Running a program in the twin

_Draft (round 14): the text is being written; the sections may still grow._

What the twin is

The machine in the window is a digital twin of the MC-1: the Twinwright controller (kernel) runs in it, the same code as on the machine, and a physical model stands for the motors, the drives, the spindle, the switches and the cut. A program runs in the twin as it would on the machine, and the twin cuts its own workpiece. Every number it gives about the physical world is an estimate.

Before you run

The twin checks a program against the setup you give it, so set it up as the real machine is:

• the machine (File > Switch machine…);
• the blank and how it is held (the Setup workspace);
• the tools (the dock's Tool table);
• the work zero (Machine > Work zero).

File > Example parts… sets all of these and loads a program in one go.

Run a program

1. Open the program: File > Open program…
2. Check it first: Program > Check program runs it through the controller's check mode, so the controller reads every line and the machine does not move.
3. Run it: Cycle start (F5). Feed hold (F6) pauses it, Cycle start goes on, Stop (F7) stops it. Single block runs it line by line, Next block runs the next line.

The twin runs in real time unless you choose otherwise: Simulation > Speed runs it from half to 200 times real time, or Fastest - as fast as this computer can - and Simulation > Pause the simulation stops the twin's time (the machine's state does not change while it stands). A click on the clock at the right of the machine bar opens the same list.

From ×100, at Fastest and while computing to the end the twin is Lite (the tag beside the clock): the chatter check and the forces may come a little later to the curves, the picture uses coarser cells and a thinner trail. What the program does - the material removed, collisions, soft limits, alarms - is judged as at ×1, and the result and report are the same. Settings > General > Full precision at high speed too (slower) turns Lite off.

When the computer cannot keep up, the clock says how fast it really runs, in amber; its tip names the busiest part of the twin (for example the material removal on fine cells) and what to do.

Simulation > Speed > Compute to the end (no playback) runs the loaded program to its end without showing it frame by frame: the machine bar shows how far it is and the time left, the 3D view stays at the start and changes when it is done, and the Analysis workspace's last run is filled in as usual. Stop beside the program stops it where it is. Connected to a real machine (shadow mode) only ×1 is possible: the twin follows the machine's time.

What the twin checks

What follows from the program and the setup's geometry:

• the controller's own alarms and errors, as the machine's controller gives them: soft limits, limit switches, a probe move that touches nothing;
• collisions of the tool holder, the collet nut and the spindle with the blank, the fixture and the table, and near misses;
• cutting in a rapid move (G0), and cutting while the spindle stands;
• cutting through the blank's bottom into the fixture or the table;
• a tool that would jam or break: a feed per tooth too high for the tool, a bending stress at the flutes' root over what the tool takes;
• the Z clearance: whether each tool, and the touch probe, clears the setup with Z at the top (see MC-1 hardware);
• with the part's design in the setup: material left on the part and cuts too deep (View > Colouring > Workpiece: deviation from the part).

The findings appear in the dock's Messages tab as they happen.

What the twin estimates

What the physical model works out:

• the cutting forces, torque and power, and the spindle's load: the spindle slows under load as a real one does, and too much load trips its drive;
• how the tool and the machine give way under the cut, in the shape cut;
• chatter: the margin along the path (View > Colouring > Trail: chatter margin), and the spindle speeds where the cut is stable (the Analysis workspace: Work out the stability lobes);
• the cutter's runout: slots and holes come out wider by it;
• the machine time, from the controller's own planning and the drives' motion;
• the cusps a ball-nose cutter leaves between its passes (their shape, not the surface's roughness).

The trail can be coloured by the feed rate, the cutting force or the chip thickness as well (View > Colouring), and the dock's Curves tab plots the run over time.

What the twin does not model

• the surface's texture and roughness;
• built-up edge, chips and getting them out;
• thin walls bending under the cut;
• the part's own heat;
• tool wear: the cut is always a sharp tool's. The twin keeps an estimate of each tool's wear - how long its edges were in the material, how far they cut, the share of an estimated life used (Taylor's law; most materials' figures are estimates) - and reports it (machine_status's tool_wear over MCP), but it does not change the cut; a new tool of the same spec starts again (setup_tool with new). Tools that cut aluminium are mostly changed first because aluminium sticks to the edge and the part shows burrs or a poorer finish;
• what happens after a tool breaks;
• details finer than the program check's grid: before anything runs, the program check looks at the workpiece in cells of about 0.1 mm, finer when the program has a tool under 2 mm across: a twentieth of its diameter, down to 0.05 mm (coarser on blanks over 100 × 100 mm: a million cells at most, four million with such a tool), so a rapid through a wall thinner than that, or a tool that grazes a wall by less, may pass the check unnoticed. Run the program in the twin to see it: the twin's own cut works in 0.02 mm cells.

Plastics and micro tools

Cutting PMMA (acrylic) with tools of 0.1–0.5 mm - microfluidic channels, for example - the twin also leaves out:

• the cut's temperature: PMMA softens towards its glass transition (about 105–115 °C), and from about 70 °C its chips weld and smear on the tool and in the channel. Keep the chip load and the speed in the range the tool's maker gives;
• the air jet or the coolant, and what it does to the heat and the chips;
• burrs along the edges of channels and pockets;
• thin walls: in PMMA, walls under about 0.15 mm thick bend or break under the cut; the twin cuts the shape the tool's path and its give leave, not the wall's own deflection;
• the damping of the cut itself: a tool this small bends at tens of kilohertz and cuts slowly, and the tool's flank rubbing the cut surface damps its vibration. The twin leaves this out, so its chatter margin for such tools is on the safe side.

The models and what they assume: Twin models.

The results and the report

• The workpiece in the 3D view is what the twin cut; its colouring shows the cut's depth or, with the part's design, the deviation from it.
• The dock's Messages tab lists the checks' findings; its Curves tab the forces, the load and the speeds over time. File > Export curves (CSV)… saves them.
• An AI client reads the same results through Studio's MCP tools: see the MCP tools reference.