Is My Masterplug RCD Safe? Please Help!

[ebee]

"Exactly. Without commenting on the actual 95%/5% figures, what the purveyors of such statistics are saying is that 95% of those who would not have survived without an RCD (i.e. '100% not survived') will survive with one."

Yes that`s what I meant

 

[ban-all-sheds]

You seem to be mixing up arithmetic with the wider subject of mathematics - I never mentioned arithmetic I simply said it was bad maths.

OK - fine.

But why did you write this:

I must admit I cannot fathom out how you got to 0.49% since adding one more working RCD to the original sample size will not make much of a dent in the overall results of the Italian data set.

 

[riveralt]

But why did you write this:

I must admit I cannot fathom out how you got to 0.49% since adding one more working RCD to the original sample size will not make much of a dent in the overall results of the Italian data set.

 

[JohnW2]

Yes I did mean that if we take a figure (any figure) as being the correct across the board failure rate, and in this instance we take the (possibly incorrect) figure of 7% then ,in this case, a failure rate of something as low as 0.49% (or certainly a lot lower than 7%) might be achievable.

Indeed. However, it's worth remembering that the mathematical basis for (using our example figures) multiplying 7% by 7% to get a probability of any two RCDs bith having failed is crucially dependent upon the assumption that the probability of the second one failing is totally independent of whether or not the first one has failed.

In the absence of that independence, that probability calculation is totally invalid. If, for example, the primary reason for RCD 'failure' were that they had been sitting in service for years and never tested or tripped (hence 'stiction'), then this might well apply almost equally to both. In other words, the conditional probability of the second one failing given that the first one had failed might be very high - ultimately approaching 100%, if both 'behaved' identically over time. If that were the case, then the probability of both RCDs failing might not be much less than that of one failing (i.e. in our example, 7%) - i.e. 0.07 x 1.0 = 0.07.

Two RCDs in tandem is worth a thought on some items and especially for outside use. For the relatively low cost of an RCD adaptor if using outdoor equipment is not lost and it just might make a bit of difference.

I agree that it's definitely worth considering - but to what extent it's worth considering depends (at least in probabilistic terms) on how 'unreliable' RCDs really are, and we don't really have anything approaching reliable statistics about that. Parachutes are pretty reliable (a hell of a lot lower a 'failure rate' than 7%), but you wouldn't catch any 'leisure parachutists'(and maybe not many professional/military ones) jumping out of an aircraft without two of them!

Kind Regards, John

 

[JohnW2]

But why did you write this:

I must admit I cannot fathom out how you got to 0.49% since adding one more working RCD to the original sample size will not make much of a dent in the overall results of the Italian data set.

I think it decidedly rotten, and totally unnecessary, for BAS you to be dragging you through this (which you described as my 'insulting your intelligence') again - but I presume he is making the same point that I did (without meaning any offence, or any reflection on your intelligence), that you clearly did not have knowledge of the mathematical process for combining probabilities.

Kind Regards, John

 

[ebee]

"In the absence of that independence, that probability calculation is totally invalid. If, for example, the primary reason for RCD 'failure' were that they had been sitting in service for years and never tested or tripped (hence 'stiction'), then this might well apply almost equally to both. In other words, the conditional probability of the second one failing given that the first one had failed might be very high - ultimately approaching 100%, if both 'behaved' identically over time. If that were the case, then the probability of both RCDs failing might not be much less than that of one failing (i.e. in our example, 7%) - i.e. 0.07 x 1.0 = 0.07."

John I agree with you entirely that if both RCDs are in the same location that "stiction" would likely effect them both in a similar manner

 

[JohnW2]

John I agree with you entirely that if both RCDs are in the same location that "stiction" would likely effect them both in a similar manner

Exactly - and, as I described, that totally invalidates the 'normal' process for calculating the probability of both failing. This is one thing to keep in mind when trying to decide whether cascading two RCDs is worth doing, particularly if they are identical devices of identical age in the same location (and even worse if they came from the same production batch).

Kind Regards, John

 

[riveralt]

In the absence of that independence, that probability calculation is totally invalid. If, for example, the primary reason for RCD 'failure' were that they had been sitting in service for years and never tested or tripped (hence 'stiction'), then this might well apply almost equally to both.

Come on John you really are stretching this beyond the limits here. You and Ebee seem to be plucking ideas out of thin air in an attempt to justify his rubbish maths.
Ebee has sought to create a probability calculation based on an assumption he cannot make because he doesn't have the data set for this type of RCD.
At least you agree the calculation is invalid.

Now he (you?) brings in the stiction issue, which is highlighted in the ESC report, as a further justification for supporting the bad maths calculation.
The his credit the OP's behaviour is quite the opposite of that highlighted in the ESC report in that he regularly tests the RCD. So that attempt to justify using the same data set is out of the window.

Furthermore, for unknown reasons both of you are now stating that this calculation can also be justified by the fact that the environmental conditions will contribute toward similar RCD failures in the home.
This has been dismissed by the report - please read it!!

If we follow Ebee's approach to its logical conclusion that means despite already having all circuit RCD protection at the board you should also be putting a plug RCD in every socket used by every appliance is ludicrous you know it. Furthermore, if one RCD goes the other will as well - absolute rubbish of the highest order.

 

[ebee]

No Riveralt.

Whilst I agree that two RCDs in the same location (same room) would be expected to suffer from similar degrees from stiction and even two RCDs in the same house/office etc etc might suffer similarly to some extent , stiction is not the only cause but it might well be a large proportion of all possible causes.
RCDs do fail and more often than we would like.
Not all causes of failure would cause an RCD to fail at the same time as another.
There are instances where one RCD might fail and another might not do so at the same time.
Different makes and different ages of unit might contribute to this too.

If someone has their own RCD for a piece of equipment I would hesitate to advise them to not replace it just because there is an upstream RCD.
If it`s causing a problem with discrimination well it might be best to not renew it, but otherwise it might just give them that second chance.

If the true figure is actually 7% then I am not insisting RCD x RCD would achieve a figure of 0.49% every time but I would expect it to shift from that 7% most of the time, perhaps significantly on some occasions.

I am not advising everybody to get additional RCDs for every appliance .
If they feel happier with two in cascade then in lots of instances it might well be safer.
RCDs are wonderful when they work.
RCDs ,sadly, would not be expected to save everybody.
There are some instances when fully functioning RCDs will not trip anyway, no matter how many you have.

 

[JohnW2]

Come on John you really are stretching this beyond the limits here. You and Ebee seem to be plucking ideas out of thin air in an attempt to justify his rubbish maths.

I have not ever tried to 'justify his rubbish maths'. I have explained what he did (since you said that you 'could not fathom out' how he'd done the calculation) but always with the qualifier that the validity of the mechanics of the calculation was dependent on the probabilities of failure of the two devices being independent.

Furthermore, for unknown reasons both of you are now stating that this calculation can also be justified by the fact that the environmental conditions will contribute toward similar RCD failures in the home. This has been dismissed by the report - please read it!!

I don't think you can have read (or understood) properly. If I understand you, I've been saying the opposite of what you suggest. I have been saying that the more commonality there is to the environmental conditions of the two devices, the less valid/justified is a calculation of the type ebee undertook.

If we follow Ebee's approach to its logical conclusion that means despite already having all circuit RCD protection at the board you should also be putting a plug RCD in every socket used by every appliance is ludicrous you know it.

It obviously depends upon one's view of the risks involved. There's nothing conceptually wrong with suggesting that one might consider (which is all he said) a 'belt and braces' approach if one's primary protection is less than 100% guaranteed to protect. As I wrote, whether or not one decides to do that depends upon one's judgement of how great the risk is in the first place, how likely the primary protection is to fail, and what one's attitude to risk is. As you say, most of us have decided not to adopt that belt and braces (even though we have virtually no useful data on which to base that decision). You are perhaps someone who would happily jump out of a plane without a reserve 'chute, and would regard it as ludicrous to even consider having one - on the basis that you already had one level of protection.

As for attitudes to risk, if it were only fatal outcomes that concerned you, then you might as well forget about RCDs completely - since the probability of dying from electrocution in the UK (about 1 in 3 million per year) is roughly ten times less than the probability of your winning the Lottery if you buy one ticket every week (about 1 in 270,000 per year).

Furthermore, if one RCD goes the other will as well - absolute rubbish of the highest order.

That's only the 'ultimate' (obviously hypothetical) case of a very valid concept - that the more commonality there is between the nature, age and 'experiences' of two devices, the more likley it is that they will both have failed by a particular point in time - and hence the less justification there would be for a calculation such as ebee undertook.

Kind Regards, John

 

[riveralt]

No Riveralt.

Whilst I agree that two RCDs in the same location (same room) would be expected to suffer from similar degrees from stiction and even two RCDs in the same house/office etc etc might suffer similarly to some extent ,

You move from 'would' to 'might' in the same sentence as if you are not sure what you are saying - come on have the courage of your convictions and provide the evidence to support this statement.

As I said much earlier please read the ESC report on their study of RCD failures - your 7% will fly out the window and be replaced by 2.8% and even that will be zero if a proper testing routine in practiced in the household.

As exemplified by the OP, for all his uncertainties and misunderstanding of why he was using an RCD for overcurrent protection, at least he regularly completed a testing routine and as a result found his faulty RCD.

 

[JohnW2]

No Riveralt. Whilst I agree that two RCDs in the same location (same room) would be expected to suffer from similar degrees from stiction and even two RCDs in the same house/office etc etc might suffer similarly to some extent

You move from 'would' to 'might' in the same sentence as if you are not sure what you are saying ...

I don't pretend to agree with him, but he seems to be sure (and clear) about what he's saying - that two RCDs in the same room 'would' (that's far too strong for me) suffer from similar degrees of stiction and that even two RCDs in the same house/office 'might' suffer rom similar degrees.

As I said much earlier please read the ESC report on their study of RCD failures - your 7% will fly out the window and be replaced by 2.8% and even that will be zero if a proper testing routine in practiced in the household.

Both ebee and myself have said that the discussion is about concepts, that the actual figures don't matter, and that neither of us believe that the true failure rate is anything like 7% (or even 2.8%).

Kind Regards, John

 

[ban-all-sheds]

I think it decidedly rotten, and totally unnecessary, for BAS you to be dragging you through this

It was neither - riveralt criticised, and continued to criticise, others for "bad maths", so I thought it would be useful to see his explanation of why he wrote "I must admit I cannot fathom out how you got to 0.49% since adding one more working RCD to the original sample size will not make much of a dent in the overall results of the Italian data set"

 

[JohnW2]

It was neither - riveralt criticised, and continued to criticise, others for "bad maths", so I thought it would be useful to see his explanation of why he wrote "I must admit I cannot fathom out how you got to 0.49% since adding one more working RCD to the original sample size will not make much of a dent in the overall results of the Italian data set"

Fair enough. My point was that I have already observed that what he wrote (as you've quoted above) indicates that he doesn't (or, at least, didn't) know how to undertake "good maths" in this situation, but he regarded that suggestion as being an insult to his intelligence.

There is also, as you will have seen, a semantic difference of opinion between riveralt and myself, since he would regard application of correct mathematical processes to 'bad data' as being 'bad maths', whereas I wouldn't.

Whatever, it has IMO really been a case of over-reaction on his part. The point originally made by ebee was not really specifically about actual (numerical) failure rate of RCDs in service (which we really not have any reliable handle on) or the mathematical process. He was merely pointing out the general truth (which is surely obvious from common sense, without any maths) that the probability of both of a pair of RCDs failing is less than the probability of a single RCD failing. That is always going to be true to some extent (but the probability of both failing will never be lower than the figure ebee calculated, as the product of the two individual probabilities).

Kind Regards, John

 

[plugwash]

Parachutes are pretty reliable (a hell of a lot lower a 'failure rate' than 7%), but you wouldn't catch any 'leisure parachutists'(and maybe not many professional/military ones) jumping out of an aircraft without two of them!

I would point out that in most cases* the RCD is the already the third layer of protection against electric shocks.

The first layer of protection against electric shocks is the regular insulation (either through use of solid insulating materials or through designing the equipment to ensure there is an air gap) arround the conductors.

The second level of protection against electric shocks is either the use of a second layer of insulation or a metal layer that is connected to earth.

The third layer is the RCD.

* TT systems being an exception.