1. Cross strait Quad-reginal radio science and wireless technology conference, Vol. 2, p.p. 993-995,2011 Application of fuzzy LS-SVM in dynamic compensation on accelerometer sensors

2. Outline Abstract Introduction Theory of the dynamic compensation for sensors The principle of the fuzzy Ls-svm Dynamic compensation simulation of accerometer sensors Conclusions References

3. Abstract A design method of the dynamic compensation for micro-silicon accelerometer sensors was presented, which was based on the fuzzy least squares support vector machine(fuzzy LS-SVM). According to the measurement data of the step response of the sensor and the reference model, the compensator was constructed by fuzzy LS-SVM regression algorithm which reduced the dynamic errors because of the limited bandwidth of micro-silicon accelerometer sensors. The method can overcome the local minima problem in dynamic compensation of neural networks and reduce the effect on model precision because of some extra essential data in standard LS-SVM which has a potential future in the field of measurement.

4. Introduction Micro-silicon accelerometer sensors are widely used at present during the real application, because the quality of the movement system is unlikely too tiny. Furthermore the increasing rigidity of spring lamination will cause sensibility of accelerometer sensors lower down, therefore upper limit of its measuring frequency is considerably restricted because that free frequency of the system is unlikely too high. The lagging of detection signal will occur and measuring accuracy will be considerably affected in the process of fast and dynamic measuring, especially under the requirements of high-accuracy. Therefore it is necessary to compensate for the dynamic characteristics of the sensors so as to meet with the requirements of fast and accurate measuring.

5. Theory of the dynamic compensation for sensors

6. The principle of the fuzzy Ls-svm A. The basic thought In the function index expression of LS-SVM, all errors ei are weighted in weighed value I, which shows that in the standard LS-SVM, all data points play the same role in training. However in many real application not all obtained data is meaningful, and the data point is not the same important to the establishment of object models.

7. The principle of the fuzzy Ls-svm

10. Dynamic compensation simulation of accerometer sensors

11. Conclusions There is no need to know the dynamic characteristics of micro-silicon accelerometer sensors for the designed methods based on fuzzy LS-SVM sensors and nonlinear dynamic error compensator. The compensator is constructed by the actual measured data of input incentive response according to the sensor and reference model. It's dynamic property and resonant frequency of tiny accelerated speed sensors can be improved and use of frequency is extended by the above compensation through simulation experiments, which has important practical value in the engineer testing field and provided a new method for other kinds of sensors' nonlinear dynamic compensation

12. References