HP Uses NVIDIA AI to Speed up Metal 3D Printing

HP’s 3D Printing division uses NVIDIA’s AI tool Modulus to improve its 3D printing processes. This AI leverages physics-informed neural networks to integrate physical laws into machine learning (ML) models, boosting their efficiency and accuracy in manufacturing. Using this technology, HP developed Virtual Foundry Graphnet, which predicts and optimizes the metal sintering process.

Virtual Foundry Graphnet helps streamline production by accurately predicting how metal powders will behave during the 3D printing process. This allows manufacturers to optimize the process, reducing errors and defects and improving part quality. By using a graph-based deep learning approach, the AI predicts part deformation during metal sintering—a step known for significant deformation. AI surrogate models quickly simulate these behaviors, providing accurate predictions and speeding up the process.

Running the well-trained metal sintering inferencing engine only takes seconds to obtain final deformation values, making the process much faster and more efficient. Moreover, by sharing its AI algorithms with NVIDIA, HP helps other manufacturers innovate and improve, making advanced AI tools accessible to everyone in the industry.

The innovative strides made by HP’s 3D Printing division with Virtual Foundry Graphnet have been detailed in a recent paper titled Virtual Foundry Graphnet for Metal Sintering Deformation Prediction. Authored by HP’s team and available on arXiv, this paper discusses developing and applying the graph-based deep learning model designed to predict part deformation during metal sintering. The study highlights how this advanced model significantly accelerates simulation times compared to traditional methods.

By employing a graph-based deep learning approach, the researchers demonstrate how Virtual Foundry Graphnet can dramatically speed up deformation simulations at the voxel level, with accuracy reaching 0.7 micrometers for a single sintering step and 0.3 millimeters for a complete cycle. These advancements are largely due to the use of AI surrogate models, which provide fast and precise predictions. This significantly boosts manufacturing yield and part quality, say the authors of the study, including HP Research Engineers Rachel Chen, Chuang Gan, Jun Zeng, and Zijiang Yang, alongside NVIDIA Senior Software Engineer Mohammad Nabian and Machine Learning Engineer Meta Juheon Lee.

HP’s commitment to open-source innovation is clear in its contributions to the NVIDIA Modulus platform. By making Virtual Foundry Graphnet available to the broader manufacturing community, HP encourages collaboration and accelerates the development of physics-informed ML applications.

The HP digital Twin team has also developed innovative physics-ML models for its manufacturing digital twin, contributing this work to Modulus. This digital twin technology allows process engineers to predict and optimize both design parameters and process control parameters, improving part quality and manufacturing yield.

HP has a long history of technology innovation, including its recent development of HP Metal Jet, a metal additive manufacturing system that delivers industrial-grade throughput and quality for novel 3D metal parts beyond traditional manufacturing capabilities. As part of its digital twin effort, HP uses Virtual Foundry Graphnet to accelerate the computation predicting metal powder material phase transitions, enabling near real-time, high-fidelity emulation of the metal sintering process.

The benefits of using Virtual Foundry Graphnet are significant. It makes production faster and more efficient, which means products can be made quicker and at a lower cost. It also helps ensure that the products are of high quality. This is important for companies that must meet strict standards and deliver reliable products to their customers.

“Our team has been developing physics simulation engines based on first principles,” said Dr. Jun Zeng, HP’s distinguished technologist heading the Digital Twin effort with HP’s 3D Printing Software Organization. “We bring experimental sensing and metrology data to calibrate these physics simulation engines so that they are grounded by the manufacturing process variability. With physics-ML, once well trained, we see orders-of-magnitude speedups, and the model can run on your laptop. Such near real-time prediction delivered by physics-ML opens doors for many new applications.”

By making its AI models available to NVIDIA, HP fosters a collaborative community where manufacturers can share knowledge and solve problems together. This collaborative approach represents a significant advancement for the industry, making powerful tools accessible to smaller companies and driving collective innovation.

HP and NVIDIA have a history of successful collaborations, leveraging NVIDIA’s advanced graphics processing units (GPUs) and AI technologies to enhance HP’s computing and printing solutions. This partnership has seen NVIDIA’s GPUs integrated into HP’s high-performance workstations, providing powerful computing capabilities for professionals in various industries, including design, engineering, and media production.

In addition to hardware integration, HP and NVIDIA have worked together on AI and ML projects. By utilizing NVIDIA’s cutting-edge AI platforms, HP has developed innovative solutions like Virtual Foundry Graphnet, which improves the efficiency and accuracy of 3D printing processes. These collaborations highlight the synergy between the two companies, driving technological advancements and setting new standards in the industry.

HP’s efforts with Virtual Foundry Graphnet show how advanced technology can be shared to benefit everyone. The result is a more dynamic and inclusive industry where manufacturers of all sizes can use advanced AI to improve their processes, boost product quality, and cut costs. This kind of open-source sharing is key to driving progress in the manufacturing industry.