UV Ultraviolet - flame detection

Erlend

Hi,

I need a sensor to detect that a gas burner has been ignited - or that it is still burning. I know that an UV sensor is best suited. IR detector is not so good, since it looks for thermal radiation from surfaces and a flame is not such a good radiator in this respect. I therefore have to find some other use for my IR detector.
My question is: where can I source an UV sensor? It normally is a photodiode with a filter for ordinary light.

Erlend

John Abshier

Not cheap $89 http://www.acroname.com/robotics/parts/R345-UVTRON-PKG.html

John Abshier

Tracy Allen

Here is the link to the Hamamatsu UV Tron:
http://jp.hamamatsu.com/products/sensor-etd/pd006

I see there is a company that sells an R2868 with a custom driver board for US$43, and they also list on ebay.

It detects UVC, short wavelengths.

erco

These "flame sensors" are $5 on Ebay: http://www.ebay.com/itm/Arduino-Flame-Sensor-Detects-Fire-Wavelength-20-100cm-/160590917205?pt=LH_DefaultDomain_0&hash=item2563f6ee55

Might just be a phototransistor. I had excellent results using phototransistor sensors at a robot contest to find & extinguish a candle flame many moons ago. My favorite was a Fairchild FPT540A, now rare as hen's teeth.

Ron Czapala

erco wrote: »

These "flame sensors" are $5 on Ebay: http://www.ebay.com/itm/Arduino-Flame-Sensor-Detects-Fire-Wavelength-20-100cm-/160590917205?pt=LH_DefaultDomain_0&hash=item2563f6ee55

Might just be a phototransistor. I had excellent results using phototransistor sensors at a robot contest to find & extinguish a candle flame many moons ago. My favorite was a Fairchild FPT540A, now rare as hen's teeth.

There is a specsheet for these http://www.dfrobot.com/index.php?route=product/product&path=36_58&product_id=195

Leon

I used to work for Irisys (http://www.irisys.co.uk/), they have had a small team working on flame detection algorithms using their sensor technology for years. I just spoke to a former colleague of mine there, and the work was apparently financed by Tyco. This is the unit they came up with:

http://www.tycosafetyproducts-europe.com/English/Products/Fire/minervadetectors/minervaflamefv300.asp

It's based on the Irisys IR sensor array.

Tracy Allen

With an infrared sensor like that one has to consider possible interferences. Here is a graph (url):

The Hamamatsu UV_TRON works in the region where the interferences have dropped way low in significance. How do robot builders deal with the interference at 940 nm?

Here is a chart from some class notes.

Taken with a spectrometer, the many dips and staircase effect is an instrument artifact, as is falloff at long wavelengths. There are a couple of strong lines around 770 nm. The overall spectrum is largely continuous like a black body, and they attribute that to incandescence of soot particles in the oxygen-starved flame.

Leon

That Irisys sensor array isn't affected by interference.

Tracy Allen

Leon, I was asking in reference to the simple 940nm phototransistor from the earlier post, and hobby robots.

The Tyco/Minerva module is very interesting and seems to be available and rebranded into several top-name security systems. The sensor array (256 pixels?) gives good options for contrast detection. That plus a CCTV on the side and its ability to see thru smoke and haze, it could be a great option for a fire-fighting robot. ($$?)

Leon

It's a 16x16 array, with 256 levels for each element. The image is focused with a germanium lens. The array is mounted on an ASIC which interfaces it to a fast processor which runs the detection algorithms. I don't know what Tyco is using, but Irisys themselves use a 600 MIPS ADI Blackfin.

I don't think that the sensor assemblies are available on their own but I'll make some inquiries tomorrow.

Erlend

Thanks for lots of tips. Originally I thought I could do this with an infrared sensor, so I bought the MLX90614 but it did not react to a flame burning close. It did register my skin temperature, though. This is a bit confusing, since several of the sensors in this thread are IR. Something must be wrong. I will set up a more controlled test where aiming at the flame will be more accurate. It has also occurred to me that I could do this much simpler. Using an ordinary light detector - before the flame is to be detected - I could measure the light level (=ambient light) and then detect the flame as more light. It is a less interesting way to do it, though.
I may go on to purchase the vacuum tube based UV detector to experient with, but I don't understand why the IR detector can not do the job.

Erlend

Tracy Allen

Leon, If you find out more about that sensor assembly, I mentioned it to some people I work with who expressed interest in it for ongoing projects.

Erland, look at the graphs and you can compare potential levels of background radiation you have to deal with in near IR at 940nm, compared with UVC at 250nm. It is a lot easier to detect a candle in a dark room! The levels of energy are small, so a scheme is needed to detect the contrast between ambient and the fire. One way to enhance the signal is to narrow the beam width and scan back and forth so that the signal becomes AC at a narrow frequency. The MLX90614 is a broadband detector out to 1000s of nm, whereas the other IR detectors discussed have been mostly IR photodiodes, peaking in near infrared at 940nm.

Leon

Tracy,

I just had a word with one of my former colleagues. They don't supply the sensor units on their own in small quantities, although I daresay that if someone had a mind-blowing application that would sell lots of units they would provide some free kit. However, their people counters occasionally turn up on Ebay, and I found one for sale in HK:

http://jensenyau.en.ec21.com/offer_detail/Sell_IRISYS_IRC3000_people_counter--8705886.html?gubun=S

It costs £990!

Here are some nice You Tube clips showing it in action:

http://www.youtube.com/watch?feature=endscreen&NR=1&v=j7ufv1w5vrg

http://www.youtube.com/watch?v=KG8r1OqGBWA&feature=endscreen&NR=1

http://www.youtube.com/watch?NR=1&feature=endscreen&v=ExTpFTEjs8M

The tracking algorithm is very effective.

Erlend

All,

What I have settled for is to order the R2869 detector with an interface board to find out if it is better at detecting the flame. I intend to ignite the gas using an electric ark (auto spark plug) and it could be that it will also detect the arc - all the better, it would mean I will also get feedback about weather that circuit works.

Thanks for all the help, this forum is fantastic!

Leon

No it isn't!

Tracy Allen

Leon, that first video reminds me of certain Gary Larson cartoons having to do with a bird's eye view of bipeds walking by under their perch. Pretty sophisticated algorithms. Someone wanting to pass unnoticed might wear an invisibility cloak with a pattern exactly matching the carpet on the floor?

I surmise the technology could be used to localize anything that can be circled against the background, not just people, fire for example. The group I am working with is measuring exposure to particulates from fire. It may be an open fire, or it may be in one of several different stoves (gas, or enclosed wood fires) which would be evident as heat or flame sources. They would want to know both which fires are being used and how many people are present in the room, and of those people, how many are children.

Leon

The flame detector uses completely different techniques from the people counter. The latter uses a Kalman filter. The sensor only responds to IR and it would be very difficult to fool it.

Tracy Allen

Leon,
Is the flame detector the same hardware as the people detector, just different processing?

Leon

They used the same sensor array and ASIC when I was working there.

I haven't a clue about how they do the flame detection, there was a different team working on that. I remember them messing about with sand-filled containers having inflammable liquid poured into them, and being set alight. It took them a few years, I think, as they were still working on it when I left.