Monitoring UV output

[Marvo]

I have designed an air purifier which we've manufactured and installed several of over the last couple of years. It consists of 3 stages of air filtration through course, medium and fine 500x500mm element filters followed by a final filter stage which has 12 cassette type HEpA filters. Before the air leaves the ecology unit there is a final ultra violet cleaning stage which contains 48 x UVC PL-L 55 watt tubes which are run by 24 electronic BAG ballasts. There are six supplies off a contactor to the ballasts which are configured in groups of four.

I know it's a lot of thinking on a Thursday night so I'll attach a sketch in a few minutes but my question doesn't really concern the layout of the system. We've had a couple of problems with the UV fail alarm. The alarm is actually 3 separate 11-pin current monitors which are monitoring two circuits each. When a tube fails the ballast current drops and the idea is that the current monitor brings in the alarm in the plant room.

Problem we've had is that the current monitors don't seem to be sufficiently accurate or more that they don't seem to hold their calibration. We've tried several different current monitors from different manufacturers and we've also tried using separate CT's rather than the current monitor monitoring directly. Neither set up has proved reliable.

I'm looking at revamping the alarm design entirely to actually monitor the UV light output itself and wondered if anyone had any experience with UV-C light sensors or if anyone has any other suggestions.

 

[Marvo]

Here's a schematic;

 

[ericks]

Hi Marvo,

Your approach is a good idea. Maybe you should just look at a custom designed unit, back in South Africa, i made borehole pump control/protection units. Part of the system was current monitoring, the advantage was that i could "fine tune" everything in software...

* just to ad, such a custom unit could control the whole air purifier!

 

[Knobhead]

You don’t say if the UV lamps are in separate chambers. Is there a minimum number of lamps the unit can safely operate with?

With your set up measuring the line current for 8 lamps, it’s looking for a relatively small variance when just one lamp fails.

Like you I’ve had problems with current and voltage relays operating inconsistently. This I put down to temperature variations and therefore moved them out of the main panel. OK temperature was a constant battle in a foundry.

 

[ericks]

I think part of the problem might come from the electronic ballasts. They might be causing noise that's affecting the current sensors. A well designed microcontroller based controller/alarm should solve the problem.With 55W lamps you will most likely be able to sense a single lamp failure. I am sure you could find someone in ZA that would be able to design/make you a low cost unit...

 

[Darkwood]

As this is my kind of thing.. 'problem solving' have you considered scrapping the temperature, noise prone sensing and housing say a small solar array in the chamber ... this will generate power dependent on UV light hitting it... you could then have this monitored and should even let you know when lamp light output has fallen too far with age.

Note I'm happy to accept a % of your profits for my input if successful

Solar array -> signal conditioner -. Analogue input -> PLC control to monitor:-

Lamp failure
Light level depreciation

Voltage fluctuations giving lighting dips and spikes can be ignored by setting in a timer of say 1 minute duration before alarm raised etc etc
This would actually hide a flickering lamp though so adding logic to detect the pattern associated with a flickering lamp wouldn't be too hard.

This system is not prone normal ballast fluctuations, temp change 'within reason' and noise its the dogs danglies!

 

[Marvo]

You don’t say if the UV lamps are in separate chambers.

The lamps are mounted within a single large chamber.

Is there a minimum number of lamps the unit can safely operate with? With your set up measuring the line current for 8 lamps, it’s looking for a relatively small variance when just one lamp fails.

Each phase is supplying 16 tubes (I didn't have room to draw all the tubes in the schematic), I need to alarm at a loss or 2 tubes or more on any phase which should be 1/8th or 12.5% reduction in current. The chamber was designed to be a certain dimension to achieve an exact air velocity through it, the UVC output from the lamps was sized to this air velocity. The surplus UVC output margin that was built into the design was 15% but it would only operate satisfactorily at 15% less light output if the failures were evenly distributed throughout the chamber. If the failures were in close proximity to each other then the discharge air quality wouldn't meet the spec required hence the alarm indicating 12.5% output loss.

I think part of the problem might come from the electronic ballasts. They might be causing noise that's affecting the current sensors. A well designed microcontroller based controller/alarm should solve the problem.With 55W lamps you will most likely be able to sense a single lamp failure. I am sure you could find someone in ZA that would be able to design/make you a low cost unit...

The current sensing is done on the phases at the supply side of the contactor. The physical separation between the current monitors and the ballasts is over a meter away, the ballasts are in a separate ventilated 316 stainless enclosure. The ballasts are very low THD and almost linear load with regard to power factor. Going by their spec sheet they look very well behaved, I've no reason to suspect it's a harmonics or noise issue. Every time we have a monitor failure we plug in a new unit and after setting up the time delay and the hysteresis it works fine at least a few months or even a year.

As this is my kind of thing.. 'problem solving' have you considered scrapping the temperature, noise prone sensing and housing say a small solar array in the chamber ... this will generate power dependent on UV light hitting it... you could then have this monitored and should even let you know when lamp light output has fallen too far with age.

The solar suggestion sounds possibly promising. Do you have a specific item in mind? The problem with the UV-C environment is it's high ozone concentrations and very highly corrosive. Plastics and organic materials of any kind have a very short life expectancy, the high UVC concentrations have a pretty similar effect to fire. PVC cabling for example lasts less then a day so any sensor or solar panel would need to be fit for purpose.

I'm really open to any and all suggestions at this stage. There are UV-C sensors that work in the 185 - 253.7 nm range but they're hellish expensive...and I mean hellish, I think they're aimed at medical or even military applications. Because of the capital outlay I've been loathed to even try one as a prototype. Likewise there's ballasts (drivers) available the have the ability to communicate with the actual lamp and can send lamp health reports over an RS485 network BUS to a central SCADA system but again it's a sledge hammer to crack a walnut and cost is prohibitive.

Voltage fluctuations giving lighting dips and spikes can be ignored by setting in a timer of say 1 minute duration before alarm raised etc etc
This would actually hide a flickering lamp though so adding logic to detect the pattern associated with a flickering lamp wouldn't be too hard.

The existing current sensors we've tried have a time delay which allows for ignoring of the starting current, we set the delay at 20 seconds. The ballasts automatically detect lamp failure and shut the lamp down entirely so there's never any flickering lamps or partially lit lamps, they're either full on or full off.

 

[Darkwood]

Its just an off the cuff suggestion see if I could get the ball rolling, using the premise that solar panels come in different types some more efficient at certain wavelengths of light than others so was just first thought in my head... Ive nothing specific to mind just batting ideas for open discussion as would be done when trying to solve any problems.

I see the issues you highlight with the Ozone so how about externally mounting the solar array behind a pane that lets through UV radiation thus not exposed to corrosive Ozone effects... again all of the cuff not sure about the strength of any of the technologies needed to do this.

 

[Marvo]

TBH I'm not an expert on UVC. I think many types of glass block shortwave UV such as UVC. Any solar panel or any type of sensor used for UVC would probably be either a naked sensor or would have a quartz glass cover.

Maybe one of the solar wizards would know what output a solar panel would give in UVC light between 185 - 253.7 nm.

I like the idea very much though. It is direct sensing ie it's sensing the actual light level which is of primary concern rather than indirect sensing of the ballast energy consumption which may or may not have a direct relationship to the quantity of light. Secondly it's retasking readily and cheaply available existing technologies which usually means it won't break the bank.

I'm definitely going to dig deeper into this option.

 

[ericks]

"The current sensing is done on the phases at the supply side of the contactor. The physical separation between the current monitors and the ballasts is over a meter away, the ballasts are in a separate ventilated 316 stainless enclosure. The ballasts are very low THD and almost linear load with regard to power factor. Going by their spec sheet they look very well behaved, I've no reason to suspect it's a harmonics or noise issue. Every time we have a monitor failure we plug in a new unit and after setting up the time delay and the hysteresis it works fine at least a few months or even a year."

Harmonics are induced through the power lines, seperation won't make any difference in those cases. But from what you are saying this is most likely not your problem, you just have devices failing... so like i said a well designed monitor would solve your problem... Using a sensor inside the unit could come with other problems. If you were to have even a small amount of dust on them, that might be enough to cause issues.