Characterizing soot in TEM images using a convolutional neural network

Sipkens, Timothy A.; Frei, Max; Baldelli, Alberto; Kirchen, Patrick; Kruis, Frank Einar; Rogak, Steven N.

doi:10.1016/j.powtec.2021.04.026

Cited by 18 publications

(8 citation statements)

References 35 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This produces a fully segmented image and a trained classifier, 24 which can subsequently be used to classify further images. Trainable segmentation is typically more accurate than thresholding methods, as we will demonstrate in our later results, but less accurate than neural networks 25 . However, compared to neural networks, the samples of training data needed for accurate classification is considerably smaller in trainable segmentation.…”

Section: Introductionmentioning

confidence: 67%

“…Trainable segmentation is typically more accurate than thresholding methods, as we will demonstrate in our later results, but less accurate than neural networks. 25 However, compared to neural networks, the samples of training data needed for accurate classification is considerably smaller in trainable segmentation. The time to train the classifier is also shorter than the time needed to train a neural network, by several orders of magnitude.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Trainable segmentation for transmission electron microscope images of inorganic nanoparticles

Bell

Treder

Kim

et al. 2022

Journal of Microscopy

View full text Add to dashboard Cite

We present a trainable segmentation method implemented within the python package ParticleSpy. The method takes user labelled pixels, which are used to train a classifier and segment images of inorganic nanoparticles from transmission electron microscope images. This implementation is based on the trainable Waikato Environment for Knowledge Analysis (WEKA) segmentation, but is written in python, allowing a large degree of flexibility and meaning it can be easily expanded using other python packages. We find that trainable segmentation offers better accuracy than global or local thresholding methods and requires as few as 100 user-labelled pixels to produce an accurate segmentation. Trainable segmentation presents a balance of accuracy and training time between global/local thresholding and neural networks, when used on transmission electron microscope images of nanoparticles. We also quantitatively investigate the effectiveness of the components of trainable segmentation, its filter kernels and classifiers, in order to demonstrate the use cases for the different filter kernels in ParticleSpy and the most accurate classifiers for different data types. A set of filter kernels is identified that are effective in distinguishing particles from background but that retain dissimilar features. In terms of classifiers, we find that different classifiers perform optimally for different image contrast; specifically, a random forest classifier performs best for high-contrast ADF images, but that QDA and Gaussian Naïve Bayes classifiers perform better for low-contrast TEM images.

show abstract

Section: Introductionmentioning

confidence: 67%

Section: Introductionmentioning

confidence: 99%

Trainable segmentation for transmission electron microscope images of inorganic nanoparticles

Bell

Treder

Kim

et al. 2022

Journal of Microscopy

View full text Add to dashboard Cite

show abstract

“…These correlations can be used then to identify or estimate the critical aspects of the process under consideration (e.g., by performing feature selection). A number of studies have successfully applied ML algorithms to predict soot output in engines and burners [29,30,31], to assist with experimental soot measurement procedures [32,33], and to classify soot in TEM images [34]. A pair of recent studies by Dworkin and co-workers [35,36] used artificial neural networks (ANN), a class of ML algorithms, to predict the soot volume fraction (f v ) in laminar coflow diffusion flames.…”

Section: Introductionmentioning

confidence: 99%

Exploring soot inception rate with stochastic modelling and machine learning

Liddo

Saldinger

Jadidi

et al. 2023

Combustion and Flame

View full text Add to dashboard Cite

“…The diameter of primary soot particles, the diameter of soot particles, fractal dimension, etc., can be measured directly from the TEM and HRTEM images using an image analysis algorithm [15]. Besides TEM and HRTEM, atomic force microscopy (AFM) [16,17] and helium ion microscopy (HIM) [18] are also used for nanoscale study of soot particles and suggest that freshly formed soot particles exhibit liquid-like features and contain structural inhomogeneity [19].…”

Section: Introductionmentioning

confidence: 99%

“…Soot particles' nuclei have two types of structures: cross-linked structures [22] and fullerene-like structures [23]. The freshly formed nucleation of soot particles exhibit liquid-like structures [18,24]. These particles undergo coagulation, agglomeration and surface growth leading to particle size and mass increase, that is, immature primary particles were carbonized into the larger soot aggregates with the more matured structure [25].…”

Section: Introductionmentioning

confidence: 99%

Effect of Temperature on Morphologies and Microstructures of Soot Particles in the Diesel Exhaust Pipe

et al. 2023

View full text Add to dashboard Cite

Insulating cotton was used to change the airflow temperature in the exhaust pipe of a diesel engine, and soot particles at different positions in the exhaust pipe under different operating conditions were collected. The morphologies and microstructures of soot particles were observed by high-resolution transmission electron microscopy (HRTEM). The characteristic parameters, including the mean primary particle diameter (dp), radius of gyration of soot aggregate (Rg), fractal dimension of soot particle (Df), carbon layer spacing (Ds), and carbon layer torsion resistance (Tf), were statistically analyzed. The changes in each characteristic parameter before and after adding insulating cotton were compared. After installing the cotton, soot particles still grew through surface chemical reactions and physical processes in the diesel exhaust pipe, the agglomeration becomes more and more prevalent, the particle size increased, and Df increased. The increase in the airflow temperature in the exhaust pipe promoted the surface growth of primary soot particles and enhanced the turbulence, which made the chain-like soot particles more likely to reunite under the action of turbulent eddies. Consequently, Rg decreased and Df increased. Furthermore, the average Ds and Tf of primary soot particles deceased, especially under high loads. This indicated that the increase in the temperature of the exhaust pipe was conducive to the graphitization of primary soot particles.

show abstract

Characterizing soot in TEM images using a convolutional neural network

Cited by 18 publications

References 35 publications

Trainable segmentation for transmission electron microscope images of inorganic nanoparticles

Trainable segmentation for transmission electron microscope images of inorganic nanoparticles

Exploring soot inception rate with stochastic modelling and machine learning

Effect of Temperature on Morphologies and Microstructures of Soot Particles in the Diesel Exhaust Pipe

Contact Info

Product

Resources

About