SCVT 010 IEEE Symposium on Communictions nd Vehiculr Technology in the Benelux 1 Wireless Vibrtion Monitoring on Humn Mchine Opertor Frederik Petré, Frnk Bouwens, Steven Gillijns, Fbien Mssé, Mrc Engels, Bert Gyselinckx, Kris Vnstechelmn, nd Christophe Thoms Abstrct Humn mchine opertors re often subject to extreme shocks nd vibrtions while operting production mchines nd vehicles. To ssess the impct on perceived comfort objectively, wireless vibrtion monitoring system is needed tht mesures whole-body vibrtions directly on the humn body. To this end, we hve developed wireless body re network consisting of low-power vibrtion sensor nodes tht hve smll nd ergonomic form fctor nd tht re esy to instll. Furthermore, we hve vlidted the BAN long with the necessry post-processing of the rw vibrtion signls on rel industril cse, i.e. the driver of forklift. Our system proves to be instrumentl in optimizing criticl tuning prmeters of mchine, exemplified by the trnsmission control prmeters of forklift. Index Terms Personl nd body re networks, Wireless condition monitoring, Whole-body vibrtion, Helth nd comfort, Forklift. F I. INTRODUCTION OR mny production mchines nd vehicles, their humn mchine opertors (or pssengers) re subject to extreme shocks nd vibrtions during opertion. Exmples of such mchines re forklifts, bulldozers, pneumtic hmmers, drill hmmers, trins [1][], combine hrvesters nd trctors [3][4]. To sfegurd humn helth, regultions such s the EU Vibrtions Directive (00/44/EC [5]) impose strict quot for these vibrtions, mking distinction between hnd-rm nd whole-body vibrtions. The mesurement of whole-body vibrtion is defined in the ISO 631 stndrd [6], which lso specifies the impct of whole-body vibrtions on humn helth nd comfort. Complince with the regultions nd the comfort for the opertor is currently checked by defining number of typicl ppliction scenrios nd by hving the humn opertor to score between 1 nd 10 on perceived comfort. Given the very subjective nture of such n pproch with observtions F. Petré, S. Gillijns, nd M. Engels re with the Flnders Mechtronics Technology Centre (FMTC), Celestijnenln 300D, B-3001 Leuven, Belgium (phone: +3-16-38053; fx: +3-16-38064; e-mil: frederik.petre@fmtc.be). F. Bouwens, F. Mssé, nd B. Gyselinckx re with the Holst Centre / IMEC-NL, High Tech Cmpus 31, 5656 AE Eindhoven, The Netherlnds (emil: frnk.bouwens@imec-nl.nl). K. Vnstechelmn, nd C. Thoms re with the Spicer Off-Highwy Product Division of Dn Holding Corportion, Ten Briele 3, B-800 Brugge, Belgium (e-mil: kris.vnstechelmn@dn.com). lrgely vrying cross different opertors, it is very hrd to obtin objective mesures nd hrd fcts. This often leds to endless discussions between supplier of certin equipment (e.g. trnsmission controller softwre, mechnics, hydrulics) nd its customer (e.g. OEM of forklifts) on perceived qulity nd comfort. To increse the objectivity of complince nd comfort testing, cler need exists for wireless monitoring system tht mesures whole-body vibrtions directly on the body of the humn opertor. The min benefit of wireless solution is tht it llows mesuring on the right plce, i.e. on the humn body, while voiding the hssle, the limittions on mobility or freedom of movement, nd the instlltion effort ssocited with cbled solution. For pure monitoring scenrio, it is nticipted tht the system will be used for registrtion of humn mchine opertor s totl exposure to whole-body vibrtions, much in the sme wy s totl rdition exposure is being monitored in nucler power plnts. Commercil wireless whole-body vibrtion mesurement systems re limited to single set sensors [7]. According to the ISO 631-1 stndrd [6], this is sufficient for helth ssessments but not for perceived comfort. For comfort mesurements, ccelertions from multiple loctions must be combined. We found preliminry version of wireless sensor network to mesure vibrtions on opertors [8], but this network hs lot of limittions (e.g. high power consumption, lrge size, no rel-time dt trnsmission) nd ws clerly not yet industry-redy. Recently, Inerti Technology [9], spinoff of the University of Twente, hs engged in the commerciliztion of wireless motion sensors, building further on the PhD disserttion of Mrin-Perinu [10]. Becuse the nodes re bsed on multi-sensor inertil pltform, including gyroscopes, the energy utonomy is limited. Finlly, s the mesurement systems re not vilble, no work is reported on the nlysis of the influence of mchine prmeters on comfort. In this pper, we demonstrte the technicl fesibility of wireless vibrtion monitoring for helth nd comfort on humn mchine opertor. To this end, we hve developed body re network (BAN) consisting of 4 sensor nodes tht cn ech mesure 3D liner ccelertions. The sensor nodes re low-power, hve smll nd ergonomic form fctor nd re esy to instll. Furthermore, we hve creted n offline post-processing tool for clculting the whole-body vibrtion from the rw vibrtion mesurements. Finlly, we hve
SCVT 010 IEEE Symposium on Communictions nd Vehiculr Technology in the Benelux vlidted the BAN nd the post-processing tool on rel industril cse, i.e. the driver of forklift, by identifying the reltion between whole-body vibrtion nd criticl design nd tuning prmeters of the forklift. Our pper is orgnized s follows. Section II reviews the mesurement nd evlution of whole-body vibrtion ccording to the ISO 631-1 stndrd. Section III describes the BAN long with its sensor nodes thtt we hve developed for performing wireless vibrtion mesurements on humn opertor. Section IV nlyzes the mesurement results we hve obtined from n in-field test cmpign on forklift driver. Finlly, Section V summrizes our mjor conclusions. II. STANDARD EVALUATION OF WHOLE-BODY VIBRATION We hve evluted the exposure of the humn driver of forklift to whole-body vibrtion ccording to the ISO 631 stndrd [6]. This stndrd defines methods of quntifying whole-body vibrtion in reltion to (i) humn helth; (ii) comfort nd probbility of vibrtion perception; nd (iii) incidence of motion sickness. The focus of the stndrd is on periodic, rndom nd trnsient whole-body vibrtion excluding extreme-mgnitude single shocks. The frequency spectrum of interest rnges from 0.5 Hz to 80 Hz for evluting the impct on helth, comfort nd perception; nd from 0.1 Hz to 0.5 Hz for evluting the impct on motion sickness. In the next subsections, we will briefly discuss the stndrd s guidelines on vibrtion mesurement nd vibrtion evlution, respectively. A. Vibrtion Mesurement In this subsection, we briefly review the stndrd s min guidelines towrds mesuring whole-body vibrtions, including the loction nd direction of mesurement s well s some generl requirements for signl conditioning. The primry quntity of vibrtion mgnitude is ccelertion, which should be mesured t the interfcee between the humn body nd the source of vibrtion. For seted persons, s is the cse for the driver of forklift, there re three principl loctions of mesurement, i.e. the supporting set surfce, the set bck nd the feet (see Figure 1). More specificlly, mesurements on the supporting set surfce should be performed beneth the ischil tuberositiess (lso known s the sitz bone). Mesurements on the set-bck should be performed in the re of principl support of the body. Mesurements t the feet should be performed on the surfce on which the feet re most often supported. Figure 1. Mesurement loctions nd directions for seted person [6]. The direction of mesurement should be ccording to bsicentric coordinte system originting from point where whole-body vibrtion is considered to enter the humn body. This is illustrted in Error! Reference source not found.. The sensitive xes of the vibrtion trnsducer my devite from the preferred xes by up to 15. The stndrd lso imposes three generl requirements for signl conditioning. First, the frequency response of the vibrtion trnsducer nd the ssocited signl conditioning prior to signl processing should be pproprite to the rnge of frequencies. Second, the dynmic rnge of the signl conditioning should be dequte for the highest nd lowest signls. Third, low-pss filter cn be pplied with cut-off frequency (-3 db) of pproximtely 1.5 times the highest frequency of interest, nmely 10 Hz, nd phse chrcteristic tht is liner within the rnge of frequencies. B. Vibrtion Evlution In this subsection, we briefly summrize the stndrd s pproch towrds evluting whole-body vibrtion mesurements. The bsic evlution method uses so-clled frequency-weighted root-men-squre (RMS) ccelertions. Before clculting the RMS ccelertion, the ccelertion time signl is filtered with the pproprite frequency weighting. The frequency weighting emphsizes certin prts of the spectrum while de-emphsizing others. The principl frequency weightings W d, W k nd W f re illustrted in Figure. Figure. Principl frequency weightings [6].
SCVT 010 IEEE Symposium on Communictions nd Vehiculr Technology in the Benelux 3 The pplicbility of the bsic evlution method depends on the so-clled crest fctor of the frequency-weighted ccelertion signl, which is defined by the stndrd s the modulus of the rtio of the mximum instntneous pek vlue of the frequency-weighted ccelertion signl to its RMS vlue. The bsic evlution method is pplicble for vibrtions with crest fctor t most equl to 9, hence excluding the presence of extreme-mgnitude single shocks. For signls with crest fctor lrger thn 9, the bsic evlution method is not sufficient nd dditionl methods re needed for dequte evlution, such s the running RMS method, the fourth power vibrtion dose method nd rtios used for comprison of the bsic nd dditionl evlution methods. In this work, however, we hve exclusively used the bsic evlution method, since this proved to be sufficient for most of our mesurements (see Section IV). Once the frequency-weighted RMS ccelertions hve been determined in the three orthogonl directions, they should be combined into single RMS vlue for tht loction. The totl vibrtion vlue of the weighted RMS ccelertion is clculted s follows: v = k x wx + where wx, wy, wz, re the weighted RMS ccelertions with respect to the orthogonl xes x, y, z, respectively, nd k x, k y, k z, re the multiplying fctors. To investigte the effects of periodic, rndom nd trnsient vibrtion on the helth of persons in norml helth condition exposed to whole-body vibrtion t work, the weighted RMS ccelertion for ech xis on the supporting set surfce should be evluted t frequencies from 0.5 Hz to 80 Hz. In specific, the frequency weightings nd vibrtion combintion should be mde s follows: X-xis: W d, k x = 1,4; Y-xis: W d, k y = 1,4; Z-xis: W k, k z = 1. Biodynmic reserch hs proven n elevted risk of helth problems due to long-term exposure with high-intensity whole-body vibrtion. Minly the lumbr spine nd the connected nervous system my be ffected. Since responses re relted to energy, wht relly mtters is the so-clled vibrtion dose or vibrtion exposure, which is the product of vibrtion intensity nd exposure durtion. This is illustrted in Figure 3, which defines helth guidnce cution zones in the weighted ccelertion versus exposure durtion plne. The helth guidnce cution zone is defined by the region in between the two dshed lines (or, lterntively, in between the two dotted lines) in Figure 3. For exposures below the zone, helth effects hve not been objectively observed. For exposures within the zone, cution with respect to potentil helth risks is indicted. For exposures bove the zone, helth risks re likely. k y wy + k z wz Figure 3. Helth guidnce cution zones [6]. To investigte the effects of periodic, rndom nd trnsient vibrtion on the comfort nd perception of seted persons in norml helth exposed to whole-body vibrtion t work, the weighted RMS ccelertion for ech xis on the supporting set surfce should be evluted t frequencies from 0.5 Hz to 80 Hz. For this cse, the frequency weightings nd vibrtion combintion should be mde s follows: X-xis: W d, k x = 1; Y-xis: W d, k y = 1; Z-xis: W k, k z = 1. The impct of whole-body vibrtion on comfort depends on mny fctors tht vry with ech ppliction. The vlues shown in Tble 1 give pproximte indictions of likely rections to vrious vibrtion mgnitudes encountered in public trnsport. However, these rections depend on pssenger expecttions with regrd to trip durtion, the type of ctivities performed by pssengers (e.g. reding, writing, eting, etc.), nd mny other fctors such s coustic noise nd temperture. Tble 1. Rection t vrious vibrtion mgnitudes in public trnsport [6]. Less thn 0.315 m/s Not uncomfortble 0.315 m/s to 0.63 m/s Little uncomfortble 0.5 m/s to 1 m/s Firly uncomfortble 0.8 m/s to 1.6 m/s Uncomfortble 1.5 m/s to.5 m/s Very uncomfortble Greter thn m/s Extremely uncomfortble III. SENSOR NODES FOR WIRELESS VIBRATION MEASUREMENTS In this section, we introduce the BAN long with its sensor nodes tht we hve developed for performing wireless vibrtion mesurements on humn opertor. The vibrtion mesurements re continuously obtined from 3D ccelerometers vi wireless link from sensor node. The UniNode, s depicted in Figure 4, is smll nd generic pltform tht is designed for BANs s described in [11]. The UniNode filters the vibrtions in the x, y, nd z-xis with bndwidth of 0 10 Hz nd smples them with frequency of 50 Hz. It trnsmits the smples t rte of 3.75 kbps vi
SCVT 010 IEEE Symposium on Communictions nd Vehiculr Technology in the Benelux 4.4 GHz str-network link to the sensor bse sttion tht is connected to the computer. The 3D ccelerometer hs rnge of -3.6G 3.6G [1] nd is smpled with 1-bit ADC, which gives sensitivity of.4 mg. The node mrks ech smple tht is trnsmitted with time stmp for synchroniztion of the dt nd future nlysis. To reduce power consumption nd increse the bttery life time of the sensor nodes, severl power mngement techniques were pplied. In this ppliction cse, the node is ble to operte for 64 hours consecutively on fully chrged bttery of 150 mah. re not chnged during the experiment. These do hve influence on the mesured results, but re not quntified in this experiment. The low-pss filtered ccelertion nd bsolute vlue of the velocity in the first test cse ( good trnsmission qulity ) re depicted in Figure 6 in time frme of 10 seconds. The consecutive ctions consist of series of forwrd nd bckwrd driving with different ccelertions. The vibrtions of the set-surfce (in ll test cses) showed most ctivity nd re depicted s blue line in Figure 10. The frequency spectr of the signls re shown in Figure 8. The ltter shows tht the vibrtions in ll three xes re well present in the rnge of 0 30 Hz. Forwrd Bckwrd Forwrd Bckwrd Figure 4. UniNode for Body Are Networks. There re three UniNodes locted on the humn body t the principl res (s noted in Subsection II.A) for seted persons: set surfce, lower bck, nd right foot. All sensors re tightly worn on the body to void movement rtifcts of the node. A fourth node is ttched to the forklift itself (nd more specificlly to its counterweight in the bck) to isolte the vibrtions cused by the device nd to perform snity check of the nodes. Additionlly, the forklift is lso equipped with tchometer to determine its speed. The xes directions of ll UniNodes re the sme s depicted in Figure 5 for ese of discussion. The sensor bse sttion, which is connected to the computer, is locted in the forklift s cbin. Front +y +x UniNode +z Bck Figure 6. Test cse 1 ccelertion nd velocity of forklift. Further processing of the dt consists of filtering (frequency weighting) the signls s shown s red signl in Figure 7. The bsic evlution both for Helth nd Comfort is selected for weighting the ccelertions, s discussed in Subsection II.B. This weighs the x- nd y-xes with W d nd the z-xis with W k s shown in Figure. Even though the z-xis seems to hve the lowest intensity in Figure 7, the post-processed results clerly shows the significnce of the z-xis. Figure 5. Orienttion of UniNode in test cses. IV. MEASUREMENT RESULTS In this section, we nlyze the mesurement results obtined from n in-field test cmpign on forklift driver. Three test cses were defined to mesure vibrtions on the humn driver of forklift under relistic conditions. The forklift hs the cpbility to chnge trnsmission controller prmeters ccording to the terrin. In the first nd second test cses the forklift drives over flt terrin with different directions (forwrd nd reverse) nd speeds. In the first test cse, the prmeters re set such tht good trnsmission qulity is obtined, while in the second test cse the trnsmission is purposely configured for degrded trnsmission qulity. In the third test cse the forklift drives over n obstcle to crete two hrsh impulses due to forwrd nd bckwrd driving direction. In ll the test cses the suspension of the mchine nd chir Figure 7. Showing corresponding (originl nd filtered) vibrtions of set-surfce UniNode in test cse 1. Vibrtions re filtered with frequency weightings defined in ISO631-1.
SCVT 010 IEEE Symposium on Communictions nd Vehiculr Technology in the Benelux 5 Comfort it is 0.8467 m/s. The severity of vibrtions for humn helth is depicted in Figure 9, indicting the helth risk hs incresed w.r.t. the first test cse with good trnsmission qulity. On the comfort level, the vibrtions re now experienced s firly uncomfortble to uncomfortble, clerly indicting the comfort level hs decresed w.r.t. the first test cse. Figure 8. Frequency Spectrum of set-surfce UniNode in Test cse 1. Highest intensity in the rnge of 0 30 Hz. From ech xis both the RMS vlue nd the crest fctor re clculted. The crest fctor for ech xis is in the rnge of 3.5 5, which justifies the use of the bsic evlution method. By using the pproprite multiplying fctors k x, k y, nd k z, the frequency-weighted RMS vlues per xis re combined into n overll RMS ccelertion for Helth (0.798 m/s ) nd Comfort (0.7515 m/s ), respectively. Using the chrt in Figure 3 for Helth, the severity of the vibrtions in this test cse is still within the sfety zone when operting less thn four hours s shown in Figure 9. On the comfort level, the vibrtions in test cse 1 re experienced by the driver s firly uncomfortble ccording to Tble 1. In test cse ( degrded trnsmission qulity ) the sme driving ctions re performed s in test cse 1 (see Figure 6). The pre- nd post-processed results re depicted in Figure 10. Notice tht only the z-xis signl hs mintined its intensity thnks to the frequency weighting s ws lso the cse in test cse 1. TC3: 1,4333 m/s TC: 0,90 m/s TC1: 0,798 m/s Figure 10. Showing corresponding (originl nd filtered) vibrtions of set-surfce UniNode in test cse. Vibrtions re filtered with frequency weightings defined in ISO631-1. The spordic vibrtions in the third test cse re depicted in Figure 11. The frequency spectrum in Figure 1 shows most intensity in the y nd z-xis in the rnge of 0 40Hz. After processing the highest intensity origintes in the z-xis when driving over the obstcles. Combining the RMS vlues of the different xes into the overll RMS ccelertion results in vlue of 1.4333 m/s for Helth nd 1.359 m/s for Comfort. The vlue for Helth is now in the dnger zone (bove the helth guidnce cution zone) when operting for longer thn 4 hours s depicted in Figure 9. Hence, helth risks re likely. From the viewpoint of Comfort, this is determined s n uncomfortble to very uncomfortble sitution for the driver. It is interesting to note tht the crest fctor is slightly bove 9 in the z-direction s it is cused by n impulse. As noted in Subsection II.B, in the strict sense the bsic evlution method is not sufficient nymore nd dditionl evlution methods such s running RMS or fourth power vibrtion dose re required. Implementing these nlysis tools ws out of scope in this experiment, but should be considered when nlyzing signls with extrememgnitude single shocks. Figure 9. Helth cution zones for test cses 1,, nd 3. We noticed tht the highest frequency intensity is lso in the rnge from 0 30 Hz similr s in test cse 1, which is s expected. From ech xis the RMS vlue is clculted tht results in crest fctor less thn 9. In test cse, the overll RMS vlue for Helth is now 0.90 m/s, while for
SCVT 010 IEEE Symposium on Communictions nd Vehiculr Technology in the Benelux 6 Forwrd drive wheels bump Figure 11. Third test cse mesurements when driving over n obstcle in forwrd nd bckwrd direction. Originl nd filtered vibrtions of set-surfce UniNode ccording to frequency weightings defined in ISO631-1. Figure 1. Frequency Spectrum of set-surfce UniNode in Test cse 3. High intensity till 40Hz in y- nd z-xes. V. CONCLUSION Bckwrd drive wheels bump Our wireless vibrtion monitoring system is suitble tool to mesure nd evlute whole-body vibrtions in reltion to humn helth nd perceived comfort of humn mchine opertor. The system ws successfully vlidted on n industril ppliction cse involving humn driver operting forklift under vrious test conditions. The wireless sensor nodes correctly cptured the 3D ccelerometer dt in different loctions on the humn body, while voiding the hssle, the limittions on mobility, nd the instlltion effort ssocited with cbled solution. The results from the field test cmpign showed tht our wireless vibrtion monitoring system cn relibly detect differences in terms of experienced helth effects nd perceived comfort between two levels of trnsmission qulity ( good versus degrded ). In ll cses, cre should be tken to minimize shocks nd void uncomfortble driving conditions nd helth risks over longer period of time. The post-processing tool implementing the bsic evlution method bsed on weighted RMS ccelertion proved to be dequte for most of our test cses. However, situtions with extreme-mgnitude shocks, e.g. cused by driving over obstcles, require dditionl nlysis techniques, such s running RMS or fourth power vibrtion dose. We believe our wireless vibrtion monitoring system should prove instrumentl in optimizing criticl tuning prmeters of mchine, s exemplified by the trnsmission control prmeters of forklift. Besides the trnsmission, other fctors such s the suspension nd the chir influence the mesured whole-body vibrtions nd the resulting helth effects nd perceived comfort. However, by creful design-ofexperiments, it should be fesible to isolte the effect of the trnsmission. This is topic for further reserch. In the long term, we envision our wireless vibrtion monitoring system to be prt of n intelligent trnsmission system tht djusts criticl prmeters in rel-time for optiml comfort ccording to the sitution. REFERENCES [1] A. R. Ismil, M. Z. Nuwi, C. W. How, N. F. Kmruddin, M. J. M. Nor nd N. K. Mkhtr, Whole Body Vibrtion Exposure to Trin Pssenger, Americn Journl of Applied Sciences, Vol. 7, No. 3, pp. 35 359, 010. [] G. Birlik, Occuptionl Exposure to Whole Body Vibrtion-Trin Drivers, Industril Helth, Vol. 47, pp. 5 10, 009. [3] Mkoto Futtsuk, Setsuo Med, Tsuks Inok, Megumi Ngno, Mshiro Shono, Tkshi Miykit, Whole-Body Vibrtion nd Helth Effects in the Agriculturl Mchinery Drivers, Industril Helth, Vol. 36, pp. 17-13, 1998. [4] Huub H.E. Oude Vrielink, Exposure to Whole-Body Vibrtion nd Effectiveness of Chir Dmping in High-Power Agriculturl Trctors Hving Different Dmping Systems in Prctice, ErgoLb Reserch, Report 1-10-009. [5] DIRECTIVE 00/44/EC OF THE EUROPEAN PARLIAMENT AND OF THE COUNCIL of 5 June 00 on the minimum helth nd sfety requirements regrding the exposure of workers to the risks rising from physicl gents (vibrtion) (sixteenth individul Directive within the mening of Article 16(1) of Directive 89/391/EEC). [6] Mechnicl Vibrtion nd Shock Evlution of Humn Exposure to Whole-Body Vibrtion Prt 1: Generl Requirements, Interntionl Orgniztion for Stndrdiztion, ISO 631-1, My 1997. [7] Cstle Group, Evec Stndrdized Mesurement nd Evlution of Whole-Body Vibrtion, Dtsheet. [8] Diogo Koenig, Mrild S. Chirmonte, Alexndre Blbinot, Wireless Network for Mesurement of Whole-Body Vibrtion, MDPI Journl on Sensors, Vol. 8, pp. 3067-3081, 008. [9] www.inerti-technology.com [10] R. S. Mrin-Perinu, Wireless Sensor Networks in Motion Clustering Algorithms for Service Discovery nd Provisioning, Ph. D. disserttion, Fculty of Electricl Engineering, Mthemtics nd Computer Science, University of Twente, The Netherlnds, 008. [11] Lindsy Brown, Bernrd Grundlehner, Jef vn de Molengrft, Julien Penders, nd Bert Gyselinckx, Body Are Networks for Monitoring Autonomic Nervous System Responses, Proceedings of 3rd Interntionl Conference on Pervsive Computing Technologies for Helthcre (PervsiveHelth 009), pp 1-3, April 009. [1] Anlog Devices, ADXL330, 3D ccelerometer dtsheet.