next up previous
Next: 4 A Simulation Algorithm Up: 3 Chattering in Physical Previous: 3.2 The Cam-Follower

3.3 Model Semantics

When a system operates in the sliding regime, the dynamics of the system along the sliding surface is not defined in a continuous sense. We adopt the notion of equivalence dynamics for the sliding regime justified by Filippov [3, 9]. Consider the switching surface as an infinitesimal band rather than a crisp border. The equivalence dynamics on the surface is defined as the limiting behaviors when the width of the band tends to zero. This construction preserves the physical meaning of the dynamics at the discontinuous boundaries. Furthermore, it serves as a basis for algorithmically determining the direction and magnitude of the sliding motion.

Assume tex2html_wrap_inline1709 is the thickness of the hysteresis band around the sliding surface (see Fig. 17). If tex2html_wrap_inline1709 is small, the fields tex2html_wrap_inline1773 and tex2html_wrap_inline1775 on either side of the surface can be represented by their instantaneous vector representations with normal components tex2html_wrap_inline1583 and tex2html_wrap_inline1585 , and tangential components tex2html_wrap_inline1781 and tex2html_wrap_inline1783 . The direction of movement is along the sliding surface, and we need to calculate the average velocity on the surface.

We compute the time the system takes to cross the tex2html_wrap_inline1709 band as tex2html_wrap_inline1787 and , and the tangential distance the system has travelled over two adjacent time intervals ( tex2html_wrap_inline1791 ) as:

(5) displaymath1840

Then we compute the average velocity of the motion on the surface:

(6) displaymath1842

where tex2html_wrap_inline1793 . Thus, the vector v is on the line connecting the end points of tex2html_wrap_inline1773 and tex2html_wrap_inline1775 , with r and 1-r as its barycentric coordinates (Fig. 18). Let c be the difference vector tex2html_wrap_inline1807 , and p be the intersection of c with the tangent vector. Let p partition c into two segments, d and e. We have tex2html_wrap_inline1821 by triangle congruence. Thus, we have shown that the barycentric coordinate for p is the same as that for v (recall tex2html_wrap_inline1827 ). This corresponds to Filippov's construction, i.e., the vector v is the same as the tangent vector.

   figure1063
Figure 17: Motion along a sliding surface.

   figure1068
Figure 18: Filippov construction of sliding motion direction and magnitude.

In fact, the formula tex2html_wrap_inline1831 can be used to compute v, where r is determined from the normal components of the two vector fields if the angle between the sliding surface and vector fields is known.


next up previous
Next: 4 A Simulation Algorithm Up: 3 Chattering in Physical Previous: 3.2 The Cam-Follower

Pieter J. Mosterman
Mon Aug 18 13:23:03 CDT 1997