Continuously validating Cableizer is a task we take serious. Our extensive test suite is a supportive tool to verify our calculations after modifications in the formulas or solver routine.
Posted on December 16, 2015
Cableizer is continuously and automatically validating its simulation results. All validation is collected in a comprehensive test suite, which allows us to ensure that modifications to our code do not affect the simulation results unexpectedly. Deviations between the calculated value and the previous result as well as between the calculated value and a reference value are analysed. The reference values are results from technical papers, books, data sheets of cable suppliers as well as results from competitors for the same cable construction and laying arrangement. The number of tests normally grows as we compare against more and more reference values.
In October, we modified the calculation of belted two- and three-core cables and introduced seven additional tests for three-core-cables. In November, we made some corrections and improvements on the calculation of screen and sheath losses. And in December, we introduced the possibility of screened three-core cables with a common sheath.
|Date||Number of tests||Maximum deviation||Standard deviation|
|September 2015||144||< 5.3 %||1.7 %|
|October 2015||151||< 11.4 %||2.3 %|
|November 2015||151||< 11.4 %||2.5 %|
|December 2015 (*)||151 (144)||< 11.8 % (5.4 %)||2.4 % (1.6 %)|
The changes in calculations for multi-core cables had a significant negative impact on the statistical data as the maximum deviation more than doubled and the standard deviation increased by 0.6 %. The deviations increased even more with the introduction of screened cables. The new tests use reference values calculated by a software from an internationally renowned competitor.
We were worried that we misinterpreted the formulas and methods given in the IEC, of which we use the new releases from 2014 and 2015. So we double-checked thoroughly and checked against the results from two technical papers from Dr. G.J. Anders from 1998 and 1999 which compared the formulas for three-core unscreened and screened cables as used in the IEC with FEM calculations and made recommendations for improved formulas for the IEC, some of which have been introduced in the new releases. The results from Cableizer are consistent with those from the two papers and we are therefore confident about the correctness and accuracy of our calculations.
We also found a significant fault in the calculation method from the competitor's software, which ignores the thermal resistivity of the filler for three-core screened cables. On the other hand, Cableizer additionally implemented the improved formulas for screened cables from the paper taking into account an effect of the thermal resistivity of the filler. This formula did not find its way into the new IEC because it is still rather cumbersome and the improvement is small since the correlation between IEC and FEM as calculated in the paper was already 0.989.References: