Skip to content
Fast Company Executive Board

The Fast Company Executive Board is a private, fee-based network of influential leaders, experts, executives, and entrepreneurs who share their insights with our audience.

Accelerating manufacturing innovation with design for automated assembly

Simulated designs perform more like real products and drive a feedback loop of continuous improvement.

Accelerating manufacturing innovation with design for automated assembly

[BMMP Studio / Adobe Stock]

Profile photo of Sviat Dulianinov

The AI server boom is pushing manufacturers to transform how they design and build high-performance hardware systems. Spearheading this shift are advancements in virtual simulation, which bridge the gap between design and manufacturing. By simulating assembly processes before building physical prototypes, manufacturers can catch and fix issues early to prevent the costly delays caused by problems being discovered on the shop floor.

The conventional approach to CPU- and GPU-based server design involves multiple physical iterations and stakeholders reviewing prototype designs after completion. But product simulation before assembly optimizes these prototype cycles through virtual validation of assembly processes. Incorporating Design for Automated Assembly (DFAA) principles serves multiple functions in the manufacturing process. Beyond evaluating designs for robot assembly, DFAA helps teams determine whether their existing automation capabilities can handle design tolerances, while also allowing early automation planning and maximizing design automatability. 

Here’s how to use these principles to accelerate production and speed time to market.

BRIDGING THE GAP BETWEEN DESIGN AND MANUFACTURING

Engineering teams using DFX (Design For X)—including design for manufacturability, assembly, serviceability, and others—face three major challenges today. Firstly, they are limited to analysis through existing 3D-modeling CAD software and the CAD models themselves, which are dimensionally perfect and do not account for real-world tolerances. Also, traditional CAD systems cannot account for specific robots’ capabilities or simulate how these machines would actually perform in the assembly process.

Secondly, the cost of initial prototype builds can be prohibitive, particularly for components like GPUs. Finally, physical prototypes require ordering parts and waiting for delivery before attempting assembly. This creates a long lead time between when a design reaches maturity and when teams can analyze how it is assembled.

Where components come from different suppliers throughout the supply chain, it becomes critical to evaluate designs early for automation compatibility. For instance, it will be necessary to validate whether a power supply unit (PSU) from one supplier will fit in the cage on the chassis of another supplier. These kinds of manufacturing tolerances can result in significant redesign costs.

PLATFORM SOLUTIONS FOR VIRTUAL DESIGN

To address these challenges, manufacturers—including my company, Bright Machines—are turning to solutions like Nvidia’s Omniverse platform and standards like the Universal Scene Description (USD) format for simulation. While CAD designs are virtual, they need to incorporate as many physical properties as possible and Omniverse provides simulation tools that attempt to bridge the “sim-to-real gap.” 

Meanwhile, the USD format—a standardized 3D scene representation—serves as a neutral file format across CAD systems that enables physics simulations, photorealistic rendering, and data synchronization between users. 

In this rich environment, stakeholders with different expertise levels can work with 3D data without needing CAD expertise. Manufacturing engineers find this particularly valuable for complex server and electronics simulations, where Omniverse’s platform-based and open-standard approach enables better interoperability in validating virtual assembly and incorporates digital twins of production equipment in the same environment.

EARLY DESIGN VALIDATION AND CONTINUOUS IMPROVEMENT

Usually within these design disciplines, issues discovered at the prototype stage trigger iterations of design changes. For example, if a robotic gripper needs to pick up and install a heatsink on the circuit board, teams need to verify there’s enough space for the end-of-arm tool to maneuver. Depending on the problem, by using CAD data, manufacturers can simulate these processes and adjust designs before creating physical prototypes. 

Without this capability, changes propagate through the development process, resulting in compounding delays. With about 70% of a new product’s cost determined by initial design choices, and up to 95% locked in by design completion, catching issues early is critical.

DFAA principles help design engineers validate robotic assembly requirements at the outset while considering other design factors such as power consumption, latency, and thermal requirements that may take higher priority. These considerations directly impact automation implementation. For instance, in computer server assembly, Bright Machines’ specialized robotic cells handle process tasks like DIMM insertion into the motherboard when the design takes into account the capabilities and constraints of such automated systems.

The data from these production lines then becomes input for the design engine, driving improvements across the DFX spectrum, alongside other design considerations, making each generation of the product more efficient to produce. Our own microfactories achieve up to 98% first-pass yield—meaning less than two units out of 100 need rework. 

APPLY VIRTUAL SIMULATION TO ROBOTIC AUTOMATION

Before implementing design for automated assembly principles, manufacturers need to evaluate their readiness across several key steps:

  • Assess how changes that occur when issues are discovered during the prototyping phase get propagated through your development process—and where time is spent in the lifecycle. 
  • Evaluate which assembly steps in your current process are the most time-consuming or error-prone, and whether these could be addressed through design modifications or automation.
  • Measure your current process “churn”—how many design iterations typically occur and where do you spend the most time making changes? 

These assessments will help designers identify where virtual simulation will have the greatest impact and validate processes virtually. The result is faster deployment with fewer physical iterations.

THE FUTURE OF AGILE PRODUCT DEVELOPMENT

Virtual simulations behave like their physical counterparts to generate highly accurate replicas as manufacturers shift from relying only on physical prototypes to optimizing designs for automated assembly processes. As manufacturers evaluate their product development lifecycle and data maturity, they will find opportunities to shift critical decisions “left”—earlier in the product development life cycle. 

For companies looking to implement DFX principles in their design process, start with a clear understanding of your current capabilities, then explore how platforms like Omniverse and data formats like USD can transform design data into manufacturing efficiency. Implementing this feedback loop of continuous improvement, from design to production and back, promises to deliver faster innovation cycles and time to market in this era of unprecedented demand.