Abstract: Genetic algorithm is widely used in engineering and optimization fields.The performance of genetic algorithm and extension of its application fields are affected by the limitation of its code.Float code is super to other codes in function and constraint optimization fields.Noise is brought by the code in selection and crossover，its influence to the performance of genetic algorithm is not noticed by researchers.In this paper，the property of the noise in float code is mostly analyzed by it in inherit operation.The mechanism of denoising in float code is researched by it with wavelet thresholding coefficient.Denoising is implemented by mutation.That wavelet theory used in float code genetic algorithm is credible for mutation denoising，the method is feasible，indicated by the results of research and experiment of it.

Key words: genetic algorithm, wavelet coefficient, threshold, mutation denoising

摘要： 遗传算法在工程和优化领域中得到了广泛地应用，但其编码的局限性影响了遗传算法的性能和应用领域的拓展。浮点数编码在函数优化和约束优化领域明显有效于其他编码，但编码在选择、交叉操作中产生的噪声对遗传算法性能的影响未被人们所重视。着重分析了遗传操作中浮点数编码噪声的性质，研究利用小波阈值化系数对浮点数编码消噪的机理，通过变异运算消噪。结果表明，将小波理论用于浮点数编码遗传算法的变异消噪是可靠的、方法是可行的。

关键词: 遗传算法, 小波系数, 阈值, 变异消噪