Abstract: In order to reduce the influence of measurement error of experimental ultrasonic attenuation spectrum on inversion results, a weighted non-negative least squares OrtWtls inversion algorithm based on optimal regularization was proposed for particle size distribution measurement by ultrasonic attenuation spectrum. Based on the prediction of McClements and BLBL model, the weighted matrix was introduced to correct the coarse error of the attenuation spectrum. The results show that the relative error between the volume median diameter of particle size distribution and the set diameter is less than 5%, and the standard deviation of multiple measurements is less than 1 μm for the ultrasonic attenuation spectrum with noise ratio of 10 dB and 20 dB respectively. The proposed OrtWtls algorithm was used to invert the experimental ultrasonic attenuation spectrum of limestone grout with mass concentrations of 10% and 20%, the relative errors compared with the image method were 7.71% and 9.64%, which proves that the OrtWtls algorithm is feasible to calculate particle size distribution.