DEM modelling and quantitative validation of flow characteristics and blending of pellets in a planar silo
Kasina, Veera Pratap Reddy
MetadataShow full item record
Blending processes in a silo minimise the fluctuations in the property of bulk solids with the blending performance being strongly influenced by the flow pattern and operating mode among other process parameters such as batch size and type of input fluctuations. An accurate prediction of flow characteristics such as flow channel boundary and velocity profiles is important for understanding and quantifying the blending performance, thereby increasing the scope for new design by minimising the number of expensive pilot scale experiments required. In this thesis, the Discrete Element Method (DEM) is deployed to predict and understand the flow characteristics and blending of cylindrical plastic pellets in a planar flat bottom silo and a multi-flow blender (a silo with an insert and a blending tube). The predictions are validated against high-resolution velocity measurements analysed using Particle Image Velocimetry (PIV) technique. A planar model silo was built to measure the flow of pellets using PIV technique. The existing GeoPIV Matlab module was customised to extract the velocity fields in the Eulerian frame of reference and its accuracy has been verified. The developed tool was then applied to quantitatively investigate the mechanism of evolution of flow in a flat bottom silo and the dependency of the state of developed flow on the depth of the planar silo. It was shown that the development of flow during discharge can be divided into two stages: a rapid upward propagation of plug flow followed by a widening of the flow channel with increasing shearing boundaries. The size of the flow channel was found to be increasing with the depth of the silo. For the 100 mm deep silo, the flow is three dimensional with significant retardation in velocity at the frontal walls, whilst a negligible retardation was found for the 20 and 40 mm deep model silos. The thickness and frontal wall friction in planar silos thus play an important role in the development of flow patterns in model silos. In this thesis, DEM model calibration relating the macro-scale bulk friction and micro- scale particle friction at different rolling friction values was developed from DEM simulations of Jenike direct shear box. During the direct shear simulation, a constant normal force was achieved with the use of a shear lid geometry made with glued spheres thereby eliminating the use of a traditional servo control function. The influence of particle rotations and rolling friction on the limiting bulk friction for different particle sliding friction coefficients was explored. The accuracy of the calibration data was assessed by simulating the flow in a flat bottom silo and comparing the model predictions of flow rate, velocity profiles and flow channel boundary with the experiments. A good quantitative agreement was found between the experiment and simulations. The DEM model predictions were also compared with the kinematic model. Following the validation of the model, it was shown that the frontal friction and rolling friction are the influential parameters in simulating the flow patterns such as semi-mass and internal flow. It was further shown that flow transits from semi-mass flow to internal flow with the increase of frontal wall friction. The drastic influence of frontal wall friction on stress, flow patterns and force chains were analysed highlighting its implications on interpretations in 2D test silos. Finally, the developed DEM and PIV tools are employed to investigate blending in a flat bottom and multi-flow blender silo for different flow patterns. The analysis showed that the blending is more effective with the internal flow when compared to semi-mass flow in a flat bottom silo, in both continuous and discontinuous modes for a variety of process conditions such as batch size, the number of recirculation and frequency of input fluctuations. An algorithm was developed to evaluate the blending performance from the spatially averaged Eulerian velocity fields. The flow in a relatively large-scale multi-flow blender comprising nearly 606,000 particles, thereby fully replicating the test silo, was simulated and the challenges in reproducing the test conditions of continuous and discontinuous modes of operation were discussed. The flow patterns and blending were first analysed from the experiments in different configurations of the insert. Using the same input parameters for the model, it was shown that the model predictions of the velocity profiles along the height of the silo are in good agreement with the experiments. Internal flow, mixed flow and mass flow were predicted for the diverging, straight and converging insert configurations respectively and the blending performance for each of these configurations suggests an optimal configuration of the blender thereby demonstrating the potential of PIV and DEM in design optimisation. The possibility of conducting the DEM simulations under increased gravity in order to reduce the computational time has also been explored.