为了解决联合收割机作业故障的非线性特征信号难以提取的问题,该研究提出了一种基于堆叠去噪自动编码器(Stack Denoising Auto Encoder, SDAE)和BP神经网络(Back Propagation,BP)融合的联合收割机作业故障监测及诊断的方法(SDAE-BP)。以转速传感器采集联合收割机脱粒滚筒转速、籽粒搅龙转速、喂入搅龙转速、杂余搅龙转速、风机转速、输送链耙转速、割刀频率以及逐稿器振动频率,并将采集的数据集作为系统的输入。利用SDAE提取输入信号的深层次特征,并由BP神经网络辨识收割机作业状态,实现联合收割机故障监测。在SDAE-BP模型训练过程中,去噪自动编码器(Denoising Auto Encode, DAE)依次经带有不同分布中心噪声的原始数据进行训练,然后将其堆叠,并通过误差反向传播算法对模型参数进行优化,以提升模型识别故障性能和泛化能力。试验结果表明,对于2018年联合收割机田间试验数据,模型的故障诊断准确率达到99.00%,与SDAE和BP神经网络相比,分别提高了1.5和4.5个百分点。将SDAE-BP故障诊断模型用2019年的试验数据进行更新,并用2018年和2019年试验数据进行测试,结果表明,更新后的模型对2018年试验数据的故障识别准确率为99.25%,对2019年试验数据的故障识别准确率为98.74%,更新后模型在2019试验数据集上的故障识别准确率较未更新模型提高了6.52个百分点。该文所建模型能够准确识别联合收割机的故障类型,且具有较好的鲁棒性,对旋转型机械故障监测及预警具有参考价值。 相似文献
Long- and short-term direct-drilling and seed broadcasting plus rotovation were examined as possible quick and cheap alternatives to conventional mouldboard ploughing and drilling. The experiment was the continuation of an existing long-term tillage experiment for spring barley. The conventional ploughing and long-term direct-drilling treatments continued on the same plots. The broadcasting and the short-term direct-drilling treatments were applied to previously chisel-ploughed and deep mouldboard-ploughed treatments, respectively. Autumn nitrogen treatments of 30 or 60 kg ha−1 and spring nitrogen treatments of 150 or 225 kg ha−1 were applied. The experiment is located on a cambisol (15% clay in topsoil) and on a gleysol (17% clay in topsoil) in south-east Scotland. Straw was removed by baling and the stubble remained when the treatments were applied.
Long-term direct-drilling yielded most over the 3 years of the experiment and was particularly successful after the unusually wet autumn and winter of the third season. Short-term direct-drilling was the lowest yielding treatment in the first season only. The success of the long-term direct-drilling treatment was associated with the development of a stable, protective surface tilth as a result of organic-matter accumulation. This was associated with some soil structural improvement deeper in the profile in the long-term direct-drilled gleysol as shown by measurements of air permeability. Crop performance apparently was not related to soil compactness or cone resistance. The relatively high rates of nitrogen applied, both autumn and spring, gave worthwhile crop responses except for the third season, when the crop lodged. Short-term direct-drilling, broadcasting with rotovation and ploughing with drilling gave similar average yields on the gleysol, but on the cambisol broadcasting with rotovation outyielded the other two treatments by an average of 0.3 t ha−1. 相似文献
Journal of Soils and Sediments - After the greatest environmental disaster in the history of Brazil and the deposition of the iron ore tailings in alluvial regions, the process of revegetation for... 相似文献
The increasing demand for biomass for food, animal feed, fibre and bioenergy requires optimization of soil productivity, while at the same time, protecting other soil functions such as nutrient cycling and buffering, carbon storage, habitat for biological activity and water filter and storage. Therefore, one of the main challenges for sustainable agriculture is to produce high yields while maintaining all the other soil functions. Mechanistic simulation models are an essential tool to fully understand and predict the complex interactions between physical, biological and chemical processes of soils that generate those functions. We developed a soil model to simulate the impact of various agricultural management options and climate change on soil functions by integrating the relevant processes mechanistically and in a systemic way. As a special feature, we include the dynamics of soil structure induced by tillage and biological activity, which is especially relevant in arable soils. The model operates on a 1D soil profile consisting of a number of discrete layers with dynamic thickness. We demonstrate the model performance by simulating crop growth, root growth, nutrient and water uptake, nitrogen cycling, soil organic matter turnover, microbial activity, water distribution and soil structure dynamics in a long-term field experiment including different crops and different types and levels of fertilization. The model is able to capture essential features that are measured regularly including crop yield, soil organic carbon, and soil nitrogen. In this way, the plausibility of the implemented processes and their interactions is confirmed. Furthermore, we present the results of explorative simulations comparing scenarios with and without tillage events to analyse the effect of soil structure on soil functions. Since the model is process-based, we are confident that the model can also be used to predict quantities that have not been measured or to estimate the effect of management measures and climate states not yet been observed. The model thus has the potential to predict the site-specific impact of management decisions on soil functions, which is of great importance for the development of a sustainable agriculture that is currently also on the agenda of the ‘Green Deal’ at the European level. 相似文献