Fast 360-Degree 3D Metrology for Directed Energy Deposition

Imagine a factory floor where massive metal parts are not cut from a block or cast in a mold, but built up from nothing, layer by layer, like a 3D printer made of fire and metal dust. This process, known as directed energy deposition, uses a powerful laser to melt a stream of metal powder, fusing it into a solid shape as it moves. It is a technique capable of creating complex components for airplanes, cars, and medical devices, or even repairing broken parts on the spot. However, the very nature of this process makes it difficult to inspect. The metal being built is often hot, uneven, and highly reflective, flashing light back at sensors in a way that blinds them. If a defect forms—a tiny hole, a warped layer, or a weak spot—it can ruin the entire part, which might cost thousands of dollars and take days to produce. To prevent this, engineers need a way to see the object clearly while it is being made, capturing a perfect three-dimensional map of its surface in the blink of an eye, before the next layer of metal is added.

For years, the tools available to measure these shiny, moving surfaces have been too slow or too limited. Traditional methods often require taking a series of pictures over time, which is impossible if the object is moving or changing shape rapidly. Others struggle with the glare of the metal, producing images full of bright spots and missing data. A team of researchers from the University of Arizona and Northwestern University has now developed a new system that solves these problems. They created a fast, all-around 3D camera setup that can scan a metal part from every angle in a single instant, filtering out the blinding reflections to reveal the true shape of the surface. Their work, detailed in a recent study, demonstrates a system capable of measuring complex metal parts with a precision finer than the width of a human hair, offering a new way to ensure quality in high-tech manufacturing.

The core of this new system is a clever arrangement of cameras and projectors placed around the printing chamber. Instead of relying on a single viewpoint, which often misses hidden corners or steep walls, the researchers positioned four cameras and three projectors in a circle around the work area. This allows them to see the entire object at once, from every side. To capture the shape, the projectors shine a pattern of light stripes onto the metal surface. When these stripes hit the uneven, bumpy surface of the growing metal part, they bend and distort. The cameras record these distortions, and a computer uses the way the lines have warped to calculate exactly how high or low every point on the surface is. This technique, known as fringe projection profilometry, is a standard way to measure shapes, but the researchers had to adapt it for the unique challenges of metal printing.

The biggest challenge was the glare. Metal surfaces, especially when they are fresh and hot, act like mirrors, reflecting the light directly back into the camera lenses. This creates bright, saturated spots that wash out the delicate stripe patterns, making it impossible to measure those areas. To fix this, the team added a layer of polarized filters to their equipment. Think of polarized light like a stream of water flowing through a fence with vertical slats; if you try to pass it through a second fence with horizontal slats, it gets blocked. The researchers placed a filter in front of the projector that forces the light to vibrate in one specific direction. They then placed a matching filter in front of each camera, oriented to block that same direction. When the light hits the metal and bounces off directly, it keeps its original vibration and gets blocked by the camera's filter. However, the light that scatters off the rough, matte parts of the metal surface gets scrambled, changing its vibration direction. This scattered light can pass through the camera's filter, allowing the system to see the true texture of the metal while ignoring the blinding glare.

To make the system fast enough to keep up with the printing process, the researchers moved away from the traditional method of taking multiple pictures in a sequence. Standard systems often need to project several different patterns and take a series of photos to build a 3D model, a process that takes too long and risks errors if the object moves even slightly. Instead, this new system uses a mathematical technique that allows it to calculate the entire 3D shape from just one single photograph. By analyzing the frequency of the light stripes in that single image, the computer can instantly determine the depth of every point on the surface. This means the entire measurement takes less than half a second, fast enough to be performed between the steps of the printing process without stopping the machine for long.

The team tested their system inside a real metal printing machine, measuring a complex part shaped like a balloon dog and a tall, thin wall. In the case of the balloon dog, the system successfully captured the height of each new layer as it was added, detecting changes as small as 700 micrometers. When they compared their measurements to those taken by a high-end industrial microscope, the difference was only 64 micrometers, a level of accuracy that proves the system can reliably track the growth of the part. The system also managed to resolve fine details, such as the tiny ripples left by the laser as it moved across the surface, showing that it can see the subtle textures of the manufacturing process.

Perhaps the most difficult test came with the thin wall, a structure where two parallel surfaces stand very close together. Conventional 3D scanning methods often struggle with such shapes, sometimes merging the two walls into a single flat surface because the software gets confused by the lack of space between them. The new system, however, used a smart algorithm to combine the views from all four cameras. It analyzed the direction the surface was facing at every point and chose the best view for each spot, discarding the conflicting data. This allowed it to reconstruct the thin wall with its true thickness intact, preserving the delicate geometry that other methods would have lost. The final result was a complete, 360-degree map of the object, accurate to within 55 micrometers in terms of precision.

This achievement marks a significant step forward for the field of metal additive manufacturing. By combining a fast, single-shot camera technique with a method to filter out glare, the researchers have created a tool that can monitor the quality of metal parts as they are being built. The system does not just see the object; it sees the object clearly, despite the heat, the motion, and the blinding reflections that usually hide the truth. While the current setup requires a brief pause between printing layers to take the measurement, the speed of the system suggests that future versions could eventually monitor the process in real-time, even while the laser is active. This capability could transform how manufacturers build critical components, allowing them to catch errors the moment they happen and ensuring that the final product is perfect, layer by layer.