CAJ | 학술논문

利用最优线性联想记忆(OLAM)神经网络方法对水泥生料样品X荧光谱进行了分析,学习谱的选择采用了两种方式:单质谱学习和混合谱学习.对这两种方式及输出结果作了对比,单质谱学习可以构建模拟谱,用于对谱漂移和漏学习的判别;混合谱学习具有信息容量大,考虑了基体效应的优点,但是学习谱之间的多重相关性可能对结果造成一定影响.
이용최우선성련상기억(OLAM)신경망락방법대수니생료양품X형광보진행료분석,학습보적선택채용료량충방식:단질보학습화혼합보학습.대저량충방식급수출결과작료대비,단질보학습가이구건모의보,용우대보표이화루학습적판별;혼합보학습구유신식용량대,고필료기체효응적우점,단시학습보지간적다중상관성가능대결과조성일정영향.
It uses OLAM artificial neural network to analyze the samples of cement raw material.Two kinds of spectrums afe used for network learning:pure-element spectrum and mix-element spectrum.The output of pure-element method can be used to construct a simulate spectrum，which can be compared with the original spectrum and judge the shift of spectrum;the mix-element method can store more message and correct the matrix effect,but the multicollinearity among spectrums can cause some side effect to the results.