d e n t a l m a t e r i a l s 2 5 ( 2 0 0 9 ) 494–499

avai lab le at www.sc iencedi rec t .com

journa l homepage: www. int l .e lsev ierhea l th .com/ journa ls /dema

In vitro evaluation of a mechanical testingchewing simulator

Martin Steiner ∗, Miltiadis E. Mitsias, Klaus Ludwig, Matthias KernDepartment of Prosthodontics, Propaedeutics and Dental Materials, Christian-Albrechts University Kiel,Arnold-Heller-Str. 16, 24105 Kiel, Germany

a r t i c l e i n f o

Article history:

Received 28 April 2008

Accepted 26 September 2008

Keywords:

Chewing simulator

In vitro

Simulation device

Force-profile

Loading

a b s t r a c t

Objectives. In this in vitro study, the force-profile of a commonly used chewing simulator

(Williytec, Munich, Germany) utilizing fixed weights for loading was evaluated.

Methods. Using piezoelectric force sensors the maximum and mean values of the applied

forces were recorded in three different load configurations in order to determine the rep-

etition accuracy in one test chamber. The variation in resulting forces of the eight test

loading-chambers when using three different loads and descending speeds was explored.

Results. The simulator showed high load repetition accuracy between the different cycle

rates. Significant differences (up to 38.2 ± 0.4 N) were observed between the different

specimen chambers. In addition, the recorded loads were generally both higher (up to

137.5 ± 0.4 N) and lower than the nominal loads defined by the static weights. The extent of

load variation at contact was highly dependent on the descending cross-speed and selected

weight. Finally there were also ringing in the load profiles attributed to vibrations of the

mechanical setup.

Significance. Studies using weight-controlled chewing simulators should consider these

effects when reporting results. In addition, calibrations should be performed to check uni-

formity of tests conditions for each test chamber.

emy

ing cycles should be performed to simulate 5 years of clinical

© 2008 Acad

1. Introduction

Preclinical testing of fatigue or wear of restorative materi-als through chewing simulation is important, because initialin vivo testing comprises ethical problems and is costly andtime-consuming. In addition, in order to achieve comparableresults, in vitro tests should replicate the physiological condi-tions of human mastication as closely as possible [1] includingclinical relevant bite forces.

Bite forces in the human mouth vary over a wide range

among individuals. Accepted values for the physiological biteforces range between 10 and 120 N during chewing of foodor swallowing [2–6]. Maximum forces are considerably higher

∗ Corresponding author. Tel.: +49 431 597 5058; fax: +49 431 597 4063.E-mail address: msteiner@proth.uni-kiel.de (M. Steiner).

0109-5641/$ – see front matter © 2008 Academy of Dental Materials. Pudoi:10.1016/j.dental.2008.09.010

of Dental Materials. Published by Elsevier Ltd. All rights reserved.

and range from 190 N to 290 N in the anterior tooth regionand from 200 N to 360 N in the molar region [7–10]. The force-profile during mastication is nearly a half-sine waveform withrepetitions of 0.2–1.5 Hz [5,11,12].

Therefore, the main requirement for a realistic chewingsimulation device is the ability to simulate properly thesehuman masticatory parameters in a uniaxial or multi-axialmovement, and to be able to perform cyclic loading with arange of bite forces. It is recommended that 1,200,000 chew-

service [13–15]. In order to test a statistically relevant numberof specimens within a reasonable time, it is desirable to usemultiple, simultaneously working test chambers. However,

blished by Elsevier Ltd. All rights reserved.

2 5

catpti[

Mcfaobbe

obsfTsbfb

sitwtclsitl

2

2

TlabToc

fcrlaaabc

d e n t a l m a t e r i a l s

omparable results can only be achieved, if the parametersre precisely controlled and do not vary considerably betweenest chambers, test cycles, and test series [1]. A definite force-rofile is not necessarily the most important parameter, buthe energy transferred into the specimens in one cycle is anmportant parameter to be reproduced for long-term loading16].

A multi-station (eight) chewing simulator (Willytec,unich, Germany) has been commonly utilized in preclini-

al testing since the last decade [17–23]. Willytec is a straightorward easy to adjust mechanical testing system. Its biggestdvantage is that it can apply the same load simultaneouslyn up to eight specimens each placed in a separate test cham-er. An additional important feature is that the specimens cane thermo-cycled in a very precise and accurate way with anlectronic control programed by the operator.

A limitation in the simulator is the inability to observer to control the forces that occur at the initial contactetween the loading stylus (metal or ceramic ball) and thepecimen surface during the chewing cycle. As the initialorce cannot be measured, it may be higher then programed.his could result in a non-physiological overload of the testpecimens [18]. Also, possible variation in contact forcesetween chambers could lead to scatter in the date. There-ore conclusions drawn from chewing simulator testing maye compromised.

Correct interpretation of the results derived from chewingimulators, requires an understanding of the forces appliedn their profiles. The purpose of this study was to examinehe contact forces and the force-profiles of the widely usedeight-controlled chewing simulator (Willytec), with varia-

ion in weight applied, stylus descending rates for each testhamber. The test hypotheses were that within physiologicaloading frequencies the resulting loading forces do not differignificantly from the applied weights and that they are notnfluenced significantly by the descending crosshead rate ofhe simulator nor do the forces differ between test chamberoad cells.

. Materials and methods

.1. The Willytec chewing simulator

he simulator has eight testing chambers with identicaload cells. Force is produced by two stepper motors whichllow computer-controlled vertical and horizontal movementsetween two antagonistic specimens in each test chamber.he masticatory load curve is programed by the combinationf the horizontal and vertical motion range, which are set andontrolled with a computer.

The lower crossbeam holding the specimen chambers isastened through a butterfly nut to one stepper motor whichontrols horizontal sliding motion of the specimen, if it isequired. The antagonistic specimens or the styli inducing theoading are embedded into the upper specimen holders whichre fixed at the lower end of vertical guide rails. The guide rails

re freely mounted within bearings in the upper crossheadnd the vertical height of the antagonistic specimen adjustedy a screw on top of the upper crossbeam. The various weightsan be mounted on top of the guide rails permitting variation

( 2 0 0 9 ) 494–499 495

of the applied chewing force. The upper crossbeam is drivenby the second stepper motor and moves the antagonistic spec-imens or styli vertically. When the upper crosshead movesdown and the upper styli touch the lower specimens, theupper crosshead moves an additional 2 mm down. This occursbecause the guide rails are freely mounted within bearings inthe crosshead. Their individual weight is fully transferred toeach specimen. The unit is additionally equipped to flood anddrain the specimen chambers with tempered water to createthermal cycling (Fig. 1A) [16,18].

2.2. Reading the force-profile

Measuring precisely the peaks in the force-profile of a chewingsimulator requires a highly accurate, fast response and sensi-tive force-detection system for the force ranges of interest.To convene these requirements a piezoelectric force sensor(9132 A, Kistler, Winterthur, Switzerland) was used to con-vert the load applied by the simulator into an electrical signal.The sensors fast response time, is because of their high stiff-ness, and also permit measurement of static load over severalminutes. The actual sensitivity of the system approaches the1 mN, and the working range is 0–2 kN. The sensor signal wasconverted to the voltage range of 0–5 V by a charge ampli-fier which was converted from an analog to a digital signalby means of a PCMCIA AD-Card (DAQCARD 6024E, NationalInstrument, Munich, Germany) and captured by software (Dia-dem 9.01, National Instrument). By calibration of the digitalvoltage values a corresponding force could be assigned. Atfirst the time series was sampled with a scanning rate of100 kHz, which allowed recording of all points and changes ofthe signal up to a frequency of 50 kHz. Then the sample ratewas reduced step-wise just before under-sampling in order toreduce the amount of data for each measurement. This pro-cess occurred without the loss of significant data. As a resultof the calibration all following measurements were made witha sample rate of 10 kHz. In order to affix and center the sensorin the specimen areas of the test chambers a special holderwas manufactured. The load was uniformly transferred tothe sensor via a brass disk of the same diameter as the sen-sor which had a centering pin at its lower surface that fittedinto the hole in the middle of the sensor (Fig. 1B). After cal-ibration under static loading (0–10–0 kg) force-profiles weremeasured.

To process the sampled data, the recorded timelines weredivided into single cycles. The start and the end points of acycle were identified manually. The steep edges at the begin-ning and the end of load contact gave an accuracy of ±1 ms tothe length of the whole cycle as well as of the time of initialload contact. From the timelines the maximum force and themean force during loading were determined, i.e. the maximumas absolute maximum over the whole loading cycle and themean force as mean of all values between the beginning andend of a load contact (Fig. 1C). To compare the recorded valueswith the selected weights for loading the measured forces, thelast were expressed in percentage of the applied loads. To sim-

plify the analysis, the acceleration of gravity was assumed tobe 10 m/s2 (instead of 9.80665 m/s2) and the selected weightwas multiplied by 10 to get the equivalent force. The errorcaused by this simplification was less than the error between

496 d e n t a l m a t e r i a l s 2 5 ( 2 0 0 9 ) 494–499

Fig. 1 – Setup and method: (A) chewing simulator Willytec with control-unit on the right. (B) Mechanical setup of sensorhead with holder for piezoelectric force sensor and coupling disc for balanced load application. (C) Principle of determiningthe maximum and the mean out of a sampled timeline. The start, the end of load section and the end of the cycle isidentified manually by the abrupt ascent and descent of the force signal. The maximum force is determined by the highest

s the

value in the timeline, the mean loading force is calculated acontact.

cycles and as a linear factor this did not affect the comparisonof values adversely.

3. Results

3.1. Comparison of cycles

To determine the differences between cycles the force-profileswere recorded with 10 kg weight and with a 20 mm/s descend-ing speed. Sensors were placed randomly in three test

chambers and each measured for three cycles. The direct com-parison in the timeline showed a comparable duration forthe load section and the total length of the cycles (±4 ms)(Fig. 2A). The recorded mean loading forces ranged from

mean of all values between the start and the end of load

106.9 N to 108.1 N, which was equal to a relative mean overloadof 6.9–8.1% compared to the desired force of 100 N (accord-ing to 10 kg weight). However, the maximum forces differedfrom 144.7 N to 148.5 N, i.e. relative maximum overloads were44.7–48.5% with a variation of 3.8 N (Fig. 2B and Table 1).

3.2. Comparison of test chambers

The force-profiles were recorded with 10 kg weight of load andwith 20 mm/s descending speed for each test chamber. Thedirect comparison in the timeline showed a varying duration

of the loading section of the cycles of 30 ± 2 ms, but the start ofeach loading cycle differed only by 5 ± 1 ms. The time averagedmean loading forces ranged from 105.8 N in specimen cham-ber 1 to 109.7 N in chamber 5, which was equal to a relative

d e n t a l m a t e r i a l s 2 5

Fig. 2 – Comparison of cycles in test chamber 4 with 10 kgweight load and 20 mm/s descending speed. (A) Timeseries of force-profile for every cycle and (B) measuredmeans and maxima for every cycle.

Table 1 – Comparison of forces between different cycles.

Cycle 1 2 3 4

MeansN 107.6 108.1 108.1 107.9% 7.6 8.1 8.1 8.0

MaximaN 144.7 145.4 144.7 148.5% 44.7 45.4 44.7 48.5

Measured in specimen chamber 4 with 20 mm/s descending speed and 10of 100 N (∼10 kg).

Table 2 – Comparison of forces between different test chambers

Chamber 1 2 3 4

MeansN 105.8 (0.1) 109.4 (0.2) 108.7 (0.2) 108.3 (0.1% 5.8 (0.1) 9.4 (0.1) 8.7 (0.2) 8.2 (0.1

MaximaN 121.2 (0.4) 138.9 (0.4) 145.4 (0.0*) 149.3 (0.4% 21.2 (0.4) 38.8 (0.4) 45.4 (0.0*) 49.3 (0.4

Measured with 20 mm/s descending speed and 10 kg weight load. The perc∗ Values are smaller than to be detected in measuring accuracy or zero du

( 2 0 0 9 ) 494–499 497

overload of 5.8–9.7% compared to the desired force of 100 N.However, the maximum forces differed from 121.2 N in spec-imen chamber 1 to 159.4 N in chamber 5. In percentage, therelative maximum overloads were 21.2–59.4% with a variationof 38.2 ± 0.4 N (Fig. 3A and Table 2).

3.3. Comparison of weight-loads and descendingvelocities

The force-profiles were recorded in specimen chamber 4 withweight loads of 1, 5, and 10 kg and using 10, 20, 55 mm/s asdescending speeds. The direct comparison in the timelineshowed a comparable duration for the load section and thetotal length of the cycles (±2 ms). The means and the maximawere always higher than the desired forces (Fig. 3B). For themeans the overloads varied between 4.7% (10 kg with 55 mm/sdescending speed) and 13.6% (1 kg with 10 mm/s descendingspeed). However, for the load maximum, the recorded val-ues represented an overloading between 47.2% (10 kg with20 mm/s descending speed) and 699.1% (1 kg with 55 mm/sdescending speed) compared to the desired loading force(Table 3).

4. Discussion

This weight-controlled chewing simulator did not show asimple force-profile as the use of simplified weight loadingmight suggest. The resulting forces were not well defined

by the applied static weights, but were greatly influenced bythe descending speed of the device. In general the recordedforces increased with an increased descending speed, whichcan be explained by the inertia of mass, the definition of

5 6 7 8

107.9 107.8 107.0 106.97.9 7.8 7.0 6.9

147.9 147.9 146.6 147.247.9 47.9 46.6 47.2

kg weight load. The percentage values are related to a desired force

.

5 6 7 8

) 109.7 (0.1) 105.9 (0.1) 109.3 (0.2) 108.2 (0.4)) 9.7 (0.0*) 6.0 (0.1) 9.3 (0.1) 8.2 (0.4)

) 159.4 (0.4) 151.0 (0.0*) 156.1 (0.6) 148.7 (0.4)) 59.4 (0.4) 51.0 (0.0*) 56.0 (0.6) 48.7 (0.4)

entage values are related to a desired force of 100 N (∼10 kg).

e to rounding issues (standard deviation).

498 d e n t a l m a t e r i a l s 2

Fig. 3 – (A) Comparison of the test chambers. The maximavalues were measured with 10 kg weight for loading and20 mm/s descending speed in the test chamber 4. (B)Comparison of weight loads. Force-profiles measured intest chamber 4 with 10 different weight loads between 1 to

10 kg.

momentum (�p = m�v) and impulse (�I = ��p = �F�t) [24]. While at

10 mm/s descending speed the recorded overloading rangedfrom one and a half to twofold, at 55 mm/s descendingspeed the excessive load could reach sevenfold of the desiredforce.

Table 3 – Comparison of forces between different descending s

Means

10 mm/s 20 mm/s 55 mm/s 10

1 kg11.4 (0.0*) 10.7 (0.0*) 12.1 (0.1) 20.113.6 (0.1) 6.7 (0.3) 21.1 (0.9) 101

5 kg54.9 (0.1) 54.3 (0.1) 53.0 (0.1) 85.09.8 (0.2) 8.6 (0.0*) 5.9 (0.2) 69.9

10 kg108.3 (0.2) 107.4 (0.1) 104.7 (0.1) 1498.3 (0.1) 7.4 (0.1) 4.7 (0.1) 49.8

Measured in specimen chamber 4. The percentage values are related to a∗ Values are smaller than to be detected in measuring accuracy or zero du

5 ( 2 0 0 9 ) 494–499

Therefore, in order to avoid the initial overloading peakat load contact the simulator would have to work at a rel-ative slow descending speed which contradicts the aim ofsimulating long-term chewing within a reasonable time span.Therefore, for future studies with weight-controlled chewingsimulators descending speed and load weights should be care-fully selected so that the desired mean and maximum forcesare not exceeded during cycling loading.

In addition, the mechanical setup of the weight-controlledchewing simulator with step motors, bearings and threadedbars moving the upper crossbeam with the weights isapparently inducing vibrations exceeding the nominal forcetwofold. The vibration peak does not appear in the initial partof the load section, but just before the vertical axis reaches thelower reversal point. In general, vibrations have less frequencyand higher amplitudes when mechanical gears are at lowervelocities [25]. These vibrational effects which are not partof the physiological chewing profiles in humans [5,11,12] willintroduce additional stress to the loaded specimens, whichmight promote microcracks or even fractures in brittle speci-mens.

Nevertheless, a hypothesis could be made that these vibra-tions would simulate bruxism. However, no data is availableas to how these vibrations might influence the outcome ofin vitro studies as compared to clinical studies. To prevent orminimize the vibrations the mechanical setup of the chew-ing simulator would have to be changed to decouple the testchambers from the drive unit. In the tested (current) versionthe chewing simulator was not able to reproduce the exactforce-profile of human mastication, but for durability tests thismight be not critical [16,18].

In general, the mean forces during load contact were morerelated to the selected weight load (only up to 21.1% of over-load) than the recorded maximum forces. It might depend onthe tested material and the test purpose, whether mean ormaximum forces are more important when examining fatigueor wear of restorative specimens [14,18].

While the cycles in an individual load-cell were repro-

ducible in duration and force, the maximum forces occurringin the eight different load-cells of the simulator differed upto 38.2 ± 0.4 N when using a weight of 10 kg. An unbalancedcrosshead assembly and a possible variation in lubrication

peeds and weight loads.

Maxima

mm/s 20 mm/s 55 mm/s

(0.0*) 32.7 (0.0*) 80.0 (0.1) N.4 (0.0*) 227.2 (0.0*) 699.1 (9.9) %

(0.0*) 79.1 (0.4) 156.8 (2.2) N(0.0*) 58.1 (0.7) 213.6 (4.3) %

.8 (0.0*) 147.2 (0.0*) 237.5 (0.4) N(0.0*) 47.2 (0.0*) 137.5 (0.3) %

desired force of 10, 50, 100 N (∼1, 5, 10 kg).

e to rounding issues (standard deviation).

2 5

cdosl

fbtcct

5

TapalTaiuw

r

contact. Dent Mater 2006;22:243–9.

d e n t a l m a t e r i a l s

are of the sliding-tubes are possible explanations for theseifferences. This variation might lead to an unbalanced loadn the specimens and therefore to misleading results [18]. Iteems necessary to eliminate these effects through improvedubrication and/or technical changes in the chewing simulator.

In existing simulators the measured values have to be ratedor the differences occurring between the eight test cham-ers. Previously published studies not evaluating and ratinghe variations between the test chambers should be reviewedonsidering these limitations. For future studies with weight-ontrolled chewing simulators it appears critical to calibratehe devices and to verify that the desired forces are applied.

. Conclusions

he weight-controlled chewing simulator, Willytec, providesn easy to use, easy to setup and a reliable method to accom-lish long-term fatigue or wear tests. However, in order tochieve physiological cycling frequencies in the long-termoading of specimens, fast descending speeds have to be used.hese cause a unique, but non-physiological force-profile withn initial excessive overloading and additional vibrations. Thenfluence of these effects on fatigue and wear test behavior isnknown, so that clinical conclusions should be drawn onlyith great caution.

e f e r e n c e s

[1] Heintze SD. How to qualify and validate wear simulationdevices and methods. Dent Mater 2006;22:712–34.

[2] De Boever JA, et al. Functional occlusal forces: aninvestigation by telemetry. J Prosthet Dent 1978;40:326–33.

[3] Eichner K. Messung der Kräfte bei Kauvorgängen. DtschZahnärztl Z 1963;18:915–24.

[4] Bates JF, Stafford GD, Harrison A. Masticatory function—areview of the literature. III. Masticatory performance andefficiency. J Oral Rehabil 1976;3:57–67.

[5] Kohyama K, et al. Effects of sample hardness on humanchewing force: a model study using silicone rubber. ArchOral Biol 2004;49:805–16.

[6] Schindler HJ, Stengel E, Spiess WE. Feedback control during

mastication of solid food textures—a clinical-experimentalstudy. J Prosthet Dent 1998;80:330–6.

[7] DeLong R, Douglas WH. Development of an artificial oralenvironment for testing of dental restoratives: bi-axial forceand movement control. J Dent Res 1983;62:32–6.

( 2 0 0 9 ) 494–499 499

[8] Fontijn-Tekamp FA, et al. Biting and chewing inoverdentures, full dentures, and natural dentitions. J DentRes 2000;79:1519–24.

[9] Morneburg TR, Proschel PA. Measurement of masticatoryforces and implant loads: a methodologic clinical study. Int JProsthodont 2002;15:20–7.

[10] Waltimo A, Könönen M. A novel bite force recorder andmaximal isometric bite force values for healthy adults.Scand J Dent Res 1993;101:171–5.

[11] Baltzer A, Kaufmann-Jinoian V. Die Beurteilung vonKaukräften. Quintessenz Zahntechnik 2002;28:928–98.

[12] Körber KH, Ludwig K. Maximale Kaukraft alsBerechnungsfaktor zahntechnischer Konstruktionen. DentLabor 1983;31:55–60.

[13] DeLong R, et al. The wear of dental amalgam in an artificialmouth. A clinical correlation. Dent Mater 1985;1:238–42.

[14] Krejci I, Lutz F. In-vitro-Testverfahren zur Evaluationdentaler Restaurationssysteme. 3. Korrelation mitIn-vivo-Resultaten. Schweiz Monatsschr Zahnmed1990;100:1445–9.

[15] Sakaguchi RL, et al. The wear of a posterior dentalcomposite in an artificial mouth. A clinical correlation. DentMater 1986;2:235–40.

[16] Kern M, Strub JR, Lü X-Y. Wear of composite resin veneeringmaterials in a dual-axis chewing simulator. J Oral Rehabil1999;26:372–8.

[17] Gu X-H, Kern M. Marginal discrepancies and leakage ofall-ceramic crowns: influence of luting agents and agingconditions. Int J Prosthodont 2003;16:109–16.

[18] Heintze SD, Zappini G, Rousson V. Wear of ten dentalrestorative materials in five wear simulators—results of around robin test. Dent Mater 2005;21:304–17.

[19] Heintze SD, et al. Influence of the antagonist material on thewear of different composites using two different wearsimulation methods. Dent Mater 2006;22:166–75.

[20] Steinebrunner L, et al. In vitro evaluation of bacterial leakagealong the implant–abutment interface of different implantsystems. Int J Oral Maxillofac Implants 2005;20:875–81.

[21] Koutayas SO, et al. Influence of design and mode of loadingon the fracture strength of all-ceramic resin-bonded fixedpartial dentures. An in-vitro study in a dual-axis chewingsimulator. J Prosthet Dent 2000;83:540–7.

[22] Heydecke G, et al. Fracture strength after dynamic loading ofendodontically treated teeth restored with differentpost-and-core systems. J Prosthet Dent 2002;87:438–45.

[23] Stober T, et al. Wear of resin denture teeth by two-body

[24] Meschede D. Gerthsen Physik, vol. 23, rev. ed. Heidelberg:Springer; 2006. p. 1162.

[25] Dresig H, Holzweißig F. Maschinendynamik, vol. 7, rev. ed.Heidelberg: Springer; 2006. p. 526.