Also, with each heating/cooling cycle mechanical tensions occure in the filament that shorten it's life
True but you are talking a few %, nothing signifigant. You will get more percentages in manufacturing error that limit the lifespan of a filament than you will create by overdriving the filament, within reason of course.
Also, filaments are not instantaneously heated and cooled, there is a clear ramping period which can be augmented by the frequency and duty cycle. Find the rampdown period of the filament and bring your frequency to the point where the falling edge-rising edge period is roughly equal to the total rampdown period. adjust the duty to bring the ramp's lowest point back up. Add a dash of flywheel diodes and debug to taste.
What you are doing is pulsing current only when needed to keep the filament hot. It's a balancing act but it does work and can conserve power. It's really just a trick and if the pulse frequency is too low you can see the strobe effect in the filament.
Perceived brightness of bulbs is proportional to temperature of the filament, temperature of the filament depends on power, so for the same perceived brightness you need to dissipate the same power, regardless is it pulsed or constant
Only if you are comparing the same type of filaments to each other. If you compare a halogen to a non halogen incandescent then all bets are off. Halogen gas concentrations, percentages of each gas, impurities in the filaments, impurities in the gas can have a measurable effect on the brightness while still consuming(wasting) the same amount of power.
What happens when you compare apples to oranges? How about a case study:
You mention perceived brightness, lumens (measured in candela), which has nothing to do with the actual power output but are generally used to quantify each other. I used 5w LEDs which have 0% IR output, 75% blue and around 24% yellow (1%other colors). The perceived brightness(with a lux meter) was measured to be nearly identical to that of a 25w halogen bulb while consuming 1/4th the current @1/2 the voltage. The LED was able to illuminate an object 20ft away with a 6deg collimator. On the other hand, the 25w halogen put out a whopping 65% IR, 20% reds, 10% yellows and the rest was a peaky mess all over the spectrum. That means that 16.25 watts of power was wasted as IR.
But that is only half the story. The human eye is MUCH more sensitive to blues than reds. This also gives humans a false sense of brightness, where a 25w halogen causes mild discomfort when pointed into the eye while a 5w LED of the same measured brightness causes actual pain in the eye of the engineer who wasn't quick enough to leave the room before being chosen as a guinea pig.
Guinea pigs were not harmed during the testing process. Rogue engineers are another matter..
:green: