APQP, FMEA & PPAP: A Real Automotive Case Study

Before we quote a price or cut any steel, we review the design. Seven DFM (design for manufacturing) checks for a high pressure die casting: uniform walls, draft angles, datum pads for machining, alloy and heat treatment, realistic tolerances, minimum machining stock and the complete tooling, plus a bonus check from the machine side. Each one comes with a concrete proposal you can send to the customer.

Resource: L03_DFM_Checklist_Die_Casting.xlsx: the seven checks plus the bonus as a ready checklist with rules of thumb, the housing example, and a wall thickness ratio check that flags hot spots.

Assignment: Write Three DFM Comments

Instructions: A customer sends you a die-cast aluminum bracket, 220 × 120 mm. Walls are 3 mm, but the M10 mounting boss is 12 mm solid. Outer walls have zero draft. The as-cast mounting face has ±0.05 mm and 2.5 mm machining stock. The drawing asks for AlSi9Cu3 with T6. Use the "Template" and "TPL Wall Check" sheets.

1. Run the wall check with a 3 mm nominal wall and a 2.0 ratio limit. Which feature fails?

2. Write three DFM comments, each with finding, proposal and benefit.

3. Which point would you also put in the assumptions list of your quote, and why?

Sample solution:

1. The 12 mm boss: ratio 4.0, NOK. It freezes last and leaves shrinkage porosity right where the thread is cut.

2. Examples: core out the boss to a 3–4 mm wall (sound thread, shorter cycle); add about 1° draft on outer walls (clean ejection, longer die life); machine the mounting face with 0.5–1.0 mm stock instead of holding ±0.05 mm as cast (realistic, fewer open pores); replace T6 with T5, or use vacuum casting with a T6-capable alloy (no blisters).

3. The heat treatment and the tolerance class. Both change the price. If the customer has not answered before you quote, write "price assumes T5 and ISO 8062-3 on as-cast surfaces".

Comments are opinions; the design FMEA (failure mode and effects analysis) turns them into decisions. We build one real line for the sealing face of our housing, following the seven AIAG-VDA steps: function, failure effect, failure mode and cause, then severity, occurrence and detection. You will see why the new handbook uses action priority instead of RPN (risk priority number), and how a special characteristic is born.

Resource: L04_Design_FMEA_Template.xlsx: AIAG-VDA style design FMEA form with automatic AP (action priority), four example lines for the housing, an empty template, and the AP lookup table.

Assignment: Rate One Design Failure Chain

Instructions: The drawing of our housing allows a 2.0 mm coolant channel wall at one corner. Use the "Template" sheet of the resource file.

1. Write the failure chain: failure effect, failure mode and failure cause.

2. Rate S, O and D with one line of reasoning each. What AP does the template give?

3. Propose one design action and re-rate. Which rating changes, and why does severity stay the same?

Sample solution:

1. Effect: coolant reaches the power electronics, the inverter shuts down, the vehicle stops. Mode: coolant leaks through the thin corner wall. Cause: the drawing allows a 2.0 mm wall, too thin for sound filling at that corner.

2. S = 8 (loss of a primary vehicle function, same as the lecture). O = 6 (no minimum wall rule on the drawing). D = 5 (prototype leak test and CT, but only after tooling). AP = H.

3. Action: minimum wall 3 mm with a generous fillet at the corner, checked by casting simulation before the die is released. New O = 3, new D = 4, AP = L. Severity stays at 8 because the effect at the customer is the same; actions lower the chance of failure, not its consequence

From an aluminum ingot to a packed part on a truck: we draw the process flow of our housing, step by step, from 10 to 120. You will learn why every later document is built on this simple map, why each step gets a number, and why a good flow also shows the paths nobody likes to draw: rework, quarantine and scrap returning to the furnace.

Resource: L05_Process_Flow_Diagram.xlsx: the full housing flow in table form, with product and process characteristics per step, plus an empty template.

Assignment: Find the Missing Paths

Instructions: A colleague drew this flow for a die-cast housing: 10 receive ingots, 20 melt, 40 die casting, 50 trim, 80 CNC machining, 100 leak test, 120 pack and ship. Compare it with the example sheet.

1. Name three missing steps and the risk each one brings if it stays missing.

2. Which paths are missing around the leak test? Draw them in the template.

3. Why do we number steps in tens, and why must the PFMEA and the control plan use the same numbers?

Sample solution:

1. Any three of: 30 degassing with density index test (hydrogen becomes porosity, no control point); 60 heat treatment (hardness not controlled); 70 blasting and deburring (burrs at assembly); 90 washing (cleanliness limit missed); 110 final inspection and data matrix code (no traceability in a complaint).

2. The rework loop R100 (one re-test, then scrap), the quarantine area for suspect parts, and scrap and trim returns going back to the furnace through dry storage.

3. Tens leave room to insert a step later (35) without renumbering. Shared numbers make the chain traceable: a step missing from the flow is missing from the PFMEA and the control plan too.

A perfect drawing, a great die and a new machine can still make bad parts. We start the process FMEA for our housing with one real step: degassing. You will define a measurable function, find the work elements with the 4M (man, machine, material, environment), and build the failure chain like a little crime story. Plus the two traps I see all the time.

Resource: L06_PFMEA_Part1_Failure_Chain.xlsx: worksheets for structure, function and failure chain, with the degassing example and an automatic flag for "operator error" causes.

Assignment: Build a Failure Chain for Step 80

Instructions: Use the "TPL" sheets. Step 80 is CNC machining of the sealing face: flatness 0.05 mm or better.

1. List at least four work elements, using the 4M.

2. Write the function of the step and of each work element, each with a measurable requirement.

3. Build three failure chains (effect, mode, cause). No "operator error" allowed.

4. A colleague writes "porosity" as the cause of a leak at the customer. What is wrong?

Sample solution:

1. Man: machine operator. Machine: machining center, fixture, cutting tool. Material: raw casting, coolant. Environment: hall and coolant temperature.

2. Step: machine the sealing face to flatness 0.05 mm or better. Fixture: clamp the housing without bending it. Tool: cut within tool life. Coolant: stay within the set temperature.

3. Effect for all: coolant leak at the customer, vehicle stops. Mode: flatness above 0.05 mm. Causes: clamping force too high bends the thin wall; tool used beyond its life; coolant warms up in the afternoon; a chip under the part on the fixture.

4. Porosity is the failure mode of an earlier step (40, die casting), not a cause at a work element. Keep each box at its own level.

The suspects go on trial. We list prevention controls (the seat belt) and detection controls (the airbag), rate severity, occurrence and detection, and find out why a perfect leak test still leaves a big red HIGH on the screen. Then we fix the cause, not the inspection, with actions that have a name and a date.

Resource: L07_PFMEA_Full_Template.xlsx: the full process FMEA form with prevention and detection controls, automatic AP before and after actions, the degassing example and the AP lookup table.

Assignment: Why Is AP Still High?

Instructions: Take the chain "a chip under the part on the fixture" from the last assignment. Today the operator cleans the fixture with an air gun, and the CMM checks one part every 2 hours. Chips are found on the fixture several times a week. Use the "Template" sheet.

1. Which current control is prevention, and which is detection?

2. Rate S, O and D. What AP do you get?

3. Your manager says: "Add a 100% CMM check and we are fine." What does the AP say? Propose a better action and re-rate.

Sample solution:

1. Cleaning with an air gun is prevention. The CMM check every 2 hours is detection.

2. S = 8, O = 6 (chips several times a week), D = 5 (parts made between two checks can escape). AP = H.

3. A 100% CMM lowers D to about 3, but with S = 8 and O = 6 the AP stays H: catching bad parts is not the same as not making them. Better: air-blow nozzles in the fixture and a seating sensor that blocks clamping when the part is not flat on its pads. New O = 3, new D = 3, AP = L.

If the process FMEA is the brain, the control plan is the hands: the one document the shop floor uses every shift. You will meet the three control plans (prototype, pre-launch and production), the prototype trap with mock-ups machined from solid, and read one production row from left to right, ending with the column that matters most: the reaction plan.

Resource: L08_Control_Plan_Template.xlsx: AIAG-style control plan with production and pre-launch examples for the housing, a comparison of the three plan types, and a template that flags any row without a PFMEA reference.

Assignment: Write One Control Plan Row

Instructions: Use the "Template" sheet and set the type to Production. The characteristic is the Ø40 H7 bore, machined at step 80.

1. Write the full row: method, sample size, frequency, control method, reaction plan and PFMEA reference.

2. A prototype housing machined from solid passed every leak test. What does that prove, and what does it not prove?

3. The production plan has a row "X-ray every 50 parts" with no PFMEA reference. What do you do?

Sample solution:

1. Step 80, CNC machining; bore gauge or air gauge; Ø40 H7 (40.000 to 40.025 mm); 1 part every hour and after every tool change; check sheet or SPC chart; reaction: stop, check tool and offset, sort all parts since the last good check, record; PFMEA reference of the bore line.

2. It proves the shape, the fit and the function. It proves nothing about die casting: a solid block has no porosity.

3. Ask why. Either a risk is missing in the PFMEA (add the line, keep the check), or the check is a leftover from pre-launch (remove it once the data supports that).

Every number you read is the part plus the noise of your measurement. We run a gauge R&R (repeatability and reproducibility) study on the CMM for the sealing face flatness: 10 parts, 3 operators, 3 trials. The first result is 34%, and the cause is a surprise. Then we look at pass/fail gauges like the leak tester, and a die casting classic that fakes a leak.

Resource: L09_MSA_Gauge_RR_Calculator.xlsx: average and range gauge R&R calculator (before and after examples, empty template) and an attribute study sheet with effectiveness, miss rate and false alarm rate.

Assignment: Read a Gauge R&R Result

Instructions: Open "Study 1 (before)" and "Study 2 (after)". Then a new study on the bore air gauge gives: tolerance 0.025 mm, EV = 0.0007 mm, AV = 0.0002 mm, PV = 0.0040 mm.

1. In Study 1, which is bigger, EV or AV? What does that tell you about the cause?

2. For the bore gauge, calculate GRR, %GRR of tolerance (6 × GRR / tolerance) and ndc (1.41 × PV / GRR). Do you accept the gauge?

3. Describe an attribute study for the leak tester, and one trap that is typical for die castings.

Sample solution:

1. AV (reproducibility) is much bigger than EV. The variation comes from people and method, not the machine: each operator clamped the thin housing differently.

2. GRR = √(0.0007² + 0.0002²) ≈ 0.00073 mm. %GRR ≈ 17.5% of tolerance: conditional, acceptable for a normal characteristic, improve it for a special one. ndc ≈ 7, above the minimum of 5.

3. Mix known good parts with master parts that leak on purpose; 3 operators test each part 3 times, blind. Targets: effectiveness 90% or more, miss rate 2% or less, false alarm 5% or less. Trap: a housing still warm from washing can fake or hide a leak, so the part temperature belongs in the method.

Lecture 10: SPC and Cpk

One good part tells you nothing; will part 40,000 be as good as part 1? We separate the voice of the customer (specification limits) from the voice of the process (control limits), build an X-bar and R chart, learn which patterns mean "react" and which mean "hands off", and calculate Cp, Cpk and Ppk with a garage-parking analogy. Then we take our sealing face from Cpk 1.33 to 1.9.

Resource: L10_SPC_Cpk_Calculator.xlsx: X-bar and R charts from 25 subgroups of 5, automatic control limits, run-of-7 and out-of-limit signals, and Cp, Cpk and Ppk, with before and after examples and a template.

Assignment: Calculate Cpk and Read the Chart

Instructions: Open the "Before" and "After" sheets first. Then a bore position has USL 120.05 mm and LSL 119.95 mm. From 25 subgroups of 5: X̿ = 120.012 mm and R̄ = 0.030 mm (d2 = 2.326).

1. Calculate σ within, Cp and Cpk. Is the process capable against a 1.67 target? What would you work on first?

2. In the "Before" sheet, subgroup 1 is flagged "Out X", but all five parts are in tolerance. Do you react?

3. Why is Ppk slightly lower than Cpk in the "Before" sheet?

Sample solution:

1. σ = 0.030 / 2.326 ≈ 0.0129 mm. Cp = 0.10 / (6 × 0.0129) ≈ 1.29. Cpk = (120.05 − 120.012) / (3 × 0.0129) ≈ 0.98. Not capable. Centering alone only lifts Cpk to Cp (1.29), so you must also reduce the variation.

2. Yes. A point outside the control limits means something changed, even if it is in tolerance. Here the point is below the lower limit (a flatter part): find out why, it may be something worth keeping.

3. The flatness drifts every afternoon. That drift adds variation between subgroups, so σ overall is bigger than σ within, and Ppk comes out lower than Cpk.

The die arrives: the moment of truth. T0 proves the tool, T1 proves the part with a full dimensional layout, X-ray and CT, and the run at rate proves the process at full speed with real operators. You will learn the toolmaker's steel-safe trick, see the correction loop from T1 to T2, and avoid the three classic mistakes, starting with the showroom run.

Resource: L11_Trial_Log_and_Run_at_Rate.xlsx: a trial log from T0 to run at rate with findings, causes and actions, and an OEE calculator that compares the run with the promise in the RFQ.

Assignment: Judge a Run at Rate

Instructions: Enter this run in the "TPL Run at Rate" sheet: planned run time 480 min, unplanned downtime 45 min, ideal cycle 50 s, measured cycle 52 s, 480 parts produced, 461 good. The RFQ promised 85% OEE and 2% scrap.

1. What are the availability, performance, quality, OEE and scrap rate? Does the run keep the RFQ promise?

2. Which loss would you attack first, and how would you find its cause?

3. The supervisor wants to repeat the run with the best operator and handpicked parts. What do you say?

Sample solution:

1. Availability 90.6%, performance 92.0%, quality 96.0%, OEE 80.0%, scrap 4.0%. Both are below the promise, so the capacity in the RFQ is not proven yet.

2. Downtime costs the most (45 min, about 9 points of availability). Make a Pareto of the stop reasons, then check whether the scrap clusters after stops, as in the lecture, where a warm-up routine fixed it.

3. No. A showroom run proves nothing about the daily process. Fix the causes, then repeat the run under normal conditions with normal operators.

From RFQ to PPAP: here we are. We open the PPAP (production part approval process) folder for our housing, sort the 18 elements into four simple piles, and go through a level 3 package: ballooned drawing, material tests, CT, leak results, MSA and a Ppk of 1.9. Then the classic rejection caused by one empty balloon, the PSW (part submission warrant), the three possible answers, and why PPAP is only a snapshot.

Resource: L12_PPAP_Checklist_18_Elements.xlsx: 18-element tracker sorted into four piles, ballooned dimensional results that flag MISSING and NOK balloons, and a PSW preparation checklist.

Assignment: Can We Sign the PSW?

Instructions: Open the example sheets of the resource file.

1. The tracker shows 87.5% ready. Which elements are open, and what exactly blocks the PSW?

2. Balloon 27 is empty. A colleague says: "It's only a datum pad, nobody cares." Your answer?

3. Six months after approval, purchasing wants a cheaper ingot supplier. What do you do first?

4. Sort into the four piles: control plan, master sample, MSA studies, design records, initial process studies, PSW.

Sample solution:

1. Element 9 (dimensional results) is open because balloon 27 has no result, and element 18 (the PSW) waits for it.

2. One empty balloon is enough for a rejection. Measure it on 5 parts from the run, or write N.A. with a reason. Never leave it blank. A datum pad also decides how the part is clamped for machining.

3. Tell the customer before changing anything, and plan for a new PPAP. A new material source changes the approved process.

4. Pile 1 (customer designed): design records. Pile 2 (we planned): control plan. Pile 3 (we measured): MSA studies, initial process studies. Pile 4 (we deliver): master sample, PSW.

Six months after PPAP, a Monday morning email: three housings failed the leak test at the customer. We solve it the structured way with 8D (eight disciplines): team, facts traced by the data matrix code, containment within 24 hours, two root causes (why did we make it, and why didn't we catch it), interlocks as permanent actions, proof that they work, read-across to every similar cell, and a thank-you to the team.

Resource: L13_8D_Report_and_5_Why.xlsx: an 8D report form from D0 to D8, a two-leg 5-why sheet that flags "operator" root causes, and a containment tracker, with the leak complaint as the worked example.

Assignment: Two-Leg Root Cause

Instructions: The customer finds 5 housings with a stripped M8 thread at assembly. The data matrix codes show all 5 came from machining center 2, Tuesday of week 22, between 13:00 and 17:00. Facts from the shop floor: the tap was changed at 13:00; the new tap came from a mixed box with the wrong thread class; all tool boxes sit in one cabinet with no ID check at tool change; the hourly thread gauge check passed, but the gauge's calibration had expired three weeks earlier and the system did not block it. Use the "TPL" sheets.

1. Write D2 with facts: what, where, when, how many.

2. Where do you contain in D3?

3. Build both 5-why legs and name one permanent action per root cause.

4. What goes into D7?

Sample solution:

1. 5 housings, M8 thread at the mounting boss strips at assembly torque; machined on center 2, week 22, Tuesday 13:00–17:00.

2. 100% thread gauge check at the customer, in transit, in our warehouse and in WIP (work in progress) from center 2 since Tuesday 13:00, with a calibrated gauge. Mark the first clean shipment.

3. Occurrence leg: thread oversize → wrong tap → mixed tool box → no ID check at tool change. Action: tool ID check at the presetter; the machine refuses an unverified tool. Escape leg: gauge check passed → worn gauge → calibration expired → nothing blocks expired gauges. Action: gauge management system that locks expired gauges.

4. Update the PFMEA thread line and the control plan with the new causes and controls, then read across: every machine that uses taps, and every gauge in the plant.

The most expensive mistake in manufacturing is the one you make twice. In the final lecture we close the APQP loop with phase 5: what to watch for the whole life of the part, how to build a lessons learned database people actually use, the six lessons our housing taught us, the foundation documents where lessons live, and how lecture 14 feeds straight back into lecture 2.

Resource: L14_Lessons_Learned_Database.xlsx: a lessons learned register with the five parts of a useful lesson, a check that each lesson changed a document, and a summary by APQP phase.

Assignment: Turn a Lesson into a Document Change

Instructions: Use the "Template" sheet of the resource file.

1. Take the 8D from the last assignment (or a problem from your own work) and write it as one lesson with all five parts.

2. Which foundation document changes? Write the exact new line: a checklist item, a PFMEA cause or a control plan row.

3. Write one good practice from your own work that should be repeated on purpose.

4. At your next RFQ, which three items from this course will you check first?

Sample solution:

1. What happened: stripped M8 threads after a tool change. Root cause: no tool ID check, and expired gauges were not blocked. Fix: tool ID check at the presetter, gauge lock. Where else: every machining cell and every gauge. Document: generic PFMEA and standard control plan.

2. Generic PFMEA, machining: cause "wrong tool loaded at tool change", prevention "tool ID verified at presetter, machine blocks unverified tools". Standard control plan: "gauge calibration status checked by system before use".

3. Answers vary. A good one is specific and names where it will be reused, like the assumptions list in the housing quote.

4. Answers vary. Strong answers name a document, for example: the assumptions list, the wall thickness ratio check and the document chain check.