Evaluation of mean grain size using the multi-scale ultrasonic attenuation coefficient

•We present a new idea of ultrasonic multi-scale attenuation coefficient.•We correlate the coefficient with grain sizes determined from micrographs.•The results show the model has less systematic error and narrower error band.

An evaluation model based on the multi-scale ultrasonic attenuation coefficient was developed to control both systematic error and random error. AISI 304 stainless steel was used to validate the presented model. Wavelet transformation was used to obtain the variation of ultrasonic signal over time and scale. Particle swarm optimization was utilized to correlate the coefficient with grain sizes. The model shows the attenuation of all scales increased with the grain size, and ultrasound attenuates faster on smaller scales. Compared with the ultrasonic velocity method and the traditional attenuation method, the proposed method has less systematic error and random error.