Approaches to the calibration and application of DEM models for cohesive bulk materials

Shaun Reid, Consulting Engineer with TUNRA Bulk Solids, shares TUNRA’s take on the current state of DEM application to industrial bulk materials handling problems. He discusses both the advantages and limitations of the method and shares key tips to ensure that DEM users are getting the most out of their analysis.

The uptake of Discrete Element Modelling (DEM) as a method for analysing bulk material handling systems has occurred at an extraordinary rate. This is largely thanks to the ever-increasing accessibility of high-performance computing hardware and advances in the functionality of simulation packages. Many advanced simulations are now possible, encompassing several million particles of complex shapes, consideration of adhesive/cohesive influences, and the ability to couple with complementary simulation techniques that enable the detailed study of fluid-particle interactions, multibody dynamics and breakage, among others.

While these advances enable the consideration of complex physical behaviour, the reality is that unless properly applied, these analyses may tend towards animation rather than simulation. In order to develop DEM simulations that yield robust physical predictions of an intended application, both sound engineering judgements and rigorous calibration are required.

DEM as an analysis tool

Any model, DEM or otherwise, is by definition an approximation of a more complex interaction. Accordingly, the development of an effective DEM approach requires appraisal of the situation to which the analysis tool is being applied and identification of potential limitations of the model in that case. The more complex the situation, the further removed the model becomes from the physics that it was designed to predict, and the more specialised calibration efforts must become. It is here that the experience of the user, not only in application of the method but also in their broader understanding of the engineering discipline, becomes key in developing practical outcomes from the modelling tool.

When appraising the suitability of DEM as a potential analysis method and setting out to develop a suitable model, several questions can be asked up front with respect to the application at hand.

  • What are the flow characteristics? Is flow occurring at high speed or low speed and is failure occurring internally or at a boundary? What pressure is acting in the regions of interest? Is cohesion or adhesion present? Are suitable flow property characterisation test results available for these conditions?
  • Is the bulk material uniform? Can the flow characteristics be adequately represented with one set of material conditions, or do several components need to be developed (representing coarse and fine components for example). Does flow tend to generate segregation or mixing? Could settling or aeration occur during handling?
  • Is the flow behaviour time dependant? Can steady state operation be achieved practically within a shortened window of simulation time?
  • What is the scale of the problem? What is the throughput and total system volume? What particle size is required to appropriately resolve the flow volumetrically? Are shaped particles necessary to develop suitable packing characteristics? Can symmetry be utilised to reduce the scale of the problem?

In considering the above, it becomes apparent that these aspects interplay in governing the most suitable approach to a simulation. Take for example the selection of particle size, this governs the number of particles required in a simulation to represent a given volume of bulk material. An instance that involves high throughputs, a long transient period and complex particle shapes, cannot be efficiently represented by the same size particle utilised for a smaller system, that reaches steady state quickly and requires only simple particle shapes.

It is also important to consider the focus of your application. Take for example a case in which extremely cohesive parameters are implemented so that chute build-up occurring over hours in practice, can be reproduced in a minute or less of DEM time. In doing so, it may be found that adverse outcomes are observed in more general flow characteristics, so that the form of the build-up could be artificially accelerated in the simulation. By instead calibrating DEM models within well understood limitations, treating the method as a tool rather than an answer, and then applying this with fundamental experience and understanding of the application, it is often the case that a more useful outcome will arise from a more simple but robust approach. Further to this, it may be necessary to partition the system of interest into sub-systems, so that appropriate calibration can be applied to each component. Where it is deemed that a physical behaviour cannot be suitably represented within the limitations of the modelling method, it is necessary to consider complementary approaches, such as physical scale modelling.

Considerations for calibration

The rise in the application of DEM has occurred alongside widespread research into calibration methods. This has resulted in some acceptance of standardised calibration approaches in less complex applications, such as the collaboratively developed white paper for the calibration of cohesionless bulk materials under rapid flow and low consolidation conditions [1]. However, once the influence of inter-particle moisture is considered, as is the case for troublesome bulk materials that exhibit adhesion and cohesion, the complex and multivariable nature of the model has seen many different calibration approaches proposed for such materials.

While various approaches for the calibration of cohesive bulk materials exist, any successful approach must replicate the handling conditions that are present in the application that is being modelled. Particular attention is required to be afforded to the dynamic interactions (at what velocity are particle interactions occurring) and the consolidation pressure acting throughout an application, so that calibration conditions can be chosen accordingly. Due to interdependencies of these aspects, it is usually the case that a model cannot be calibrated adequately for generalised conditions – the simplifications (in size/shape and mechanics) at play mean that a model calibrated correctly for a low consolidation/fast-flow application such as a transfer chute is unlikely to adequately predict the behaviour of a high-consolidation/quasi-static application such as drawdown from a hopper.

Related stories:

While DEM is often applied for the prediction of macro material behaviour (characteristic of the stream), its implementation requires the input of micro parameters (characteristic of an individual, simplified particle). Some of these parameters, for example a rolling resistance factor, are non-physical (cannot be measured in a laboratory). Therefore, it is generally the case that a calibration test is required to produce a macro characteristic that is dependent on the nature of the bulk material, an example being the angle of repose for material of a given moisture content. The test would then be repeated within the simulation domain, with necessary particle simplifications, and the micro-parameters optimised to yield the desired macro result/behaviour.

In consideration of the above, the selection of tests upon which numerical calibration is based will vary with application. In most applications, tests will be required to govern the calibration of interparticle, particle-boundary and bulk density/packing parameters. The nature of these interactions, for example whether they involve sliding or impact, must be considered as appropriate to each case. A common set of testing performed for the calibration of DEM parameters in low consolidation conditions (as required for a typical transfer chute analysis) may involve:

  • Shear-box/slump angle and repose angle testing for the calibration of interparticle parameters.
  • Adhesion or sliding versus inclination angle testing for the calibration of particle-boundary interactions.
  • Bulk density test, given that the packing of simplified DEM particles is typically less volumetrically efficient than occurs in practice and the particle density likely needs to be increased to compensate.

The above tests could be performed in either a static condition or a dynamic condition, depending on the nature of the final application. In calibration testing scopes undertaken with TUNRA, it is also commonplace to supplement calibration specific testing with a basic flow properties test regime, for the identification of worst-case moisture characteristics, compressibility, and wall friction data.

In addition to a dedicated testing and calibration procedure, it is advised that the resulting DEM material model is validated and refined as necessary, by implementation in a full-scale system that may be benchmarked to behaviour observed on site. This may involve analysis of an as-built piece of equipment, ensuring that flow characteristics and handling difficulties reported on site can be replicated in DEM, before applying the same model to assess changes in design that are aimed to improve operating performance.

In summary

It is paramount that when developing analysis scopes within which DEM is to be utilised, that the user is able to appraise the suitability of the method for the given application. By familiarising oneself with the advantages and limitations of the modelling tool, the engineer utilising DEM simulations is able to make informed design and operating decisions. It is recognised that research into calibration methods and underlying DEM models must continue, so that a more rigorous framework for material model development can be realised. However, it is critical that this development occurs in such a way that it contributes to solving industrial problems pragmatically, rather than creating additional complexity in the translation from the simulation realm to the real world. 

[1] Katterfeld, André & Coetzee, C.J. & Donohue, Tim & Fottner, Johannes & Grima, Andrew & Ramírez-Gómez, Álvaro & Ilic, Dusan & Kačianauskas, Rimantas & Necas, Jan & Schott, Dingena & Williams, Kenneth & Zegzulka, Jiri. (2019). Calibration of DEM Parameters for Cohesionless Bulk Materials under Rapid Flow Conditions and Low Consolidation. 10.13140/RG.2.2.26318.31048/1.

Send this to a friend