基于BP神经网络的柴油发动机特性建模

建立连续的发动机燃油特性和调速特性数学模型作为液压机械无级变速器虚拟试验平台的动力源。根据虚拟平台对不同特性区域的精度需求对柴油发动机不同特性区域的试验数据进行不同的密度选取、乱序和归一化处理,采用单隐层BP神经网络对试验数据进行训练,对比不同隐层节点数网络的训练误差和测试误差,选取误差最小的网络,求解出网络的数学表达式。通过该方法以ISLe310柴油发动机为例建立燃油特性和调速特性的连续数学模型,这两个简单的数学表达式准确反映了发动机万有特性和外特性,连续模型避免了虚拟试验中出现信号的突变和奇异点。通过和经典的最小二乘法拟合得到的最优特性模型进行对比,其具有更小的误差、更强的泛化能力,能够更好地反映柴油发动机的相关特性。     

Assessing the utility of topographic variables in predicting structural complexity of tree stands in a reforested urban landscape

The transformation of natural landscapes into impervious built-up surfaces through urbanization is known to significantly interfere with the ecological integrity of urban landscapes and accelerate climate change and associated impacts. Although urban reforestation is widely recognised as an ideal mitigation practice against these impacts, it often has to compete with other lucrative land uses within an urban area. The often limited urban space provided for reforestation therefore necessitates the optimization of the ecological benefits, which demands spatially explicit information. The recent proliferation of tree stands structural complexity (SSC) and topographic data offer great potential for determining the ecological performance of reforested areas across an urban landscape. This study explores the potential of using topographic datasets to predict SSC in a reforested urban landscape and ranks the value of these topographic variables in determining SSC. Tree structural data from a reforested urban area was collected and fed into a tree stand structural complexity index, which was used to indicate ecological performance. Topographic variables (Topographic Wetness Index, slope, Area Solar Radiation and elevation)- were derived from a Digital Elevation Model (DEM) and used to predict SSC using the Partial Least Squares (PLS) regression technique. Results show that SSC varied significantly between the topographic variables. Results also show that the topographic variables could be used to reliably predict SSC. As expected, the Topographic Wetness Index and slope were the most important topographic determinants of SSC while elevation was the least valuable. These results provide valuable spatially explicit information about the ecological performance of the reforested areas within an urban landscape. Specifically, the study demonstrates the value of topographic data as aids to urban reforestation planning.     

基于连续投影算法和光谱变换的冬小麦生物量高光谱遥感估算

为探索基于全波段冠层高光谱以及变换光谱的冬小麦地上部生物量的遥感估算方法，以2016、2017年冬小麦田间试验为基础，通过对冠层光谱和地上部生物量的相关性分析，筛选拔节期、抽穗期的冬小麦冠层光谱、一阶导数光谱、对数变换光谱和连续统去除光谱对地上部生物量的敏感波段，并结合偏最小二乘法(PLS)分别建立拔节期和抽穗期基于SPA算法的冬小麦地上部生物量估测模型，再与基于任意两波段组合的最佳归一化光谱指数、比值光谱指数、差值光谱指数和已报道光谱指数的冬小麦地上部生物量估测模型进行比较。结果表明：(1)SPA算法较好地利用了全波段冠层光谱信息，并显著降低了光谱维度，不同变换光谱的地上部生物量敏感波段个数在4~14之间；(2)拔节期和抽穗期冠层光谱与地上部生物量的相关性高于开花期和灌浆期，各生育时期一阶导数光谱与地上部生物量之间的相关性优于连续统去除光谱、对数变换光谱和光谱指数；(3) 利用抽穗期一阶导数光谱敏感波段建立的预测模型和验证模型达到了较高的精度，其预测模型的决定系数和均方根误差分别为0.78和0.87 t·hm^-2，验证模型的决定系数和均方根误差分别为 0.84和0.69 t·hm^-2，预测相对偏差为2.74。这说明，抽穗期是估算地上部生物量的最佳生育时期，且基于冠层一阶导数变换光谱，结合连续投影算法和偏最小二乘回归方法所构建抽穗期地上部生物量估算模型具有最优的精度和预测能力，可用于地上部生物量的定量估算。     

2003-2017年陕西省NDVI时空变化及其影响因素

[目的] 研究影响陕西省植被覆盖主导气候类和非气候类因子，为区域生态文明建设提供科学依据。[方法] 基于2003—2017年MODIS NDVI数据，利用趋势分析和地理探测器模型等方法，研究了气候类和非气候因素对陕西省植被覆盖的影响。[结果] ①2003—2017年陕西省NDVI空间分布总体上为改善趋势，但在不同植被类型区和生态区有所差异。②降水是影响陕西省植被覆盖空间分布的主导气候类因素，其他因素对当地植被覆盖影响程度有所差异。③降水与气温、日照、风速、相对湿度的交互作用对陕西省植被覆盖空间分布起主导作用。且气温的作用只有在和降水的交互作用下才能体现出来。④植被类型和地貌是陕西省植被覆盖空间分布的主导非气候类因素，其他因素的影响有所差异。⑤植被类型、地貌和土壤类型之间的交互作用对陕西省植被覆盖空间分布起主导作用。人口、GDP的影响也只有在与其他因素的交互作用下才能显现出来。[结论] 气候类因素的影响大于非气候类因素，气候类和非气候类因素的共同作用能够充分的解释植被覆盖空间分布。     

Technical efficiency of different farm sizes for tilapia farming in China

China is the world's largest tilapia producer. This study uses the stochastic frontier function in the Cobb–Douglas model to evaluate a production frontier for tilapia aquaculture and analyses the relationship between farm size and production efficiency in China. A random sample of 300 tilapia farmers was surveyed in the main tilapia‐producing areas to study their economic efficiencies. The percentage distribution of tilapia farm technical efficiency was determined, with an average efficiency of 79%, which showed that the sample tilapia farmers operate at 21% below the production frontier and hence that they still have a chance to achieve targeted yields. The technical efficiencies of two categories (≤1 ha and >1 ha) were 78.82% and 79.27% respectively. A second‐stage analysis investigated the relationship between farm size and technical efficiency and showed a positive relationship between farm size and production efficiency based on the ordinary least‐squares model. The optimal farm size was found to be 7.50 ha. This study proposes strategies such as enlarging the farming scale moderately, thereby achieving the advantage of economies of scale, and enhancing the competitiveness of the tilapia industry in China.     

Feasibility analysis of NIR for detecting sweet corn seeds vigor

This paper explores the feasibility of particle-based detection and grading of seed vigor based on a self-built seed single-granulation device using near infrared spectroscopy (NIRS). Sweet corn with uniform kernel size was used for this study. The seed samples were divided into three types, they were normal seeds, artificially aged seeds and heat-damaged seeds. A 2-part spectral acquisition of each seed were performed, one for the collection of seeds that fall into the detection zone within the separation pipe, another was on the static platform, whose collection was performed on 5 faces of each seed. Partial least squares discriminant analysis (PLS-DA) was used to classify the original data of the seeds. In the 2 parts, the discriminant results of the unprocessed normal seeds and the artificial accelerated aging seeds, the untreated normal seeds and the heat-damaged seeds showed that classification accuracy was higher than 98%. The research indicates that the spectral data of different positions of seeds can reflect their activity information, and it is feasible to detect and classify seeds in real time in the detection area of the separation pipeline.     

宁夏玉米营养成分近红外分析模型建立与应用

研究采集宁夏区内各市县(固原、海原、同心、中卫、中宁、永宁、平罗)抽取玉米样品120批,用近红外分析仪扫描定标样品集获得玉米近红外光谱图,利用偏最小二乘法建立模型,并分别经过无预处理、均值中心化、标准正态变量转换、一阶导数、标准正态变量转换结合去趋势校正(SNV+D)预处理光谱,获得水分、粗蛋白、粗脂肪、粗纤维预测模型,分别是经过MC、SNV+D、SNV+D、SNV+D光谱预处理后效果最好。预测决定系数R² _p分别为0.951、0.976、0.728、0.897,相对分析误差RPD除粗脂肪为2.62外,其他均大于3的近红外预测模型,通过预测模型验证集验证后,并对模型预测值与实测值进行U检验,结果为差异不显著(P>0.05)。该近红外预测模型对玉米中水分、粗蛋白、粗纤维具有较佳预测效果,粗脂肪的预测精度有待进一步提高。     

基于PLSR-BP复合模型的红壤有机质含量反演研究

对红壤地区土壤有机质进行快速预测,以满足智慧农业与精准施肥的需要。以江西省奉新县北部为研究区域,采用1 km×1 km标准格网划分研究区进行采样,共得到红壤样本248个。对土壤光谱进行了包含分数阶导数在内的3种数学变换方法,将经过P=0.01显著性检验的波段用于模型的构建,选用偏最小二乘回归(PLSR)和BP神经网络建立土壤有机质含量预测模型。结果表明:当对红壤光谱数据进行1.5阶导数变换后再使用PLSR-BP复合模型对土壤有机质含量进行预测时的结果为最优,训练集R~2=0.89,RMSE=4.68g·kg~(-1),验证集R~2=0.87,RMSE=5.55g·kg~(-1),RPD=2.75。1.5阶导数对红壤光谱数据的变换能够更好地突出与有机质相关的特征信息,有助于其含量预测。PLSR-BP复合模型预测精度优于单一模型,能够较好地预测红壤有机质含量,为精准农业快速监测红壤有机质含量提供了新的途径。     

Whole wheat bread: Effect of bran fractions on dough and end-product quality

Consumption of whole-wheat based products is encouraged due to their important nutritional elements that benefit human health. However, the use of whole-wheat flour is limited because of the poor processing and end-product quality. Bran was postulated as the major problem in whole wheat breadmaking. In this study, four major bran components including lipids, extractable phenolics (EP), hydrolysable phenolics (HP), and fiber were evaluated for their specific functionality in flour, dough and bread baking. The experiment was done by reconstitution approach using the 2⁴ factorial experimental layout. Fiber was identified as a main component to have highly significant (P < 0.05) and negative influence on most breadmaking characteristics. Although HP had positive effect on farinograph stability, it was identified as another main factor that negatively impacted the oven spring and bread loaf volume. Bran oil and EP seemed to be detrimental to most breadmaking characteristics. Overall, statistical analysis indicates that influence of the four bran components are highly complex. The bran components demonstrate multi-way interactions in regards to their influence on dough and bread-making characteristics. Particularly, Fiber appeared to have a high degree of interaction with other bran components and notably influenced the functionality of those components in whole wheat bread-making.