Researchers at the HSE FCS AI and Digital Science Institute have created CAD2TechSpec, a framework that translates 3D models of mechanical parts into detailed machining process plans for computer-controlled machine tools. The system addresses a significant bottleneck in industries like aircraft manufacturing, where engineers currently spend considerable time manually developing these plans for complex parts such as turbine blades.
CAD2TechSpec first renders each model from 28 different angles, then utilizes RAG, a retrieval system acting as a reference guide, to draw on ISO standards and existing solutions. “Experiments have shown that the quality of the generated process plans depends not only on the model itself but also on the context provided to it,” notes co-author Maxim Minets, Research Assistant at the Laboratory of Methods for Big Data Analysis at HSE FCS.
CAD2TechSpec Framework Converts 3D Models to Machining Process Plans
The CAD2TechSpec framework achieves plan ratings of 86 to 89 points out of 100, according to expert evaluations of its generated machining instructions; this performance suggests a high degree of accuracy in translating 3D designs into actionable manufacturing steps. This level of detail is critical for industries demanding precision, as the system’s output directly informs the operation of CNC machines responsible for creating complex parts.
Error analysis indicates further gains are possible with expanded training datasets, demonstrating a clear path for refinement and increased reliability. This design choice acknowledges the critical need for oversight in manufacturing processes, particularly when dealing with high-value components. The system’s application extends to precision engineering and aircraft manufacturing, sectors where extensive technical documentation traditionally consumes significant engineering time.
The framework’s ability to generate process plans stems from its multimodal approach, combining visual data with textual prompts, HSE FCS says. This allows the AI to not only “see” the part’s geometry but also “understand” the intended manufacturing process. CAD2TechSpec differs from purely automated systems by prioritizing efficiency in preparation, rather than complete autonomy, a pragmatic approach to implementation in complex industrial settings.
RAG and ISO Standards Enhance Neural Network Accuracy for CNC Machining
The CAD2TechSpec framework achieves heightened accuracy in generating machining process plans through the integration of Retrieval-Augmented Generation, or RAG, and adherence to International Organization for Standardization (ISO) guidelines. RAG functions as an external knowledge base, supplementing the neural network’s internal data with validated solutions and established manufacturing protocols. This approach moves beyond purely generative AI, allowing the system to reference existing, proven methods when applicable, rather than always creating plans from scratch, according to HSE FCS.
To assess performance, the system processed 1,236 models sourced from an open database, with evaluations conducted by both automated systems and human mechanical engineering experts. The evaluation criteria focused on the structural correctness of the output, the precision of the proposed manufacturing process relative to the part’s geometry, and the appropriate application of ISO standards. Results indicated the Mistral Large 3 model, when paired with the ISO knowledge base, delivered the strongest performance across these metrics.
The incorporation of ISO standards is not merely a compliance measure; it directly influences the quality of the generated plans, highlighting the importance of external validation.
See today’s quantum computing news on Quantum Zeitgeist for the latest breakthroughs in qubits, hardware, algorithms, and industry deals.
