The auditor sits across from you and asks: "Do you have a current process capability study (PFU) for this bolted joint?" If you hesitate at this point, you are not alone. The process capability study according to VDI/VDE 2645 Part 3 is one of the most frequently requested - and most frequently performed incorrectly - types of evidence in screw assembly.
This guide walks you through how to correctly plan, carry out, and document a PFU - in a practical way, without getting lost in a jungle of standard clauses.
What is a PFU and why is it required?
VDI/VDE 2645-3 describes procedures for process capability studies (PFU) for breakaway torque of bolted joints with preload. That may sound dry, but it has a very clear practical meaning: you demonstrate that your fastening process consistently and reliably meets the required tightening parameters - not just today, not just in a test, but under real series production conditions.
A PFU is typically required by:
- OEM quality management and Tier-1 customers as part of PPAP or APQP
- Auditors in IATF 16949, AS9100, or VDA 6.3 audits
- Internal stakeholders, when quality deviations or complaints have arisen on bolted joints
- In the case of process changes (new tool, new material, new tightening strategy)
The PFU provides key data for assessing and continuously improving the screwdriving process under series conditions - including identifying systematic influences, assessing tolerances, and defining intervention limits for control charts.
MFU vs. PFU: The crucial difference
Before you get started, you need to understand what differentiates a PFU from an MFU - because this difference determines how you must measure.
| Feature | MFU - Machine Capability Study | PFU - Process Capability Study |
|---|---|---|
| Goal | Assessment of pure machine precision | Evaluation of the overall manufacturing process |
| Conditions | Controlled laboratory conditions | Real production conditions |
| Influencing factors | Machine only | People, Material, Method, Machine, Environment (5M) |
| Time period | Short-term study (1 shift) | Long-term study (multiple shifts/weeks) |
| Metric | Cmk ≥ 1.67 | Cpk ≥ 1.33 (Automotive: ≥ 1.67) |
| Sample size | Min. 50 consecutive parts | Min. 50-125 parts over a representative period |
| Typical application | Machine acceptance, startup | Series monitoring, audit evidence |
| Standard reference | VDI/VDE 2645 Sheet 2 | VDI/VDE 2645 Sheet 3 |
The core idea: Unlike a machine capability study (MFU), a PFU considers not only the machine influence, but also the influence categories man, material, method, and environment.
In practical terms, this means: In an MFU you exclude all disturbance factors - same operator, same batch, no interruptions. In a PFU, you deliberately allow these disturbance factors, because you want to know: How capable is my process under real-world conditions?
This has direct consequences for your sampling plan.
Step by step: How to carry out your PFU
Define in writing the screw-tightening scenario, target value, and tolerance limits (OSG/USG). Determine the sampling plan: at least 50 measurements, ideally 125 over a representative period (several shifts, lot changes). Ensure that all measuring devices are properly calibrated — an expired calibration renders the entire PFU unusable.
Before you begin the actual PFU: Validate your measurement system. Check whether the employed analysis tool is sufficiently accurate (ÜG ≤ 5% of the tolerance). Only a demonstrably capable measurement system yields usable PFU data. Document the device type, serial number, and calibration date.
Start the measurement under real production conditions: real operators, real batches, normal shift sequence. Record for each fastener the breakaway torque. Document all operating conditions: date, time, batch, operator, tool ID.
Before calculating Cpk, your measurement data must be checked for normal distribution (e.g., using the Anderson-Darling test or graphically via a Q-Q plot). VDI/VDE 2645-3 requires a normal distribution of the measurement values. If the data are not normally distributed, you must apply alternative evaluation methods.
Calculate mean (μ), standard deviation (σ), Cp, and Cpk from your measurement data. The minimum Cpk is 1.33 for standard applications—automotive customers often require ≥ 1.67. If the Cpk is below the limit, immediately initiate a root cause analysis.
Create a comprehensive PFU report including: process description, sampling plan, measurement data (tabular and graphical), distribution diagram, Cpk calculation and result evaluation. The report forms the basis for process release by the customer and for future audits.
Interpreting Cpk and Cmk correctly
You have your measurement data - now comes the evaluation. The key question: What do the capability indices actually mean?
The formula behind it
The Cpk value relates the tolerance range to the actual process variation while taking into account the position of the process mean:
- Cp = (USL - LSL) / (6 × σ) -> Pure consideration of spread
- Cpk = Minimum of [(USL - x̄) / (3σ)] and [(x̄ - LSL) / (3σ)] -> Spread and centring
Important: A high Cp value with a low Cpk value means your process is precise, but systematically offset from the target value - which is just as problematic as high variation.
Overview of limit values
| Cpk value | Assessment | Theoretical Scrap | Typical Measure |
|---|---|---|---|
| < 1,00 | ❌ Not capable | > 2.700 ppm | Stop the process immediately, perform root cause analysis |
| 1,00 - 1,32 | ⚠️ Conditionally capable | 318 - 2.700 ppm | Increased monitoring, initiate improvement actions |
| 1,33 - 1,66 | ✅ Capable (Minimum requirement) | 63 - 318 ppm | Regular monitoring, implement SPC control charts |
| ≥ 1,67 | ✅✅ Good capability (Automotive) | < 63 ppm | Standard for safety-critical connections |
| ≥ 2,00 | 🏆 Excellent | < 0,001 ppm | Six-Sigma level, Aerospace & Defense |
A Cpk value of 1.67 results in a theoretical defect rate of less than 0.57 ppm (parts per million). This corresponds to the quality level that automotive OEMs demand for safety-critical Class A bolted joints as defined in VDI/VDE 2862. In medical technology or aerospace, even higher Cpk values are often required than in general mechanical engineering.
Interactive Cpk/Cmk calculator
Enter your process parameters and immediately receive an assessment of your fastening process:
Typical pitfalls in your first PFU - and how to avoid them
From practical experience, we know the most common mistakes that lead to a PFU having to be repeated or not being accepted in an audit:
Pitfall 1: Measurement system not validated
Common mistake: Many companies start the PFU without validating the measurement system in use beforehand. If the analysis or inspection tool itself is not calibrated, all measured Cpk values are worthless — no matter how carefully the measurement was performed. Calibration first, then measurement.
Pitfall 2: Sample not representative
Many teams take all 50 measurements in a single shift - and call it a PFU. In reality, this is an MFU under a different name. A true PFU must reflect shift changes, batch changes, and different operators. Distribute your samples across at least three shifts and, if possible, several days.
Pitfall 3: Wrong measurement parameter
VDI/VDE 2645-3 refers to breakaway torque (also "breakaway moment" or "loosening torque") - not tightening torque. Breakaway torque is measured after embedding and relaxation effects have occurred and indicates the preload that still remains in the joint. If you measure tightening torque, you are answering a different question.
Pitfall 4: Normal distribution not checked
The Cpk calculation is based on the assumption of normally distributed measurement values. Many skip this verification step - with the result that Cpk values are statistically unsound. To assess process capability, customer requirements are compared with process results using a probability model based on the normal distribution. Use at least a graphical Q-Q plot or the Anderson-Darling test.
Pitfall 5: Incomplete documentation
The PFU report must stand up to an audit. If calibration certificates, batch information, or the process description are missing, the evidence is worthless to an auditor - even if the Cpk values are fine. Define before you start measuring what must be documented.
Practical tips for a valid PFU
On sample size:
- Minimum requirement in VDI/VDE 2645-3: at least 50 measurements over a period that is representative of series production
- Practical recommendation: 125 measurements (5 × 25 from different production times)
- Automotive customers often explicitly require ≥ 100 measurements
On measurement conditions:
- Normal production environment - no special cleaning of the workstation before measurement
- Deliberately include different operators - this is intended, not a mistake
- Any batch or lot changes that occur in normal operation must be reflected in the PFU
On documentation:
- Include the calibration certificate for the measuring tool with a valid date
- Keep raw measurement data (not just the calculated Cpk)
- Record date, time, operator, batch, and tool ID for every measurement point
- Provide a process description with a photo of the joint location
How GWK tools support PFU execution
A PFU stands or falls with the quality of the measurement data. Two factors are crucial: the accuracy of the analysis tool and the reliability of its calibration.
QUANTEC MCS® - analysis tool for your data foundation
The QUANTEC MCS® analysis tool with patented angle sensor technology records every tightening process in full - torque and angle in real time. For the PFU, this means: for each of the 50-125 measurements, you obtain complete tightening curves from which the breakaway torque can be precisely extracted. The integrated data interface (Wi-Fi, ZIGBEE) transfers the measurement data directly to your evaluation software - no manual transfer, no transcription errors.
This is particularly relevant if you need to monitor several bolted joints at the same time or ensure seamless traceability for future audits.
Q-CHECK® - calibration verification of the tools in use
No PFU result is more reliable than the measurement system used to produce it. With the Q-CHECK® calibration test system, you can quickly and precisely check whether your screwdriving tools are still within their calibration tolerances - before you start the PFU measurements. Q-CHECK® documents the test results automatically, including tool data and timestamp.
GWK also operates its own DAkkS-accredited calibration laboratory - for situations where traceable in-house or DAkkS calibration is required as evidence. Mobile calibration services enable calibration directly at your site, minimising production downtime.
Conclusion: A PFU is not just a paperwork exercise
A properly executed process capability study according to VDI/VDE 2645-3 is more than an audit document. It provides robust figures on how stable and reliable your fastening processes really are - and reveals where systematic optimisation potential lies.
The key points at a glance:
- PFU ≠ MFU: Real production conditions with all 5M influences are mandatory
- Make the sample representative: At least 50 measurements across multiple shifts/batches
- Validate the measurement system first: Calibration is not optional, it is a prerequisite
- Check normal distribution before calculating Cpk
- Cpk ≥ 1.33 for standard applications, ≥ 1.67 for automotive and safety-critical classes
- Make documentation audit-proof: raw data, tool calibration, process description
How many measurements do I need for a PFU according to VDI/VDE 2645-3?
The standard recommends at least 50 measurements, which must be distributed over a representative production period. In practice, 125 measurements (5 samples of 25 parts each) are ideal to also capture shift and batch changes. Automotive customers often require a minimum sample size of 100 measurements.
When must a PFU be repeated?
A repetition is necessary for: significant changes to the screwdriving tool, changes to screw type or material, change of assembly method, after complaints or internal quality deviations, as well as at regular intervals according to your internal quality plan (typical: annually or after major production ramp-up phases).
What is the difference between Cp and Cpk?
Cp describes whether the process dispersion fundamentally fits into the tolerance—disregarding the position of the mean. Cpk additionally considers whether the process is centered. A process with Cp = 1.8 and Cpk = 1.1 disperses little, but is closer to a tolerance limit than desirable. For the proof of compliance, the Cpk value is always decisive.
What happens if the Cpk value does not meet the requirement?
A Cpk < 1.33 means: The process is considered not capable. You must perform a root cause analysis (e.g., with an Ishikawa diagram) and initiate corrective actions. Typical actions are: check/calibrate the screwdriving tool, change the tool type, adjust tightening parameters, or train the operators. After that you must repeat the PFU.
Can I perform a PFU without specialized software?
Basically yes - for manageable datasets an Excel spreadsheet with built-in formulas for mean, standard deviation, Cp and Cpk is sufficient. In practice, however, specialized tools are recommended that record measurement data automatically and analyze them. Analysis tools like the QUANTEC MCS® capture torque and angle data directly at every screw operation and prepare them for the PFU evaluation - this saves substantial time and eliminates manual transcription errors.
