Research On Equation-based Modeling And Simulation Of Cyber Physical Systems

Cyber-Physical Systems (CPS) is an integration of computation, communication, control with physical systems and physical processes. Embedded computers and networks monitor and control the physical processes, usually with feedback loops where physical processes affect computations and vice versa. Because of complicated hierarchy of CPS, and various disciplines CPS involves, CPS research is faced with several technical challenges, such as basic disagreements of information science and engineering science, lack of cooperative design and analysis, and complicated system verification etc. Modeling and simulation (M&S) of CPS is one of the foundational technologies. Modelica is an equation-based object-oriented modeling language, with semanteme for non-causal modeling, hybrid modeling and multi-domain physical systems modeling, which means Modelica has a great potential in the M&S of CPS. The key issues of M&S of CPS based on Modelica are studied in this dissertation.Due to the large-scale equation systems from CPS model, time consumption of numerical simulation is expensive. Symbolic reduction and decomposition technology is performed to reduce coupling of system, which means cost of computation is also reduced. Classical reduction and decomposition technology will meet with state space expansion problem, when performing on multi-dimensional physical system model of CPS. A formalization description method with type-based weighted bipartite graph (WBG) is proposed to represent structure constraint of equation system of hybrid vector space physical system model (HVSPSM). Basical algorithms for WBG are discussed and a new reduction and decomposition strategy is designed for HVSPSM. State space expansion problem is avoided by the new strategy, which decomposes equation systems of HVSPSM to serialized small-scale sub-systems directly and improves the simulation efficiency.Time definition and synchronization are two fundamental problems of M&S of CPS. Continuous time and discrete time model described by Modelica is ambiguous, and overabundant state event detection and iteration results in inefficient performance of simulation. By expanding semantics of Modelica with clock mechanism, interaction between clock-based model and time-based model is clealy defined. Thus unified modeling technology of CPS is proposed. Clock-infer rules are designed to determine boundary between infor part and physical part of CPS model. Reduction and decomposition method is discussed, to determine constrains for synchronization. Finally, unified simulation strategy for CPS model is designed, based on synchronous data-flow principle and global simulation clock. Clock event and state event are processed separately with each other, which promotes simulation efficiency for CPS.CPS model is usually large scale and of high resolution, to obtain more accurate dynamic system behaivior. Taking full advantage of computing resource is an actual requirement to improve efficiency of large system simulation. Howerver, it is a heavy and complicated work to build a CPS simulation model for parallel computing. An automated parallelization method is proposed. Considering non-causal feature of equation-based modeling language, data dependency of equation systems is analysed by reduction and decomposition method, and task graph with equation-granularity is built. Based on task graph, scheduling algorithm is designed. Also, generateing parallel simulator is discussed. Thus, equation-based CPS model can be parallelized automatically by this method, and CPS designer doesn’t have to manually build a parallel simulating model to run on parallel computer.