CAD workstation AI heralds a new era in engineering

Related Vendors

Dell GmbH

Taiwan Machine Tool & Accessory Builders' Association (TMBA)

Artificial intelligence has arrived in engineering and is changing the way designers work as well as the definition of the profession: specialists must acquire new skills and use powerful workstations to meet the demands of AI.

Peter Beck, Field Product Manager Workstations and Rugged at Dell Technologies in Germany

Body and aircraft parts, as well as highly complex machines for the manufacturing industry, share the fact that their development and design no longer take place on the drawing board. Digital processes via Computer Aided Design (CAD) and Computer Aided Engineering (CAE) have become standard, as they offer a decisive advantage: they reduce the effort for designers and engineers from the initial idea to final production. AI-assisted systems have long been facilitating work in CAD and CAE—now, generative AI (GenAI) is on the rise as a tool that will provide a real productivity boost.

Natural language for communication

GenAI and AI assistants operating on this model, like ChatGPT from OpenAI or Microsoft Copilot, are booming—and the market for available digital helpers, which sometimes deliver astonishing results, is steadily growing. The foundation for generative AI applications are Large Language Models (LLMs), meaning "large language models." On one hand, they function as knowledge databases containing billions of data sets, forming the basis for the assistant's precise responses. On the other hand, they give digital helpers the ability to understand, process, and output natural language. This enables a seemingly natural conversation between human and machine, making it easier to intuitively interact with the new technology. With cameras as a substitute for eyes, certain assistants can now even process visual impressions, as well as acoustic inputs and outputs with human speech.

AI-based engineering

Generative AI, when meaningfully integrated into CAD or CAE applications, increases efficiency for designers: Instead of manually modifying parts of the design using complex formulas, experts can use generative AI to perform calculations and desired modifications in their project using natural language. When properly programmed, the AI automatically considers all abstract and known factors.

Even more possibilities arise when AI-supported image generation comes into play. Designers and engineers often consider factors such as feasibility, manufacturability, and potential costs in problem-solving. AI, however, initially disregards such limitations and attempts to generate the most efficient solution for a specific problem. In designing a new wing for an aircraft, this might mean that no strut of the final product is straight in the model. Therefore, correct and meaningful training of AI-based design applications is essential to incorporate these factors.

No limitless AI

Limits are important for both the use of AI in CAD and CAE applications and for tools that autonomously create designs—they need clear guidelines and contextual knowledge. While this somewhat curtails the creativity of GenAI implementations, it still leaves enough room to generate theoretically possible designs and processes. For example, AI can help fully exploit new manufacturing methods for components. This way, a truly feasible end product is created with the help of GenAI.

The prerequisite for this is that the AI can consider technical feasibility. To ensure this, it needs extensive knowledge in areas such as materials science, process engineering, and aerodynamics. Ideally, the data should not come from a single manufacturer and be based solely on their research and development. A more holistic knowledge database with information based on the current state of technology provides significantly better results. Based on this, a corresponding AI-based program could design an aircraft that differs significantly from current modes of transport, yet remains functional. GenAI tools should also have the ability to independently continue learning (Widening AI). Technologically, there are already approaches for this, but there is still a lack of appropriate regulations.

The right hardware for engineers

Even though many AI applications are primarily cloud-based, clouds will play a subordinate role in the context of engineering and design for the foreseeable future. The use of cloud servers is rarely cost-efficient enough, as CAD applications often work with very large amounts of data. Additionally, they are based on an established kernel that excludes the distribution of workloads across multiple CPUs, necessitating the use of special workstations. Ideally, these should contain single CPUs with high clock rates of up to 6 GHz—multiple computing cores do not mean more performance in the CAD context.

It's a different story with graphics accelerators, as engineers can never have enough of them: Especially with very large designs composed of many components, they have a positive impact on computation time. Manufacturers like Dell Technologies offer work devices specifically designed for engineers, such as the Precision Workstations. The models of the 3000 series, for example, are specifically tailored for engineering and are available both as stationary towers and mobile laptop versions. They contain a small number of processor cores with very high clock rates and rely on high graphic performance.