多重群体遗传规划在高性能混凝土强度预测中的应用.pdfVIP

四川建筑科学研究 第37卷 第4期 21O Sichuan Building Science 2011年8月 多重群体遗传规划在高性能混凝土强度预测中的应用 赵胜利。，宋玲玲 ，张东原。 (1．河北农业大学城建学院，河北保定 071001； 2．保定供水总公司，河北保定 071001； 3．河北建设集团混凝土分公司，河北保定 071000) 摘 要：针对影响高性能混凝土强度的主要因素作为输入因子，28 d抗压强度作为输出变量，应用遗传规划理论(GP)建立了 高性能混凝土强度预测的非线性显式数学解析式模型。为了更好地保持进化过程中的遗传多样性，提高求解此问题的效率， 提出了多重群体遗传规划理论。通过实测数据进行验证，并分别与线性回归模型和神经网络模型相比较，结果表明，多重群 体遗传规划(MGGP)模型具有更高的拟合精度和更好的预测效果，在高性能混凝土强度预测方面有很强的实用价值。 关键词：遗传规划；高性能混凝土；强度预测；多重群体 中图分类号：TU528．31 文献标识码：A 文章编号：1008—1933(2011)o4—210—03 Application of Multi-Group genetic programming in strength forecast of high performance concrete ZHA0 Shengli ，SONG Lingling ，ZHANG Dongyuan (1．Rural and Urban Construction College，Agricultural University of Hebei，Baoding 071001，China； 2．Baoding Water Company，Baoding 071001，China； 3．Concrete Branch，Construction Group of Hebei Province，Baoding 071000，China) Abstract：The main factors that influence the strength of high performance concrete are considered as input variables of genetic programming(GP)model and 28一day compressive strength is seen as output variable．A visible nonlinear-mathematical—model of high performance concrete strength is舀ven by appl~ng the method of improved GP，in which creating initial population and fitness function are adjusted mming at the special problem．In order to keep the variety and improve the efficiency，Multi·Group GP(MGGP)is put forward．The performance of the proposed MGGP model is analysed and the result of it is compared with that of the linear regression and BP neural network model which show that the MGGP model have higher fitting precision with the experiment data and better prediction effect，thus，the proposed MGGP model is useful to be used to predict the strength of high performance concrete． Key words：genetic programming；high performance concrete；strength forecast；Multi-Group