CAPP-GPT: A Large Multimodal Model via a Custom Encoder-Decoder Architecture for Macro-Computer Aided Process Planning
Bibliographic record
Abstract
In this article, a Large Multimodal Model (LMM) is developed for the macro-CAPP for Hybrid Manufacturing (HM), which leverages the flexibility of metal-based additive manufacturing and metal cutting. A Generative Pretrained Transformer (GPT) architecture is developed to address the problem at hand. CAPP-GPT uses an architecture with an encoder-decoder layout, where a Part Encoder processes part data and a Plan Decoder generates the plan sequence. The LMM's Part Encoder handles CAD data, converting it into token embeddings with positional information. It then performs geometric feature recognition followed by mapping each geometric feature into its processing steps or features. Using the geometric and processing features as supportive encoder input, the Plan Decoder autoregressively predicts the sequence of operations on processing features, starting with a special token and iterating until the sequence is complete. The developed architecture allows for robust and systematic process planning for hybrid manufacturing. A framework employing a hybrid approach combining Operations Research and Machine Learning (ML) to generate the corpus for the pretraining of GPT has been developed. It can independently create process plans without commercial datasets, using a hybridized heuristic that considers additive and subtractive manufacturing's unique requirements. Constrained clustering, interpretable ML Logical Analysis of Data (LAD), and mathematical programming are integrated to iteratively develop process plans by identifying promising patterns through LAD in an autoregressive manner by rule mining.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.002 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.002 | 0.001 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.008 | 0.002 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".