论文部分内容阅读
提出了新的图谱预处理方法,消除近红外光谱中的物理信号,应用于样品的化学组分、结构分析、近红外建模等。将光谱信号分解为物理吸收和化学吸收两大类,提出理论假设,样品光谱中存在一段平滑的物理特征曲线,并可以进行消除。试验中选择注射用头孢哌酮钠粉针剂进行分析验证,共选择4个厂家5个批次的样品,每批次样品扫描10张光谱,通过计算标准偏差,和常规的基线校正预处理方法进行比较,然后通过加速试验,对比分析近红外光谱与有关物质含量的相关性,验证预处理方法的合理性。扣减样品光谱中物理特征曲线,得到重复性更好的图谱,标准偏差明显小于基线校正的结果,有关物质含量与近红外光谱正相关。
A new method of spectrum pretreatment was proposed to eliminate the physical signal in near infrared spectroscopy, which was applied to the chemical composition, structure analysis and near infrared modeling of samples. The spectral signal is decomposed into two categories: physical absorption and chemical absorption, and a theoretical assumption is proposed that there is a smooth physical characteristic curve in the sample spectrum and can be eliminated. In the experiment, cefoperazone sodium for injection was selected for analysis and verification. A total of 5 batches of samples from 4 manufacturers were selected and 10 spectra were scanned for each batch of samples. The standard deviation was calculated and compared with the conventional baseline correction pretreatment method. Through accelerated test, comparative analysis of the correlation between near-infrared spectroscopy and related substances, verify the rationality of the pretreatment method. By subtracting the physical characteristic curve of the sample spectrum, a better repeatability spectrum is obtained, the standard deviation is obviously smaller than that of the baseline correction, and the related substance content is positively correlated with the near infrared spectrum.