利用TS与GA的混合算法（TSGA）求取剩余静校正量

静校正问题是一个具有多参数、多极值的全局优化问题,当大量未知参数存在时,常规的遗传算法(GA)几乎必然存在早熟收敛现象,很难保证全局收敛。为此,在Glover理论的基础上,本文提出一种GA与TS(禁忌搜索算法)的混合策略——TSGA。通过把TS独有的记忆功能引入到GA进化搜索过程之中,构造了新的重组算子TSR,针对GA爬山能力差和易早熟的缺陷,把TS作为GA的灾变算子TSM,即当GA陷于局部极值时,用TS对GA进行适当规模灾变,这样既保持了GA的已有搜索成果,又可以使GA跳出可能的局部极值陷阱,最终使搜索向全局极值前进。TSGA通过TS算法克服了GA爬山能力差的弱点,综合了GA具有多出发点、TS具有记忆功能和爬山能力强的优点,较好地解决了剩余静校正量求取的复杂非线性问题。模型数据处理结果表明,文中方法具有适应能力强、能快速收敛于大静校正量最优解的优点,是一项实用的求取复杂地形条件下静校正量的方法。

Using hybrid algorithm (TSGA) of TS and GA for computation of residual statics.

The issue of static corrections is global optimum issue with multi-parameters and multi-extreme-values. The ordinary genetic algorithm (GA) almost certainly appears premature convergence when a great deal of unknown parameters exists and is difficult to ensure global convergence. For that reason, based on the Glover theory, the paper presented a strategy mixed GA with TS-TSGA. The new recomposed operator is constructed by introducing the TS-distinct memorial function into GA evolution searching process; in view of the disadvantages of poor GA climbing ability and premature, taking TS as catastrophe factor TSR of GA, i. e. using TS to conduct the catastrophe of GA in an adequate scale when GA is finished in local extreme values, that can both keep the GA-searched results and skip out the possible trap of local extreme values and the search can finally progress to global extreme value. The TSGA better solved complicated non-linear problem of residual statics computation by overcoming poor climbing ability of GA through using TS algorithm and integrating the advantages of multiple starting points of GA with good memorial function and strong climbing ability of TS. The processed results of modeling data showed the method in the paper has advantages of strong adaptability and fast converging to the optimum solution of larger statics, which is a practical approach computing statics in complex topographic conditions.