PAI-150 · Education

Manufacturing Physical AI

Take a part from a clean CAD model to a humming, optimized production line inside a browser-native virtual-factory twin: deciding how to make it, generating its process plan, taming tolerance and variation, balancing the line, and optimizing a whole factory against throughput, cost, and yield.

Beginner → Skilled·5 modules·15 lessons
▶ Start the course ← All courses
THE HORIZON

Where this sits, and what moves it.

Binding constraint · Variation. Every process has a distribution, not a value, and yield is what happens when a tolerance stack meets that distribution.

Was impossible

Connecting a tolerance decision to a yield number meant building the line and counting the scrap. The loop from a design choice to its cost was measured in months and capital, which put it out of reach of anyone learning.

Is probable

A virtual factory closes that loop in a browser in seconds, and the physics of the stack-up is exact. What a twin does not give you is the process distribution itself -- that still comes from a real machine on a real day, and a twin fed a guessed sigma returns a confident wrong yield.

Becomes possible

The unlock is a shared, measured library of process distributions that a twin can cite rather than assume, the way a structural engineer cites material data. Until then the right discipline is the one this course teaches: state where your sigma came from, and treat a yield figure without one as unfinished.

Every hard thing was impossible until the constraint that made it impossible was named. How we read a frontier →

Module 1

From Design to Manufacturable

Read a CAD part and decide how it could actually be made: choose a material and a process, and flag the geometry that fights manufacturing.

  1. L0What "manufacturable" meansYou probe a CAD part and find a barrel wall at 0.80 mm. What is the manufacturability problem?Given the sensor-mount CAD part, identify at least 3 manufacturability problems and state why each is a problem, each mapped to the correct defect class on a fixed grader seed.→
  2. L1Choosing a material and a processAluminium and CNC gives a feasible part at about $10.10 against a $3.50 target. You need 5,000 of them. What changes?For the sensor-mount part and a stated requirement (strength, cost, count), select a material+process pairing and justify it against at least two alternatives, using the twin's estimated cost, time, and feasibility.→
  3. L1Designing for the process (DFM)You thicken a wall inward to clear two DFM violations. The bore regenerates at 11.4 mm against a 12.00 target. What happened?Modify the part's geometry to remove ≥2 flagged DFM violations while preserving its functional dimensions, and verify the inspector now passes.→
Module 2

CAD → CAM: Toolpaths & Process Plan

Turn a manufacturable design into the concrete instructions that make it: toolpaths, cutting parameters, and a datum-consistent operation sequence.

  1. L1From geometry to toolpathYou cut a pocket with 4 mm corner radii using a 10 mm tool. What is left behind?Generate a valid roughing + finishing toolpath for a pocket and explain how tool diameter and stock define what the path can and cannot reach.→
  2. L1Process parametersYou rough at 6 mm depth of cut and the spindle hits 138 percent load. Which parameter do you change first?Choose spindle speed, feed, and depth-of-cut that complete the cut with no overload and an acceptable surface-finish score, and state the trade-off each parameter drives.→
  3. L1Sequencing a process planYour plan drills four M4 holes referenced to the top face, and faces the top later. What fails?Order a multi-operation plan into a valid sequence, choosing setups, and justify why the order is forced by datums and the surfaces each operation creates.→
Module 3

Tolerances, Process Simulation & Variation

Understand and quantify why real parts deviate from their nominal design, and specify tolerances that keep parts functional despite that variation.

  1. L2Tolerances: the band a part must live inYour smallest bore is 0.005 mm larger than your largest pin. Is that a fit?Assign tolerances to ≥2 critical dimensions and predict how tightening or loosening each changes cost and fit, reaching a functioning fit at the lowest cost.→
  2. L2Where variation comes fromA batch runs 5% scrap, parts outside the spec band. The fastest LEGITIMATE way to get scrap below 2% is to…Run a batch, read the output distribution, identify which process parameter drives the spread, and reduce it.→
  3. L2Tolerance stack-upThree stacked parts each meet their own tolerance. Will the assembled gap ALWAYS stay in spec?Predict whether a chain of toleranced parts will assemble using both worst-case and statistical (RSS) stacks, and state when each method applies.→
Module 4

Assembly & the Production Line

Compose individual parts into an assembly and a balanced production line, reasoning about sequencing, throughput, and bottlenecks.

  1. L2Assembly sequencingStep 4 places the left finger, but the servo horn it fits onto has not been placed. What kind of constraint is this?Produce a feasible assembly order for a multi-part product and explain at least one constraint (access, fastening, fit) that forbids an alternative order.→
  2. L2Throughput & the BottleneckA line has four stations. You install a faster machine that halves the time of a NON-bottleneck station. Line throughput…Find a line's bottleneck (the station with the slowest per-machine cycle) and fix it so the line meets its throughput target.→
  3. L2Balancing the LineYou find the slowest station on a production line and speed it up. What happens to the line’s throughput?Balance a multi-station line so no single station dominates, including the SECOND bottleneck that appears once you fix the first, to hit a high throughput target.→
Module 5

Quality, Yield & Optimization + Capstone

Connect tolerance, process, and line decisions to yield and cost, and optimize a full virtual factory against competing throughput, cost, and yield targets.

  1. L2Yield, scrap & process capabilityA process is already perfectly centered in its tolerance band. To halve the scrap rate you should…Compute a line's yield and process-capability index (Cp/Cpk) from a batch, and raise yield above a target by centering and/or narrowing the process.→
  2. L2Optimizing Under Competing TargetsYou fix the bottleneck station and throughput rises. Where does it stop?Hit the throughput target at MINIMUM cost. Every machine costs money, so add capacity only where it actually buys throughput (the bottleneck), staying within budget.→
  3. L2Capstone: Run & Optimize a Virtual FactoryA five-station line must hit 10 a minute, so every station must clear 6 seconds. Two stations exceed it. What is the minimum-cost fix?Take a full five-station line from below target to meeting both a throughput spec and a tight machine budget: the whole manufacturing toolkit (bottleneck, balancing, cost) at once.→