摘要
旋转行波超声电机的转速预测器,基于在线最小二乘支持向量机建立,并采用输入增维,化多值映射为多对一映射。其输出位置信息使用增量式圆光栅编码器进行测量,转速则由编码器位置信息求差分获得。输入数据每个点为相关的维向量,输出数据每个点为标量,同时在线更新的数据集。最后令LSSVM在线完成一步预测,由得到的原始数据和预测结果,可获得误差曲线和预测结果的迟滞环。实验验证该预测器有效。
Based on the least squares support vector machine （LSSVM）, a predictor is established to forecast the behavior of traveling-wave ultrasonic motor （TWUSM）, in which the speed is controlled by transforming voltages of two-phase driving sine wave. And to transform the multi-valued mapping of hysteresis into one-to-one mapping, the dimension of input space of the predictor is increased. The angle position of rotor is measured by photoelectric encoder, and the rotate speed can be acquired through computing the difference of angle position. With the multi-dimension input space and one dimension output space, the data in the data window are updated on-line. Finally, the results of one-step-ahead prediction based on the on-line LSSVM are obtained, and the validity of this method is confirmed.
出处
《兵工自动化》
2007年第9期64-65,70,共3页
Ordnance Industry Automation
基金
国家自然科学基金(60572055)
广西科学基金(0339068)
关键词
旋转行波超声电机
转速预测器
最小二乘支持向量机
迟滞
Traveling-wave ultrasonic motor
Rotate speed predictor
Least squares support vector machines
Hysteresis
