Defect Detection in Additively Manufactured Metal Parts Using In-Situ Steady-State Ultrasonic Response Data

The increasing implementation of additively manufactured parts into safety critical applications within a wide range of industries is accelerating the demand for non-destructive evaluation as a means of quality control for defect detection during the additive manufacturing build process.  The need for in-process quality control is driven by both cost and performance.  Identification of a critical defect in the part before its completion can enable early termination of the build, resulting in both reduced material costs and machine time.  Additionally, current post-build inspection techniques, such as x-ray computed tomography, have limited resolution capability as the size of the part increases.  As such, an ideal inspection technique for metal additive manufacturing would be capable of:

·       direct part measurements performed on a layer-by-layer basis as the part is being built,

·       fast enough to not alter the thermal history and subsequent material properties of the part,

·       non-destructive evaluation measurements over the whole part and whole build plate,

·       identifying common defects that form during the build process, e.g. porosity, inter-layer cracks, part separation from the build plate due to thermal stress.

The application of acoustic wavenumber spectroscopy is proposed to address these inspection needs in metal additive manufacturing.  Acoustic wavenumber spectroscopy is an acquisition and processing method created at Los Alamos National Laboratory that provides a means for rapid, non-destructive evaluation of structures using steady-state ultrasonic measurements. This technology is two orders of magnitude faster than conventional ultrasonic inspection techniques and has successfully been utilized in a wide range of applications, including detecting weld cracks, corrosion of large thin-walled parts, and delamination in carbon-fiber reinforced panels.

An adaptation of existing acoustic wavenumber spectroscopy technology was implemented into a direct laser metal sintering machine to perform direct-part measurements during the fabrication of 304L stainless steel parts. A scanning laser Doppler vibrometer was used to record the surface response of each layer resulting from steady-state ultrasonic excitation of the parts through the build plate. The result is a unique, three-dimensional inspection volume of the entire build.  The processing of data for each layer is quick enough to be performed during the lasing of the subsequent layer, enabling the potential for early termination of the build when critical defects are identified.

Various processing techniques were used to detect changes in the steady-state response that indicate different defect types. For example, the response to the steady-state ultrasonic excitation differs between bulk porosity areas and inter-layer cracking. The data analysis and machine-learning techniques used to identify features found in the response data to locate regions of defects is presented. This work ultimately demonstrates a practical means for in-situ monitoring of additive manufacturing parts to ensure quality control for safety-critical applications.