阻燃包装纸研发技术优化研究

目的 为探索机器学习算法利用检验大数据快速鉴别复合包装膜袋材质的可行性。方法 以不同复合层数、不同功能层材质、不同食品接触层材质的10种复合包装膜袋共计1333个样本作为数据集，将韧性向拉伸强度、刚性向拉伸强度、韧性向断裂标称应变、刚性向断裂标称应变、水蒸气透过率、氧气透过率、厚度等7个维度的性能测试数据作为特征值，利用人工智能机器学习算法进行复合包装膜袋材质鉴别。结果 综合比较决策树、逻辑回归、支持向量机、K近邻、神经网络、高斯朴素贝叶斯等6种学习算法后，发现决策树算法的准确率和kappa系数最高，运行速度也很快。经参数优化后，决策树算法的鉴别结果准确率为95.4%，kappa系数为93.2%。结论 决策树算法在复合包装膜袋材质鉴别中具有一定优势。

Optimization of Research and Development Technology of Flame Retardant Packaging Paper

The work aims to explore the feasibility of machine learning algorithms in the quick identification of composite packaging film materials with inspection big data. 1333 samples of ten composite packaging films with different numbers of composite layers, different functional layer materials, and different food contact layer materials were used as data sets and the inspection values of tensile strength in toughness direction, tensile strength in rigid direction, elongation at break in toughness direction, elongation at break in rigid direction, water vapor transmission rate, oxygen transmission rate, and thickness were used as characteristic values. Then, the machine learning algorithms of artificial intelligence were used to identify the materials of composite films. After comprehensively comparing the six learning algorithms of Decision Tree, Logistic Regression, SVM, K Neighbors, MLP, Gaussian Naive Bayesian, the Decision Tree algorithm was found to have high accuracy, kappa coefficient and running speed. After parameter optimization, the accuracy and kappa coefficient of Decision Tree algorithm for material identification were 95.4% and 93.2%, respectively. Therefore, the Decision Tree algorithm has a certain advantage in the identification of composite packaging film materials.