Biosystems Engineering最新文献

Assessing the feather condition of broilers is crucial for monitoring the animal welfare status and detecting the occurrence of feather pecking activities. Currently, the feather condition of individual broilers is manually scored by trained experts. To provide a more objective and efficient tool for feather condition scoring, a novel deep learning-based model, named Feather Condition Scoring Network (FCS-Net), was proposed based on RGB and thermal infrared images. The FCS-Net model combined the ResNet18 architecture with the proposed Dense Feature Fusion (DFF) module, which can effectively learn the feature mapping relationship between RGB and thermal infrared images. Before inputting the images into the network, an image registration process was conducted to align the RGB and thermal infrared images. The results showed that the FCS-Net model had a good performance for feather condition scoring, with the Accuracy of 97.02%, the Precision of 96.99%, the Recall of 97.04%, the F₁ of 97.01%, and the Inference speed of 15.34 fps. Compared to the ResNet18_RGB model, which only utilise RGB images, the FCS-Net model showed notable improvements in Accuracy by 4.02%, Precision by 3.90%, Recall by 4.08%, and F₁ by 4.01%. Moreover, it was observed that the FCS-Net model focused more on the back region of the broilers through heatmap visualization. Furthermore, the algorithm was compared with six typical image recognition algorithms including VGG16, ResNet18, SE-ResNet18, DenseNet121, Mobilenet_V2, and Shufflenet_V2_x1_0, as well as the state-of-the-art (SOTA) feather condition assessment methods. The results showed that the FCS-Net model achieved better performance than the six algorithms and the SOTA feather condition assessment methods. This study provided a valuable reference for automated monitoring of feather condition scoring of broilers in smart farming.

Sowing depth is a critical factor in crop growth and is determined by both the soil conditions and the force of the opener. The trend for the future is to control sowing depth based on soil dynamic parameters. Therefore, this paper developed a downforce measurement and control system based on the ‘T’-shaped furrow opener and investigated the influence of soil dynamic parameters and opener downforce on sowing depth. A test-rig was constructed and the accuracy of the system in measuring downforce and controlling downforce and sowing depth was verified. The study shows that at different sowing depths, soil moisture, bulk density and their interaction have a significant effect on downforce (P < 0.01). As the moisture content decreases and the bulk density increases, the required downforce increases for the same sowing depth. A mathematical model of downforce-sowing depth-soil bulk density-soil moisture content was established using experimental data, with an R² of 0.916, VIF <5 and a Durbin-Watson value of 1.628. Field experiments show that, at an operating speed of 6 km h⁻¹, the control strategy based on the soil dynamic parameters predicted by downforce theory significantly outperformed the strategy of adjusting the downforce in response to perceived changes in downforce. This indicates that after dynamic and rapid measurement of soil bulk density and moisture content during field operations, sowing depth can be accurately controlled based on the directed downforce of the opener. The mathematical model provides a theoretical basis for sowing depth control based on soil dynamic parameters.

Sensor technologies were integrated into a commercial sensor-guided hoeing system to counteract the force of gravity and reduce crop damage caused by the offset of hoeing in maize fields on sloping terrains. For this study, a hoe was equipped with a contact disc, sensors, an electric cylinder, and a decision support system. The offset of the hoe could be compensated in real time based on the automatic adjustment angle of the support wheel. In maize, three field experiments were conducted over two years to evaluate the system on three different slope gradients (between 4 and 12°). Plant populations were measured in each plot one day before and during hoeing to evaluate crop damage. However, for support wheel angle, Slope Compensation Intensity (SCI) 2 and 3, there were no significant crop plant losses in any trials. As a result, there was no hoe drifting during the sensor-based guidance along the rows. It has been verified that the development presented is functional and can counteract the force of gravity on slopes. This development aims to optimise the use of precision mechanical weed control and support farmers during hoeing on hilly terrain.

Pressurised sand filters used in drip irrigation need periodic backwashing to flush the contaminant particles out of the porous media. This process consumes high amounts of energy and water. The selection of more efficient backwashing operational conditions requires accurate information of the pressure drop and the bed expansion, the latter being not measured in commercial filters. An experimental study with a scaled filter that used a window to observe the bed expansion was conducted with three porous media types (glass microspheres and two silica sands), two packed media bed heights (200 mm and 300 mm) and four nozzles (one commercial and three prototypes). The 24 combinations of the filter experimental configuration were investigated for different superficial velocities. Both data and video recordings for all the 705 tests conducted were carefully analysed to obtain mean values and standard deviations of the height of the expanded bed. The behaviour of the fluidised bed dynamics was characterised. Results indicated that the nozzle design had a strong influence on the pressure drop, and, in consequence, on the power required for backwashing. It also had an observable impact on the fluidised bed dynamics although its effect on determining the overall height of the expanded bed was limited, this being more dependent on the type of the porous media. The most effective combination in terms of energy efficiency and porosity of the expanded bed was obtained with microspheres, though its retention efficiency might be questionable from the literature review, and the frustoconical nozzle geometry.

In the liquorice-soil composite shear test, flexible thick roots bend to create soil resistance which makes measurement results inaccurate. However, the liquorice pulling force characterises the contact strength of the liquorice-soil composite and can be used to study the root-soil interactions. This paper proposed a three-part modelling method to model the liquorice-soil composite at harvesting period. The mechanical parameters of soil particles were calibrated using the soil unconfined compressive strength test. The calibration results showed that the errors of peak force and peak displacement for soil unconfined compressive strength tests were 1.09% and 1.64%, respectively. The flexible liquorice model was constructed based on 3D scanning and particle filling methods, and the simulation model was calibrated based on compression properties. The relative errors in calibration of the flexible liquorice's radial and axial compression forces were 1.35% and 3.9%, respectively. Simplifying liquorice pulling force and liquorice surface area into a linear correlation effectively supports the general modelling method. The contact parameters between soil and liquorice were determined using liquorice pulling force as the target value, and the proportional calibration method was used to improve simulation efficiency. The calibration error for the liquorice pulling force is 4.39%. In addition, the results of the pulling force for the different surface areas show that the calibrated parameters are valid within a liquorice surface area of 0.0075–0.0181 m². This study provided a general and accurate simulation method to the liquorice-soil composite, which can be used as the reference for modelling the long root-soil composite, and provide methodological support for developing root crop harvesters.

Investigations into the use of urease inhibitors for reducing ammonia emission in dairy farming have been published in several papers. The aim of this study is to expand the existing knowledge on the use of urease inhibitors for reducing ammonia emissions in fattening pig houses. In this respect, in addition to the proven standard application approach using a backpack sprayer, the investigation was extended to include different application techniques.

Urease inhibitor was applied on two farms over six experimental periods throughout the year using three different application techniques: a backpack sprayer, and a semi-automatic system that applies the inhibitor both on-floor and under-floor. Two identical compartments, alternated between treatment and control, were used on each farm. A linear mixed model with repeated measurements was used to quantify the reduction effect of the urease inhibitor.

The use of the backpack sprayer led to a reduction in ammonia emissions of 22.9% (standard error, SE: 4.9%). The on-floor application system reduced the emissions by 16.6% (SE: 4.9%), and the under-floor application system resulted in no significant reduction.

The development of the semi-automatic application system can be considered beneficial for reducing emissions. However, further development and improvement of this application system is necessary for its widespread practical use, especially regarding the distribution accuracy of the application liquid, contamination issues, and the manual workload. In addition, the effects of the presence of the animals during the application process need to be investigated in more detail.

In the food packaging industry, ventilated corrugated paperboard boxes are crucial for sustainable transport of fresh products. While these boxes' ventilation holes advance air circulation, they also impact the material's compression or buckling strength. Variations in hole geometry and location affecting this strength are explored, considering the composite material, multi-layered structure. Traditional mechanical analyses, which often require simplifications, may not fully capture this complexity, leading to less accurate predictions of the paperboard's strength. To address these challenges, a machine learning (ML) approach was utilized, employing the Light Gradient Boosting Machine (LGBM) to develop a predictive model for the buckling strength of corrugated paperboard boxes with ventilation cutouts. This physics-informed ML model, trained on a compression dataset resulting from experimental tests for plates with a single cutout in three shapes and Finite Element Method (FEM) simulations for plates with various patterns of circular cutouts, provides highly accurate estimates of the plates' buckling strength. It achieved 91.7% accuracy on experimental data and 94.68% on FEM simulation data, showcasing its reliability. A new tool for predicting the buckling strength of corrugated paperboard is provided by this research, along with insights that can inform the design of more sustainable packaging solutions. Furthermore, the methodology and findings have broader applications, potentially benefiting sectors like aerospace and construction, where similar structural materials are used.

It is important to develop low-cost, fast and portable meat adulteration detection systems to ensure the meat authenticity and safety in complex market environments. A multi-channel spectral detection system for meat adulteration was developed in this study. The core hardware of the system mainly includes a designed spectral module and a Raspberry pi controller. The spectral module consists of three multi-channel spectral sensors and LED lamps with specific wavelengths, containing 18 channels covering a range of 410–940 nm. The software was developed based on PyQt5. After completing the construction of the system, the detection distance was discussed and determined to be 4 mm. Based on the spectral data collected by the developed system, the models for classifying pure mutton, pure pork, mutton flavour essence adulteration, colourant adulteration and adulterated mutton with pork were established and compared. Four intelligent optimisation algorithms were further used to improve the model performance. The results of the test set showed that the support vector classification (SVC) model optimised by a sparrow search algorithm (SSA) obtained the best classification performance, with an accuracy of 97.59% and a Kappa coefficient of 0.9696. After the SSA-SVC was incorporated into the sensor software, the system performance was evaluated using external validation samples. The overall accuracy of the system was 94.29%. The system took about 5.31 s to detect a sample, and the total weight of the system was 1.55 kg. Overall, the developed portable spectral system has considerable potential to rapidly and accurately discriminate adulterated mutton in the field.

Quantifying seedling emergence pressure or forces (soil impedance to seedling) during the process of plant emergence is difficult in a practical setting. In this study, a mechanical seedling testing device was designed and calibrated to measure seedling emergence pressures experienced by conical or spherical mechanical seedling in soil with varying compaction levels. The data were analysed to generate regression models for predicting seedling emergence forces. Results showed a high correlation between the seedling emergence pressure and soil resistance. The resultant regression model produced a coefficient of determination (R²) of 0.99. After incorporating the morphological characteristics of soybean cotyledon and maize coleoptile into the model, the predicted seedling emergence forces increased with the soil compaction level. During the emergence process, average emergence force of the soybean seedlings was 11.8 N for the lowest compaction level and 28.5 N for the highest compaction level, and the corresponding values of the maize seedlings were 0.2 N and 0.6 N. In a non-compacted field plot, maize crop had a 95.4% emergence rate and soybean crop had 97.2%, whereas for a compacted plot, the corresponding emergence rates were decreased to 19.1% and 60.5%. Inferences made from the study provide information on the dynamics of soil-seedling interaction, which have important implications for managing soil compaction in crop production.

In-situ leaf chlorophyll content (LCC) estimation based on hyperspectral imaging (HSI) is crucial to track the growth status of crops for field management. However, spatial and spectral features of HSI data, suffering from interference of growth dynamic effect and soil, pose the challenge on accuracy and robustness of LCC estimation in several years and growth stages. Therefore, a joint spectral-spatial feature extraction method was proposed by cascade of three-dimensional convolutional neural network (3DCNN) and long short-term memory (LSTM) to reduce the interference for optimising the LCC estimation. Firstly, crop pixels were separated from soil with vegetation index segmentation method. Secondly, when raw images and segmented pixels were input, sensitive bands were selected by random frog (RF bands), and 3DCNN-LSTM was used to extract the joint spectral-spatial features. Finally, models established by RF bands, 3DCNN and 3DCNN-LSTM were compared, and robustness in individual years and stages was validated. Results showed that RF bands and 3DCNN obtained R_P² of 0.76 and 0.84 when not segmented. After segmentation, performance of 3DCNN improved (R_P² = 0.85) compared to RF bands (R_P² = 0.80). Spectral-spatial features by 3DCNN reduced the interference of soil. 3DCNN-LSTM without and with segmentation obtained good performance with R_P² of 0.95 and 0.96, and the proposed method could reduce the image segmentation process. The optimal model achieved R_P² above 0.93 in individual years (R_P² = 0.96 in 2021, R_P² = 0.94 in 2021) and R_P² in the range of 0.87–0.97 at individual stages. This paper provides a method to track growth variability between soil and crop for the LCC estimation optimisation.