Detailseite
Projekt Druckansicht

Entwicklung von Methoden zur Speicherung von Simulationsergebnissen aus der Diskrete-Elemente-Methode

Fachliche Zuordnung Mechanische Verfahrenstechnik
Förderung Förderung von 2016 bis 2020
Projektkennung Deutsche Forschungsgemeinschaft (DFG) - Projektnummer 328985984
 
Erstellungsjahr 2019

Zusammenfassung der Projektergebnisse

Advanced methods to store simulation results which are obtained after modelling of granular materials with the discrete element methods have been developed. These methods have been implemented using C++ programming language as a stand-alone programming module for the data saving. According to the developed saving concept, a combination of lossy and lossless compression methods and separate handling of each single particle and its properties during the results saving stage are proposed. Since each single property of each object in the modelling system is treated separately, the application of lossy compression methods with removing of unnecessary saved time points allows one to remain only required data. Consequentially, in the simulations with a relatively small saving time intervals, such approach allows one to avoid the loss of important information. Moreover, due to the usage of the combination of various compression approaches the volume of saved data is reduced significantly. To perform lossy compression, several compression algorithms based on data-segmentation techniques have been implemented in the developed saving module: the top-down time-ratio and sliding-window methods. Moreover, another compression method which is based on the wavelet-transformation procedure has been also investigated. Additionally, to perform lossless compression, the DEFLATE algorithm has been used. To increase the performance of the proposed lossy compression methods, data buffering into RAM and CPU-parallelization approach have been used in the developed subsystem. Thus, all time-dependent data generated during DEM simulations are buffered into previously allocated RAM-buffer organizing separate data blocks matched with each property of each object. When some of the data blocks are filled, parallelized lossy compression method is performed using all available cores of CPU. Therefore, a significant increase in performance is achieved. To provide fast access to the compressed data, an additional extraction method has been developed. According to this extraction method, all time-dependent data stored in the compressed file is extracted according to time order using constant unpacking time step into a separate decompressed file. Subsequently, it allows one to decrease time delays caused during the data analysis process. The developed saving module was integrated into the DEM simulation framework MUSEN. To analyse the efficiency of the developed saving module several case studies have been selected using the developed classification of problems solved with DEM. It has been observed, that the proposed saving approaches can be effectively applied only to compress certain types of the investigated scenes. Thus, the saving of scenes with homogeneous spatial and time dynamics (e.g., mixing) low compression efficiency is observed. Contrary to this, during the saving of scenes with heterogeneous spatial and/or time dynamics (e.g., particle impact, bunker emptying, agglomerate breakage) higher compression efficiency is reached. Concerning to the scenes with high frequency of the discrete events (e.g., dilute flow) much higher compression efficiency can be reached when some of the datasegmentation-based compression methods are used for lossy compression. When the wavelet-based compression method is used, significantly different results are obtained. Such distinctions are observed due to the different working principle of implemented approaches. Overall, it can be concluded that the main goals planned in the project proposal have been reached and new methods for storing of DEM simulation results have been proposed and their efficiency analyzed.

Projektbezogene Publikationen (Auswahl)

 
 

Zusatzinformationen

Textvergrößerung und Kontrastanpassung