A Decoupling Method of Multiaxis Accelerometers Based on Segmentation Surface Fitting

The coupling of a multiaxis accelerometer mainly depends on two input variables: acceleration and its frequency, which is different from a multiaxis force sensor with a single input variable and changes the coupling function from 2-D to 3-D. As a result, new challenges arise in the decoupling process. In view of this situation, a decoupling method of multiaxis accelerometers suitable for unknown input variables is proposed in this article. The method uses calibration data as input variables to deduce a coupling model. In the process of decoupling, based on a segmentation surface fitting method, two solving methods of the coupling model are proposed: the direct surface fitting and the point-line surface distribution fitting. The two fitting methods are used to solve the coupling function of a triaxial MEMS accelerometer. The experimental results show that the two fitting methods can reduce coupling error without increasing the highest order of the fitting function. In cases of strong coupling, the decoupling method proposed can reduce the coupling error to 1.47% of that before decoupling, the coupling interference was reduced significantly. The coupling error can be reduced to less than 0.005 g, and the decoupling effect is stable. Compared with the decoupling method based on backpropagation (BP) neural network, the decoupling effect is also improved. In addition, the single-decoupling time of the decoupling method is less than 0.561 ms.