The approach used in this paper looks to couple the powerful multi-objective algorithm NSGAII with an electrical circuit model representation, coined (GAECM). However it was not a true multi-objective algorithm, nor did the actual values come from the previously designed filter. The approach proved successful for the set of targets outlined, in this instance to match certain values for both mass, stiffness and damping of a single resonator device. Even so an approach was outlined to allow the automatic synthesis of a physical device in this case utilising an analytical model of a folded flexure resonator and linking it with the powerful approach of GA’s. Though successful a large number of functional evaluations were required (2.6 million) and no respective circuit values were given and therefore it is not possible to derive whether the actual designs were physically feasible. Past attempts towards MEMS filter design optimisation have looked to couple the powerful approach of genetic programming with a bond graph representation, coined (GPBG). This can be partitioned into two sets of goals, firstly the design of a suitable circuit level topology consisting of RCL tanks that matches our target response, and secondly using the values from this circuit model to derive suitable targets for the design of a physical device layout, in this case of a folded flexure resonator. The design and optimisation of a MEMS bandpass filter forms the basis of our multi level approach. Similar approaches to this have been undertaken, however these have focused on each level separately with little regard to linking them through the use of equivalent circuit equations such as those above. This allows a direct link between the system and device levels and as a result allows designers to derive both function and fabrication to one particular instance of the MEMS filter design. Using these equations it is possible to derive resistor, capacitor and inductance values from the damping, stiffness and mass values of the resonator and equivocally vice versa. pn is the dc bias voltage, is a constant that models additional capacitance due to fringe field electrics, ε o is the permittivity of air, h is the structural layer thickness, N fin is the number of comb drive fingers and d is the comb finger gap spacing.
0 kommentar(er)
