Friday 1st Jul, 2022

Optimising a wet screening operation using computational modelling tools

A collaboration of Newcastle businesses, including TUNRA Bulk Solids, HIC Services, and Lindsay Dynan, worked together to improve the throughput of a transfer chute at a prominent Hunter Valley coal mine by using 3D modelling technology.

During a plant optimisation analysis, a client identified a wet screening process as one of the points that required further work. This case study focuses on material flow from the scrubber onto the screen deck.

Site observations at a wet screening operation indicated a consistent loading bias of slurry onto the northern side of the screens, while a more balanced presentation of material, utilising the whole available screen area, was expected to optimise the screening process, reduce misplacement of particles and improve product moisture contents.

The system under analysis comprises a wet scrubber followed by a chute discharging onto a wet screen, shown in Figure 1.

Figure 1: Scrubber and wet screen.

 

One of the challenges is that the scrubber rotation adds considerable asymmetry to the flow, shown in Figure 2, such that the main body of slurry is situated south of the centre. This is highlighted in blue in the image. This part of the flow is relatively slow moving but is surrounded by rapid-moving slurry driven by the scrubber rotation. The lifters elevate the slurry along the scrubber wall on the northern side of the scrubber, which, upon reaching a critical height, cascades back down with high velocities.

This rapid downwards flow appears to displace the main slurry volume into the position south of the centre, with a distinct boundary line along the length of the scrubber. The area marked in blue shows a level surface, which may indicate that some separation of constituents happens within the scrubber, with liquids concentrating in the southern half and solids accumulating preferentially in the northern half and along the circumference of the scrubber.

Figure 2: Scrubber flow pattern.

The circumferential currents within the scrubber introduce non-symmetric slurry outflow across the weir. This is shown in Figure 3 in which the outflow is grouped into two regions, with a fast-moving stream on the southern side fed by the main slurry body, and a second stream on the northern side which appears to curl around towards the weir from the reversed flow returning along the north side of the scrubber hull.

Figure 3: Non-symmetric slurry flow.

Further to observations of the flow inside the scrubber, images of the subsequent screen loading were made available and demonstrated a distinct loading bias towards the northern side of the screen deck.

Numerical modelling approaches

As long-time users of numerical modelling as engineering tools, TUNRA Bulk Solids’ engineers understand that all models are limited, and identifying suitable models in spite of limitations is often an arduous job. Concerning the assessment of modelling tools, George Box coined the phrase ‘all models are wrong but some are useful’, in 1976. Bearing this aphorism in mind, two numerical approaches were investigated during this project, with the aim of identifying the most applicable one to the system under analysis. This was conducted through the evaluation of a baseline model given the complexity of the flow dynamics during the scrubber discharge and limited site data available. It is important to highlight that neither modelling approach has been intended as a predictor of operating performance, but rather, the simulations were intended for the development of relative findings between different designs when operating under identical conditions.

The minimum requirements for the baseline model include the ability to model the undesirable loading bias, as well as observable proxies of the flow dynamics assumed to be the underlying cause of the non-centrality. The baseline model was assessed based on the following qualitative criteria:

  • Biased loading outcome comparable to that observed on site
  • Relative velocities at the scrubber weir outlet comparable to that observed on site
  • Flow dynamics within the scrubber comparable to that observed on site, including:
  • One main body with a level surface;
  • Circumferential flow and lifter interaction characterised by distinct boundaries between slurry bodies.

Modelling Approach #1: Discrete Element Method (DEM) for Fluids

DEM is typically applied in the modelling of “dry” granular materials or powders, simulating systems composed of discrete particles. This case study used the DEM software Rocky 4, which typically uses a hysteretic linear spring model for normal force interactions [1] and an elastic-frictional force model in the tangential direction. Particle rolling resistance is implemented according to the type C models [2]. A linear cohesion model similar to that described by Singh et al. [3] was used to include attractive forces between particles. This method is widely used in industry and one of its main advantages for modelling the presented slurry flow is its ease of application.

DEM has been utilised successfully for a range of fluid modelling projects, highlighting the fact that not all systems require accurate modelling of all aspects of fluid motion for successful assessment of flow. The major shortcomings of DEM when attempting to model liquid flow include the inability to model velocity-dependant transfer of shear stress, surface tension and liquid incompressibility.

For the DEM material model, velocity-dependant damping was simplified to a constant damping (or internal friction) value, which may yield useful results provided the system is not overly dynamic. Surface tension, on the other hand, does not play a significant role in governing flow when the liquid volume is sufficiently large to make gravity the major driving force. Furthermore, to approximate the behaviour of the two-phase slurry, multiple material models were established with frictionless to near frictionless characteristics in both inter particle and particle-boundary interactions to represent the liquid portion of the slurry. These models were assessed both individually and also coupled with a second phase of high-density particles included to be representative of the lump within the slurry, in order to assess how higher density particles may be transported throughout the slurry.

The DEM baseline model after calibration and sensitivity analysis is shown in Figure 4. It considers the ‘as-built’ geometry of the scrubber discharge chute and therefore has enabled a comparison between the simulated results and site data.

Figure 4: Non-symmetric slurry flow.
Figure 4: Non-symmetric slurry flow.

The following observations have arisen from the comparison:

  • The simulated flow is more continuous than the observed in site images;
  • The main body of slurry north of the scrubber centre was not observed;
  • There was no clear separation between north and south volumes; rather, a single continuous angle of repose was formed throughout the scrubber, which indicates that the simplifications made have greater impact on the resulting flow dynamics than desired.

Such aspects have led to unrealistic flow patterns within the weir, deviating from site observations. This can be shown by the opposite locations of the fast/slow moving streams when compared with the site images, as observed on the detail in Figure 4.

Based on these significant differences between the site observations and the flow observed in the simulation, the DEM model was not pursued further.

Modelling Approach #2: Smoothed Particle Hydrodynamics (SPH)

The second modelling approach selected was SPH, as implemented by Crespo et al. [4], with the open-source software DualSPHysics. SPH is typically used for simulating the mechanics of continuum media, and uses Navier-Stokes equations, which describe the motion of viscous fluids.

A two-phase consisting of water and sediments was used to model flow behaviour. Whilst the water phase is governed by known macro parameters (viscosity and density), the solids phase required calibration to derive the material parameters. These included:

  • Yield stress (m): to ensure that stresses may be transferred when material is at rest;
  • Power law index (n): controls the behaviour of shear stress as a function of shear rate;
  • Density;
  • Viscosity.

The general flow characteristics predicted by the SPH approach are included in Figure 5 and were deemed to represent site observations satisfactorily. The flow effects described can be observed in the simulation and a bias of the solid components of the slurry flow towards the northern side of the screen can be observed. Figure 5 also highlights the comparable flow characteristics observed over the weir in both the physical and simulated cases, with a distinct high velocity, turbulent region on the southern side.

Compared to site images, two main effects appear to differ in the SPH model. Firstly, while the lifters appear to lift the main flow up to approximately the correct height within the scrubber, the reverse flow is less established than it appears to be in the site images. Secondly, no droplets are transported past the nine o’clock position and the downward spray of droplets is not captured accordingly. Sensitivity analysis showed that both the reverse flow and the droplets could be simulated using finer resolution; however, the increase in resolution from the original 50 mm to the 12 mm necessary to resolve these effects significantly increased computational expense.

Figure 5: Baseline SPH scrubber model.
Figure 5: Baseline SPH scrubber model.

Based on these preliminary observations, the SPH model was selected to be pursued further and additional analysis was performed in order to form a basis for comparison with any subsequent designs.

Figure 6 shows the location of the throughput control volumes used for loading analysis. The measuring volumes are highlighted in red. The solids mass within these boxes north and south of the centre was measured and their ratios were determined. A northern loading bias of 65 per cent was observed through the weir. This bias increased to 85 per cent in the centre control position and then decreased to 60 per cent at the chute exit point.

Figure 6: Location of throughput analysis boxes.
Figure 6: Location of throughput analysis boxes.

Design Improvements

After assessing the validity of the SPH baseline model, the model was used to assess the effects of changes to the transfer between scrubber and weir aimed at counter-acting the non-central flow conditions.

A suitable chute geometry was shown to significantly reduce biased loading throughout the parts of the chute and subsequently onto the screen. The improved design is currently awaiting installation on site.

Conclusions

Although modelling approaches can be a useful tool to advance engineering practises, all models have limitations. In this case-study, multiple numerical models were developed and assessed based on their usefulness for the problem at hand. An SPH model was found to map the material flow leading to non-central screen loading with sufficient accuracy to be useful in developing solutions to the problem. After assessing the validity of the SPH baseline model, the model was used to assess the effects of changes to the transfer between scrubber and weir aimed at counter-acting the non-central flow conditions. A suitable chute geometry was shown to significantly reduce biased loading throughout the parts of the chute and subsequently onto the screen, with the improved design now awaiting installation on site.  

References:

[1] Walton, O. and Braun, R., 1986. Stress calculations for assemblies of inelastic speres in uniform shear. Acta Mechanica, 63(1-4), pp.73-86.

[2] Ai, J., Chen, J., Rotter, J. and Ooi, J., 2011. Assessment of rolling resistance models in discrete element simulations. Powder Technology, 206(3), pp.269-282.

[3] Singh, A., Magnanimo, V., Saitoh, K. and Luding, S., 2014. Effect of cohesion on shear banding in quasistatic granular materials. Physical Review E, 90(2).

[4] Crespo, A., Domínguez, J., Rogers, B., Gómez-Gesteira, M., Longshaw, S., Canelas, R., Vacondio, R., Barreiro, A. and García-Feal, O., 2015. DualSPHysics: Open-source parallel CFD solver based on Smoothed Particle Hydrodynamics (SPH). Computer Physics Communications, 187, pp.204-216.

This article, written by Dr Jens Plinke, Consulting Engineer at TUNRA Bulk Solids, was originally published and presented at the Iron Ore Conference 2021 in Perth, WA, under the original title “Where DEM and SPH Collide – Wet Screening Optimisation Tools”.

JOIN OUR NEWSLETTER

JOIN OUR NEWSLETTER
Close