基于注意力机制的时空卷积数控机床热误差模型研究

为了提高数控机床热误差模型的精度与泛化性,提出了基于注意力机制的长短时记忆卷积神经网络(Long short term memory convolutional neural network based on attention mechanism, AM-CNN-LSTM)热误差模型。利用卷积神经网络提取高维数据空间状态特征的能力和长短时记忆网络提取长时间序列状态特征的能力,构建具有2个支路的热误差模型,分别提取特征后输入到注意力机制中进行特征重要性重构,建立原始数据与热误差的特征映射,最后通过全连接层进行热误差预测。采用G460L型数控机床进行实验数据采集,将不同季节采集到的温度数据和热误差作为模型输入,采用循环学习率与正则化优化方法对模型进行训练。与LSTM、ConvLSTM和CNN-LSTM热误差模型对比,结果表明,AM-CNN-LSTM模型对特征还原能力最强,残差波动范围最小,其残差范围较最大值下降62.09%,模型预测精度在2.4μm以内。     

Effect of short‐term supplementation with rumen‐protected fat during the late luteal phase on reproduction and metabolism of ewes

This study was designed to study the effect of short‐term supplementation with rumen‐protected fat during the late luteal phase on reproduction and metabolism of sheep during breeding season. Seventy‐six ewes (Rahmani, Barki and Awassi × Barki) were allocated to two groups considering genotype: the control ewes (C‐group) received a maintenance diet, and the fat‐supplemented ewes (F‐group) received the maintenance diet plus 50 g/head/day of rumen‐ protected fat (Megalac) for 9 days during which oestrus was synchronized. The latter had been accomplished using double intramuscular injection of prostaglandin F_2α (PGF_2α) 11 days apart. Ovarian activity, serum concentration of cholesterol, glucose, insulin and reproductive performance variables were recorded. Data were analysed considering treatment (group) and genotype. Supplementation had positive effects on the overall mean serum concentrations of cholesterol (p < 0.05), glucose (p < 0.05) on day 6 of nutritional treatment and insulin (p = 0.07) on day 8. Fat supplementation did not affect the total number of follicles, follicle populations and ovulation rate. However, fat‐supplemented Rahmani ewes tended to have higher ovulation rate compared with other breeds (treatment × breed interaction, p = 0.06). Treatment also did not affect the mean concentration of serum estradiol or progesterone. Supplemented ewes had higher conception (p = 0.06) and lambing rates (p < 0.05) compared with control. In conclusion, short‐term supplementation with rumen‐protected fat as a source of energy around breeding time improved metabolism, conception and lambing rates of ewes without effects on steroidogenic capacity and ovarian activity being apparent.     

Carbon Sequestration in Soil Humic Substances Under Long-Term Fertilization in a Wheat-Maize System from North China

The changes in humic substances （HS） is fundamental in detecting soil carbon sequestration mechanisms in natural and cultivated environments. Based on a long-term trial, the amounts of water dissolved substances （WSS）, humic acid （HA）, fulvic acid （FA） and humin （HU） were determined to explore the impact of long-term fertilization on HS. Increases in the amounts of WSS, HA, FA and HU were significant different among the treatments with manure. A significant correlation was found between the increased soil organic carbon （SOC） and HS （R^2=0.98, P〈0.01）. The change in the E4/E6 ratio was significantly correlated with the increased SOC （R2=0.88, P〈0.01）, HA （R^2=0.91, P〈0.01）, FA （R^2=0.91, P〈0.01） and HU （R^2=0.88, P〈0.01）. The cluster was mainly divided into two parts as manure fertilization and inorganic fertilization, based on the increases in HA, FA and HU. These results suggest that long term fertilization with manure favours carbon sequestration in HS and is mainly stabilized as HU, while the HA becomes more aliphatic. We conclude that increases in SOC can be linked to changes in the molecular characteristics of HS fractions under long term fertilization.     

食品安全标准的贸易效应分析——以中国农产品出口为例

食品安全标准已经成为中国农产品出口的主要障碍。运用扩展引力模型,使用面板数据分析了食品安全标准对中国农产品出口的影响。结果表明,短期内,食品安全标准抑制贸易,长期内则促进贸易。因此,应充分理解食品安全标准对中国农产品出口的潜在影响,短期内可开拓食品安全标准较宽松的市场,使出口市场多元化;长期内需要不断提高产品的质量水平。     

生猪交易物流运作模式的分析

本文根据我国生猪产业发展的现状、生猪物流系统内部发展及运行环境,从物流主体的组织化程度、流通环节、信息的传播、资产专用性等角度来分析了我国现有三种生猪物流运作模式的特点,并在此基础上设计了适应未来生猪市场的以第四方电子交易平台为主导的新型物流模式。     

长期施肥对渭北旱塬麦田土壤线虫群落特征的影响

为探究长期定位施肥对黄土高原渭北旱塬麦田土壤线虫群落的影响,以长武国家黄土高原农业生态试验站的长期定位试验（1984—2018年）为平台,调查裸地（L）以及不施肥（CK）、施氮磷肥（NP）、单施有机肥（M）、氮磷肥配施有机肥（MNP）的小麦田土壤线虫群落数量、组成结构和生态功能指数,并分析其与土壤理化性质的关系。结果表明：（1）小麦田土壤线虫数量显著高于裸地,而CK、M和MNP处理使小麦田中土壤线虫数量比NP处理提高了31.63%~56.20%;（2）施肥减缓了长期种植小麦导致的食细菌线虫（特别是头叶属）相对丰度的下降,单施有机肥的处理（M）植食性线虫相对丰度显著降低,同时,相比于CK,施用有机肥的M和MNP处理杂食/捕食线虫相对丰度分别增加了18.4%和8.24%,表明有机肥对杂食/捕食线虫的促进作用;（3）长期施肥处理土壤线虫Shannon多样性指数（H）和指示土壤健康状况的瓦斯乐卡指数（WI）分别为1.80~2.19和0.36~0.68,相较于裸地（H=2.36;WI=1.57）有所下降,但M和MNP处理的线虫成熟度指数（MI）和结构指数（SI）均高于其他处理（L、CK和NP处理）,表明施用有机肥有助于土壤食物网维持复杂的结构和成熟稳定的状态;（4）不同施肥措施下土壤全氮、有机碳、微生物生物量碳和氮以及可溶性碳的变化是影响土壤线虫数量和群落特征的重要环境因素。     

Analysis of series of variety trials with perennial crops

Field trials with perennial crops give rise to repeated measurements taken on the same plot on several occasions. It is important to account for serial correlation among repeated measurements in such trials. This study illustrates the use of mixed models for this purpose. We consider the analysis of trials conducted at several locations and the combination of trials with different starting years. A key issue in the analysis is the distinction between effects of calendar years, which are associated with external environmental variation, and harvest years, which represent internal yield formation processes of the perennial crop. Two methods of two‐stage analysis are compared with single‐stage analysis. It is shown that results of two‐stage analysis are very similar to those of single‐stage analysis, if serial correlation is properly taken into account. Program code for the MIXED procedure of the SAS system is given in an Appendix S1 for all considered cases.     

厚壁毛竹六个节气笋芽发育的转录组分析

厚壁毛竹(Phyllostachys edulis‘Pachyloen’)是目前发现的毛竹多个变异类型中唯一一个有用材价值的优良新品种,但其分子遗传机制目前还未知。利用第二代高通量测序技术从霜降节气开始,对从霜降、大寒、冬至、雨水、春分、谷雨6个节气采集的厚壁毛竹笋芽分别进行转录组测序,共捕获30 555个基因,基因平均覆盖率为79%。将不同节气的竹笋转录组进行两两比较,首次证实厚壁毛竹笋在不同节气基因表达有差异,其差异表达基因数从2 288~12 991,进一步将这些差异表达基因进行了GO功能注释和KEGG Pathway富集性分析,并初步报道了DNA复制、类胡萝卜素生物合成、淀粉和蔗糖代谢3个代谢通路的基因表达模式,在一定程度上解析了厚壁毛竹"竹壁特厚"等种质性状的分子调控模式及其机制。     

框架与框架填充墙结构的统计损伤诊断

提出了一种基于递推随机有限元方法(RSFEM)的随机结构损伤识别方法来识别框架与框架填充墙结构的损伤.在定义了随机损伤指数概念的基础上,考虑模型误差的不确定性和测量噪声的影响,建立了关于随机损伤指数的控制方程.然后,利用RSFEM得到了结构随机损伤指数的统计特性.数值算例的结果显示,新的方法与两步法结合能在考虑模型误差和测量噪声的情况下对结构损伤进行有效识别,且在结构随机参数有较大涨落情况下,该方法与蒙特卡洛模拟法同样有效.     

运用混沌理论制作长期灾害预报模型初探

为了提供准确的农业灾害预报，本研究运用混沌理论，以北京历年重大旱灾预报为例，构建了长期灾害混沌预报模型。该模型采用便于描绘混沌事件的三阶差分方程与多因子状态方程联立，模拟出北京市1955－2001年的历次旱灾事件发生过程。按照该模型规定的法则做重大旱灾预测比用单纯统计预测的准确率有显著提高。灾害混沌预报模型可作为长期灾害预测的新途径。