ISO 9001 certification in progress RoHS material control Vietnam country-of-origin documentation rfq@saolamfg.comEmail WhatsApp +84 86 501 7914

Guide

What does a useful PFMEA example look like from a machining or die casting supplier?

Short answer

A PFMEA is a process failure mode and effects analysis: a structured, living record in which the supplier walks every step of its process flow, asks how each step could fail to produce the feature you drew, what that failure would do to your assembly, what causes it, and which controls prevent or catch it. This guide shows a worked PFMEA example for a die-cast and machined housing and explains how to tell a real one from a template that was filled in for the paperwork.

By Saola Manufacturing engineering team · Published · Updated

It is written for the supplier quality engineer who has asked for a PFMEA alongside a control plan and now has to judge what came back, and for the design engineer who wants to know which of the drawing’s features the supplier actually treated as risky. The method and terms follow the AIAG & VDA FMEA Handbook and SAE J1739; the review checklist and RFQ wording are engineering guidance, not part of either document. Where a PFMEA fits inside a full approval package is covered in the guide to PPAP for non-automotive parts; this page goes one level down, into the content of the analysis itself.

On this page

What goes into a process FMEA under the current method?

A current-method PFMEA is built in seven steps, from planning and a structure tree of process steps down to a failure chain of effect, mode and cause for each step, with prevention and detection controls rated on separate scales. The output is a ranked list of actions, not a score sheet.

The 2019 handbook, published jointly by AIAG and the German VDA, replaced the older AIAG fourth-edition approach and is the basis many automotive and industrial customers now reference. Its seven steps are planning and preparation, structure analysis, function analysis, failure analysis, risk analysis, optimization and results documentation. In practice, three of them decide whether the document is useful.

  • Structure analysis. The process steps come straight from the supplier’s process flow diagram, and under each step sit the work elements that can cause trouble, usually grouped as man, machine, material and environment, sometimes with method and measurement added. If a step exists on the router but not in the PFMEA, it was never analyzed.
  • Failure analysis. Each step gets a failure chain: the failure mode (what the step produces wrong, such as a bore oversize), the effect (what you see at your assembly or in the field) and the cause (which work element produced it, such as an insert worn past its life).
  • Risk analysis. Severity rates the effect, occurrence rates the cause in light of its prevention controls, and detection rates how well the detection controls catch the mode or the cause before the part leaves. All three use 1-to-10 scales defined in the handbook.

The handbook also retired the risk priority number, the product of the three ratings, as the way to rank work. It uses an action priority table instead, which weights severity first, then occurrence, then detection, so a high-severity failure with weak controls is flagged even when its product looks modest. An older PFMEA ranked by risk priority number is not wrong, but it is a sign the document predates the current method and may not have been reviewed since. The ASQ overview of FMEA gives the general logic for readers outside automotive, and the IEC 60812 standard covers the same technique in a sector-neutral form.

What does a real PFMEA example look like for a die-cast and machined housing?

A real one names the specific feature, the specific cause on that machine or tool, and a control that exists on the floor, as in the rows below. Every row should be traceable to a drawing characteristic and to a line on the control plan.

The part is an aluminum die-cast housing with a gasket face, a bearing bore, tapped mounting holes and a cross-drilled port, plated after machining with the threads masked. These rows are an illustration of the level of detail to expect, not data from any production process. A full PFMEA would carry many more rows per step, plus the function column, responsibility and target dates for actions.

Illustrative PFMEA rows for a die-cast aluminum housing, machined and finished after casting · severity (S), occurrence (O) and detection (D) on the 1-to-10 scales per SAE J1739 and the AIAG & VDA FMEA Handbook (2019); example ratings, not data from any process
Process stepFailure modeEffect at your assemblyCause (work element)Prevention controlDetection controlSOD
Die castPorosity in the gasket-face zoneLeak path opens once the face is machinedShot parameters or die temperature drift (machine)Parameters recorded at setup; gate and overflow position fixed at toolingMachined face compared with a porosity boundary sample856
Mill datum face AFace not flat, step at fixture locatorGasket does not seal; bore position shiftsChips trapped under a locator (man, machine)Blow-off and locator check in the work instructionFlatness on a surface plate at first piece and at a set frequency746
Bore bearing seatBore oversizeBearing loose, noise, early wearInsert worn; offset not reset after insert change (machine, man)Tool-life counter forces the change; offset changes loggedIn-machine probe gauges the bore on every part and stops the cycle when it is out of limit743
Drill and tapThread short of full depthScrew cannot be installed; your line stopsBroken tap not noticed (machine)Tap replaced on a counted lifeGo/no-go thread gauge to depth on every part636
DeburrBurr left at the cross-holeBurr breaks loose into the fluid pathHand step rushed or skipped (man)Defined tool and picture-based work instructionVisual check of the intersection with a light, method not yet proven758
Mask and plateCoating inside tapped holesScrews bind or gall at assemblyMasking plug missing (man, material)Plug count per rack recordedGo gauge after finishing546
PackOld revision mixed into the cartonWrong part reaches your lineTwo revisions in process at once (method)Line clearance at each revision changeLabel and count check against the router626

Three things in the table are worth copying into your own review. First, every cause is physical and local: an insert, a locator, a plug, not “operator error” or “machine fault”. Second, the prevention and detection columns are different things; a gauge is detection, a tool-life counter is prevention, and a PFMEA that lists “100 percent inspection” as a prevention control has confused the two. Third, the deburr row carries the weakest detection rating and is the row that should drive an action, which is exactly what the action priority logic is meant to surface. The die-cast row is also the one where the PFMEA should lean on casting-specific knowledge; the guide to porosity that shows up after machining explains why the gasket face is the classic exposure point.

How can you tell a supplier’s PFMEA is boilerplate?

A boilerplate PFMEA could be stapled to any part number: its steps are generic, its causes are vague, its ratings are uniform and nothing in it has ever produced an action. Check it against your drawing and the supplier’s own process flow, and the gaps show quickly.

Use these checks in order. Each one takes minutes and each one fails a template outright.

  1. Part number and revision. The header should carry your part number, the drawing revision it analyzes and its own revision history. A PFMEA dated before your latest drawing revision has not seen the change.
  2. Step match. Lay the PFMEA beside the process flow diagram and the router. Outside processing, rework loops, washing and packing are the steps most often missing.
  3. Your critical features. Every characteristic you marked as critical or special on the drawing should appear as a failure mode with its own row, and its severity should reflect what you told the supplier about function. If you never told them, a generic severity is your gap, not theirs.
  4. Specific causes. “Operator error”, “machine malfunction” and “wrong material” are categories, not causes. A real cause names something that can be controlled.
  5. Rating spread. If most rows carry the same occurrence and detection values, nobody rated them. Real ratings vary with the controls in place.
  6. Controls you can see. Pick two controls and ask for the evidence: the gauge, the tool-life setting, the check sheet. On a site visit or a live video walk-through, look for them at the machine.
  7. Closed actions. A mature PFMEA shows actions taken, with dates, and re-rated rows after them. Lessons from returns and corrective actions should feed back into it.

One caution in the other direction: a long PFMEA is not a better one. A document with hundreds of rows of handling and labeling failures on a simple bracket usually means the template was copied in full. Depth belongs on the features that carry function.

How should severity, occurrence and detection be rated on machining and casting steps?

Severity is set by the effect on your product and should come from you, occurrence by how strong the prevention control is, and detection by how reliably the check catches the defect before shipment. Most rating disputes come from mixing those three up.

Per the AIAG & VDA FMEA Handbook, the top of the severity scale is reserved for effects on safety or regulatory compliance, and the supplier rates severity at the highest level it knows about, which may be the effect at your plant or at the end user. A build-to-print supplier cannot know that a bore feeds a gas valve or a dental handpiece unless the drawing or the RFQ says so. Give them the function of the critical features, or expect severity to be guessed low.

Occurrence should drop only when prevention improves. On a machining step, a tool-life counter, a fixture that physically cannot accept the part in the wrong orientation, or a locked program revision justifies a lower rating; a promise to be careful does not. On a casting step, stable recorded parameters and scheduled die maintenance are the prevention controls that count.

Detection depends on what the check can actually find. In the handbook’s process detection table, human inspection or manual gauging rates 6 when the method has been proven effective and 8 when it has not, while automated checks that stop the process or hold the part rate between 2 and 4. An attribute gauge on every part catches an undersize thread far better than a visual check on a sample, and a feature that is only inspected at first article has almost no detection once production is running. Two errors are common: rating detection on a check that happens after the defect is already hidden, such as a plated-over burr, and rating it low because the gauge exists when the measurement system has never been studied. The guide to go/no-go thread gauging covers what an attribute check can and cannot see.

How should the PFMEA connect to the control plan and the drawing?

The process flow, the PFMEA and the control plan should share the same step numbers, and every detection control in the PFMEA should reappear in the control plan with a method, sample size and frequency. If the two documents disagree, the control plan is what the floor runs.

The chain works like this. Your drawing marks the characteristics that matter. The PFMEA rates the risk on each and decides which controls are needed. The control plan turns those controls into instructions: which gauge, how many parts, how often, and the reaction when a check fails. The inspection records then show the controls were run. Break any link and the PFMEA becomes a desk exercise.

Where the PFMEA calls for a capability study on a characteristic, the control plan should say how the data are collected, and the guide to Cpk vs Ppk explains which index fits a launch and which fits ongoing production. Where it relies on first-piece checks after each setup, the first article inspection guide covers what that report should contain. And because both documents live and change, ask that each carries a revision level and that the PFMEA is reviewed whenever the drawing, the tooling, the process or the site changes, and after any significant nonconformance.

What should the RFQ and purchase order say about a PFMEA?

State the method, the scope and the features, and ask for the process flow and control plan in the same package, so the supplier can price the work and you can review the three together. A one-word requirement of “PFMEA” invites a template.

Wording along these lines works on an RFQ or a supplier-quality clause on the purchase order:

  • “Supplier shall provide a process flow diagram, process FMEA and control plan for this part number at drawing revision C, prepared per the AIAG & VDA FMEA Handbook (2019) or SAE J1739. Rows shall be traceable to the characteristic numbers on the ballooned drawing.”
  • “Characteristics 4, 7 and 12 are critical: loss of seal on a pressurized housing. Supplier shall rate severity accordingly and identify prevention and detection controls for each.”
  • “The process FMEA shall be updated and resubmitted with any change of drawing revision, tooling, process step, outside processor or manufacturing site.”
  • “The process FMEA may be reviewed on screen or on site if the supplier considers it proprietary.”

The ballooned numbering matters more than it looks: it is what lets you trace a PFMEA row to a characteristic in seconds, and the guide to ballooned drawings and inspection reports shows how to set it up. Scope the requirement to the parts that carry function. A full PFMEA on every washer and bracket costs time on both sides and is rarely read.

For a part that is die cast and then machined, ask for one PFMEA that covers casting, machining, finishing and packing as one flow, because most of the interesting failure chains cross those boundaries: porosity created at the die-casting machine is only found at the machining step, and coating in a thread is only found at assembly.

What process documents does Saola provide today?

Saola’s published documentation set does not list a PFMEA as a standard document, so if your approval needs one, say so on the RFQ and ask for it to be answered with the quote. What the quality page does list is a process routing sheet on request, a first-article inspection report, a dimensional record with every lot, and work to a control plan you supply.

The plant runs die casting, CNC milling of its own castings, CNC turning and surface finishing on one site, so a single process flow can cover the whole routing. The quality page lists the process routing sheet per part number on request, a first-article report covering every toleranced dimension, a dimensional inspection record with every lot, salt-spray results at first article and on request, and material certification on request. It also states that dimensions you mark as process-capability control points are checked more often than the rest of the print, and that the plant builds its inspection around a control plan or cosmetic standard you send. ISO 9001 certification is in progress; no third-party quality audit report exists yet.

The factory page shows the kind of risk reasoning a PFMEA formalizes: for a die-cast base with drilled and tapped holes, the inspection plan was written before the first shot, with machined holes checked on every part because a drill can wander on any cycle, and cast features checked at a fixed interval because they move only as the tool wears. If you need that reasoning delivered as a PFMEA in the AIAG & VDA format, put it on the quote request together with the drawing and the critical characteristics.

Frequently asked

Who owns the PFMEA, the supplier or the buyer?

The supplier owns the process FMEA because it owns the process; the buyer reviews it and supplies the inputs only the buyer has, mainly the function and severity of critical features. The design FMEA belongs to whoever owns the design, so a build-to-print supplier is normally not asked for one.

Is the risk priority number still acceptable?

Many non-automotive customers still accept it, but the AIAG & VDA handbook replaced it with action priority because the product of three ratings can rank a severe, poorly detected failure below a trivial, frequent one. If your quality manual says RPN, consider updating it to name the method you actually want.

Can a supplier refuse to send its PFMEA?

Some treat it as proprietary because it describes their process in detail. A reasonable compromise is review on screen or during a visit, with the control plan and process flow sent in full. Agree this before the order, not at the approval stage.

How often should a PFMEA be updated?

Whenever the drawing, tooling, process, outside processor or site changes, after a significant nonconformance or customer complaint, and at a periodic review the two parties agree. A PFMEA with no revision since launch on a part that has had corrective actions has not been maintained.

Does a simple machined part need a PFMEA at all?

Often not. For a low-risk part, a process flow, a control plan and a first article report carry most of the value. Reserve the PFMEA for parts with sealing, pressure, safety or regulatory functions, or for processes with several hand-offs. The PPAP guide shows how to scope the rest of the package the same way.

Sources

  1. AIAG & VDA FMEA Handbook, 1st edition, Automotive Industry Action Group and VDA QMC, 2019
  2. SAE J1739, Potential Failure Mode and Effects Analysis (FMEA) Including Design FMEA, Supplemental FMEA-MSR, and Process FMEA, SAE International, revised January 2021, stabilized May 2026
  3. IEC 60812:2018, Failure modes and effects analysis (FMEA and FMECA), International Electrotechnical Commission, 2018
  4. Failure Mode and Effects Analysis (FMEA), ASQ, accessed 2026
  5. Production Part Approval Process (PPAP), 4th edition, Automotive Industry Action Group, 2006

Have a drawing for this? Send it for a quote.

STEP or PDF, one part or a whole family. You get a unit price at your volume, the tooling cost, and a short manufacturability note.