Key takeaways
- Welding is a molten bridge that solidifies across the interface. Sticking is a weld weak enough to hold until a later operation tries to open it.
- Inrush loads, DC arcs and contact bounce cause most welding failures, not steady-state current.
- Material choice helps, but force, opening speed and arc suppression decide whether the weld forms.
- Test with the real load and the real closing force. A resistive bench test will not reproduce the failure.
Welding and sticking are not the same failure
Contact welding is a metallic joint formed between two contacts when the interface melts and then solidifies. The joint can be strong enough to hold the mechanism closed after the coil is de-energised. Sticking is the same family of failure, but the joint is weaker: the contacts open on one operation and hold on another, or they open slowly and leave a rough, welded mark.
The distinction matters for diagnosis. A hard weld usually points to a high-energy event, such as a short circuit or a capacitive inrush. Intermittent sticking usually points to marginal force, bounce or a surface that is partially fused. The two need different fixes, even though they look similar on a returned part.
A contact that has welded often shows a torn surface, a matte patch or a small crater where the bridge pulled apart. A contact that has eroded without welding shows a rougher, more uniform loss. Photograph the face before cleaning it, because the evidence is easy to destroy.
How a weld forms
Current passes through a small number of contact spots. The local current density is high, and the temperature at those spots rises above the melting point even when the bulk contact stays cool. If the contacts are opening, the last spot to separate forms a molten bridge. If the bridge cools before it breaks, the two surfaces are joined.
At make, the same thing happens in reverse. The contacts approach, a small area touches first, and the current concentrates there before the full face closes. If the mechanism bounces, the contact makes and breaks several times in a few milliseconds, and each break produces an arc and a molten spot.
The energy available at the interface is the product of voltage, current and arc duration. That is why a load with a high inrush and a long time constant is more dangerous than a resistive load at the same nominal current. The nominal current sets the heating. The inrush and the arc duration set the welding.
Loads that make welding more likely
Capacitive loads draw a very high initial current as the capacitor charges. Lamp loads draw a high cold-filament current that falls as the filament heats. Motor loads draw a starting current several times the running current. All three produce a current pulse at make that is much higher than the steady value.
DC loads are harder than AC because the arc does not get a current zero to extinguish it. The arc duration is set by the circuit time constant and the opening speed, and a long arc transfers more energy into the contact. Inductive loads with a long time constant are the worst case.
A short circuit is the extreme case. The contact may weld during the fault, and the mechanism may not be able to break the resulting joint. That is one reason switchgear uses contact materials and mechanisms designed for the prospective fault current, not just the rated load.
| Load | Why it is hard | First measure |
|---|---|---|
| Capacitive | Very high inrush at make | Pre-charge or series resistor |
| Incandescent lamp | Cold filament inrush | Rate for the cold current |
| Motor | Starting current and inductance | Check the starting duty |
| DC inductive | Long arc, no current zero | Freewheel diode or suppressor |
| Transformer | Inrush and stored energy | Switch at the right point |
| Resistive | Lower risk at rated current | Check the actual overload |
The table ranks typical risk. The circuit, the opening speed and the contact material can change the order.
Material choice and force
Materials that resist welding have a high melting point, a dispersed oxide phase or a graphite phase. Silver tungsten, silver graphite and formulated silver tin oxide are the usual choices for high-inrush duty. Fine silver and silver nickel are more prone to weld because the molten bridge is strong and the matrix is soft.
Material alone will not fix a mechanism problem. A weld forms when the interface melts, and the force at the interface controls how much real area is available and how easily the molten bridge is broken. Too little force gives a small contact area, high local heating and a weak mechanism for breaking a weld. Too much force can increase the weld area, but it also reduces bounce, which is usually the larger effect.
The separating force matters as much as the closing force. If the spring can only just open the contacts at the end of life, a small weld will hold them. Check the opening force at the worst-case position, with the spring at its tolerance limit and the mechanism at its wear limit.
Bounce, speed and overtravel
Bounce is the most common mechanism cause. The contacts touch, the armature rebounds, and the contacts separate for a fraction of a millisecond before closing again. Each separation draws an arc at a current that is often higher than the steady current. Reducing bounce is usually more effective than changing the contact material.
Opening speed sets the arc duration. A faster opening stretches the arc and brings it to a higher voltage sooner, which limits the current and shortens the energy pulse. A slow opening, or a mechanism that hesitates, keeps the arc alive and puts more energy into the face.
Overtravel is the margin between the point where the contacts first touch and the point where the mechanism reaches its rest position. It provides the force that resists welding and the travel that absorbs wear. A mechanism with too little overtravel loses force as the contacts erode, and the welding risk rises through life.
Check the spring at end of life, not at the start
A new relay may have twice the force it needs. After wear, relaxation and erosion, the margin can disappear. Set the minimum force at the end of life and design the spring to hold it there.
Circuit measures that reduce the energy
Arc suppression attacks the energy at the source. An RC snubber across an AC contact reduces the rate of rise of voltage and shortens the arc. A metal oxide varistor clamps the inductive kick. A freewheel diode across a DC coil or load provides a path for the stored current, but it also slows the release, so check the effect on the mechanism.
Zero-cross switching turns an AC load on or off near the voltage zero, which reduces the inrush for capacitive and transformer loads. It is not a universal fix, because some loads need to be switched at a particular point in the cycle, and because the contact still has to break a fault current.
A series resistor or a pre-charge circuit limits the capacitor inrush. A soft starter does the same for a motor. These are system changes, not contact changes, and they are often more effective than a harder contact material.
Testing for welding
Use the real load. A resistive test at the rated current will not reproduce a capacitive inrush, a cold filament or a motor start. Build a test circuit with the actual load, the actual cable and the actual source impedance, and run enough operations to see a trend.
Measure two things: whether the contacts open after each operation, and how much force is needed to separate them. A simple continuity check after each cycle catches a hard weld. A force gauge or a current-limited pull test catches a weak weld that a continuity check would miss.
Record the contact resistance before and after the run. A rising resistance with a stable face suggests erosion or film growth. A sudden drop followed by sticking suggests a weld that has been pulled apart and re-formed. Photograph the face at intervals so the progression is visible.
A prevention checklist
Work through the list in order. Most welding problems are solved before the material is changed.
- Rate the contact for the inrush, not the steady current.
- Reduce bounce with a damped armature, a pre-travel stop or a different spring rate.
- Set the minimum contact force at end of life, including wear and relaxation.
- Check the opening force against the expected weld strength.
- Add arc suppression or zero-cross switching where the load allows it.
- Choose a material with a proven welding resistance for the load class.
- Test with the real load and record the separation force, not only continuity.
Frequently asked questions
Why do relay contacts weld when the current is within rating?
The rating is usually a steady-state value. A capacitive, lamp or motor load can draw many times that current at make, and a DC inductive load can hold an arc for a long time. The welding energy comes from the inrush and the arc duration, not the nominal current.
Does a harder contact material stop welding?
It can reduce it, but it is not a complete fix. Welding depends on the molten bridge, the force and the arc duration. A harder material with the wrong force or a bouncy mechanism can still weld.
Can a welded contact be repaired?
The contact face has been melted and torn, so its surface and material distribution have changed. Replacing the contact is the reliable option. If the part is a sealed relay, replace the relay.
How do I test for sticking that only happens occasionally?
Run a long test with the real load and check the separation force after every operation, not just continuity. Intermittent sticking often needs a force measurement to detect, because the contacts may open on the next cycle.
Sources and further reading
- IEC 61810-7, Electromechanical elementary relays, test and measurement procedures
- IEC 60947-4-1, Low-voltage switchgear and controlgear, contactors and motor starters
- Qumek contact life test and failure analysis notes
Qumek Engineering Desk
Engineering team
The engineering team behind Qumek drawings, DFM notes, tolerances and inspection standards. Every guide is reviewed against the drawings and processes used on live programmes.
Chasing a contact that will not open?
Send the load profile, the contact material and the spring force. The engineering desk will reply with a failure analysis plan and the changes worth testing first.
Send the failure details