Inside the world of Gen AI and copilots in manufacturing

Artificial intelligence (AI) influences nearly every aspect of our day-to-day lives. It helps us generate communications, summarize documents, engineer code, translate languages—and much more. It’s difficult to overestimate the widespread impact AI has had, and it’s hard to find areas of life untouched by this disruption—including manufacturing.

As resistant as some people may be, generative AI (Gen AI) is reshaping the way people and software interact—and companies that fail to adapt may risk falling behind their competition.

Emerging Gen AI and copilot tools are enhancing engineering tasks from design to manufacturing. In design, generative capabilities in CAD systems are evolving, while on the shop floor, AI is improving robotics and machine building. The most exciting development is AI driven by deep learning algorithms and neural networks, capable of synthesizing complex data into useful engineering information.

And the industry has been eager to show off its advances.

Innovations in Gen AI for manufacturers and engineers were a key trend at this year’s Hannover Messe, where providers showed off their new copilots and AI assistants integrated into new and existing technologies to support the role of the engineer.

According to a recent report from ABI Research, the added revenue from Gen AI use cases in manufacturing will be US$10.5 billion by 2034. As well, ABI found that the overall sentiment toward deploying Gen AI solutions in manufacturing is to take a slow-and-steady approach. To achieve this goal of steady progression, Gen AI suppliers are listening to the day-to-day problems faced by manufacturers and are working backward on how AI can be a solution.

“The current focus of vendors is to make Gen AI available and useful to as many employees as possible, regardless of job title or function,” explains James Iversen, industrial and manufacturing industry Analyst at ABI Research. “This holistic approach is supported by the industry trend of free or freemium Gen AI models. Vendors are more focused on garnering users than profitability at this moment.”

In early 2024, IT consulting provider SoftServe commissioned a global study to evaluate the effectiveness of organizations’ current Gen AI adoption strategies. The study revealed that while organizations continue to invest and pursue many use cases in search of ones that will deliver the biggest impact, they struggle with internal data readiness, governance and skill development.

Of the many insights in the survey, only 42 per cent of organizations have the capabilities to train Gen AI models—and a staggering 89 per cent face difficulties preparing data for Gen AI use.

While organizations are trying to keep up with emerging AI tools, many are struggling to find the right use cases and build up the foundation of data necessary to leverage AI effectively. Design Engineering spoke with four organizations to uncover pragmatic use cases for AI in manufacturing—and to look ahead at the potential that AI has to revolutionize the role of the engineer.

Explore all sections of the feature:

• AI assistance and code generation
• Shortening design-to-manufacturing timeline
• Streamlining the design process
• Equipment health checks in real time

Beckhoff’s TwinCAT Chat harnesses AI-supported engineering with automated code optimization, code documentation and restructuring. (Credit: Beckhoff)

AI assistance and code generation

Andy Burleigh, engineering manager at Beckhoff Canada, highlighted that many AI-focused technology showcases and product announcements are still in the early stages.

“AI-assisted tools in automation are in their infancy,” he says. “But we do see that our customers are already using tools like ChatGPT to assist them in many aspects of their work to become more efficient—from research, to design and even code development. I would say the current use-cases are explorative, and it is typically younger engineers that are rushing to embrace this technology.”

Burleigh detailed TwinCAT Chat, a large language model (LLM) integrated with the TwinCAT XAE engineering environment. “TwinCAT Chat is trained on automation generally, and Beckhoff products specifically,” he explains. This tool, accessible within the TwinCAT IDE, supports engineers by providing instant answers and generating code, HMI configurations and hardware setups. “Code objects created by TwinCAT Chat can be reviewed in the Chat window and added to the project with a click.”

“The response has been very positive,” he noted, referencing the interest generated since Beckhoff’s technology showcase in early 2023. Engineers appreciate AI tools that are easily accessible and contextually sensitive within familiar development environments. “Code generation is a primary theme in the feedback we’ve received.”

Discussing AI’s complement to engineering, Burleigh emphasizes productivity enhancements. “At a basic level, AI-driven documentation searches will simplify access to information and increase the pace of learning. But beyond that, tools which create and optimize automation solutions—whether that is hardware configurations, algorithms or code—will increase the output and quality of that output, of individual engineers.”

He notes how engineering is really an ideal use case for AI technologies because engineering is an act of creation. “AI tools can be used very effectively to allow a creator to achieve their vision more quickly.” But he says one key challenge will be in validating the results of AI-assisted engineering. “Can the results be tested using existing methods, or will new ones be required?” Additionally, security and privacy concerns need addressing, similar to other technology sectors.

Looking ahead, Burleigh expects rapid improvements in LLM training. “Every automation software developer will benefit from virtual assistants creating code, HMI layouts or performing research,” he predicts. Future LLMs will recognize efficient design patterns, reducing development time and producing more robust solutions, thus continuously optimizing automation systems.

InfinitForm supports CNC part design, assuming the above five-axis machining directions and tool diameters. The ghosted part is the allowable design space. (Credit: InfinitForm)

Shortening design-to-manufacturing timeline

Incorporating AI into the product development processes may be a complicated endeavour—but it’s an inevitable one. 

InfinitForm founder and CEO Michael Bogomolny has jumped into the market, eager to shake up the design engineering process. “The CAD industry hasn’t really had any real innovation for decades,” he says. “Most of them do all the same things.” Current CAD systems still rely heavily on manual drafting—and a lot of trial and error when it comes to manufacturability. What’s more, the existing design-for-manufacturing process is challenged by silos, intricate workflows and time-intensive procedures. This process can take weeks to months, involve a multitude of expensive software tools, workflows, and require highly skilled personnel with cumbersome coordination.

The current design engineering process involves creating shapes while considering manufacturing constraints, often without knowing all the details in the early stages.

Design engineers draft initial shapes, iterating between simulation and design before undergoing a manufacturing review. This process typically involves three to four engineers and can take weeks or months, depending on the complexity. CAD and modeling tools are used manually to create these shapes, making it challenging for engineers to design optimal parts right away, especially when applying the right manufacturing constraints and materials from the start.

“It’s difficult for a human to design an optimal part right away, to be able to apply the right manufacturing constrains and materials right away,” Bogomolny says. The iterative nature of this process highlights the need for more efficient solutions to streamline design and manufacturing integration.

According to Bogomolny, it’s time to integrate assistance and automation into the design process, leveraging Gen AI to enhance efficiency and innovation. His company, InfinitForm, conducted market research, identifying manufacturers’ top three priorities as speed of innovation, part cost and manufacturing cost. He envisioned a platform that integrated all these elements effectively.

With tools like InfinitForm, the goal is also to integrate its AI copilot seamlessly into existing software environments, allowing engineers to work smoothly without needing to switch platforms. From concept to production, users benefit from advanced Gen AI design, simulation and optimization tools, streamlined collaboration features, and an intuitive user interface.

Integrating Gen AI into the design process can reduce the tedious parts of design work. AI can help in designing better assemblies of different parts, enabling engineers to create concepts without getting bogged down in detailed work or other lower-level design tasks.

Bogomolny details how AI offers two primary layers of assistance.

Firstly, AI can help set up the problem correctly, ensuring that initial design parameters are accurately defined. And secondly, during the design generation process, AI can produce machinable and editable products. This approach helps make sure the final products are conceptually sound and ready for manufacturing. Generative design has the potential to overcome human biases, often creating models that meet all specifications in unexpected ways—resulting in stronger and lighter parts.

Algorithms can ensure manufacturability from the start, taking constraints into account and optimizing designs for production. This automation allows engineers to concentrate on the more innovative aspects of their projects, improving overall productivity and innovation in the design process.

For the design engineering industry to innovate, adopting new technologies is crucial, despite the strict regulations there may be in sectors like aerospace, automotive and defence. Integration into existing platforms is key to ensuring a smooth transition.

Siemens’ Industrial Copilot shares relevant information on the live status of a robotic arm and other operations in progress. (Credit: Siemens)

Streamlining the design process

Gen AI is revolutionizing various industries, and Siemens is at the forefront with its Siemens Industrial Copilot—designed to enhance human-machine collaboration and support the role of engineers. The Siemens Industrial Copilot aims to boost productivity and efficiency by automating specific tasks through Gen AI.

Gorve Rekhi, national business development with Digital Enterprise at Siemens, says their Industrial Copilot currently supports two main domains: engineering and operations. In engineering, its capabilities include PLC code generation, human-machine interface visualization generation and providing answers to automation engineering-related questions.

For operations, the copilot assists with predictive maintenance by generating responses to user questions related to operations and maintenance, which is integrated with their predictive maintenance tool, Senseye. When Senseye triggers a maintenance notice, users can inquire about the root cause, service instructions and required tools.

“In the future, Siemens envisions a suite of generative AI-powered industrial copilots across the entire value chain—design, planning, engineering, operations and services,” Rekhi says. “These features will work across multiple industries, including automotive, infrastructure, transportation and health care.” And as manufacturers accumulate more data, the potential for these copilots to improve and expand their capabilities will grow.

In the realm of design, the Industrial Copilot aims to streamline the design process by assisting with brainstorming and generating mechanical design concepts based on specific criteria like performance, materials and manufacturing methods.

Rekhi explains that the copilot introduces the possibility of generating innovative designs that may not have been considered by humans, aiming to create high-performance, lightweight, sustainable parts that meet critical criteria on an accelerated timeline. Additionally, it aids in creating initial prototypes and mock-ups, providing design variations for faster iteration and experimentation.

SoftServe’s Industrial Copilot uses large language models of manufacturing processes, manuals and equipment to provide workers with operations support, advice and guidance. (Credit: SoftServe)

Equipment health checks in real time

Vasyl Mykhalchuk, senior R&D engineer at SoftServe, emphasizes that AI advancements are not merely hype but practical tools delivering measurable benefits today. He notes, “AI plays a crucial role in optimizing weight, cost and performance designs.” For instance, topology optimization tools refine structures to use less material while maintaining strength, which is essential for industries like aerospace and automotive.

Integrated AI assistants in CAD software, such as SolidWorks’ XDesign, automate routine tasks, provide suggestions and detect potential errors early. This not only enhances productivity but also reduces costly design mistakes. Mykhalchuk highlights real-world applications, mentioning that Airbus uses AI for generative design to develop new aircraft components, reducing weight and improving fuel efficiency. Collaborative robots, or cobots, from companies like Universal Robots, use AI to learn tasks and interact safely with human workers, making manufacturing processes more flexible and efficient. “AI is indispensable in robotics and automation,” he says, “and optimizes supply chains by predicting demand, managing inventory and identifying bottlenecks.”

SoftServe’s Gen AI Industrial Copilot, powered by Gen AI technologies and the NVIDIA NeMo framework, offers precise insights into maintenance processes. “It automates equipment health checks and aids in preventing malfunctions by analyzing historical KPIs,” Mykhalchuk explains. The integration of the NeMo framework allows the copilot to process complex queries quickly and accurately, enhancing decision-making in maintenance operations.

At its heart is its extensive knowledge base, built from detailed equipment information such as operation and maintenance manuals. This enables meticulous searches and the provision of contextualized, accurate answers based on the retrieval-augmented generation (RAG) approach. “The RAG approach significantly speeds up the search process and reduces the reliance on manual efforts, providing relevant and actionable insights.”

The Industrial Copilot is a game-changer for troubleshooting and planning by leveraging real-time IoT equipment data. “It continuously collects and analyzes data from sensors embedded in machinery and equipment, allowing for immediate detection of anomalies and potential issues,” Mykhalchuk says. Understanding equipment performance patterns and predicting future needs helps schedule maintenance during non-peak times, ensuring resources are used efficiently.

The use of dynamic 3D models and digital twins by SoftServe’s Industrial Copilot significantly improves production process visualization and problem-solving. Digital twins create precise digital replicas of physical assets, allowing for real-time monitoring and analysis. “To achieve photorealistic visualization of the equipment within the SoftServe’s Industrial Copilot, we utilize the cutting-edge NVIDIA RTX renderer bundled in NVIDIA Omniverse,” he says. “Furthermore, we harness the power of OpenUSD to seamlessly integrate data and 3D assets from various sources, CAD formats and tools into a cohesive 3D scene.”

When integrated with real-time data, engineers can simulate different scenarios to predict outcomes and test solutions without disrupting actual production. “This capability significantly improves troubleshooting, as potential problems can be identified and addressed before they affect the physical equipment,” Mykhalchuk says.