面向可重构系统的局部永久故障恢复方法研究

发布时间：2018-04-09 13:20

本文选题：FPGA　切入点：可重构系统　出处：《哈尔滨工业大学》2017年硕士论文

【摘要】：可重构计算的兴起,为处理系统的功能灵活性和计算性能的平衡提供了良好的解决方案,可编程器件FPGA的普及和广泛应用,更是极大地推动了这一技术的发展。基于FPGA的可重构系统被广泛应用到空间计划,深海探测等项目中,但是由于恶劣环境的影响,引发FPGA产生各类故障,降低了系统的可靠性,特别是难以恢复的局部永久故障,更是对系统性能以及芯片寿命产生了严重影响。为了提高可重构系统的可靠性和可用性,本文提出一种面向可重构系统的局部永久故障恢复方法,在本文提出的方法中,使用FPGA构建可重构系统,利用FPGA的动态部分重构功能,结合冗余恢复技术的快速性和演化硬件恢复技术的灵活性,解决可重构系统的局部永久故障恢复问题。首先,研究基于模块多样化算法的冗余恢复技术。使用模块多样化算法确定冗余配置矩阵,为每一个配置矩阵生成相应的配置信息,形成初始冗余配置信息集。模块多样化算法满足完整性条件,能保证在发生单点永久故障的情况下,快速修复故障。同时模块多样化算法满足多样化条件,可以最大限度的减少每个模块的故障压力。其次,研究基于自适应遗传算法的演化硬件恢复技术。自适应遗传算法结合模拟退火算法的优势,缓解简单遗传算法的早期早熟现象和后期停滞现象;使用自适应适应度评估方法,提高算法的性能。在使用冗余恢复技术无法实现故障恢复的情况下,使用演化硬件恢复技术翻新冗余配置信息,解决更为严重的多点永久故障修复问题。为验证提出的方法的性能,基于Xilinx公司的Zynq-7000系列Zedboard开发板,充分利用Zynq系列可编程逻辑和处理系统协同的架构,使用集成软件开发环境ISE14.7,实现了提出的面向可重构系统的局部永久故障恢复方法。通过故障注入的方式模拟单点和多点永久故障,进行局部永久故障恢复实验。实验结果表明,本文提出的方法能100%恢复可重构系统中出现的单点永久故障,即使在苛刻的故障假设情况下,仍然能以较高的概率实现可重构系统的多点永久故障的自主恢复。
[Abstract]:The rise of reconfigurable computing provides a good solution for the balance between the functional flexibility and computational performance of the processing system. The popularity and wide application of the programmable device FPGA has greatly promoted the development of this technology.The reconfigurable system based on FPGA is widely used in space planning, deep-sea exploration and other projects. However, due to the bad environment, it causes various kinds of faults in FPGA, and reduces the reliability of the system, especially the local permanent fault, which is difficult to recover.It also has a serious impact on system performance and chip life.In order to improve the reliability and availability of reconfigurable systems, a local permanent fault recovery method for reconfigurable systems is proposed in this paper. In the method proposed in this paper, the reconfigurable system is constructed with FPGA, and the dynamic partial reconfiguration function of FPGA is utilized.The problem of local permanent fault recovery of reconfigurable systems is solved by combining the rapidity of redundant restoration technology and the flexibility of evolutionary hardware recovery technology.First of all, the redundancy recovery technology based on modular diversity algorithm is studied.The modular diversity algorithm is used to determine the redundant configuration matrix, and the corresponding configuration information is generated for each configuration matrix, and the initial redundant configuration information set is formed.The modular diversity algorithm satisfies the integrity condition and can quickly repair the fault in the case of a single point of permanent fault.At the same time, the algorithm of module diversification meets the requirements of diversification, which can minimize the failure pressure of each module.Secondly, the evolutionary hardware recovery technology based on adaptive genetic algorithm is studied.The adaptive genetic algorithm (AGA) combines the advantages of simulated annealing algorithm (SA) to alleviate the early precocious phenomenon and anaphase stagnation phenomenon of simple genetic algorithm, and the adaptive fitness evaluation method is used to improve the performance of the algorithm.In the case of failure recovery can not be realized by using redundant recovery technology, the redundant configuration information is renovated by using evolutionary hardware recovery technology to solve the more serious problem of multi-point permanent fault recovery.In order to verify the performance of the proposed method, based on the Zynq-7000 series Zedboard development board of Xilinx Company, it makes full use of the Zynq series programmable logic and the cooperative architecture of processing system.An integrated software development environment ISE 14.7is used to implement a local permanent fault recovery method for reconfigurable systems.The local permanent fault recovery experiment is carried out by simulating single point and multi point permanent fault by fault injection.The experimental results show that the proposed method can recover 100% of the single point permanent faults in the reconfigurable system. Even under the harsh fault assumption, the autonomous recovery of the multi-point permanent fault of the reconfigurable system can be achieved with a high probability.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN791;TP18

【参考文献】