In automated fastening projects, the first thing a customer's quality department asks in the kick-off meeting is rarely the machine model or the cycle time. It is which class a given tightening point belongs to, whether the whole process is monitored, and whether the records can be retrieved. Behind that question sits VDI/VDE 2862, the classification framework for tightening joints. It is not a paper exercise: once a joint is classified, the tool accuracy, control method, depth of recording and inspection frequency all follow from it. Get the classification right and the equipment specification has a basis; get it wrong and the investment may land entirely in the wrong place.
1. Classification follows the consequence of failure, not the bolt size
VDI/VDE 2862 sorts bolted joints into three classes according to what happens when the joint fails. The German original calls them KAT A, KAT B and KAT C; in Chinese practice they are simply class A, B and C:
- Class A (safety-relevant): failure may endanger people, for example steering, braking, seat and seat-belt anchorages. The highest requirements apply, with monitoring and records for every single fastener.
- Class B (function-relevant): failure causes loss of function or an obvious fault, for example wipers, window regulators, pump bodies and bracket joints. Requirements sit in the middle.
- Class C (customer complaint): failure affects neither safety nor function but generates complaints, for example interior trim panels and decorative fixings. Requirements are the lowest, though correct torque is still required.
Two mistakes account for most of the trouble in practice. The first is classifying by bolt size or torque level: an M6 joint is not automatically less critical than an M10, and the consequence of failure is what matters. The second is classifying downwards on the basis of experience, turning a safety-relevant point into class C. The latter is normally a major nonconformity in an IATF audit, because once the class is wrong, every requirement downstream is wrong as well.
The correct order is to look first at the customer's special requirements (CSR) and at the function of the system, and only then judge the consequence of failure. Most vehicle manufacturers list their safety characteristics in the CSR; anything on that list is class A and needs no internal debate. In industries without a CSR to follow, work through severity of consequence and have the conclusion confirmed in writing by the customer's quality department.
2. Minimum requirements of the three classes
The differences between the classes concentrate on four dimensions: control method, data recording, process capability and tool accuracy. Taking the classes one at a time is clearer than laying them side by side, and easier to map onto specific stations.
Class A: the controlled and monitored variables must differ
- Control torque or angle, monitor the other; the two must not be the same, and one of them must be measured directly;
- Monitor and record every fastener, with storage that cannot be altered and traceability by workpiece;
- Indicate pass or fail explicitly; a failed fastener that has not been reworked must not reach the next station;
- Verify measuring accuracy and repeatability at defined intervals;
- Cpk for torque and angle typically not below 1.67, with tool accuracy around ±5%.
Class B: the controlled variable must be measured directly
- Control torque or angle, with the controlled variable measured directly by transducer;
- Process monitoring need not cover every fastener, but records must exist;
- Some class B points carry additional monitored variables and traceability requirements; confirm in the CSR;
- Cpk typically not below 1.33, with tool accuracy around ±10%.
Class C: the lowest requirements, but verification still applies
- Control torque or angle, measured directly or indirectly;
- Curve records and per-fastener evaluation are not mandatory; basic parameters plus visual confirmation;
- Dynamic verification at defined intervals is still required;
- Tool accuracy around ±20%, with no capability index requirement.
The accuracy figures and capability indices above are the common industry reading; actual values follow the customer's CSR and the part drawing, and individual manufacturers add or relax requirements within the same class. Based on the company's current project experience, the argument on class A points most often lands on Cpk: the machine meets its specification, but incoming parts vary too much for the index to reach target. In that case the fix is upstream, in the parts and the thread quality, not in machine parameters.
3. Mapping accuracy grades onto real equipment
Once ±5%, ±10% and ±20% are put against hardware, the choice is fairly clear:
- ±5% means a sensorised servo tightening spindle: torque measured by transducer, angle captured simultaneously by encoder, closed-loop control, curve stored. Class A points essentially have no other route.
- ±10% means a sensorised electric tool or a servo system with torque feedback: it satisfies the class B requirement that the controlled variable be measured directly.
- ±20% means current-controlled or clutch-type electric screwdrivers and pneumatic tools: torque is governed by current or a mechanical clutch. Repeatability may be adequate, but there is no direct torque measurement and no curve, so these can only carry class C.
Chinese projects often refer to the accuracy grades I, II and III of QC/T 518, which broadly correspond to ±5%, ±10% and ±20% for classes A, B and C. The correspondence is useful when drafting a proposal, but the customer's CSR still governs.
The most common mismatch is using a clutch-type screwdriver on a class A point. Its repeatability may even fall within ±5%, but without direct torque measurement, a per-fastener curve and result storage it does not formally meet class A and will be treated in an audit as unmonitored. No parameter adjustment fixes this; the tool has to change.
On an automated fastening machine, a class A point is normally configured like this: a servo spindle with dual-channel torque and angle acquisition, one result record per screw, bound to the workpiece barcode and station number; when the evaluation fails, the machine outputs an NG signal interlocked with the line so that nothing proceeds until rework is done. Class B may record at station level only, and class C can be covered by torque sampling at a single station.
4. The four things an audit actually checks
1. Calibration and machine capability
An auditor looks at calibration records, not merely at whether calibration happened. Tightening systems are normally verified for measuring accuracy and repeatability against the VDI/VDE 2645 series, and the reference torque transducer must itself be within its validity period with a range that matches the point being checked. Comparing static torque only, without a dynamic repeatability check, is a frequent weak spot.
2. Residual torque sampling
Class A and B points are normally sampled for residual torque once per shift, more often for class A. Using the re-tightening method on a hard joint, the acceptance window is commonly taken as 0.8 to 1.2 times the process torque M, with the lower limit relaxed on soft joints. That window is only a starting point: the working window should be set from the station's own history so that each point has its own limits. Class C points are generally not sampled for residual torque.
3. Curve windows and detection of false passes
Class A requires multi-stage window monitoring. Final torque alone is not enough; the angle progression and gradient during tightening matter as well. A seat face that never closed, a cross-threaded screw and a second tightening of the same screw can all end with a final torque inside the window while the curve shape is wrong. Being able to detect these false passes is the dividing line between a class A system and an ordinary power tool.
4. Whether records can be traced and whether they can be altered
The requirement is a record per fastener, not editable, retrievable by workpiece. The details that come up on site include whether the record is bound to a unique workpiece identifier, how long it is retained, whether data is complete across a power interruption or a shift change, and whether a reworked screw leaves its own trace. All of this has to be settled at proposal stage; retrofitting it usually means reworking the line's data path.
5. What a wrong classification costs
A downgraded classification is a major nonconformity in an IATF audit and directly affects a supplier's rating. The more practical risk is a safety-relevant point without per-fastener monitoring: once a nonconforming part escapes, the cost of a recall and claim dwarfs the price difference on the equipment. The opposite error costs too: configuring every class C point to class A standard wastes both investment and cycle time, since per-fastener recording and curve evaluation consume acquisition and storage resource and add takt. The point of classification is to spend where spending is required.
6. Three steps from classification to equipment configuration
- Build the list: enumerate every tightening point on the product with its joint function, consequence of failure and whether it appears on the customer's safety characteristic list. Missing a point here invalidates everything afterwards.
- Classify and confirm: produce the A/B/C list against the CSR and system function, and have it confirmed in writing by the customer's quality department. The classification belongs in the proposal file as the basis for configuration and acceptance.
- Configure equipment and records by class: class A takes a servo spindle with per-fastener records and NG interlock; class B takes a sensorised tool with station-level records; class C is consolidated into a high-throughput station. Where one machine must cover several classes, configure it to the highest class present, otherwise a changeover also means changing tools.
Mixed-model lines have one further detail: classification follows the product, so the same station may carry different classes on different products. At proposal stage, identify the most demanding case, size the machine for it, and then call the appropriate evaluation window per product in the program.
7. Frequently asked questions
Q1: Do we need classification outside the automotive industry?
The method itself is general. Appliance, 3C and energy storage industries do not mandatorily reference VDI/VDE 2862, but the logic of a customer audit is the same: for safety-critical parts you must be able to show that every screw was acceptable. Running the points through an A/B/C review at least makes clear which points need monitoring and which can be sampled.
Q2: Does class A always require a record for every fastener?
The standard calls for per-fastener monitoring and per-fastener records that cannot be altered. Individual customers may relax the recording granularity in their CSR, but the monitoring itself is not relaxed. Follow the customer's written requirement and do not lower it on your own initiative.
Q3: What if Cpk cannot reach 1.67?
Separate machine capability from incoming variation. On the machine side, check torque repeatability and calibration status; on the material side, check thread quality, seat-face flatness, burrs and applied adhesive. In most cases that fall short, the root cause is the incoming part, and machine parameters cannot compensate for it.
Q4: If a machine is rated at ±5%, does it meet class A?
No. Accuracy is only one item. Class A also requires that the controlled and monitored variables differ, per-fastener records, result evaluation, NG interlock and periodic calibration. A machine that meets the accuracy but has no recording and interlock capability still does not satisfy class A.
Q5: What has to change to raise an existing clutch-type line to class A?
The essential change is at the execution end: replace clutch-type screwdrivers with sensorised servo spindles, add torque and angle acquisition, per-fastener records and NG interlock, and feed results into the line or the MES. Station structure and fixtures can usually be retained, with the work concentrated on the spindle, the controller and the data path. The exact scope depends on the workpiece and the state of the existing line; the above is stated on the basis of the company's current project experience, and the final solution is subject to the actual workpiece and confirmed proposal.
— End of Article —
Author: Chisu Automation Technical Team
Published: November 9, 2026
Keywords: tightening class, VDI/VDE 2862, class A B C, accuracy grade, residual torque, tightening data traceability