COMSOL ANALYSIS OF WHEELCHAIR PUSHRIM SIX DOF LOAD CELL FOR QUANTIFYING PROPULSION EFFICIENCY BY SENSEWHEEL
The SenseWheel is a lightweight force-sensing handrim, comprised of three identical load cells. The load cells are
interposed between the pushrim and drive wheel. The initial design SenseWheel Mk 1 measured the three orthogonal
forces Fx, Fy and Fz, and axial torque Tx, applied at each load cell . It has been constructed, calibrated, and used in a
limited clinical trial. SenseWheel Mk1 is: 1) light weight (less than 100g), 2) able to fit s different diameter wheel or hand
rim and 3) Waterproof. This paper describes the development of an improved Mk2 version of Sensewheel which uses
wireless 6DOF load cells to measure Fx, Fy, Fz, Tx, Ty, and Tz. This will simplify the coupling to the wheel and handrim.
Each load cell will transmit its data via Bluetooth to a master tablet/SmartPhone carried by the user or held remotely.
Methods: COMSOL was used to redesign the 6DOF load cell, using 8 quarter-bridge strain gauges equi-spaced around
the inner circumference of the widest part of the load cell for good strain sensitivity and selectivity. This provided some
redundancy in case of non-optimal gauge placement in the actual load cell. Each load cell is 40mm diameter at its widest,
to accommodate coin cell, Bluetooth module and flexible printed circuit. The optimum orientation of the gauges was
determined using COMSOL Multiphysics® with a 3D axis-symmetric finite element model generated from a 2D cross
sectional model. The load cell was designed as two halves, to be screwed together after assembly, and this was
modelled in COMSOL as one part. COMSOL was set to output direct strains (eX, eY, and eZ) as well as shear strains
(eXY, eYZ, and eZX) in response to applied forces Fx, Fy, Fz, and torques Tx, Ty, and Tz. The strains were calculated at
5 deg intervals along the circumference of the inside wall of the load cell. These strains were then converted from the
global axis of the model to the local axis at each strain gauge site using a rotation transformation matrix. The resulting
strains were used to investigate both the sensitivity of each gauge to each load type, and also the selectivity of each load
type. The gauge angle w.r.t. the longitudinal axis of the load cell was varied, and the strain sensitivities logged. The
gauges of one semicircle of gauges were angled opposite the other semicircle to allow for separation between Fx and
Mx, as previously used in an instrumented shoulder implant . The correlation coefficient between force/moment pairs
was then calculated to predict the optimum angle, which would give best separation between these load types.
Importantly, this analysis demonstrated that the strain response, of 8 strain gauges at 45deg intervals around the
circumference, to each applied load direction. is clearly distinguishable from the other load types.
Results: The peak sensitivity of the strain output by radial position, for several strain gauge angles w.r.t. the longitudinal
axis, and the correlation coefficient between same axis force/moment pairs, is shown in Table 1. Figure 1 shows the
strains at 5 deg increments around the circumference, for a gauge angle of 45deg, which gave a compromise between
good strain sensitivity to each applied load, and adequate selectivity between load types. The designed load cells will
now be manufactured, and following assembly each load cell will be calibrated to relate each strain output to each load
type applied via a cross-sensitivity matrix, and measured loads are then combined to find the resultant force system on
Conclusion: The optimum angle for all gauges was found to be about 45 deg with respect to the longitudinal direction.
This provided good sensitivity and separation of force components. Based on this COMSOL analysis, a wireless version
of the SenseWheel will be developed, on a mobile platform. A musculoskeletal model is being developed to infer the
shoulder forces, to be animated by these pushrim forces.
References:  Gutierrez et al. The Relationship of Shoulder Pain Intensity to Quality of Life, Physical Activity, and
Community Participation in Persons With Paraplegia. J Spinal Cord Medicine, vol. 30, no3, p. 251, 2007.
 Tatsuto Suzuki, Catherine Holloway, Stephen JG Taylor, COMSOL Analysis to Determine Optimum Strain Gauge
Locations for SENSEWHEEL, COMSOL Conference 2014 Cambridge
 Taylor, S., Lambert, S., Bayley, I., Blunn, G. (2008). A telemetrised glenoid prosthesis for reverse- anatomy shoulder
replacement. Journal of Biomechanics, 41 S448