API St 2560:2003(2010) pdf download

API St 2560:2003(2010) pdf download.Reconciliation of Liquid Pipeline Quantities.
6.5.6 Meter factors usually behave in a predictable way. If operating conditions are essentially constant and wear is not excessive. meter factors may be plotted on conventional con trol charts with warning, action and tolerance limits. However. if meters are subject to variable operating conditions and’or liquids with ditk,vnt physical properties. their control charts will exhibit enough natural variation to dilute thc value of warning and action limits.
6.5.7 Meter factor patterns on control charts should be reviewed to determine if a meter (1) is about to go out of’ tolerance or (2) is developing an abnormal pattern or trend. If either of these occurs, the meter should be inspected for wear or damage. Some companies set a fixed meter factor tolerance for mandating meter repair.
6.5.8 For multi-functional meters. interpretation of control charts is not straight forward. The patterns on the charts are composites of several sub-patterns which are dictated by flow rate. temperature. pressure and liquid properties. Insofar as possible. the data for such meters should he broken into separate plots of meter factor segregated by one variable, such as liquid type. with other conditions being as nearly constant as possible.
6.5.9 Even when charts arc broken out by crude type. conventional control charts may not be adequate. because in order to get enough data with one crude type. it may be necessary to accumulate single meter factors or small groups of meter factors for a given crude which are separated from each other by significant lengths of time. As a result, each subsequent factor or group of factors may be affected by meter use and wear between factors. This leads to a trending situation. and trending charts may be required to depict the data.
6.6 TRENDING CHARTS
6.6.1 Trending charts may be used when data exhibit a def’ mite upward or downward trend and do not hover around a simple horiontal mean value. Such charts may be shown as a trending run ckLrt merely to show a trend in the data, or may resemble a control chart with lines representing average per. fni-rnance (similar to “mean”) and control limits that follow the upward or downward trend of the data.
6.6.2 Meter factor charts are often trending charts, as meter lictors generally tend to increase in a regular fashion with time due to wear in a meter.
6.6.3 An example of a trending control chart is shown on Figure X.
6.6.4 Mean and Control limit values cannot be represented by tixed-valuc hcwitontal lines on a trending control chart since the normal trend of data would soon move past the control limits. With a normal meter factor control chart, this would signal a need for some sort of action. However, with a trending chart, the system may be quite all right and the data are simply following a normal trend. Hence, mean and control limits must be calculated in a ditlerent fashion. This can be done with a mathematical procedure called ‘linear regression.’ Many computer spreadsheet programs and sonic types of hand-held calculators (e.g.. “Scientific’. ‘lngineering”. “Statistical”, etc.) have linear regression programs and can be used simply by keying in the data. A method for hand-calculating a linear regression by the “Least Squares” method is given in the Appendix A.
Linear regression yields an equation of the form:
In which y is the dependent variable (e.g., meter factor), “a” is a constant (called the zero intercep. “h” is a constant (called the X coefficient), and .v is the independent variable (e.g.. month, proving sequence. etc.).
The values of “a” and “1,” are derived from the data set and are unique to the particular data set.
The mean and control limits of trending data are represented by equations rather than fixed values.