Use of Component-Based Software Architectures in Industrial Control Systems

Component-based software engineering (CBSE) denotes the disciplined practice of building software from pre-existing smaller products, generally called software components, in particular when this is done using standard or de-facto standard component models. The popularity of such models has increased greatly in the last decade, particularly in the development of desktop and server-side software. The main expected benefits of CBSE are increased productivity and timeliness of software development projects. The last decade has also seen an unprecedented interest in the topic of software architecture in the research community as well as among software practitioners. CBSE has notable implications on a system’s architecture, and an architecture that supports CBSE, e.g. by mandating the use of a component model, is called a component-based software architecture.This thesis investigates the benefits and problems related to the use of such architectures in industrial control systems, which are computer-based systems that control physical processes and equipment. The investigation is mainly performed through an industrial cases study of a global company developing a new generation of control systems, intended to replace several existing systems. To leverage its global development resources and the competency of different development centers, the company decided to adopt a component-based software architecture that allows certain functionality to be realized by independently developed components. The architecture incorporates a limited version of a standard component model.The process of redesigning the software architecture is presented in this thesis, along with the experiences made during and after the project. An analysis of these experiences shows that the component-based architecture effectively supports distributed development and that the effort required for implementing certain functionality has been substantially reduced. The use of the selected component model in real-time systems is furthermore analyzed from a more general perspective. It is shown that adopting the model means that real-time requirements can still be satisfied in most cases, but that this may require certain precautions to be taken.