Deconvoluting acoustic beamforming maps with a deep neural network

Localization and quantification of noise sources is an important scientific and industrial problem, the use of phased arrays of microphones being the standard techniques in many applications. Non-physical artifacts appears on the output due to the nature of the method, thus, a supplementary step known as deconvolution is often performed. The use of data-driven machine learning can be a candidate to solve such problem. Neural networks can be extremely advantageous since no hypothesis concerning the environment or the characteristics of the sources are necessary, different from classical deconvolution techniques. Information on the acoustic propagation is implicitly extracted from pairs of source-output maps. On this work, a convolutional neural network is trained to deconvolute the beamforming map obtained from synthetic data simulating the response of an array of microphones. Quality of the estimation and the computational cost are compared to those of classical deconvolution methods (DAMAS, CLEAN-SC). Constraints associated with the size of the dataset used for training the neural network are also investigated and presented.

Overall Profile Measurements of Tiny Parts with Complicated Features with the Cradle-Type Five-Axis System

There are generally complex features with large curvature or narrow space on surfaces of complicated tiny parts, which makes high-precision measurements of their three-dimensional (3D) overall profiles a long-lasting industrial problem. This paper proposes a feasible measurement solution to this problem, by designing a cradle-type point-scanning five-axis measurement system. All the key technology of this system is also studied from the system construction to the actual measurement process, and the measurement accuracy is improved through error calibration and compensation. Finally, the feasibility is proved by engineering realization. The measurement capability of the system is verified by measuring workpieces such as cross cylinders and microtriangular pyramids.

A fingerprint of a heterogeneous data set

In this paper, we describe the fingerprint method, a technique to classify bags of mixed-type measurements. The method was designed to solve a real-world industrial problem: classifying industrial plants (individuals at a higher level of organisation) starting from the measurements collected from their production lines (individuals at a lower level of organisation). In this specific application, the categorical information attached to the numerical measurements induced simple mixture-like structures on the global multivariate distributions associated with different classes. The fingerprint method is designed to compare the mixture components of a given test bag with the corresponding mixture components associated with the different classes, identifying the most similar generating distribution. When compared to other classification algorithms applied to several synthetic data sets and the original industrial data set, the proposed classifier showed remarkable improvements in performance.

Learning to Approximate Industrial Problems by Operations Research Classic Problems

Operations research (OR) practitioners are accustomed to dealing with variants of classic OR problems. Indeed, an industrial problem often looks like a traveling salesman problem, a vehicle routing problem, a shortest path problem, etc., but has an additional constraint or a different objective that prevent the use of the powerful algorithms produced by decades of research on the classic OR problems. This situation can be frustrating, notably when we realize that the classic problem catches most of the structure of the variant. In “Learning to approximate industrial problems by operations research classic problems,” Axel Parmentier introduces a machine learning approach to use the algorithms for the classic OR problems on the variant. The idea is to leverage structured learning to obtain a mapping that approximates an instance of the variant by an instance of the classic problem.

A Gaussian Process Regression approach within a data-driven POD framework for engineering problems in fluid dynamics

<abstract><p>This work describes the implementation of a data-driven approach for the reduction of the complexity of parametrical partial differential equations (PDEs) employing Proper Orthogonal Decomposition (POD) and Gaussian Process Regression (GPR). This approach is applied initially to a literature case, the simulation of the Stokes problem, and in the following to a real-world industrial problem, within a shape optimization pipeline for a naval engineering problem.</p></abstract>

JUSTIFICATION OF PARAMETERS FOR DRYING GIANT GRENADIER (ALBATROSSIA PECTORALIS) DURING PRODUCTION OF DRIED-CURED PRODUCTS

Giant grenadier (Albatrossia pectoralis) is an underused commercial object, and improvement of its pro-cessing technology presents an urgent scientific and industrial problem. The solution to this problem, to a certain extent, is associated with the development of an effective technology for production of dried-cured products from this raw material. The study results of drying giant grenadier dynamics under various conditions, degree of proteins denaturation, organoleptic characteristics of dried products proved the rational modes for its processing in convective dryers. A three-stage drying was used at a temperature of 20ºC in the production of dried fillets, the duration of the first stage was 6 hours and leveling for 4 hours, the second stage was 6 hours and leveling for 4 hours and, finally, drying for 4 hours. During production of dried-cured products in the form of flakes, the fillet (i.e. decapitated fish) was divided into pieces (flakes) after the se-cond drying. Then they were soaked in a flavored filling and dried for 4 hours.

The bacterial lifecycle in cotton and polyester textiles

&lt;p&gt;Colonization of textiles and subsequent metabolic degradation of sweat and sebum components by axillary skin bacteria cause the characteristic sweat malodor and discoloring of dirty clothes. Once inside the textile, the bacteria can form biofilms that are hard to remove by conventional washing. When the biofilm persists after washing, the textiles retain the sweat odor. In addition to posing a huge industrial problem, textile biofilms constitute an interesting case study of bacterial behavior in periodically wetted and dried substrates with varying surface hydrophobicity. Here we aim to study the bacterial behavior in each of the four stages of the bacterial lifecycle in textiles: adhesion, growth, drying and washing. To accomplish this, we designed a novel in vitro model to mimic physiological sweating while wearing cotton and polyester textiles. The hydrophobic polyester adhered bacteria more strongly and absorbed more sebum, the bacteria&amp;#8217;s primary nutrient source. Bacteria were therefore initially more active in polyester textiles than in cotton. However, polyester did not bind water as well as cotton. The increased water content of cotton allowed the bacteria to retain a higher activity after the textile had dried. However, neither of the textiles retained enough water upon drying to prevent the bacteria from irreversibly adhering to the textile fibers by capillary action. This demonstrates that bacterial colonization depends on the hydrophobic and hygroscopic properties of the colonized material while highlighting the possibility of controlling bacterial behavior by either changing the surface properties or the surrounding environment.&lt;/p&gt;