Читать книгу Encyclopedia of Renewable Energy - James Speight G., James G. Speight - Страница 19

Analysis is the means by which any fuel (fossil or alternative, especially a biomass-derived) fuel is characterized and accuracy and precision are required otherwise the analytical data are suspect and cannot be used with any degree of certainly. This is also especially true of analytical data that is reused for commercial operations where the fuel may be sold on the basis of heat content.

Thus, the accuracy of a test is a measure of how close the test result will be to the true value of the property being measured. As such the accuracy can be expressed as the bias between the test result and the true value. However, the absolute accuracy can only be established if the true value is known. In the simplest sense, a convenient method to determine a relationship between two measured properties is to plot one against the other. Such an exercise will provide either a line fit of the points or a spread that may or may not be within the limits of experimental error. The data can then be used to determine the approximate accuracy of one or more points employed in the plot. For example, a point that lies outside the limits of experimental error (a flyer) will indicate an issue of accuracy with that test and the need for a repeat determination.

However, the graphical approach is not appropriate for finding the absolute accuracy between more than two properties. The well-established statistical technique of regression analysis is more pertinent to determining the accuracy of points derived from one property and any number of other properties. There are many instances in which relationships of this sort enable properties to be predicted from other measured properties with as good precision as they can be measured by a single test. It would be possible to examine in this way the relationships between all the specified properties of a product and to establish certain key properties from which the remainder could be predicted, but this would be a tedious task.

An alternative approach to that of picking out the essential tests in a specification using regression analysis is to take a look at the specification as a whole, and extract the essential features (termed principal components analysis).

Principal components analysis involves an examination of set of data as points in n-dimensional space (corresponding to a number (n) of original tests) and determines (first) the direction that accounts for the biggest variability in the data (first principal component). The process is repeated until n principal components are evaluated, but not all of these are of practical importance since some may be attributable purely to experimental error. The number of significant principal components shows the number of independent properties being measured by the tests considered.

Following from this, it is necessary to establish the number of independent properties that are necessary to predict product performance in service with the goals of rendering any specification more meaningful and allowing a high degree of predictability of product behavior. On a long-term approach it might be possible to obtain new tests of a fundamental nature to replace, or certainly to supplement, existing tests. In the short term, selecting the best of the existing tests to define product quality is the most beneficial route to predictability.

On the other hand, the precision of a test method is the variability between test results obtained on the same material, using a specific test method. The precision of a test is usually unrelated to its accuracy. The results may be precise, but not necessarily accurate. In fact, the precision of an analytical method is the amount of scatter in the results obtained from multiple analyses of a homogeneous sample. To be meaningful, the precision study must be performed using the exact sample and standard preparation procedures that will be used in the final method. Precision is expressed as repeatability and reproducibility.

The intra-laboratory precision or the within-laboratory precision refers to the precision of a test method when the results are obtained by the same operator in the same laboratory using the same apparatus. In some cases, the precision is applied to data gathered by a different operator in the same laboratory using the same apparatus. Thus, intra-laboratory precision has an expanded meaning insofar as it can be applied to laboratory precision.

Repeatability or repeatability interval of a test (r) is the maximum permissible difference due to test error between two results obtained on the same material in the same laboratory.

The repeatability interval (r) is, statistically, the 95% probability level or the differences between two test results are unlikely to exceed this repeatability interval more than five times in a hundred.

The inter-laboratory precision or the between-laboratory precision is defined in terms of the variability between test results obtained on the aliquots of the same homogeneous material in different laboratories using the same test method.

The term reproducibility or reproducibility interval (R) is analogous to the term repeatability but it is the maximum permissible difference between two results obtained on the same material but now in different laboratories. Therefore, differences between two or more laboratories should not exceed the reproducibility interval more than five times in a hundred.