Fractionation of palm kernel oil （PKO） by short path distillation （SPD） at two feed flow rates （135 g/h and 195 g/h） and six distillation temperatures, TD,s （200, 210, 220, 230, 240 and 250 ℃） was investigate...Fractionation of palm kernel oil （PKO） by short path distillation （SPD） at two feed flow rates （135 g/h and 195 g/h） and six distillation temperatures, TD,s （200, 210, 220, 230, 240 and 250 ℃） was investigated. Other distillation parameters, such as vacuum pressure （0.001 mbar）, blade rotation speed （400 rpm） and temperature of the feed material （60 ℃） were kept constant. The fractionated products, known as residue and distillate, were analysed for physico-chemical properties including fatty acid composition （FAC）, triacylglycerol （TAG） composition, slip melting point （SMP）, thermal analysis by differential scanning calorimetry （DSC） and solid fat content （SFC）. Product yield was measured as well. Crystallisation behaviour of PKO and the fractionated products were studied by measurement of isothermal crystallisation, Tc,. at 0, 5, 10, 15, 20 and 25 ℃. The distillates, collected at all fractionation temperatures, were enriched with caprylic, capric and lauric acids. These fractions were also concentrated with low molecular weight and C36 TAGs. Distillates obtained at higher TDis （230-250 ℃） exhibited higher in SMP and SFC. On the other hand, the residual oils collected at all fractionation temperatures contained higher amount of long-chain fatty acid and palmitic acid. These fractions were enriched with high molecular weight TAGs. Residues obtained at lower Tois （200-220 ℃） were low in SMP and comparable SFC with PKO. Changes in fatty acid and TAG composition resulted in different crystallisation behaviour of the fractions. Distillates collected at all fractionation temperatures crystallised in a sharper peak while residues obtained at higher T Dis （230-250 ℃） showed broader crystallisation peaks, as shown by the DSC thermograms.展开更多
针对油气领域知识图谱构建过程中命名实体识别使用传统方法存在实体特征信息提取不准确、识别效率低的问题,提出了一种基于BERT-BiLSTM-CRF模型的命名实体识别研究方法。该方法首先利用BERT(bidirectional encoder representations from...针对油气领域知识图谱构建过程中命名实体识别使用传统方法存在实体特征信息提取不准确、识别效率低的问题,提出了一种基于BERT-BiLSTM-CRF模型的命名实体识别研究方法。该方法首先利用BERT(bidirectional encoder representations from transformers)预训练模型得到输入序列语义的词向量;然后将训练后的词向量输入双向长短期记忆网络(bi-directional long short-term memory,BiLSTM)模型进一步获取上下文特征;最后根据条件随机场(conditional random fields,CRF)的标注规则和序列解码能力输出最大概率序列标注结果,构建油气领域命名实体识别模型框架。将BERT-BiLSTM-CRF模型与其他2种命名实体识别模型(BiLSTM-CRF、BiLSTM-Attention-CRF)在包括3万多条文本语料数据、4类实体的自建数据集上进行了对比实验。实验结果表明,BERT-BiLSTM-CRF模型的准确率(P)、召回率(R)和F_(1)值分别达到91.3%、94.5%和92.9%,实体识别效果优于其他2种模型。展开更多
文摘Fractionation of palm kernel oil （PKO） by short path distillation （SPD） at two feed flow rates （135 g/h and 195 g/h） and six distillation temperatures, TD,s （200, 210, 220, 230, 240 and 250 ℃） was investigated. Other distillation parameters, such as vacuum pressure （0.001 mbar）, blade rotation speed （400 rpm） and temperature of the feed material （60 ℃） were kept constant. The fractionated products, known as residue and distillate, were analysed for physico-chemical properties including fatty acid composition （FAC）, triacylglycerol （TAG） composition, slip melting point （SMP）, thermal analysis by differential scanning calorimetry （DSC） and solid fat content （SFC）. Product yield was measured as well. Crystallisation behaviour of PKO and the fractionated products were studied by measurement of isothermal crystallisation, Tc,. at 0, 5, 10, 15, 20 and 25 ℃. The distillates, collected at all fractionation temperatures, were enriched with caprylic, capric and lauric acids. These fractions were also concentrated with low molecular weight and C36 TAGs. Distillates obtained at higher TDis （230-250 ℃） exhibited higher in SMP and SFC. On the other hand, the residual oils collected at all fractionation temperatures contained higher amount of long-chain fatty acid and palmitic acid. These fractions were enriched with high molecular weight TAGs. Residues obtained at lower Tois （200-220 ℃） were low in SMP and comparable SFC with PKO. Changes in fatty acid and TAG composition resulted in different crystallisation behaviour of the fractions. Distillates collected at all fractionation temperatures crystallised in a sharper peak while residues obtained at higher T Dis （230-250 ℃） showed broader crystallisation peaks, as shown by the DSC thermograms.
文摘针对油气领域知识图谱构建过程中命名实体识别使用传统方法存在实体特征信息提取不准确、识别效率低的问题,提出了一种基于BERT-BiLSTM-CRF模型的命名实体识别研究方法。该方法首先利用BERT(bidirectional encoder representations from transformers)预训练模型得到输入序列语义的词向量;然后将训练后的词向量输入双向长短期记忆网络(bi-directional long short-term memory,BiLSTM)模型进一步获取上下文特征;最后根据条件随机场(conditional random fields,CRF)的标注规则和序列解码能力输出最大概率序列标注结果,构建油气领域命名实体识别模型框架。将BERT-BiLSTM-CRF模型与其他2种命名实体识别模型(BiLSTM-CRF、BiLSTM-Attention-CRF)在包括3万多条文本语料数据、4类实体的自建数据集上进行了对比实验。实验结果表明,BERT-BiLSTM-CRF模型的准确率(P)、召回率(R)和F_(1)值分别达到91.3%、94.5%和92.9%,实体识别效果优于其他2种模型。
