Intelligent denoise laser ultrasonic imaging for inspection of selective laser melting components with rough surface

Abstract

The random microdefects are inevitable during the Selective laser melting (SLM) process due to the principle of discrete-stacking. The rough surface induced strong background noise reduces the probability of detection of traditional laser ultrasonic testing system. In this study, an intelligent denoise laser ultrasonic imaging method was developed to inspect the micro defects on the rough surface of SLM components: (1) a non-contact laser ultrasonic scanning setup was established for data acquisition of multiple ultrasonic signals; (2) a denoising algorithm based on unsupervised machine learning was designed and trained by abundant ultrasonic data to enhance the signal to noise ratio (SNR); (3) a signal matching algorithm based on the cross-correlation and self-normalized method was established to match the amplitude and arrival time of Rayleigh waves from different scanning points. The performance of developed method was verified using micro hole defects on the rough surface of a SLM part. The results indicated that the established denoising algorithm could increase the average SNR from 27.0 dB to 35.2 dB. All holes with diameter of 50 μm and 100 μm can be detected and sized based on the high SNR image without removing the rough surface. The conclusion can be drawn that the proposed intelligent denoise laser ultrasonic imaging method is a very potential way for the online inspection of SLM.