Powder Technology 200 (2010) 105–116

Contents lists available at ScienceDirect

Powder Technology

j ourna l homepage: www.e lsev ie r.com/ locate /powtec

Study of the effects of feed frames on powder blend properties during the filling oftablet press dies

Rafael Mendez a, Fernando Muzzio b, Carlos Velazquez a,⁎a Department of Chemical Engineering, University of Puerto Rico at Mayaguez, PO Box 9000 Mayaguez PR 00681, Puerto Ricob Department of Chemical and Biochemical Engineering, Rutgers University, 98 Brett Road, Piscataway, NJ 08854, United States

⁎ Corresponding author.E-mail address: carlos.velazquez9@upr.edu (C. Velaz

0032-5910/$ – see front matter © 2010 Elsevier B.V. Aldoi:10.1016/j.powtec.2010.02.010

a b s t r a c t

a r t i c l e i n f o

Article history:Received 15 September 2009Received in revised form 2 February 2010Accepted 15 February 2010Available online 26 February 2010

Keywords:Flow phenomenaPowder feedersDie fillingShear rate

This paper studies systematically the effect of a feed frames, a device used in rotary tablet presses to drive thepowders into compression dies, on the properties of the powders entering the tablet press dies. The workfocused on the effect of blend composition, feed frame parameters (blade speed, residence time), and rotarydie disc parameters (die disc speed, die diameter) on the flow pattern, uniformity of die filling, applied shear,and the flow properties of pharmaceutical blends. The flow pattern suggests a stratified filling of the dies andtherefore, non-uniform properties of the tablets. The amount of powder entering the dies depended on blendflow properties, feed frame speed, and dies disc speed. In addition, blend properties changed significantly asthe powder flowed through the feed frame. The flowability of lubricated blends improved significantly as thenumber of feed frame blade passes increased, decreasing in turn the RSD of the die filling weight.

quez).

l rights reserved.

© 2010 Elsevier B.V. All rights reserved.

1. Introduction

Powder compression in rotary tablet presses is used as a majorstep in the manufacture of pharmaceuticals, cosmetics, catalysts,personal care products, military ordnance, etc. The value of allproducts manufactured by compression operations is enormous; justfor pharmaceuticals, more than 70% of all pharmaceutical productssold worldwide are tablets, having an estimated market value inexcess of $300 B/yr.

In all of these manufacturing processes, consistent filling of thetablet press die with a uniform weight of powder is often both criticalto quality and rate-limiting for the entire process, and therefore,uniform die filling is a crucial control variable. The amount of powderentering the die prior to compression determines the weight of thecompressed product unit, and, for pharmaceutical products, theoverall drug content of individual tablets.

Moreover, the effect of inconsistent die filling on quality is felt inless obvious ways. Most tablet presses are constant displacementmachines, where the run of the punches is fixed. In such systems, theforce at which a product unit is compressed depends exponentially onthe amount of powder in the die. As a result, several final properties ofthe compact, including its density and porosity and the amount ofelastic stress stored in the compact, are affected by the amount ofpowder in the die. For pharmaceutical tablets, which is the mainapplication examined in this paper, several other critical qualityattributes are affected by the die filling weight and the compression

force, including mechanical strength, size (thickness), incidence ofmechanical defects, chemical and physical stability, distortion,cracking, and most notably, drug release characteristics [1].

The filling of the dies of rotary presses could occur as acombination of three-flow patterns: nose flow, bulk flow, andintermittent flow [2–4]. Nose flow dominates the filling process forfree-flowing materials, low die disc speed, and larger dies. Bulk flowoccurs for high die disc speed. Intermittent flow also occurs for highdie disc speed and for more cohesive material, such as fine AvicelPH101 with particle size of 50 µm [5]. Die filling has been simulated[6] for cohesive material where it has been demonstrated thatcohesive powders take longer to fill the dies and hence could be apotential cause of tablet weight variability.

The flow behavior into die cavities depends also on powdercharacteristics, such as size, shape, and morphology [7]. For irregularparticle shapes, the filling rate decreases [8]. When the size of the diedecreases, particle rearrangement is more difficult, and the fill densitydecreases [9].

Feed frames (a.k.a, “feed shoes”) are common mechanical devicesused to force-feed powders into tablet press dies. These devices forcethe powder into the cavity and/or aerate the powder so it behavesrather like a liquid. There exist different designs of feed frames interms of number, dimensions, and shapes of the paddles, number ofchambers, and heights. Typically, a feed frame (Fig. 1) is a boxcontaining one of more paddles. Powder enters through the top, andthe paddles force the powder to exit through a slit at the bottom,pushing it into the dies.

It is well known to operators and to process design scientists thatoperating parameters, such as the size of the die cavity (diameter anddepth), the speed of the rotating tablet press, the shape and number of

Fig. 1. Feed frame equipment and dimensions.

106 R. Mendez et al. / Powder Technology 200 (2010) 105–116

paddles, and speed of the paddles in the feed frame can have animpact on tablet weight uniformity and on other tablet properties.Other factors include: die disc speed, distance between dies, and dieshape [1]. Performance depends on powder characteristics, such asparticle size distribution, powder density, and powder cohesion.Another important parameter is the history of applied stress [5].

There are a few studies where feed frames have been examined.Sinka et al. [5] studied a feed frame and a model shoe-die-fillingsystem to understand the flow properties of powders using airpressure and vacuum to fill the dies. Large differences in the criticalvelocity, which is the maximum velocity of the filling device to fillcompletely the die were found [3,5,10,11]. A correlation between thecritical velocity needed and the powder flow properties wasestablished.

Given the preceding discussion, it would be logical to expect thatthorough studies of feed frames would have been conducted, and thatquantitative design criteria for selecting feed frame design andoperating conditions for a given powder mixture having specificflow properties (akin, for example, to the information availableregarding fluid pumps) would be available. Surprisingly, while diefilling has been examined in several contexts [1–9], feed frames havebeen seldom studied systematically.

Fig. 2. Feed frame and

In this paper, using a simulated feed frame/rotating die disccombination, we examine experimentally the effect of powder flowproperties, die size, rotary disc speed, and feed frame paddle speed ondie filling uniformity. Moreover, we begin to examine the effect ofthese process parameters on the properties of typical pharmaceuticalpowders as they flow through the feed frame. Section 2 describes theapparatus and methods used in this paper. Section 3 describes themain experimental results observed. Finally, Section 4 is devoted toconclusions.

2. Experimental procedure

2.1. Simulated feed frame and die disc system

The apparatus used in this study is composed of a standard two-stage controlled feed frame taken from a Manesty Betapress and aPlexiglass die disc with actual dimensions. The powders are fed to thefeed frame at a controlled rate using a loss in weight feeder, and exitthrough three holes of different sizes (see top left, Fig. 1). Two teflon-wheels with eight rectangular paddles stir the powders, rotatingconstantly forced by a 1/15 hp variable speed Bodine Electric gearmotor with a dc-motor controller. The motor is connected to the

die disc system.

Table 1Summary of die disc and feed frame system results for 6.9 mm die diameter.

Material Die disc speed(rpm)

Feed frame speed(rpm)

Average die weight(mg)

Average residence time(min)

Average shear(# pp)

Powder flow index Powder dilation

FF lactose Untreated 36.916 21.07329 24 321.0 61.7 197.4 35.552 18.92829 48 352.8 56.4 361.0 35.015 19.23129 72 368.4 53.6 514.6 40.605 19.54257 24 212.7 47.7 152.6 32.999 18.17457 48 258.6 39.0 249.6 33.755 17.79957 72 277.3 36.1 346.6 34.815 18.246

FF lactose+1% MgSt Untreated 36.309 20.22429 24 302.4 70.2 224.6 35.740 19.28529 48 331.6 63.9 409.0 32.740 17.66429 72 377.5 56.5 542.5 28.180 10.52357 24 210.0 50.8 162.5 31.130 19.22157 48 252.5 42.2 334.3 30.890 17.92057 72 285.2 37.9 364.2 28.930 15.165

FF lactose+3% APAP+1% MgSt Untreated 36.475 22.75829 24 296.00 67.600 216.30 – 19.23829 48 330.00 59.300 379.50 – 17.97829 72 362.40 55.700 534.90 – 13.15157 24 204.00 51.800 165.90 – 18.57857 48 234.60 45.700 292.60 – 17.01257 72 252.40 43.000 413.10 – 12.453

FF lactose+30% APAP+1% MgSt Untreated 37.136 27.47029 24 185.40 115.60 369.90 31.010 23.06329 48 221.75 95.100 608.90 25.910 21.65829 72 245.80 88.300 848.00 22.980 12.58657 24 155.60 67.300 215.40 31.660 21.49657 48 173.40 63.900 408.70 30.680 19.57657 72 198.20 54.200 520.10 27.330 14.285

107R. Mendez et al. / Powder Technology 200 (2010) 105–116

transmission and the shaft to a pinion to move the two eight-paddlewheels at the same speed. Powders move from one chamber to theother one within the feed frame by convective transport forced bythese blades.

The simulated die disc consists of a 1/8 hp Dayton DC gear motordriving an acrylic disc, which moves in clockwise direction atcontrolled speeds between 0 and 94 rpm. The radius and thicknessof the acrylic disc is 16.2 cm and 17.8 mm respectively, with 22 holessimulating dies, as illustrated in Fig. 2. The die disc moves below thefeed frame and each “die” (each hole in the rotating disc) passesacross exits 1, 2, and 3. Below the rotating disc, a stationary acrylic lidis used to close the holes for half of their rotation. As the disc rotatesbeyond the lid, the powders in the holes flow out of the dies by gravityand are collected using a variable speed conveyor belt.

2.2. Materials

The materials used in this study include Foremost Farm fast flowlactose 316 monohydrate N.F., which is a common direct compression

Fig. 3. A cross sectional front view of the feed frame and die disc at the exits.

ingredient, magnesium stearate N.F. non Bovine as lubricant (sievedbefore use), and Mallinckrodt semi-fine acetaminophen USP/para-cetamol Ph Eur powder (APAP). The particles of fast flow lactose werespherical, ranging in size from 60 to 120µm (as specified by thesupplier). Different combinations of materials were chosen to studythe effect of the flow properties in the powder behavior inside thefeed frame. A blue-colored tracer material used for residence timetests was prepared by wet granulation with fast flow lactose and bluepowder dye. After being dried, the blue tracer material was sieved toobtain a particle size distribution between 150 and 250 µm.

2.3. Blending procedure

All experiments required powder blends with known and uniformproperties to isolate the effect of shear, exerted by the feed frame, onpowder properties. Blends were mixed in a GEI Buck bin blender (ref.NC1056-A01). The first two blends consisted of fast flow lactose andtwo different concentrations of magnesium stearate (0.5 and 1%). The14 kg formulation was blended at 6 rpm for 500 s (total of 50revolutions). The second two blends consisted of fast flow lactosewith two different concentrations (3 and 30%) of APAP and 1% ofmagnesium stearate. Initially, lactose and APAP were blended at9 rpm for 768 s (total of 128 revolutions). Subsequently, themagnesium stearate was added to the mixer and blended using thesame procedure as for the first blends. After the blending procedure,the material was characterized in terms of bulk and tap density,powder flow index, and dilation.

2.4. Flow patterns and mixing effects characterization

The flow pattern inside the two chambers was the firstcharacteristic of the feed frame examined. The feed frame hopperwas filled continuously with fast flow lactose using a SchenckAccurate gravimetric feeder (Modpharma 2007). The lactose rancontinuously through the feed frame. To examine flow patterns andresidence times, a small amount of the blue powder was added to thelactose entering the feed frame at a discrete time while the feed frame

Fig. 4. Powder flow patterns inside the feed frame as function of time.

108 R. Mendez et al. / Powder Technology 200 (2010) 105–116

was rotated at 24 rpm and the die disc at 29 rpm. A Canon EOS DigitalRebel digital camera (2000 frames per second) was used to takeimages of the flow patterns. This procedure was also used forestimates of residence time; following introduction of the bluepowder into the feed frame, samples were collected from the diesas a function of time tomeasure the concentration of the blue particlesusing the Near Infrared (NIR) technique.

2.5. Experimental procedure for examination of die filling and appliedshear effects

As before, the gravimetric feeder (Modpharma 2007) was used tomaintain a continuous feed of the blends to the feed frame hopper.

Fig. 5. Feed frame mixing effect areas.

The design of experiments calls for three feed frame speeds (24, 48and 72 rpm), and three die disc speeds (29, 43 and 57 rpm). Allexperiments started by setting up the feed frame and die disc at itsrespective speeds and then the powder was fed at the necessary feedrate to maintain a constant level in the feed hopper. The steady feedrate depended on the die diameter, feed frame speed, and thesimulated die disc speed. Depending on these variables, the range offeeder flow rates for the experiments was between 30 and 100 lb/h.After the system reached steady state, approximately 2.5 kg ofpowder was collected to characterize powder properties. At the endof the experiment, the feed frame and die disc were stopped and fiveindividual samples were collected from five dies to measure theweight of the powders in the die cavity.

2.6. Quantification of powder cohesion: powder flow index

A previously developed experimental method, gravitationaldisplacement rheometer (GDR) [12–15], was used to characterizeflow properties of materials under unconfined conditions. Themagnitude of avalanches is characterized by measuring the shift inthe center of gravity of the powder bed as it avalanches within theGDR, which is composed of a rotating cylinder and its drivemechanism, both of which are mounted on a pivoted table supportedby a load cell. As avalanches take place, the shift in the center ofgravity of the powder bed alters the distribution of forces between thepivot and the load cell. The weight variations recorded by the load cellare the combined effects of the avalanche size and total displacement.

Fig. 6. Outlet blue lactose concentration response as a function of time for a step changein inlet blue lactose.

Fig. 7. Exponential model applied to the residence time experimental data.

109R. Mendez et al. / Powder Technology 200 (2010) 105–116

The cohesive forces and static friction between particles determinedthe size of the avalanches and total displacement, while the speed ofthe avalanche is controlled by dynamic friction at the shear plane andthe tensile cracking mechanism [16].

The GDR was used to quantitatively measure the flow character-istics of blends before and after passing through the feed frame. Thedrum was a cylinder measuring 20.3 cm in diameter and 25.4 cm indepth. The entire cylinder was constructed of transparent plexiglas,which allowed for observation of the flow dynamics within the drum,even though transparency is not necessary for data acquisition [12].As the drum rotates, a load cell (5 lb subminiature compression loadcell, type 13/2443-06 by Sensotec, Ohio, USA) measures the change inmoment of inertia of the powders as it avalanches. Data was acquiredfor speeds from 5 to 30 rpm to capture the relevant dynamics flow. Inall of the tests, the cylinder was loaded with powder to approximately40% by volume and rotated at 5, 10, 15, 20, 25, and 30 rpm, and theload cell sampled at a frequency of 2000 Hz for 100 s. (Data collectionbegins approximately 5 min after the cylinder is first started in orderto get the steady-state flow data and ignore the initial transient phasein which avalanches are noticeably larger.)

2.7. Powder dilation

Dilation is an important phenomenon in fine powder flowdynamics. Once any granular assembly is perturbed, it has to dilate(expand in volume) in order for the flow to commence. As flowdynamics depend on the state of consolidation of powders, the degreeof dilation a powder will exhibit as it commences flow from aconsolidated state provides an accurate and convenient measurementof its flow properties, and their effect on many processes, such as, forexample, mixing [16].

The dilation phenomenon is easily visualized in the cylinder usedfor the GDR. Once a material has failed either through tensile crackingor by sliding, it flows down the surface of the cascade. Powdercohesion plays a strong role in determining the type/size ofavalanches/failures and the flow dynamics. For completely free-flowing materials run at moderate rotation rates, the flow iscontinuous on the surface, and the flowing surface remains nearly

flat. Particles flow as individual grains, enter the cascade, flowdownward, and then re-enter the area of solid-body rotation. Dilationis minimal, just enough to allow flow [16].

For slightly cohesive materials (fast-flow lactose), particles flowfreely, resulting in nearly continuous flow in the form of waves (thatis, weak avalanches). The free surface remains nearly flat, nearlysmooth. As cohesion increases distinct avalanches characterize flow.Large portions of the material fail and flow down the surface in largeavalanches. The cohesive forces and static friction between particlesdetermine the size of avalanches, while the speed of the avalanche iscontrolled by dynamic friction at the shear plane. Faqih et al. [16]proposed that an increase in cohesion plays a key factor in increasingdilation by increasing the porosity of the bed. This suggests thatdilation is heavily dependant on cohesive forces.

Dilationmeasurements were obtained using a cylinder of the samesize as before loaded with powder again to approximately 40% byvolume. The cylinder with the powder was placed horizontally touniformly distribute the powder. The powder was compacted using areciprocating table initially designed for tap density measurements.For all tests, 500 taps were used to pre-condition the powder prior tothe dilation measurements. The cylinder was then put in the Labmill-8000 roller motor system to rotate counter-clockwise at 15 rpm. Adigital video camera was used to record the drum for approximately12 revolutions. The video was analyzed to obtain images everyrevolution and measure the changes in volume related to the originalvolume. The change in bed volume in Eq. (1) represents the powderdilation. Table 1 includes the summary of the dilation results for allthe material before entering and after leaving the feed frame.

% Increase inBedVolume =Volumenew−Volumeinitial

Volumeinitial× 100: ð1Þ

3. Results and discussion

3.1. Flow patterns inside the feed frame

The total powder flow through the feed frame should be visualizedas a combination of different sub-flows. Fig. 3 shows a cross sectionalfront view of the feed frame and the die disc and the different powder

Fig. 8. Average residence time distribution.

110 R. Mendez et al. / Powder Technology 200 (2010) 105–116

flow zones generated inside the studied feed frame. The redrectangles at exits 1 and 2 of the feed frame are open spaces betweenthe paddles and the die disc.

Fig. 4 shows how the blue powder (marker) is distributed as afunction of time (from a to f) from chamber 1 to chamber 2. The blueparticles enter the feed frame and form a circular pattern in the firstchamber. This circular movement indicates that part of the material isexposed to more than one revolution before being forced to move tochamber 2. In addition, the powder re-circulating in the first wheelmixes with the fresh material entering the system. The powder in thefirst stage becomes visibly more consolidated than the powder in thesecond one. The powders in the open space over the second wheelmoved continuously backward over the paddles mixing with thepowder pushed forward by the paddle behind creating a quasi backmixing effect. Since the height of the second chamber is twice as highas the first one, this open space allowed the backward movement. Inthe first chamber on the contrary, the powders do not have the openspace, therefore they become more compacted and hence have lesspossibility of mixing.

Fig. 9. Weight in the dies for fast flow lactose, as function of the feed frame speed andfor three different die speeds using 6.9 dies diameter.

In the area swept by both blades, the powders from chamber 1entering chamber 2 mix with the powders re-circulating in thissection of chamber 2 (Fig. 5). After aminute, amore uniform colorwasobserved in the second chamber (Fig. 4f), indicating a significantmixing effect compared to chamber 1. The possible back mixing effectdepended on the feed frame and die disc speed and the level of thepowder in this chamber.

3.2. Step change approach for residence time

Fig. 6 presents the concentration of blue lactose exiting the feedframe as a function of time. These results, which were used toestimate the residence time, demonstrate the effect of the flowpattern inside the feed frame (Fig. 4). A quick “rise” is observed inFig. 6, corresponding to a “short circuit” through exit 1, followed by along response time, indicating that most of the material has a largeresidence time. Combining the curve in Figs. 6 and 4a,b,c, it is clearthat part of the blue tracer powder leaves quickly through exit 1before any of it has time to pass to chamber 2. However, the overall

Fig. 10. Weight in the dies for fast flow lactose, as function of the feed frame speed andfor three different die speeds using 9.6 dies diameter.

Fig. 11. Weight in the dies for blends with fast flow lactose and two differentconcentrations of APAP and 1% of magnesium stearate as function of the feed framespeed and for two different die speeds using 6.9 die diameter.

111R. Mendez et al. / Powder Technology 200 (2010) 105–116

residence time is several times larger than the time spent by thepowder leaving through the short circuit (exit 1). For instance, in thefirst minute, only 63% of the blue powder has exited the feed frameand, after 2min, 92% of the blue powder has exited. Assuming that allthe incoming powder has the same RTD, this means that 37% of thefast flow lactose in the feed frame before entering the blue powderhad spent more than 1min inside the feed frame, being exposed toshear applied by the paddle.

Since powders passed through feed frames are often lubricatedandmixed with glidants, which are known to be shear sensitive, thereis the possibility that the properties of the powder exiting quicklyafter entering the feed frame could be different to those of thepowders spending more time in the feed frame. This, in turn, affectsthe uniformity of the powders in the die and suggests the possibilitythat the die was filled in layers with powders with different totalapplied shear (more details below).

Fig. 12. Die filling RSD as function of the feed frame and die disc speeds.

The experimental data of concentration vs. time was fitted to atwo-parameter exponential model as in Eq. (2).

τ = τ0ekC ð2Þ

where τ is the residence time, C is the concentration and τ0 with unitsof time and k of concentration−1 are the two adjustable parameters ofthe model. By linearization of the model and using the experimentaldata, the two parameters were estimated as τ0=16.46 s andk=106.4 with a coefficient of determination R2 equal to 0.964.Fig. 7 shows the experimental values and the exponential model.

3.3. Average residence time

The best way to measure the average residence time would be toweight the amount of powder inside the feed frame and divided it bythe mass flow rate. However, this approach was not practical, so aslightly different method was employed. A volumetric approach wasused to calculate the average residence time. The average residencetime was considered to be approximately equal to the feed frameempty volume divided by the volumetric powder flow rate into thedies. This procedure uses the same criteria used to calculate residencetimes in liquid flow system. If density remains nearly constant withinthe feed frame, this procedure is strictly correct. If, on the other hand,the powder compresses in a steady fashion, this procedure yields avalue within a constant of the actual residence time. Despite, it wasassumed that the feed frame was completely filled with powders andnot significant powder compaction effect inside the feed frameoccurred, which allow to use the bulk density of the powder. Eqs. (3)and (4) were used to calculate the average residence time in the feedframe.

τ̃ =Vν̇

=VFF−Vpad

ν̇ð3Þ

ν̇ =Ndie⋅νdie⋅Wdie

ρbdð4Þ

where τ̃ is the average residence time, VFF is the volume of the feedframe, Vpad is the paddle volume, V is the powder volume in the feedframe, ν̇ is the volumetric powder flow per unit of time, Ndie is thenumber of dies in the rotary disc, νdie is the disc speed, Wdie is theaverage weight in the dies, and ρbd is the powder bulk density.

Fig. 8a and b shows the average residence time as a function of thefeed frame and die disc speeds. Fig. 8a shows results obtained usingpure fast flow lactose and two different die diameters. Results inFig. 8b correspond to fast flow lactose with APAP and 1% ofmagnesium stearate. In both cases, the average residence timedecreased as the feed frame and die disc speeds increased. However,for smaller die diameter, the variation in average residence timeincreased. Therefore, the residence time distribution of the powder inthe feed frame depends on the feed frame and dies disc operatingconditions, the flow pattern inside the feed frame, the quantity ofpowder in the die, and may be on the design parameters, such as theexits geometry.

3.4. Die filling as a function of FF speed, die disc speed, die diameter, andcohesion of incoming powder

Figs. 8–10 depict the die filling weight as a function of feed frameand die disc speeds and material characteristics. Figs. 9 and 10correspond to 6.9 and 9.6 mm die diameter respectively and showresults obtained using fast flow lactose. As expected, the larger thefeed frame speed, the larger the fill weight. On the other hand, thelarger the die disc speed, the smaller the filling weight.

Fig. 13. Total shear applied to the powder as function of the operating conditions.

112 R. Mendez et al. / Powder Technology 200 (2010) 105–116

Comparing Figs. 9 and 10, it can be noticed that the larger diediameter (Fig. 10) exhibited a larger reduction in filling weight as thedie disc was increased. The larger reduction is observed for thesmallest feed frame speed. All of these results suggest that die filling isflow-rate limited; the larger the dies or the faster the die table, thelarger the flow rate into the dies that is needed to achieve completefilling. As the feed frame speed is increased, more blade passes areavailable to push powder into the dies during the brief amount of timewhen the dies are exposed to the discharge ports of the feed frame.

Similar behavior was observed (Fig. 11) for the formulation oflactose, 1% of magnesium stearate, and two different concentrations ofAPAP (3% and 30%) for the same operating conditions. As is known topractitioners, die filling is affected by powder flow properties. For theformulation with 30% APAP, which is more cohesive, the quantity ofpowder in the dies decreased approximately 33% relative to the blendwith 3% of APAP. This behavior indicates that the weight of powder in

Fig. 14. Standard deviation of the weight signal as function of the drum rotary speed forthree feed frame speed: 29 rpm die speed.

the dies depended also on the material flow properties. The additionof 30% of APAP increased the cohesion (the dilation values of the 3%and 30% APAP after passing through the systemwith 29 rpm as the diedisc speed and 24 rpm as the feed frame speed were 19.2 and 23.1respectively, which correspond to an increment of 20% in powderdilation), reduced the permeability of air moving outside the dies, anddecreased the flowability of the material.

Fig. 12 depicts the reduction in relative standard deviation (RSD)of the filling weight for lubricated powder as the feed frame speed isincreased for a constant die diameter and material composition. Thefeed frame speed increases the feeding force and the total shearapplied to the material. These two effects increase the weight in thedie and reduce the die weight variability. For the material tested, thedie disc speed seems to have no impact on the uniformity of the diefilling weight.

3.5. Shear strain as a function of FF speed, die disc speed, die diameter,and cohesion of incoming powder

The total applied shear to the powder is related to the number ofpaddle passes, Eq. (5). Assuming that the powder was impacted byone paddle wheel at a time and that the total shear was a result of theapplied shear to the powder in the first and second chambers, theaverage total shear in the feed frame is computed as:

Npp = Np ⋅FFrpm⋅τ̃ ð5Þ

where Npp is number of paddle passes, Np is the number of paddles,FFrpm is the feed frame speed, and τ̃ is the average residence time.

Fig. 13a and b shows an increase in applied shear as a function ofincreases in feed frame speed. The applied shear decreased forincreases in die disc speed. Fig. 13a shows that the applied sheardecreased also for larger die diameter (dotted vs. solid lines). Fig. 13bshows that higher APAP concentration increased the residence timeand hence the applied shear (dotted vs. solid lines).

The red rectangles at exits 1 and 2 (Fig. 3) include open spacesbetween the paddles and the die disc where the powders are depositedand exposed to a significant normal force and shear as the paddles forcethe powder to move inside the die cavities. The normal force is acontribution of the normal component of the shear applied by the bladesand gravity. At exit 3, the separation between the paddle and the die discis smaller, thus thenormal forcedue to the applied stress shouldbe larger.

Fig. 15. Effect of the feed frame and die speed on powder flow index: (a) non-lubricated and lubricated blends of fast flow lactose; (b) lubricated blends of fast flow lactose and APAP.

113R. Mendez et al. / Powder Technology 200 (2010) 105–116

The flow pattern (Fig. 4) indicates that the powder leaving the feedframe through exit 1 was subjected to lesser shear than the powdersleaving through exit 3, which have been subjected to shear in bothchambers. The backmixing effect at the intersection of both chambersshould produce a larger applied shear to the material since in thissection the paddles move in opposite direction over the other paddle.The back mixing effect increased the residence time of the powders inthe second chamber relative to the first one, therefore it increased theshear applied to the powders in this chamber.

Another key area is the one demarked by the blue rectangle overthe paddle wheel in the second chamber. In this zone the powdersmoved in a more dilated condition, experiencing lesser shear than thepowder over the paddle wheel in the first chamber. All these differentshear zones inside the feed frame may cause different effects on theproperties of the material entering the die cavities, whichmight causea non-uniform distribution of material properties in the tablets. This,in turn, might cause non-uniform performances of the tablets. Infuture work, this layering effect will be studied.

3.6. Powder flow index and dilation

The flow properties of powders after passing through the feedframe are listed in Table 1 and graphically depicted in Fig. 14. Asexplained, samples were collected and tested in the GDR before andafter passing through the feed frame-die table combination. Forunlubricated blends there is essentially no effect of the feed frame anddie table speed on the flow properties of the powder. On the otherhand, for lubricated blends, there is a strong effect; powder that haspassed through the feed frame has significantly lower flow index anddilation, indicating a significantly lower cohesion, than powder testedprior passing through the feed frame.

The flow index (Fig. 15) is calculated as the average of the standarddeviation of the load cell weight signal when the cylinder in the GDR isrotated between 5 and 20 rpm. In all cases, for all lubricated blends,the flow index and the dilation decreased markedly as the feed framespeed increased (which increases both the shear rate and the totalstrain, Fig. 15). For non-lubricated blends, on the other hand, the flowindex showed a small increase effect as is observed in Fig. 15a.

Fig. 16 shows the powder dilation (percent of change of bedvolume from consolidated to unconfined flowing state) for non-lubricated lactose blends, lubricated lactose blends, and blends with

APAP with respect to the number of revolutions. The dilation resultsfor non-lubricated blend in Fig. 16 remained relatively constant as thefeed frame speed increased. This demonstrates that the shear appliedto non-lubricated blend with this particle size distribution does notaffect significantly the powder cohesion.

For lubricated fast flow lactose, Fig. 16b and c, at high feed framespeed the dilation of the powder decreased, indicating that the largerfeed frame speeds caused the cohesion of the powder to decrease. Thesame behavior occurred for the blends with fast flow lactose, APAPand MgSt (Fig. 16d and e).

Fig. 17 demonstrates for lubricated blends a markedly dilationdecrease as the feed frame speed increased (which increases both theshear rate and the total strain), and remained constant for the non-lubricated. The same behavior occurred for lubricated blend withAPAP. The reason for this behavior is clear: as it is well known, manyblends lubricated with MgSt are shear sensitive. Large amounts ofapplied strain reduce their cohesion as well as the dissolution rate ofdrugs and the hardness of tablets made from them. The results in thissection demonstrate that such blends can experience enough shear ina feed frame to “overlubricate” them, leading to significant changes inflow properties.

3.7. Die filling RSD as a function of to flow properties of outgoingpowders

As shown in the preceding section, the increased shear strainexperienced at large residence time decreased the cohesion oflubricated materials. Correspondingly, Figs. 18 and 19 show areduction in relative standard deviation (RSD) of the die fillingweight with the decrease in flow index and dilation for lubricatedmaterial with 1% of MgSt. It is important to notice that this lower RSDin filled weight occurs, coincidently, at the higher feed frame speedand the lower die disc speed. This condition corresponds to the highershear applied and residence time of the powder inside the feed frame,which are also the conditions that maximize the exposure time of thedies to the fill region and the number of blade passes available to pushthe powder into the dies. The fact that these effects are co-linear raisesa problematic issue: in order to maximize die filling uniformity, weneed to operate the system under conditions that also maximize theprobability of overlubrication.

Fig. 16. Percent of volume change as function of the drum revolutions: (a) unlubricated blends with fast flow lactose, (b) lubricated blends with fast flow lactose and 0.5% MgSt,(c) lubricated blends with fast flow lactose and 1% MgSt, (d) lubricated blends with fast flow lactose and 3% APAP, and (e) lubricated blends with fast flow lactose and 30% APAP.

114 R. Mendez et al. / Powder Technology 200 (2010) 105–116

Fig. 17. Effect of the feed frame and die speed on powder dilation: (a) unlubricated and lubricated blends with fast flow lactose; (b) lubricated blends with fast flow lactose and APAP.

115R. Mendez et al. / Powder Technology 200 (2010) 105–116

4. Conclusions

Effect of feed frame operation parameters on powder flowproperties and filled weight uniformity were examined experimen-tally. The feed frame and die disc speeds, feed frame geometry, andmaterial characteristics affected the powder flow pattern inside thefeed frame and into the dies. The flow pattern and the mixing effect inthe second stage in the feed frame affected the residence timedistribution of the powder and therefore the shear applied to differentportions of the material. Given the fact that powder exiting throughdifferent discharge positions experience different amounts of strain,the dies could then be filled with powder having variation in itsproperties along the height. Since portions of the powder experienc-ing larger strain can be overlubricated and have measurably differentcohesive properties, this phenomenon could cause axial failure of thetablets.

Fig. 18. Die filling RSD correlation with the flow index.

The average residence time and the average total shear are afunction of the material and the operating conditions. The powderweight in the dies increased for increases in feed frame speed anddecreased for increases in die disc speed. The powder weightdepended also on the powder flow properties. The uneven flow ofthe powder though different zones in the feed frame, and the differentshear rates applied to the powder in these different areas possiblychanged the powder properties of the material leaving the feed framerelative to the inlet properties. Flow index and dilationmeasurementsshowed that as higher strain is applied by increasing the speed in thefeed frame and decrease the speed of the die disc, the cohesion of thelubricated materials decreased, as did the RSD of the filled weight. Thefact that weight uniformity is maximized under conditions thatmaximize tendency to overlubricate is potentially problematic andwill be addressed in future studies seeking to decouple these twoeffects.

Fig. 19. Die filling RSD correlation with the powder dilation.

116 R. Mendez et al. / Powder Technology 200 (2010) 105–116

Acknowledgement

This work was supported by the NSF ERC grant number EEC-0540855, ERC for Structured Organic Particulate Systems.

References

[1] X. Xie, V.M. Puri, Uniformity of powder die filling using a feed shoe: a review, Part.Sci. Technol. 24 (2006) 411–426.

[2] C. Wu, L. Dihoru, A.C.F. Cocks, The flow of powder into simple and stepped dies,Powder Technol. 134 (2003) 24–39.

[3] C. Wu, A.C.F. Cocks, O.T. Gillia, D.A. Thompson, Experimental and numericalinvestigations of powder transfer, Powder Technol. 138 (2003) 216–228.

[4] L.C.R. Schneider, I.C. Sinka, A.C.F. Cocks, Characterization of the flow behavior ofpharmaceutical powders using a model die-shoe system, Powder Technol. 173(2007) 59–71.

[5] I.C. Sinka, L.C.R. Schneider, A.C.F. Cocks, Measurement of the flow properties ofpowders with special reference to die fill, Int. J. Pharm. 280 (2004) 27–38.

[6] A. Mehrotra, B. Chaudhuri, A. Feqih, M.S. Tomassone, F.J. Muzzio, A modelingapproach for understanding effects of powder flow properties on tablet weightvariability, Powder Technol. 188 (2009) 295–300.

[7] I.C. Sinka, F. Motazedian, A.C.F. Cocks, K.G. Pitt, The effect of processing parameterson pharmaceutical tablet properties, Powder Technol. 189 (2009) 276–284.

[8] C. Wu, A.C.F. Cocks, O.T. Gillia, Die filling and powder transfer, Int. J. PowderMetall. 39 (4) (2003) 51–64.

[9] E.R. Rice, J. Tengzelius, Die filling characteristics of metal powder, Powder Metall.29 (3) (1986) 183–194.

[10] K. Kachrimanis, M. Petrides, S. Malamataris, Flow rate of some pharmaceuticaldiluents through die-orifices relevant to mini-tableting, Int. J. Pharm. 303 (2005)72–80.

[11] S. Jackson, I.C. Sinka, A.C.F. Cocks, The effect of suction during die fill on a rotarytablet press, Eur. J. Pharm. Biopharm. 65 (2007) 253–256.

[12] A.M.N. Faqih, A. Mehrotra, S.V. Hammond, F.J. Muzzio, Effect of moisture andmagnesium stearate concentration on flow properties of cohesive granularmaterials, Int. J. Pharm. 336 (2007) 338–345.

[13] A.M.N. Faqih, B. Chaudhuri, A.W. Alexander, C. Davies, F.J. Muzzio, M.S.Tomassone, An experimental/computational approach for examining unconfinedcohesive powder flow, Int. J. Pharm. 324 (2006) 116–127.

[14] A.M.N. Faqih, A.W. Alexander, F.J. Muzzio, M.S. Tomassone, A method forpredicting hopper flow characteristics of pharmaceutical powders, Chem. Eng.Sci. 62 (2007) 1536–1542.

[15] A.W. Alexander, B. Chaudhuri, A.M.N. Faqih, F.J. Muzzio, C. Davies, M.S.Tomassone, Avalanching flow of cohesive powders, Powder Technol. 164(2006) 13–21.

[16] A.M.N. Faqih, B. Chaudhuri, A.W. Alexander, S.V. Hammond, F.J. Muzzio, M.S.Tomassone, Flow-induced dilation of cohesive granular materials, AIChE j. 52 (12)(2006) 4124–4134.