【作者】 林剑；

【导师】 许力；

【作者基本信息】 浙江大学 ， 控制理论与控制工程， 2013， 博士

【摘要】 近几年来,人工智能技术得到了长足的发展,特别是仿生优化算法更是不断涌现,仿生算法的提出大多是受自然界中生物群体行为的启发,这些优化算法的发展为复杂优化问题的求解提供了有效的解决渠道。本文针对目前使用比较广泛,同时也是比较典型的两种仿生优化方法(和声搜索优化算法和生物地理优化算法),分析两种算法各自的特点,引入局部搜寻机制,研究提高算法性能的方法与策略,并在此基础上提出相应的改进算法,用于解决平版印刷中的实际问题。总的来讲,本文主要的工作内容与相关研究成果可归纳如下：(1)对仿生优化算法的框架模型进行了总结,给出了几种常用的改进策略和机制。在此基础上,以生物地理优化算法为例,提出了一种混合算法,该算法采用对立学习机制来产生较优初始群体,引入和声搜索算法中“调音”操作来提高其局部搜索能力,从而实现算法全局探索和局部开发能力的平衡,提高算法整体寻优能力。为验证所提混合算法的性能,将其应用于混沌系统的参数优化估计。仿真分析分别以典型的无时滞混沌系统和时滞混沌系统为对象,结果表明,在无噪声和含噪声条件下,所提混合算法可以得到比现有典型算法更好的参数估计结果,验证了所提算法的有效性和鲁棒性。(2)研究分析了平版印刷中专色油墨配色算法存在的问题,提出一种基于光谱的专色油墨配色模型。针对该配色模型配方计算过程中存在的问题,将其转换为一带多约束复杂优化问题。在所提基于生物地理优化的混合算法基础上,进一步引入约束处理机制,用于解决该优化问题。实验通过随机选取已知配方的测试样本,与现有典型算法进行比较,结果表明,在所有测试样本中,所提算法均表现出更优的收敛性能。为了进一步验证所提专色油墨配色模型的性能,将其与目前使用较为广泛的FM Ⅲ配色软件进行比较,实验结果表明,在大部分测试样本中,所提模型能得到比FM Ⅲ配色软件精确更高的配方,相应色差值也更小,并且所得结果的色差值均在允许规定范围以内。(3)在分析平版印刷工艺及相应输墨系统特点的基础上,提出了一种全局和声搜索优化的平版印刷油墨预置方法,该方法采用最小二乘支持向量机建立油墨预置问题中网点面积率与相应墨区墨键开度之间的非线性模型,并利用全局和声搜索算法对模型参数进行优化,继而用于解决平版印刷中的油墨预置问题。仿真实验结果表明,提出的全局和声搜索算法具有简单,执行速度快、优化能力强的特点,同时,将全局和声搜索算法应用于基于最小二乘支持向量的平版印刷油墨预置模型的参数优化,经与曼罗兰ROLAND700机型实际生产所得样本的测试比较,比该机自带功能具有更高的油墨预置精度。更多还原

【Abstract】 The artificial intelligent technology is addressed a remarkable advance in recent years, especially, the bionic optimization algorithms. The proposing of these algorithms are mostly inspired by biological group behaviors in the nature, the development of the optimization algorithms provide an effective alternative for solving complex optimization problem. In this dissertation, two typical and widely used bionic optimization algorithms, namely, harmony search optimization algorithm and biogeography based optimization algorithm, are discussed. Based on the analysis of each algorithm, some local searching strategies are employed to improve its performance, and further applied to deal with some practical problems in offset printing. In general, the main works and contributions of this dissertation can be summarized as follows:(1) The framework of bionic optimization algorithm model is summarized, and some improvement strategies and mechanisms are presented. Furthermore, the biogeography-based optimization (BBO) is employed as an example, and a BBO based hybrid algorithm is proposed. In this hybrid algorithm, the opposition based learning (OBL) approach is utilized to generate a better initial population, and the pitch adjusting mechanism in HS is employed to enhance the local searching ability. By fusing the operators in BBO with the OBL and pitch adjusting mechanisms, the exploration and exploitation capability may be enhanced and well balanced, and thus result in a more effective algorithm for global optimization. The proposed hybrid algorithm is further applied to the parameter estimation of chaotic systems to verify its performance. Numerical simulations are conducted on some typical chaotic systems with or without time-delay, the results show that, whatever the noise is concerned or not, the proposed scheme can obtain better estimated results than some typical existing algorithms, and the effectiveness and robustness of the proposed scheme are also demonstrated by the results.(2) The problem in the spot color matching (SCM) algorithm is studied and analyzed, and a spectrophotometric based SCM model is proposed, and can be converted into a multiple constraints complex optimization problem for its feature. The aforementioned BBO based hybrid algorithm is further improved by employing a constraint handling method, and applied to solve this problem. The experiments are conducted on some randomly selected testing samples, and the comparisons between the proposed algorithm and some other typical methods show that, for all of the samples, the proposed algorithm exhibits better convergence performance. Further comparisons are conducted between the proposed SCM model and the widely used spot color matching FM III software, the experimental results show that, in most cases, the proposed model can achieve more accuracy color matching results than the FM III software in terms of color difference, and all of the color differences obtained by the proposed model, in fact, are lower to the acceptability limit.(3) On the analysis of the offset printing technique and the corresponding ink system, a global harmony search optimization (GHSO) based ink preset method is proposed, and the nonlinear relationship model between the dot area rate and the corresponding ink key opening degree in each ink area is also established by the least squared support vector machine. The GHSO is used to optimize the parameters in the model, which is further applied to solve the ink preset problem of offset printing. The experimental results show that, the proposed GHSO has the features of easy complementation, fast running and strong optimization ability. The results of comparison with real printing data proved the validity of the proposed scheme, and the GHSO based ink preset model also has higher precision than the preset device of Roland printing machine.更多还原