Application of an Adaptive Neuro-Fuzzy System in the Prediction of HPC Compressive Strength

摘要

In this paper, application of Adaptive Network Based Fuzzy Inference System (ANFIS) in estimation of compressive strength of high performance concrete (HPC) was investigated. ANFIS is a hybrid structure that is based on fuzzy If-Then Rules that are represented in a network. To predict the compressive strength of HPC with ANFIS, totally 429 records collected from four different resources. Records were randomly divided into two sets. One set is called training set that is used to train the ANFIS models and another set called testing pairs was used for evaluation of the models. Totally 56 different ANFIS models with various membership functions were used to predict the compressive strength of HPC. These models were evaluated with root means square (RMS) and correlation factors (CFs). Finally with comparison of the capabilities of the models, some models were proposed as optimum models. It was found that model with triangular membership function (trimf) and "St334333" structure (4 membership function for super plasticizer and three membership function for other concrete mix components) is the best model to predict the HPC's compressive strength regarding its components.