Page 73 - IJAMD-1-1

International Journal of AI
for Material and Design Machine learning for gripper state prediction

Table 1. Isotropic linear elastic material properties of various face labeled “A” in Figure 4. To bend the finger, the string
components used in the soft gripper model is pulled by applying a displacement of x = 25 mm (y = z =
0) to the face of the string on the opposite side of the knot,
Component Materials Storage Poisson’s
modulus (MPa) ratio as indicated by the red-colored face labeled “B” in Figure 4.
Conductive PLA at 40°C 1575 0.23 Quasi-static structural analysis was performed using the
PLA backbone PLA at 45°C 1550 0.23 static structural module of Ansys Workbench, with thermal
PLA at 50°C 1500 0.23 conduction properties not modeled. Instead, the influence
of temperature on the modulus of the cPLA material was
PLA at 55°C 1400 0.23
accounted for by simulating with the appropriate storage
PLA at 60°C 1225 0.23 modulus values corresponding to different temperatures,
PLA at 65°C 888 0.23 as outlined in Table 1.
PLA at 70°C 375 0.23 It is important to note that the main focus of the
Soft TPU finger TPU 15 0.48 current work is to develop a framework for the predictive
String Nylon 2700 0.4 model rather than to accurately replicate the physics of
Abbreviations: PLA: Polylactic acid; TPU: Thermoplastic polyurethane. the problem. Thus, the simulation model was heavily
simplified with various assumptions, which include, but
material using a dynamic mechanical analyzer (Q800, TA not limited to:
Instrument, United States of America). The dimensions of (i) The absence of gravity effects, though preliminary
the coupon for the dynamic mechanical analysis (DMA) studies found that gravity has insignificant effects on
test are 17.50 mm × 15.17 mm × 2.07 mm. Stress relaxation the bending angle.
was employed as the test method. (ii) The assumption of linear elastic material properties
While Young’s modulus and storage modulus are without plastic deformation, as the determination
conceptually distinct, with the former quantifying the of temperature-dependent plastic properties,
stiffness of material under uniaxial stress and the latter would require much more extensive material
representing the elastic component of a viscoelastic characterization.
material under cyclic loading, their responses to (iii) The absence of thermal conduction within the material,
temperature variations exhibit significant similarities. as the determination of temperature-dependent
As evidenced in Figure S1, both moduli exhibit parallel thermal properties, requires extensive material
trends in response to thermal changes. This correlation is characterization. In this case, we assumed no change
grounded in the fundamental thermomechanical behavior in the modulus of the TPU material when the cPLA
of polymers, where increased temperature generally leads joint sections were heated. Figure S2 demonstrates
to decreased stiffness due to enhanced molecular mobility. that the effect of modulus change in the TPU material
29
In this context, the storage modulus effectively captures the due to temperature variation has minimal impact on
temperature-dependent elastic behavior of cPLA, which is joint kinematics.
a critical factor in determining its mechanical properties. Upon proving the viability of the predictive model,
While employing Young’s modulus in the simulation more accurate models can be obtained by using more
would allow for an accurate determination of stress accurate simulation models.
distribution and loads within the gripper, our primary Due to the large deformation in the finger, non-
objective is to analyze how variations in joint stiffness linearity was expected in the analysis. Consequently,
influence joint angles. In this specific context, the similarity the large deflection option was activated, and automatic
in the temperature-dependent trends of both moduli is time stepping was utilized with an initial sub-step time
sufficient for our purpose. In addition, Al-Rubaiai et al. equivalent to 2% of the total step time.
have shown that Young’s modulus and storage modulus of
the Protopasta cPLA material exhibit similar trends and To calculate the bending angles of each joint, we first
are comparable in magnitude. Thus, using the storage obtained the y and z coordinates of six nodes on the
30
modulus of the cPLA material as a surrogate for its Young’s symmetric plane, labeled as points “A,” “B,” “C,” “D,” “E,”
modulus is arguably justified for simulating the changes and “F” in Figure 5. Subsequently, the joint angles could be
in joint bending angles in response to varying actuation easily calculated using trigonometry.
conditions. To investigate joint angle within the operating window
A fixed boundary condition was applied to the face on for this 2-joint finger example, we identified several essential
the opposite side of the knot, illustrated by the green-colored factors influencing joint bending angles: the thickness

Volume 1 Issue 1 (2024) 67 https://doi.org/10.36922/ijamd.2328