Free Courses Fluids Structures Electronics 3D Design Optics Photonics Materials Installation Python Optimization Safety & Cybersecurity Embedded Software MBSE Digital Mission Engineering STEM NVH Student Teams Semiconductors Indo Kursus Partner Courses My Courses

[FAQ]

The accuracy of an implicit analysis depends on the convergence tolerances specified in *CONTROL_IMPLICIT_SOLUTION. The ultimate way to judge if your converged solution is accurate enough is to rerun the model with the DCTOL or/and ECTOL reduced by let’s say a factor of 10 and then compare the history of a critical response quantity to the history that was obtained with the larger tolerances. If the results are similar, it means that your original tolerances were likely good enough. If the results are different, it means that you need to lower the tolerances further. In other words, you need to see how sensitive your analysis results are to tighter tolerances. Note that it is recommended that RCTOL=1e10 and ABSTOL=1e-20 in *CONTROL_IMPLICIT_SOLUTION. The iteration data is written in the d3hsp and message files. Setting NLPRINT=3 in *CONTROL_IMPLICIT_SOLUTION will write the iteration data in greater detail, as described in the manual. The d3hsp can be loaded to LS-PrePost to visualize the implicit convergence statistics by selecting Misc>D3hsp View>Load>(pick d3hsp file)>Implicit Statistics. This allows, for example, for the user to plot the number of iterations required per time step, or the residual force norm, as well as the displacement and energy norms, at convergence per time step, or the norm values with respect to the iterations in a single time step. Visualizing the iteration data can help the user judge and adjust the convergence tolerances specified in *CONTROL_IMPLICIT_SOLUTION. For example, if an analysis takes too little iterations (like 2 or 3) to converge, it can indicate that there is room to lower the tolerances. Or, if the analysis takes too many iterations to converge for a given tolerance, it may show that the settings of *CONTROL_IMPLICIT_AUTO need to be adjusted or that the tolerances are too low. Furthermore, the absolute residual should be orders of magnitude lower than the applied load in your model. In addition, there is also the option to write the residual forces and other quantities to the binary database d3iter at every iteration of each output step. Fringing the residual forces may help identify in which areas of the model the convergence is weaker. This output is activated by setting D3ITCTL>1 in *CONTROL_IMPLICIT_SOLUTION and RESPLT=1 in *DATABASE_EXTENT_BINARY. The implicit solver will then write the d3iter database that is similar to d3plot. After the d3iter file is loaded into LS-PrePost (through File > Open > LS-DYNA Binary Plot > select the d3iter), the residual forces can be fringed through Post>FriComp>Ndv>Resultant Residual Forces.