mechanism for the real-time utilization of plant models for solving those problems, and describe the resultant new type of plant operations. In later section we discuss our work with mini-plant experimental systems.

MODELING AND SIMULATION TECHNOLOGY

Figure 4 Prediction Simulator

Yokogawa has provided integrated solutions including plant dynamic simulations and steady-state simulations. These solutions have been primarily designed for oil refineries and petrochemical plants. We have also developed training simulators to provide training on tasks such as plant startup, shutdown, and handling unexpected conditions. These simulators are designed for new workers and operators who have had fewer opportunities to perform atypical tasks, because reliability of systems and equipments increased and opportunities of handling unsteady conditions decreased. There is strong demand for training simulators, and they have significantly achieved actual performance at petrochemical plants.

Below we discuss aspects of our work aimed at improving simulator precision.

1. Parameter identification
   In petrochemical plants, advances in process modeling for simulation and physical property calculations have made it possible to reproduce actual states of the plants with considerable precision. However, process-specific parts such as reaction characteristics must still be handled by adjusting the individual parameters based on operation data. In the batch polymerization reactor shown in Figure 1, the approximate mathematical model and parameters can be obtained in advance from physicochemical laws and data on physical properties. However, a number of parameters must be identified from operational data, as shown in Table 1¹.

   Table 1 Parameter Identification

   Parameter	Operations Data for Identification
   Reaction heat from monomer to polymer	Relation between total amount of reaction and that of heat removed by cooling per one batch process
   Reaction rate constant from monomer to polymer	Relation between calculated value of concentration and amount of heat removed by cooling
   Thermal conductivity	Relation between flow rate and amount of heat removed by cooling
   Lag time constant of cooling	Response waveform of control system

    

   

   Figure 5 Plant Operations in Networked Environment

   In the batch polymerization reactor, a monomer such as ethylene, propylene, or vinyl chloride is introduced at first, then heated together with catalysts and additives to cause a polymerization reaction. This is an exothermic reaction. Heat generation stops and the batch processes are completed when all of the monomers have reacted. The amount of reaction heat is equal to the amount of heat removed by cooling, and the amount of heat removed by cooling can be calculated as follows:
   Cooling Water [(Exit Heat Quantity) - (Entrance Heat Quantity)] x (Flow Rate)

   Figure 6 Mini-plant Experimental System 1 (Simple Cooling Unit)

   The total amount of heat removed by cooling in a single batch process can be divided by the amount of raw materials monomer to determine the reaction heat per unit quantity. In addition, the reaction speed is known to decline as the polymer increases. In our reaction rate model, the reaction amount and monomer concentration are calculated from the amount of heat removed by cooling moment by moment, and the monomer concentration and monomer conversion ratio (the percentage which reacted) are plotted in a graph. Independently of this, the thermal conductivity of a heat exchanger is known to be dependent on flow rate, so flow rate and conductivity are plotted to model the flow rate- dependence of thermal conductivity.
   Finally, we developed a first-order lag model by identifying the delay in temperature measurement from the vibrational response waveform of the temperature control system due to PID control. Because the operating conditions of a batch reactor change every second as described above, even characteristics such as the concentration-dependence of the reaction speed and the flow rate-dependence of thermal conductance can be calculated from the operation data for a single batch process.
   It should be noted that in continuous processes when there is only a small amount of information, often users will focus on a few steady states and fit parameters based on balance data. For real-time utilization, it is necessary to have a means of adjusting parameters successively because the simulator is used continuously for an even longer time period.
2. Determining initial values of states
   In order to perform a dynamic simulation, it is necessary to prepare initial values for all states within the model. In steady states, it is possible to determine initial values using a steady state simulator. However, it is difficult to determine initial values for non-steady states. We have proposed a method for determining the initial values of all state variables by minimizing an evaluation function, which is sum of squares of differential coefficients, based on measured values². The evaluation function is closer to a balanced state. For example, in the case of the distillation column model in Figure 2, it was necessary to provide initial values for state variables such as the composition and temperature at all trays in the tower. The reference² presents an example of a simpler three-cascade tank system. In the example, initial values are determined using the proposed technique, and a future prediction simulation provided predictive subsequent response which were very close to the actual response.
   Various techniques for identifying parameters and determining initial values were discussed above. In the next section, the tracking simulator and prediction simulator handle this problem in more general, and we propose a method for real-time utilization of a plant model.

TECHNOLOGIES TO BE REALIZED

Figure 3 presents an overview of a real-time operations support system based on a tracking simulator.

1. Figure 7 Example of Tracking Simulator (Simple Cooling Unit)

   Tracking simulator
   The tracking simulator we propose runs a plant model simultaneously with the actual plant on a real-time basis. Model parameters are gradually adjusted in order to reduce differences between the measured values from the actual plant and the calculated values from the tracking simulator. Note that the tracking simulator should contain a PID control system, and also incorporate all of the actual plant's field conditions and environmental conditions. If a copy of the actual plant can be realized on a computer using the tracking simulator, then it can be used in a variety of ways, such as for estimating internal states and predicting and optimizing future responses, as shown in Figure 3.
   The tracking simulator is also used as a tool for identifying unknown parameters when the simulator is installed. If the plant's physical properties change over time, then the simulator can track changes in characteristics.
2. Prediction simulator
   The prediction simulator predicts future states if current operations are continued. It can be used to confirm whether the states are approaching target values. Additionally, it can help ensure safe and secure operations by predicting accidents in advance such as uncontrollable or dangerous state when the plant is operated close to operation and control limits.
   The prediction simulator is operated by fixing the parameter and input values from the time when the trigger is set, accelerating by several tenfold or hundredfold speed. As shown in Figure 4, the tracking simulator must always operate simultaneously and in parallel with the actual plant. Therefore, for the prediction simulator, we prepare an exact copy of the tracking simulator, and the tracking simulator's internal states (parameters, variable values, etc.) are copied simultaneously with a trigger to the prediction simulator for initial states. Because the prediction simulator's initial values are copied from the tracking simulator, it is not nece