Effect of Wheels, Casters and Forks on Vibration Attenuation and Propulsion Cost of Manual Wheelchairs

Manual wheelchair users are exposed to whole-body vibrations as a direct result of using their wheelchair. Wheels, tires, and caster forks have been developed to reduce or attenuate the vibration that transmits through the frame and reaches the user. Five of these components with energy-absorbing characteristics were compared to standard pneumatic drive wheels and casters. This study used a robotic wheelchair propulsion system to repeatedly drive an ultra-lightweight wheelchair over four common indoor and outdoor surfaces: linoleum tile, decorative brick, poured concrete sidewalk, and expanded aluminum grates. Data from the propulsion system and a seat-mounted accelerometer were used to evaluate the energetic efficiency and vibration exposure of each configuration. Equivalence test results identified meaningful differences in both propulsion cost and seat vibration. LoopWheels and SoftWheels both increased propulsion costs by 12-16% over the default configuration without reducing vibration at the seat. Frog Legs suspension caster forks increased vibration exposure by 16-97% across all four surfaces. Softroll casters reduced vibration by 11% over metal grates. Wide pneumatic ‘mountain’ tires showed no difference from the default configuration. All vibration measurements were within acceptable ranges compared to health guidance standards. Out of the component options, softroll casters show the most promising results for ease of efficiency and effectiveness at reducing vibrations through the wheelchair frame and seat cushion. These results suggest some components with built-in suspension systems are ineffective at reducing vibration exposure beyond standard components, and often introduce mechanical inefficiencies that the user would have to overcome with every propulsion stroke.


I. INTRODUCTION
H UMANS are often exposed to vibrations through extra- 36 neous sources during activities of daily life within and 37 outside the household, and especially in the workplace. In cer-38 tain doses, vibration exposure has been correlated to positive 39 effects on physiological health [1], [2]. In other situations, 40 deleterious health effects are instead attributed to the vibration, 41 namely in the form of neck and lower back pain [3], [4] and 42 fatigue [5] in seated persons in the workplace. To combat 43 the health risks from workplace vibration exposure, orga-44 nizations including the International Standards Organization 45 (ISO) and the Health and Safety Executive (HSE) in the 46 United Kingdom have developed guidelines and thresholds for 47 potential health risks associated with an 8-hour duration of 48 whole-body vibration (WBV) exposure [6], [7]. Magnitudes 49 and frequencies of the vibrations, as well as the durations 50 of exposure, are considered in these assessments. However, 51 as these guidelines were developed around the comfort of only 52 seated, able-bodied humans in the workplace, they may not 53 appropriately reflect permissible vibration exposure levels for 54 non-able-bodied individuals performing everyday activities. 55 Manual wheelchair (MWC) users are exposed to WBV 56 directly from the use of their wheelchair. While there are 57 no conclusions about the full extent of adverse health effects 58 from MWC-induced WBV [8], some common ailments and 59 comorbidities of MWC users such as lower back pain [3], neck 60 pain [9] and fatigue [10] may be explained by constant WBV 61 exposure, and many wheelchair users have spinal impairments 62 that could explain symptom onset for a low amount of vibra-63 tion exposure. Several articles contend that WBV exposure 64 levels of MWC users are non-negligible [8], [10], [11], [12], 65 [13], [14]. However, research suggests that the duration of 66 time moving in a manual wheelchair is very low, on the 67 order of 1 hour per day [15], [16], [17], compared to the 8-68 hour exposure period used by ISO and HSE. Since vibration 69 exposure is based upon time of exposure, the implication for 70 users remains unclear. 71 Few studies to date [12], [13], [14], [18] have 72 reported harmful MWC vibration exposure levels. 73 Garcia-Mendez et al. [14] calculated WBV from total 74 daily occupancy time (13 hours). This implies 12 hours of 75 vibration exposure, on average, occurred when the user was 76 occupying but not propelling the wheelchair. While MWC 77 users may also experience vibrations when they are not in 78 This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/ propelled across three test tracks with standardized wheels and 138 tires. Their findings support the vibration attenuation capabil-139 ities of carbon fiber and, though no significant difference in 140 propulsion cost were reported, negative correlations present in 141 the data analysis suggest that frame-based energy absorption 142 reduces vibration exposure at the likely expense of mechanical 143 propulsion efficiency [32]. Optimization of MWCs require 144 careful assessment of the anticipated environments of use, 145 as propulsion cost and vibration exposure can vary across 146 surfaces, as well as between configurations of tires and weight 147 distributions.

148
The objective of this study is to assess the impact of five 149 commercially-available components on propulsion cost and 150 vibration exposure, as measured at three locations on the 151 wheelchair frame. The breadth of measurement is vital to 152 identify the potential benefit of vibration attenuation in relation 153 to the potential for added propulsion cost. Four common 154 indoor and outdoor surfaces were selected that represent a 155 wide range of commonly-traveled surfaces: decorative brick, 156 expanded aluminum grates, sidewalk, and linoleum tile. These 157 surfaces were chosen because they induce steady-state vibra-158 tions (i.e., experienced every single time the wheelchair is 159 in motion), rather than the comparatively infrequent low-160 frequency, high-magnitude shocks or impacts experienced by 161 users (e.g., traversing thresholds, curb-drops, and potholes). 162 A wheelchair-propelling robotic testbed was used to drive each 163 configuration with highly repeatable trajectories to standard-164 ize the travel path, maneuver speed, and weight distribution 165 over the components to permit accurate measurements of the 166 propulsion costs as reported in prior work [33].  Figure 1, is a wheelchair-propelling robot that was used 173 to maneuver the wheelchair in this study. Its propulsion 174 subsystem permits highly repeatable, configurable propulsion 175 patterns to be deployed across a wide range of chair configu-176 rations. Its construction mimics a seated person in size, shape, 177 and mass distribution to apply realistic loads to the frame and 178 wheels as per the wheelchair test dummy standard defined by 179 ISO 7176-11 [34], scaled to a total mass of 80 kg to more 180 closely represent the average occupant mass reported in [35]. 181 Wheelchair propulsion is controlled by motors attached to 182 custom-made ring gears replacing the push-rims of each drive 183 wheel. A high-powered motor speed controller (HDC 2460, 184 RoboteQ Inc.) utilizes velocity-based feedback system to 185 impart discrete and highly repeatable pushes. Between pushes, 186 the motors are mechanically disconnected from the push-187 rims, allowing the wheelchair to freely coast. Rotational       tires (Primo Orion, Xiamen Lenco Co., Ltd.) inflated to the 206 recommended 75 psi on metal spoked drive wheels, and 207 solid 5 × 1" urethane caster wheels (Primo 5 × 1, Xiamen 208 Lenco Co., Ltd.). These components served as the basis of 209 comparison as they are common standard options offered by 210 manufacturers at no additional cost to the user.

211
The chosen 'energy-absorbing' components were selected 212 based on their assumed dampening capabilities, and would 213 incur additional monetary cost to add to any user's wheelchair 214 configuration. These included: 24 × 2.   When testing a drive wheel, the standard 5 × 1" caster 230 was used, and when testing a caster component, the standard 231 24 × 1.3/8" pneumatic tired was used. Mass and weight 232   the direction most closely associated with discomfort [37] and 256 physiological injury [5], [9].

257
The accelerometer near the caster captured the most severe 258 vibrations [11], [38], as the casters are the first component Tests were conducted on four surface types: decorative 297 brick, expanded aluminum grates, sidewalk, and linoleum tile, 298 collectively seen in Figure 6. To account for any slopes or 299 inconsistencies in each surface, 6 trials were run in opposing 300 directions along the same path for a total of 12 trials per 301 surface per configuration, or 288 trials overall. Data from 302 the motor armature current sensors, motor encoders, and 303 wheel-mounted encoders were collected at 40 Hz during the 304 over-ground trials. These data were processed in MATLAB 305 (R2020a, The Mathworks Inc.).  Ratios with values of 1.00 reflect similar performance 363 between the test and reference configurations. Ratios less than 364 1.00 indicate that the test configuration exhibited better per-365 formance (i.e., lower propulsion cost or vibration exposure).

366
To assess the meaningfulness of these ratios, the 95% CIs 367 were compared to pre-defined upper and lower equivalence 368 limits (UEL, LEL). The limits for propulsion cost were 369 informed from several published studies on wheelchair propul-370 sion efforts where human subjects were asked to propel a wide 371 variety of wheelchair configurations including power-assisted 372 wheels [44], lever-driven wheels [45], sports wheelchairs [46], 373 chairs with weights added to the frame [47], [48], and with 374 under-inflated tires [49], [50]. Across these studies, the average 375 difference between the biomechanical outcome variables of 376 the studies was calculated to be 9.4%. Therefore, we defined 377 the equivalence limits as ±5% based on the assumption 378 that mechanical testing is more precise than human subject 379 investigation. Similarly, UEL and LEL values were defined 380 for vibration exposure from published studies on wheelchair 381 seat vibration measurements with human subjects propelling 382 one wheelchair across different styles of sidewalk [13] and 383 wheelchairs of different frame materials over various sur-384 faces [32], as well daily exposure using folding or rigid MWC 385 frames [14]. The average difference between r.m.s. vibration 386 values for the independent variable groups was calculated 387 to be 6.2%. Therefore, the seat, caster, and frame vibration 388 equivalence limits were set to ±6%.

389
The extents of each 95% confidence interval were compared 390 against equivalence limits (±5% for propulsion cost, ±6% for 391 all vibration metrics). Confidence intervals that are completely 392 below 0.95 (propulsion cost) or 0.94 (vibrations) were classi-393 fied as 'Superior' as the test configuration experienced more 394 preferable performance (i.e., lower propulsion costs or vibra-395 tions than the Default configuration). Conversely, confidence 396 intervals completely above 1.05 (propulsion cost) or 1.06 397 (vibrations) were classified as 'Inferior'. Confidence intervals 398 that cross one or both limits, or that reside completely within 399 the limits, are considered 'Comparable' as the test and Default 400 configurations exhibited similar performances. Examples of 401 these classifications are shown in Figure 7.   Not unexpectedly, the five components exhibited disparate 460 performances in the main outcome variables (propulsion cost 461 and seat vibration) over different surfaces. The significant 462 differences summarized in Table VI imply that many of the 463 components specifically designed to reduce vibration exposure 464 of MWC users were not effective at this task. In particular, 465 SoftWheels and LoopWheels appear ineffective at vibration 466 attenuation, even incurring 12-17% greater vibrations at the 467 seat over most surfaces, with increased propulsion costs.     Secondary outcome variables (frame and caster vibrations) 501 also differed across components. These variables offer inter-502 esting mechanical insights to the designs of the components 503 and of the wheelchair as a complete system. Energy-absorbing 504 components must be engineered according to the types of 505 accelerations that are intended to be attenuated. With respect 506 to wheelchairs, components will have different performances 507 according to the speed or surface. Transient impacts might 508 be handled differently than the more steady-state vibrations 509 experienced during WBV measurements. The commercial 510 manufacturers of the components evaluated in this study do 511 not fully disclose the types of surfaces, obstacles, or speeds 512 for which the designs were optimized.

513
Vibration is a continuous, periodic perturbation to a system, 514 like rolling down a tiled hallway floor, whereas shock is a 515 sudden and infrequent impulse like impacting a door threshold. 516 To assess shock-induced vibrations, the fourth-power vibration 517 dose value (VDV, expressed in m/s 1.75 ) is often calculated 518 as a complementary analysis to the basic evaluation method 519 (r.m.s. acceleration), as described in ISO 2631-1 [6]. VDV is 520 more sensitive to acceleration peaks than r.m.s. and is used 521 when intermittent shock exposure is present. As a cumulative 522 force to be used as a fatigue-to-failure test apparatus [52]. . One bicycle simulation using 572 SoftWheels showed significant (17-26%) reductions of vibra-573 tion over asphalt and small shock-inducing obstacles, with 574 an estimated 14% loss in efficiency at cruising speeds [54]. 575 The cyclic compression and re-extension of the three damper 576 spokes were identified as a periodic energy loss in rolling, 577 as well as a cause of non-negligible pitch perturbation [54]. 578 This periodic energy absorption likely explains the propulsion 579 cost increases for SoftWheels and LoopWheels.

580
The MtnWheel was expected to reduce vibration exposure 581 due to its wider contact patch and lower tire pressure compared 582 to the Default pneumatic wheels. This proved to be untrue. 583 However, MtnWheels are designed for outdoor environments 584 (e.g., dirt and grass), where their intended benefit is the 585 added tread and traction, rather than vibration attenuation. 586 Overall, MtnWheels had comparable propulsion costs and 587 seat vibrations to the Default configuration. Over smooth tile, 588 the miniscule recurring impacts from the 'knobby' tire tread 589 contacting the floor appeared to cause an increase in frame 590 vibration. Vibrations transmitted from the frame to the seat 591 must have been dampened by the cushion, as the seat vibra-592 tions were comparable between MtnWheels and Default. This 593 suggests some vibrations from the tires are permissible without 594 impacting the WBV exposure of the user. In future efforts, 595 an assessment of frequency-based vibration magnitudes could 596 be used to investigate if this damping effect from the frame, 597 seat, and cushion shifted the acceleration frequencies present 598 in the measured vibration. As in [19], some tradeoffs may exist 599 where the component reduces overall shock or vibration, but 600 shifts the vibration into frequencies most commonly associated 601 with injury and discomfort in the human body.

602
Finally, the Softroll caster was expected to exhibit similar 603 energy loss parameters as the default caster [30] with a 604 small effect on propulsion cost [33]. Wheels lose energy 605 primarily through hysteresis, or the cyclic deformation 606 and restoration of infinitesimal elastomeric radial elements 607 that make up the tire [55]. Unlike the discrete elastomeric 608 elements of the suspension wheels and caster fork, Softroll 609 deformation is simultaneous and omni-directional, providing a 610 smoother ride without potential misalignment. The dampening 611 benefit of Softrolls over the default casters is likely due to 612 a combination of the wider contact patch and material 613 properties, like pliability. and satisfaction might also be a reason to consider energy-632 absorbing components, but this decision will be improved by Manual wheelchair mobility exposes the user to whole-680 body vibrations that may have adverse health effects. Vibration 681 exposure can be minimized without sacrificing the mechanical 682 efficiency of the vehicle, though some component options 683 have negative effects on vibrations and/or propulsion cost. 684 In this study, robotic wheelchair propulsion permitted a 685 repeated-measures test on wheelchair propulsion cost and 686 vibration exposure over four common surfaces at a clinically-687 relevant speed. Six components with energy-absorbing char-688 acteristics were compared to standard pneumatic drive wheels 689 and casters. Wheels with built-in suspension did not have 690 any significant difference on vibration exposure, yet they did 691 significantly increase the propulsion cost, effectively making 692 the wheelchair more energetically-expensive to propel. Simi-693 larly, elastomeric suspension forks significantly increased the 694 vibration. Out of the component options, 'soft roll' casters 695 show the most promising results for ease of efficiency and 696 effectiveness at reducing vibrations through the wheelchair 697 frame and seat cushion. Users may define other reasons to 698 consider energy-absorbing components, but this decision will 699 be improved by understanding the performance under typical 700 conditions which dominate everyday mobility.