提出了一种利用偏最小二乘回归系数矩阵筛选光谱波段的算法。该算法利用偏最小二乘回归系数作为筛选光谱波长的依据,参考(root-mean-squares error of cross-validation,RMSECV)曲线,使初选波长数大大降低。在此基础上通过循环选择将无效信息光谱波长剔除,同时增强了所建模型的预测精确性。通过生产过程的Raman光谱数据验证,该算法比传统的利用回归系数筛选波长的算法更好地提高了模型的精确性,同时降低了模型的复杂程度,是一种高效实用的算法。
description of magnetization curve has important effect on ferroresonance. In most of earlier ferroresonance studies the magnetization curve is modelled as a 3rd or 5th order polynomial. However, it is not comprehensive. This paper investigates the chaotic ferroresonance behaviour exhibited by a non-autonomous circuit which contains a nonlinear flux-controlled inductance. The ferromagnetic characteristic of this nonlinear inductance represented by a magnetization curve could be expressed as an nth order two-term polynomial. By varying the value of exponent n, the circuit can assume a diverse range of steady-state regimes including fundamental and subharmonic ferroresonance, quasi-periodic oscillations, and chaos. A detailed analysis of some simulations demonstrates that the probability of chaos increases as the exponent of the magnetization curve rises. The effect of varying the magnitude of the source voltage on the chaotic behaviour of the circuit is also studied.
