Accelerometer calibration based on improved particle swarm optimization algorithm of support vector machine

Aiming at the problem that the deterministic errors caused by non-orthogonal installation, calibration factor, zero bias and other factors in production and in the use of accelerometers need to be calibrated by high-precision instruments, support vector machine regression is used to process the original data output by the accelerometer, and the processed data of each axis are used to establish a parameter calibration model without reference datum through the relationship between the output value of each axis of accelerometer, gravity acceleration and coaxial reversal in the paper. Then, the adaptive mutation rate is used to dynamically adjust the number of reverse learning particles, and the particles of particle swarm optimization algorithm are selected and adjusted according to the reverse learning, which solves the problems that particle swarm optimization algorithm tends to fall into localoptimum and the convergence speed is slow, through which a fast, accurate and simple calibration can be realized, and the performance of particle swarm optimization algorithm is improved. The calibration experiment shows that the improved particle swarm optimization algorithm has higher accuracy and faster convergence speed than the particle swarm optimization algorithm, and the calibration parameter accuracy is higher than that of the least square method, which does not need the datum of each axis. The calibration model proposed in this paper can realize a benchmark-free calibration outside the laboratory. At the same time, the improved particle swarm optimization algorithm can obtain calibration parameters with higher accuracy and faster speed in the rapid calibration, which provides the idea of a new model for accelerometer calibration and expands the application environment of accelerometer.