Structural Forensic and Statistics

Post date: Jan 14, 2018 12:30:41 AM

Structural forensic is one of the hardest discipline in civil and structural engineering. Structural forensic is never about making judgment from results of non-destructive and destructive test conducted for in-situ structure. Raw data have no use and without any capacity to make judgement. It doesn't render any sense by just looking at the data.

In order to understand the overall condition of the structure, it goes beyond statistical analysis. Many engineer will stop after sum all data as in the overall mean for the structure. Some will go to the level of determining their likely characteristic strength.

My advice is not to stop there since a datum such as characteristic strength does not sum up the overall issue of the structure. Mean, median and mode tell a story about the possible skewness of the graph which tell you the accuracy of quantitative population distribution of the sample. The kurtosis highlight the precision and/or accuracy of the sample values within a range. This indicate the range of population based on sample taken (which must be representative).

On the other hand after deriving all data, it is best to calculate and correlate the the distribution to respective percentile or quantile of the sample population. This will only give you picture perfect information about your population based on your samples.

This is a general method to summarize the global sampling for structure strength for ultimate limit state. Another method which was not discussed in most literature is the sampling method based on time series analysis and construction sequencing. This is important when tracing for local delaminations due to workmanship and material handling/batching.

Subsequently this will provide a picturesque tell-tale information and very conclusive when undertaking proper structural rehabilitation processes and the consideration for material safety factor for existing structure as part of restrengthening works.