In precision assembly, it is easy to assume that once the specified torque is reached, the job is done. In many applications, that assumption holds. In others, it becomes the source of hidden problems.

This was the situation faced by a manufacturer of advanced CO₂ detectors assembling a component in which an internal optical element had to be perfectly aligned and mechanically balanced. The element was secured using four screws, each fitted with a soft O-ring.

On paper, the process was straightforward. Apply the specified torque to each screw and verify that the values fall within tolerance. In practice, torque correctness turned out to be only part of the equation, causing major issues and delays in production and culminating in a request for Kolver's help.

Soft joints do not behave symmetrically

Soft elements such as O-rings do not compress instantaneously or uniformly. Their behavior depends on the tightening order, the amount of torque applied at each step, and how load redistribution is allowed to occur during assembly.

In the original process, each screw was tightened directly to one hundred percent of the target torque. The first screw fully compressed its O-ring. The following screws were then tightened against a system that was already partially distorted and caused a tilting effect.

Final torque values were correct and repeatable, but internal stresses and overall balance/alignment were not.

From the outside, assemblies appeared identical; internally, they behaved very differently.

Validation revealed the problem, but too late

Each assembled unit was subjected to a functional validation cycle lasting between five and ten minutes. Only at the end of this test was the part approved or rejected.

With previous product generations, rejection rates reached roughly thirty percent. Every rejected unit represented lost validation time, disassembly, rework, and disruption to production flow.

The tightening data showed no obvious anomalies. Torque values were within specification. Despite this, functional performance remained inconsistent.

Precision was necessary, but not sufficient

Introducing our K-DUCER NT fastening system, with its ultra-precise embedded transducer, eliminated one source of variability (since we found that the torque tools they had been using were not regularly and properly calibrated), but it did not solve the problem on its own. The turning point came from changing the tightening logic rather than focusing solely on tool accuracy.

Three adjustments proved decisive.

First, the screws were tightened using a star pattern. This distributed compression more evenly across the component and reduced asymmetric loading.

Second, torque was applied incrementally instead of in a single step. Tightening was performed at approximately thirty percent, then sixty percent, and only then to final torque. This allowed the O-rings to compress progressively and equalize before the assembly was fully locked.

Third, attention shifted away from the final torque value alone. The focus moved to how each screw behaved while reaching torque, and whether all four followed the same pattern.

Linking the fastening process to validation results

To better understand the remaining failures, a simple but disciplined method was introduced. When a unit was rejected during validation, its unique identifier was recorded so the corresponding tightening data could be traced.

After collecting a statistically meaningful sample of at least one hundred units, and preferably more than two hundred, tightening curves from approved and rejected parts were compared.

The most significant differences did not appear at final torque. They emerged during the running torque phase, where O-ring compression and internal alignment effects are most evident.

From correction to prevention

Once these patterns were understood, it became possible to act earlier in the process. Abnormal tightening behavior could be detected during assembly, allowing suspect parts to be rejected before entering the validation machine. In parallel, tightening strategies could be refined to avoid the problematic conditions altogether.

At that point, quality was no longer something verified after the fact. It became something controlled at the source.

The broader lesson

This case highlights a principle that applies well beyond this specific application.

Reaching the correct torque does not guarantee a correct assembly. When soft elements are involved, tightening order, progression, and dynamic behavior matter as much as accuracy.

When validation is slow and costly, relying on end-of-line testing alone is inefficient. The most effective approach is to ensure that assemblies are already correct before they ever reach validation.

Precision remains essential. Without the right tightening strategy, however, precision alone is not enough.