Retrieval of size and refractive index of spherical particles by multiangle light scattering: neural network method application

Real-time detection method for bulk bubbles in optics based on deep learning

The existence of bulk bubbles could decrease the laser-induced damage threshold of optics and affect the beam quality, so the detection of bulk bubbles is an essential step for quality assurance. Currently, the inspection of bubbles in optics relies on manual work, which is not recommended because of the low precision and inconsistency. To improve the quality evaluation process, a real-time detection method for bubbles inside the optics based on deep learning is proposed. Our method can implement bubble detection at 67 fps with a recall of 0.836. As for retrieval of the radius, it costs 58.8 ms on each bubble, and the absolute deviation is 3.73% on average. Our method conducts real-time and accurate detection of the positions and radii of the bubbles in the optics, thus, having significant potential for the manufacturing process.

An Artificial Neural Network Assisted Dynamic Light Scattering Procedure for Assessing Living Cells Size in Suspension

Dynamic light scattering (DLS) is an essential technique used for assessing the size of the particles in suspension, covering the range from nanometers to microns. Although it has been very well established for quite some time, improvement can still be brought in simplifying the experimental setup and in employing an easier to use data processing procedure for the acquired time-series. A DLS time series processing procedure based on an artificial neural network is presented with details regarding the design, training procedure and error analysis, working over an extended particle size range. The procedure proved to be much faster regarding time-series processing and easier to use than fitting a function to the experimental data using a minimization algorithm. Results of monitoring the long-time variation of the size of the Saccharomyces cerevisiae during fermentation are presented, including the 10 h between dissolving from the solid form and the start of multiplication, as an application of the proposed procedure. The results indicate that the procedure can be used to identify the presence of bigger particles and to assess their size, in aqueous suspensions used in the food industry.

Non-spherical particle size estimation using supervised machine learning

The inverse scattering problem of non-spherical particle size estimation is solved using a series of supervised machine learning models trained on a library of light scattering data. By establishing a large library with spheres and spheroids as fundamental shapes and through optimization of model hyperparameters, the trained models are able to accurately estimate a precise equivalent volume sphere radius of particles from an external database and simulations, with root mean square errors of 2.6% and 1.9% for the external and simulated particles, respectively. It was found that classification via a k-nearest neighbor model and refinement via a trained ensemble regression model performed best for equivalent volume measurements.

Determining optical constants of 2D materials with neural networks from multi-angle reflectometry data

Synthetically generated multi-angle reflectometry data is used to train a neural network based learning system to estimate the refractive index of atomically thin layered materials in the visible part of the electromagnetic spectrum. Unlike previously developed regression based optical characterization methods, the prediction is achieved via classification by using the probabilities of each input element belonging to a label as weighting coefficients in a simple analytical formula. Various types of activation functions and gradient descent optimizers are tested to determine the optimum combination yielding the best performance. For the verification of the proposed method’s accuracy, four different materials are studied. In all cases, the maximum error is calculated to be less than 0.3%. Considering the highly dispersive nature of the studied materials, this result is a substantial improvement in terms of accuracy and efficiency compared to traditional approaches.

Miniaturized sensor for particles in air using Fresnel ring lenses and an enhanced intensity ratio technique

We present a miniaturized particle sensor collecting scattered light in two solid angle intervals by Fresnel ring lenses. The particle size is determined from the ratio of both scattering amplitudes (intensity ratio) in addition to a linear diversity combining technique, generating a 3D particle size matrix that reduces the ambiguity by the index of refraction on the particle size identification. A signal-to-noise ratio of 30.3 was achieved for 147 nm sized polystyrene latex particles. Measurements of polydisperse particle size distribution show good agreement with the results by a scanning mobility particle sizer.

Generalized eikonal approximation for fast retrieval of particle size distribution in spectral extinction technique

In retrieving particle size distribution from spectral extinction data, a critical issue is the calculation of extinction efficiency, which affects the accuracy and rapidity of the whole retrieval. The generalized eikonal approximation (GEA) method, used as an alternative to the rigorous Mie theory, is introduced for retrieval of the unparameterized shape-independent particle size distribution (PSD). To compute the extinction efficiency more efficiently, the combination of GEA method and Mie theory is adopted in this paper, which not only extends the applicable range of the approximation method but also improves the speed of the whole retrieval. Within the framework of the combined approximation method, the accuracy and limitations of the retrieval are investigated. Moreover, the retrieval time and memory requirement are also discussed. Both simulations and experimental results show that the combined approximation method can be successfully applied to retrieval of PSD when the refractive index is within the validity range. The retrieval results we present demonstrate the high reliability and stability of the method. By using this method, we find the complexity and computation time of the retrieval are significantly reduced and the memory resources can also be saved effectively, thus making this method more suitable for online particle sizing.