Point cloud based online detection of geometric defects for the certification of additively manufactured mechanical metamaterials

Abstract

Unlike conventional materials, the properties of a mechanical metamaterial are governed by its periodically-arranged unit cells. It is critical to precisely realize the designed unit cell geometry so that desired properties can be achieved. Owing to its high-level freedom in design and fabrication, additive manufacturing (AM) has been increasingly leveraged for the production of mechanical metamaterials. However, geometric defects largely exist in AM products. The gap between “as-designed” and “as-built” unit cell geometries will compromise the properties of the printed metamaterial and even cause safety hazards for mission-critical applications. This study develops a novel framework that harnesses in-situ point cloud sensing and analytics for the detection and characterization of geometric defects in additively manufactured mechanical metamaterials. Specifically, upon the printing of each layer, a point cloud is obtained to capture the geometries of unit cells. A recurrence plot based approach is developed to characterize how each printed unit cell departs from the as-designed counterpart. After that, we integrate low-rank tensor decomposition with one-class classification to determine whether the unit cells on the printed layer are associated with geometric defects. Results of simulations and real-world experiments have shown that the developed framework effectively handles the complexity of mechanical metamaterials and characterizes geometric defects for online quality assurance.