New Zealand Engineering 1999 March

Food & Bioprocess

Commercial Aspects of the Measurement and Analysis of the Structural Properties of Timber

Trade in timber used to be based on prices relative to sets of visual grading rules. Usually these grading rules were prescriptive and specified in terms of visual assessments of defects. It is now internationally accepted that this method of assessment was poor in sorting structural properties.

Since 1990 with the introduction of ‘limit state’ or ‘ultimate strength’ designs, there has been a move toward terms of trade being dependent on the measurement of the properties in performance terms.

The difficulties arise in the mismanagement and/or misunderstanding of the clear, precise and accurate required properties specified by the design engineers as they are translated into terms of supply and payment for consignments of timber.

The ‘trade terms’ of supply and payment are really the documentation supporting the shipping and the payments, as in letters of credit or some equivalent. These particular trade terms are necessarily definitive and absolute in all respects, down to the most minute detail. If any detail is not perfect then payment is withheld by the bank concerned.

As an example, let us consider the lower 5th percentile of the Modulus of Rupture. For timber design it is accepted that the 5th percentile should be estimated with a confidence of 75 percent. So far that sounds all very well. However, it means that there is a 75 percent chance that any estimation of the 5th percentile will be below the ‘certain value’ (population 5th percentile).

Now if the ‘terms of trade’ are worded so that the ‘certain value’ shall be achieved with 100 percent certainty, then this behoves that the supplied timber will have to have grade properties well above that specified by the original design engineer’s required properties.

In terms of Australian and New Zealand conditions, if MGP10 were specified then it would be necessary to supply MGP12 or perhaps MGP15. (In European terms - see Eurocode ENV : 1995 and EN338 : 1995 - it would be necessary to supply C27 say, to satisfy a design requirement C22).

To understand the implications of the estimation of 5th percentile, one should refer to the paper by Hunt and Bryant1 of Auckland University. Estimates of the 5th percentile depend on many factors including the sample size, variation and method of estimation. In Figure 1, the confidence bounds for estimates for the 5th percentile are plotted against sample size. These samples were taken from a population with a coefficient of variation of 40 percent. For a sample size of 6 the estimate of the population value can range from less than 0.35 to greater than 1.6 times the population value. Even for a sample size of 100 the estimate can range from less than 0.82 to greater than 1.12 times the population value.

Figure 1 (click here!): Estimates of five percentile lower from a log-normal population with a coefficient of variation of 40 percent.

In Figure 2 confidence bounds for estimates are plotted against sample size for a population with a coefficient of variation of 10 percent. For a sample size of 6 the estimate now varies from less than 0.75 to greater than 1.15 times the population value. There is still a significant range with a sample size of 100. Sawn timber members are likely to have a coefficient of variation of approximately 30 percent, therefore the range of estimates is going to be closer to that of Figure 1 than Figure 2.

Figure 2: (click here!) Estimates of five percentile lower from a log-normal population with a coeficient of variation of 10 percent

Once suggestion of a solution is that the supply and payment documents must be drafted in a manner that avoids any mention of absolute values.

Perhaps the most effective way to achieve the desired design performance would be centred around average values of Modulus of Elasticity within certain limits, ie. �xGPa depending on the circumstances.

I recommend that a discussion centred on Bryant and Hunt’s paper would be starting point towards a better understanding.

I know of timber supply contracts which have devastated trade agreements for otherwise perfectly good timber designers and perfectly good timber suppliers.

Reference:

Hunt, R, Bryant, A: Statistical Implications of Methods of Finding Characteristic Strengths. Journal of Structural Engineering, ASCE, Vol 122, No 2, February 1996.

Pat Simperingham, Professional Engineering Services