[SY-C11] A Multi-Scale Dislocation Language - Data Mining, Statistical Analysis, and Steps Towards a Community-Driven Data Base
Invited
Dislocations are simplistic objects: they are one-dimensional, their motion is constrained by the crystallography and they are surrounded by a stress field that decays with 1/r. However, once dislocations start to interact with themselves or with other microstructures, their collective behavior becomes extremely complex. This - despite the apparent simplicity of the individual object - is still far from completely being understood.
Simulation methods, such as the Molecular Dynamics (MD) or the Discrete Dislocation Dynamics (DDD) have been very successfull in predicting the evolution of dislocation microstructures along with the resulting structure-property relations. However, up to now, there is no common "language" that allows to directly compare different dislocation structures - not even if they are obtained from the same simulation method. This makes statistical analysis and data mining on the level of the dislocations difficult.
An overview over state of the art methods for analyzing systems of dislocations will be given. Subsequently, the "Multi-scale Dislocation Language" (MuDiLingo) will be introduced, which allows to extract important geometrical properties of dislocations along with the corresponding energies. Based on examples from MD and DDD it will be shown (i) how this approach can be used to quantitatively compare dislocation microstructures, (ii) how valuable input for simulation methods on higher length scales can be provided by data-mining, and (iii) how MuDiLingo will enable us to create a data base of dislocation data, which might be of great benefit to the community.
Simulation methods, such as the Molecular Dynamics (MD) or the Discrete Dislocation Dynamics (DDD) have been very successfull in predicting the evolution of dislocation microstructures along with the resulting structure-property relations. However, up to now, there is no common "language" that allows to directly compare different dislocation structures - not even if they are obtained from the same simulation method. This makes statistical analysis and data mining on the level of the dislocations difficult.
An overview over state of the art methods for analyzing systems of dislocations will be given. Subsequently, the "Multi-scale Dislocation Language" (MuDiLingo) will be introduced, which allows to extract important geometrical properties of dislocations along with the corresponding energies. Based on examples from MD and DDD it will be shown (i) how this approach can be used to quantitatively compare dislocation microstructures, (ii) how valuable input for simulation methods on higher length scales can be provided by data-mining, and (iii) how MuDiLingo will enable us to create a data base of dislocation data, which might be of great benefit to the community.