GERMANY VIDEO REVIEW · KNOWLEDGE REUSE
Teaching AI to Program with Your Own History: A Walkthrough of DELMIA's Knowledge-Reuse CAM Video
01What this demo video is about
marista is a German-speaking (Swiss) partner of Dassault Systèmes. This demo video, released in June 2026, is titled "CAM-Programmierung mit KI (CAM programming with AI)" and demonstrates DELMIA Machining[1].
The description lays out the flow unusually clearly[1]: taking a complex workpiece as an example, the software uses manufacturing knowledge already present in past projects, recognizes similar machining features, and from that automatically generates machining strategies, tools and NC operations; the resulting manufacturing process is then simulated and verified. The target figure: through the intelligent reuse of manufacturing knowledge, CAM programming is made up to 70% faster—the vendor's claim, flagged as usual when cited.
02Breaking down the flow: from historical knowledge to NC operations
Break the description's flow down step by step, and each step has a clear theoretical counterpart:
- Feature recognition: Recognizing the holes, slots and faces on the new workpiece—deep-learning feature recognition is already a mature capability[3].
- Similarity matching: Finding "similar features you've done before" in the historical project library—this step is the core of the knowledge-reuse route, corresponding to the most data-efficient analogical reasoning in the machine-learning literature[4].
- Solution transfer: Transferring the historical job's strategy, tools and parameters onto the new feature to generate NC operations—the CAPP literature calls this variant planning, as opposed to generative planning[2].
- Simulation verification: The transferred solution still has to pass simulation—because "similar" isn't "identical," and verification can't be skipped.
Sharp-eyed readers will notice: this is exactly the convergence of ideas that has run through the whole Germany series—OPEN MIND's similar-workpiece search, CINTEG's part-family method, Kawasaki Heavy Industries turning drawings into assets—and DELMIA has built it into an end-to-end product flow.
03Two AI CAM routes: knowledge reuse vs. from-scratch generation
| Knowledge reuse (like DELMIA) | From-scratch generation (like CAM Assist) | |
|---|---|---|
| Knowledge source | Your shop's historical projects | A generic model trained by the vendor |
| Strong scenario | Shops where similar parts recur | New part types, parts with no history to check |
| Cold start | Needs an existing digitized case library | Usable on day one |
| Result style | Close to your shop's conventions | Generic best practice |
| Shared precondition | The output must pass simulation and verification, with a human doing the final gate-keeping | |
The two routes aren't opposites but a spectrum: in practice the best systems mix them—reuse when there's history to follow, generate when there isn't, all passing the same verification. To judge which end your shop leans toward, one number is enough: over the past year, what share of new jobs could find a 70%-alike reference in your historical cases?
04A reading framework and preconditions for Taiwanese shops
Two takeaways this video offers Taiwanese shops:
- Ask clearly about the route during selection: When a vendor says "AI programming," first ask whether it reuses your knowledge or generates a generic solution, whose hands the knowledge accumulates in, and what the cold start needs. The two routes have completely different adoption tempos and long-term value.
- Whichever route, the precondition is a digitized case library: Knowledge reuse needs historical jobs to check, and even the output of from-scratch generation should be stored back as a future reference—the accumulation of case data is the shared foundation of both routes.
This is exactly the intent behind BestAI CAM's flow design: while every job runs through "AI reads the 2D drawing → builds the 3D model → generates code according to the shop's tool library and machine limits → cutting simulation → independent AI dimensional cross-verification," the drawing, operations, program and verification records are automatically retained as a structured case library (see The Complete Guide to CNC Automated Programming)—today it makes the prep of every new job faster, and accumulated, it becomes your shop's own "reusable manufacturing knowledge." The future the DELMIA video demonstrates has its foundation laid from the very first digitized drawing.
05FAQ
Is "programming up to 70% faster" credible?
This is the vendor's claim in a demo video, and it comes with a clear precondition: the benefit comes from "knowledge reuse of similar features"—the richer your historical case library and the higher the similarity between new jobs and history, the more obvious the acceleration; brand-new part types don't enjoy this dividend. The real number for your shop is what you get by testing on your own mix of parts before adopting.
How do I choose between the knowledge-reuse route and generative AI CAM?
It depends on your order structure: for a shop where similar parts recur (part families, lots of variants), knowledge reuse has high long-term value; for a shop with a high proportion of new part types, the cold-start advantage of the generative route matters. The two can be mixed, and their shared precondition is that the output must pass simulation and verification and that case data must keep being digitized and accumulated.
Our shop's historical jobs are all in folders and in the masters' heads—how do we start?
You don't have to do a big migration first—start accumulating from today's new jobs: run every job through the AI flow of reading the drawing, modeling, generating code, simulating and verifying, and the case data is structured and retained automatically. In half a year you'll have a comparable, reusable digital case library, and the knowledge-reuse capability will connect to it naturally.
We'll let you know when new articles and video walkthroughs go live—no inbox flooding, one-click unsubscribe. (newsletter in Chinese)
YOUR HISTORY, YOUR AI
Starting from your first digitized drawing, build your own manufacturing knowledge base
AI reads the drawing, generates code and verifies, and every job leaves behind structured case data—speeding up prep today, becoming a reusable knowledge asset tomorrow.
Contact an onboarding consultant Training courses📋 Free download: "CNC Outsourcing Checklist" (printable) (中文)
06References
- marista ag (YouTube). CAM-Programmierung mit KI (published 2026-06-24, DELMIA Machining demo). youtube.com/watch?v=EgE5vksRKCs
- Xu, X., Wang, L., & Newman, S. T. (2011). Computer-aided process planning — A critical review of recent developments and future trends. International Journal of Computer Integrated Manufacturing, 24(1), 1–31.
- Zhang, Z., Jaiswal, P., & Rai, R. (2018). FeatureNet: Machining feature recognition based on 3D Convolutional Neural Network. Computer-Aided Design, 101, 12–22.
- Wuest, T., Weimer, D., Irgens, C., & Thoben, K.-D. (2016). Machine learning in manufacturing: advantages, challenges, and applications. Production & Manufacturing Research, 4(1), 23–45.