UV, IR, UV/IR and other possible sensors for detecting fires

Hello,

Trying to make an effective way of detecting flames from 2 meters away. A number of articles suggest I use either a combination of temperature, smoke, UV (to detect UV-C), IR (multi-IR to remove false alarms) and UV/IR (also for false alarms). Current flame source is from a candle.

I currently have an MLX90640 (temp) and AS7331 UV sensor board (UV) from Sparkfun.

A number of issues I seeing so far:

  1. From a candle, the flame I'm getting is over 200 to a bit over 500 °C but it changes the farther I move the candle. Not really concerned on the high accuracy but just that it can detect the presence of those temperature
  2. UV-C is detected from some flames but only from a small distance

Current resolutions I'm trying but with little progress:

  1. Get the difference between the warmest point in the array versus the surrounding arrays but only if there exist a surrounding array value of over 100°C
  2. Parabolic reflector to reflect light to the sensor. I can see a number of reflectors from Mouser but not sure if I should get that or those generic once for photography.

What could I do to improve my fire detection capabilities?

Then for sensors, some questions:

  1. For smoke detection, do I get the PM 2.5 sensors over the AQI sensors such as those in Adafruit/Sparkfun? I was hoping to get something smaller/less bulky than PMS5003. If AQI sensor, which sensor can I get? currently eyeing ENS160 sensor and SGP30 sensor
  2. For multi-IR, what sensor can I use based on my range requirement. Currently looking at Triad Spectroscopy Sensor or AS7263 NIR sensor board (both from Sparkfun)? My issue is that the limited effective range of these sensors are the same as the AS7331 UV sensor board I have.
  3. For UV detection, I can see a lot of groups and even from fighter-fighting robots using those Hamamatsu UVtron UV sensors R2868 (or similar variants). My issue is that I dont know where to get that photodiode

Please let me know what I can do to improve my chances of detecting flames at 2 meters based on my requirements.

There is a lot of information missing, what type of flame, etc. There are some white papers, check them out.

I have seen this link from a few posts here about fire detection in this forum already haha. My basis would be the Class A, B and C in our fire code


So burning plastics, paper for A, burning LPG/Butane gas for Class B and burning wires for Class C. I hope that helps.

It's natural. Sensors like this have a lobe, like a cone, meaning that the more far away the greater the area included is. It increases. What the sensor sees is a sum up of the radiation withing the loob area.
Once built a simple IR thermometer and it's easy to verify how the reading decreases when the distance increases.

How do you compensate for that?

Why not get some commercial units and then reverse engineer the one(s) that work? Flame detection is not an exact science but an art.

Art? How so?

Not easy with a student's budget and it's not something I can easily get from where I'm from. (The fire code above already gave that away)

A bit of guidance/wisdom would really be helpful.

No, Class C isn't burning wires. It's energized electrical equipment. The only distinction between it and (usually) Class A and (occasionally) Class D is that it is energized.
In other words, a Class C ceases being Class C once the power is cut.

Understood. So I'll just remove Class C fires from my requirements and focus on Class A and B flames

I don't. I make sure that the entire view of the sensor is the object, like in an ear, close to a heated metal plate etc.

Using IR radiation focus lenses, if they are available, ought to help. Look at IR camera drones!

1 Like

Yup, I agree. The flame is visible to the sensor from my experiments. No obstruction between and inside the house without interference from sunlight. Only issue is the distance since my max detecting distance is 2 meters.

Will check these out in the meantime.

Put this in your search engine "Flame sensing 2 feet" If you find one you like get a copy of its manual. From that download the patents they claim. After reading them you should have a good idea of what you want.

You're referring to which device?

First, let's be certain this is for a school project or such in a controlled, constantly monitored setting with no exposures by someone with ready access to fire fighting equipment such as a 2-A:10-B:C fire extinguisher and trained how to use it. Yes, that absolutely includes experimenting with a candle.

Not easy, not at any useful range for anything other than demonstration purposes of the sort of technology used, within any modest budget. At 2 metres, the equipment can't be trusted to sense what you want to sense. Too close and of course, you melt the equipment in any practical demonstration. IR sensors are for fast flaming fires and invisible fires such as alcohol fires. In other words, as I'm sure you're aware, they detect IR light.
A temp sensor may work for your purposes if it's used for convection detection (ie over the candle as the candle heats the air and the hot air rises) but not so much for radiant heat (ie the side). Adafruit makes a thermal imaging camera that uses IR detection and may suit your needs
https://learn.adafruit.com/adafruit-mlx90640-ir-thermal-camera/arduino-thermal-camera

UV detection is used to make longer range thermal cameras day blind such as those used in forestry firefighting. The thermal cameras in those are still based on IR detection, largely, at least the ones I've used.

Smoke, pick any MOS tin dioxide ribbon (iirc) type, general purpose smoke sensor module or roll your own from just the sensor and make it into any circuit you like. The MQ-2 is inexpensive and works for non-life safety applications. You can make these more or less sensitive by adjusting the load resistor, some modules I think even let you adjust the sensitivity with a potentiometer.

The difference between these and the professional sensors comes down to how the circuit is wired (for professional detection - one example being Wheatstone bridge configuration vs simple voltage divider), what else is used at the same time (to compare against), the standards they were tested to (ULC for example), sensor life expectancy and drift (less drift is better), the standards of the attached alarming device (how loud, how bright) and the ability of the device to continue to alarm under fire conditions (off the top of my head, I'm sure there's more).

UV light, if it's in the scope of your project, is typically used in the bulb of a photo ionization detector (PID), which you might well call a VOC detector, since it's an easy way to remember what they're most often used for. The UV part isn't a sensor, but a lightbulb in this case to ionize the air in the detector chamber and change the current across the detection circuit.

Well, not necessarily. A rated extinguisher for class C fires is that they may be used on energized electrical equipment and not conduct electricity back to the person using the extinguisher. It's still a point worth considering.

Why did you omit Class D, if I may ask? Flammable metal fires are becoming an increasing concern in the fire service with the increased popularity of EVs, not just for their batteries, but for high strength, lower weight chassis materials such as magnesium that pose extra concerns since they are notoriously difficult to extinguish, particularly with standard hose streams (water makes it worse but boy are the fireworks fun! - photo not me). This is what happens when the hose stream hits magnesium.

Industrial flame sensors like "fireye" brand (AKA purple peepers), sense the UV emitted by combustion, not heat or visible light. Very accurate (and expensive).

https://www.fireye.com/

Yes, and class B fires made and put out by firemen in our local fire department. So safety's first and foremost.

Yes, I actually do have an MLX90640 board already. Though I do have concerns as stated in my first post. But do I need say like another IR sensor board sensitive to a specific wavelength?

I actually have that and I will try but would say an ENS160 board be better? I was wondering what's a good indicator of smoke detection, AQI or resistance or existence of PM 2.5?

Yes, these are standards and safety features in placed for certified products which I can try adding in future revisions. Though, I'd like to make the project work first which I find I'm struggling especially with the detection range.

I think that's the case with Hamamatsu UVtron sensor but is that the same principle for AS7331. Current issue is really the detection range.

Class A and B fires are mostly common here. Not a lot of EVs right now where I'm from. Battery fires, sure but I think (I maybe wrong) it classifies as chemical fires.

This was actually recommended in a number of previous posts related to flame detection. Though I think these sensors use a UVtron sensor and a special lens/parabolic reflectors inside (maybe to focus the light to the sensor) which is actually what I want to do with my AS7331 to increase the UV-C detection range but cant yet managed to do so far. It's a hit-or-miss actually and I'm still open to suggestions

Excellent

Both unless you're monitoring for some specific condition and fuel load.

Is this your application? If so,

Without knowing anything about your requirements beyond a candle flame at 2 meters, I would use a thermal imaging camera with enough range.

Whatever detects at the lowest level while avoiding false positives and continues to work under the condition you're testing for. Also depends on your application I would think. Air quality and particulate matter have more to do with human health exposure to drifting wildland fire smoke and AQ especially was something discussed a lot during covid as it pertained to certain workplaces. Short answer from a firefighting perspective is as I say. Keep small fires from growing with best detection available. A good thermal imaging camera accomplishes that goal, ie finding heat well above what's expected in the environment you're in.

Even that doesn't have a simple answer and the ongoing debate depends on who you ask, what type of lithium battery, its quantity, whether still energized or not and so on. It's not often that fires occur in controlled settings like a lab, so in practical terms, it has considerations that cross into every major classification.

I was just wondering but it doesn't matter if it's flames you're really looking for which is characterized most by heat and light of course, so you're still in a great starting place by using the best thermal camera you can afford.

As to specific cameras, I can't say from your selection since I didn't design the ones I use. Our go to TIC for general purposes is made by MSI
https://us.msasafety.com/Thermal-Imaging/c/105?locale=en
and the camera we use for general purposes (tons of adjustments available in software) on our DJI Matrice 300 drone is called the Zenmuse XT
https://www.dji.com/ca/zenmuse-xt
so maybe you could sort of reverse engineer what is already standard in the fire service.