Graph-based database architecture and data preprocessing for analysis of 2D dynamic systems

This article proposes the computational approach for the analysis of dynamic systems. Traditionally such systems are presented in transfer function or state matrices forms, the second one being especially convenient for computation with tools such as MATLAB environment. The problem with such approach arises when very large number of forms of dynamic system need to be compared and analysed, as finding similarities and properties is time-consuming or outright impossible due to computational complexity problems. Due to such limitations it is impossible to analyse all possible solutions for given dynamic systems and determine their usability. As solution to such problem alternative approach is proposed, where all possible representations of dynamic system are created using fast GPGPU (General-Purpose computing on Graphics Processor Units) compatible algorithm and stored in the form of directed graphs (digraphs). Creating efficient database architecture for storage of large number (numbering in millions) of such graphs in graph-based database like Neo4j allows us to efficiently analyse a whole set of solutions to find similarities, hidden properties and limitations, while at the same time retaining the ability to store the obtained data for further analysis (especially of more complicated systems) and expand the solution size, efficiently building new solutions on base of simpler solutions stored in the database.