Reinforced concrete beams are often prone to cracking, in particular when being subjected to severe operating conditions. It is therefore imperative to investigate the damage properties and hence prevent beam damage in a timely manner. The acoustic emission signal contains useful information
for effective damage property extraction and recognition. However, acoustic emission suffers from noise. To eliminate the noise, this paper presents a new approach to identify the beam damage properties using acoustic emission (AE) and non-linear independent component analysis (NICA). A radial
basis function (RBF) neural network was used to estimate the mixed non-linear noise in the original acoustic signal, and then the NICA was employed to separate the true damage signal with the noise removed. Lastly, the short-time Fourier transform (STFT) was adopted to extract the distinct
damage features from the separated signal in the time-frequency domain to recognise the beam damage properties. A three-point bending loading test has been carried out to evaluate the proposed method. The experiment results show that the mixed noise can be eliminated effectively by the use
of non-linear ICA, and the identification precision is acceptable. Moreover, the performance of the proposed method is superior to the method without ICA processing.