Personalized, Sequential, Attentive, Metric-Aware Product Search

The task of personalized product search aims at retrieving a ranked list of products given a user’s input query and his/her purchase history. To address this task, we propose the PSAM model, a Personalized, Sequential, Attentive and Metric-aware (PSAM) model, that learns the semantic representations of three different categories of entities, i.e., users, queries, and products, based on user sequential purchase historical data and the corresponding sequential queries. Specifically, a query-based attentive LSTM (QA-LSTM) model and an attention mechanism are designed to infer users dynamic embeddings, which is able to capture their short-term and long-term preferences. To obtain more fine-grained embeddings of the three categories of entities, a metric-aware objective is deployed in our model to force the inferred embeddings subject to the triangle inequality, which is a more realistic distance measurement for product search. Experiments conducted on four benchmark datasets show that our PSAM model significantly outperforms the state-of-the-art product search baselines in terms of effectiveness by up to 50.9% improvement under NDCG@20. Our visualization experiments further illustrate that the learned product embeddings are able to distinguish different types of products.

Colliding respiratory jets as a mechanism of air exchange and pathogen transport during conversations

Air exchange between people has emerged in the COVID-19 pandemic as the important vector for transmission of the SARS-CoV-2 virus. We study the airflow and exchange between two unmasked individuals conversing face-to-face at short range, which can potentially transfer a high dose of a pathogen, because the dilution is small when compared to long-range airborne transmission. We conduct flow visualization experiments and direct numerical simulations of colliding respiratory jets mimicking the initial phase of a conversation. The evolution and dynamics of the jets are affected by the vertical offset between the mouths of the speakers. At low offsets the head-on collision of jets results in a `blocking effect', temporarily shielding the susceptible speaker from the pathogen carrying jet, although, the lateral spread of the jets is enhanced. Sufficiently large offsets prevent the interaction of the jets. At intermediate offsets (8-10 cm for 1 m separation), jet entrainment and the inhaled breath assist the transport of the pathogen-loaded saliva droplets towards the susceptible speaker's mouth. Air exchange is expected, in spite of the blocking effect arising from the interaction of the respiratory jets from the two speakers.

Edge Race-Tracking during Film-Sealed Compression Resin Transfer Molding

Edge race-tracking is a frequently reported issue during resin transfer molding. It is caused by highly permeable channels and areas between the preform edge and cavity, which can significantly change the preform impregnation pattern. To date, information is scarce on the effect of edge race-tracking in compression resin transfer molding (CRTM). To close this gap, laboratory equipment was developed to study the CRTM preform impregnation via flow visualization experiments. The preform was thereby encapsulated in thin thermoplastic films sealing its impregnation. Film-sealed compression resin transfer molding (FS-CRTM) experiments of preforms with a small geometrical aspect ratio showed fast filling of the injection gap and a subsequent through-thickness preform impregnation. Creating an edge race-tracking channel, an additional lateral in-plane flow from the channel towards the preform center was observed, initiating soon after the injection started and caused by the spatial connection between the injection gap and the race-tracking channel. To diminish edge race-tracking, a passive flow control strategy was implemented via a split design of the upper tool to spatially isolate the injection gap from the channel and to pre-compact the preform edge. A delayed and reduced lateral race-tracking flow was observed, showing that the passive flow control strategy increases the process robustness of FS-CRTM regarding edge race-tracking effects.

The World’s First Test Facility That Enables the Experimental Visualization of Cavitation on a Rotating Inducer in Both Cryogenic and Ordinary Fluids

It is well known that cavitation breakdown, which is a phenomenon in which the pump head suddenly drops with a decrease in the inlet cavitation number, occurs in turbopumps. Especially in cryogenic pumps, cavitation breakdown occurs at a lower inlet cavitation number than that of ordinary fluids such as water. This phenomenon is referred to as a thermodynamic effect, as Stepanoff reported. The thermodynamic properties of the working fluid affect the sizes of cavitation elements, and the sizes affect cavitation breakdown; therefore, experimental flow visualization is an effective approach to realize a more efficient and more reliable cryogenic pump. In 2010, the author and colleagues developed the worldߣs first test facility to enable the visualization of cavitation on a rotating inducer in both cryogenic and ordinary fluids. At that time, only two reports on the flow visualization of a rotating cryogenic impeller had been published: one on flow visualization in liquid hydrogen by NASA in 1967 and the other in liquid nitrogen by JAXA in 2010. The present facility employs a triple-thread helical inducer with a diameter of 65.3 mm and a rotation rate of up to 8000 rpm with both liquid nitrogen and water available as working fluids. Unsteady visualization experiments for cavitation on an inducer in liquid nitrogen and water have revealed the characteristics of tip vortex cavitation, backflow vortex cavitation, and cavitation element size based on comparisons between cryogenic fluids that exhibit a stronger thermodynamic effect and ordinary fluids such as water.

Vortex interaction on a LEX configuration

Flow visualization experiments were conducted in a water channel and a low-speed wind tunnel at Reynolds number of 1.54 $$\times$$ × 104 to 1.2 $$\times$$ × 105 for a leading-edge extension model, which is referred to as a NASA TP-1803 model in this study. In addition, particle image velocimetry velocity measurements were taken in the water channel to obtain the quantitative information about the three-dimensional velocity field over the strake and wing surfaces. The results obtained at low, medium and high angles of attacks represent three distinct cases of interaction between the strake and wing vortices. Namely, at α = 5o and 10° the strake and wing vortices were developed over the wing surface without significant interaction noticed; at α = 20°, the strake vortex strongly interacted with the wing vortex in an intertwining manner, which was sensitive to Reynolds number; at α = 30°, the breakdown of the strake vortex took place close to the junction of the strake and the wing; thus, the interaction of the strake and the wing vortices appeared to be less significant than the case of α = 20°. Graphic abstract

Elastic energy storage in seahorses leads to a unique suction flow dynamics compared to other actinopterygian

Suction feeding is a dominant prey-capture strategy across actinopterygians, consisting of a rapid expansion of the mouth cavity that drives a flow of water containing the prey into the mouth. Suction feeding is a power-hungry behavior, involving the actuation of cranial muscles as well as the anterior third of the fish's swimming muscles. Seahorses, which have reduced swimming muscles, evolved a unique mechanism for elastic energy storage that powers their suction flows. This mechanism allows seahorses to achieve head rotation speeds that are 50 times faster than fish lacking such a mechanism. However, it is unclear how the dynamics of suction flows in seahorses differ from the conserved pattern observed across other actinopterygians, nor how differenced in snout length across seahorses affect these flows. Using flow visualization experiments, we show that seahorses generate suction flows that are 8 times faster than similar-sized fish, and that the temporal patterns of cranial kinematics and suction flows in seahorses differs from the conserved pattern observed across other actinopterygians. However, the spatial patterns retain the conserved actinopterygian characteristics, where suction flows impact a radially symmetric region of ∼1 gape diameter outside the mouth. Within seahorses, increases in snout length were associated with slower suction flows and faster head rotation speeds, resulting in a trade-off between pivot feeding and suction feeding. Overall, this study shows how the unique cranial kinematics in seahorses are manifested in their suction feeding performance, and highlights the trade-offs associated with their unique morphology and mechanics.

SAVSDN: A Scene-Aware Video Spark Detection Network for Aero Engine Intelligent Test

Due to carbon deposits, lean flames, or damaged metal parts, sparks can occur in aero engine chambers. At present, the detection of such sparks deeply depends on laborious manual work. Considering that interference has the same features as sparks, almost all existing object detectors cannot replace humans in carrying out high-precision spark detection. In this paper, we propose a scene-aware spark detection network, consisting of an information fusion-based cascading video codec-image object detector structure, which we name SAVSDN. Unlike video object detectors utilizing candidate boxes from adjacent frames to assist in the current prediction, we find that efforts should be made to extract the spatio-temporal features of adjacent frames to reduce over-detection. Visualization experiments show that SAVSDN can learn the difference in spatio-temporal features between sparks and interference. To solve the problem of a lack of aero engine anomalous spark data, we introduce a method to generate simulated spark images based on the Gaussian function. In addition, we publish the first simulated aero engine spark data set, which we name SAES. In our experiments, SAVSDN far outperformed state-of-the-art detection models for spark detection in terms of five metrics.

Particle Removal in Ultrasonic Water Flow Cleaning Role of Cavitation Bubbles as Cleaning Agents

Visualization experiments are performed to examine the role of acoustic cavitation bubbles that appear in 0.43-MHz ultrasonic water flow spreading over glass surfaces in the context of physical cleaning. The cleaning performance is evaluated using glass samples on which small silica particles are spin-coated. The visualization suggests that acoustic cavitation bubbles play a major role in particle removal as in the case of conventional cleaning with ultrasonic cleaning baths.