Support vector machines regression on - line modelling and its application 支持向量機(jī)回歸在線建模及應(yīng)用
Because of the periodicity of the hub oscillatory load and the nonlinearity of the i / o relationship of the smart rotor , the neural control methods with both off - line and on - line modeling in frequency domain on the smart rotor and their realization procedures are presented 飛行中直升機(jī)的槳轂交變載荷以周期性的為主,而智能旋翼的輸入輸出關(guān)系具有非線性特征。據(jù)此,本文提出離線建模的和在線建模的智能旋翼頻域神經(jīng)控制方法,并給出實(shí)現(xiàn)的具體方案細(xì)節(jié)。
The procedure for finding the optimal identification is simplified , and both the premise and consequent parameters are identified simultaneously by using the sofia . because of reduction of computational requirement for identifying a takagi - sugeno ( t - s ) fuzzy model by efficient parameter and structure identification , this algorithm can be used in on - line modeling 改進(jìn)后的算法簡(jiǎn)化了前提結(jié)構(gòu)辨識(shí)的過程,并使前提參數(shù)辨識(shí)和結(jié)論參數(shù)辨識(shí)同時(shí)完成,極大的減少了參數(shù)辨識(shí)和結(jié)構(gòu)辨識(shí)的計(jì)算量,能夠保證在線辨識(shí)的要求。
In accordance with the problem that the fcm algorithm is quite time - consuming for search out cluster cancroids and may not be suitable for on - line modeling and control . this dissertation proposed an improved fuzzy identification method based multistage random sampling fuzzy c - means clustering algorithm ( mrfcm ) . it has higher approximate precision and the cpu time has slowed down sharply compared with the common fuzzy Johnyen和liangwang介紹了幾種應(yīng)用于模糊模型的信息優(yōu)化準(zhǔn)則,本論文在此基礎(chǔ)上對(duì)統(tǒng)計(jì)信息準(zhǔn)則進(jìn)行一些改進(jìn),并與快速模糊聚類和正交最小二乘方法結(jié)合,提高了模型的辨識(shí)精度和泛化能力。