Use of Web Services for next-generation automation systems

This paper outlines the shortcomings of current automation architectures as well as the technological opportunities that will allow facing the requirements of contemporary and future automation systems. In particular, it highlights the value proposition of applying the service-oriented architecture (SOA) paradigm implemented through Web Services technology, adopted at all levels of the automation hierarchy. The promises and initial results of this approach are presented, together with an outline of the standardization landscape. Issues and challenges Present-day industrial automation systems are no longer in line with market demands and requirements. • Market dynamics. Under today's global competition, production lines must be fast to set up and to reconfigure in order to promptly respond to varying market demands, to allow for mass customization and for very small lot sizes. • Lack of interoperability. Today's industrial installations are often rigid patchworks of technology islands with poor scalability, due to a lack of widely accepted standards. • Engineering practices. Today, the engineering of manufacturing systems focuses on the construction of specialpurpose machines for a single product and for a constant output flow and virtually every new piece of automation has its own unique control system. As a result, nearly every machine is built as "one-of-a-kind" and 80% of the engineering effort is devoted to re-implementing the control and related electrical systems each time a new machine is implemented. Engineering tasks cannot be run in parallel, so that application testing cannot take place before final installation at the end user site. The use of a plethora of incompatible tools for disparate technologies further reduces efficiency, both at design time and during operation. Since engineering cost accounts for 70% of the overall cost of an automation project, "automating the automation engineering work" is strongly needed. Hardware only accounts for 30% of the overall cost of an automation project, the remaining 70% being personnel cost. Consequently, the main challenge is not to reduce the device costs, but to improve the cost-effectiveness of the automation engineering work. As illustrated in Figure 1, a higher degree of modularity would allow engineering activities to take place in parallel and time-to-market to be reduced. Figure 1 – Serial vs. parallel engineering tasks • Disconnect between the shop floor and the top floor. The convergence of enterprise resource planning (ERP) systems and manufacturing execution systems (MESs) is hampered by the fact they come from different horizons and are separated by different underlying technologies. Therefore, the exchange of data between the production control level (shop floor) and the enterprise application level (top floor) is often done either manually or semi-automatically, e.g., by exporting a production plan to a spreadsheet file that is then fed into the MES. This integration gap between the enterprise level and the plant level results in untimely and errorprone data exchanges, which may have very severe cost implications. • Loss of process know-how. The current personnel aging phenomenon is leading to a substantial shortfall of skilled labor among end users. This spurs a growth in supplier-provided services over the next several years, as end users will increasingly tend to outsource their automation, demanding an increasing level of vertical industry expertise from their suppliers. An open, flexible and agile environment with plug-and-play connectivity is desperately needed, to enable rapid reconfiguration and re-use of higher-level modules – integrating a variety of heterogeneous devices into an interoperable network of resources – to account for the introduction of new product models, as well as the ability to dynamically provide new value-added services and efficient diagnosis and maintenance. Trends and opportunities Attempts to address the need for more configurable production systems better able to meet the requirements of agile manufacturing have led to a growing interest in automation paradigms that model and implement production systems as sets of collaborating production units. Major trends, enabled by technological advances and evolutions, support this movement. • Smarter, networked devices. State-of-the-art electronics provide unprecedented computing horsepower in very tiny components. Devices can thus be made "smart" and self-reliant, moving intelligence away from central controllers. If network-connected and using universally adopted, vendor-neutral networking standards, such devices can become members of the "Internet of things" or "Web of objects" that is gradually pervading the world. Furthermore, usage of a uniform communication paradigm will greatly contribute to the creation of an open, flexible and agile environment, and hence facilitate the elimination of the existing technology islands. • Higher-level modularity. On this basis, it becomes feasible to build automation systems out of higher-level, network-connected modules, each consisting of mechanical, electrical and control elements. Such a module is a controllable, reusable and reconfigurable part of a system; it encapsulates process-oriented functionality, exposed through well-defined interfaces of the highest possible level of abstraction, hiding its implementation to the outside world. A module may be used in its own right or be combined with other modules to form composite modules, according to the "Russian dolls" composition paradigm. As documented in [1], the notion of module covers different automation concepts: a module can be a standard machine, a cell, a workstation or a unit. The latter concept of "unit" or "mechatronic module", as the lowest level of autonomous unit of a production system – each with its specific process-oriented functionality, such as positioning, drilling, transportation, etc. – is the enabler of new, distributed control system architectures. Modules can be pre-tested as stand-alone components and such tested and proven components can be re-used across applications and systems. A similar line of thinking can be found in [2], even if it mostly addresses logical units and relies on proprietary technology. • Reconfigurability. Only systems capable of being rapidly reconfigured with very short design-developmentramp-up-produce life cycle phases can stand the pressures of a continuously changing market. Reconfigurability goes hand in hand with dynamicity, the capability to automatically add or remove devices, machines, functions, sub-systems, etc. at any time. Reconfigurability is strongly favored by plug-and-play connectivity and by the use of tested and proven components. • Maintainability. One of the first requirements expressed by virtually all industrial customers is "improve the maintenance". The use of plug-and-play smart devices that encapsulate their own complexity has the potential to substantially improve the effectiveness of device and system maintenance. To a large extent, smart devices shall be able to diagnose themselves, thus obviating the need for exposing a multitude of disparate information items, as is the current practice. Also, structuring devices in a modular fashion helps zooming in on diagnostics in a top-down manner. Many connectivity issues can be overcome by using open interoperable protocols. Leveraging SOA based on Web Services in the device space In just a few years, the concept of SOA has gained substantial traction in business system environments. In the context of automation systems, the characteristics of SOA closely match the technical and business level requirements outlined above. In a nutshell, SOA is an architectural paradigm for building systems from autonomous yet interoperable components. A service provides business-meaningful, self-contained functionality that is not tied to particular usage scenarios. A service only exposes its interface ("contract"), which fully encapsulates the complexity of its implementation. Services can be published and discovered dynamically. SOA is characterized by coarse-grained service interfaces, loose coupling between service providers and service consumers, and message-based, asynchronous communication. Leveraging the SOA paradigm allows for services to be re-used across processes and systems, and systems to be "built for change". Reliability is improved as applications and systems can be made up of tested and proven components. SOA offers the potential to provide the necessary system-wide visibility and interoperability in complex systems subject to frequent changes and operating in a multi-vendor environment. The net result is a substantial cost benefit together with increased agility, both from a technical and a business point of view. The use of open, vendor-neutral standards, in particular those of the Web Services family, allows for implementing SOA in a platform-agnostic fashion, a key asset for gaining widespread adoption. In order to apply the SOA paradigm in the device space using Web Services technology, the Devices Profile for Web Services (DPWS) was defined [3]. In addition to the core Web Services standards, such as SOAP, WSAddressing, WSDL and XML Schema, DPWS includes WS-Discovery for plug-and-play device discovery and WSEventing for publish-subscribe asynchronous event notification. DPWS was designed in 2004 by an industry consortium led by Microsoft and is natively integrated in Microsoft's Windows Vista platform, thus enabling intelligent devices, including PCs, to communicate across a network using Web Service protocols. In the context of the ITEA/SIRENA project [4], Schneider Electric delivered the first implementation of DPWS for simple, low-cost, embedded devices and demonstrated that DPWS connectivity could be implemented using a 4€ processor chip. The SIRENA project was granted the ITEA Achievement 2006 award for this pioneering effort and the succ