A Reusable Framework for Rolling Mills

Esprit Project 22897
Software Technologies
to home page

Innovation in Process Automation

Next Generation Process Automation System for Rolling Mills
Press release from December 1998

In December 1996 the EU Commission signed a project contract with the REFORM consortium in the frame of the Information Technology Program Esprit. REFORM (Reusable Framework for Rolling Mills) aims to develop and implement a reusable object-oriented software architecture (framework) for process automation of rolling mills.

REFORM involves a total cost of approximately 3 Million ECU over 3 years. The project consortium is built of the enterprises Siemens (Germany, coordinator), Mandator (Sweden), UNI SOFTWARE PLUS (Austria), Voest-Alpine Stahl(Austria) and the Universities of Linz and Hamburg. The partners were selected in order to maximize the outcome by contributing the complementary skills in domain knowledge, software engineering, modern software architectures including design patterns, software development process modeling, machine learning and constraint logic programming. That the REFORM project has a good economic potential can be clearly demonstrated by the fact that currently process automation systems are rewritten from scratch for every new rolling mill, in spite of the possibility that 70% of the code of such systems could be reused. Analyzing the cost/benefit ratio with regard to the market volume and forecasted market shares, a return on investment period of less than 3 years is expected. Nevertheless the technical feasibility was probable but not proven at the project start. Even taking into account that REFORM has its base in highly innovative technologies in its domain, it needed to be demonstrated that a system of this complexity can be implemented in the proposed generic way represented by an object-oriented framework. To minimize this risk, the framework is structured into subframeworks, the development process is strictly prototyping-oriented, prototypes are fully tested in a simulation environment semiannual reviews with independent experts are conducted by the project coordinator of the European Commission and the REFORM results will be verified and validated at the trial site of Voest-Alpine Stahl under real production conditions. This outlines, that REFORM has the typical characteristic of a research project but is demand-oriented with a clear exploitation strategy. Consequently funding of more than 50% of the total costs was accepted by the European Commission.

The first two years of project work have confirmed the technical feasibility of the undertaking. Moreover, process automation can be seen as object-oriented in nature. The project is in the stage of consolidation, tuning and refinement. The remaining year will involve the assimilation of the current results into the trial automation system at Voest-Alpine Stahl in Linz in order to demonstrate that the REFORM results not only speed up software development and time to market but also improve its quality and the quality of the finished products of the rolling mills.

The consortium is confident that its commercial partner Siemens can exploit the REFORM results immediately after the project end in November 1999 and use the framework as a basis for its next generation of process automation systems.

From a course-grained perspective, one hot rolling mill looks like any other; a pushing furnace to heat the slabs, heavy edge, rolling stands, and a cooling section transforming the slabs into metal strips. From a finer grain, namely the interface to the basic automation system, two hot rolling mills can look like completely different plants. Signals, measured values and control information come in certain frequencies and need to be processed on a higher level in order to optimize the process towards product quality, energy consumption and scheduled time frames. The difficulty of the traditional approach with procedural languages like C in process automation for rolling mills lies in the transformation of cooperating tasks of components, aggregates and measurement devices in a certain plant configuration into a procedural paradigm. It is therefore intelligible that developers reuse knowledge, but in many cases instead of analyzing and adapting existing procedural code rewrite all functions completely. Adaptations need offline "compile, link and go cycles" before a new functionality can be installed at the user site.

Paradigm Shift

A hot rolling mill is a special configuration of collaborating units (aggregates, controlled devices and systems, measurement units etc.) managing the process by passing messages to each other and continuously observing the material to be produced. The most natural decomposition is not functional, procedural or data-driven but object-oriented. For example, it is rather complex to have a central material tracker storing the position, velocity and acceleration of each strip in the plant. It helps to master the complexity if the material itself knows its position and necessary velocity and acceleration and communicates these to the stands, roller tables and other plant components;in implementation terms, they act as objects.

The most difficult development task in this approach is to find the right key abstractions in order to make the implementation of the components as general and customer-independent as possible. The REFORM architecture uses components based on object-oriented frameworks. As semi-finished software systems, frameworks implement the customer-independent parts of the problem domain. Nevertheless, these parts are open for customer-specific extensions. As prebuilt functional parts of the process automation system, components allow users to configure their software even at run time.

Software Principles

The use of modern techniques of component-oriented system architectures and object-oriented software allows reuse on a high level. Instead of individual procedures, the whole logic of systems and subsystems can be reused. Components like model calculations for rolling forces, temperatures, etc. can easily be plugged into the kernel system containing coordination and administration tasks. The framework divided into subframeworks such as measured value processing, material tracking, mill pacing and machine learning is the basis for the implementation of these components. The communication is handled via CORBA or via abstract telegram handlers. Because of the object-oriented structure, the abstract interfaces, and the customer-independent value handling, for example, the component triggering a model calculation does not need knowledge of the internal structure of the model, and even traditionally implemented models can be wrapped in an object-oriented manner. Tightly connected with value handling is the mill dictionary. This dictionary contains standardized names for measured and calculated values. Additionally, to exploit best practice cases from past productions, multi-strategy machine learning techniques -- including but not limited to fuzzy logic, neural networks and genetic algorithms -- support model calculations in an adaptive way or characterize significant process properties related to the quality of the finished products. Machine learning functionality is also provided in an object-oriented framework. In order to guarantee optimized schedules, constraint logic programming techniques are applied. The REFORM project is committed to de facto standards like Windows NT, C++ and CORBA. High-level development tools like hypertext tools for source code structuring and tracing tools for framework understanding have also been developed in the project. These tools are used in addition to commercially available development environments like SNiFF+.

Organization

Because of the distributed locations of the REFORM partners, an Internet-based virtual project office was established. To exploit the complementary know-how of the different groups, temporary task forces relating to such topics as integration, tools and models have been established; these task forces usually cooperate on site. In critical phases of the project, developers from all partners work together either in Erlangen or in Linz. Every three months the consortium meets for several days in order to discuss the progress of the project and plan corrective action as necessary.

The consortium has agreed upon a common software development environment, technical guidelines and coding conventions, documentation, and configuration management. Common office tools and shared document templates facilitate cooperation. Additionally, development is guided by organizational provisions like a development process model.

Nevertheless, the partners' need to reach a maximum of cohesiveness in such a collaborative project was underestimated at the beginning.

Consortium

The foreseen roles of the partners and the real responsibilities have progressively changed and improved. The prime contractor Siemens, represented by the research department and the automation division, together with the software engineers and architects from the University of Linz developed the kernel of framework and are responsible for the integration of all pieces. Mandator from Sweden developed the mill pacing framework and wrote the simulation environment to test process automation systems. UNI SOFTWARE PLUS together with the Fuzzy Logic Lab Linz created the multistrategy machine learning framework and in parallel analyzed real data from VASL production. Voest-Alpine Stahl's process automation group planned and implemented the demonstrator and additionally rearranged their basic automation system in order to test the flexibility of the framework under real conditions. The University of Hamburg, originally responsible for guidelines, will also coordinate the whole packaging (documentation, advertisement, etc.).

A consortium for such a complex project needs to reach a degree of cooperation where the reallocation of tasks and efforts can be managed across partner borders, in agreement with the whole consortium.