A psychophysical study to determine acceptable limits for repetitive hand impact severity during automotive trim installation

International Journal of Industrial Ergonomics 26 (2000) 625–637

A psychophysical study to determine acceptable limits for

repetitive hhand impact severity during automotive triminstallation

Jim R. Potvina,*, Jim Chiangb, Chris Mckeanc, Allison Stephensd

aFaculty of Human Kinetics, University of Windsor, Windsor, Ont., Canada N9B 3P4bSandalwood, Southfield, MI, USA

cSchool of Human Biology and Nutritional Sciences, University of Guelph, Guelph, Ont., CanadadVehicle Operations, Ford Motor Company, Dearborn, MI, USA

Received 23 July 1999; accepted 11 May 2000

Abstract

The purpose of the current study was to use a psychophysical methodology to establish acceptable impact severitylevels for this automotive trim installation. Two studies were conducted. In the first study, 17 male and 12 femalesubjects (6 assembly line workers and 23 students) performed 5 hand impacts/min on a device that simulated the processof seating push pins during door trim panel installation. In both studies, subjects were asked to impact the simulationdevice as hard as they found acceptable without causing injury, numbness or pain. Subjects were trained for 11 h. Forceand hand acceleration time-histories were recorded from the simulation device and a hand-mounted accelerometer,respectively. The magnitude of each impact was quantified with eight dependent measures: peak, time-to-peak, load rateand impulse, from both the force and acceleration transducers. Statistics were used to determine the e!ects of gender,skill level and impact location on acceptable impact severity. In the second study, 8 male and 8 female subjectsperformed repeated hand impacts on a wall-mounted force plate at three di!erent frequencies (2, 5 and 8 impacts/min)over three separate sessions. Force measures and statistics were the same as in Study 1. In the first study, impactlocation did not appear to have a consistent e!ect on the acceptable impact severities and there was no significantdi!erences observed between male and female values. For both force and acceleration, impulse was the most reliablevariable followed by the peak. In the second study an increase in impact frequency was observed to result in asignificant decrease in the acceptable levels of peak force and force impulse. This e!ect was largest when going from 2 to5 impacts/min and was less pronounced when going from 5 to 8 impacts/min (especially for force impulse). Malesubjects demonstrated significantly higher acceptable impulse levels. Based on the combined results from both studies,acceptable limits were recommended for peak force and impulse that would be acceptable to 75% of the population fora range of frequencies. These limits were observed to range from 181 (8/min) to 259N (2/min) for peak force and 2.53(8/min) to 3.52N s (2/min) for force impulse. It was concluded that force impulse and peak force were the variablesmost likely being controlled by the subjects.

Relevance to industry

Automotive assembly includes a number of tasks that involve hand impacts within the manufacturing process. Onesuch task is the door trim panel installation process where the base of the hand is used to impact the door trim panel

*Corresponding author.E-mail address: [email protected] (J.R. Potvin).

0169-8141/00/$ - see front matter # 2000 Elsevier Science B.V. All rights reserved.PII: S 0 1 6 9 - 8 1 4 1 ( 0 0 ) 0 0 0 3 1 - 7

and drive fastening push pins through holes in the metal door frames. The current study provides tolerance limits sothat industrial tasks involving hand impacts can be evaluated for their injury risk. # 2000 Elsevier Science B.V. Allrights reserved.

Keywords: Hand impact; Psychophysics; Automotive assembly; Ergonomics

1. Introduction

The incidence of cumulative trauma disorders(CTDs) of the upper extremity has increased overthe last decade (Hadler, 1992; Higgs et al., 1992;Kroemer, 1992), and has become a major problemfacing industry (Flinn-Wagner et al., 1990; Hadler,1992; Statistics Canada, 1993). Repetitive occupa-tional tasks can result in chronic tissue injuries.Previous research has provided safety thresholdvalues (i.e. tissue capacity) for some tissues undersome occupational conditions. For example, limitshave been suggested for compression forces actingon the intervertebral joints of the lower backduring manual load handling (NIOSH, 1981;Mital et al., 1993). Maximum acceptable loadsfor lifting have been assumed to be sensitive toload location, lift/lower distance and frequency(Snook, 1978; NIOSH, 1981). In addition, strengthlimits are available for various joints during brief(Cha"n and Andersson, 1991), intermittent(Rhomert, 1973) or sustained isometric contrac-tion (Manenica, 1986).

A number of factors have been identified asincreasing the risk of sustaining CTDs, includingforce, frequency, duration and posture(Armstrong et al., 1986; Putz-Anderson, 1988;Moore et al., 1991). These factors can also interactto magnify the risk of injury (Silverstein et al.,1986, 1987). Many occupational tasks involverepeated hand impacts and such tasks can put aworker at risk of injury. Unfortunately, tissuecapacity values have not been determined for thesekinds of tasks. McAtamney and Corlett (1993)have identified shock, or rapid force increase, as aspecific risk factor in their rapid upper limbassessment (RULA) method. The etiology of handand wrist disorders has been associated with stressconcentrations over the base of the palm (Kendall,1960; Armstrong, 1983). This type of mechanicalstress may be associated with manual striking of

parts when the hand is used as a hammer.Unfortunately, while it is possible to quantifyforce, frequency and limb posture, it is verydi"cult to directly quantify their safe levelsindividually, let alone when they are combined invarious ways. Some e!orts have been made toidentify cadaveric tissue strengths under single andrepetitive loading conditions, but this is onlypossible with passive tissues like bone, ligaments,tendons, etc. (Frankel and Nordin, 1980).

Psychophysics is one method that can be used toestimate acceptable loads under a variety of force,repetition, posture and duration conditions. Psy-chophysics is the study of the relationship betweensensations and their physical stimuli (Snook et al.,1970) and relies on the assumption that individualscan identify working conditions that they perceiveas having an acceptable level of stress for them,based on an integration of biomechanical andphysiological sensory feedback. The primaryadvantage of the psychophysical approach is thatit permits the realistic simulation of industrialwork, allowing aspects such as workspace dimen-sions and task frequencies to be altered accord-ingly (Snook, 1985). Snook (1978) and Snookand Ciriello (1991) have used psychophysicsto determine maximal acceptable weights andforces for various lifting, lowering, pushing, pull-ing and carrying tasks. Ayoub et al. (1978)have also provided similar guidelines for liftingcapacity.

More recently, a psychophysical methodologyhas been used to set acceptable exposure limits forthe upper limbs. Armstrong et al. (1989) wereamong the first to do this by correlating subjectiveassessments with objective measurements of handtools used in automotive assembly. Psychophysicshas been used to determine maximum acceptablefrequencies of drilling for di!erent wrist postures(Kim and Fernandez, 1993; Davis and Fernandez,1994; Marley and Fernandez, 1995), and varying

J.R. Potvin et al. / International Journal of Industrial Ergonomics 26 (2000) 625–637626

applied forces (Kim and Fernandez, 1993)and of gripping under di!erent conditions of gripforce and duration (Dahalan and Fernandez,1993). In addition, psychophysics has been usedto set guidelines for repetitive wrist flexion andextension exertions (Snook et al., 1995) and wristulnar deviation (Snook et al., 1997). They con-cluded that psychophysics was an acceptabletechnique for establishing acceptable exposurelimits and that these limits were sensitive tochanges in motion, frequency, duration and handgrip type.

Automotive assembly includes a number oftasks that involve hand impacts within themanufacturing process. One such task is the doortrim panel installation process where the base ofthe hand is used to impact the door trim panel anddrive fastening plastic push pins through holes inthe metal door frames. The purpose of the currentstudy was to use a psychophysical methodology toestablish acceptable impact severity levels for thistask. The first study was designed to determine theacceptable impact severity during a simulated doortrim panel installation task performed at 5 im-pacts/min. The second study was designed todetermine the relative e!ect of impact frequency(ranging from 2 to 8/min) on the acceptable impactseverity.

2. Methods

2.1. Subjects

All the subjects, in both studies, were screenedfor previous upper limb injuries prior to selectionand informed consent was obtained from eachparticipant. Both studies were approved by theUniversity’s Human Ethics Committee. InStudy 1, six subjects were assembly line workers(termed ‘‘skilled’’) and 23 subjects were universitystudents (termed ‘‘unskilled’’). Each ‘‘skilled’’subject had at least two years experience on anautomotive assembly line performing tasks thatrequire repetitive hand impacts. A total of 17 male(5 skilled and 12 unskilled) and 12 female (1 skilledand 11 unskilled) subjects participated in Study 1.

Eight male and eight female university studentsparticipated in Study 2.

2.2. Experimental protocols

In both Studies 1 and 2, subjects participated in apsychophysical methodology where they were askedto perform hand impacts with their magnitudes thatthey found acceptable (i.e. no signs of discomfort,numbness or injury after the session). The instruc-tions to the subjects were modified from Snook et al.(1995) and are provided in the appendix.

Study 1: Subjects were instructed to use theirhand to repeatedly strike a mechanical door trimpanel simulation device with their maximumacceptable force assuming that the impact levelsthey selected would be sustained repetitively overan 8 h work shift. Subjects could modify theresistance of the simulation device to movement.Skilled subjects were trained for 4 h in a single day.Unskilled subjects were trained for 8 h over twodays with at least 48 h between sessions (4 htraining the first session and 4 h the second). Itwas assumed that skilled subjects would requireless training as they were very familiar with thetask. In addition, it was only possible to have themin the lab for one day. Unskilled subjects weregiven more training to familiarize them with whatwas a novel task. The data indicated that allunskilled subjects converged on stable acceptablelevels after 8 h of training. No data were collectedduring the training sessions. All 4 h sessionsincluded two 15min rest periods. The timing ofthe session was 90min, 15minute rest, 60min,15min rest, 60min. The collection session was 4 hin duration and took place at least 24 h after thesecond training session. Subjects were required tostrike the simulation device repeatedly at afrequency of 5 impacts/min. Impacts were clus-tered at the beginning of each minute with theparticipant striking the plate at a self-determinedrate. This procedure was used to simulate whatwas observed on the assembly line. The impactsurface was placed in one of four locationcombinations relative to the body. These locationswere either high (130 cm) or low (108 cm) andeither close to the body (45 cm from the ankle) orat a reach for the hand (75 cm from the ankle). In

J.R. Potvin et al. / International Journal of Industrial Ergonomics 26 (2000) 625–637 627

both studies the impact locations and frequencieswere selected to simulate those values observed inautomotive assembly tasks. The order of presenta-tion of the impact location was randomized andsubjects were rotated to a new location every15min. Subjects were free to select the hand/armposture at impact. However, each selected aposture where the arm movement was parallelwith the horizontal plane and the impact was onthe palm of the hand, with the fingers pointinganteriorly along the body’s midline, the thumbpointing upwards and the elbow slightly bent.

Study 2: Subjects were required to repeatedlystrike the vertical surface of a wall-mounted forceplate, at approximately chest height, with the palmof their dominant hand while standing to the non-dominant hand’s side of the plate. These subjectswere trained in a preliminary 2 h session, spending40min at each of three impact frequencies(randomly ordered). Impacts were spaced evenlyover time and the frequencies were 2, 5 and 8/minute (one impact every 30, 12 and 8.5 s,respectively), as controlled by an audible metro-nome. The training session was used to familiarizethe subjects with the psychophysical methodologyand with the consequences of the selected impactseverities. Subjects returned for three separate datacollection sessions of 2 h each. In each session,they impacted the force plate at either 2, 5 or8 impacts/min for the entire session. The order ofthe presentation of these sessions was randomizedwithin each subject. Each session in Study 2 wasseparated by at least 48 h.

2.3. Data collection

For both studies, data were collected at 2000Hzand processed with LabVIEW software (version4.1, National Instruments). The collection of eachimpact was triggered by the sudden increase inforce. Trials were collected from 50ms before thetrigger to 150ms after. Force time-histories werecollected for each impact during the collectionsessions.

Study 1: Force, acceleration and air pressuredata were collected on a PC-compatible computer(Intel P166MMX, 32MB RAM) with a 16 bitmultifunction I/O board (AT-MIO-16XL, Na-

tional Instruments). A mechanical simulationdevice (Fig. 1) was designed to simulate the time-history of the hand impact forces required to insertpush pins during the door trim panel installation(Fig. 2). During actual door trim installation, thisbi-phasic force resulted from the rapid increase inforce during initial wedging of the conical push pinthrough the hole followed by the low-resistancemotion of a small stem followed by the rise in forceas the trim panel made contact with the metal doorframe.

The simulation device was constructed from a6mm thick 6061 aluminum plate. The verticalimpact surface was a 15! 15 cm2 square coveredwith a thin layer of foam and vinyl (4mm thick) toprovide some cushioning between the hand andplate on impact and to simulate the feel of anautomortive door trim panel impact. The verticallyoriented impact surface was connected to ahorizontal linear motion slider connected to a tabletop. The resistance to posterior motion of theimpact plate and linear motion slider was con-trolled by a pneumatic cylinder (SMC NCDMB-075) connected behind the impact plate. The rangeof motion of the slider was limi-ted posteriorly bya metal stop, and anteriorly by the end range ofthe pneumatic cylinder. A force transducer (Trans-ducer Techniques MLP-500-CO) was placed be-tween the impact plate and pneumatic cylinder. Anair compressor supplied pressure to the pneumaticcylinder. Air line pressure was monitored by apiezo-resistive sensor (Motorola SenseonMPX570-GP) and air pressure was controlledwith a precision regulator (Norgren 11-018-100),Device timing was controlled by a custom timingcircuit utilizing a 556 timer IC for post-impactdelay and momentary return air. The timingcircuit was connected to a pneumatic valve (Acso8320Gfd174) via a solid-state relay. On/o!switch-ing of air flow was controlled by the valve.

The functioning of the impact device was asfollows: Prior to impact, the valve was on therebypressurizing the pneumatic cylinder and pushingthe impact plate forward. Upon impact the sliderwas forced backward. After 3mm of travel amagnetic position switch was tripped which turnedthe valve o! and removed air pressure from thecylinder. The plate and slider travelled backwards

J.R. Potvin et al. / International Journal of Industrial Ergonomics 26 (2000) 625–637628

an additional 6mm where it came into contactwith a metal stop. At this point a second magneticswitch was tripped which began the delay timer ofthe timing circuit. The delay was approximately0.5 s, after which a second timer was triggered,switching the valve on. This pressurized thecylinder and extended it back to the startingposition so that the next impact could occur.

Throughout all sessions, subjects were awarethat they were free to increase or decrease theresistance setting of the pneumatic cylinder byturning an unmarked dial. The experimenter

changed the resistance level every 15min by eitherincreasing or decreasing pressure by an unspecifiedand randomly selected amount. This forced thesubjects to re-establish the acceptable resistancelevels at regular intervals. For each subject, handimpact acceleration data were collected with a tri-axial accelerometer (Crossbow CXL75M3). Theaccelerometer was mounted to the back of theimpacting hand via a modified wrist/hand bracewhich the participants wore during the sessions.Force time-histories were collected for all subjectsand acceleration time-histories were collected for12 male and 10 female subjects.

Study 2: Force data were collected using a PC-compatible computer (Intel 486/66, 8MB RAM)with a 12 bit multifunction I/O board (AT-MIO-16, National Instruments) A Kistler piezoelectricforce plate (Type 9281B) was rigidly mountedvertically on a wall so the target impact locationwas 1.15m from the floor. A 10mm thick sheet offoam was fastened to the surface of the force plate.The foam in Study 2 was thicker than the 4mmfoam used in Study 1 because there was nodeformation in the rigid force plate and this extrathickness allowed for impacts that better simulatedthe feel of a door trim panel installation.

2.4. Data analysis

In both studies the dependent variables werepeak, time-to-peak, rate of loading and impulse

Fig. 1. Schematic representation of the door trim panelinstallation simulation device. The top diagram illustrates thefunctional components of the device as seen from a side view.The bottom figure indicates the mounting of the device on theplatform and the location of the unmarked, cylinder pressureadjustment dial.

Fig. 2. Sample force time-history for one impact with anindication of how each independent variable was determinedfrom the force transducer.

J.R. Potvin et al. / International Journal of Industrial Ergonomics 26 (2000) 625–637 629

(see Fig. 2) for force and acceleration in Study 1and for force in Study 2. The variables used in thisstudy were similar to those commonly used toquantify the severity of foot impact during runningand walking (see Nigg et al., 1987; Lafortune,1991; Hennig et al., 1993). The rate of loading wascalculated to be the linear regression slope of theforce or acceleration curve between 30% and 70%of the peak. Impulse was calculated as the integralof the time history (N s for force, m/s foracceleration).

Study 1: Force and hand acceleration werecollected for each impact. Only data from the lasthour of the 4 h collection session were analyzed(15min at each location). The two impactsfollowing any pressure adjustment were removedfrom analysis because they were not considered torepresent a normal trial. Both the force andacceleration data were digitally filtered using afourth order, dual-pass Butterworth filter with alow cut-o! frequency of 500Hz to remove aliasing.For both the force and acceleration time-histories,over 95% of the signal power was found to bebelow 115Hz. The resultant of the tri-axialaccelerometer data was calculated and was usedfor all acceleration processing. The force andacceleration curves were windowed independently.The length of the window for both curves was50ms. The start of the window for both the forceand acceleration data was determined indepen-dently using a threshold. In addition, the resis-tance setting was recorded for each impact.

Study 2: Only data from the last 30min of the2 h collection session were analyzed. Only forcemeasures were made in this study and these signalswere processed as described for Study 1.

2.5. Statistical analysis

For both studies, statistical analyses wereperformed with the SuperANOVA software pack-age (version 1.11, Abacus Concepts, Inc). Sig-nificance was set at p50:05 . Means and standarddeviations were calculated for each factor. Coe"-cients of variation (CV: standard deviation as apercentage of the mean) were calculated for alldependent variables to determine the consistency,within subjects, for each condition of both studies.

Averages of these within-subject CVs were calcu-lated across subjects for each variable.

Study 1: A three factor (2! 2! 4) ANOVA withrepeated measures was used to determine the maine!ects of gender (male and female), skill level(skilled and unskilled) and impact location (high–close, high–far, low–close, low–far) for ninedependent variables (peak, time-to-peak, rate ofloading and impulse for both force and accelera-tion, in addition to resistance setting). Post hoccontrasts were performed on dependent variablesthat showed a location e!ect or interactionbetween position and either gender or skill level.

Study 2: A two factor (2! 3) ANOVA withrepeated measures was used to determine the maine!ects of gender (male and female) and impactfrequency (2, 5 and 8/min) for four dependentvariables (force peak, time-to-peak, rate of loadingand impulse). Post hoc contrast were performedon dependent variables that demonstrated a fre-quency e!ect and/or gender-frequency interaction.

3. Results

3.1. Study 1

In the first study, both skill and gender werefound to have a significant e!ect on resistancesetting. Otherwise, no independent variables de-monstrated a statistically significant e!ect on anyof the other dependent variables. The location ofthe impact surface, relative to the body (high vs.low, far vs. close), did not appear to influence theacceptable impact severity as measured with theforce and acceleration independent variables.None of these variables showed any e!ects of skillor gender. In addition, there were no interactionsfound between independent variables. It wasdi"cult to compare the skilled and unskilledsubjects because only 6 skilled subjects were tested(compared to 23 unskilled subjects) and accelera-tion data were obtained from only 3 skilledsubjects (compare to 19 unskilled subjects).

3.1.1. Force variablesDuring the hand impacts with the simulation

device, there was usually a first peak in the force

J.R. Potvin et al. / International Journal of Industrial Ergonomics 26 (2000) 625–637630

record, associated with the initial contact with thedevice, and then a second peak associated with thesimulation of the door trim panel coming intocontact with the door. The first peak was almostalways higher than the second, and the first peakforces were an average of about two times higherthan that for the second peak. Thus, the first peakwas determined to be the most hazardous of thetwo peaks and only its data were presented in theresults. A sample plot is presented in Fig. 3showing all force impulses over the course of atypical 4 h trial.

The pooled results for the force variables inStudy 1 are presented in Table 1. Despite the lackof statistically significant e!ects of gender, malesdemonstrated a trend towards having higher peakforces (by 17%), load rates (by 19%) and impulses(by 23%) and lower time-to-peak (by 7%), whencompared to the average values for females. Fromthe force time-histories, the most reliable variablewas impulse (CV=12%), followed by peak force

(17%), time-to-peak force (18%) and load rate(34%).

3.1.2. Hand acceleration variablesThe pooled results for the acceleration variables

in Study 1 are presented in Table 1. There wasno significant e!ect of gender, location or skilllevel on any of the acceleration variables. Itwas observed that the time-to-peak accelera-tion was less than the time-to-peak force by anaverage of 0.62ms. This di!erence resultedfrom the di!erent windows used for the forceand acceleration time-histories. In fact, the peakforce always occurred before the peak accelera-tion. For the hand acceleration time-histories, themost reliable variable was acceleration impulse(14%), followed by peak acceleration (15%), time-to-peak acceleration (21%) and rate of accelera-tion (28%).

Fig. 3. Sample time-history for the force impulse (* ) and resistance setting (* ) over the course of one 4 h collection session in Study1. Pressure was adjusted (up or down) and posture was changed every 15min. There was generally a high variability in each measuredvariable for the first few impacts after an adjustment as the subjects reset the pressure to an acceptable level. These impacts were notincluded in the statistical analysis. There was a 15min break after 90 and 150min. Statistics were performed on the four posturesassociated with the last 60min of the session (indicated with solid vertical lines).

J.R. Potvin et al. / International Journal of Industrial Ergonomics 26 (2000) 625–637 631

3.1.3. Resistance settingThe pooled average for resistance setting was

30.8 PSI such that the average static force of theresistance was 60.4N (Table 1). Male subjectsselected resistance levels that were significantlyhigher than female values (by 20%, p50:05 ).Unskilled subjects selected resistance levels thatwere significantly higher than those of skilledsubjects (by 36%, p50:05 ). Force impulse was thevariable with the highest correlation to theresistance setting (r2 " 0:67 ), followed by peakforce (r2 " 0:50 ). No other variable had acorrelation with the resistance setting that wasabove r2 " 0:25 .

3.2. Study 2

The pooled results from Study 2 are presented inTable 2. Impact frequency was found to have asignificant e!ect on peak force (p50:01 ), load rate

(p50:05 ) and impulse (p50:01 ). The averagepeak force at a frequency of 2/min was higher than5/min (by 20%) and 8/min (by 27%) (both atp50:01 ). The average load rate at a frequency of2/min was higher than 8/min (by 52%, p50:05 ).Although the di!erence was not significant, therewas a trend for load rates at 2/min to be higherthan 5/min (by 36%) and for 5/min to be higherthan 8/min (by 12%). The average impulse for 2/min was found to be higher than 5/min (by 12%,p50:01 ) and 8/min (by 10%, p50:01 ). Fig. 4illustrates the e!ect of impact frequency on peakforce, load rate and impulse, relative to the meanvalues at 5 impacts/min. For each variable, therewas a larger decrease in acceptable values whenfrequency increased from 2 to 5/min than when itincreased from 5 to 8/min.

Male impulse values were found to be signifi-cantly higher than female values (by 40%, p50:05 ). Although not significant, the average male

Table 1Mean (and standard deviation) values for each variable measured in Study 1 with the simulation device. Data have been pooled acrossall conditions (All Data), within each impact location, within both genders and within both skill levels. Resistance setting refers to thepressure set in the pneumatic cylinder (PSI) and the static force required to overcome the resistance (N)

Variable All Data Impact location Gender Skill level

Low High Male Female Skilled Unskilled

Close Far Close Far

First peak force (N) 235.2 229.6 234.3 235.1 241.9 250.1 214.2 243.3 233.1(60.8) (63.8) (59.8) (68.3) (66.5) (67.2) (45.1) (61.6) (64.1)

Time to first peak (ms) 5.39 5.42 5.10 5.83 5.21 5.54 5.17 4.86 5.53(1.14) (1.16) (1.01) (1.56) (1.19) (1.19) (1.08) (1.2) (1.11)

Rate of loading (N/ms) 81.1 79.1 85.3 74.2 85.9 87.3 73.4 96.4 77.7(37.9) (40.9) (35.3) (45.1) (39.8) (45.8) (24.8) (56.7) (33.3)

Impulse (N s) 3.50 3.42 3.50 3.56 3.53 3.79 3.09 3.12 3.60(1.03) (0.99) (1.03) (1.13) (1.08) (1.08) (0.83) (0.89) (1.06)

Resistance setting (PSI) 30.8 29.9 31.2 30.8 31.3 33.1 27.5 23.9 32.6(10.8) (11.5) (11.1) (12.5) (12.7) (9.4) (12.3) (8.7) (10.8)

Subjects (force data) 29 29 29 29 29 17 12 6 23Peak accel (m/s2) 490.4 492.5 500.5 469.7 498.7 472.9 511.2 384.8 507.0

(119) (127) (127) (128) (119) (124) (117) (134) (112)Time-to-peak accel (ms) 4.77 4.90 4.62 5.01 4.57 4.74 4.81 4.53 4.81

(1.07) (1.34) (0.9) (1.19) (1.19) (1.18) (0.99) (0.88) (1.11)Rate of accel (m/s2/ms) 189.6 188.9 192.2 175.7 201.7 181.6 199.4 152.9 195.4

(72.4) (82.2) (73.3) (70.7) (78) (76.3) (70.3) (93.6) (69.9)Accel impulse (m/s) 4.23 4.28 4.31 4.07 4.27 4.34 4.10 3.77 4.30

(0.78) (0.82) (0.84) (0.79) (0.85) (0.96) (0.5) (1.03) (0.74)Subjects (accel data) 22 22 22 22 22 12 10 3 19

J.R. Potvin et al. / International Journal of Industrial Ergonomics 26 (2000) 625–637632

peak forces values showed a trend towards beinghigher than female values (by 20%) (Table 2).

Coe"cients of variation were calculated withdata pooled within subjects for each variable.These data were used to determine the consistencyand reliability of each measured variable. Fromthe force time-histories, the most reliable wasimpulse (CV=9%), followed by time-to-peak(9%), peak force (14%) and load rate (22%).

3.3. Recommended acceptable limits

Seventy fifth percentile values have been pro-posed by Snook (1978) and NIOSH (1981) asreasonable levels to set acceptable limits for

exposure to occupational loads. The results fromStudies 1 and 2 were combined to create theacceptable limits for peak force and force impulse.The values provided in Table 3 are based on theaverage values obtained in Study 1 at 5 impacts/min, scaled to the ratios obtained in Study 2 (seeFig. 4). A linear interpolation of the ratios from 2to 5/min and 5 to 8/min was used. Accelerationdata was not collected in Study 1; so, a similarcalculation cannot be made for this variable.However, the 75th percentile values at 5 impacts/min for peak acceleration and acceleration impulsewere 410m/s2 and 3.7m/s, respectively.

3.4. Comparing the results of Studies 1 and 2

A comparison was made between the forcevariables recorded at 5 impacts/min with both thesimulation device (Study 1) and the force plate(Study 2). This comparison was used, in additionto the reliability of each variable, to assist in adetermination of the factor(s) that the subjectsmay have been controlling or attending to whensetting acceptable impact severities. The averagepeak force and rate of loading from Study 2 werefound to be higher than the Study 1 averages by161% for both variables. Conversely, the averagetime-to-peak force and impulse values from Study2 were lower than those from Study 1 by only 8%and 17%, respectively.

Table 2Mean (and standard deviation) values for each variable measured in Study 2 with the force plate. Data have been pooled across allconditions, within each frequency and within both genders

Variable All Data Frequency (impacts/min) Gender

2 5 8 Male Female

Peak force (N) 644.9 739.4 613.8 581.5 702.8 587.0(318.3) (403.5) (306.4) (279.1) (337.1) (309.5)

Time-to-peak (ms) 4.89 4.71 4.95 5.02 5.08 4.71(0.93) (0.97) (1.03) (0.97) (0.57) (1.21)

Rate of loading (N/ms) 229.8 287.9 211.6 189.8 236.4 223.1(170.9) (267.9) (155.3) (129.9) (192.5) (159.5)

Impulse (N s) 3.03 3.25 2.89 2.95 3.54 2.52(0.95) (1.10) (0.93) (0.90) (0.97) (0.64)

Subjects 16 16 16 16 8 8

Fig. 4. Ratio of peak force, load rate and impulse variables at 2and 8 impacts/min when compared with the results at 5 impacts/min for Study 2.

J.R. Potvin et al. / International Journal of Industrial Ergonomics 26 (2000) 625–637 633

4. Discussion

The purpose of the current studies was to use apsychophysical methodology to establish accepta-ble impact severity levels during door trim panelinstallation. Study 1 was designed to determine theacceptable impact severity during a simulated doortrim panel installation task performed at 5 im-pacts/min. Study 2 was designed to determine thee!ect of impact frequency (ranging from 2 to 8/min) on the acceptable impact severity. Using therelative e!ects of impact frequency observed inStudy 2 and the impact severities observed inStudy 1 at 5 impacts/min, limits were recom-mended for peak force and impulse that would beacceptable to 75% of the population for a range offrequencies. The need for dynamic testing washighlighted by the observation that the averagepeak dynamic force in Study 1 (235.2 N) wasalmost four times higher than the minimal averagestatic force required to overcome the set resistanceand move the pneumatic cylinder (60.4N) in thesimulation device (Table 1).

The limitations of the psychophysical metho-dology and the assumptions made in these studiesmust be recognized. Psychophysical has become anaccepted method for setting acceptable exposurelimits in industry. The primary advantage to thepsychophysical approach is that it permits therealistic simulation of industrial work (Snook,1985). However, it should be noted that subjectsrequire training to do this reliably and errors canbe made at the extreme frequencies (Karwowksiand Yates, 1986). In the current study, unskilledsubjects were trained on the simulation device for

two 4 h sessions before their final 4 h collectionsession, in which data was processed only from thelast hour. In this way they had 11 h of exposure tothe device and the task before an evaluation wasmade of their acceptable limits. The lack ofconsistent di!erences between the skilled andunskilled subjects supports a conclusion that thetraining was su"cient in this study.

It was assumed that the device used in Study 1provided a reasonable simulation of the insertionof push pins during door trim panel installation.The device was designed to replicate the rapidincrease in force required to accelerate the doorpanel and to wedge the push pin through the hole.Following this, there was a rapid decrease inresistance to replicate the movement of the pushpin stem through the hole. Finally the devicereplicated the second rapid rise in force as the doorpanel’s posterior surface came into contact withthe sheet metal.

The observation that acceptable force levelsdecreased with increasing exposure frequency(Table 2 and Fig. 4) was consistent with previouspsychophysical studies of lifting (e.g. Ayoub et al.,1978; Snook and Ciriello, 1991), wrist flexion andextension (Snook et al., 1995) and drilling (Kimand Fernandez, 1993; Dahalan and Fernandez,1993; Davis and Fernandez, 1994; Marley andFernandez, 1995). In Study 2 no attempt was madeto simulate any particular automotive manufac-turing task. Instead, a generic hand impact taskwas performed on a wall-mounted force plate todetermine the relative e!ect of impact frequencyon impact tolerance. Frequencies of 2, 5 and 8impacts/min were selected to represent a typical

Table 3Recommended threshold limit values for peak force and force impulse based on 75th percentile values. ALL refers to a combinedgroup with males and females. Female and male limits are intended only when one of those groups is doing the task exclusively

Variable Group Impacts/min

2 3 4 5 6 7 8

Peak force (N) ALL 234 221 207 194 191 187 184Female 208 200 192 184 183 182 181Male 259 241 223 205 199 193 187

Impulse (N s) ALL 3.15 3.03 2.92 2.80 2.80 2.8 2.8Female 2.74 2.67 2.60 2.53 2.53 2.53 2.53Male 3.52 3.37 3.21 3.06 3.06 3.06 3.06

J.R. Potvin et al. / International Journal of Industrial Ergonomics 26 (2000) 625–637634

range found in automotive assembly. It is pro-posed that the results of Study 2 may be applicableto a wide variety of industrial tasks for identifyingthe frequency dependence of peak force andimpulse limits. Snook et al. (1995) used acceptablevalues obtained for various wrist flexion andextension frequencies (ranging from 2 to 20motions/min) performed two days per week toscale values obtained at one frequency (15/min)performed five days per week. Similarly, the resultsfrom the three frequencies from Study 2 were usedto estimate acceptable limits for a range offrequencies during door trim panel installationbased on the one frequency (5/min) tested in Study1. The larger relative decreases observed asfrequency was increased from 2 to 5/min, com-pared to 5 to 8/min, may be due to the largerrelative change in the frequencies form 2 to 5(150% increase) compare to 5 to 8 (60% increase).This non-linear relationship between frequencyand impact severity is consistent with many otherobservations in psychophysics as characterized byStephen’s Power Law (Stevens, 1967).

Observation of the door trim panel installationprocess on automotive assembly lines indicatedthat individuals perform a number of handimpacts while standing in one location and thenstep forward. While standing, they are observed toimpact both high and low on the panel and toimpact close to the body and at some reach beforestepping. These motions were simulated with thefour impact locations relative to the subjects’bodies. Based on the results of the current study,it did not appear that impact location and armposture had any systematic e!ect on the acceptableimpact severities selected by the subjects. An e!ectmay have been observed if more extreme postureshad been studied but, generally, the workers wereobserved to step before adopting such postures.

In both studies, males demonstrated a trendtowards accepting impacts with higher peakforces, impulses and load rates and lower time-to-peak than females (Tables 1 and 2). Each ofthese di!erences indicated that males were tolerantof more severe impacts than females. In the firststudy, both genders demonstrated changes in eachof the four force variables indicating a decrease inseverity from 2 to 5 impacts/min. This decrease

was relatively larger for males. However, onlymales showed any trend towards decreasingseverity from 5 to 8 impacts/min. It was concludedthat males were more sensitive to the e!ects ofincreased impact frequency. It is possible thatfemales selected a more conservative estimate ofacceptable force and maintained that limit at allfrequencies. It should be noted that the variabilitywas high across both the male and female groupsand there was not a clear distinction between thegenders. This accounts for the relative absence ofstatistically significant gender e!ects in spite of thetrends that were observed.

Given the relatively low number of skilledworkers used in this study, it was di"cult to assessthe di!erences that may have existed with theunskilled student subjects. In the first study, 6 ofthe 29 subjects (26%) were experienced employeesfrom automotive assembly lines (consideredskilled) and only 3 of the 22 subjects (16%) testedwith the accelerometer were skilled. However, forthe force variables there was no consistent trendobserved to distinguish the two groups. Thus, itwas assumed that the 11 h of training (precedingdata collection) were su"cient to allow theunskilled subjects to e!ectively ascertain theiracceptable limits and the data was subsequentlypooled across skill levels when recommendationswere made. The average acceleration values forskilled subjects must be viewed with caution asthey represent only 3 subjects.

A number of variables were used to characterizethe severity of each impact. Previous studies haveused similar variables to summarise components ofthe force and acceleration time-histories that wereconsidered to be most directly related to thedevelopment of a cumulative trauma disorders. Itwas assumed that the variables showing thegreatest consistency within subjects may have beenthe variables that were being controlled by thesubjects. Impulse represents the change in momen-tum of the arm occurring during the impact phaseand provides a representation of the velocitythe arm has just before contact. Impulse was themost repeatable variable for force in Studies 1 and2 and for acceleration in Study 1, with averagecoe"cients of variation of 9%, 12% and 14%,respectively. This low variability indicates good

J.R. Potvin et al. / International Journal of Industrial Ergonomics 26 (2000) 625–637 635

reliability for each of the impulse variables. Acomparison was made between the peak force,time-to-peak, load rate and impulse as measuredin Studies 1 and 2. The average acceptable forceimpulse and time-to-peak were relatively consis-tent between Studies 1 and 2 when compared topeak force and load rate (Tables 1 and 2). Inaddition, the force impulse was the variable thathad the highest correlation (r " 0:82 ) with theresistance setting being controlled by the subjectson the simulation device. Based on these findings,it was concluded that force impulse was thevariable most likely being controlled by thesubjects during each impact.

In each case, the peak and time-to-peak vari-ables also demonstrated acceptable reliabilityalthough the load rate variables generally demon-strated high coe"cients of variation within sub-jects and were not considered to be reliable. Thepeak force also demonstrated a relatively highcorrelation to most of the variables measured inStudy 1. However, in Study 2 the average peakforce was 614N compared to 235N in Study 1(161% higher). This indicates that the subjectswere willing to accept relatively high impact forceson the force plate but the forces did not reach theselevels on the simulation device. This discrepancymay indicate that peak force was not the variablesubjects attended to when setting acceptableimpact limits in Study 2.

Snook (1978) established a threshold limit value(TLV) at the 75th percentile for both males andfemales based on insurance data indicating thatthe risk of low-back injury was three times higherif a lifting task that was acceptable to less than75% of the working population. The TLV valueswere calculated for each of the variables measuredin Study 1 (Table 3). Because of the trendsindicating di!erences between males and females,these 75th percentile values were estimated withdata pooled across all subjects, as well as malesand females for the force variables. The di!erencesbetween male and female acceleration values werenot su"cient to warrant separate recommendedlimits and, thus, acceleration TLV values areprovided only for the data pooled across allsubjects. The data from Studies 1 and 2 werecombined to estimate the recommended threshold

limit values for peak force and force impulseacross frequencies ranging from 2 to 8 impacts/min. The limits were found to range from 181Nand 2.53N s (females, 8/min) to 259N and 3.52N s(males, 2/min) for the peak force and force impulsevariables, respectively.

Acknowledgements

This research was funded by Ford MotorCompany. The authors would like to acknowledgethe contributions of Mike Vitek and Dan Drouinto this research.

Appendix. Subject instructions

Your job is to strike the impact device surfacefive successive times after each time you hear thebeeper. Each time you will strike the device withenough force to move it backwards until it stops.We strongly encourage you to maintain the pace of5 impacts/min and to move the impact surface thefull distance with each impact.

Instructions for adjusting the impact resistance:This task is intended to simulate an assembly linetask being performed 5 times/min for an 8 h shift,which would include 7 h of work and 1 h of breaks.We want you to perform the task in such a waythat you strike the device with the highest amountof force that would not cause pain, discomfort ornumbness in your hand or arm after an 8 h period.

You will adjust your own resistance to impactby turning the knob located beside the device.Adjusting your own resistance is not an easy task.Only you know how you feel.

If you feel the resistance is too high, reduce theresistance by turning the knob clockwise.

However, we do not want you hitting too lightlyeither. If you feel you can increase the resistance,turn the knob counter clockwise.

Do not be afraid to make adjustments. Youhave to make enough adjustments so that you geta good feeling for what is too hard and what is tooeasy. You can never make too many adjustments,but you can make too few.

Remember, this is not a contest. Everyone is notexpected to strike the device with the same amountof force.

J.R. Potvin et al. / International Journal of Industrial Ergonomics 26 (2000) 625–637636

We want your judgement on how hard you canstrike the device without developing pain, dis-comfort or numbness in the hand and arm.

References

Armstrong, T.J., 1983. An Ergonomic Guide to Carpal TunnelSyndrome. American Industrial Hygiene Association, Akron,OH.Armstrong, T.J., Punnet, L., Ketner, P., 1989. Subjectiveworker assessments of hand tools used in automobileassembly. American Industrial Hygiene Association Journal50, 639–645.Armstrong, T.J., Radwin, R.G., Hansen, D.J., Kennedy,K.W., 1986. Repetitive trauma disorders: job evaluation anddesign. Human Factors 28, 325–336.Ayoub, M.M., Dryden, R.D., McDaniel, J.W., Knipfer, R.E.,Aghazadeh, F., 1978. Modeling of lifting capacity as afunction of operator and task variables. In: Drury, C.G.(Ed.), Safety in Manual Materials Handling. DHEW/NIOSHPublication No. 78–185. NIOSH, Cincinnati, OH.Cha"n, D.B., Andersson, G.B.J., 1991. Occupational Biome-chanics. Wiley, New York, Toronto.Dahalan, J.B., Fernandez, J.E., 1993. Psychophysical fre-quency for a gripping task. International Journal ofIndustrial Ergonomics 12, 219–230.Davis, P.J., Fernandez, J.E., 1994. Maximum acceptablefrequencies for females performing a drilling task in di!erentwrist postures. Journal of Human Ergonomics 23, 81–92.Flinn-Wagner, S., Mladonicky, A., Goodwin, G., 1990.Characteristics of workers with upper extremity injurieswho make a successful transition to work. Journal of HandTherapy 3, 51–55.Frankel, V.H., Nordin, M., 1980. Basic Biomechanics of theSkeletal System. Lea and Febiger, Philadelphia.

Hadler, N.M., 1992. Arm pain in the workplace: a small areaanalysis. Journal of Occupational Medicine 34, 113–118.

Hennig, E.M., Milani, T.L., Lafortune, M.A., 1993. The use ofground reaction force parameters for the prediction of peaktibial accelerations in running. Journal of Applied Biome-chanics 9, 306–314.

Higgs, P., Yount, V.L., Seaton, M., Edwards, D., Freely, C.,1992. Upper extremity impairment in workers performingrepetitive tasks. Plastic and Reconstructive Surgery 90,614–620.

Karwowksi, W., Yates, J.W., 1986. Reliability of the psycho-physical approach to manual lifting of liquids by females.Ergonomics 29, 237–248.

Kendall, D., 1960. Aetiology, diagnosis, and treatment ofparaesthesiae in hands. British Medical Journal 2, 1633–1640.

Kim, C.H., Fernandez, J.E., 1993. Psychophysical frequencyfor a drilling task. International Journal of IndustrialErgonomics 12, 209–218.

Kroemer, K.H.E., 1992. Avoiding cumulative trauma disordersin shops and o"ces. American Industrial Hygiene Associa-tion Journal 53, 596–604.

Lafortune, M.A., 1991. Three-dimensional acceleration of thetibia during walking and running. Journal of Biomechanics24, 877–886.

Manenica, I., 1986. A technique for postural load assessment.In: Corlett, N., Wilson, J., Manenica, I. (Eds.), TheErgonomics of Working Postures. Taylor & Francis,London, pp. 270–277.

Marley, R.J., Fernandez, J.E., 1995. Psychophysical frequencyand sustained exertion at varying wrist postures for a drillingtask. Ergonomics 38 (2), 303–325.

McAtamney, L., Corlett, E.N., 1993. RULA: a survey methodfor the investigation of work-related upper limb disorders.Applied Ergonomics 24, 91–99.

Mital, A., Nicholson, A.S., Ayoub, M.M., 1993. A Guide toManual Materials Handling. Taylor & Francis, London.

Moore, A., Wells, R., Ranney, D., 1991. Quantifying exposurein occupational manual tasks with cumulative traumadisorder potential. Ergonomics 34 (12), 1433–1453.

Nigg, B.M., Bahlsen, H.A., Luethi, S., Stokes, S., 1987. Theinfluence of running velocity and midsole hardness onexternal impact forces in heel–toe running. Journal ofBiomechanics 20, 951–959.

NIOSH Technical Report, 1981. Work practices guide formanual lifting. DHHS (NIOSH) Publication No. 81, 122pp.

Putz-Anderson, V., 1988. Cumulative Trauma Disorders:Manual for Musculoskeletal Diseases of the Upper Limbs.Taylor & Francis, Bristol, PA.

Rhomert, W., 1973. Problems in determining rest allowances.Part 1: use of modern methods to evaluate stress and strain instatic muscular work. Applied Ergonomics 4, 91–95.

Silverstein, B.A., Fine, L.J., Armstrong, T.J., 1986. Hand–wristcumulative trauma disorders in industry. British Journal ofIndustrial Medicine 43, 779–784.

Silverstein, B.A., Fine, L.J., Armstrong, T.J., 1987. Occupa-tional factors and carpal tunnel syndrome. American Journalof Industrial Medicine 11, 343–358.

Snook, S.H., 1978. The design of manual handling tasks.Ergonomics 21 (12), 963–985.

Snook, S.H., 1985. Psychophysical considerations in permis-sible loads. Ergonomics 28 (1), 327–330.

Snook, S.H., Ciriello, V.M., 1991. The design of manualhandling tasks: revised tables of maximum acceptableweights and forces. Ergonomics 34 (9), 1197–1213.

Snook, S.H., Irvine, C.H., Bass, S.F., 1970. Maximum weightsand work loads acceptable to male industrial workers.American Industrial Hygiene Association Journal 31, 579–586.

Snook, S.H., Vaillancourt, D.R., Ciriello, V.M., Webster, B.S.,1995. Psychophysical studies of repetitive wrist flexion andextension. Ergonomics 38, 1488–1507.

Snook, S.H., Vaillancourt, D.R., Ciriello, V.M., Webster, B.S.,1997. Maximum acceptable forces for repetitive ulnardeviation of the wrist. American Industrial Hygiene Associa-tion Journal 58 (7), 509–517.

Statistics Canada, 1993. Work injuries 1990–1992. Report72–208.

Stevens, S.S., 1967. Intensity functions in sensory systems.International Journal of Neurology 6 (2), 202–209.

J.R. Potvin et al. / International Journal of Industrial Ergonomics 26 (2000) 625–637 637