Sagging roof 2

[JStructsteel]

So I looked a small house this am for a roof that is sagging. from the outside you can see the sheathing bowing between the rafters, and some noticeable bow in the rafters. This house is under contract for sale, and the inspector wrote it up. I checked it out, the rafters are 2x4 spanning about 12'-0" from ridge to eave. How do you justify that its been that way for 70 years, the wood is just creeping, and that its fine, when you know the design would never check.

The wood is in very good condition, i suspect the sheating is 3/8" or 1/2" max, and is just creeping over time.

Structurally Its in no worse shape than the 100 other houses built in the neighborhood. Just need to find a way to say in a report.

Suggestions?

 

[KootK]

..it can be a fact by judgement.

In my book, a fact by judgement is somewhere between:

1) An oxymoron and;

2) A professional opinion.

And that's really what I've been driving at. I think it overly simplistic to be giving the impression that historic evaluation work is binary stuff where an ethical engineer need merely "say what they can prove to be fact". Were that the case, we wouldn't be saying much of value.

I would temper such statements to something more along the lines of "say what you believe to be accurate and can provide plausible justification for". Structural evaluation of existing building involves a very real dimension of "guesswork". I feel that it is misleading to suggest otherwise.

 

[KootK]

It's not the Monty Hall problem.

I didn't say that it was the Monty hall problem. I said that it shares one important feature of the Monty Hall problem: new information is added along the way. That information being:

1) A successful performance history in the case of an existing structure and;

2) Where the car is not in the case of the Monty Hall problem.

Snow events are not fixed goats and prizes behind doors that you progressively open.

I disagree. The goats and the prizes are not in fact fixed until after they are revealed. Rather, they exist as outcomes with associated probabilities just like Schrodinger's cat and next year's snowfall.

 

[IRstuff]

2) With each additional year that passes where the actual snow load experienced does not exceed the specified snow load from #1, we have more confidence that the ultimate limit state snow load will exceed the specified snow load by a factor less than was originally assumed when the structure had no available history of performance.

Not sure I buy this line of reasoning; there have been lots of instances where nothing significant happens for many years, and then, we get hit with multiple "100 yr" events. Moreover, given climate change, it's arguable that the previous weather events were less challenging than weather events going forward.

TTFN (ta ta for now)
I can do absolutely anything. I'm an expert!

https://www.youtube.com/watch?v=BKorP55Aqvg

faq731-376 forum1529 Entire Forum list

http://www.eng-tips.com/forumlist.cfm

 

[KootK]

Not sure I buy this line of reasoning; there have been lots of instances where nothing significant happens for many years, and then, we get hit with multiple "100 yr" events. Moreover, given climate change, it's arguable that the previous weather events were less challenging than weather events going forward.

Such is the nature of probabilistic events that pretty much anything can happen at any roll of the metaphorical dice. At worst, this just means that recent and future extreme events will need to be incorporated into our statistical models going forward. And such updating was always the plan.

 

[Tomfh]

I disagree. The goats and the prizes are not in fact fixed until after they are revealed. Rather, they exist as outcomes with associated probabilities just like Schrodinger's cat and next year's snowfall.

The goats and prizes are chosen in advance, are fixed, and are known to the presenter, and are progressively used up like a deck of cards - which is the essence of the Monty Hall problem. It is probability without replacement. The future probabilities are strengthened as a door is opened.

Random quantum mechanics outcomes and future snow events are random events. Probability with replacement. You don't change the future odds by knowing the result of this year. You do potentially improve your statistical model by adding to your data set, but that's all.

 

[KootK]

Due to my own, limited knowledge of statistics, I'm likely to run out of ammunition with which to support my positions here. That said, the National Building Code of Canada contains a procedure that does just what I've described: it reduces load factors based on a demonstrated history of non-overload. I encourage anyone who questions the approach to check out both the procedure and the references that accompany it for more information. It's neat stuff.

 

[r13]

given climate change, it's arguable that the previous weather events were less challenging than weather events going forward.

It is true regionally only. Climate change causes extreme weather pattern changes around the globe. Where I live has just experienced the snowy winter and coldest spring in decades.

 

[Tomfh]

That said, the National Building Code of Canada contains a procedure that does just what I've described: it reduces load factors based on a demonstrated history of non-overload.

Yes, the structure has already been loaded to working loads and thus a proportion of the original safety factor is now redundant. We know it's good for at least working loads. So we needn't add the same amount of fat as a completely untested structure.

 

[phamENG]

KootK - maybe I'm just tired, but I'm a little confused by your wording in number 2. I'm going to take my own shot at rephrasing it. Instead of:

2) With each additional year that passes where the actual snow load experienced does not exceed the specified snow load from #1, we have more confidence that the ultimate limit state snow load will exceed the specified snow load by a factor less than was originally assumed when the structure had no available history of performance.

Should we say:

2) With each additional year that passes where the actual snow load experienced does not exceed the specified snow load from #1, the probability of a design snow even randomly occurring within the remaining service life of the structure is reduced.

This approach makes sense to me. As Tomfh mentioned, we're dealing with probability with replacement. It can occur an unknown number of times, at unknown intervals, and with an unknown upper limit. However, we can use the probabilistic forecasting to determine approximately what interval and intensity we think will occur. The probability of an event occurring within a specified period of time is proportional to the interval considered. For instance, what's the probability of a 100 year snow occurring in the next 5 minutes? Extremely low. This winter? Higher, but still pretty low. Then it'll shift until you get to considering 100 years - we'd be looking at something closer to 99% that it would occur in our considered time period. So going back to our building - as we progress through the service life the remaining service life is getting shorter. So we can say that the probability of one of these random events occurring within the remaining service life is being reduced. The chances of it happening in a given year hasn't changed, but the probability of that year being in our subject time frame does change. Statistics isn't my strongest subject either, so I don't remember the academic term for this - compounding probabilities? Something like that.

As for using up the load duration, it does add an interesting dimension to the evaluation of existing structures, doesn't it? Might even warrant its own discussion. Load duration factors do, indeed, exist to prevent creep rupture, but what happens when a building outlasts it's "design life" and what is the standard design life? In most cases, it seems like it won't matter. See page 29 and 30 of this PDF. Way up north in the ice and snow, however, you may have a problem if you regularly see design snow events. If you're regularly below the full design snow load, then I wouldn't worry about it. If you are, and there's a legitimate chance you're approaching your "limit" then you can analyze the existing for a Cd=0.9 and reinforce as required.

 

[Tomfh]

So going back to our building - as we progress through the service life the remaining service life is getting shorter.

That’s valid too. You generally design for about 10x your expected service life. 50 year life you design for about 1 in 500 year event. Less remaining life means you can design for less extreme event for the remaining period.

It’s no way near linear though. It’s not until you’re almost thru the lifespan that it makes much difference.

 

[Ron]

jstruct.....you mentioned the house is 70 years old. If it has plywood or OSB sheathing, then those were added at a much later date. Houses built 70 years ago had individual dimensional lumber as the roof substrate.....similar to the photos KootK showed.

[ ]

 

[KootK]

Firstly, I feel that a public service announcement is in order here regarding the NBCC Chapter L provisions. Based on the turn that this conversation has taken, one might get the impression that it's principally about justifying lower specified loads. It's not. Rather, it's a wholistic approach that integrates many aspects of structural reliability including:

1) load variation
2) excess structural capacity
3) performance history
4) field / plan verification that details aren't load path abominations
....
....
75) in a very real way, society effectively deciding "meh, maybe we just don't care that much after all for this particular situation".

I'm going to take my own shot at rephrasing it.

Your point is interesting and valid in my opinion. It is not, however, an accurate restatement of my position. I'll try again, by way of example.

1) Assume a structure located where the code flat roof design snow load is 30 PSF.

2) Assume a structure where analysis has indicated that the roof is only capable of supporting a 20 PSF snow load.

3) With each passing year of successful performance, does it not become increasingly probable that some combination of the following is true:

a) The reserve capacity of the roof has been underestimated and/or;

b) The flat roof snow load specific to this site and this building should be lower than the 30 PSF value determined elsewhere in the state?

That's it. Similar thinking is apparent in the NBCC Chapter L provision shown below.

In my opinion, it is not correct to be asking:

"What will be the peak snow load next year at the meteoritical station where we collect our data?"

But, rather.

"What site specific, peak snow load should be considered given that this structure has 20 years of service lift left and we know that it hasn't collapsed in the last 30 yrs even though analysis would indicate that it is 10 PSF under capacity?"

Those are very different questions, the latter incorporating new information only made available with the introduction of a successful building performance history.

 

[Ron247]

If the house you are describing is in an area controlled by Roof Live rather than Snow, that is the primary reason it has stood for 70 years. Where I live, I have seen at least a 100 roofs that would have failed in a 20 psf snow load area but have stood here for many years. Some of these roofs are less than 10 years old. See the photo below of a newer house. Note the ridge board running front to rear only has rafters on one side. Has not failed and shows no obvious signs of problems at this time. However, it has never had more than the weight of shingles, decking and probably one person on it since the temporary bracing was removed. While it has some vertical braces (not many) it has nothing to inhibit sideways movement.

I am fairly sure 2x4 rafters are maxed out at about 7'. In the past I have had to pass judgement on roofs such as you depict.

"It has functioned for years with the loads that have been imposed on it but that is no guarantee larger loads will not be imposed on it. It currently only provides x% of what a modern code requires."

 

[LittleInch]

Depends who you're writing it for to a certain extent as to how you frame the recomendation.

The seller wants a simple clean bill of health, nothing wrong, no problems, you don't need to do anything type of report.

The buyer / mortgage company wants something similar, but with an obvious sag, you may want to add something along the lines of

"Doesn't / wouldn't comply with current code design, no current signs of failure / distress, but may benefit in in future from some strengthening to arrest any further sagging which may occur over time or after a particularly high snow load". This likely to cost in the region of $xxx.

Then leave the two to haggle a bit over the price.

I bought a house with a visibly sagging roof / valley beam and cracked ceilings due to replacement of slate tiles with concrete ones ( 4 x weight) and minimal reinforcement of the roof structure. Got a report which said something like that and we negotiated a reduction in price to accommodate the works which we got done a few months later.

Remember - More details = better answers
Also: If you get a response it's polite to respond to it.

 

[r13]

Structurally Its in no worse shape than the 100 other houses built in the neighborhood. Just need to find a way to say in a report.

Lack of care and maintenance.

 

[phamENG]

KootK - I see where you're going with it now, and I don't disagree - but I feel like there's a bit of a gap that has to be bridged before I completely agree.

I certainly agree that site specific micro-climate effects can reduce (or increase) probabilistic snow load, and if possible to incorporate that information it should be done. But it seems like a leap to say that, because a design snow event hasn't taken place in the first 30 years of the buildings service life, it must be estimated as too high. I think there needs to be more information. Did the measuring station for which this region is based experience the design snow, or did other surrounding areas experience it? If so, then I'm with you 100%. This site should likely be reduced. ASCE 7 uses a 50 year MRI (2% annual probability of being exceeded), so it seems there would be no reason to reduce the loading at 30 years unless there's supporting evidence that it has seen it's local peak.

I can buy the reserve capacity argument. It doesn't "feel right" but focusing on the logic - it makes sense.

 

[JStructsteel]

Not to hijack my own post, here is a picture too of a 'bracing scheme' or attempt to truss the roof. the bracing shown and the metal strap are mirrored about the center line. The come down to meet on a metal clip. I see these a somewhat original to the house, as the hardware is older. (perhaps someone had old hardware, square nuts, etc laying around)The lumber does not seem to match the rafters exactly like it was from the same lot.

The buyer is asking for money to add these braces at each rafter. In my mind that does not fix the bowing, as they are all bowed, even those with bracing.

 

[phamENG]

The only thing that will fix the bowing is to replace the roof. To prevent future, additional bowing, trussing out all of the rafters may be beneficial. Right now, the trusses are stiffer than the rafters. So as the rafters deflect, the load is transferred to the trusses through the sheathing (there is a bit of load sharing through the deck). So the truss sees a disproportionate amount of the roof load. So then it deflects. If you stiffen all of the rafters by turning them into trusses, then you'll require less load sharing to get the load to a stiff element and the aggregate load path will be a bit more spread out.

 

[KootK]

I can buy the reserve capacity argument. It doesn't "feel right" but focusing on the logic - it makes sense.

But it seems like a leap to say that, because a design snow event hasn't taken place in the first 30 years of the buildings service life, it must be estimated as too high.

The reality of what the NBCC procedure is doing is likely to feel even "less right" than what you're thinking:

1) Firstly, the method does get into load testing and, as one would think, treats that as the gold standard.

2) When load testing isn't done, they still give you a way to use the lowered load factors, albeit to lesser effect.

3) Although the relaxations are expressed as reductions to load factors, my understanding is that those reduction are meant to include the effects of a number things in addition to a possible reduction of applicable loads. The load factor was just a convenient bucket into which to slop everything into. Maybe a better expression would have been:

phi_resistance x nominal resistance >= alpha_load x specified load / kappa_voodoo. Truth in advertising.

4) The NBCC procedure, as far as I can tell, actually makes no attempt to separate or quantify the relative impacts of reserve capacity and excessive snow loads in the absence of load testing. It just says "hey, the numbers predict a failure that isn't occurring so something must be up".

As you know, statisticians are able to do something resembling magic in that they can often extract meaningful results from data without actually knowing all of the the details of what's going on. Like the difference between Newton and Schroedinger. Rest assured that I find it as intellectually unsatisfying as you do. I've tried to read some primers on structural reliability but I'm afraid the math bucks me before I get to the reveal and I lose my grasp of the real world meaning long before that.

For all its dark magic character, I do respect the NBCC method and find it to be uncommonly pragmatic for code stuff. I did my first ten in Wisconsin with lots of historic Milwaukee renovations and all the usual struggles with quantifying capacity and being the bearer of bad/expensive news. When I returned to Canada, I discovered chapter L and was like "Where has this been all my life??". I go into most renovations with a gut feel for whether or not I really think that invasive work is necessary. With chapter L, I find that the calculated answer jives with my gut feel far more often. And I like that.

 

[phamENG]

Load testing! Excellent. Good to know that's in there and actually makes it feel a little better. The link I posted early to the Timber Frame Engineering Council gets into some load testing, though it cautions against some of the accepted standards as being overly conservative and potentially harmful to the structure (going back to the cumulative damage/load duration/creep rupture discussion).

In the absence of load testing, you're right - probabilistic voodoo. Statistics is intriguing, infuriating, fascinating, and frustrating.

I keep waffling on my master's research - maybe I'll play with something in this vein as I do a good bit of historic work here in Virginia and this is a constant issue.