Reaearch on Modeling and Optimization Strategy in Industry Multi-service Real-time Networks
|Course||Control Science and Engineering|
|Keywords||EPA DRP multi-service network model stability FRT(FastReal-Time)system scheduling algorithm performance analysis|
With the increasing integration of informatization and automation in industries like petroleum, chemical, energy, metallurgy, pharmacy, transportation, shipbuilding and manufacturing, we are now building systems with larger scale and higher complexity wherein more service data needs to be transmitted. Among those services there may exist traditional low-speed process control services, high-speed motion control services or even fault detection and recovery services. The challenges to real-time, stability and reliability of communication networks are becoming more and more severe. In traditional industry network systems, different services are usually transmitted in separated networks, and the research on them are also independent. These independent networks lead to information islands that have become a major obstacle to information integration. The real-time requirements of the network are also hard to be satisfied. As a result, it is important to integrate all these services into a single network to form a multi-service real-time network that ensures mutual access, interoperability and information fusion.However, most of the existing researches are restricted to single-service industrial Ethernet. There are few works on multi-service industrial networks and no substantial application solution is presented either. A research on this topic is thereotically important and practically urgent.After a thorough review of researches on RTE (Real-Time Ethernet) modeling, scheduling and optimization for different industrial services, this paper defines a new conception of industrial multi-service real-time network for the first time. The services are divided into three categories named RT (Real-Time), FRT(Fast Real-Time) and DR (Distributed Redundancy) services. Based on the real-time protocol EPA (Ethernet for Plant Automation) proposed by China, this paper studies the classification and modeling of services, the communication model of RTE, the scheduling and optimization of FRT services, the stability of control loops in RTE, the modeling of DR services with an analysis of recovery times, and the structure of the multi-service fusion network with a scheduling scheme. Specific results can be summarized as follows:1. The categories of RT, FRT and DR services are established based on EPA. The scheduling mechanism of EPA networks is analyzed. As a first attempt to quantitatively model EPA, the data, scheduling and timing models are built based on a comprehensive analysis of the distributed EPA protocol. The scheduling algorithm and criterion of schedulability of EPA networks are given.2. Some shortcomings inherent in legacy EPA protocol are analyzed and modifications are made to improve the real-time performance of EPA networks in FRT scenarios. Optimization schemes for communication process, synchronization, network segmentation and static time window allocation are proposed. A multi-segmented matrix network structure with a set of scheduling rules is presented.3. A new model of Markovian jump discrete-time networked control systems (NCS) is proposed based on the analysis of the influences that the packet transmission delay and the precision of clock synchronization exert on EPA control systems. The system stability conditions that must be satisfied by the transmission delay and the precision of clock synchronization are studied. An NCS based on EPA is modeled. An algorithm is proposed to obtain the shortest communication macrocycle for an EPA system to be stable.4. The original DRP (Distributed Redundancy Protocol) is modified and a new DR service is defined to accommodate the RT service which is not allowed for in existing researches. The recovery time and service occupancy rate models are given based on the network calculus. The performance of DRP is studied while the RT service is considered. The optimization scheme of DRP is proposed.5. A network structure of the multi-service fusion network is proposed. A service fusion scheme and a dynamic macrocycle scheduling algorithm are given. The models of data, flow and effective bandwidth utilization are developed for the multi-service network. A multi-level network structure is proposed according to different real-time requirements of the services. A data aggregation policy, a scheduling scheme and a conception of optimal tree are proposed to solve the problems of network structure optimization and timeslot allocation in the multi-level network.