Ad d itive m anufacturing of p rototyp e elem entsw ith p rocess interfaces for continuously op erating m anufacturing lines

Cosima Hirschberg,Mikkel Schmidt Larsen,Johan Peter Bøtker,Jukka Rantanen

Department of Pharmacy,University of Copenhagen,Universitetsparken 2,2300 Copenhagen,Denmark

Keywords:3D printing Continuous m ixing Near-infrared spectroscopy Additive m anufacturing

A B S T R A C T

1. Introd uction

The interest for continuous m anufacturing of solid dosage form s is grow ing w ithin pharm aceutical sciences.However,the transition from traditional batch to batch m anufacturing to continuous manufacturing is challenging,although continuous m anufacturing has w ell docum ented advantages[1–4].The regulatory burden related to im plem entation of these principles is m aking m any pharm aceutical com panies hesitant to change tow ards continuous m anufacturing[1].

Engineering of the m echanical parts of continuous m anufacturing lines for solid dosage form s have been investigated w ith a focus on the pow der blending process,the granulation or drying process and the tableting or capsule f illing part[4–10].It is crucial to identify critical engineering parameters to effectively implement continuous m anufacturing and potentially,experim entally investigate different engineering solutions.Continuous m ixing has been intensively studied,taking into consideration environm ental factors,like controlling humidity and tem perature,the effect of material properties and the inf luence of the setup(feeder,m ixing chamber and screw)on the m ixing process[7,11].Modifying the existing engineering solution of m echanical parts is challenging and often limited to the already existing options available from the m anufacturer of the specif ic production line.Rapid prototyping provides a fast approach for investigating different engineering solutions at sm aller scale and im plem entation of m anufacturing innovation based on com putational approaches.

Previous w ork has been focusing on assessing how different process param eters and pow der properties inf luence the m ixing result.Mean residence tim e and residence tim e distribution(RTD)are the key engineering term s used,to characterize the continuous m ixing processes.[12].Gao et al.found that the RTD is more sensitive to changes in blade speed than to the pow der feed rate into the system or the blade conf iguration.They further identif ied that a low blade speed and a low er f low rate lead to a higher m ean residence time and a w ider RTD,w hile w ith increasing feed rate the RTD becom es m ore narrow[5].A narrow RTD and a long m ean residence tim e increase the chance of achieving a consistent hom ogeneous pow der m ixture[7].Different types of formulations have different challenges during the m ixing process.High dose form ulations have typically been show n to be m ixed w ell at high m ixer speeds[13],w hereas,low dose form ulations are a challenge in pharmaceutical development and production as the concentration can be too low to be detected by in-line process analytical tools.How ever,it w as show n that during continuous m ixing the paddle orientation and the m ixer speed did not have an inf luence on the critical quality attributes of the tablets w ith a low content[14].Besides the design of the m ixer,the m aterial properties largely dom inate the m ixing perform ance.Especially to break clumps or agglomerates of cohesive pow der,a certain m inim um of shear rate is required[7].During the continuous m ixing process,real-tim e m onitoring of the m ixing is im portant to ensure a hom ogeneous blend and,if necessary,adjustm ent of the process param eters to ensure the quality of the blend.Real-tim e inline m onitoring w ith spectroscopic m ethods such as nearinfrared(NIR)spectroscopy[2,15]and Ram an spectroscopy have been proven to show good results[16].NIRspectroscopy has a w ide range of applications during solid dosage form m anufacturing.It can be used indirectly to m onitor continuous pow der f low in the feeders,by analyzing the density of the pow der[17].Granulation and drying processes can be m onitored using NIR spectroscopy[18–20].The pow der mixture can be m onitored right before tableting by placing a NIRprobe before the tablet punches[21]and the content uniform ity of tablets can be determ ined using NIRspectroscopy[22–24].Previous work has show n that the process measurements w ith an in-line NIRspectrom eter during continuous m ixing results in com parable results w ith m easurem ents after the m ixing process[25].It is,how ever,crucial that the NIRprobe is properly positioned in the system,and it might even be needed to identify and evaluate m ultiple positions in the m ixing chamber[26–29].Developing a quantitative m odel for the concentration of a given analyte is an im portant part of a functional continuous production line.Alam et al.found that a NIR calibration based on only the raw m aterial in the formulation gave a sim ilar prediction perform ance,com pared to a full factorial calibration set w ith eleven different samples[30].Also m oving the pow der blend m anually in a vessel before taking each spectrum resulted in a good calibration m ethod[31].Evaluation of new ideas for m echanical parts(screw design)or a m odif ication of a continuous mixing setup can be challenging and expensive using existing solutions.

In silico design and m odelling of pharm aceutical processes are gaining increasing interest[32].The behavior of pow der and granular m aterial can be m odelled during different unit operations and the expected outcom e can be predicted[33–35].Another approach is the in silico design of process geom etries to optimize the design and to m odify it to its ow n needs and by m odeling,identify the optim al geom etry.3D printing provides a platform for rapid prototyping of the designed geom etries and testing these in a sm all scale[36].Additionally,interfaces for process analytical tools and the behavior of m aterial of interest at the process m easurem ent interface can be optim ized w ithout destroying an existing equipment.Furtherm ore,these different in silico designs can be directly printed and new prototype ideas tested w ith m inim al cost[32].

In this study,com puter aided design(CAD)and 3D printing were f irst used to design a novel calibration setup for NIR spectroscopy.This new calibration setup enables dynam ic m ixing of the pow der sam ple during the m easurem ent of the calibration sam ples w ith a f inal goal to reduce the spectral variation between the calibration samples.Additionally,a prototype production geom etry for a continuous m ixing line w as designed and 3D printed w ith tw o interfaces for an NIR probe to m onitor the pow der blending process in this continuous mixer.This approach enables an eff icient optim ization of production geom etries including dim ensions of equipm ent and further,location of process analytical interfaces.

2. Material and m ethod s

2.1. Material

Flow Lac 100 SD(spray dried lactose m onohydrate,lactose monohydrate,Meggle Pharm,Wasserburg,Germany)and Pearlitol 100 SD(spray dried m annitol,mannitol,Barentz Ap S,Breda,The Netherlands)w ere used as m odel excipients in this study.Polylactic acid(PLA,Innof il 3D,Em m en,Netherlands)and polyvinyl alcohol(PVA,Ultim aker,Gelderm alsen,Netherlands)w ere used to 3D print the prototype elem ents of continuous production lines.

2.2. Methods

2.2.1. Sample preparation

All used pow der sam ples w ere sieved(500μm)before they w ere used in an experim ental setup.For the NIR calibration sam ples,lactose monohydrate and mannitol were weighed in different w eight ratios:0/100,20/80,40/60,60/40,80/20 and 100/0,respectively,w ith a total m ass of 100 g.The m ixtures w ere blended in a Turbula m ixer(Type T2F,System Schatz,Willy A.Bachhofen AG,Maschinenfabrik,Sw itzerland)for 2 m in at 32 rounds per m inute(rpm).

2.2.2. Powder characterization

The pow der f low ability and the w all friction w ere analyzed using a ring shear tester(RST-Xs.,Dr.Dietmar Schulze,Wolfenbüttel,Germ any).A shear cell type XS-Mr.w as used for the f low ability m easurem ents and a shear cell type XS-WM for the w all friction m easurem ents(both:Dr.Dietm ar Schulze,Wolfenbüttel,Germany).The pow der f low ability w as analyzed using a preshear of 1 k Pa w ith norm al stresses set to 200,500 and 800 Pa.The f low function coeff icient(ffc)and the bulk density w ere analyzed w ith this m ethod.The w all friction w as analyzed against a steel w all plate and a 3D printed w all plate,printed from the sam e m aterial as the later used setups.The applied norm al stresses ranked from 1400 to 200 Pa w ith a 200 Pa distance.The orientation of the print layers of the wall friction plate is not parallel to the shear m ovem ent as the w all plate w as printed using horizontal nozzle m oves as opposed to circular nozzle m oves.

The f low rate was further analyzed using a Flow Pro(SAYgroup,Finland).Approxim ately 5 m L of sam ple w as f illed in the sam ple holder and tapped against the top of the instrum ent w ith a frequency of 1 s-1.The m ass of a sam ple released through a 3 mm hole on a balance w as recorded.

The particle size distribution w as analyzed using a laser diffraction particle size analyzer(Master Sizer 2000,Malvern Instrum ents,Worcestershire,UK)equipped w ith a dry powder feeder(Scirocco 2000,Malvern Instruments,Worcestershire,UK).An air pressure of 3 bar w as used to disperse the pow der sam ples.Scanning electron m icroscopy(SEM,TM3030 Tabletop Microscope,HITACHI,Tokyo,Japan)gave an indication about the particle shape.The sam ples w ere prepared on carbon sticky stubs and coated w ith gold(Cressington sputter coater 108,Cressington Scientif ic Instrum ents,Watford,UK).The picture was taken w ith a voltage of 15 k V and 300×magnif ication.

The dynam ic vapor sorption of the excipients w as analyzed using a VTI-SA+(TA-instruments,Worcestershire,UK).The sam ple w as dried at 60°Cfor a m aximum tim e of 180 m in or until w eight equilibrium w as established(＜0.001 w t%in 5 m in).The relative hum idity w as increased from 0 to 95%and decreased to 0%in 10%steps(5%step from 90%–95%relative hum idity).Equilibrium tim e for each step w as set to 150 m in or until w eight-equilibrium w as reached(＜0.001 w t%in 5 m in).Data points w ere recorded every 2 m in or at a w eight change of 0.010 w t%.

2.2.3. 3D printing

The structures w ere designed by com puter aided design(CAD)using the online platform tinkercad.com(Autodesk,CA,US)and w ere exported as stl.f iles.A Maker Bot Replicator 2(Maker-Bot,New York,NY,US)w as used to print structures from PLA,w ith a layer height of 200μm.An Ultim aker 3 extended(Ultimaker,Geldermalsen,Netherlands)was used to print structures w ith PLA and PVA as support structure,w ith a layer height of 150μm.No post processing w as perform ed after printing.Before printing,the f iles w ere converted using Cura 2.4.0(Ultimaker,Gelderm ase,Netherlands)to create a readable f ile for the printer.

2.2.4. Near infrared spectroscopy

For spectroscopic analysis a NIR spectrom eter(NIR-256L-2.2T2)equipped w ith a ref lectance probe(7-400-SMA,f iber diam eter 400μm)and the light source LS-E-NIR(all Control Developm ents Inc.,IN,US)w as used.A spectroscopic range of 1100–2200 nm was applied to the sample.The integration tim e per spectrum w as set to 0.01 s and an average of 32 m easurem ents w as used to record one spectrum.The spectra w ere recorded using Spec 32(version 1.5.4.2,Control Developm ent Inc,IN,US).After each measurement cycle,it w as visually verif ied that pow der residues w ere not blocking the probe.

2.2.5. Calibration setup

Calibration sam ples for the NIRspectroscopic evaluation w ere prepared according to Section 2.2.1 and em ptied on a f lat surface.Spectra w ere m easured recorded at nine different positions in the pow der sam ple,by placing the NIR-probe in a 90°angle in the pow der.This method is referred to as the static m ethod.

Using the 3D printed calibration setup(Fig.1),the prem ixed sam ples w ere f illed into the sam ple cham ber.The rotation of the screw was controlled w ith a power supply(303DD,ISO-TECH,Ahaus,Germ any)connected to a m otor 919D series(MFA/Com o Drills,Kent,UK),set to 10 V,corresponding to 17 rpm.Process m onitoring w as perform ed using NIRspectroscopy and nine spectra were m easured.

2.2.6. Powder mixing

A prototype continuous m ixing setup w as designed using CAD and 3D printed(Fig.2).In the dynam ic m ixing experim ents,this setup w as used to m ix lactose m onohydrate and m annitol.Tw o identical hoppers w ere attached to the m ixing chamber and f illed w ith pow der.The rotation w as controlled w ith a power supply(303DD,ISO-TECH,Ahaus,Germany)connected to a m otor 919D series(MFA/Com o Drills,Kent,UK),set to 15 V,corresponding to 25 rpm.The sam e w eight of lactose m onohydrate and m annitol(100 g each)w as f illed into funnels,w hich could be attached to the m ixing cham ber.The integrity of the mixing screw w as evaluated visually after every experiment,to detect possible deform ations or defects in the screw.Applying an axial load of 1 kg to the screw,resulted in a deform ation of approx.1%.

Fig.1–Calibration setup left:CAD of the calibration setup,right:3D printed calibration setup,consisting of cham ber,lid,screw and backstopp er.

Fig.2–left:CAD of the continuous single-screw m ixing p rototyp e cham ber and screw,right:3D printed single-screw m ixing prototyp e setup for continuous m ixing.

2.2.7. Data processing

The obtained spectral data were transformed using MATLAB R2014a(Mathw orks,Natick,MA,US)and m ultivariate data analysis w as perform ed using SIMCA(Um etrics,Um eå,Sw eden).The spectra w ere pre-treated by calculating the f irst derivate,perform ing standard normal variate(SNV)normalization and m ean centering.A spectral range of 1440–2200 nm w as chosen.Principal com ponent analysis(PCA)w as perform ed to capture the difference betw een the used calibration sam ples and to evaluate the recording during the m ixing experim ents.Global m odels taking all calibration spectra and the m easured spectra of the sam ple w ere alw ays created including the calibration spectra.Graphs were plotted using Origin Pro 9.1(Origin Lab Cooperation,MA,US).

3. Results and d iscussion

3.1. Pow der characteristics

To understand the pow der behavior during processing it is necessary to know a range of properties at the particulate level and on the bulk level of the pow der.Different properties of the tw o model pow ders(lactose m onohydrate and mannitol)w ere assessed.The size and the shape of the particles inf luence the bulk pow der behavior.When com paring m annitol w ith lactose m onohydrate,it can be observed that m annitol had a larger particle size(d(50),Table 1).The SEM im ages(data not show n)display that both lactose m onohydrate and m annitol had spherical particle shapes.The dynam ic vapor sorption prof iles showed that both m aterials adsorb less than 0.2%w ater below a relative hum idity(RH)of 70%.The deliquesce points of lactose m onohydrate and m annitol are at 95%RH and 96%RH,respectively[37].It can be expected that below the deliquescence point the physicochemical properties of these m aterials are not dram atically affected by changes in the relative hum idity.

Table 1–Sum m ary of pow der characteristics for lactose m onohyd rate and m annitol,all results are ind icated as m ean±standard deviation(n=3/n=5 for f low rate).

The f low ability m easurem ents(ffcand f low rate)indicate that m annitol had better f low ability than lactose m onohydrate(f low rate:P＜0.05,α=0,05)(Table 1).The f low ability of the model pow ders w as essential in this study,because in the later m ixing experim ents(Section 3.3)no active pow der feeding w as used.Pow der f low in the dynam ic experim ental setup occurred by gravitational force from tw o com parable hoppers w ith the sam e design and a sam e hopper opening.Due to the higher f low rate of m annitol(Table 1),a higher concentration of m annitol can be expected in the dynam ic blending experiments.The friction between the pow der and the surrounding w all m aterial inf luences to a high degree the pow der behavior w ithin the geom etry[38].In this study,the setups w ere 3D printed using polym eric m aterials and not m anufactured,as common industrial practice,from stainless steel.Therefore,it w as im portant to understand the behavior of the m odel powders against the printing polym er,PLA.A w all plate f itting for the w all friction cell of the used ring shear tester w as designed and 3D printed.

The w all friction experim ents indicated that there w as less friction betw een lactose m onohydrate and both w all m aterials(PLA and stainless steel)than betw een m annitol and the respective wall material.With increasing normal stress during the w all friction m easurem ents,the difference in w all friction betw een the stainless steel w all plate and the 3D printed plate increases.At low norm al stresses(200 Pa),there is only a small difference between the two different wall materials(Fig.3),indicating that at low stresses,as can be present in sm all scale setups[39],the choice of w all m aterial has no practical inf luence on the pow der behavior.In this study no post treatm ent w as perform ed to sm oothen the surface of the printed w all plate.By perform ing post treatm ent of the printed w all plate and printed geom etries using solvent vapor based approaches or mechanical approaches,a smoother surface can be obtained and thereby the w all friction can be decreased[40].

A sim ilar bulk behavior of both m aterials is expected in later experim ents.Due to the lower particle–particle interactions(a higher ffcvalue),m annitol m ight express an advanced f low over lactose m onohydrate.How ever,due to the sim ilar physicochem ical and bulk properties a good m ixability of lactose monohydrate and mannitol can be expected[7,41]

3.2. Calibration setups

The spectral range of 1440–2200 nm show ed m ost variation between lactose monohydrate and mannitol and,therefore,w as identif ied as the m ost useful area for m ultivariate m odeling(Fig.4).In this range,lactose m onohydrate has the relatively sharp absorption m axim a related to w ater(1900–1950 nm).

Fig.3–Wall friction of lactose m onohyd rate and m annitol against stainless steel and 3D p rinted w all p lates p rinted w ith PLA,m ean±standard deviation(n=3).

Fig.4–SNV norm alized and m ean centered NIR sp ectra of the raw m aterials,lactose m onohydrate and mannitol at the sp ectral range 1440–2200 nm.

The calibration setup was 3D printed using 3D printers based on the fused deposition m odelling m ethod.The resolution that can be reached w ith those printers is m oderate.How ever,as the sam e printers,w ith the sam e settings w ere used throughout the study,no differences in print quality can be expected betw een the geom etries.

Calibration sam ples for NIR spectroscopic analysis w ere m easured w ith the tw o different experim ental setups.The m ain difference betw een the tw o m ethods to m easure the calibration spectra is that in the static m ethod,the NIR probe is m oved in a pow der sam ple and w hen using the calibration setup,the NIR probe is on a f ixed position,w hile the powder sam ple is being m oved.Both datasets w ere preprocessed by calculating the f irst derivate,subsequent SNV normalization and m ean centering of the data.This com bination of pre-treatm ent has been show n to be suitable for NIR spectra analysis[2].A PCA plot w as then calculated and plotted,using the f irst two principal components,to compare the perform ance of the tw o m ethods to record the calibration sam ples(Fig.5).PCA scores for NIR spectra have been show n to be a good w ay to com pare the perform ance of different recorded calibration sample sets[42].

Fig.5–A:princip al com p onent analysis(PCA)of the calibration blend s left:NIRsp ectra of the p rem ixed p ow d ers m easured at different p ositions of the blend right:NIR spectra of the m oving calibration p ow d er blend s,m easured in the 3D printed calibration d evice;B:load ings of the princip al com p onents of the correspond ing PCA m od els.

As for both the used m ethods,the f irst tw o principal components explain over 99%of the variation observed in the spectra,further com ponents w ere not taken into consideration.The loadings for both methods to record the spectra of the calibration sam ples are com parable and are related to the sam e variation in the sam ples(Fig.5B).The largest variation in the loading can be seen betw een 1900 and 2000 nm,w hich corresponds to a com bination band of w ater,related to the w ater in lactose m onohydrate.The scores plot of the PCA indicate that the data points of the spectra recorded using the 3D printed device are closer to each other in the PC1–PC2 space,com pared to the spectra m easured on a static powder bed(Fig.5A).This indicates that dynam ic m ovem ent of the pow der w hile m easuring the calibration sam ples can result in a better calibration model for NIRspectroscopy,which can result in a low er prediction errors for quantitative analysis,e.g.,w hile m onitoring pow der m ixing using for exam ple PLSregression of the data.Movem ent of pow der w hile recording the samples was show n to improve the calibration of a NIR m ethod before[31].While Berntsson et al.m oved their pow der by shaking the vessel m anually;the approach presented in this study autom ates the process and thereby m akes the recording of spectra of calibration samples independent from the operator.Furtherm ore,the 3D printed device is m imicking the m ixing cham ber in w hich the pow der blend is m onitored later,giving the sam e conditions for recording the calibration spectra and the spectra w hile mixing the pow der.

3.3. Continuous pow der mixing in the single-screw mixing prototype geometry

A continuously operating m ixing line has several specif ic requirem ents that lim it the design and setup of the line:powder is constantly fed into the blender and the blend is constantly released from the process and directly transferred into the next unit operation.Most commonly a loss in weight feeder is used to control the feed of pow der into the m ixing chamber[12].In this study,a gravitational feeder is used instead.Since it is expected that no interruption w ill occur during the production,in-line process m onitoring and control of the pow der blending process are crucial.Near IR(NIR)spectroscopy is a well-established method and broadly applied in different applications,as it is a non-destructive m ethod and can m easure through a sight glass w ithout disturbing the m aterial f low[15].In this study,a 3D printed single-screw prototype of a m ixing line w as used w ith tw o interfaces for a NIR probe(Fig.2,Position 1 and Position 2).The pow der feed into the system w as not w eight controlled,but based on gravitational f low from two similar hoppers with a similar hopper opening.Due to the sim ilar bulk properties of lactose m onohydrate and m annitol,a sim ilar behavior during the process and a good m ixability w as expected.During the experim ents,it w as observed that m annitol had better f low properties in this geom etry into the m ixing cham ber com pared to lactose m onohydrate,w hich can be explained by the better f low ability of m annitol(see Section 3.1).

Fig.6–Principal com ponent analysis(PCA)of the NIR data recorded during continuous m ixing in the continuous single-screw mix ing p rototype geom etry;Ieft:position 1 for the NIRprobe and right:position 2 for the NIRprobe.Red sym bols referring to dynam ic m ix ing sam p les and other sam p les referring to the calibration sam ples.

The NIRprobe was positioned in two different positions in the m ixing setup to m onitor the pow der blend and to explore the role of process interfacing of an in-line PAT tool.In position 1 the observed NIR spectra w ere m ainly related to m annitol and only low concentrations of lactose monohydrate can be observed in the blend(Fig.6A,left).The m annitol w as fed into the m ixing setup on the sam e side,the NIR probe w as positioned at,w hich further explains the poor quality of m ixing at this point and indicates that this position should not be used for process m onitoring.After sw itching the lactose and the m annitol feed to opposite hoppers,a high concentration of lactose could be observed(data not show n).This further indicates the w rong location of the NIRprobe and conf irm s that the tw o pow ders w ere not yet m ixed w ell at Position 1.At position 2,the lactose m onohydrate content in the blend w as up to 40%(w/w).It can be concluded from these observations,that the positioning of the NIR probe is crucial for m onitoring of continuous processes,w hich is com parable w ith other studies,w here a NIRprobe was placed at different positions in the vessel during blending[26–29].

During this study,the feed of raw m aterial w as not controlled and thereby the blend ratio w as at a sam e volum etric ratio.By choosing raw m aterials w ith similar bulk properties fast m ixing w as expected.The observed higher am ount of m annitol in the blend w as expected,due to the superior f low ability of m annitol(Table 1).Overall,the use of 3D printed prototypes is a prom ising approach for rapid prototyping of continuously operating production geom etries.The design can be changed and adapted to specif ic needs fast at a very low cost.Multiple interfaces for process analytical technology can be im plem ented in the process,thereby helping to im prove the design w ithout destroying expensive existing equipm ent.

4. Conclusion

It w as dem onstrated that com puter aided design(CAD)and 3D printing is a valuable starting point for designing and prototyping continuously operating production geom etries.These 3D printed geom etries enable fast testing of different continuous pow der m ixing setups.By designing a process interface w ith a possibility for measuring,e.g.,the NIRspectra in a dynam ic m ode gives an opportunity to m im ic different process dynam ics and directly m easure the effect on the process perform ance.The pow der m ovem ent and dynam ic blending during the measurement of calibration sam ples can also be m odif ied in custom-m ade calibration geom etries,w hich improves the calibration m odels.Furtherm ore,it w as show n that a prototype of a continuously operating single-screw m ixing line can be used to follow the m ixing process in this 3D printed setup.Hence,3D printing provides a possibility for rapid prototyping of different process interface and production equipment geom etries guiding the design of new setups and process innovation.

Conf licts of interest

The authors declare that there is no conf lict of interest.The authors alone are responsible for the content and w riting of this article.

Acknow led gm ents

This study w as funded by Innovation Fund Denm ark;Project:High Quality Dry Products w ith Superior Functionality and Stability–Q-Dry;File No:5150-00024B