Compression of families of simulation data

The project aims a data compression of simulation data from finite element and finite volume simulations. Examples are crash tests and CFD simulation in the automotive industry. The challenges are on one hand the compression of non regular and adaptively refined meshes and on the other hand a compression of functions that are defined on nodes, edges or cells as, for example, velocities, plastic stresses, temperatures and flows. It shall be investigated which lossy predictor methods can exploit the correlation in space, time and different simulation results. Examples for such methods are PCA, high pass quantisation, mesh deformations as well as lossless methods on integers (Lorenzo predictor) with a previously executed quantisation. In cooperation with an automobile company it was shown that compression rates up to a factor of 100 can be achieved depending on the given realistic accuracy.