针对近红外光谱分析技术中未充分利用预测模型光谱数据的问题,提出了一种可充分利用光谱数据和有效预测蚕丝含量占比的新方法。以5种类型共145个样本的蚕丝含量占比以及相应的所有蛋白质基光谱数据为研究对象,将这些样本分别划分为校正集和验证集,并采用偏最小二乘回归(Partial Least Squares Regression,PLSR)方法和提出的偏最小二乘回归多模型(multi--model Partial Least Squares Regression,multi--PLSR)方法建立了预测模型。然后对比和观察了两种方法的预测效果。以类型2的蚕丝样本为例,选用13个主成分并对比两种模型后发现,multi--PLSR模型的相关系数由0.594增至0.9784,平均相对误差由0.4866降至0.1384。实验结果表明,新方法充分利用了光谱数据中的信息,提高了蚕丝含量占比预测模型的精度,为建立近红外光谱预测模型提供了一种新思路。
In this paper,symplectic schemes and symmetric schemes are presented to simulate Nonlinear Schrodinger Equation(NLSE)in case of dark soliton motion.Firstly,by Ablowitz–Ladik model(A–L model),the NLSE is discretized into a non-canonical Hamiltonian system.Then,different kinds of coordinate transformations can be used to standardize the non-canonical Hamiltonian system.Therefore,the symplectic schemes and symmetric schemes can be employed to simulate the solitons motion and test the preservation of the invariants of the A–L model and the conserved quantities approximations of the original NLSE.The numerical experiments show that symplectic schemes and symmetric schemes have similar simulation effect,and own significant superiority over non-symplectic and non-symmetric schemes in long-term tracking the motion of solitons,preserving the invariants and the approximations of conserved quantities.Moreover,it is obvious that coordinate transformations with more symmetry have a better simulation effect.
