Deep learning makes X-ray CT inspection of 3D printed parts faster and more accurate



A new deep learning framework developed at the Department of Energy’s Oak Ridge National Laboratory is accelerating the process of inspecting additively manufactured metal parts using X-ray computed tomography, while increasing the accuracy of results. Reduced time, labor, maintenance and energy costs are expected to accelerate the expansion of additive manufacturing or 3D printing.

“Scanning speed significantly reduces costs,” said ORNL principal investigator Amir Ziabari. “And the quality is higher, so post-processing analysis becomes much easier.”

The framework is already being integrated into the software used by business partner ZEISS in its machines at the DOE’s Manufacturing Demonstration Facility at ORNL, where the companies are perfecting 3D printing methods.

ORNL researchers had previously developed a technology capable of analyzing the quality of a part during its printing. Adding a high level of imaging precision after printing provides an extra level of confidence in additive manufacturing while potentially increasing production.

“With this, we can inspect every part coming out of the 3D printing machines,” said Pradeep Bhattad, ZEISS Business Development Manager for Additive Manufacturing. “Currently, CT is limited to prototyping. But this tool alone can propel additive manufacturing towards industrialization.”

X-ray computed tomography is important to certify the solidity of a 3D printed part without damaging it. The process is similar to medical CT scans. In this case, an object placed inside a cabinet is slowly rotated and scanned at every angle by powerful X-rays. Computer algorithms use the resulting stack of two-dimensional projections to construct a 3D image showing the density of the structure internal to the object. X-ray computed tomography can be used to detect defects, analyze failures, or certify that a product is of intended composition and quality.

However, X-ray computed tomography is not widely used in additive manufacturing because current scanning and analysis methods are time-consuming and inaccurate. Metals can totally absorb low-energy X-rays in the X-ray beam, creating image inaccuracies that can be further multiplied if the object has a complex shape. The resulting flaws in the image can hide the cracks or pores that the scan is supposed to reveal. A trained technician can correct these issues during the scan, but the process is time-consuming and labor-intensive.

Ziabari and his team have developed a deep learning framework that quickly provides clearer and more accurate reconstruction and automated analysis. He will present the process his team developed at the Institute of Electrical and Electronics Engineers International Conference on Image Processing in October.

Training a supervised deep learning network for CT usually requires many expensive measurements. Since metal parts pose additional challenges, it can be difficult to obtain the proper training data. Ziabari’s approach offers a leap forward in generating realistic training data without requiring extensive experiments to collect it.

A Generative Adversarial Network, or GAN, method is used to synthetically create a realistic-looking dataset to train a neural network, leveraging physics-based simulations and computer-aided design. GAN is a class of machine learning that uses neural networks to compete with each other like in a game. It has rarely been used for practical applications like this, Ziabari said.

Because this X-ray CT frame requires scans at fewer angles to achieve accuracy, it reduced imaging time by a factor of six, Ziabari said — from about an hour to 10 minutes or less. . Working so quickly with so few viewing angles would normally add significant “noise” to the 3D image. But the ORNL algorithm taught on the training data corrects for this, even improving the detection of small defects by a factor of four or more.

The framework developed by Ziabari’s team would allow manufacturers to quickly refine their builds, even when changing designs or materials. With this approach, sample analysis can be done in a day instead of six to eight weeks, Bhattad said.

“If I can inspect the whole room very quickly in a very cost-effective way, then we have 100% confidence,” he said. “We are partnering with ORNL to make CT an accessible and reliable inspection tool for the industry.”

ORNL researchers evaluated the performance of the new frame on hundreds of samples printed with different scan settings, using complex and dense materials. These results were good, and ongoing trials at MDF aim to verify that the technique is equally effective with any type of metal alloy, Bhattad said.

This is important, because the approach developed by Ziabari’s team could significantly facilitate the certification of parts made from new metal alloys. “People don’t use new materials because they don’t know the best printing parameters,” Ziabari said. “Now, if you can characterize these materials so quickly and optimize the parameters, it would help bring these new materials to additive manufacturing.”

In fact, Ziabari said, the technology can be applied in many fields, including defense, automotive manufacturing, aerospace and electronics printing, as well as nondestructive evaluation of electric vehicle batteries.

Source link


Comments are closed.