【作者】 王永翔；

【导师】 王立德；

【作者基本信息】 北京交通大学 ， 电力电子与电力传动， 2009， 博士

【摘要】 列车通信网络技术是现代列车关键技术之一,本文在铁道部重点课题“TCN底层协议及关键技术研究”的基础上展开并进行了延续和扩展。列车通信网络的应用是保证列车控制有效性、安全性及旅客舒适性的必要措施。作为列车控制系统的重要组成部分,要求列车通信网络具有较高的实时性、可靠性及安全性。我国在列车通信网络方面的研究起步较晚,虽然取得了一定的进步,但较一些发达国家还有一定差距。本文着眼于TCN的底层协议,围绕列车通信网络在性能评估、优化及具体实现等关键问题展开研究,所完成的主要工作及获得的创新性成果如下:1、针对列车通信网络动态性能分析的需求,研究了现场总线建模的方法,指出Petri网是适用于对现场总线这种分布、离散、并发、随机系统进行建模的方法。广义随机Petri网(GSPN)、确定与随机Petri网(DSPN)、随机着色Petri网(SCPN)是针对网络底层协议的不同抽象层次进行建模的有效工具。分析了Petri网特性与被建模网络协议特性之间的关系以及采用随机Petri网的稳态概率进行网络协议性能分析的方法。随着计算机软硬件水平的发展和Petri网复杂性的不断提高,软件模拟与仿真方法将是基于Petri网分析的主要方法。针对MVB总线性能分析的不同需求,建立了MVB总线的GSPN模型、事件仲裁模型和MVB总线的着色Petri网模型,采用Petri网计算机辅助分析软件TimeNET和CPN-Tools进行仿真分析,给出了仿真分析结果。MVB总线的GSPN模型用于分析其单纯过程数据的通过性能,事件仲裁模型分析了偶发相时间宽度、设备地址相似度及偶发相均匀度对于事件仲裁性能的影响,从而提出了优化周期扫描表的问题。在对比分析了随机着色Petri网建模与单纯离散事件仿真方法不足的基础上,提出了基于OMNET++与SCPN相结合的TCN仿真分析方法,利用该方法建立计算机仿真分析平台,有效利用了二者的长处,更加适合于TCN的仿真分析与验证。2、为了达到在保证车载控制网络中过程数据可靠调度的前提下尽量提高消息数据通信性能的目的,提出了MVB总线周期扫描表的优化设计方法。首先提出了衡量周期扫描表相关基本指标的概念,讨论了按照IEC61375-1标准所构造的周期扫描表的缺陷,之后针对是否报文定时相关提出了两种优化设计算法。对于报文定时无关的可以采用逐步填空法完成,而对于报文定时相关的周期扫描表(TPPT)构成问题,则采用混合遗传算法进行求解。建立了TPPT问题的数学模型,给出TPPT问题的目标函数,及编码、选择、交叉和变异的方法。结果表明这两种算法可以较好的优化MVB周期扫描表的构造。为了克服混合遗传算法易于陷入局部最优的问题,进一步提高优化速度,提出了基于S-Tent映射的混沌混合遗传算法。讨论表明S-Tent映射具有均匀的遍历特性,给出了TPPT问题的染色体尺度变换方法。从最终的性能评估可以看出,基于S-Tent映射的混沌混合遗传算法是对TPPT问题实现最优配置的有力手段。3、在自主设计的总线访问专用IP核的基础上,提出了基于SOPC技术的总线控制器和应用处理器的融合方法。MVB总线控制器是实现MVB总线网络功能的关键器件,负责访问MVB总线,并提供与微处理器的通信接口,实现数据传输。在应用MVB总线控制器时,通常需要与外置处理器共同完成总线通信任务。本文给出了MVB总线控制器的接口设计方案,完成用于总线访问的专用IP核,采用SOPC技术将应用处理器同MVB总线控制器集成在同一块FPGA芯片之中,从而提出了一种具备更高集成度的MVB总线接口实现方案。4、建立了满足TCN网络实验及测试需求的TCN网络半实物仿真平台。半实物仿真平台可以提供网络设备的验证平台,同时可以为TCN网络的性能测试提供实验环境。在该平台搭建过程中重点研究了典型电力机车的信号控制电路,将一些司机室的实物设备通过TCN网络与PC机连接。PC机模拟其它大型的复杂设备,从而构成半实物仿真平台。在该平台上进行的相关实验,可以验证本文提出的网络模型、周期扫描表的优化方法及基于SOPC技术的TCN接口实现方法等的正确性。最后,在总结全文的基础上,给出了论文研究过程中得出的思考和结论,提出了一些需要进一步深入研究的问题。更多还原

【Abstract】 Train Communication Network is the core technology in the modern train. This paper is based on the major research project - "Research of TCN low-layer protocol and key technology" which is carried out by the Ministry of Railways. The application of the train communication network has been the necessary steps to provide a more safe and faster railway experience. As a main part of the train control system, the train communication network has to have high real-time and high reliability and high safety performance. We have been late on the research of the TCN, and we still lag far behind the foreign vendors. This paper mainly studies the key problems related with the performance evaluation, optimization and the interface implementation method. It can be divided into four main parts:1. Aiming at the dynamic performance evaluation of the train communication network, the paper presents the methods to model the field bus, and it points out that the Petri net is the best way to model the distributed, discrete, concurrent, stochastic systems. General Stochastic Petri Network (GSPN), Determentic and Stochastic Petri Network (DSPN) and Stochastic Colored Petri Network (SCPN) can be used in the different level in modeling the field buses. The method to analyze the performance of field bus is talked. With the development of the hardware and software, the method based on the simulation will be the main method to solve the Petri net. To analyze the performance of the MVB, the GSPN model of MVB, and SCPN model of the MVB events arbitration, and the CPN model of the MVB are presented in the paper, and the results are also given. The GSPN model is used to analyze the performance of pure process data throughput. The model of events arbitration can be used to analyze the relationship between the event arbitration performance and the time width of stochastic phase, the similarity of the device address and the uniformity of stochastic phase. From the analysis, the optimization problem of the period polling table is proposed. After comparing the Petri nets modeling method and the discrete event simulation method, the TCN simulation method combined with SCPN and OMNET++ is proposed. This method is more suitable to analyze the real TCN.2. To maximize the message transmitting performance on the premise of reliability of process data scheduling, an optimization method of the MVB period polling table is presented. Based on the basic definition of the period polling table, the shortage of the recommended method in IEC61375-1 standard is talked. Two methods have been proposed to solve the two different problems. The first named filling step by step method can be used to construct the period polling table that was time independent. And the time dependent problem is solved by the hybrid genetic algorithm. The model of the TPPT problem has been presented, and the target function, and other related problem have also been given out. The computing result presented in the paper proved that these methods can optimize the MVB period polling table efficiently. To overcome the shortage of the genetic algorithm and increase the optimization speed, the hybrid chaos and genetic algorithm based on the S-Tent map is proposed. The S-Tent map has better ergodicity, and the scaling method of the chromosome is also given. The result showed that this method is an effective method to optimization the period polling table.3. The fusion method of the mvb bus controller and application processor using SOPC technology is presented, based on the independently designed bus access IP core. The MVB bus controller is the main parts to implement the bus access function. The controller worked with an application processor commonly. This paper has given the interface implementation method of the MVB bus controller, the application processer and the controller is integrated into one FPGA chip using the SOPC technology.4. The TCN hardware-in-the-loop simulation platform is created to complete the experiment and test of the TCN. On this platform the interface implementation method proposed in this paper can be verified, and the performance measurement can be done. This platform includes the equipment on the driver’s desk and the other large equipments are simulated by the PC software. The experiments can verify the correctness of the model, the optimization method and the implementation method proposed in this paper.In the last part, the paper presents the research conclusions, proposes the innovative ideas, and points out the remaining problems that need further research.更多还原