Field calibration of a neutron moisture meter in a cracking grey clay

Summary A neutron moisture meter was field calibrated in a cracking grey clay prepared for furrow irrigation at Narrabri, N.S.W. Neutron counts were taken in successive 0.1 m increments between the 0 and 1.5 m depths. Concomitant measurements using undisturbed soil cores provided independent estimates of volumetric water content. Separate linear calibrations were required for the 0–0.1 m, 0.1–0.2 m and 0.2–1.5 m depth increments. Correction for bias due to cracking and changes in bulk density slightly improved the calibrations. The accuracy of predicted soil water content was improved relative to previous calibrations. A precision of ±0.01 m³m^–3 required 3 samples per mean by the neutron method or 11 samples per mean by the core sampling method.     

Suspended sediment load prediction of river systems: An artificial neural network approach

Information on suspended sediment load is crucial to water management and environmental protection. Suspended sediment loads for three major rivers (Mississippi, Missouri and Rio Grande) in USA are estimated using artificial neural network (ANN) modeling approach. A multilayer perceptron (MLP) ANN with an error back propagation algorithm, using historical daily and weekly hydroclimatological data (precipitation P_(t), current discharge Q_(t), antecedent discharge Q_(t−1), and antecedent sediment load SL_(t−1)), is used to predict the suspended sediment load SL_(t) at the selected monitoring stations. Performance of ANN was evaluated using different combinations of input data sets, length of record for training, and temporal resolution (daily and weekly data). Results from ANN model were compared with results from multiple linear regressions (MLR), multiple non-linear regression (MNLR) and Autoregressive integrated moving average (ARIMA) using correlation coefficient (R), mean absolute percent error (MAPE) and model efficiency (E). Comparison of training period length was also made (4, 3 and 2 years of training and 1, 2 and 3 years of testing, respectively). The model efficiency (E) and R² values were slightly higher for the 4 years of training and 1 year of testing (4 * 1) for Mississippi River, indifferent for Missouri and slightly lower for Rio Grande River. Daily simulations using Input 1 (P_(t), Q_(t), Q_(t−1), SL_(t−1)) and three years of training and two years of testing (3 * 2) performed better (R² and E of 0.85 and 0.72, respectively) than the simulation with two years of training and three years of testing (2 * 3) (R² and E of 0.64 and 0.46, respectively). ANN predicted daily values using Input 1 and 3 * 2 architecture for Missouri (R² = 0.97) and Mississippi (R² = 0.96) were better than those of Rio Grande (R² = 0.65). Daily predictions were better compared to weekly predictions for all three rivers due to higher correlation within daily than weekly data. ANN predictions for most simulations were superior compared to predictions using MLR, MNLR and ARIMA. The modeling approach presented in this paper can be potentially used to reduce the frequency of costly operations for sediment measurement where hydrological data is readily available.     

基于神经网络预测的无线传感器网络田间射频信号路径损耗

为解决应用无线传感器网络技术监测农田信息时无法快速预测射频信号路径损耗的问题,基于神经网络理论研究了田间路径损耗与其影响因素间的关系。试验中选取915和2 470 MHz 2个载波频率,在冬小麦的不同生长阶段测量射频信号在田间各影响因素作用下的路径损耗,建立和验证基于神经网络的射频信号田间路径损耗预测模型。所建立模型模拟值与实测值的相关系数为0.92,应用建立的神经网络预测田间射频信号路径损耗并与实测值对比,最大预测误差绝对值为4.186 dB,最大预测标准差为2.759 dB,预测准确度为94.2%。所建立的BP网络可以对田间射频信号路径损耗进行预测。     

农机总动力预测的灰色神经网络新方法

为提高农机总动力变化趋势的预测精度,将pGM(1,1)模型与RBF神经网络相结合,建立了基于pGM(1,1)-RBF神经网络的农机总动力预测模型,并以中国农机总动力数据预测为例,验证了该模型精度高、可行有效,适用于农机总动力预测.     

基于人工神经网络的消失模铸造充型过程数值模拟

本文建立了一种铸件消失模铸造充型过程数值模拟的人工神经网络模型。并利用此模型进行铸件的消失模铸造充型过程计算机数值模拟 ,其模拟充型过程结果与相应的实测值基本一致。     

基于神经网络的离心泵汽蚀性能预测

介绍了离心泵汽蚀性能预测的研究现状,分析了离心泵汽蚀性能预测的主要研究方法．根据设计流量下离心泵汽蚀余量的影响因素,确定人工神经网络的拓扑结构．应用MATLAB的神经网络工具箱,建立单级单吸离心泵汽蚀性能预测的BP神经网络（Back Propagation Neural Network）和RBF神经网络（Radial Basis Function Neural Network）两种人工神经网络模型．用工程实践中得到的57台离心泵几何参数和试验数据作为样本来训练建立好的网络,并用6台离心泵的数据来测试网络．预测值与试验值的相关性分析表明,BP和RBF网络的预测结果均较好,其中BP网络预测模型的平均相对偏差为5．69％,RBF网络预测模型的平均相对偏差为6．32％,可满足工程应用的要求．     

Estimating plant root water uptake using a neural network approach

Water uptake by plant roots is an important process in the hydrological cycle, not only for plant growth but also for the role it plays in shaping microbial community and bringing in physical and biochemical changes to soils. The ability of roots to extract water is determined by combined soil and plant characteristics, and how to model it has been of interest for many years. Most macroscopic models for water uptake operate at soil profile scale under the assumption that the uptake rate depends on root density and soil moisture. Whilst proved appropriate, these models need spatio-temporal root density distributions, which is tedious to measure in situ and prone to uncertainty because of the complexity of root architecture hidden in the opaque soils. As a result, developing alternative methods that do not explicitly need the root density to estimate the root water uptake is practically useful but has not yet been addressed. This paper presents and tests such an approach. The method is based on a neural network model, estimating the water uptake using different types of data that are easy to measure in the field. Sunflower grown in a sandy loam subjected to water stress and salinity was taken as a demonstrating example. The inputs to the neural network model included soil moisture, electrical conductivity of the soil solution, height and diameter of plant shoot, potential evapotranspiration, atmospheric humidity and air temperature. The outputs were the root water uptake rate at different depths in the soil profile. To train and test the model, the root water uptake rate was directly measured based on mass balance and Darcy's law assessed from the measured soil moisture content and soil water matric potential in profiles from the soil surface to a depth of 100 cm. The ‘measured’ root water uptake agreed well with that predicted by the neural network model. The successful performance of the model provides an alternative and more practical way to estimate the root water uptake at field scale.     

Assessment of a 4-input artificial neural network for ETo estimation through data set scanning procedures

Evapotranspiration is a complex and non-linear phenomenon that depends on the interaction of several climatic parameters. As an alternative to traditional techniques, artificial neural networks (ANNs) are highly appropriate for the modeling of non-linear processes. In general, in the most common ANN applications, the available climatic series are usually split up into 3 data sets: one for training, one for cross-validating, and one for testing. Up to now, the studies regarding ANN-models for reference evapotranspiration estimation and forecasting consider usually only a single chronological assignment of data for the definition of these 3 data sets. In these cases, the ANN performance can only be referred to this specific data set assignment. This paper analyzes the performance of a simple ANN model, a temperature-based 4-input ANN, taking into consideration a complete scan of the possible training, cross-validation, and test set configurations using ‘leave one out’ procedures. The results of a comparative analysis between both methodologies show that the performance results achieved with the traditional methodology can be misleading when evaluating the real ability of a model, as they are referred to the single specific data set assignment assumed.     

基于神经网络和粒子群算法产流量模型构建

为了研究坡面产流量与各影响因子间定量关系,分析野外多尺度人工径流小区实测数据,采用人工神经网络及粒子群算法,建立了坡面产流量预测模型,产流量与坡长、坡宽、坡度、前期土壤含水量间可采用二次抛物线关系进行描述,与植被覆盖度、降雨量间分别采用幂函数和线性函数进行描述.另外采用加权相对差距和法确定了产流量BP神经网络模型最优拓扑结构及网络参数,建立了产流量BP神经网络模型,该模型模拟值与实测相对误差在±20%以内,预测精度较高.同时基于产流量与各单因子定量关系,建立了产流量经验模型,采用粒子群算法推求了模型未知参数,该经验模型相对误差主要在±30%以内,其精度略逊于BP神经网络模型.     

Estimation of daily pan evaporation using artificial neural network and multivariate non-linear regression

Measurement of evaporation (E) rate from various natural surfaces is known as the key element in any hydrological cycle and hydrometeorological studies. Due to the shortage of pan evaporation (E _P) data, the estimation of E _P for such studies seems necessary. The main aim of this paper was to estimate daily E _P using artificial neural network (ANN) and multivariate non-linear regression (MNLR) methods in semi-arid region of Iran. Five different ANN and MNLR models comprising various combinations of daily meteorological variables, that is, relative humidity (RH), air temperature (T), solar radiation (SR), wind speed (U) and precipitation (P) were developed to evaluate degree of effect of each of these variables on E _P. The comparison of models estimates showed that the ANN 5 model characterized by Delta-Bar-Delta learning algorithm and Sigmoid activation function which uses all input parameters (T, U, SR, RH, P) performed best in prediction of daily E _P. The sensitivity analysis revealed that the estimated E _P data are more sensitive to T and U, respectively. A comparison of the model performance between ANN and MNLR models indicated that ANN method presents the best estimates of daily E _P.     

A manually controlled TDR soil moisture meter operating with 300 ps rise-time needle pulse

Summary The principle of operation of a simple, manually controlled Time Domain Reflectometer (TDR meter) for soil moisture measurements, which operates with a needle pulse of 300 ps rise-time, is described. A block diagram and constructions are also given. Construction of a switchless multiple sensor probe, having an inherent delay reference, is presented. Results of measurements of the soil dielectric constant as related to water content, for soils having different bulk densities, textures and humus content show a high correlation. The results agree closely with other investigators measurements with different, more expensive, TDR instruments. The general principle of microprocessor-controlled TDR operated soil moisture meter is considered.Patent pending     

基于人工神经网络的离心泵叶轮边界涡量流预测

根据边界涡量动力学理论,从边界涡量流在离心泵叶轮内表面的分布情况,可获知叶轮的受力状况,进而改进叶轮设计.以BP神经网络和径向基神经网络为建模手段,以叶轮内表面的边界涡量流为预测目标,通过高精度的CFD计算获得70个离心泵叶轮内表面的BVF分布,建立可用于训练人工神经网络的初始样本集;再利用63个初始样本建立离心泵叶轮几何参数和边界涡量流的非线性映射关系,并用剩余的7个校对样本进行测试.根据神经网络预测结果和数值模拟计算结果的误差分析,确定最适用于离心泵叶轮边界涡量流预测的神经网络类型.研究表明:径向基(RBF)神经网络的预测精度高于BP神经网络,其训练时间更短、运行稳定性更高;径向基函数的宽度对RBF神经网络的预测性能有较大影响,当径向基函数宽度取0.3时,RBF神经网络的预测性能最佳,预测误差仅0.020 3;RBF神经网络预测所得叶轮内表面的边界涡量流分布,可以作为评价叶轮水力设计优劣的重要指标,进而指导叶轮机械的优化设计.     

人工神经网络结合SWAT模型在河道径流量预测中的应用

径流量预测的常用方法具有不确定性大、未考虑大气变化及人类活动等因素影响的缺点,因此提出人工神经网络结合SWAT模型的预测方法来对其进行优化.应用该方法对辽宁省哈巴气水文站的降雨量及径流量进行模拟,预测结果与实测值吻合度较高,证明了该方法的合理性.此外,应用该方法对该站未来15年的径流量变化情况进行了预测,为该地区的水资源规划提供基础资料.     

Non-water-stressed baselines for irrigation scheduling with infrared thermometers: A new approach

Surface temperature measured with infrared thermometers is an important tool for irrigation scheduling which has been in practice for some decades. Several indices have been developed to time irrigation events. The most useful is the Crop Water Stress Index (CWSI). Its use, however, relies on a non-water-stressed baseline that, although having a theoretical basis, is to be determined experimentally given the uncertainties related to the surface resistance of the crop. The drawbacks of this procedure, besides the non-transferability of the lines from place to place, are that the surface temperature measurements have always to be made under similar weather conditions. A new definition of a non-water-stressed baseline theoretically based and driven by weather variables that can easily be measured and/or estimated is proposed that allows measurements at any time of the day and whatever the weather conditions, thus simplifying the task of the irrigator. Received: 19 April 1999     

Simulation of artificial neural network model for trunk sap flow of Pyrus pyrifolia and its comparison with multiple-linear regression

Trunk sap flow of tree is an important index in the irrigation decision of orchard. On the basis of the measured sap flow (SF) of pear tree (Pyrus pyrifolia) in the field, the multiple-linear regression for simulating the SF was obtained after analyzing the relationships between the SF and its affecting factors in this study and an artificial neural network (ANN) technique was applied to construct a nonlinear mapping to simulate the SF, then the simulated SF by two models was, respectively, compared to the measured value. Results showed that trunk SF had significant relationship with the vapour pressure deficit (VPD) in the single-variable analysis method but with soil volumetric water content (θ) using the ANN models with default of different variables. The correlation coefficient (R²), mean relative error (MRE) and root mean square error (RMSE) between the measured and simulated sap flows by the ANN model developed by taking VPD, solar radiation (S_r), air temperature (T), wind speed (W_s), θ, leaf area index (LAI) as the input variables were 0.953, 10.0% and 5.33 L d⁻¹, respectively, and the simulation precision of ANN model was superior to that of multiple-linear regression due to its better performance for the nonlinear relationship between trunk SF and its affecting factors, thus ANN model can simulate trunk sap flow and then may help the efficient water management of orchard.     

Comparison of artificial neural network models and empirical and semi-empirical equations for daily reference evapotranspiration estimation in the Basque Country (Northern Spain)

Reference evapotranspiration (ETo) determination is a key factor for water balance and irrigation scheduling. Evapotranspiration can be measured directly by high-cost micrometeorological techniques, or estimated by mathematical models. The combination equation of Penman–Monteith, modified by Allen et al. [Allen, R.G., Pereira, L.S., Raes, D., Smith, M., 1998. Crop evapotranspiration. Guidelines for computing crop water requirements. FAO Irrigation and Drainage, Paper no. 56. FAO, Rome] (PM56), is the reference equation for ETo estimation. This method is also appropriate for the calibration of other ETo estimation equations. The utilization of these calibrated ETo equations is recommended in the absence of data of any of the meteorological parameters necessary for the application of PM56. In addition to the use of classic ETo equations, the adoption of artificial neural network (ANN) models for the estimation of daily ETo has been evaluated in this study. ANNs are mathematical models, whose architecture has been inspired by biological neural networks. They are highly appropriate for the modelling of non-linear processes, which is the case of the evapotranspiration process. Seven ANNs (with different input combinations) have been implemented and compared with ten locally calibrated empirical and semi-empirical ETo equations and variants of these equations (with estimated meteorological parameters as inputs). The comparisons have been based on statistical error techniques, using PM56 daily ETo values as a reference. ANNs have obtained better results than the locally calibrated ETo equations in the three groups of evaluated methods: temperature and/or relative humidity-based methods (0.385 mm d⁻¹ of root mean square error (RMSE)), solar radiation-based methods (0.238 mm d⁻¹ of RMSE), and methods based on similar requirements to those of PM56 except for the estimation of solar radiation and/or relative humidity (0.285 mm d⁻¹ of RMSE).     

Estimation of water levels in a main drainage canal in a flat low-lying agricultural area using artificial neural network models

The Chiyoda basin is located in the Saga Prefecture of the Kyushu Island, Japan, and lies next to the tidal compartment of the Chikugo River, into which excess water in the basin is drained away. This basin has a total area of approximately 1100 ha and is a typical flat and low-lying agricultural area. The estimation of the water levels at the gates and along the main drainage canal is a crucial issue that has recently been the subject of much research. At these locations farmers and managers need to control the operation of the irrigation and drainage systems during periods of cultivation. An attempt has been made to apply a feed-forward artificial neural network (FFANN) to model and estimate the water levels in the main drainage canal. The study indicated that the artificial neural network (ANN) could successfully model the complex relationship between rainfall and water levels in this flat and low-lying agricultural area. Input variables and the model structure were selected and optimized by trial and error, and the accuracy of the model was then evaluated by comparing the simulated water levels with the observed ones during an irrigation period in July 2007. The water levels at two locations, located upstream and downstream of a main drainage canal, were investigated by using a time series at intervals of 20, 30, and 60 min. At these intervals, rainfall and tide water levels in the Chikugo River were measured, and the backward time-step numbers of the input variables of rainfall and tide water level were searched. For the upstream location, the optimal combination yielding good agreement between the observed and estimated water levels was obtained when the interval of the time series was 60 min. The number of backward time-steps of the input variables of rainfall and tide water level were 5 and 4, respectively. In contrast to the downstream location, the optimal combination was obtained for the interval time series of 20 min with 4 backward time-steps for both the input variables of rainfall and tide water level. The present study could provide farmers and managers with a useful tool for controlling water distribution in the drainage basin, and reduce the cost of installing water level observation points at many locations in the main drainage canal.     

Artificial neural networks approach in evapotranspiration modeling: a review

The use of artificial neural networks (ANNs) in estimation of evapotranspiration has received enormous interest in the present decade. Several methodologies have been reported in the literature to realize the ANN modeling of evapotranspiration process. The present review discusses these methodologies including ANN architecture development, selection of training algorithm, and performance criteria. The paper also discusses the future research needs in ANN modeling of evapotranspiration to establish this methodology as an alternative to the existing methods of evapotranspiration estimation.     

Artificial neural network for modeling reference evapotranspiration complex process in Sudano-Sahelian zone

The major problem when dealing with modeling evapotranspiration process is its nonlinear dynamic high complexity. Researchers developed reference evapotranspiration (ET-ref) estimation models in rich and poor data situations. Thus, the well-known Penman-Monteith (PM) model always performs the highest accuracy results of ET-ref from a rich data situation. Its application in many areas particularly in developing countries such as Burkina Faso has been limited by the unavailability of the enormous climatic data required. In such circumstances, simple empirical Hargreaves (HARG) equation is often used despite of its non-universal suitability. The present study assesses the artificial neural network (ANN) performance in ET-ref modeling based on temperature data in Bobo-Dioulasso region, located in the Sudano-Sahelian zone of Burkina Faso. The models of feed forward backpropagation neural network (BPNN) algorithm type ANN and Hargreaves (HARG) were employed to study their performance by comparing with the true PM. From the statistical results, BPNN temperature-based models perform better than HARG. Beside, when wind speed is introduced into the neural network models, the coefficient of determination (r²) increases significantly up to 9.52%. While, sunshine duration and relative humidity might cause only 3.51 and 6.69% of difference, respectively. Wind is found to be the most effective variable extremely required for modeling with high accuracy the nonlinear complex process of ET-ref in the Sudano-Sahelian zone of Burkina Faso.     

风力机故障诊断神经网络特征参数确定方法

为了诊断风力发电机组的故障,在搭建故障诊断神经网络平台时,选择合适的输入层特征参数搭建小波神经网络以达到网络训练时的稳定收敛.通过对风力发电机组故障诊断神经网络系统输入层特征参数的选择研究.发现风力发电机的齿轮箱、转子、叶片为独具代表性的易故障部件.分别对3个典型故障部件的一般故障类型和其产生机理进行了分析,得出齿轮箱的频率特性可以用来表征其故障类型,不同的转子故障会对应于不同的轴心轨迹,而叶片的故障诊断则可以运用声发射系统.根据分析的结果,提出了输入层特征参数的确定方法.齿轮箱按照其故障的时-频特性来确定输入层特征参数;转子利用其轴心轨迹能够反映故障类型的这一特性,来确定输入层特征参数;而风机的叶片则是通过“声发射系统”测量叶片表面性能时产生的特性数据作为输入层的特征参数.该方法可为风电机组故障诊断神经网络的建立提供参考.