The Challenge: Design Freedom Creates Complexity
nTop’s implicit modeling engine enables parametric designs such as lattices, ribbed structures, organic shapes, and computational design workflows that break traditional CAD tools. But this design freedom comes with a challenge: an exponentially larger design space.
Consider a simple ribbed panel with four parameters: arc count, cell length, unit cell type, and height. Even with modest ranges, you’re looking at thousands of possible combinations. Traditional approaches fall short:
- Manual iteration is slow and biased toward familiar solutions
- Brute-force DOE (grid search) is computationally prohibitive
- Random sampling wastes compute on uninteresting regions
Engineers need a smarter way to navigate this space. One that learns from each simulation and focuses on the designs that matter.
The Solution: AI-Guided Exploration
Bayesian optimization offers a fundamentally different approach. Instead of exhaustively sampling the design space, it builds a surrogate model of the objective function and intelligently selects the next design to evaluate. Each simulation informs the next, rapidly converging toward optimal regions.
The key ingredients:
- Intact.Simulation for nTop: Meshfree FEA that runs directly on implicit geometry, eliminating the meshing bottleneck that breaks automation
- nTop Automate: Headless execution of parametric nTop workflows
- PyIntact: Headless execution of automated simulation at scale
- Python ML tools: Bayesian optimization libraries that guide the search
Together, these tools create a closed-loop system: the optimizer proposes a design, nTop generates the geometry, Intact.Simulation evaluates performance, and the results feed back to refine the next proposal.
Figure 1: AI-guided design exploration workflow. The user defines the initial parameters and ranges, optimizer proposes parameters, nTop generates geometry, Intact.Simulation evaluates performance, and results inform the next iteration.
Case Study: Multi-Objective Ribbed Panel Optimization
To demonstrate this workflow, we optimized a ribbed panel structure with two competing objectives:
- Minimize maximum displacement (stiffness)
- Minimize additional mass (weight)
Design Variables
These parameters control the ribbed pattern on the panel, creating a rich design space with diverse structural behaviors.
Figure 2: Sample designs from the parameter sweep showing the diversity of ribbed panel configurations.
The Optimization Script
The workflow is remarkably concise. Here’s the core objective function:
The optimizer handles the rest—selecting the next design, balancing exploration vs. exploitation, and building the Pareto front.
Results: A Complete Pareto Front in 300 Trials
After 300 trials, the optimizer discovered a well-defined Pareto front spanning the full trade-off between stiffness and weight.
Figure 3: Pareto front showing the trade-off between maximum displacement and mass. Each point represents a validated design.
Key observations:
- Clear trade-off curve: The Pareto front reveals the fundamental relationship between mass and stiffness for this design family
- No manual meshing: Every simulation ran automatically on the implicit geometry
- Every point is validated: Unlike surrogate-only approaches, each Pareto-optimal design has been fully simulated
Efficiency: Convergence in Hours, Not Weeks
The progression chart reveals how quickly the optimizer finds good solutions:
Figure 4: Running minimum of each objective over trials. The optimizer rapidly discovers high-performing regions within the first 50–100 trials.
The study completed in approximately 12 hours on a standard workstation—an average of ~2.5 minutes per trial including geometry generation, simulation, and post-processing. A comparable brute-force grid search would require 10,000+ evaluations to achieve similar Pareto coverage.
Interactive Exploration
Results are only valuable if engineers can explore them. We built an interactive dashboard using Panel that connects the Pareto front to the underlying designs:
Figure 5: Interactive dashboard for exploring optimization results. Clicking a point on the Pareto front displays the corresponding geometry and design parameters.
This click-through capability transforms abstract data points into tangible design insights. Engineers can:
- Identify designs at specific trade-off points
- Compare parameter combinations across the Pareto front
- Export promising candidates for further refinement
The Bigger Picture
This ribbed panel study is a proof of concept, but the implications extend far beyond a single geometry. The same workflow applies to:
- Lattice structures for lightweighting
- Heat exchangers with thermal-structural coupling
- Brackets and mounts with multi-load requirements
- Any parametric nTop design where simulation guides decisions
The combination of meshfree simulation, implicit geometry, and AI-guided optimization represents a new paradigm for computational design. Instead of manually iterating toward a solution, engineers define the design space and let the system discover optimal regions automatically.
Intact.Simulation for nTop and PyIntact removes the meshing bottleneck that has historically prevented this kind of automation. When simulation is as simple as a Python function call, it becomes a building block for intelligent design systems.
Get Started
Ready to explore AI-guided design exploration for your workflows?
- Intact.Simulation for nTop: Learn about our nTop integration
- PyIntact Documentation: Explore the Python API
- Contact Us: Discuss your application