How to Detect Seam Leaks in Welded Waterproof Bags

Why Seams Fail Before Fabric

A welded seam is a joint, not a continuation of the material. The parent fabric arrives with a coating or laminate that has been verified to a hydrostatic head under ISO 811 or AATCC 127, but the seam is created on the production floor by melting or softening two layers together under heat, pressure and time. Every one of those three variables has a window, and every window drifts as the machine warms up, as the electrode wears and as operators change shift.

The failure modes are predictable. Insufficient dwell leaves a cold weld that looks closed and leaks under load. Contamination from release agent, dust or hand cream prevents fusion at the interface. Variation in film thickness changes the energy needed at the joint, so a roll at the thick end of tolerance welds differently from a roll at the thin end. Geometry compounds all of it: corners, curvature transitions and points where three layers meet concentrate stress and are the first places to open.

Stitched construction has a different but related problem. The needle punches holes through the barrier, and water travels along the thread by wicking even when the stitch line looks tight. That is why stitched waterproof seams are always backed by tape or by a welded membrane, and why the tape interface becomes the new weak point. Understanding which mechanism applies to your product determines which detection method you should be paying for.

Our own sequence starts long before the leak test: incoming material verification, a first-off weld check at every setup, and in-process inspection at the station. The leak test then confirms what the process data already suggests. Buyers who only test the finished article are paying for information they could have had earlier and cheaper, and they discover problems after the lot is already built. See how we structure that flow on the manufacturing and welding capability page.

Visual and Tactile Inspection

Visual inspection catches the majority of gross seam defects and costs almost nothing, which is why it belongs at the station rather than in a laboratory. The operator looks for an uneven weld line, discoloration that indicates overheating, a weld that has squeezed out too much or too little material, and any gap where the two layers have not met. Tactile inspection adds information vision cannot give: a fingernail drawn along the seam edge finds a cold weld that looks acceptable.

The limitation is real and should be stated plainly. Visual and tactile inspection cannot find a micro-channel, a partially fused section that holds together in the hand but opens under pressure, or a defect buried inside a multi-layer joint. Those are exactly the defects that generate field returns. Visual inspection is therefore a screen, not a release criterion, and any specification that treats it as one is relying on luck.

Make the screen effective by defining it. Give the station a defect catalog with photographs or physical reference samples, set a lighting level, and require the operator to check both faces of the seam along its full length. Then measure the screen: track how many units the station rejects and how many leak tests subsequently fail, because the gap between those two numbers is the true performance of your visual inspection.

• Uneven or intermittent weld line along any part of the seam length.
• Discoloration or scorching that indicates excessive heat or dwell.
• Insufficient squeeze-out, which usually means the layers never fused.
• Excessive squeeze-out, which thins the joint and weakens it under peel load.
• Visible gaps, blisters or delamination at corners and layer transitions.
• Any seam that separates when a fingernail is drawn along its edge.

Water Dip With Internal Absorbent

The water dip is the most intuitive leak test and, done properly, one of the most reliable. The bag is packed with a dry absorbent material such as tissue or paper wadding, closed in its service configuration, and immersed in water for a defined period. The absorbent is then removed and inspected. Any wet spot marks the location of the leak, which makes this method uniquely valuable for diagnosis as well as for pass or fail.

Four parameters control the result, and all four belong in the work instruction: immersion depth, immersion time, water condition and closure configuration. Immersion depth sets the pressure at the seam, so a shallow dip is a gentler test than a full submersion and the two must never be compared. Immersion time matters because a slow leak through a micro-channel needs time to deliver a visible volume of water.

Add a wetting agent to the water if you want a more aggressive test, because plain water has high surface tension and resists entering a fine channel. Understand that you are changing the severity when you do this, and record whether surfactant was used. The method is destructive in the sense that a failed unit is wet inside and needs drying before it can be reworked, which is the main reason high-volume programs supplement it with air testing.

Air-Decay Testing of Sealed Volumes

Air testing inverts the problem. Instead of waiting for water to find its way in, you seal the bag, introduce a small controlled pressure, and watch whether that pressure holds. Because air is far less viscous than water and the pressure difference can be measured precisely, the method detects leaks that a short water dip would miss, and it does so in seconds rather than minutes. It also leaves the unit dry, so a passed bag can go straight into packing.

The method returns a number rather than a wet spot, which is what makes it suitable for statistical process control. Record the decay for every unit and you get a distribution; a distribution drifting upward tells you the welding process is degrading before any unit actually fails. That early warning is worth more than the pass or fail decision itself, and it is the main argument for choosing air testing on a high-volume program.

The trade-off is that you lose location information. A decay reading tells you there is a leak but not where it is, so a failed unit still has to be diagnosed, usually by a water dip or by sectioning the seam. Plan for that: air testing for volume screening, water dip for diagnosis. We cover the air method in detail in our dedicated guide to air pressure leak testing.

Dye-Penetrant for Stitched Seams

Where the construction is stitched rather than welded, the leak path is usually along the thread. Dye penetrant makes that path visible. A low-surface-tension liquid carrying a visible dye is applied to one side of the seam, and the seam is then examined or sectioned to see how far the liquid has traveled through the stitch holes and along the thread. Wicking distances of several millimeters are common and are invisible from the outside.

The technique is destructive in practice, because sectioning is the only reliable way to measure wicking distance, and stained units cannot be sold. Use it on first-off samples, on periodic audit units and during root-cause investigation rather than as a routine release test. A few units per week, tested to destruction and photographed, generate more useful process information than a hundred non-destructive passes.

The most valuable output is a comparison between constructions. Run the same dye test on a stitched-and-taped seam and on a welded seam from the same material and the difference in penetration is usually obvious. That is the evidence you need when deciding whether your product should be built welded, stitched and taped, or as a hybrid, and it directly informs the discussion in our comparison of RF welding versus stitch and tape.

Sampling Rate by Lot Size

Seam leak testing is applied statistically, not to every unit, unless the value of the contents justifies a one hundred percent check. The standard framework is ISO 2859-1 with an acceptance quality limit of 2.5, which sets the sample size from the lot size and general inspection level. You then apply your chosen leak method to every unit in the sample and accept or reject the lot against the accept and reject numbers.

The table below gives commonly used values for general inspection level II at AQL 2.5, and it is the reference our release teams work from. Note how slowly the sample size grows relative to the lot size: going from a 500-piece order to a 20,000-piece order roughly triples the sample, not forty-folds it. That is deliberate, and it is why AQL sampling remains the economical choice for most programs.

Choose your inspection level deliberately. Level II is the default and the right choice for established programs with stable process data. Move to a tighter level when a new construction is being qualified, when a supplier is new, or after a failure, and relax back once the data supports it. Document the level in the inspection plan, because switching levels without telling the supplier is a common source of disputes at pre-shipment inspection.

Whatever sample size you land on, keep the leak test result in the same lot file as the dimensional and cosmetic findings, so a single release decision is made on all the evidence at once. Our AQL 2.5 inspection plan is built around that single-file principle, and it is what makes a pre-shipment dispute resolvable without re-opening the whole lot.

Repairability and Rework Rules

Not every leaking seam should be repaired, and not every repair should be accepted. A short unmelted section on a straight run can usually be re-welded with a patch or a second pass, and with the right setup the repair is as strong as the original. A leak at a three-layer corner, a leak caused by contamination across the whole joint, or a leak that recurs after one repair attempt generally means the unit should be scrapped.

Set the rules before production starts. Define which defect types are repairable, how many repair attempts are permitted on one unit, what the repaired area must look like, and whether a repaired unit has to be retested by the full method or by a reduced one. Write the retest requirement explicitly, because a repaired seam that is never retested is simply an unchecked seam with extra history.

Watch the rework rate as a process signal rather than a cost line. A stable rework rate is normal; a rising one during a run usually means electrode wear, a material roll change or a drifting machine parameter. Stop and fix the cause when the rate moves, because reworking units is far more expensive than adjusting the machine, and reworked units carry a higher field-failure risk even when they pass.

• Repairable: short unmelted sections on straight runs, single pinholes, edge welds with clear margins.
• Scrap: multi-layer corner failures, contamination across the joint, repeat failures after one repair.
• Every repaired unit must be retested by the same leak method before it rejoins the lot.
• Cap the number of repair attempts per unit, typically one, and record each attempt.
• Escalate when the rework rate rises materially during a run rather than reworking to the end.

Recording Seam Failures by Station

The final discipline is attribution. Log every seam failure against the station, the machine, the operator, the shift and the material roll that produced it. Without that attribution you know the lot failed but you do not know why, and the same failure returns on the next order. With it, a Pareto chart of failures by station usually points at one machine or one setup within a single production week.

Keep the record simple enough that operators will actually maintain it. A station code, a defect code from the catalog, a count and a time stamp is enough. Photograph representative failures and attach them to the defect code so the catalog stays grounded in reality rather than in descriptions written once at product launch and never revised.

Close the loop with a review cadence. A weekly review of station failures, with the welding technicians present, converts data into parameter changes: a dwell adjustment here, an electrode replacement schedule there, a change to how a roll is staged before welding. That cadence, more than any single test method, is what separates a supplier with a stable seam yield from one that tests harder and still ships leaks.

For buyers, the practical ask is straightforward: request the station-level failure summary for your last three production lots before you commit to a new one. A supplier who can produce it is running a controlled process. Detailed guidance on what else to request is set out in our pre-shipment inspection planning notes and in the custom dry bag manufacturing overview.

Frequently Asked Questions

Q1: Why do welded seams leak when the fabric itself tests well?

A: The fabric value and the seam are different things. A hydrostatic head result under ISO 811 describes flat material, while the seam is a joint created by heat, pressure and dwell on the production floor. Cold welds, contamination, thickness variation and stress at corners all produce leaks that have nothing to do with the fabric rating.

Q2: Can visual inspection alone be used to release a lot?

A: No. Visual and tactile inspection finds gross defects such as uneven weld lines, scorching and open gaps, but it cannot detect a micro-channel or a partially fused section that holds together in the hand and opens under load. Use it as a station-level screen and release the lot on a physical leak test with statistical sampling.

Q3: What is the most reliable way to find where a leak is located?

A: A water dip with dry absorbent packing inside the bag is the most informative, because the absorbent shows the wet spot and therefore the location. Immerse the closed bag for a defined time at a defined depth, then unpack and inspect. Air testing is faster but only tells you a leak exists, not where it is.

Q4: How long should a water dip test run?

A: Long enough for a fine channel to deliver a visible volume of water into the absorbent, which is typically several minutes rather than seconds. Too short a dip produces false passes on slow leaks. Fix the duration in the work instruction and never compare results between dips of different lengths or depths.

Q5: Should a wetting agent be added to the dip water?

A: Adding a surfactant makes the test more aggressive because plain water resists entering fine channels due to surface tension. That is useful when you want a harsher screen, but you must record whether it was used, because results with and without surfactant are not comparable and mixing them makes trend data meaningless.

Q6: Why is air-decay testing preferred for high-volume programs?

A: It is fast, it leaves the unit dry so a passed bag can go straight to packing, and it returns a numeric decay value rather than a simple pass or fail. That numeric record supports statistical process control, so a drifting distribution warns you about process degradation before units actually start failing.

Q7: When should dye penetrant be used instead of other methods?

A: Use it on stitched and taped constructions where the leak path runs along the thread, and on first-off samples or root-cause investigations rather than routine release. The test is destructive because measuring wicking distance requires sectioning, and stained units cannot be sold, so keep volumes low and photographic records high.

Q8: What sample size should be used for seam leak testing?

A: Use ISO 2859-1 at AQL 2.5 with general inspection level II as the default. That gives a sample of 80 for a lot of 501 to 1,200 pieces and 200 for a lot of 3,201 to 10,000 pieces. Move to a tighter level for new constructions, new suppliers or after a failure, and document the level in the plan.

Q9: Which seam defects can be repaired and which should be scrapped?

A: Short unmelted sections on straight runs and single pinholes are usually repairable with a second weld pass or a patch. Multi-layer corner failures, contamination spread across the joint, and any unit that leaks again after one repair should be scrapped. Every repaired unit must be retested before rejoining the lot.

Q10: How many repair attempts should be allowed on one unit?

A: One, as a general rule. A seam that fails twice has usually failed for a systemic reason such as contamination or geometry, and repeated reworking rarely fixes it while adding cost and field risk. Record every attempt, because a rising rework rate is a process signal rather than a cost line.

Q11: Why should seam failures be recorded by station?

A: Attribution turns a failed lot into a fixable cause. Logging failures against station, machine, operator, shift and material roll produces a Pareto chart that usually identifies one machine or setup within a week. Without attribution you know the lot failed but not why, and the same failure returns on the next order.

Q12: Does a stitched and taped seam need a different leak test?

A: Yes, because the leak mechanism is different. Stitching punches holes through the barrier and water wicks along the thread, so a dye-penetrant test that measures wicking distance gives information a water dip or air test does not. Keep the routine leak test as well, but add dye penetrant at first-off and audit points.

People Also Ask