We propose a technique for classifying paints with time-dependent properties using a new method of merging principal-component analyses (the "PCA-merge" method) that utilizes shifting of the barycenter of the PCA score plot. To understand the molecular structure, elemental concentrations, and the concentrations in the evolved gaseous component of various paints, we performed comprehensive characterizations using Fourier transform infrared spectroscopy, inductively coupled plasma mass spectrometry, and head-space-gas chromatograph/mass spectrometry while drying the paint films for 1-48 h. As various detected intensity- and time-axis variables have different dimensions that cannot be handled equally, we normalized those data as an angle parameter (θ) using arctangent to reduce the influence of high/low intensity data and the various analytical instrument. We could classify the paints into suitable categories by applying multivariate analysis to this arctangent-normalized data set. In addition, we developed a new PCA-merge method to analyze data groups that include different time components. This method merges the PCA data groups of each time-component axis into that of specific-component axes and distinguishes each sample by utilizing the shift in the barycenter of the PCA score plot. The proposed method enables the simultaneous utilization of various data groups that contain information about static and dynamic properties. This provides further insight into the characteristics of the paint materials via shifts in the barycenter of the PCA scores without requiring numerous peak identifications.

中文翻译：

---

反正切归一化和主成分分析合并方法，利用随时间变化的材料数据对特性进行分类。

我们提出了一种对具有时间相关属性的涂料进行分类的技术，该技术使用一种新的合并主成分分析方法（“PCA 合并”方法），该方法利用 PCA 得分图重心的移动。为了了解各种涂料的分子结构、元素浓度和逸出气态成分的浓度，我们在干燥涂料的同时，使用傅里叶变换红外光谱、电感耦合等离子体质谱和顶空气相色谱/质谱进行了全面的表征。漆膜1-48小时。由于各种检测到的强度轴和时间轴变量具有不同的维度，无法同等处理，我们使用反正切将这些数据归一化为角度参数（θ），以减少高/低强度数据和各种分析仪器的影响。通过对反正切归一化数据集应用多元分析，我们可以将油漆分类为合适的类别。此外，我们开发了一种新的 PCA 合并方法来分析包含不同时间成分的数据组。该方法将每个时间分量轴的PCA数据组合并到特定分量轴的PCA数据组中，并利用PCA得分图的重心偏移来区分每个样本。所提出的方法使得能够同时利用包含静态和动态属性信息的各种数据组。这可以通过 PCA 分数重心的变化进一步了解涂料材料的特性，而无需进行大量峰值识别。