Powders and their mixtures are elemental for many industries (e.g., food, pharmaceutical, mining, agricultural, and chemical). The properties of the manufactured products are often directly linked to ...the particle properties (e.g., particle size and shape distribution) of the utilized powder mixtures. The most straightforward approach to acquire information concerning these particle properties is image capturing. However, the analysis of the resulting images often requires manual labor and is therefore time-consuming and costly. Therefore, the work at hand evaluates the suitability of Mask R-CNN—one of the best-known deep learning architectures for object detection—for the fully automated image-based analysis of particle mixtures, by comparing it to a conventional, i.e., not machine learning-based, image analysis method, as well as the results of a trifold manual analysis. To avoid the need of a laborious manual annotation, the training data required by Mask R-CNN are produced via image synthesis. As an example for an industrially relevant particle mixture, endoscopic images from a fluid catalytic cracking reactor are used as a test case for the evaluation of the tested methods. According to the results of the evaluation, Mask R-CNN is a well-suited method for the fully automatic image-based analysis of particle mixtures. It allows for the detection and classification of particles with an accuracy of 42.7% for the utilized data, as well as the characterization of the particle shape. Also, it enables the measurement of the mixture component particle size distributions with errors (relative to the manual reference) as low as −2±5 for the geometric mean diameter and −6±5% for the geometric standard deviation of the dark particle class of the utilized data, as well as −8±4% for the geometric mean diameter and −6±2% for the geometric standard deviation of the light particle class of the utilized data. Source code, as well as training, validation, and test data publicly available.
This work is a first direct numerical simulation of a configuration closely related to the SpraySyn burner (Schneider et al. in Rev Sci Instrum 90:085108, 2019). This burner has been recently ...developed at the University of Duisburg-Essen to investigate experimentally nanoparticle synthesis in spray flames for a variety of materials. The present simulations are performed for ethanol and titanium tetraisopropoxide as a solvent and precursor, respectively, in order to produce titanium dioxide nanoparticles. In the direct numerical simulations, the complete scenario leading to the production of well-defined nanoparticles is taken into account, including evaporation of the liquid mixture (solvent and precursor) injected as a spray, multi-step kinetics for gas-phase combustion, and finally nanoparticle synthesis. The employed models are described in this article. Additionally, the impact of the inlet velocity of the pilot flame on the nanoparticle synthesis is investigated. It has been found that increasing this speed delays spray flame ignition, decreases nanoparticle concentration, but leads to a narrower size distribution at early stage.
The application of the Monte Carlo (MC) simulation technique for the modelling of nucleation processes with an existing background particle concentration is presented in this paper. Next to the ...nucleation of novel particles, the coagulation of an existing particle population as well as the condensational growth and evaporation of unstable particles (whose diameter is smaller than the critical Kelvin diameter) are included into the simulation. The usage of statistically weighted MC particles allows the description of particle size distribution (PSD), whose concentrations differ in several orders of magnitude. It is shown, that this approach allows to model the complex interplay between freshly nucleated particles and an existing background particle population. In this work, the nucleation of novel particles is modelled by three different nucleation theories discussed by Girshick, S. L. and C.-P. Chiu (1990), The Journal of Chemical Physics 93, which comprise of (1) the classical nucleation theory, (2) a mathematical correction to (1) and (3) a self-consistency correction of (2). For the chosen simulation conditions, the resulting PSDs are independent of the used nucleation theory for longer simulation times, in which the simulations are described by the coagulation mechanism only. The time-frame is identified for which relevant discrepancies of the PSDs have to be taken into account.
The SpraySyn burner is a new system recently developed at the University of Duisburg-Essen to investigate experimentally nanoparticle synthesis in spray flames for a variety of materials. The current ...project aims at performing direct numerical simulations with detailed physicochemical models of configurations closely related to this burner. The effect of using different solvents to produce titanium-dioxide (TiO
2
) nanoparticles is discussed in this work. The two solvents considered are o-xylene and ethanol mixed in liquid state with tetraisopropoxide to form TiO
2
. The liquid is injected into a pilot flame as dispersed spray with a carrier flow (dispersion gas). The resulting particle size distribution is examined as well. It is in particular observed that using ethanol leads to faster agglomeration and larger nanoparticles. This effect is qualitatively similar to that found when injecting smaller liquid spray droplets.
A population balance method based on weighted Monte-Carlo droplets is used to investigate breakup phenomena in single droplet combustion and spray flame synthesis (SFS). Particle shell formation in ...conjunction with superheating of liquid components is shown to be a plausible cause for droplet breakup. The breakage rate is calculated based on temperature and particle concentration profiles inside individual droplets, whereas the breakage function was determined using shadowgraphy imaging of droplets and image analysis. This enabled the simulation of consecutive breakup events in single droplet combustion and SFS. Furthermore, in the context of population balance simulations, an adaptive grid method is introduced for the calculation of particle accumulation at the droplet surface. This method is applicable to quickly evaporating droplets where particle transport by advection is fast compared to transport by radial diffusion. The new adaptive grid method is evaluated by comparison with seven single droplet combustion experiments from literature. More than that, parallel algorithms for a computationally efficient implementation using graphics processing units (GPUs) are discussed.
•Population balance modelling of droplet breakup.•Shadowgraphy imaging of droplet fragments.•Particle shell formation.•Spray combustion simulation.•Parallelization on GPU.
There is a high demand for fully automated methods for the analysis of primary particle size distributions of agglomerates on transmission electron microscopy images. Therefore, a novel method, based ...on the utilization of artificial neural networks, was proposed, implemented and validated.
The training of the artificial neural networks requires large quantities (up to several hundreds of thousands) of transmission electron microscopy images of agglomerates consisting of primary particles with known sizes. Since the manual evaluation of such large amounts of transmission electron microscopy images is not feasible, a synthesis of lifelike transmission electron microscopy images as training data was implemented.
The proposed method can compete with state-of-the-art automated imaging particle size methods like the Hough transformation, ultimate erosion and watershed transformation and is in some cases even able to outperform these methods. It is however still outperformed by the manual analysis.
Display omitted
•Artificial neural networks (ANNs) can help automate PSD measurements of TEM images.•ANNs can compete with and even outperform other automated methods (e.g. watershed).•Image synthesis can produce lifelike TEM images, suitable for ANN training.
Two-component aggregative mixing of initially bidisperse particle populations results in a Gaussian-type compositional distribution function, which can be fully described by the overall mass fraction ...of component A (ϕ) and the mass-normalized power density of excess component A (χ, indicating the mixing degree). χ will reach a steady-state value once the self-preserving size distribution is attained. The relation between the steady-state value of χ (χ∞) and its initial value (χ0) has not been investigated before. This paper applies population balance modeling to gain insight into the dependence of χ∞ on initial feeding conditions. By model fitting from hundreds of systematically varied simulations, it is found that χ∞/χ0, which depends on ϕ, can be formalized by the Gaussian-type function for Brownian aggregation in the free-molecular regime as well as in the continuum regime, however with different geometric standard deviations. The present work can help to optimize mixing by properly selecting the initial mass and number concentrations of components A and B in the feed.
•TiN nanoparticles are in-situ injected into a growing CrN thin film.•The approach enables the synthesis of an artificial nc-CrN/nc-TiN nanocomposite.•The embedded nanoparticle content is influenced ...by the bias supply mode and value.•The bias-voltage can be used to attract selected nanoparticle sizes.•The injected nanoparticles act supplementary to the effect of the bias-voltage.
The formation of a nanocomposite structure is thermodynamically driven by spinodal decomposition in at least two phases. With respect to the suppression of solid solution formation, artificial nc-TiN/nc-CrN composites were deposited using a novel hybrid-process, in which TiN nanoparticles and CrN thin film were separately synthesized and simultaneously deposited during composite growth. The bias-voltage is known as a crucial deposition parameter concerning the structural and mechanical properties in thin film technology. However, it is still unclear whether an externally injected nanoparticle jet is influenced by the bias-voltage applied to the substrate. In this work, composite thin films were DC sputtered applying bias-voltages of 0 V, -100 V and -200 V in DC mode, as well as -100 V in MF and HiPIMS mode. TEM-investigations reveal the successful embedment of the nanoparticles in the film. Growth defects in the interface between nanoparticle and thin film can be reduced using a pulsed bias-voltage. Based on 2D GI-XRD experiments using synchrotron radiation, a bias-voltage of -200 V DC and -100 V MF enables the reinforcement of a higher nanoparticle content in the thin films. Similar to an increased bias-voltage, the injection of nanoparticles results in a decrease of the crystallite size. In principle, the residual stresses are increased by the nanoparticle embedding, as is the case for an increasing bias-voltage. In the event of a pulsed-bias voltage, however, the residual stresses can be reduced by the embedding of the nanoparticles. The mechanical properties of the CrN thin films can be maintained when nanoparticles are injected.
Display omitted