Direct solution

This is a standard Cholesky solution based on a decomposition of the matrix of the system. The advantage is that it can solve several right sides at the same time. This type of solution is effective especially for small and middle-size problems when disk swapping is not necessary. The limit depends on the size of the problem and on the size of available RAM memory.

It can be said that this solution is more convenient for most of problems.

Disadvantage of this solution may emerge with extremely large problems. The calculation time may rise significantly if RAM size is insufficient. What’s more, if the available disk space is not large enough, the problem cannot be solved at all.

If the problem is excessive and of poor numerical condition, the rounding error may be so big that it exceeds the acceptable limit. This may result in imbalance between resultants of load and reactions. The difference between the total sums of loads and reactions should not be greater than about 0.5%. But even the value of 0.1% suggests that the results may be suspicious.

Generally, the direct solver should be used only for beam structures (without any 2D members) or planar structure composed of 2D members (i.e. a plate or a wall). In other cases the direct solver should be used as a default solution method. The application of iterative solution depends on the total number of nodes, band width and memory size of the particular computer. If the direct solution leads to an excessive disk swapping, the process is slowed down sifgnificantly and the iterative solution must be employed. This solver does not require so much memory – 150 000 nodes needs about 250 MB RAM. Other reason for the application of iterative solution may be poor determinetness of the equation system. These numerical problems can result in a discrepancy between the total load and sum of reactions. If this difference is greater than 5%, a warning is issued and the direct solver should be replaced by the iterative one.