基于BP神经网络的超高性能混凝土抗压强度预测

The Prediction of Compressive Strength of Ultra-High Performance Concrete Based on BP Neural Networks

摘要: 超高性能混凝土(UHPC)因其优异的力学性能和耐久性，广泛应用于高层建筑、桥梁等结构中。抗压强度是评估其力学性能的重要指标，掌握其规律并准确预测抗压强度是研究UHPC的关键。然而，传统的抗压强度预测方法往往存在复杂性、计算量大和缺乏自适应能力的问题。相较之下，BP神经网络能够通过学习大量实验数据，自主调整网络权重，从而优化预测精度。本文基于BP神经网络，提出了一种用于预测UHPC抗压强度的新方法。通过收集不同配比下的实验数据，设计了一个包含输入层(10个节点)、双隐含层(第一层12个节点、第二层24个节点)和输出层(1个节点)的神经网络结构。结果表明，BP神经网络能够有效捕捉UHPC的非线性特征，并具有较高的预测精度。研究表明，该方法为UHPC抗压强度的快速预测提供了有效的解决方案，并为其配比设计与工程应用提供了理论支持。

Abstract: Ultra-high performance concrete (UHPC) is widely used in high-rise buildings, bridges, and other structures due to its excellent mechanical properties and durability. Compressive strength is a key indicator for evaluating its mechanical performance, and understanding its patterns and accurately predicting compressive strength are crucial for UHPC research. However, traditional methods for predicting compressive strength often suffer from complexity, large computational demands, and a lack of adaptability. In contrast, BP neural networks can learn from large amounts of experimental data and autonomously adjust the network weights, thereby optimizing prediction accuracy. This paper proposes a novel method for predicting the compressive strength of UHPC based on a BP neural network. By collecting experimental data from different mix proportions, a neural network structure consisting of an input layer (10 nodes), two hidden layers (12 nodes in the first layer and 24 nodes in the second layer), and an output layer (1 node) was designed. The results indicate that the BP neural network can effectively capture the nonlinear characteristics of UHPC and achieve high prediction accuracy. The study demonstrates that this method provides an effective solution for the rapid prediction of UHPC compressive strength and offers theoretical support for its mix design and engineering applications.