Several workers are exposed to mechanical vibration and the harmful effect of the exposure is well documented in literature. The induced vibrations to the extremities of the limbs may cause neurovascular problems, namely, Raynaud syndrome, even resulting in disabilities. These problems have been linked to higher frequencies than the ones affecting the human body as a whole. Therefore, the first part of this dissertation describes the mechanical design and development of the actuators system able to generate the triaxial vibration needed to characterize the response of the standing human subject from 1 to 80 Hz. The machine is based on the architecture of the linear delta robot with vertical actuators. Once the machine was optimized for the defining dimensions, the appropriate actuators were selected, and custom components were drawn to be produced. The tests done on the machine after its realization confirm that it can generate pseudo-random noise up to 80 Hz across the three mutually perpendicular spatial axes simultaneously. The average root-mean squared error between the modulus of the spectrum of generated signal and the signal measured directly on the platform is 6,9 mm/s² across the three axes, and it is 5,5 mm/s² when a person is standing on the platform, i.e. lower than 1% of the imposed acceleration. A generalized mathematic method to directly link the design parameters of a linear delta robot with its kinematic performances was developed.

The second part of this dissertation describes the development of a planar lumped parameters mechanical model of the foot-ankle system. Given that the data regarding the response of the human foot to multiaxial vibration are not available yet, the model was validated with the objective of reproducing the transmissibility of vertical vibration from the ground to five different positions on the foot. The model also reproduced the apparent mass of a standing person subjected to vertical whole-body vibrations. Results evidenced that the model reproduces the transmissibility with errors that are small in comparison with the inter- and intra-subject variability.