Effect of 85 Years of Service on Mechanical Properties of Timber Roof Members. Part III. Reliability Study

The first two parts of this series (Fridley et al. 1996a, b) presented a comprehensive experimental and analytical duration of load (DOL) study of Purdue University's Peirce Hall, an 85-year-old, turn-of-the-century building razed in 1989. In this part of the research, we approach the structure from a probabilistic standpoint. We treat the material properties and loading as random variables in order to answer the following questions: Would we have ever seen any DOL effects in Peirce Hall? If not, and recognizing that the structure was in general conformance with current design practice, are we overly conservative in current design considering DOL? If yes, how often, and how does this relate to the level of safety we currently assume in design? We found that Peirce Hall, with randomness of loading and material properties, had a finite probability of experiencing a DOL failure, but not as high as assumed in current design. That is, the probability of Peirce Hall experiencing a DOL failure was less than what would be predicted using current DOL-analysis procedures. The major reason for this lower probability of failure is not due to an "over-designed" Peirce Hall, but is contributed to differences in real loading and assumed loading. Specifically, the snow load process assumed in current DOL-analysis procedures utilizes a rectangular representation of snow events, when a triangular representation is more appropriate. The rectangular event produces greater damage and higher probabilities of failure than the triangular event.