The Next Frontier of Plastics Injection Moulding Simulation
Introducing Cadmould AI Solver, the world's first neural physics engine for plastic injection moulding. This technology is set to revolutionize design for manufacturing by combining the accuracy of physics engines with the speed of neural networks to deliver simulation results in seconds.
Simulation Today
Wait hours to validate 10–20 design variants.
Simulation Tomorrow with Cadmould AI Solver
Automate the exploration of up to 10,000 to 20,000 alternatives in a single day.
1000x
Shift from Validation to True Optimization
Research Preview Capabilities
This research preview represents the initial deployment of the world's first transformer-based neural physics engine for injection moulding—a foundational step toward covering the full scope of classical simulation.
Unsurpassed Performance
On a mid-tier GPU, the Research Preview typically achieves around 200× speedup over classical numerical simulation. On high-end GPUs, 1,000× and beyond are achievable.
True Geometric Generalization
True Geometric Generalization: The model does not merely memorize data; it learns the physics of flow. None of the geometries showcased in this research preview were part of the training dataset, demonstrating the engine's ability to generalize to unseen topologies with competitive accuracy.
Fully Differentiable Physics
Unlike standard solvers, this model understands the "gradient of change." It does not just calculate a result; it mathematically understands the relationship between input and output, laying the groundwork for hyper-efficient, automated optimization.
Multi-Dimensional Physics
The simulation goes beyond simple filling patterns. It simultaneously calculates shear rates, temperature fields, and pressure distribution in a single run, providing a complete physical picture.
Adaptive Accuracy
The architecture is designed for continuous improvement. If a specific geometry class shows deviation, it can be added to the training set to refine the model, ensuring the system becomes smarter and more precise with every iteration.
Real-Time Interactivity
The speed of the engine enables a new "Cause-and-Effect" workflow. You can adjust critical process parameters—like melt temperature or flow rate—and observe the immediate physical consequences, allowing for intuitive exploration of the process window.
Unsurpassed Performance
On a mid-tier GPU, the Research Preview typically achieves around 200× speedup over classical numerical simulation. On high-end GPUs, 1,000× and beyond are achievable.
True Geometric Generalization
True Geometric Generalization: The model does not merely memorize data; it learns the physics of flow. None of the geometries showcased in this research preview were part of the training dataset, demonstrating the engine's ability to generalize to unseen topologies with competitive accuracy.
Fully Differentiable Physics
Unlike standard solvers, this model understands the "gradient of change." It does not just calculate a result; it mathematically understands the relationship between input and output, laying the groundwork for hyper-efficient, automated optimization.
Multi-Dimensional Physics
The simulation goes beyond simple filling patterns. It simultaneously calculates shear rates, temperature fields, and pressure distribution in a single run, providing a complete physical picture.
Adaptive Accuracy
The architecture is designed for continuous improvement. If a specific geometry class shows deviation, it can be added to the training set to refine the model, ensuring the system becomes smarter and more precise with every iteration.
Real-Time Interactivity
The speed of the engine enables a new "Cause-and-Effect" workflow. You can adjust critical process parameters—like melt temperature or flow rate—and observe the immediate physical consequences, allowing for intuitive exploration of the process window.
Current Constraints
The research preview supports a fixed selection of geometries not seen during training. To benchmark on your own parts, join the Partner Program.
The research preview is limited to the filling phase. It does not yet model cooling or warpage.
Benchmark the AI Solver on your own parts
Cadmould AI Solver Partner Program
As a partner, you get direct access to our development team, benchmark reports on your geometries, and early access to new capabilities.
Benchmarks on Your Parts
Exclusive Insights
Roadmap Influence
From Linear Validation to Real-Time Intuition
Current simulation workflows are often a bottleneck. Because traditional numerical solvers take minutes or hours to compute, the number of design variants engineers can realistically explore remains limited — even when simulation is used early in the process. This disconnect between setting a parameter and seeing the result breaks the creative flow, comparable to shooting on film where you must wait days to see if your settings were correct.
Cadmould's AI Solver removes this latency. By delivering results in seconds, it creates an instant feedback loop. Engineers can adjust a temperature or pressure slider and immediately see the physical consequence on the part. This allows users to actively navigate the process window and identify failure limits in real-time, turning simulation into a true design companion rather than just a validation gate.
Cadmould's AI Solver removes this latency. By delivering results in seconds, it creates an instant feedback loop. Engineers can adjust a temperature or pressure slider and immediately see the physical consequence on the part. This allows users to actively navigate the process window and identify failure limits in real-time, turning simulation into a true design companion rather than just a validation gate.
Join the AI Solver Waitlist to get exclusive Insights
Product Updates: Be among the first to hear about new capabilities and milestones as development progresses.
Behind the Scenes: Receive updates on our progress — from new model versions to engineering insights.
Priority Access: When we open the Cadmould AI Solver to a wider audience, waitlist subscribers are first in line.
How Cadmould's AI Solver works
Unlike generative AI tools that create images based on visual patterns, our AI Solver does not "guess" what a filling pattern looks like. It predicts the physical forces acting on the polymer. Here is the scientific process behind the speed:
Training on Ground Truth
The model learned injection moulding physics from over a million transient simulation trajectories generated by our proven Cadmould Flex solvers, spanning thousands of materials and systematically varied process conditions. This physics-based training foundation ensures that predictions stay grounded in real fluid dynamics and thermodynamics — not pattern guessing.
Geometric Compression and Calculation
When you input a design, the neural network functions as a sophisticated compression engine. It translates complex 3D geometries and process parameters into a compact, abstract representation. In this highly efficient state, the network calculates flow behavior and thermal interactions almost instantly at a fraction of the computational cost of traditional numerical solvers. The model is fully differentiable, allowing for future gradient-based optimization workflows.
Millisecond Decompression
The calculated data is immediately decompressed back into a full 3D simulation result. Rather than solving the filling phase element-by-element over several minutes, the system predicts the complete physical outcome in under a second, delivering accurate pressure, temperature, and shear rate distributions.
A Partnership of Expertise and Innovation
We are bridging the gap between established industrial engineering and cutting-edge artificial intelligence. This research preview is the result of a strategic collaboration between two distinct leaders in their fields, combining decades of domain knowledge in injection moulding simulation with the rapid advancements of modern deep tech.
The company behind Cadmould Flex
With over 35 years of experience in injection moulding simulation, SIMCON provides the physics-based foundation. We contribute the validated "ground truth"—millions of high-fidelity data points and deep domain expertise ensuring that the underlying physics remain accurate and reliable.
Specialists in Large Engineering Model (LEMs)
Emmi AI are experts in the architecture and training of Large Engineering Models — domain-specific transformer based neural networks trained to predict physical behaviour from engineering inputs.
Questions about the Cadmould AI Solver?
We are ready to answer them
The Cadmould AI Solver represents the cutting edge of speed, but you might have immediate production needs today. Whether you are curious about joining the Partner Program or need the proven, high-fidelity validation of Cadmould Flex, let's discuss the right path for your engineering team.
