逐一混合遗传算法在大坝变形预报中应用

建立了大坝变形预报的逐一混合遗传模型,并将其与整体遗传模型、逐一Levenberg-Marquardt (LM)模型、整体LM模型进行了比较. 工程实例表明,在建模样本相同,预报因子相同,且结构参数不变的条件下,逐一混合遗传模型和整体遗传模型的预报精度分别高于逐一LM模型和整体LM模型,且预报结果稳定;逐一混合遗传模型和逐一LM模型的预报精度分别高于整体遗传模型和整体LM模型;随着样本的积累,逐一混合遗传模型的预报精度不断提高, 并具有实时性的优点, 可以准确、有效地应用于大坝变形监测量的实时预报.

Application of seriatim hybrid genetic algorithm to prediction of dam deformation

A new model named seriatim hybrid genetic algorithm model based Levenberg-Marquardt algorithm(SGA-LM) is developed to predict dam deformation.The model is compared with whole genetic algorithm model based Levenberg-Marquardt algorithm (WGA-LM),seriatim Levenberg-Marquardt model(SLM) and whole Levenberg-Marquardt model(WLM).Case study shows that SGA-LM and WGA-LM are superior to SLM and WLM respectively,and SGA-LM and SLM are superior to WGA-LM and WLM respectively in prediction accuracy under the circumstances of the same modeling samples,forecasting factors and structure parameters.The SGA-LM has characteristics of real-time and stable forecasting results.With accumulation of samples,the forecasting precision of SGA-LM is continually improved.Thus,SGA-LM is comparatively effective in real-time prediction of dam deformation.