ARR - Annual Report

Executive Summary

This report summarizes the main achievements of the first year’s effort of development of new crystallization technologies for improved crystal size and shape control during the crystallization process. The successful crystallization process and system design requires an interdisciplinary effort, which ranges from population balance model (PBM) development of the system concept, through efficient implementation of model equations to soft-sensor development, which is required for the model predictive control (MPC) design as well. This report contains practically the first steps of these interdisciplinary developments, which already crossed each-other in certain points, but it will consist organic parts of the final, integrated system.

The system concept analyzed consists in implication of wet-milling during the crystallization, which is supposed to widen the achievable crystal size domain, and this is applied as an external equipment linked to the crystallizer by a recirculation line. The recirculation flowrate also serves as an important design parameter that can be optimized. In order to make the optimization feasible, the nonlinear equation system must be implemented efficiently. In this work the implication of the graphical processing units (GPU) is employed to speed up the solution of the high-resolution finite volume method (HR-FVM). The GPU ensured roughly 100% speedup for the simulation of integrated batch-crystallization external wet mill system for the 1D case. Also, a generic, portable crystallization modeling platform for 1D and 2D batch and continuous crystallization systems have been developed. The dynamic optimization of the integrated crystallizer – wet mill system revealed an unexpected, sequential operation: the primary nucleons created in the crystallizer are transferred to the wet mill, where the (very uncertain) population is milled down to the minimal size, during which in the crystallizer complete dissolution occurs. Then, the seed crystals are fed back to the crystallizer dynamically in controlled manner, which is able to achieve broad variety of CSD shapes.

A forward transformation, with great accent on real time applicability, is developed for 2D rod-like crystals to simulate the most probable chord length distribution (CLD) and aspect ratio distribution (ARD) that would be measured. In order to speed up the CLD and ARD calculation of arbitrary 2D CSDs artificial neural networks (ANN) are implied. This will be applied in a CLD based N-MPC of crystal size and shape, which is subject of further development.

Executive Summary

This report summarizes the work performed during the last 12 month primarily by the project student, Mr. El Hebieshy. It primarily covers a comprehensive study of size and density induced segregation.

Segregation mechanisms of powder mixtures during die filling were experimentally investigated. Binary mixtures of powders with either similar particle density but different mean particle sizes or similar particle size but different particle densities were considered to investigate segregation due to size or density differences. A segmented shoe and die were constructed to quantitatively explore, the size induced segregation and density induced segregation during die filling. The results indicated that, during die filling, depending on the shoe speed, different segregation mechanisms came into effect; the sieving segregation mechanism often occurring in hoppers was observed during all condition, where air could freely escape. The fluidization segregation mechanism was observed when the air inside the die, while being displaced by the powder, was only able to escape by permeating through the powder bulk in the shoe. It was also found that the composition ratio of the mixtures only had a minor effect on the segregation tendency for the materials considered in this study.

In addition, progress was also made in exploring rotary die filling, for which a rotatory die filling system was constructed and an extensive experimental work was conducted to compare die filling behaviours using the linear and rotatory die filling systems. A large quantity of data was collected and is currently being analysed and will be reported at IFPRI 2018 AGM. Furthermore, a suction filling system was designed and is currently under construction, it is expected that the system will be ready for research work in early 2018 and preliminary work will be reported at IFPRI 2018 AGM.

EXECUTIVE SUMMARY

Currently, there is no first-principles, general theory of intermediate dry granular flow that predicts its rheological response as a function of particle size, shape, and friction (even leaving aside adhesion, which is more challenging still). It is an open question what constitutive equations best describe such flows. Therefore, there is a need for experimental data which tests these theories, and thereby provides an improved understanding of how particle properties control the rheology of granular materials, independent of the flow geometry. Rather than using empirical relations fit to bulk data for a particular flow geometry and particles, we aim to connect grain-scale parameters to macroscale behaviors.

In this second year of effort, we have have continued our laboratory testing of one nonlocal theory cooperative model, Kamrin and Koval 2012, and added a comparison to a second model gradient model, Bouzid et al. 2013. To test the efficacy of these two models across different packing fractions and shear rates, we performed experiments in a quasi-2D annular shear cell with a fixed outer wall and a rotating inner wall, using photoelastic particles. The apparatus is designed to measure both the stress ratio µ (the ratio of shear to normal stress) and the inertial number I through the use of a torque sensor, laser-cut leaf springs, and particle-tracking. We obtain

µ(I) curves for several different packing fractions and rotation rates, and successfully find that a single set of model parameters is able to capture the full range of data collected once we account for frictional drag with the bottom plate. Our measurements confirm the prediction that there is growing lengthscale at a finite value µs, associated with a frictional yield criterion. Finally, we newly identify the physical mechanism behind this transition at µs by observing that it corresponds to a drop in the susceptibility to force chain fluctuations.

We have begun experiments testing the influence of other particle-properties, starting with particle material and shape. Our preliminary investigations have revealed that the shape of the interparticle contacts (rounded vs. angular) is an important control on µs, separate from material properties such as the coefficient of friction or elastic modulus. In addition, we observe that the model parameters will require adjustment in order to fit µ(I) and velocity profile data collected for different particle types, as expected.

Project resources: A first PhD student (Wael Ebrahim) started on the project in April 2017, his work and that of supporting Master’s student projects is reported here. Funding was also secured from the University of Leeds for a second student who will start in November 2017 (Tien Nguyen), his primary focus will be the introduction of mechanical models into the single drop drying framework. Funding has also been received from the EPSRC for a collaborative project, with Durham and Bristol Universities, looking at the fundamentals of structure formation when single drops are dried. This project has just started and wil

Objectives:

This project seeks to develop experimental and modelling methods that enable dried particle structure, properties and drying rate to be predicted based on droplet drying history. The project will focus on effects driven by boiling and look to develop material independent models which capture behaviors of industrial interest. In particular it will look to address key limitations in current understanding:

1. the impact of a non-isothermal drying history on particle structure and consequently drying rate;
2. improved measurements of material properties under the non-equilibrium conditions experienced during drying;
3. extension of models to include the mechanics of structure formation.

Approach:

Initially, material classes and materials will be identified. Novel experimental rigs and methods will be developed to allow unsteady drying behaviour to be investigated and material properties to be measured under relevant conditions. Established models will be extended to include better material models and mechanical deformation. We will bring this together into a regime map(s) which links material properties, particle size, initial moisture content and drying history to morphology.

Recent Technical Progress:

• Benchmarking dryer/drop tube operation – the pilot spray dryer from ProCepT was investigated to check the accuracy of the inlet temperature measurement, check the inlet temperature distribution and to investigate the heat loss. A significant offset in inlet temperature was recorded and a distinct profile in the inlet temperature across the system was measured. Large drops in heat loss were mitigated by insulating the dryer. The dryer will be used as a basis for the drop tube.
• Evaluation of mono-dispersed atomization technologies – several mono-dispersed atomisers were assessed, a customised atomiser from Leeds will be tested on the drop tube.
• Behaviour of HPMC (one of the model systems identified) has been mapped - the ProCept spray dryer was used to dry HPMC droplets across a range of droplet sizes and dryer temperatures. At all temperatures we saw highly deformed structures, there was no clear, discontinuous, change when the droplet exceeded the boiling temperature. However two distinct structures were noted, a highly deformed, deflated, structure and a smoother spherical structure.
• Single drop drying rig development – overall design has been defined and support and pipe work structure complete. The heater control is due to be finished in January and initial commissioning will start then.
• Models to estimate the composition distributions of atomised slurry droplets have been developed – a stochastic model has been developed that enables the composition distribution of slurry droplets to be estimated. At low concentrations of solids the solid particle number in each droplet shows a Poisson like distribution.
• Models to investigate the prediction of boiling have been developed – a CDC model has been used to estimate the bulk moisture concentration at the droplet boiling temperature. This will be extended to diffusion based models to enable the link between material properties and diffusion to be explored.
• Initial experiments into microwave droplet drying have been made – a method for material characterisation by puffing droplets using microwaves has been explored. Initial attempts have proved unsuccessful and work is in progress to develop the technique.

Future Focus - 2017

• Rig build and commission: The initial version of the single drop rig is due for commissioning Jan 2017. A mono-dispersed nozzle for the drop tube will be tested in the spray dryer Jan 2017.
• System mapping: The work with the HPMC highlighted the need to do this for a mono-dispersed drop tube system; consequently this will begin once capability in place. Opportunities to work with alternative techniques at collaborators are also being investigated.
• Material properties: Measurement and technique development due to start Q2 2017.
• Model development: Approach defined for incorporation of mechanical properties by end of Q2.
• Associated work: CFD modelling of the ProCept pilot scale spray dryer which will enable better understanding of the droplet trajectories and drying histories. (Masters project completion May 2017).

Executive Summary

Segregation model development holds promise for translation of academic research into industrial practice. Two significant issues that hamper the applicability of models in industry, however, are (1) the inherent difficulty in measuring segregation rates (especially in an experimental setting) for validation purposes and (2) the significant dearth of validated scale-up studies for these models.

In this project, we seek to alleviate these two shortcomings of segregation research through a combined theoretical, computational, and experimental program. One unique aspect of our work is that we use flow perturbations to establish an “equilibrium” between segregation and mixing in free surface granular flows in order to alter the steady-state distribution of particles. By achieving this balance between the rate of segregation and the perturbation rate, we can combine the model expressions that we are interested in testing with dramatically simplified experiments to ultimately deduce the segregation rate (and validate the expressions). Moreover, by exploring a novel view of the interplay between granular rheology and segregation, we aim to introduce a new way of structuring segregation rate models that will make them inherently more scalable than any models previously reported. As the project progresses, we expect to yield – either via adoption from the literature or through new theoretical development – (experimentally) validated segregation models that can be incorporated into device-level transport equations in order to supply quantitative prediction of segregation at process scale.

One of the long term barriers to advanced and accurate modelling of particulates is the lack of a suitable set of test particles that can be used to validate particle models. Generally the approach has been to take a specific, simplified particle system, measure the mechanical and surface properties as accurately as possible, and input these parameters into a model. The model is then used to estimate a property of the agglomerate – for instance agglomerate strength – and compared to experimental measurements on the simplified particle system. Whilst some of these approaches have produced some elegant simulation results, they often fail to produce an accurate prediction of the full distribution of behaviour of the simple particle system, let alone the behaviour of far more complex industrial powders.

There are two key limitations with the existing approach. Firstly, we collapse our experimental data to an average particle shape, average roughness, average surface energy etc. and eliminate the complexity of real particles very early in the process. The final model becomes "an average of averages", and the important effects of the structure, interactions and distributions are lost. Secondly, the destructive experimental tests can only ever test a single agglomerate in a single test condition and a single (usually unknown) orientation. The structural details of the agglomerate and the test conditions (particularly orientation) are never precisely modelled and the experimental test can never be replicated with an identical particle under identical conditions. Thus we are never sure if the model is insufficient to describe the behaviour, or if the model was accurate but the number of experimental testing replicates was insufficient to statistically converge to the average behaviour predicted by the model.

Real agglomerates produced by spray fluidised beds and high shear mixers in industrial processes have complex structures and irregular shapes which are difficult to study directly. At a basic level, general terms such as porosity or its complementary solid fraction are used to define the structure of the agglomerates [1-3]. These terms can also be related to variables such as coordination number or envelope density [4-6]. However, more advanced and useful analytic tools, such as X-ray microtomography technique, are recently available to study the structure of the agglomerates. Farber et al. [7] used X-ray microtomography to characterise pharmaceutical granules. Total porosity, pore size distribution and geometric structure were obtained by this technique. Rahmanian et al. [8] have also used X-ray microtomography to characterise the granule structure evolved in a high shear granulator. Due to complexity of the agglomerate breakage analysis in some cases, such as characterisation of internal stresses by experimental work, numerical simulation using Distinct Element Method (DEM) has been widely used by different researchers to provide a basis for sensitivity analysis of different factors affecting the agglomerate structure, and hence the breakage of agglomerates [9-11]. Golchert et al. [12, 13] for the first time studied the failure of the agglomerates with their structures characterised by X-ray micro-tomography, and their strength analysed by DEM models. The 3D spatial locations of particles of real agglomerates were obtained and implemented into the simulation code to generate simulated agglomerates. The effects of agglomerate shape and structure on breakage patterns during compression were analysed. A similar piece of work was carried out by Moreno-Atanasio et al. [14]. More recently, Dadkhah et al. [15, 16] characterised the internal morphology of agglomerates produced by a spray fluidised bed using X-ray micro-tomography. The 3D volume images of agglomerates were analysed in terms of porosity, coordination number, coordination angle. For the first time, they separated the solidified binder morphology of these agglomerates using this imaging technology. Although structural details of agglomerates can be obtained by X-ray micro-

Karen Hapgood ARR67-02 Page 4

tomography, the destructive breakage tests can only be carried out on a single sample under a single orientation.

The effect of structure details has hardly been investigated and the breakage test can never be replicated under identical conditions. The complexity of the agglomerate structure, arising from different parameters such as primary particle size distribution, void fraction, inter-particle bond characteristics and material properties of both primary particles and bonds, makes it difficult to establish a full map of agglomerate breakage regimes. Overall, the agglomerates can break in different patterns, depending on their properties and loading conditions leading to various failure modes. Several pieces of work have been done on classification of patterns of agglomerate breakage [17, 18]. Subero and Ghadiri [17] made agglomerates using glass ballotini as primary particles bonded together by bisphenol-based epoxy resin. In order to explore the effect of agglomerate structure on agglomerate impact strength, the agglomerates were made with different levels of porosity by making different number and size of the macro-voids. The particles were impacted at different impact velocities and angles. In order to elucidate the fracture patterns, the shapes of the fragments were observed. They reported different patterns of breakage for agglomerate impact breakage obtained in their work, such as localised damage, fragmentation, multiple fragmentations with localised damage and disintegration. In order to study the effects of structure on agglomerate breakage, it is desirable to produce multiple identical test agglomerates with controlled structures, and then study their breakage behaviour in detail with the aid of mathematical models and experimental instruments.

In this project, 3D printer-Objet Connex 500 is used to print multiple customised agglomerates. The Objet 500 is an eight jet "PolyJet" printer which can print multiple materials simultaneously in a single print run, including rigid or rubber-like flexible materials with well-defined mechanical properties. Liquid photopolymer is printed on a build tray to form the object and cured with UV light. It can also print a removable support gel to support overhangs and/or complicated geometry. PolyJet prints simultaneously different materials with varied mechanical properties to represent the particles and/or dried liquid bridges between the particles. There are five broad material classes available, some with sub-variations: rigid opaque materials (2 variations); rubber-like materials (3 variations); transparent materials (2 variations); a polypropylene-like material and a high temperature material. The properties of each material are well defined and detailed datasheets are available [19], specifying density, hardness, tensile strength, elongation at break, elastic modulus, water adsorption and glass transition temperature Tg (where relevant), and other properties as well as the ASTM test method used to measure each of these properties. This permits a broad spectrum of agglomerates to be produced with "tuneable" physical properties.

In year one, quasi-static compression tests and drop weight impact tests were carried out using a spherical symmetrical agglomerate to investigate the agglomerate breakage behaviour at different strain rates. Preliminary experiments to determine the influence of agglomerate orientation, bond properties and strain rates were conducted to demonstrate "proof of principle" for the approach. An updated description of this work is included in this report, and the first "proof of principle" paper has recently been published in Powder Technology in late 2016. In year two, two different agglomerates were designed (cube shaped tetrahedral agglomerate, and a spherical shaped randomly structured agglomerate) and both breakage tests and DEM modelling were conducted. This report summarises the progress to date and the remaining work for year 3.

Executive Summary

Milling is commonly deployed in many industrial sectors for intended particle size reduction. In this project, we aim to develop a robust methodology to link material grindability with particle dynamics in a mill in order to provide an innovative step--change in mill fingerprinting and optimization. This involves characterizing the stressing events that prevail in a milling operation and establishing material grindability in the context of the stressing events. The material grindability will require a detailed study of the fundamental fracture and breakage mechanisms of individual particles under different loading regimes, and how they relate to the mechanical properties and the final size distribution. This will provide the fundamental scientific basis for developing appropriate grindability measure capable of analysing particle breakage subjected to particle impact, compression, and shear etc. pertaining to a milling process, which in turn will provide the basis for an improved particle breakage model calibrated against the defined grindability.

The centrifugal impact pin mill has been selected as the first mill to be studied for this project, in collaboration with Hosokawa Micron Ltd. The work performed in Year 4 of the project is summarized here. UPZ100 pin mill experiments were conducted with the effect of rotary speed and feed rate examined. Six parameters including particle size distribution, median product size, relative size span, bond’s grinding energy, size reduction ratio and specific surface area are chosen to characterize the milling results. In particular, the relationship between relevant parameters is investigated. The grinding energy approximately follows a linear relationship with the size reduction ratio. The alumina particle requires more grinding energy as compared to the zeolite particle. The population balance modelling (PBM) is used to predict the product size distribution in the milling impact tests. As indicated by the general form of PBM, it shows that two functions, i.e. selection and breakage functions have to be considered. The capacity of PBM is exemplified by predicting product size distribution in the impact pin mill considering two simple selection and breakage functions. The coupling framework of PBM-DEM is presented considering the deficiency of PBM, which forms the platform for the follow-on work.

The focus of the recent work was on the generation of different types of structure. Random

close structured granules as presented in previous years were investigated more

detailed regarding the packing density and primary particle distance distribution inside

a granule. Additional random loose structures were generated allowing porosity as additional

phase. Porous systems were generated using two different methods: fluid bed

granulation and a sintering method. Layered systems or core/shell systems were generated

using a fluid bed; a fine particle suspension was sprayed on larger core particles.

Additionally the particle-particle distance of coarse primary particles that was presented

as suitable structure measure last year was investigated more detailed. The goal was to

develop a theoretical calculation of particle-particle distance based on the known primary

particle size distribution. Results show a good agreement between values calculated

from X-ray micro-tomography images and values from theoretical calculations. Therefore

this theoretical particle-particle distance was used as structure measure in random close

structured granules.

The dissolution measurements of different sets of experiments showed that the dissolution

speed is dependent on two factors: amount of soluble material (fraction of surface

occupied by soluble material) and pore size or phase width of soluble materials. These

two factors are strongly connected and it is difficult to investigate one without the other.

Together a high fraction of soluble material and a high phase width leads to a faster dissolution

compared to granules with smaller phase width (or pore size) and fraction.

The mechanical strength of random close structured granules showed no explicit results

and was difficult to evaluate. Random loose structured granules on the other side showed

a higher strength for lower effective porosities. These results were as expected, the development

of a network of solid bridges between primary particles leads to stable granules

if more and thicker bridges are build.

Executive Summary

The understanding and control of crystallographic polymorphism and crystal habit of organic as well as inorganic compounds is scientifically and technologically important to a number of industries. To date, however, the experimental control of polymorphs (crystalline solids with different arrangements of the same constituents) is difficult. Since a polymorph is determined at the nucleation of a crystal, methods that lead to an advanced understanding of early crystal formation pathways and mechanisms are highly desirable. Towards this aim, in this project we employ arrays of self-assembled monolayers (SAMs).

Self-assembled monolayers (SAMs) are well-defined surfaces that can be used to study the relationship between the nucleation event and the final polymorph selection. Furthermore, by tuning the substrate-crystal interface energy, potentially crystalline order of SAMs can promote the nucleation of polymorphs not accessible via solution methods. It is these two advantages, i.e. the establishment of scientific correlations between nucleation and observed polymorph and access to polymorphs not accessible via solution methods, that have led us in this project to choose heterogeneous surface nucleation via SAMs as the primary means to study polymorph selection.

In the first-year of work, we have selected three types of SAMs, two hydrophilic (carboxylic acid terminated surface and hydroxyl terminated surface) and a hydrophobic (methyl terminated) surface to investigate their ability to influence the nucleation, crystal growth, and polymorph selection of a common drug, acetaminophen (ACM). It turns out that the hydrophilic surface tends to promote the formation of the monoclinic form of ACM, while the hydrophobic surface induces the formation of the less thermodynamically stable orthorhombic form of ACM. We hypothesize that this selection is due to the energetic preference of certain crystal facets interacting with the chemically specific SAMs surface. By studying the known relationships between the structure of the crystal and the nucleating surface, we will gain insights into molecular-scale recognition events that can lead to polymorphism which is a promising step to the final goal: understanding early formation pathways of crystallographic polymorphs.