【作者】 康波；

【导师】 陈光（礻禹）； 吕炳朝；

【作者基本信息】 电子科技大学 ， 测试计量技术及仪器， 2003， 博士

【摘要】 混沌控制与混沌优化是当前混沌学研究领域的一个重要课题。本论文在对混沌控制与混沌优化的相关理论和方法进行较为全面、系统的研究基础上，提出了一些新的混沌控制与混沌优化方法，并针对混沌优化与混沌神经网络在组合电路测试生成问题中的应用展开了详细研究，提出了基于混沌优化的组合电路测试生成方法，得到了有价值的成果。本文的主要内容与创新贡献包括以下几个方面：1．基于一类混沌系统的一维返回映射图，研究了混沌系统的一类外部周期性脉动控制方法，包括常量脉动控制方法与反馈脉动控制方法，并针对非线性阻尼摆中混沌行为的控制问题进行了详细研究，仿真实验结果验证了脉动控制方法的有效性与好的鲁棒性。将神经网络与内模控制相结合，提出了一种用基于BP神经网络的内模控制方案进行混沌控制的方法，并重点针对Duffing振荡器的混沌控制问题进行了详细设计与实验研究。仿真实验表明，基于神经网络的混沌控制方法不仅能实现对混沌行为的有效控制，而且还具有强的鲁棒性与自适应能力。2．将混沌搜索与Hopfield网络的梯度算法结合，提出了一种基于混沌搜索的梯度算法，并基于组合电路测试生成的Hopfield神经网络模型，用上述算法来求网络能量函数的最优解，提出了一种基于混沌搜索的测试生成算法，仿真实验表明，该算法能有效克服梯度算法易于陷入局部极小的不足，并具有较快的优化速度，提高了测试生成的效率。3．通过在组合电路的Hopfield神经网络模型中引入自反馈项，从而构成了组合电路的混沌神经网络模型，提出了基于混沌神经网络的测试生成方法，并针对混沌神经网络自反馈项的控制策略，分别提出了具有固定退化规律的一般算法和随搜索进程适当调整的改进算法，仿真实验表明，一般算法具有很高的测试生成速度与测试生成概率，而改进算法能够将测试生成概率提高到100％，明显优于Hopfield神经网络。4．将混沌动态与遗传算法结合，利用混沌序列来控制遗传算法中交叉和变异操作，即通过引入混沌交叉与混沌变异的概念，提出了一种改进型遗传算法——混沌遗传算法。基于组合电路测试生成的Hopfield神经网络模型，并针对<WP=6>测试生成问题的具体特点，将混沌遗传算法用于求解网络能量函数的最优值，提出了一种基于混沌遗传算法的组合电路的测试生成方法，实验结果表明所提出方法具有良好的全局搜索能力，加快了测试生成过程，提高了测试生成效率，具有重要的实用价值。5．针对故障测试集最小化问题的具体特点，引入一种特殊的混沌搜索算法与混沌遗传算法，分别提出了基于混沌搜索与基于混沌遗传算法等两种测试集最小化方法，实验表明，前一种方法极大地提高了最小完备测试集生成速度，能很好地解决中等规模测试集最小化问题，而后一种方法具有更快优化速度，适合于大规模复杂故障测试集的最小化问题。更多还原

【Abstract】 Chaos control and chaotic optimization is one of important fields in scientific chaos research. In this dissertation, firstly, the theories and methods about chaos control and chaotic optimization are studied. Based above, some new approaches for chaos control and chaotic optimization are proposed. Furthermore, the application of chaotic optimization for test generation of combinational logic circuits is investigated in detail, and new methods for test generation are proposed based on chaotic optimization. Several valuable and important results are achieved.The main contribution and valuable results of this dissertation can be listed as following:1. Based on one type of one-dimensional return maps of chaotic systems, two impulsive control methods for chaos suppression, namely, constant impulsive control and feedback impulsive control method, are studied. With these methods, controlling chaos in nonlinear damping pendulum is discussed in detail. Simulation results demonstrate the effectiveness and robustness to chaos suppression. With combination of neural networks and internal model control, an internal model control scheme based on BP neural networks is proposed for controlling chaos. The simulation results based on a typical chaotic system, namely, Duffing oscillator, illustrate the effectiveness and strong robustness and adaptability.2. Combined chaotic searching with gradient algorithm of Hopfield neural networks, an optimization algorithm based on chaos is proposed. Based on the Hopfield neural network model for combinational circuit test generation, a new test generation algorithm with chaotic searching is developed. The algorithm can overcome the shortcoming of gradient algorithm and has a globally searching ability. Simulation results demonstrate that the algorithm can enhance the efficiency of test generation.3. A chaotic neural networks model for combinational circuit is derived from the Hopfield neural network model by adding a self-feedback term. Based on the chaotic neural network model, a test generation approach is proposed. A simple<WP=8>algorithm and an improved algorithm to controlling the self-feedback term is given respectively. Experiment results show that both algorithms has a high efficiency for test generation, and the test generation probability of the improved algorithm can get up to 100%.4. Combined chaotic dynamics with genetic algorithm, an improved genetic algorithm —chaos genetic algorithm with a new crossover and mutation scheme controlled by a chaotic series is proposed. A test generation method based on chaos genetic algorithm is proposed, in which a chaos genetic algorithm is used for minimizing the energy function of the Hopfield neural network model for combinational circuit test generation. Simulation results show that the scheme has a globally searching ability, speeds up test generation and enhances the efficiency of test generation.5. To account for the problem of minimizing for complete fault test set, a special chaotic searching algorithm and chaos genetic algorithm is designed. Two minimizing approaches for fault test set based on chaotic searching and chaos genetic algorithm are proposed respectively. Simulation results show that both algorithms improve significantly globally searching speed and the method based on chaos genetic algorithm is especially fit for minimizing complex large scale fault test set.更多还原