I removed names and companies but this is a report of an end grain installation that has gone wrong. It is ancient yellow birch end grain 1/2" thick. My recommendation is possibly afix it to a plywood subfloor and use Bostich Ultra Set adhesive. The end gran sucks up moisture and the block expands. We have 1/8" granulated cork in an amulsive based (like silicone) adhesive between each block. Then on 32" centers aluminum strip in T shape one direction and straight 1/8" x 1/2" in the other. It is a real nightmare.
Wood flooring failure: March 26, 2003
Based upon observations and reasonable assumptions the following is what I have observed and what I believe needs to be considered by whoever must make the decision of how we are to proceed. It is important to note that the material that was installed was checked over one hundred times randomly and the moisture level of that installed was between the ideal moisture ranges of 9% and 10%. The only known material that measured dryer was the first sample batch and this was not installed anywhere. The site conditions have been checked at least twice a day and have been between the ideal ranges of 38% and 42% relative humidity. The site temperatures have fallen between acceptable ranges of 70 – 75 degrees. The Ardex topping slab moisture content was below 12% when installation commenced. I believe that any or all of the issues below have contributed to the situation as it sets right now at the Karen Millen Atlanta site:
1. The installation of this material at this thickness was known by all to be an experiment all along.
a. This along with concerns of two wood flooring installers that were bidding the installation caused our concern and was the very reason that we made the moisture tests, micrometer measurements and initiated the long series of conversations, memos and e-mails.
b. We placed the responsibility into the manufacturers representative’s hands to do the installation. This was because they (Kaswell Company) were experts in end grain flooring installation but not in this exact material as this is the first know attempt to use this material in this thickness. XXXXX had direct contact with the manufacturer. This was done in an effort to reduce the risks and assure a proper installation.
2. This material was noted upon receipt of the sample material to be very sensitive to moisture. It expands and contracts at incredible ratios and that it was measuring below 6% moisture content.
a. I demonstrated to the architect on one random piece subjected to moisture for one or two seconds and left to dry grew by about 1/8” in five minutes. This piece was measuring less than 6% moisture prior to this test. A few hours later the same piece returned to its original size but then slightly cupped and diagonally slightly different. See figure C below. The black line (square) being the tile when first measured, the red line (square) being the tile when dunked in water, surface dried and let set in the open air for five minutes and the dashed blue line (parrelligram) being the tile after completely drying and returning and seemingly stabilized to +/- 9.5% moisture content.
3. The wood tiles are adhered to the Ardex topping slab. The slab does not expand and contract anywhere close to the expansion and re-contraction that the wood tiles do. Effectively the slab does not expand or contract at all for our application. When a good bond has been achieved on the bottom side of the wood tile and the tile expands and contracts due to sanding heat or presumably the introduction of moisture in the water based finish, this causes the glue between the wood and the slab to stretch back and forth along with the wood on the top side and not move at all on the bottom side. In this case there is a tremendous shearing action. Note: In all cases there has been an apparent good bond to the Ardex material.
4. On the sample that failed, it did not fail at the MDF backing board or on the bottom side of the wood tiles but rather within the glue itself. It therefore seems to indicate that the adhesive did not have sufficient elastomeric ability to accommodate this shearing action. Possibly the adhesive was subjected to a radical temperature change as a result of sanding and this affected the adhesives chemistry.
5. In the field failures it has been my observation that the failures are generally along the aluminum T. See figure A below:
a. Figure A:
6. I do not think that the aluminum T is the cause but rather an indicator of a further issue (#9 below). The material measures about 35/64” thick. (3/64” thicker than ˝”). Aggressive sanding is required to expose the top surface of the aluminum T and this may have caused the glue at these locations to have heated more than it would at other tiles that were less aggressively sanded.
7. I very strongly question #6 above as it would have tended to be more of an issue closer to the aluminum T than at the opposite edge. I observed that the failure was consistent the entire width of the tile and then absolutely no such failure on blocks directly adjacent.
8. Many times but not on all that we pulled that had failed there is a visible residue of what we are calling CNC glue on the failed tiles. This is the glue that was used to fix the tiles to the CNC laser bed when fabricating. It looks that one or more boxes that had this residue were among those that were selected to cut the kerf to allow for the bottom flange of the aluminum T. When I first seen the white residue, I asked XXXXX from XXXXX what it was and would it go away in the sanding and he said that he had seen it before on other installations and that yes it would go away when sanded. Later I seen that it in fact was removed when sanded and no discoloration was remaining.
9. In many cases where we found failures the wood tile was seen to be a pronounced measurement below the adjacent aluminum T. It is clear that the wood rose up when the sander hit it and that the wood later returned back down. There were many places that it would have been impossible for this to not be the case. The heat that is generated when sanding very aggressively would have transmitted into the aluminum. Simple physics tells us that heat will radiate away from the heat source and heat sink away from the heat within the material being heated. See figure B below.
a. I believe that the bottom flange superheated the wood tile laying on it and “boiled” the wood to rapidly expand up and the surface was removed as it rose. This is the only possible explanation
10. In that the material was cut on a CNC laser and was said to be 3-7/8” x 3-7/8” x ˝” and that it actually measured less than that in length and width and almost always 3/32” thicker than that in thickness, either the programming was wrong when the material was cut or it changed after it was cut. While it could have been an error in length and width, ˝” is such an easy number to enter in a program that it is curious at least that it was found to be consistently thicker. I have no idea what this information might mean to someone but if this is ever analyzed, it might be a factor that will lead to the cause for the failures that we have witnessed.
11. In all cases less one single tile, the failures were where the adhesive set less than 48 hours prior to sanding. Where the greatest failures were found is where the adhesive was set less than 24 hours.
a. This all might be coincidences but more likely, there must be a reason for this.
b. It might however be that the failure just happened sooner where this was the condition and will happen eventually everywhere, as it did on the sample.
12. Possibly the installation of thin end grain wood blocks should not be tried over a hard surface but rather over a plywood sub-floor. I have no evidence that this is an issue but it is a possibility and must be part of the consideration if an end grain replacement material is decided upon.
XXXXXXXXXX is happy to assist XXXXXXXXX, their consultants and suppliers by communicating all of our observations and by assisting in testing. We however cannot recommend that the material and installation be or not be accepted.
It seems that the following strikes are against keeping the end grain birch blocks now installed:
1. There is no way to determine where the CNC glue is on the bottom side of the installed floor tiles.
2. Expansion and contraction has been observed to be radical and on the sample it is clear that the adhesive failed within itself, due apparently to shearing action.
3. If accepted and it later fails, the walls and glass on the walls would need major rework to allow for demolition of the flooring and reinstallation of a replacement.
4. If accepted and it fails after the store has opened, the entire rear half of the store would be impossible for shopping for at least one week and maybe two weeks if complete replacement is decided to be needed. Also it would be dusty, noisy, and as stated in #2 it would require major repairs to adjacent finished surfaces.
5. There is no known track record on this flooring.
The reasons that weigh in favor of keeping the current floor and making minor repairs is as follows:
1. One week has already been lost in the schedule effective this Sunday March 30th. If the floor is removed and reinstalled at least another week will have been lost. These weeks are not recoverable.
2. It may work fine if the repairs are made. This however is quite hopeful as if almost any of the above issues are true, future failure might occur.
3. An extended warranty of 2-1/2 years could be sought and in speaking to XXXXXXXX he was not against this. His opinion, which seems logical, is that if it lasts one year, that there is no reason to believe that it would not last much longer.
Further testing and observations will be presented as it comes to our attention. We await a decision and direction.