How would you test the lifetime of a light bulb?

[tony_engin]

Hi all.
I'm wondering how people determine the lifetime of a light bulb. Do they really turn on the bulb and wait till it go down? Or there are special methods to shorten the duration of the testing experiment? Anyone know?

 

[PerennialII]

I think even though we're talking about wolfram/what ever alloys they use nowadays which I know pretty much nothing about, the problem is essentially the viscoplasticity & creep of the material. So in a simplified sense if we assume the failure to obey a typical creep failure law, the failure is characterized by a combination of time-stress-temperature (and a couple other parameters to specify the basic law). So I'd do the tests at tougher conditions (higher stresses and temp -> decreased time to failure) and verify that the law I've selected applies for the material of the bulb -> use the law to predict the long term behavior. This is the common approach to creep testing, the testing times can be taken down from, say, 100k hours (which is a typical design criterion for high temperature components) to 100 - some thousands of hours (in high temperature use of materials).

 

[waht]

re

i suppose you could apply more power to the light build and it would burn out quickly. I think there would be a non-linear relationship with the amount power applied and maximum time befere it burns out.

 

[Cliff_J]

The lifetime can be approximated from the voltage at a very large power.

http://members.misty.com/don/bulb1.html#pay

http://en.wikipedia.org/wiki/Light_bulb#Voltage.2C_light_output.2C_and_life

 

[tony_engin]

But what if I got 3 bulbs that need testing of their lifetime.
I can't use directly the relation to get guess the lifetime, i would need to actully find out the lifetime of each bulb...
What should be the standard method used by the industry?

 

[tony_engin]

I think even though we're talking about wolfram/what ever alloys they use nowadays which I know pretty much nothing about, the problem is essentially the viscoplasticity & creep of the material. So in a simplified sense if we assume the failure to obey a typical creep failure law, the failure is characterized by a combination of time-stress-temperature (and a couple other parameters to specify the basic law). So I'd do the tests at tougher conditions (higher stresses and temp -> decreased time to failure) and verify that the law I've selected applies for the material of the bulb -> use the law to predict the long term behavior. This is the common approach to creep testing, the testing times can be taken down from, say, 100k hours (which is a typical design criterion for high temperature components) to 100 - some thousands of hours (in high temperature use of materials).

I'm sorry...I'm such an idiot that I don't really understand the concept..
Could you please explain the general idea of it?

 

[PerennialII]

I'm sorry...I'm such an idiot that I don't really understand the concept..
Could you please explain the general idea of it?

No worries, sometimes I can hardly grasp any of what I'm saying myself, , the basic idea is that the time to failure is dependent on the combination of stress and temperature (this is quite a simplification, but many creep laws do that with success (at least what comes to precision required in most engineering evalutions)). So increasing either will decrease the time to failure. So we assume that the failure life is government by a creep law, do the testing at tougher conditions (higher stress or temperature) to evaluate the creep law parameter(s) for the material in question, and use the evaluated parameter(s) to compute the life to failure for the actual service conditions (the ones with lower stress / temperature etc.).

Naturally this sort of approach requires some analysis of the situation itself and working with material parameters, it is the common way research & integrity evaluation of this field are carried out. One typical parameter applied in this context is the Larson-Miller one (there are still some questions about the high temperature behavior of alloys used in light bulbs, I for one can't recollect seeing any actual analysis with respect to what material laws their follow etc., some "caution" here, but would expect them to behave as other materials do),

http://pvmdb.nims.go.jp/jpvrc/pvm/jst0151e.htm
http://info.tuwien.ac.at/IAA/news/icpvt/panels/syz11.ppt (slide 7)
http://www.mece.uAlberta.ca/groups/reliability/papers/Paper-PDF-Files/2000-Zuo-Chiovelli-Nonaka-AMSE-JPVT.pdf
http://www.specialmetals.com/documents/Nimonic%20alloy%20263.pdf

Most of the stuff is grad level and beyond so no hesitation in your part is in order if you need any further info, it ain't the simplest stuff around.

 

[Integral]

It is my opinion that the life of a light bulb should be measured in On/Off cycles rather then hours of operation. Another factor seems to be ambient temperature. Higher temperature environments seem to mean lower life.

 

[PerennialII]

It is my opinion that the life of a light bulb should be measured in On/Off cycles rather then hours of operation. Another factor seems to be ambient temperature. Higher temperature environments seem to mean lower life.

Great observation, could as well be the cyclic loading with this information about the material & loading conditions.

 

[brewnog]

I would definitely try to include the cyclic nature of light bulb operation in my answer. A light bulb which is constantly turned on and off will last less long, in terms of actual bulb lifetime, than a bulb which is constantly on.

Personally, I would try and model the failure of a light bulb based upon some kind of fatigue model, although I'm not sure whether the low stresses present would warrant the use of a low-cycle model (as the use/temperature parameters would dictate).

 

[PerennialII]

I would definitely try to include the cyclic nature of light bulb operation in my answer. A light bulb which is constantly turned on and off will last less long, in terms of actual bulb lifetime, than a bulb which is constantly on.

Personally, I would try and model the failure of a light bulb based upon some kind of fatigue model, although I'm not sure whether the low stresses present would warrant the use of a low-cycle model (as the use/temperature parameters would dictate).

On this basis the problem ought to be addressed using low & high cycle fatigue models, a creep model, and then essentially a creep interaction model (creep - fatigue interactions bringing us to a completely new set of models) ... seems like the diverging of approaches warrants some closer investigation in the actual mechanism(s) of failure in order to avoid added work. One thing so far neglected is aging of the filament material bringing some added spice to the equation (and local deformations (necking) likely have a sort of a role at least in the end of the failure process, not forgetting the solder points)...

and googling the problem adds evaporation to the equation :

http://www.seabay.org/articles/let_there_be_light_1.htm

this one below (interesting source) suggests that the traditional (steady - state) creep approach could master it (would consider the fatigue aspect though) :

http://www.everything2.com/index.pl?node=creep

Creep in materials is a slow continuos process where the material slowly deforms over time. To creep the material has to be under load - temperature only accelerates the process. Creep will occur under any load - the stress does not have to be over the yield strength of the material.

A common example of creep is light bulbs - when lit, the tungsten filament heats up to around 2000 degrees C. While this is way below the melting point of tungsten (above 3000 deg C), it is hot enough to initiate creeping in the material. At this temperature it will slowly bend downwards under the force of gravity, and permanently elongate. When it reaches it's maximum elongation, it breaks off - and the light goes out.

The critical temperature when creep starts is typically 30-40% of the melting temperature. Therefore, ice creeps rapidly even at very cold temperatures (-30 C) this is because -30 C is really 88% of the melting temperature of ice. At room temperature we also see that lead is at 50% of melting temperature, which explains why lead pipes deform with age.

There are two main consequences of creep:

* At constant load, the material deforms over time. This is typical with glaciers, turbines, and structures. The material is subject to the same load (most often plain old gravity), and therefore will bend permanentely - given enough time.
* At constant displacement, the load will decrease. Bolts and other pre-tensioned fasteners are often affected by this. The bolts doesn't go anywhere, but they get longer - and therefore relaxes. Bolts on engine blocks needs tightening from time to time. A funny thing here is that the more you tighten the bolt, the quicker it will creep, and the faster it will need re-tightening...

Final fracture occurs when the creep has elongated the defects inside the material so much that the rest material cannot handle the load. This is why most turbine blades are made out of a single grain. A single grain structure minimizes the defects inside the material, and it becomes much less suceptible to creep fracture.

To avoid creep you need to select materials with a high melting point. As stated, the working temperature needs to be under 30% of the melting temperature. If this cannot be avoided, one should alloy the material to make it less creepy (the technical term is to maximise obstructions to dislocation motion...).

As a quick glance in the theory behind creep, here is the formula for steady state creep (power-law creep):

εss = B · σ n

ε = Strain (Elongation)
ss = steady state
B = a material constant
σ = Stress in material
n = A material constant - usually between 3 and 8.

A temperature dependent equation for creep can be found in Arrhenius's Law.

So in trying to arrive at a sort of a conclusion I'd use steady state creep approach as in post #7 and do a creep-fatigue analysis (which will likely be in the high cycle region just considering the number of on/offs and gravity loading). The other intricacies are better left alone if not craving for a huge workload (viscoplastic transient & steady - state FEA combined to creep and fatigue laws, sure could do some analytic work first). Since none of this may not be all that straightforward methods proposed in #3 & #4 might be the practical ways of attaining some empirical information.

Example about a "simple" problem getting some dimension .

 

[brewnog]

Good work PerennialIII.

However, I'd probably suggest that the cyclic nature of the problem would be in the low cycle region rather than the high cycle region. In my house, light bulbs probably last around 6 months, being turned on and off maybe four times a day, which is well in the low cycle range (say 200 days, 4 cycles per day, - N=1000).

 

[PerennialII]

Good work PerennialIII.

However, I'd probably suggest that the cyclic nature of the problem would be in the low cycle region rather than the high cycle region. In my house, light bulbs probably last around 6 months, being turned on and off maybe four times a day, which is well in the low cycle range (say 200 days, 4 cycles per day, - N=1000).

Thanks & yeah, the cycle count is actually quite low compared to what high cycle is usually reserved for, and come to think of it, considering the strain based nature of low cycle methods they seem more fitting for this sort of a case with the high temperature, how the failure occurs and all.

 

[Antiphon]

You have to burn out all three and measure the times.

Three lamps is not enough to draw conclusions about a larger population
of lamps. You need around 20 or more to do that.