Tuned vibration absorbers (TVAs) are prevalent in many vibration control applications such as in aircraft fuselages due to their low cost and well-established vibration absorption capabilities. TVAs are extremely effective in cases where there is a single, known frequency excitation to be controlled. Its limitations lie in that it is only effective in a very narrow bandwidth. While active vibration controllers have been developed to control vibrations over a large bandwidth, their use as vibration control mechanisms have been limited for several reasons. Active vibration controllers, while they can be highly effective, possess costly and highly sophisticated control algorithms. In addition, due to their real-time property changing scheme, an active vibration absorber subjected to an unanticipated excitation, or one that is improperly controlled, can actually add energy into the system and thus drive it into instability.
The semiactive absorber, or in our case a state-switching absorber (SSA), is a hybrid of the reliable TVA and the more effective active vibration controller. The SSA is capable of switching one or more of its properties, in our case its spring stiffness, to fundamentally increase its operational bandwidth. It is much more stable and simple than the active vibration absorber, though, because the control algorithm only allows switching to occur at discrete times and to discrete states. In this way, the risk of adding energy to the system is eliminated since between states the SSA behaves as a classical, stable TVA.
Just because the SSA can control multiple vibration frequencies does not necessarily make it a TVA competitor. For the SSA to absorb vibration better than a TVA, the excitation source must be a variable-frequency one. In addition, the SSA requires external energy to enable switching, both to the switching mechanism and to the sensors that feed into the control algorithm; this additional energy must be minimal. The SSA must be comparable in size and mass to the TVA; a large or bulky system will not do. It must also be able to operate within the same frequency range. Perhaps one of the most limiting comparisons is also cost - the SSA, including the sensors, control mechanisms, and construction, must be similar to that of a TVA.
The research here has an ultimate goal of producing a semi-active vibration absorber and control algorithm to effectively suppress vibration that is as compact and light as a conventional tuned vibration absorber that is operational in the sub-100 Hz range, and at least a 50% increase in the natural frequency. The compact and light requirements translate into specifications of The control algorithm to be implemented will be able to control vibration from any noise source, beginning with white noise, better than the best TVA alternative. The focus here is on developing a magnetorheological (MR) spring-mass-damper device operational under 100 Hz with at least a 50% natural frequency change.
Iron particles are added to a two-part silicone prior to curing. This composite liquid is then placed in a specially-designed mold to cure. The purpose of this mold is to generate a strong magnetic flux path that goes through the elastomer during the cure process. This causes the iron particles within the composite to align along the lines of flux and then be solidified into these chains. This enhances the stiffness change effect once cured.
The cured silicone composite under the influence of no magnetic flux density is soft, whereas variable amounts of magnetic flux can increase its stiffness. The SSA must be designed such that a magnetic flux path can be generated to run through the elastomer. For this reason, two half-cylinders of low-carbon steel were constructed as base and absorber masses. Roughly 300 turns of gage-25 magnet wire are wrapped around the absorber mass. With this setup, direct current through the magnet wire can generate a magnetic flux path through each silicone half.
Of critical importance to the development of the SSA is the volume percentage of iron in the elastomer. The test setup to determine this can be seen below. To test a composite silicone, a sample is first cut in half lengthwise, and then epoxied to two iron half-circles. Accelerometers are attached to each half-circle, and the mass-spring system is mounted to a shaker. The shaker excites the system with band-limited white noise, measured by the top accelerometer. The response of the SSA is measured by the bottom accelerometer.
The larger experimental setup can be seen below. The excitation signal is chosen on the computer and sent to Siglab (flat black box), which is then run through an amplifier and finally out to the shaker. The signals from the accelerometers are sent through the accelerometer power supply (light blue box) and then into Siglab, which is then recorded by the computer. The DC power supply (large light grey box in center) connects the magnet wire in parallel with a small circuit suppressing transients from the power supply.
Once time data has been collected, fast Fourier transforms are calculated, comparing the SSA response to that of the excitation. By looking at the 90 degree phase shift point and the width of the FFT peak, natural frequencies and damping coefficients were found by best-fit. In this way, an elastomer of 30-35% iron by volume was found to yield the highest natural frequency change.
There are several current control algorithms in consideration
at present. The goal is for the SSA algorithm to always minimize
the energy in the base mass better than the best-tuned TVA. Prior
work has established a maximum work-extraction criterion (Cunefare,
Sadegh). This algorithm states that at any candidate switch point,
the choice between two possible states is determined by which
spring would either take the most energy out of the base, or else
put the least amount of energy into the system. The candidate
switch points are limited to those where a state change would
not shock the system. For our case with a switchable spring, this
occurs when k1x^2=k2x^2, where x is the deformation and k1 and
k2 are the respective spring stiffnesses. Another strategy, developed
by Walsh and Lamancusa, dictates that at a candidate switch point,
one should tune the SSA to the frequency of the excitation.
Although both of these strategies are highly effective in certain situations, applying these algorithms to broadband excitation has been only moderately successful. Future work will involve combining these strategies with a predictive Kalman filter in attempts to "guess" a future signal