Key points
Segregation is observed in material handling in many industries. It creates problems in many different handling steps and can lead to dust emission, and in some situations, the safety of the plant. Therefore, prediction of segregation is often considered important.
However, available modelling approaches and simulation techniques are not adequate due to computational power limitations. Therefore, The Wolfson Centre for Bulk Solids Handling Technology has recently developed Cellular Automata (CA) based modelling approach to predict segregation in bulk material handling.
The CA modelling approach has previously been used in many different other applications such as computer science, mathematics, physics, biology and micro-structure modelling. The CA model calculates property values for cells in a grid that is projected to cover the space of the silo The properties of a cell in the grid are calculated by considering the influences of the cells in close proximity.
Close proximity cells are recognised as neighbours. Generally, the CA considers all possible neighbouring cells to calculate the property values. However, influences of the neighbourhoods are much more limited and highly directional in bulk solids behaviour.
Influential segregation mechanism in free-flowing materials
The motion of bulk materials that are occupied by a cell are driven to move only by gravity, either vertically downwards or diagonally downwards on an angle of repose. Pellets and fines neither can move upwards, nor can they move sideways horizontally. Thus, for bulk solids modelling, a new neighbourhood arrangement was introduced.
To calculate the property values (in this case the level of fines) to a cell, a set of rules should have to be created. In order to formulate this set of rules, the main influential segregation mechanism should be identified. Sieve/percolation segregation mechanism was identified as influential segregation mechanism in free-flowing materials.
In the CA model, silo filling and discharge were dealt with individually but using identical principles and algorithms. In this model development, a numbering system is introduced to represent fines quantity/particle size distribution in a cell. The set of rules to imitate fines transfer mechanism is rather simple.
When a cell moves diagonally, the model checks the fines contents of the layer of the cell underneath to the cell in motion. If the underneath cell layer is saturated with fines, no fines are transferred from the cell in motion. If the underneath cell layer has less than saturated fines content, fines are transferred until the layer saturates or until the cell in motion is devoid of fines.
Results of Cellular Automata (CA) based modelling approach
Results of the CA models have shown promising outcomes. Given how simple the model was, this was a really satisfying outcome. Hence, the CA models could be an effective tool in predicting the actual segregation pattern in wood pellets heap formation. Computer solutions for 2D models were generated less than a minute.
Comparison of discharge model outcome with the experimental data showed that the developed models are extremely well performing in predicting fines profile during the discharge. This is important in determining spikes in fines during discharge, where the method can be developed to control these abrupt fines peak.
The validation of the 3D model was conducted against industrial silos which were located in Immingham port in Yorkshire, England. Two silos were selected for the validation. Simulation results: the cross sections of a silo are shown in Figure 1. Dark brown areas in figures show higher fines accumulation compared to the light brown areas. Cells with very pale brown colour are more or less coarse materials.
The output arrays of the simulations were captured and after the calibration, the calibrated model predicted outcomes against experimental data were plotted. The solid line in Figures 2 is the model predicted fines content and red stars are the experimental data points.
According to Figure 2 it is clear that the experimental data points are closely matched with the model predicted outcomes. This is significant compared to the size and quantity of the material used in the trial. For these cases, models were simulated within 3-4 hours.
In conclusion, the developed CA models are capable of predicting segregation in large industrial silos. The set of rules were applied in the construction of 2D and 3D CA models are logical but complemented by mathematical rules.
The model predicted outcomes were calibrated and compared with the experimental data and statistical test showed results of discharge experiments were agreeing to a 95 % confidence against the calibrated model predicted results.
This modelling approach could handle multiple inlets, multiple outlets, and variable fines content in the inflow streams, without further modifications to the set of rules in the models. Other than that, the 3D models are capable in handling different shaped vessels such as cylindrical, cornered edge shaped, elliptical and square.
The advantages of Cellular Automata (CA) based modelling approach
Some advantage of the modelling is identified and those are (i) This set of logical rules and arithmetic calculation can be modified easily. (ii) Since this method works at a bulk level not a particle level, the size of the cells in the modelling domain can be made relatively large, save the number of elements that have to be modelled compared to discrete element method. (iii) Since this modelling approach does not solve computationally expensive differential/partial differential equations, computation time is very low compared to the other commercially available. Newly introduced neighbourhood could also be used to deal with the other bulk solid handling related problems as well.
For more information please contact Dr Jeff Dissanayake, Research and Consultant Engineer, Wolfson Centre for Bulk Solids Handling Technology at the University of Greenwich.
S.C.D.DissanayakeMudiyanselage@greenwich.ac.uk or call 020 8331 8646
