Semiconductor equipment
Micius learns three engineering tools, then builds one workbench that models, costs and renders
Talk to an engineer
- 3 tools
- One run: Micius first does a typical task in a 3D CAD tool, a renderer and a spreadsheet
- 174 parts
- One run: a mechanical keyboard built part by part from a blank project, alongside a 9-part desktop robot arm
- 1 change
- Widen the upper arm by 10 mm and geometry, mass (0.9202 → 1.0066 kg), cost and render update together
A designer working on a desktop robot arm usually needs three tools: 3D CAD for the model, a spreadsheet for mass and material cost, and a renderer for the image. Every dimension change has to be carried through all three by hand. Micius, Zhiman’s model, learned how each tool works, then wrote a single workbench where a change updates everything at once.
Challenge
One change, three tools
A desktop robot arm has a base, a shoulder, an upper arm, a forearm, a wrist and two grippers, and each part needs a size and a material. The designer then has to answer three questions, usually in three different tools. 3D CAD builds the solids. A spreadsheet takes each part’s volume, density and unit price and works out mass and material cost. A renderer applies materials and lights and produces the image.
Change is where this breaks down. Widen the upper arm by 10 mm in CAD and its volume changes, so the spreadsheet has to be recalculated and the model re-imported and rendered again. The numbers move between the tools by hand. Miss a step and the weight and cost you report no longer match the model. How much one person gets through in a day depends largely on how fluently they work these tools.
Workbench
Learning the tools, then writing one workbench
Micius, Zhiman’s model, started by using each tool the way an engineer would: opening it, finding the menu, filling in values. In the CAD tool it changed a fillet radius on a solid’s edges from 1.0 mm to 1.5 mm. In the renderer it set three area lights on the robot arm to 10 W, 5 W and 15 W. In the spreadsheet, which lists each part’s material, volume, density and unit price, it cut the upper arm’s material price from ¥32 to ¥14 per kilogram and watched the cost column follow.
The point was to learn what each tool actually does. A fillet is a feature applied to edges; a light has a type, colour, power and shape; mass is volume times density, and cost is mass times unit price. With that understood, Micius wrote a workbench that combines those functions.
The workbench keeps geometry, materials and cost, and the rendered scene in a single project, in millimetres, kilograms and yuan. A part tree lists every part. Select one and you can set its dimensions, add a through-hole, repeat it as an array, or round its edges, then pick anodised aluminium, brushed steel or engineering plastic. Each part attaches to a parent at a set position and angle, so the assembly builds up level by level. The 3D view has an exploded-view slider; a bar along the bottom keeps a running total of parts, mass and material cost; and the bill of materials lists quantity, material, volume, density, mass and cost for every part. The render comes out of the same project.
Modelling
Two products from a blank project
Micius then built two products in the workbench, one part at a time. The robot arm has 9 parts in steel, aluminium and plastic, each attached to the one above it, with mass and cost in the bottom bar updating as each part goes in. The mechanical keyboard is far busier: a base plate, side walls, a locating grid and rows of keycaps, 174 parts in all. Identical parts went in as arrays, a whole row at a time.
From the finished geometry, the bill of materials put the arm at 0.9202 kg and ¥17.33, and the keyboard at 0.8482 kg and ¥18.96. These are material costs only, at sample unit prices set in the project, not quotes.
Results
A dimension changes, and everything follows
Then came the test this started with. Micius widened the arm’s upper arm from 22 mm to 32 mm. Total mass went from 0.9202 kg to 1.0066 kg, material cost from ¥17.33 to ¥19.92, and the render was redone at the new size. On the keyboard, raising the fifth row of keycaps from 8 mm to 14 mm took mass from 0.8482 kg to 0.8801 kg and cost from ¥18.96 to ¥19.73.
Illustrative flowThe same change: carried between three tools before, updated in one place here.
The figures can be checked by hand. The upper arm is 20 mm deep and 160 mm tall, so 10 mm more width adds 32,000 mm³. At 2.7 g/cm³ for anodised aluminium that is 0.0864 kg, exactly the increase in total mass. Every part’s volume, density and price is in the bill of materials, so an engineer can check any line.
The results have clear limits. Both products are made of boxes, cylinders, holes, arrays and fillets, with no complex surfaces. Mass is nominal volume times density; cost covers material only; the render is a preview, not an engineering drawing. The workbench was written for these two products and doesn’t replace a team’s professional tools. Where a team’s own tools are already installed, a task can also be described to Micius and worked through across them.
Micius did the tool work and the coding. Which dimensions to change, which materials and prices to use, and whether to trust the numbers stay with the engineer.
The same approach also works for
-
3D models of equipment structures
Build a structure part by part from its description and give each part’s mass and material cost for an engineer to check.
-
Comparing materials
Try one structure in several materials, cost each from the team’s own price list, and set them side by side for an engineer to choose.
-
Running simulation tools
Set up, run and export a CAE or other simulation from an engineer’s conditions, and say which steps were done and which weren’t.
-
Small tools for repeated hand-offs
Where the same data keeps moving between tools, learn how each one works, then write a small tool that brings those steps together.
Do your design changes still go through several tools?
Tell us about a typical part, the tools you use and what you need to work out. We’ll walk through it with you and show what Micius takes on and what your engineers check.
Related use cases
Predictive maintenance
Reads sensor and maintenance data to flag a tool before it fails, with engineers deciding on the repair.
AOI inspection for packaging
Calls defects on site, next to the inspection tool, and passes the uncertain ones to a person.
Equipment assembly SOPs
Reads an assembly drawing, extracts the BOM and drafts step-by-step assembly instructions for people to sign off.
Industrial software development
Describe the engineering tool you need in a paragraph; Micius breaks it into capabilities, writes a working tool and runs an example through it.
