Abstract: In order to understand the evolution regularity of freeze-thaw damage of foam concrete and accurately identify its damage type, uniaxial compression-acoustic emission joint tests were carried out on foam concrete with different freeze-thaw times at different loading rates. The acoustic emission signal label data were obtained by Gaussian mixture model (GMM) clustering, and the cross clusters in the data were separated by support vector machine (SVM). A new method of acoustic emission signal damage pattern recognition based on GMM-SVM algorithm was proposed. The results show that freeze-thaw action will accelerate the transformation of foam concrete failure characteristics from brittleness to ductility. The tensile stress dominates the damage mode of foam concrete in the early stage of loading, and the shear stress increases gradually with the loading process and reaches the maximum in the unstable failure stage, and the maximum decrease in the proportion of tensile cracks is 41.25%. Through the GMM-SVM algorithm, the cross-cluster of tensile-shear fracture can be accurately distinguished. It is verified that this method can correspond to different damage modes of foam concrete throughout the full loading process.

Key words: foam concrete, freeze-thaw damage, acoustic emission, crack classification, Gaussian mixture model, support vector machine