基于XGBoost算法的仿真混凝土材料配合比设计方法

Design Method of Simulated Concrete Material Mix Ratio Based on XGBoost Algorithm

针对传统仿真混凝土配制过程周期长、耗费成本高的问题,提出了一种快速设计仿真混凝土材料配合比的方法。基于现有仿真混凝土的研究成果,构建了仿真混凝土材料配比数据库;利用XGBoost算法对数据库中的数据信息进行训练,以不同配合比混凝土材料的抗压强度、弹性模量的预测值与实测值之间的决定系数作为目标函数,利用交叉验证和网格寻优建立仿真混凝土配合比计算模型;最后利用Python开发了仅需输入期望的抗压强度、弹性模量及允许误差即可设计出仿真混凝土材料配合比的程序。实际应用表明,仿真混凝土材料配合比设计程序运算速度快、精度高,计算的预测值与实测值吻合好,是一种方便快捷的新方法,并可大大缩短模型试验周期、降低试验成本。

Aiming at the problems of long cycle and high cost in the preparation process of the traditional simulated concrete,a method for quickly designing the mixing ratio of simulated concrete materials was studied.Based on the previous research results of the simulated concrete,a database of simulated concrete material ratios was constructed.The data information in the database was trained using the XGBoost algorithm.Taking the determination coefficient between the predicted values of the compressive strength and elastic modulus of the mix ratios and the actual measured values as the objective function,the cross-validation and grid optimization were used to establish a simulation concrete mix calculation model.Finally,Python was used to develop a simulation concrete material that only needs to input the desired compressive strength and elastic modulus and allowable errors.Practical application shows that the simulation concrete material mix design program has fast calculation speed and high accuracy,and the predicted value is in good agreement with the actual measured value.This method provides a convenient and fast method,which greatly shortens the model test cycle and reduces the test cost.