Structural Health Monitoring Using Smart Sensors - ideals ...

1
1
Normalized accumulated stress
measurement 3 measurement 2 measurement 1
0.5
0.3
0
1
0.5
0.3
0
1
0.5
0.3
Normalized accumulated stress
measurement 3 measurement 2 measurement 1
0.5
0.3
0
1
0.5
0.3
0
1
0.5
0.3
0
3 4 5 6 7 8 9 1011 7 8 9 10 111213 1415 1112 1314 151617 1819
Element ID
(a)
3 4 5 6 7 8 9 1011 7 8 9 101112131415 111213141516171819
Figure 8.13. Damage detection (damaged element: Element 8): (a) mass perturbation
DLV; and (b) SDLV.
0
Element ID
(b)
3, 4, and 5 monitor for the damaged element for the other cases. The experiment is
repeated at least three times for each case to assess the repeatability of damage detection.
The SDLV method is employed as the DLV method on Imote2s, while the damage
detection capability of the mass perturbation DLV method is also examined on a PC. As
stated in section 8.4, the DCS for SHM using known mass perturbation has been
implemented on Imote2s, and the same numerical operations on Matlab on the centrally
collected truss acceleration response data has been confirmed to produce the same
calculation results as those on Imote2s. Therefore, for convenience, the mass perturbation
DLV method is applied on Matlab to the centrally collected data to investigate the damage
detection capability.
Figures 8.13 to 8.21 show the damage detection results using the two DLV methods.
For most of the experiments, both of the DLV methods can detect the damaged element,
i.e., it has a normalized accumulated stress smaller than 0.3. In some cases, however,
false-positive and/or false-negative damage detection is observed.
False-positive damage detection can be theoretically explained. The DLV methods
identify elements with small stress as candidates for damaged elements. Undamaged
elements are not necessarily excluded from the small stress elements. Bernal (2002) and
Gao (2005) reported that such false-positives are occasional in their simulations and
experiments. In the Imote2 experiments, Figure 8.15 (a) shows false-positive damage
detection at element 11 in the second measurement. This false-positive is not observed in
the first and third measurements. In Figure 8.14 (a), however, false-positives are observed
consecutively. Element 8 has small stress in all cases. Though consecutive false-positives
do not contradict the theoretical derivation of the DLV methods, they are not reported in
Bernal (2002) or Gao (2005). Consecutive false-positives are likely to indicate that factors
such as a structure’s nonlinearity and observation noise affect the damage detection
capability in these experiments.
150