A Two-Phase Optimization Model Based on Inexact Air Dispersion Simulation for Regional Air Quality Control

This study proposes a two-phase optimization model for regional air pollution control. To predict the pollutant concentrations at receptor zones, an interval Gaussian plume model is advanced to facilitate the generation of optimal pollution control policies. Results from the case study indicate satisfactory performance of the proposed model in handling uncertainties in parameters expressed as intervals and in stipulations associated with pollutant emission and ambient air quality. Compared with conventional models, it has advantages of generating compromised management strategies according to decision makers’ preference. This would be useful when the guarantee of satisfying all constraints is inapplicable or too costly. The proposed model is capable of identifying key factors and/or input conditions that may intensely affect system outputs and thus facilitating decision makers in adjusting current system status to benefit future management. The results also reveal a significantly enhanced satisfactory level would be obtained compared with conventional “single-phase”-based optimization models.     

Field tests of a linear gradient system for exposure of row crops to S02 and HF

An exposure system for exposing plants to gradients of gaseous air pollutants in the field was tested using sulfur dioxide SO₂ and hydrogen fluoride (HF). Well defined linear gradients of SO₂ concentration and HF flux were easily produced and were repeatable from exposure to exposure. The gradients were altered by wind speed and direction, but in tightly closed canopies, the alterations were minor. This system has many advantages: a graded series of exposures can be conducted in a small area, plants may be grown using accepted cultural practices, the cost of the apparatus is low, and more than one pollutant can be used, either concurrently or countercurrently. Disadvantages include the requirement for an intensive air monitoring network and an unnatural vertical pollutant profile in the canopy. The gradient system should be used as a supplement to open-top chambers, and not as a replacement for them.     

不同进风方案下隧道烘干窑热风流场CFD模拟和优化

隧道烘干窑内同一横截面的热风均匀性影响着物料干燥均匀性和产品质量,而烘干窑入口风速分布直接影响着窑内热风流场的均匀分布。为了解决单一风机直进风隧道烘干窑存在的风速不均匀问题,提出了多种进风结构设想,并利用计算流体力学方法对实际生产的隧道窑进风流场进行数值模拟,研究3种不同的进风方案(4风机、6风机和9风机)对隧道窑内热风流场均匀性的影响。模拟结果表明:6风机方案下隧道窑入口处进风均匀,热风扩散距离短,窑内热风流场整体均匀性较佳,综合性价比最高。研究结果为隧道窑入口进风的设计提供参考。     

葡萄差压通风预冷影响参数的试验研究

果蔬差压通风预冷包装箱的开孔大小及包装箱两侧的压差是影响预冷速度、冷却均匀性的重要参数。本文对不同开孔大小、不同压差工况下葡萄差压通风预冷过程进行了试验研究，分析了它们对葡萄冷却速度、冷却均匀性的影响。结果表明，开孔面积的增加有利于加强冷风的横向流动扩散；压差的增大有利于提高冷风的径向渗透性。增大开孔直径和压差可以不同程度地缩短葡萄的7/8冷却时间、在一定条件下改善冷却均匀性。对于该试验采用的葡萄包装箱，综合考虑预冷速度、冷却均匀性以及预冷后葡萄的温升速度、干耗等因素，开孔直径在28 mm左右、压差保持在100 Pa左右为宜。     

采用轻量级网络MobileNetV2的酿酒葡萄检测模型

为提高高分辨率田间葡萄图像中小目标葡萄检测的速度和精度,该研究提出了一种基于轻量级网络的酿酒葡萄检测模型（Wine Grape Detection Model,WGDM）。首先,采用轻量级网络MobileNetV2取代YOLOv3算法的骨干网络DarkNet53完成特征提取,加快目标检测的速度;其次,在多尺度检测模块中引入M-Res2Net模块,提高检测精度;最后,采用平衡损失函数和交并比损失函数作为改进的定位损失函数,增大目标定位的准确性。试验结果表明,提出的WGDM模型在公开的酿酒葡萄图像数据集的测试集上平均精度为81.2%,网络结构大小为44 MB,平均每幅图像的检测时间为6.29 ms;与单发检测器（Single Shot Detector,SSD）、YOLOv3、YOLOv4和快速区域卷积神经网络（Faster Regions with Convolutional Neural Network,Faster R-CNN）4种主流检测模型相比,平均精度分别提高8.15%、1.10%、3.33%和6.52%,网络结构分别减小了50、191、191和83 MB,平均检测时间分别减少了4.91、7.75、14.84和158.20 ms。因此,该研究提出的WGDM模型对田间葡萄果实具有更快速、更准确的识别与定位,为实现葡萄采摘机器人的高效视觉检测提供了可行方法。     

Particle Deposition and Decay in a Chamber and the Implications to Exposure Assessment

Accurate modeling indoor aerosol deposition and decay is an important step for exposure assessment. High deposition rate reduces indoor pollutant concentration and results in lower inhalation exposure. Many of indoor surfaces have random roughness protrusion scales up to several millimeters which it may significantly affect the deposition loss rate. Aerosols deposition onto most indoor surfaces can be considered as “rough’ surface deposition. However, particle deposition from anleaflong indoor environment is frequently modeled by assuming that surfaces are smooth. In this work, experimental results for deposition velocity onto drywall surfaces for supermicron particles from 1 to 7 μm are presented. Deposition velocity is significantly higher than that predicted by a previous published model. To illustrative the influence of particle deposition on exposure, two hypothetical room sizes and two air exchange rates were considered. Inhalation transfer factor was adopted as a risk exposure index. Taking into account of high deposition velocity onto drywall surfaces, inhalation transfer factors are 8 to 35% lower than that predicted by the model and this significant difference is important to the exposure assessment.     

基于深度学习的葡萄果梗识别与最优采摘定位

针对葡萄采摘机器人在采摘作业中受果园环境干扰,难以准确识别与分割葡萄果梗及定位采摘点的问题,该研究根据葡萄生长的特点提出一种基于深度学习的葡萄果梗识别与最优采摘定位方法。首先通过改进掩膜区域卷积神经网络(Mask Region with Convolutional Neural Network,Mask R-CNN)模型对果梗进行识别与粗分割;然后结合阈值分割思想对果梗的色调、饱和度、亮度(Hue Saturation Value,HSV)色彩空间进行分段式提取,取每段色彩平均值作为该段果梗基准颜色阈值,利用区域生长算法对果梗进行精细化分割;最后计算果梗图像区域的质心,并以临质心点最近的果梗水平两侧中心作为最终采摘点。试验结果表明,在不同天气光照下该方法对葡萄果梗的检测精确率平均值为88%;在果梗成功识别后最优采摘点定位准确率达99.43%,单幅图像的果梗采摘定位平均耗时为4.90s,对比改进前Mask R-CNN检测耗时减少了0.99 s,F1-得分提高了3.24%,检测效率明显提升,该研究为葡萄采摘机器人提供了一种采摘点定位方法。     

On setting air quality criteria

A method is proposed to relate simple biological response modeling to air quality criteria. We also suggest a procedure using the biological response curve in conjunction with measured pollutant concentrations for setting emission levels to ensure compliance with air quality standards. A byproduct of our analysis is art air pollution index which avoids the arbitrariness of commonly used indices.     

Future directions in photochemical air quality modeling

Photochemical air quality models provide the most defensible method for relating future air quality to changes in emission, and hence are the foundation for determining the effectiveness of proposed control strategies. However, strategies, based primarily on controlling reactive organic gas emissions, have not provided the expected benefits. This raises the question of what have been the deficiencies in previous studies utilizing these tools? Furthermore, what changes are necessary, and desired, to improve upon past efforts? The current generation of models have matured within their original frameworks to represent, relatively accurately, the important physical and chemical processes affecting pollutant dynamics in urban atmospheres. The ability to follow regional dynamics is less well demonstrated. Current regional models have a single horizontal resolution scale. Multiscale models will enable detailed treatment of urban chemistry, and also effectively follow long range transport and chemistry. Improved computational capabilities will allow more detailed chemistry and heterogeneous processes to be followed within the models. The practice of photochemical modeling will benefit greatly from recent and future intensive field studies. The advancements in both the model framework and practice will allow much more accurate evaluation of proposed control strategies, and lead to a much improved understanding of pollutant dynamics.     

自然环境下贴叠葡萄串的识别与图像分割算法

针对自然环境下贴叠葡萄串难以识别与分割的问题,该文首先提取HSV颜色空间中的H分量,获取贴叠葡萄串区域,分析该区域长宽比从而判定葡萄串的贴叠性质;提取葡萄串图像轮廓信息,获取轮廓拐点与类圆心点信息;利用拐点与中心点之间的斜率判定目标葡萄串所在位置。然后,利用Chan-Vese模型进行葡萄串的迭代识别,并结合拐点信息获得重叠边界的轮廓信息。最后,将重叠边界轮廓与图像轮廓进行融合,实现目标葡萄串识别。试验结果表明,该文方法的平均精准度为89.71%,平均假阳率为4.24%,识别成功率为90.91%,与现有方法相比,该文方法可实现完整目标葡萄串的识别与分割,并提高了识别与分割的精准度,为葡萄采摘机器人成功采收贴叠葡萄串提供切实可行的算法。     

A New Imitative Model to Predict the Impact of Air Pollutants on Scots Pine Health and Radial Increment

In order to address the problem of pollutant effects on trees against the background of environmental fluctuations a simulation model using radial increment as an indicator of tree productivity is proposed. Ring times series of large trees provide one of the most acceptable biological indicators for the assessment of pollutant impact. This research focused on the tallest severely damaged and healthy trees growing close to each other in the surroundings of Jonava Mineral Fertilizer plant, which primarily emits nitrogen (N) and sulphur (S) compounds. The annual increment was found to be strongly correlated with mean September air temperature of the previous year and S load. The correlation coefficient (k=0.961) between modelled and observed tree ring times series of severely damaged trees proves the impact of S load on growth and vitality of a tree and enables fulfillment of tree condition prognosis.     

Modeling of first plume encounters with precipitation

A regional air pollutant transport model was used to simulate the fate of S emissions in the northeast United States. Hourly calculations were analyzed to describe trajectory characteristics and mass balances as pollutant trajectories first encounter precipitation during transport away from emission sources. Model results include air concentrations of SO₂ and sulfate, dry deposition values, and sulfate concentrations scavenged by precipitation, along with trajectory statistics. Results of sensitivity tests are compared to base case simulations which consider all precipitation events as a means of suggesting priorities for future regional transport model development.     

基于通道特征金字塔的田间葡萄实时语义分割方法

复杂环境下葡萄的快速检测识别是智能采摘的关键步骤,为解决目前葡萄识别精度低和实时性差的问题,该研究提出一种轻量级葡萄实时语义分割模型（Grape Real-time Semantic Segmentation Model,GRSM）。首先,利用通道特征金字塔（Channel-wise Feature Pyramid,CFP）模块进行特征提取,该模块通过1?3和3?1空洞卷积的跳跃连接,在减少模型参数量的同时提取葡萄图像的多尺度特征和上下文信息;然后,采用池化卷积融合结构完成下采样,增加可训练参数以减少信息损失;最后,利用跳跃连接融合多种特征恢复图像细节。试验结果表明：该研究所提出的模型在田间葡萄测试集上达到了78.8%的平均交并比,平均像素准确率为90.3%,处理速度达到68.56帧/s,网络结构大小仅为4.88MB。该模型具有较高分割识别精度和较好实时性,能满足葡萄采摘机器人对视觉识别系统的要求,为葡萄的智能化采摘提供了理论基础。     

Effects of varying air trajectories on spatial and temporal precipitation chemistry patterns

This study was designed to determine if judicious use of synoptic data and an operational trajectory model could identify probable source regions of anthropogenic pollutants in northeastern United States precipitation and thus relate receptor measurements to emissions data without consideration of the complex intervening meteorological and chemical processes. The storm event of April 8 to 10, 1979, was selected for intensive study. Precipitation chemistry data were obtained from event samples at six MAP3S sampling sites and from hourly samples at Brookhaven National Laboratory. Concentrations of hydrogen, sulfate, nitrate and ammonium ions were used as receptor data. Some emissions data for SO_x and NO_x were obtained from the MAP3S emissions inventory. Surface and upper air meteorological data were analyzed. Backward trajectories ending at each of the sampling sites during the precipitation period were computed with the Heffter Interactive-Terminal Transport Model using selected transport layers. Results show that concentrations of pollutant species in event precipitation samples were much higher at stations at end points of trajectories passing through the Ohio River valley than at stations with other trajectories. Likewise, concentrations at Brookhaven were much higher during the end period of a trajectory through the same region than with more northerly and more southerly tracks. The model produced back trajectories consistent with synoptic flows. Concentrations of air pollutants in precipitation were roughly proportional to the number of major pollutant sources along the trajectory. These results suggest that a larger number of studies might identify more restricted source areas or even establish a quantitative relationship between source emissions along a trajectory and concentrations in precipitation at receptor sites.     

Impact assessment of urban waste processing systems using a multibox model

A set of 14 household waste management system scenarios have been ranked according to the environmental impact of their respective refuse outputs. Each system has been characterized by the annual release of 8 chemical elements, namely: S, Cl, Fe, Cu, Zn, Cd, Hg and Pb. The method relies on the determination of a load factor, which is defined as the ratio between the actual pollutant concentration and the accepted norm for this concentration. These figures are calculated via a three box model, where the three boxes are air, water and soil. The model considers the pollutant output from each scenario, the dilution or accumulation in each box as well as dry deposition from air to water and soil. The pollutant load factors are then aggregated in order to obtain a score. These scores allow a final scenario grade to be determined, which can then be used to rank the various scenarios. The final grades are strongly dependant on the presence of deposition terms as well as on the values chosen for the individual pollutant concentration norms, whereas they are reasonably independent of all other parameters.     

Identification of risk factors for sub-optimal housing conditions in Australian piggeries: Part 1. Study justification and design

We undertook a literature search related to pig production facilities with two major aims: first, to review all the likely benefits that might be gained from air quality improvements; and second, to review previous research that had identified statistically significant factors affecting airborne pollutants and environmental parameters, so that these factors could be considered in a multifactorial analysis aimed at explaining variations in air pollutant concentrations. Ammonia, carbon dioxide, viable bacteria, endotoxins, and inhalable and respirable particles were identified as major airborne pollutants in the review. We found that high concentrations of airborne pollutants in livestock buildings could increase occupational health and safety risks, compromise the health, welfare, and production efficiency of animals, and affect the environment. Therefore, improving air quality could reduce environmental damage and improve animal and worker health. To achieve a reduction in pollutant concentrations, a better understanding of the factors influencing airborne pollutant concentrations in piggery buildings is required. Most of the work done previously has used simple correlation matrices to identify relationships between key factors and pollutant concentrations, without taking into consideration multifactorial effects simultaneously in a model. However, our review of this prior knowledge was the first important step toward developing a more inclusive statistical model. This review identified a number of candidate risk factors, which we then took into consideration during the development of multifactorial statistical models. We used a general linear model (GLM) to model measured internal concentrations, emissions, and environmental parameters in order to predict and potentially control the building environment.     

Diffusion and emission of smoke from agricultural burning in Hawaii

Agricultural burning has been common practice for many years in Hawaii. Cane fields are burnt before harvesting and the crowns of pineapple are cut-off during harvest, left in the field to dry and subsequently burned. While this practice has many advantages, it produces air pollution levels which at least one study finds affect public health. In 1974 open burns of cane and pineapple were simulated in a burn tower where samples were burned while monitoring weight loss and smoke density. From these measurements emission values were calculated. However, because questions were raised as to whether the vigorous field fires can be simulated. in a burn tower, a field survey was undertaken and a diffusion model, basically a limited Gaussian line source model, was developed and run using the data collected during the burns. The calculated emission rate for particulates is orders of magnitudes larger than the burn tower rate. This is attributed to dust drawn into the fire. When emission values from a survey of the open burn literature and from the present study were plotted against wind speed, low intensity fires such as grass, straw and pineapple residue fires showed decreased emissions with increasing wind speed while cane fires showed increased emission. This suggests that emission characteristics are different for low and high intensity fires. For low intensity fires increased air flow acts mainly to increase the combustion efficiency producing less particulates. For high intensity fires, such as cane fires, increased air flow results in hotter fires lofting more dust, which is the major pollutant from such fires. Revised emission values are recommended.     

基于迁移学习和改进CNN的葡萄叶部病害检测系统

为建立高效、准确的葡萄叶部病害检测系统,引入迁移学习机制,利用大型公开数据集对VGG16模型预训练,保持模型前端13个层的参数和权重不变,对全连接层和分类层改进后利用新数据集微调训练模型,包括对训练优化器、学习率和中心损失函数平衡参数的优选试验,最后将模型部署在Android手机端。试验表明,在微调训练阶段选择Adam优化器、初始学习率设为0.001、中心损失函数平衡参数设为0.12时,改进的VGG16模型性能最优,对葡萄6类叶部图像的分类平均准确率为98.02%,单幅图像平均检测耗时为0.327s。与未改进的VGG16模型相比,平均准确率提高了2.82%,平均检测耗时下降了66.8%,权重参数数量减少了83.4%。改进后的模型综合性能优于AlexNet、ResNet50和Inceptionv3等模型。将模型跨平台部署在Android手机端,自然环境下验证的平均准确率为95.67%,平均检测耗时为0.357 s。该研究建立的基于迁移学习和改进卷积神经网络的病害检测系统可实现对葡萄叶部病害的快速、智能诊断,为葡萄病害的及时防控提供依据。     

Study on Winter Air Pollution Control in Lanzhou,China

Air pollution control is one of the most important issues in thecities of the developing countries. Based on an analysis of the local economic and technological development the paper puts forward 9 air pollution control schemes. They are 4 briquette schemes, 2 coal gas schemes, 1 centralized heating system and 2 comprehensive schemes. A comprehensive diffusion model combined with a 3-D advection-diffusion equation and the Gaussian puffmodel is developed and verified. According to the calculation ofpotential reduction of pollutant discharge and the correspondingair quality improvement made by each air pollution controlscheme, the paper then works out a feasible and comprehensiveair pollution control scheme, which is characterized by lowinvestment and better environmental benefits. When the scheme isefficiently carried out, 19.3% SO₂ and 24.4% sootdischarge is to be reduced annually in the city. There is aclear descent of air pollution concentration. The daily mean concentration of SO₂ and TSP declines 24.6–75.4%and 10.4–45.4%, respectively. SO₂ concentation of the whole city does not exceed the second grade of the Ambient Air Quality Standard of China except at a few grid points.     

Atmospheric Stability and Meteorological Scenarios as Inputs to Air Pollution Transport Modeling

Meteorological scenarios concerning the data required for engineering applications of pollutant transport modeling in the low layers of troposphere are defined through a simple methodology. This involves only data at soil surface and substantially relies on the determination of atmospheric stability carried out through fictitious vertical profiles of air temperature considered as representative of the actual vertical profiles in a study area. This assumption is supported by comparisons of fictitious vertical profiles, obtained by measurements of air temperature at soil surface but at different heights above sea level, and the temperature vertical profiles observed by the radio-sounding station closer to the study area. The fictitious profiles are first used to derive the usual classes of atmospheric stability as unstable atmosphere, neutral atmosphere, stable atmosphere, thermal inversion at the surface and thermal inversion in the upper layers. Then, each scenario is determined through the classes of atmospheric stability observed at 06.00 GMT and 15.00 GMT together with the experimental data of air temperature, relative humidity, wind speed, wind direction, and cloud cover, all averaged in the period of investigation of a few years. An area of Central Italy, where the meteorological measurements for a period of 7 years were available, has been selected for this study.