Particle Size Distribution Determination from Spectral Extinction Using Neural Networks

Li, Mingzhong; Frette, Thor; Wilkinson, Derek

doi:10.1021/ie000826+

Cited by 10 publications

(8 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…with concentrations can be predicted in flight non-intrusively. Simulations and experiments have illustrated that it is feasible to use a neural network to obtain the parameters of a particle size distribution by training the neural network to decipher log normal particle data [27]. The method has an advantage of simplicity of use, instantaneous delivery of results and suitability for real time particle size analysis.…”

Section: Extension Of Technique To Water Aerosols and Silica Particlesmentioning

confidence: 99%

A Diagnostic Technique for Particle Characterization Using Laser Light Extinction

Barboza

Lowe

et al. 2015

Volume 6: Ceramics; Controls, Diagnostics and Instrumentation; Education; Manufacturing Materials and Metallurgy; Honors and Aw

View full text Add to dashboard Cite

Increased operations of aircraft, both commercial and military, in hostile desert environments have increased risks of micro-sized particle ingestion into engines. The probability of increased sand and dust ingestion results in increased life cycle costs, in addition to increased potential for performance loss. Thus, abilities to accurately characterize inlet sand would be useful for engine diagnostics and prognostic evaluation. Previous characterization studies were based on particle measurements performed a posteriori. Thus, there exists a need for in situ quantification of ingested particles.The work presented in this thesis describes initial developments of a line-of-sight optical technique to characterize ingested particles at concentrations similar to those experienced by aircraft in brownout conditions using light extinction with the end goal of producing an onboard aircraft diagnostic sensor. By measuring the extinct light intensity in presence of particles over range of concentrations, a relationship between diameters, concentration and light extinction was used for characterization. The particle size distribution was assumed log-normal and size range of interest 1-10 μm.To validate the technique, particle characterization in both static and flow based tests were performed on polystyrene latex spheres of sizes 1.32 μm, 3.9 μm, 5.1 μm, and 7 μm in monodisperse and poly-disperse mixtures. Results from the static experiments were obtained with a maximum relative error of 11%. Concentrations from the static experiments were obtained with a maximum relative error of 18%. Mono-dispersed and poly-dispersed particle samples were iii sized in a flow setup, with a maximum relative error of 12% and 10% respectively across all diameter samples tested. Uncertainty in measurements were quantified, with results indicating a maximum error of 17% in diameters due to sources of variability and showed that shorter wavelength lasers provide lower errors in concentration measurements, compared to longer wavelengths.For real time, on-board measurements, where path lengths traveled by light are much larger than distances traveled in initial proof of concept experimental setups, requirements would be to install sensitive detectors and powerful lasers to prevent operation near noise floors of detectors. Vibration effects from the engine can be mitigated by using larger area collection optics to ensure that the transmitted light falls on active detector areas.Results shown in this thesis point towards validity of the light extinction technique to provide real time characterization of ingested particles, and will serve as an impetus to carry out further research using this technique to characterize particles entering aircraft engine inlets.iv Dedicated to Nicole whose love, encouragement, support and honesty has made this possible for me.v

show abstract

Section: Extension Of Technique To Water Aerosols and Silica Particlesmentioning

confidence: 99%

A Diagnostic Technique for Particle Characterization Using Laser Light Extinction

Barboza

Lowe

et al. 2015

Volume 6: Ceramics; Controls, Diagnostics and Instrumentation; Education; Manufacturing Materials and Metallurgy; Honors and Aw

View full text Add to dashboard Cite

show abstract

“…Simulated and experimental results showed the ability of NNs for estimating the parameters of a log-normal PSD (i.e, the mean diameter and the standard deviation) from turbidity measurements [40]. An important limitation of the method is that the particle refractive index must be accurately known, or alternatively, well-characterized PSDs should be used to experimentally train the NN [40]. Finally, multi-level NNs with linear activation functions were used to estimate the average size, the aspect ratio, and the orientation of prolate spheroidal particles with equivalent radii in the range 3.00 · 10 -7 -1.50 · 10 -6 m (300-1500 nm) [42].…”

Section: Introductionmentioning

confidence: 99%

“…The method was recently validated on the basis of independent measurements by scanning electron microscopy [5]. Several neural network (NN) methods have also been applied for solving inverse problems in light scattering systems to estimate shape, size, and orientation of polymer particles [39][40][41][42][43][44]. For instance, radial basis function NNs proved adequate for simultaneously estimating the average radius and the particle refractive index in a system with homogeneous spheres on the basis of ELS measurements [39].…”

Section: Introductionmentioning

confidence: 99%

“…For instance, radial basis function NNs proved adequate for simultaneously estimating the average radius and the particle refractive index in a system with homogeneous spheres on the basis of ELS measurements [39]. Simulated and experimental results showed the ability of NNs for estimating the parameters of a log-normal PSD (i.e, the mean diameter and the standard deviation) from turbidity measurements [40]. An important limitation of the method is that the particle refractive index must be accurately known, or alternatively, well-characterized PSDs should be used to experimentally train the NN [40].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

A Neural Network Model for Estimating the Particle Size Distribution of Dilute Latex from Multiangle Dynamic Light Scattering Measurements

Gugliotta

Stegmayer

Clementi

et al. 2009

Part & Part Syst Charact

View full text Add to dashboard Cite

The particle size distribution (PSD) of dilute latex was estimated through a general regression neural network (GRNN) that was supplied with PSD average diameters derived from multiangle dynamic light scattering (MDLS) measurements. The GRNN was trained with a large set of measurements that were simulated from unimodal normal-logarithmic distributions representing the PSDs of polystyrene (PS) latexes. The proposed method was first tested through three simulated examples involving different PSD shapes, widths, and diameter ranges. Then the GRNN was employed to estimate the PSD of two PS samples; a latex standard of narrow PSD and known nominal diameter, and a latex synthesized in our laboratory. Both samples were also characterized through independent techniques (capillary hydrodynamic fractionation, transmission electron microscopy, and disc centrifugation). For comparison, all examples were solved by numerical inversion of MDLS measurements through a Tikhonov regularization technique. The PSDs estimated by the GRNN gave more accurate results than those obtained through other conventional techniques. The proposed method is a simple, effective, and robust tool for characterizing unimodal PSDs.

show abstract

“…Для преодоления указанных выше проблем в настоящее время разрабатываются перспективные подходы на основе методов машинного обучения и нейросетевых алгоритмов с обучением на синтетических данных [Ulanowski et al, 1998;Li et al, 2001;Guardani et al, 2002;Berdnik et al, 2004;Berdnik, Loiko, 2005;Deriemaeker, Finsy, 2005;Chicea, 2017]. Подобные методы были успешно применены для нахождения распределения частиц по размерам в случае монодисперсной среды с крупными частицами (500-1000 нм).…”

Section: Introductionunclassified

A multilayer neural network for determination of particle size distribution in Dynamic Light Scattering problem

Shabanov¹,

Petrov²,

Чикиткин³

2019

CRM

View full text Add to dashboard Cite

Solution of Dynamic Light Scattering problem makes it possible to determine particle size distribution (PSD) from the spectrum of the intensity of scattered light. As a result of experiment, an intensity curve is obtained. The experimentally obtained spectrum of intensity is compared with the theoretically expected spectrum, which is the Lorentzian line. The main task is to determine on the basis of these data the relative concentrations of particles of each class presented in the solution. The article presents a method for constructing and using a neural network trained on synthetic data to determine PSD in a solution in the range of 1-500 nm. The neural network has a fully connected layer of 60 neurons with the RELU activation function at the output, a layer of 45 neurons and the same activation function, a dropout layer and 2 layers with 15 and 1 neurons (network output). The article describes how the network has been trained and tested on synthetic and experimental data. On the synthetic data, the standard deviation metric (rmse) gave a value of 1.3157 nm. Experimental data were obtained for particle sizes of 200 nm, 400 nm and a solution with representatives of both sizes. The results of the neural network and the classical linear methods are compared. The disadvantages of the classical methods are that it is difficult to determine the degree of regularization: too much regularization leads to the particle size distribution curves are much smoothed out, and weak regularization gives oscillating curves and low reliability of the results. The paper shows that the neural network gives a good prediction for particles with a large size. For small sizes, the prediction is worse, but the error quickly decreases as the particle size increases.

show abstract

Particle Size Distribution Determination from Spectral Extinction Using Neural Networks

Cited by 10 publications

References 17 publications

A Diagnostic Technique for Particle Characterization Using Laser Light Extinction

A Diagnostic Technique for Particle Characterization Using Laser Light Extinction

A Neural Network Model for Estimating the Particle Size Distribution of Dilute Latex from Multiangle Dynamic Light Scattering Measurements

A multilayer neural network for determination of particle size distribution in Dynamic Light Scattering problem

Contact Info

Product

Resources

About