When you have completed the configuration of EbsIdent, you may start the identification phase. For the identification, you have to repeat the following sequence several times:
Depending on the number of available data sets you can either load all data sets at one time in different profiles, or you can load one data set after the other.
The identification run is invoked by the EbsScript function
ebsIdent ();
There is no corresponding menu command for this procedure, because it would be too inconvenient to perform a large number of calculations manually.
When invoked the first time, EbsIdent creates the intermediate files required to collect data and results. The data is collected in the ".bel" files. For each identification item, a multi-dimensional array is provided, according to the specified limits and grid widths. Note that the size of this array corresponds to the factor of the grid steps of all variables. In the example mentioned above, we have 7 grid steps (from 0.5 to 1.2, with a width of 0.1) for each variable, hence, 49 grid cells. If you have more variables and a finer grid, you can easily get thousands of cells.
Each time when "EbsIdent" is called now, the input variables are checked with respect to the specified grid. If a variable lies outside of its range, EbsIdent discards this entry and does not do a calculation. If all variables are within the range, the entry is stored in the array, unless the corresponding array cell is already occupied by more than three entries. In this case, the entry is discarded as well.
If the entry was stored in the array and the number of available entries exceeds the minimum number, an adaptation calculation is done to calculate the coefficients of the requested polynomial. The new coefficients and the grid occupancies are stored in the ".ref" file. The coefficients are written (in ASCII form) in the log file as well.
During the identification phase, the whole range of the variables should be covered. Otherwise, the calculated polynomial is not valid for the complete range. On the other hand, if you use measured values, the time range of the data collection should be smaller than the typical times of changes of the components.