Performance improvement of Deep Learning Models using image augmentation techniques

The major barrier while using deep learning models is lack of large number of images in the training dataset. In fact, there is a need of thousands of images in each image categories based on the complexity of problem. Prior studies have shown that picture augmentation techniques can be used to enhance the number of images in a training dataset artificially. These techniques can aid in improving the overall learning process and performance of a deep learning model. Hence, to address this problem we have proposed three algorithms. Firstly, two image acquisition algorithms have been proposed to systematically obtain real field images for testing and images from public datasets for training a model. Secondly, an algorithm is proposed to describe the procedure how the augmentations can be applied to enhance the datasets. During this study, we have investigated 52 augmentations that can allow enhancing the size of input dataset by improving the quantity of images. To perform the classification process of four maize crop diseases, a new convolutional neural network model is developed and several experiments have been performed to prove its effectiveness. Firstly, two tests were carried out using the original dataset from Kaggle public repository and the augmented dataset. When compared with the original dataset, the model improved by 5.14% with the augmented dataset. Secondly, three experiments carried out to evaluate the performance of proposed augmentation method. Experimental results demonstrated that the proposed approach outperforms the existing three approaches by 27.38%, 3.14%, and 1.34% during the classification process. The proposed IPA augmentation method has been compared with six existing methods: Full Stage Data Augmentation Framework, LeafGAN, Novel Augmentation method based on GAN, Wasserstein Generative Adversarial Network (WGAN), Activation Reconstruction-GAN, and Step-by-Step Data Augmentation Method and experimental results show that performance is better than existing methods by 28.31%, 19.76%, 20.18%, 13.75%, 2.42%, and 12.68% respectively.

     

Low-Temperature Synthesis of Calcium Hexa-Aluminate

Calcium hexa-aluminate (CaO·6Al₂O₃) has been prepared from calcium nitrate and aluminum sulfate solutions in the temperature range of 1000°–1400°C. A 0.3 mol/L solution of aluminum sulfate was prepared, and calcium nitrate was dissolved in it in a ratio that produced 6 mol of Al₂(SO₄)₃·16H₂O for each mole of Ca(NO₃)₂·4H₂O. It was dried over a hot magnetic stirrer at ∼70°C and fired at 1000°–1400°C for 30–360 min. The phases formed were determined by XRD. It was observed that CaO·Al₂O₃ and CaO·2Al₂O₃ were also formed as reaction intermediates in the reaction mix of CaO·6Al₂O₃. The kinetics of the formation of CaO·6Al₂O₃ have been studied using the phase-boundary-controlled equation 1 − (1 − x )^1/3= K log t and the Arrhenius plot. The activation energy for the low-temperature synthesis of CaO·6Al₂O₃ was 40 kJ/mol.     

Gravitational Search Based Algorithm for Optimal Coordination of Directional Overcurrent Relays Using User Defined Characteristic

Due to dependency on Plug Setting (PS), Time Dial Setting (TDS), size of the network, more than one back-up relays for one primary relay and other technical constraints, coordination of Directional Overcurrent Relays (DOCRs) is an extremely constrained and nonlinear optimization problem. In this paper, a new Gravitational Search (GS) based algorithm is presented for achieving optimal coordination of directional overcurrent relays. The proposed algorithm utilizes user defined characteristic for inverse time overcurrent relays than the predefined standard curves. The user defined relay characteristic deals with constants that control the shape of the characteristics as variable adjustable values which are optimally chosen along with TDS and Plug Setting Multiplier (PSM). The performance of the proposed algorithm has been evaluated on 8-bus, 15-bus and IEEE 30-bus distribution network at different fault locations (near-end, far-end and middle point). In addition, the time of operation of some of the primary relays at different fault locations on IEEE 30-bus distribution network is also presented. At the end, comparative evaluation of the proposed algorithm with other optimization algorithms having different relay characteristics presented in the literature clearly indicates its effectiveness and superiority in terms of the sum of total primary relays operating time.     

Experimental investigation of particle migration in suspension flow through bifurcating microchannels

Experimental measurements of velocity and concentration profiles were carried out to study transport of non‐colloidal suspension in bifurcating micro channels for both diverging and converging flow conditions using a combination of mirco‐particle image velocimetry and particle tracking velocimetry techniques. Migration of particles across the streamline was observed and symmetric velocity and concentration profile in the inlet branch becomes asymmetric in the daughter branches. Further migration of particles toward the center of the channel in the outlet branch make the profiles again symmetric. The evolution of velocity and concentration profiles was observed to be different in the symmetric and asymmetric bifurcation channels. The comparison of the streamlines for the fluid and the particles showed significant deviation near the bifurcation region. This may explain why there is unequal flow and particle partitioning during flow of suspension in asymmetric bifurcating channels as reported in many previous studies. © 2018 American Institute of Chemical Engineers AIChE J, 64: 2293–2307, 2018     

An electron microscopy investigation of the structure of porous silicon by oxide replication

The morphology of porous silicon is studied by scanning electron microscopy (SEM) by making an oxide replica of the pore structure. Highly branched n-type porous silicon samples were prepared and a replica was formed by oxidation of the pores followed by selective removal of the silicon substrate to expose the oxide pores. Scanning and transmission electron microscopy images confirmed many previously held assumptions about porous silicon formation, including the fractal structure and crystallographic propagation; they also provided a clearer understanding of the details of pore formation. The replica procedure also provides a platform for a more facile and comprehensive analysis of the porous silicon morphology.     

Thermal stresses in aluminum 6061 and nylon 66 long fiber thermoplastic (LFT) composite joint in a tailcone

The present paper involves a metal/polymer joint in a tailcone in a kinetic energy penetrator (KEP), one of the ammunition types used by the military. It is currently made of aluminum 7075 alloy, which could be partly replaced by long fiber thermoplastic (LFT) composite. Two different types of aluminum insert geometries were considered, viz., beaded and threaded. Thermal stresses set in during cooling of the tailcone from the processing temperature mainly because of the difference in the values of coefficients of thermal expansion and differential cooling between the aluminum and the LFT composite. Finite element (FE) modeling was done to predict the temperature profile during the cooling of the tailcone from the processing temperature. FE results showed that the LFT composite part of the tailcone cooled faster than the aluminum insert. Experimental verification of this temperature profile was obtained by infrared (IR) thermography. Based on the temperature profile, thermal stresses at the metal/LFT composite interface were estimated using an FE model. Different magnitudes of thermal stresses were present at the aluminum/LFT composite interface owing to the nature of distribution of fibers around the insert. Magnitude of thermal stresses in the case of a beaded insert was approximately 2.5 MPa whereas in the case of a threaded insert, it was approximately 12 MPa.

---

High-Frequency Fatigue Behavior of Woven-Fiber-Fabric-Reinforced Polymer-Derived Ceramic-Matrix Composites

The monotonic and high-frequency (100 Hz) fatigue behavior of two Nicalon-fabric-reinforced SiCON matrix composites was investigated at room temperature. The matrix composition was varied by the addition of BN and SiC particulate fillers, to contain shrinkage from processing by polymer infiltration and pyrolysis (PIP). The composites had strong fiber/matrix bonding, which resulted in substantially less frictional heating than observed with weakly bonded composites. Both composites exhibited fatigue runout at 10⁷ cycles at ∼80% of the monotonic strength. Comparison with existing fatigue data in the literature (for the same composites) at 1 Hz shows no change in fatigue life; i.e., no frequency effect was observed. Most of the stiffness reduction in the composites occurred in the first fatigue cycle, whereas subsequent decreases in moduli were attributed to limited fiber cracking. The major driving force for failure was the localized debonding of transverse and longitudinal plies at the crossover points in the fabric, which, when linked, resulted in interlaminar damage and failure in the composite.     

Some thermodynamic and kinetic properties of the system PETG-CO2, and morphological characteristics of the CO2-blown PETG foams

The solubility of CO₂ in PETG, a glycol-modified PET, was measured at different temperatures and over a broad pressure range, and diffusion coefficients were derived at the corresponding conditions. The solubility of CO₂ is quits high. For example, almost 15 wt% CO₂ can be dissolved in PETG at 35°C and 6.0 MPa. Consequently, CO₂ is good blowing agent for PETG. Cellular foams in the density range of about 0.04 to 1.2 g/cm³ and diameters in the range of about 10 to 150 µm were produced. The foam density and the cell size were found to depend on the foaming temperature and time, with larger cells obtained at higher temperatures or when the sample was foamed for a longer time. The foam density decreased with an increase in the foaming temperature to about 90°C, beyond which the density tended to increase slightly due to the cell collapse or coalescence. The density reduction also depended on the pressure at which the polymer was saturated with CO₂; the higher the saturating pressure at a given temperature, the greater the density reduction.     

The fracture toughness and fatigue behavior of DRA

‘In this study, the fracture toughness and elevated-temperature tensile and fatigue behavior of discontinuously reinforced aluminum (DRA) were examined. The effects of heat treatment and specimen thickness on fracture toughness were studied in a 7093/SiC_p composite. The toughness of the DRA was lowest in the peak-aged condition, but increased considerably in the overaged condition. The highest toughness was obtained at a critical value of specimen thickness; this critical value was used to fabricate a laminated composite consisting of alternating layers of DRA and unreinforced alloy. Elevated-temperature tensile and fatigue behaviors were investigated in a 2080-T6/SiC_p composite at different volume fractions and particle sizes. Increasing reinforcement volume fractions resulted in increases in tensile and fatigue strength. Exposure tests for 300 h at 150°C produced no significant reduction in ultimate tensile strength or yield strength, indicating good thermal stability of the 2080 matrix. For more information, contact A.B. Pandey, Air Force Research Laboratory, Materials and Manufacturing Directorate, Wright-Patterson AFB, Ohio 45433; (937) 255-1320; fax (937) 255-3007; e-mail awadh.pandey@ml.afrl.af.mil.     

Managing more physical with less virtual

Recent explosive growth in physical memory configurations inverts the problem faced by traditional operating systems, which were built around the assumption of limited physical memory and plentiful virtual memory. With the introduction of the Pentium Pro, there is an aberration in the historical trend towards increased virtual space. Such processors have 36‐bit physical and 32‐bit logical addressing; with traditional algorithms, the smaller virtual space limits the number of concurrent processes as well as the total physical memory usable by an application. To address these problems, two general principles can be applied to any operating system to reduce its virtual memory needs: first, multiplexing different physical addresses over the same virtual—as opposed to multiplexing different virtual over the same physical, where the optimization was to share the physical memory across multiple processes; second, optimize virtual consumption by kernel pagepool data, by organizing data structures to extend their physical memory reach using the same amount of virtual space. Copyright © 2000 John Wiley & Sons, Ltd.