Silicon Volley: Designers Tap Generative AI for a Chip Assist
The work demonstrates how companies in highly specialized fields can train large language models (LLMs) on their internal data to build assistants that increase productivity.
The paper details how NVIDIA engineers created for their internal use a custom LLM, called ChipNeMo, trained on the company’s internal data to generate and optimize software and assist human designers. Long term, engineers hope to apply generative AI to each stage of chip design, potentially reaping significant gains in overall productivity, said Ren, whose career spans more than 20 years in EDA. After surveying NVIDIA engineers for possible use cases, the research team chose three to start: a chatbot, a code generator and an analysis tool.
On chip-design tasks, custom ChipNeMo models with as few as 13 billion parameters match or exceed performance of even much larger general-purpose LLMs like LLaMA2 with 70 billion parameters. In some use cases, ChipNeMo models were dramatically better.
New software solution to accelerate manual CAM programming time by 80%, enabling manufacturers to be more productive
CAM Assist – currently available as a plug-in within Autodesk’s Fusion 360 software platform – uses advanced computer science techniques to generate professional machining strategies for 3-axis parts in seconds, which could take CNC machine programmers hours or even days to manually create. As a result, the amount of time it takes to program a CNC machine to make a component – a bottleneck in many factories, due to a global skills shortage – is reduced by up to 80%, compared to the previous manual programming process.
Previously, depending on complexity, it could take a CAM programmer between an hour to several days to determine the best strategy to CNC machine a new component. This includes selecting the correct tools, toolpaths, and techniques – determining between hundreds of thousands of potential variables and approaches.Instead, CAM Assist uses advanced computational optimisation and AI inference techniques to rapidly determine a professional strategy and toolset needed to manufacture a part, along with the most appropriate cutting speeds and feeds from the user’s library.
Simplify Your Thermal Simulation With Immersed Boundary Method
However, common bottlenecks to simulation have been CAD preparation and the numerical discretization of that model (meshing). Both consume time and manual intervention. The advent of advanced physics solvers and novel meshing techniques, such as the immersed boundary method, means that engineers spend less time making their CAD models simulation-ready and more time on insight-driven design. Skipping the time-intensive CAD preparation also opens up the possibility of doing simulations very early when some components are still in the draft stage and comparing many variants that otherwise would have required repeated CAD simplification efforts.
The Immersed Boundary method addresses the core of this dilemma. It completely removes the CAD preparation or reduces it to a few minutes at most. At the same time, the physics-driven meshing avoids high mesh resolutions on detailed CAD features that are insignificant to the system’s thermal behavior. Yet, it resolves physically relevant regions like power sources or flow channels to the level the user requires. This level might differ significantly based on the current simulation intent.
Unlocking the Value Potential of Additive Manufacturing
Transitioning to AM requires not only a change in mindset but more importantly, the ability to quickly and easily identify which parts are best suited for the additive manufacturing process. This is where AI and machine learning are now bridging the gap between traditional AM –where most of its value materializes in the form of functional prototypes – and more advanced additive manufacturing operations. “We have upwards of a million part numbers,” said Werner Stapela, head of global additive design and manufacturing at Danfoss – an international leader in drives, HVAC and power management systems. “So, it would be impossible for us to manually analyze each one to determine whether additive manufacturing would either add value or reduce costs.”
“We have been utilizing 3D printing for decades, mostly for prototyping, but the Castor3D software allows us to focus on our end components and more specifically the costs associated with that,” added Stapela. The software’s algorithm and machine learning can scan thousands of parts at once by analyzing CAD files. It evaluates five factors: materials, CAD geometry, costs, lead time and strength testing to identify suitable parts for AM. The software can also make design for additive manufacturing (DfAM) suggestions regarding part consolidation and weight reduction opportunities.
What is CAD: the technological foundations of CAD software
In 3D Computer Aided Design systems, the most important building block is the 3D geometry. This is created by a 3D geometry kernel, which is a software component, responsible for geometry calculations. Some examples of the operations that a geometry kernel can provide are boolean operations, extrusions, sweeps, lofts and many more. Typically, a geometry kernel has thousands of geometry operations implemented. By the ’80s, boundary representation or b-rep became the industry standard mathematical model for 3D manufacturing applications. While it would be great if standard meant not just a standard for mathematical principles, but a standard implementation, this could not be further away from the truth. Over time different CAD vendors implemented plenty of different boundary representation engines (aka. geometry kernels), which is one of the main reasons for poor compatibility of different CAD systems even today.
Geometry kernels are not easy to write nor to replace. One of the most expensive software issues in history was when Dassault Systemes replaced CATIA’s kernel upgrading from V4 to V5. The incompatibility issues that this version introduced cost Airbus an estimated $6.1 billion due to delays in production.
CAD programs usually use different tricks to display a huge amount of geometry. One of the typical tricks is called LoD (Level of Detail). LoD simply means that when you look at something from further away, a lower resolution mesh is going to be rendered, but when you zoom in, the CAD system will load a higher resolution version of the geometry. This way when you zoom out and look at a large assembly, the rendering performance will not degrade. It’s worth mentioning that this tessellation data is responsible for a significant amount of the memory consumption of CAD systems - sometimes more than 50% of the memory that a CAD system is using is allocated to the tessellation data.
🚙 Application of optimized convolutional neural network to fixture layout in automotive parts
Fixture layout is a complex task that significantly impacts manufacturing costs and requires the expertise of well-trained engineers. While most research approaches to automating the fixture layout process use optimization or rule-based frameworks, this paper presents a novel approach using supervised learning. The proposed framework replicates the 3-2-1 locating principle to layout fixtures for sheet metal designs. This principle ensures the correct fixing of an object by restricting its degrees of freedom. One main novelty of the proposed framework is the use of topographic maps generated from sheet metal design data as input for a convolutional neural network (CNN). These maps are created by projecting the geometry onto a plane and converting the Z coordinate into gray-scale pixel values. The framework is also novel in its ability to reuse knowledge about fixturing to lay out new workpieces and in its integration with a CAD environment as an add-in. The results of the hyperparameter-tuned CNN for regression show high accuracy and fast convergence, demonstrating the usability of the model for industrial applications. The framework was first tested using automotive b-pillar designs and was found to have high accuracy (≈ 100%) in classifying these designs. The proposed framework offers a promising approach for automating the complex task of fixture layout in sheet metal design.
CAD-based data augmentation and transfer learning empowers part classification in manufacturing
Especially in manufacturing systems with small batches or customized products, as well as in remanufacturing and recycling facilities, there is a wide variety of part types that may be previously unseen. It is crucial to accurately identify these parts based on their type for traceability or sorting purposes. One approach that has shown promising results for this task is deep learning–based image classification, which can classify a part based on its visual appearance in camera images. However, this approach relies on large labeled datasets of real-world images, which can be challenging to obtain, especially for parts manufactured for the first time or whose appearance is unknown. To overcome this challenge, we propose generating highly realistic synthetic images based on photo-realistically rendered computer-aided design (CAD) data. Using this commonly available source, we aim to reduce the manual effort required for data generation and preparation and improve the classification performance of deep learning models using transfer learning. In this approach, we demonstrate the creation of a parametric rendering pipeline and show how it can be used to train models for a 30-class classification problem with typical engineering parts in an industrial use case. We also demonstrate how our method’s entropy gain improves the classification performance in various deep image classification models.
How to Design Furniture in Fusion 360: Everything You Need to Know
Recently, we hosted an extensive, week-long workshop on designing furniture in Fusion 360. If you’ve been curious about how to take advantage of workflows that include parametric design and production automation — or even just want to learn the basics — read on because this article is for you. We’ll share each video in the series, along with a brief recap of what you’ll learn from watching. Let’s get started.
Manufacturing Manakins for Medical Simulation and Training
Human patient simulators may mimic the human body with varying degrees of realism—or fidelity—and can be used in almost every aspect of healthcare education. The most effective medical training devices are those that have the ability to create accurate modeling of the underlying structures of the human body and replicating them digitally and physically, noted Alban. It is why Simetri’s anatomical models and medical training aides integrate electronic, mechanical and computational components and turns to materials science for innovations in soft and skeletal tissue.
The roadmap to digitization for Simetri, said Alban, started first on the mechanical side, when mechanical models started to go from sketches to using SolidWorks and 3D models, and then embedding sensors to capture data before writing the related software and then advancing the software development capability.
In another development, software can monitor when skin has been cut, and when and if the correct fascia (connective tissue encasing the muscle) has been cut. That data is transmitted digitally to the manakin, and the physiology model of that manakin is updated as a result of that new data and, therefore, displays new vital signs. “If you will have done it the right way, you will lose pulse at the foot, but if you do this procedure correctly, you will gain back pulse at the foot because you’re allowing circulation to flow through,” explained Alban.
How Data-driven Manufacturing Unlocks Speed and Transparency during Injection Molding Process
Many of the parts we manufacture have at least one measurement that’s mission critical. Maybe the parts won’t work in an assembly unless a planned hole is within spec. Typical CAD models provide an opportunity to include specific dimensions, but what if you could tell your manufacturer early-on that dimension X is the one that makes or breaks a part? That’s where Critical-to-Quality (CTQ) comes in.
The CTQ specifications that you include in your quote and CAD model help to guide us during manufacturing, saving another critical dimension: TIME. We can often tell you if it’s possible for us to make your part before the mold is cut.
CTQ is also an important element of our digital manufacturing processes because we use these specs to evaluate initial runs of your parts. Let’s say that your parts require sample qualification or part validation. CTQ becomes even more crucial at that point because the data that flows from those initial shots can predict the future tolerances for those critical dimensions, revealing the suitability of end-use parts for a given assembly.