Local Epochs Inefficiency Caused by Device Heterogeneity in Federated Learning

Federated learning is a new framework of machine learning, it trains models locally on multiple clients and then uploads local models to the server for model aggregation iteratively until the model converges. In most cases, the local epochs of all clients are set to the same value in federated learning. In practice, the clients are usually heterogeneous, which leads to the inconsistent training speed of clients. The faster clients will remain idle for a long time to wait for the slower clients, which prolongs the model training time. As the time cost of clients’ local training can reflect the clients’ training speed, and it can be used to guide the dynamic setting of local epochs, we propose a method based on deep learning to predict the training time of models on heterogeneous clients. First, a neural network is designed to extract the influence of different model features on training time. Second, we propose a dimensionality reduction rule to extract the key features which have a great impact on training time based on the influence of model features. Finally, we use the key features extracted by the dimensionality reduction rule to train the time prediction model. Our experiments show that, compared with the current prediction method, our method reduces 30% of model features and 25% of training data for the convolutional layer, 20% of model features and 20% of training data for the dense layer, while maintaining the same level of prediction error.

Performance-Enhancing Sulfur-Doped TiO2 Photoanodes for Perovskite Solar Cells

High-performance electron transport layer (ETL) anode generally needs to form a uniform dense layer with suitable conduction band position and good electron transport properties. The TiO2 photoanode is primarily applied as the ETL because it is low-cost, has diverse thin-film preparation methods and has good chemical stability. However, pure TiO2 is not an ideal ETL because it lacks several important criteria, such as low conductivity and conduction band mismatch with compositional-tailored perovskite. Thus, TiO2 is an inefficient photo-anode or ETL for high-performance perovskite devices. In this study, sulfur as dopant in the TiO2 photo-anode thin film is used to fabricate solid-state planar perovskite solar cells in relatively high humidity (40–50%). The deposited S-doped thin film improves the power conversion efficiency (PCE) of the device to 6.0%, with the un-doped TiO2 producing a PCE of 5.1% in the best device. Improvement in PCE is due to lower recombination and higher photocurrent density, resulting in 18% increase in PCE (5.1–6.0%).

Plant disease prediction using convolutional neural network

Every year India losses the significant amount of annual crop yield due to unidentified plant diseases. The traditional method of disease detection is manual examination by either farmers or experts, which may be time-consuming and inaccurate. It is proving infeasible for many small and medium-sized farms around the world. To mitigate this issue, computer aided disease recognition model is proposed. It uses leaf image classification with the help of deep convolutional networks. In this paper, VGG16 and Resnet34 CNN was proposed to detect the plant disease. It has three processing steps namely feature extraction, downsizing image and classification. In CNN, the convolutional layer extracts the feature from plant image. The pooling layer downsizing the image. The disease classification was done in dense layer. The proposed model can recognize 38 differing types of plant diseases out of 14 different plants with the power to differentiate plant leaves from their surroundings. The performance of VGG16 and Resnet34 was compared. The accuracy, sensitivity and specificity was taken as performance Metrix. It helps to give personalized recommendations to the farmers based on soil features, temperature and humidity

High-temperature corrosion behavior of TP91, C22 alloy and C22 based on laser coatings in simulated reducing corrosion environment

To explore corrosion characteristics of TP91, C22 alloy and C22 laser coatings in reducing environment caused by low-nitrogen combustion at 500°C-600°C, a synthetic corrosive medium containing 0.2 vol. % H2S-0.1 vol. % O2-N2 were selected. Results showed that the order of corrosion resistance is: C22 laser coatings>C22 alloy>TP91. 550°C was the limit corrosion temperature for C22 alloy and C22 laser coatings. The reason for the strong corrosion resistance of C22 alloy and C22 laser coatings is that a dense layer of Cr-rich corrosion products due to the is produced in the inner layer, thus protecting the matrix from direct corrosion by corrosive gas. Finer grains before corrosion tests and the formation of dense Cr2O3 inner layers in much speedier process were the main reasons for the best corrosion resistance of C22 laser coatings.

Exogenous Application of Melatonin and Methyl Jasmonate as a Pre-Harvest Treatment Enhances Growth of Barhi Date Palm Trees, Prolongs Storability, and Maintains Quality of Their Fruits under Storage Conditions

Fresh date palm fruits (cv. Barhi) have received much attention due to their sweet taste and popularity in marketing. There is a critical need to prolong their storability, as well as maintain their quality during the postharvest and marketing periods. In this study, the effects of spraying date palm trees with melatonin (Mt) and/or methyl jasmonate (Mj) at 10, 20, and 50 ppm, on their growth and yield were investigated. In addition, impacts on quality and storability of the fruits were also studied. In general, application of Mt was mostly more effective than that of Mj, even at 50 ppm, with regard to all evaluated parameters. However, the dual treatment at 50 ppm recorded the highest relative chlorophyll and nutrient content in date palm leaves, as well as the yield and its components. Regarding the date palm fruits stored at 4 °C for 28 days, this dual treatment recorded the lowest weight loss and fruit decay values (0.14 and 2%, respectively), the highest firmness (6 g·cm−2), total soluble solids content (36 °Brix), total sugar content (32.5 g/100 g fresh weight), and the lowest total acidity (0.16 g citric acid/100 mL juice). Moreover, the highest total phenolic content and activity of peroxidase and polyphenol oxidase enzymes in the stored fruits were also recorded for the dual treatment. In contrast to the untreated fruits, scanning electron microscopy observations showed that the sprayed fruits had a very good microstructure, showing intact and thick exocarp tissue with a dense layer of epicuticular wax. The mesocarp tissue showed a normal and clear cellular framework with well organized and arranged cells, after 28 days storage at 4 °C. Based on these results, we can conclude that application of the dual treatment (Mt + Mj) at 50 ppm is a promising way to prolong the storability of date palm fruits and maintain their quality during storage periods.

Non-Solvent Influence of Hydrophobic Polymeric Layer Deposition on PVDF Hollow Fiber Membrane for CO2 Gas Absorption

The implementation of hydrophobicity on membranes is becoming crucial in current membrane technological development, especially in membrane gas absorption (MGA). In order to prevent membrane wetting, a polypropylene (PP) dense layer coating was deposited on a commercial poly(vinylidene fluoride) (PVDF) hollow fiber membrane as a method of enhancing surface hydrophobicity. The weight concentration of PP pellets was varied from 10 mg mL−1 to 40 mg mL−1 and dissolved in xylene. A two-step dip coating was implemented where the PVDF membrane was immersed in a non-solvent followed by a polymer coating solution. The effects of the modified membrane with the non-solvent methyl ethyl ketone (MEK) and without the non–solvent was investigated over all weight concentrations of the coating solution. The SEM investigation found that the modified membrane surface transfiguration formed microspherulites that intensified as PP concentration increased with and without MEK. To understand the coating formation further, the solvent–non-solvent compatibility with the polymer was also discussed in this study. The membrane characterizations on the porosity, the contact angle, and the FTIR spectra were also conducted in determining the polymer coating properties. Hydrophobic membrane was achieved up to 119.85° contact angle and peak porosity of 87.62% using MEK as the non-solvent 40 mg mL−1 PP concentration. The objective of the current manuscript was to test the hydrophobicity and wetting degree of the coating layer. Hence, physical absorption via the membrane contactor using CO2 as the feed gas was carried out. The maximum CO2 flux of 3.33 × 10−4 mol m−2 s−1 was achieved by 25 mg modified membrane at a fixed absorbent flow rate of 100 mL min−1 while 40 mg modified membrane showed better overall flux stability.

Preparation of Dual-Layered Core–Shell Fe3O4@SiO2 Nanoparticles and Their Properties of Plasmid DNA Purification

The rapid purification of biomaterials such as DNA, RNA, and antibodies has attracted extensive attention, and research interest has increased further with the COVID-19 pandemic. In particular, core–shell-structured superparamagnetic nanoparticles have been continuously studied for their application as biopurification materials. It has been reported that Fe3O4@SiO2 nanoparticles are one of the most promising candidates for separating nucleic acids via a simple and rapid process. This study proposed a fabrication method for dual-layered Fe3O4@SiO2 nanoparticles, in which the density of the SiO2 shell was controlled using an intermediate surfactant during the SiO2 coating. After the fabrication of dual-layered Fe3O4@SiO2 nanoparticles, structural, morphological, and magnetic analyses were conducted. The results showed that the Fe3O4 nanoparticles were surrounded by a dense layer 15.6~27.9 nm thick and a porous layer 24.2~44.4 nm thick, and had superparamagnetic properties with high saturated magnetization at room temperature (86.9 emu/g). Then, the optimal conditions for the biopurification material were suggested based on analysis of the selective separation of plasmid DNA.

Influence of Plasma Electrolytic Oxidation coating on corrosion characteristics of Friction Stir Welded ZE41 rare earth Magnesium alloy

Magnesium and its alloys are extensively used in the defence, aerospace and automotive industries, owing to their excellent mechanical properties. But there is a requirement of corrosion prevention treatment before using it for applications, since the corrosion resistance of Mg alloy is poor. In this work, a study on the mechanical properties of friction stir welded ZE41 rare earth magnesium alloy thick plates of 15mm and 25mm, was carried out. A metallographic examination and mechanical tests were conducted on both the welds. Furthermore, Plasma Electrolytic Oxidation (PEO), a surface conversion treatment was carried out on both the base material as well as on the friction stir butt joints. X-ray diffraction and scanning electron microscopy analysis were carried out on all the samples, with and without coating. The coated and uncoated samples were assessed by potentiodynamic polarisation (PDP) tests and electrochemical impedance spectroscopy (EIS) to study the electrochemical corrosion behaviour. An outer porous layer and an inner dense layer, have been identified from the SEM images. From the corrosion tests results it was revealed that there was an improvement in the corrosion resistance of both the base metal as well as the weldments by the PEO coating process.

Leveraging Large Subsurface Data and Associated Uncertainties to Build High Quality 3D Structural Model

Objectives/Scope At brownfield development stage, dealing with diverse and large amount of data makes it challenging to integrate them all in a consistent manner to build a prime structural model. Like many others, the studied field consists of several-stacked reservoirs featuring many faults and close to a thousand drilled wells with vertical, slanted and horizontal trajectories. On top of that, many horizontal wells are targeting thin carbonate layers for which tightly spaced data points often result in conflicting observations. Consequently, horizontal and deviated wells are commonly discarded from structural modelling, leaving substantial and valuable information unused. Some of these wells may be indirectly accounted through the introduction of pseudo-wells, making the modelling workflow tedious, user-dependent and therefore difficult to repeat. Methods, Procedures, Process ’It's better to be approximately right than exactly wrong’ quoted by Carveth Read, 18th century. Accordingly, every physical measurement, even from the most modern and sophisticated tools, is subject to some uncertainty. Therefore, assessing the uncertainty related to each input data is paramount in this method. Integrated teamwork between geologists, geophysicists and drilling specialists lead to a thorough analysis of each data feeding the process of structural model building while providing best uncertainty estimates. The ranges were specified for ∼1000 well trajectories, ∼16000 geological markers, 3 seismic travel time maps, 3 interval velocities and 59 thickness maps. All available data are used in a consistent manner to minimize the depth uncertainty. The accuracy is further improved by linking together all surfaces in a multi-layered model. In addition, this methodology considers both large scale spatial continuity capturing structural trends and more local scale incorporating inter-well variations of thickness due to sedimentological controls. Results, Observations, Conclusions After following this approach, all subsurface data started to come in agreement and resulted in more geological architectures. As an example, Figure 1 shows a cross-section along a well that drilled in B4 target layer which average thickness of 6 ft. As illustrated in the left figure, classical workflow using vertical wells and some pseudo-wells resulted in an anomalous pull-up structure and overall wavy non-geological geometry. Moreover, the well shows that it is in non-reservoir dense layer even though the well in the reservoir based on the zone log interpretation. However, the right figure shows that considering horizontal wells and uncertainties help to integrate all subsurface data with improved consistency where the structure model is smoother & more geological, plus the well is correctly placed in the targeted reservoir. Novel/Additive Information This approach will make the studied field one of the first brownfields that incorporated all data in consistent manner without pseudo-wells to build 3D structural model. It will bring considerable value to reduce uncertainties during subsequent property and dynamic modelling stages.

Microphysical process of precipitating hydrometeors from warm-front mid-level stratiform clouds revealed by ground-based lidar observations

Mid-level stratiform precipitations during the passage of warm fronts were detailedly observed on two occasions (light and moderate rain) by a 355 nm polarization lidar and water vapor Raman lidar, both equipped with waterproof transparent roof windows. The hours-long precipitation streaks shown in the lidar signal (X) and volume depolarization ratio (δv) reveal some ubiquitous features of the microphysical process of precipitating hydrometeors. We find that for the light-rain case precipitation that reaches the surface begins as ice-phase-dominant hydrometeors that fall out of a shallow liquid cloud layer at altitudes above the 0 ∘C isotherm level, and the depolarization ratio magnitude of falling hydrometeors increases from the liquid-water values (δv<0.09) to the ice/snow values (δv>0.20) during the first 100–200 m of their descent. Subsequently, the falling hydrometeors yield a dense layer with an ice/snow bright band occurring above and a liquid-water bright band occurring below (separated by a lidar dark band) as a result of crossing the 0 ∘C level. The ice/snow bright band might be a manifestation of local hydrometeor accumulation. Most falling raindrops shrink or vanish in the liquid-water bright band due to evaporation, whereas a few large raindrops fall out of the layer. We also find that a prominent δv peak (0.10–0.40) always occurs at an altitude of approximately 0.6 km when precipitation reaches the surface, reflecting the collision–coalescence growth of falling large raindrops and their subsequent spontaneous breakup. The microphysical process (at ice-bright-band altitudes and below) of moderate rain resembles that of the light-rain case, but more large-sized hydrometeors are involved.