Prediction and Regulation of SCC's Shrinkage Using the PSO-BPNN Model

The shrinkage deformation is a significant risk to self-compacting concrete (SCC)-filled steel tube structures. It was essential to understand the concrete autogenous shrinkage strain before being regulated in order to determine compensation shrinkage measures. In this study, A PSO-BPNN model was constructed, which is based on the Particle Swarm Optimization-Back Propagation Neural Networks (PSO-BPNN), and the autogenous shrinkage strain of SCC was predicted based on PSO-BPNN before being regulated. Moreover, some experiments about compensating for shrinkage by expansion and by a combination of expansion and contraction were investigated. Based on this prediction, a series of experiments was conducted on the regulation of the shrinkage deformation of SCC for an actual bridge project. The results indicated that a good consistency of PSO-BPNN between predicted and measured values, demonstrating that PSO-BPNN is a model with high accuracy in predicting concrete autogenous shrinkage strain before regulation, and as a guidance for regulation to compensate for shrinkage. The prediction error was less than 10% for 28-day self-shrinkage, and the experimental workload was reduced. The PSO-BPNN is a convenient tool for predicting the shrinkage of SCC, enabling the determination of dosages of expansion agent and reducing shrinkage agent to achieve SCC's shrinkage regulation.