Provisions for Adaptability in Distributed Intelligent Cyber Physical Systems

Abstract

Increasing system complexity and physical dispersion of automation systems mandates a paradigm shift away from centralized monolithic control systems to highly distributed, granular control. The Internet of Things concept envisions extreme levels of distribution granularity; where every physical object is embedded with some computational power. Rapid technological progress has resulted in constant system dynamism and a need for developers to design applications with change in mind. Components are constantly added and removed, failures are expected and computational power between similar applications can vary greatly. Distributing control applications brings immediate benefits such as mitigation of total system failures and localized control for highly dispersed systems. The design of automation control systems is a co-ordination between functional and nonfunctional requirements. Trade-offs between the two often result in centralized control for performance oriented systems and distributed control for more dynamic, flexible systems. Software for these dynamic systems cannot continue to adhere to traditional development paradigms. In order for a distributed system to achieve identical behavior to a centralized system, a number of challenges must be overcome. Application logic must be separated into distributed communicating modules with a limited local view of data. Effectively deploying software to a distributed network is another challenge. Although a distributed software application can be separated onto a number of devices, the optimal device allocation can constantly vary during runtime. In addition to these, maintaining operational timing performance is often crucial. When control operations execute on a single device, there is only a single source of clocks and synchronization is implicit. On separate networked devices, timing and synchronization become a problem especially when devices have different performance characteristics and new computational power is added and removed dynamically. This thesis addresses these challenges by utilizing the IEC 61499 standard to improve the design, operation and maintenance stages of a distributed automation application lifecycle. Adaptability is focused on as a key factor for improving the application lifecycle as this often directly affects the amount of developer effort. The provisions to achieve adaptability are through commoditization of a number of key design elements of distributed industrial automation systems: the mechatronic component, the computational devices and the software. In this thesis, the mechatronic component is commoditized by developing a systematic framework for embedding intelligence at the component level. Control, simulation and visualization are all integrated using the Intelligent Mechatronic Component (IMC) concept; following a cyber-physical design paradigm where software structure is closely tied and coordinated to hardware structure. The digital ecosystems concept was then introduced to commoditize the PLC and its software. Rather than treating the control system as a static configuration it is assumed to be dynamic, with potential failures and addition of additional devices during runtime. Using IEC 61499 reconfiguration mechanisms, a methodology to allow software migration between devices further extends the digital ecosystems vision of dynamic hardware and software. To further support the digital ecosystems concept, precise software timing characteristics as well as hardware invariance to timing requirements were desired. The Time-Complement Event-Driven architecture was proposed in a joint work in order to achieve these criteria and behaviors. This architecture was then unified with the digital ecosystems software design paradigms. Reconfiguration difficulties such as control logic complexity and the absence of a time notion were explored and addressed with the Time-Complemented Event-Driven Reconfiguration architecture. This architecture was demonstrated to provide significantly more reliable reconfiguration simplicity compared to traditional control logic. A number of implementations have been developed and it has been demonstrated that IEC 61499 is a feasible implementation layer for achieving adaptability in the lifecycle of distributed automation applications. Baggage handling system and sorting station running examples have been used throughout this thesis to demonstrate feasibility of each newly introduced architecture or design artefact. It has been shown that the ideas presented in this thesis add value to the distributed application lifecycle in a number of stages, including design, operation and maintenance.