Bombardier Recreational Products (Canada)

Transformers are becoming the state-of-the-art in multiple Computer vision (CV) and Natural language processing (NLP) tasks. As for hyperspectral target detection, a Transformer architecture named SpectralFormer2 has been developed and demonstrated improved performance over the previously state-of-the-art Convolutional neural network (CNN) architecture on widely studied classification tasks. The SpectralFormer was adapted from a CV architecture, the Vision Transformer (ViT). Concurrently, still in CV, a hierarchical and multi-scale version of the ViT, named Shifted windows (Swin) Transformer, is gaining momentum and is already the stateof-the-art on multiple tasks. In this paper, we adapt the Swin Transformer for hyperspectral classification and rare sub-pixel target detection. We apply this new architecture to commonly studied classification benchmarks public datasets such as Pavia University and Centre datasets, and to a new, large-scale airborne sub-pixel target detection dataset we developed. This new dataset is composed of over 100 M pixels taken over three collect days at three locations thousands of kilometers apart and in different climates. The proposed model reaches competitive performance on all of those tasks while reducing memory usage by 63 %.

The low conductivity of Na+ electrolytes in solid polymer electrolytes (SPEs) curtails the development of Na polymer batteries. In this study, NaClO4 (3-24 wt %, 90-9:1 O:Na) is dissolved in statistical copolymers of ethylene oxide (EO) and propylene oxide (PO) (0-20 mol %). Remarkably, the conductivity of these SPEs increases as the concentration of Na+ decreases, thus departing from the usual Nernstian behavior. Using a combination of calorimetric measurements and molecular dynamic simulations, this unusual phenomenon is attributed to the presence of physical cross-links generated by Na+ . As a result, polymers containing a low salt concentration (3 wt %) display a drastically enhanced ionic conductivity (up to 0.2 10-4 S cm-1 at 25 °C), thus paving the way for the design of all-solid-state PEO-based sodium batteries operational at room temperature.

Two methods of filamentation control for remote applications were studied. The first one consisted in an adaptive optic system, implemented in a specially designed focusing beam expander, which corrected for wavefront aberrations. Using this setup we are able to generate extraordinarily strong nitrogen signals at a distance as far as 90 m using 40 mJ laser pulses. Moreover, the filaments produced were used in a remote sensing scheme to detect and identify multiple targets such as trace hydrocarbon gases, solid metallic targets and aqueous aerosol clouds. The other method consists in a filament regularization method where a circular aperture of variable diameter is centered on the laser propagation axis, prior to filamentation. Longer filaments with higher ionization densities were produced in air and studied via the backscattered N2 fluorescence. 3D + time stochastic numerical simulations has shown that the optimum aperture size corresponds to the case of multiple filament 'squeezing' around the propagation axis forming the regularized elongated structure with higher overall amount of plasma.

High-energy femtosecond laser pulses propagating in the atmosphere undergo self-focusing resulting in the appearance of the phenomenon of filamentation. We observed an extra-attenuation of such (terawatt) femtosecond laser pulses propagating in the atmosphere when compared with long pulses (200 ps) with the same energy. This is because, in contrast to the linear propagation of the long pulse, the input femtosecond laser pulse is attenuated owing to either absorption through multiphoton ionization/tunnel ionization or to scattering on the laser-induced plasma; self-phase-modulation and self-steepening further convert partially the energy initially contained in the fundamental bandwidth into the broad side bands of the laser, becoming eventually a white-light laser pulse (supercontinuum). The experimental data allow us to extract an effective extra-attenuation coefficient for an exponential decay of the input pulse energy with the propagation distance. Such a coefficient allows us to estimate an upper bound of the filament length under the experimental conditions used. More generally, our observation leads to a new technique to remotely detect light filaments in the atmosphere.

<div class="htmlview paragraph">The Oil Certification Committee of the National Marine Manufacturers Association has developed FC-W<sup>®</sup>, a new standard for crankcase lube oil used in four-stroke cycle inboard, outboard, and sterndrive marine engines. A sub-committee representing the marine engine industry, the oil industry, oil testing laboratories, and the NMMA engineering standards group was formed to study the lubrication and corrosion prevention needs of four-stroke cycle engines. The sub-committee developed a rust test and an engine test as well as adopting 3 industry standard bench tests. These tests, together with formulation restrictions are used to identify oils that meet the requirements for use in four-stroke cycle marine engines. This paper gives an overview of the development of the new tests, formulation restrictions, and product approval system.</div>

Carbon Monoxide (CO) has become a pressing issue for the recreational marine industry. An increasing number of boating incidents have been linked to CO poisoning caused by emissions from gasoline-powered marine engines. Measurements by the National Institute for Occupational Safety and Health (NIOSH) and the U.S. Coast Guard have confirmed potentially hazardous CO concentrations near many of these engines. The measurements have also shown much lower CO concentrations for Evinrude® two-strbke direct-injected engines. This paper reviews national and international CO emission regulations for marine engines and discusses CO formation and reduction mechanisms. The differences between homogeneous- and stratified-charge combustion systems on CO formation, resulting from design and calibration criteria, are analyzed. The primary driving factors for CO formation under high-load homogeneous operation are driven by a desire to maximize power output and a necessity to control the thermal loading of internal engine components. Part throttle and low-load operation are largely driven by emissions and run-quality requirements. Several distinct advantages for direct-injected engines lead to the formation of significantly lower CO concentrations in the exhaust. The unavoidable over-scavenging of air aids in diluting cylinder-out CO levels. This, combined with high exhaust gas temperatures, can generate a significant post-oxidation effect. The post-oxidation effect was investigated in detail and some engine test results are presented. Stratified operation at low loads with high air-fuel ratios result in CO concentrations up to 100 times less than homogeneous-charge engines.

<div class="htmlview paragraph">Thanks to its unsurpassed power-to-weight ratio, its low package space and low-maintenance design, the loop-scavenged two-stroke engine with conventional mixture preparation is still being used in some sectors of vehicle engineering, such as boat drives, snow mobiles and motor scooters, as well as in hand-held applications.</div> <div class="htmlview paragraph">To maintain the potential of the 2-stroke engine for the future it is necessary to take adequate steps against the system-dependent disadvantage of the simple 2-stroke engine, namely that of higher emissions compared to 4-stroke engines. One possible solution is gasoline direct injection. Its more frequent use will increase the production numbers, making it an interesting technology even in the above-mentioned cost-sensitive applications.</div> <div class="htmlview paragraph">The current report presents various concepts of direct injection in 2-stroke engines, from air-assisted injection through to high-pressure direct injection, and compares them with traditional techniques of mixture formation. The report describes the experiences with different direct-injection systems, in particular in high-performance engines, from their development up to series application, presenting both simulation results and measurement data from tests for illustration. It also deals with the feasibility of exhaust aftertreatment and with the particularly critical subjective assessment of smoke and smell. In-depth investigations using 3D-CFD were made to develop and examine the different combustion methods and injection systems, including spray and combustion simulations with CFD codes specially adapted for two-stroke engines, taking the movement of the specific two-stroke charge into account. Flow investigations were made to evaluate the simulation and were compared with the tests and the simulations by means of measuring techniques specially adapted for two-stroke engines.</div> <div class="htmlview paragraph">The report concludes with an outlook on the effects of gasoline direct injection on fleet emissions and with the evaluation of future gasoline direct-injection technologies as to their feasibility and potential.</div>

<div class="htmlview paragraph">This paper presents a numerical study on air-fuel mixing in a two-stroke direct injection spark ignition engine under homogeneous operation. The simulated engine is loop scavenged and uses an outwardly opening swirl injector. A generic mesh-snapping algorithm is developed to enable the moving piston to snap through transfer ports with complicated geometry. A spray model based on Linear Instability Sheet Atomization is used to describe the primary breakup of fuel sprays, and the initial rotational velocity of the conical sheet is determined from a CFD simulation of the nozzle internal flow. A wall film model accounting for the effect of contacting area is also developed to avoid the severe grid-dependence of the original film model in KIVA. Comparisons between simulations and experiments were made for sprays in quiescent ambient conditions, and a good agreement of the spray characteristics was obtained. The simulations were performed for four different injection timings. The model was shown to be capable to predict engine-out unburned hydrocarbon emission. It was also found that the mixture preparation around the spark plug is directly related to the engine-out emission, and leaner mixture tends to yield higher engine-out emission.</div>

Aluminum nanopowders, because of their larger surface area, can increase the burning rate of propellants. It has been suggested that the powders could also enhance the detonation properties of certain explosives. For these reasons, an experimental study was undertaken to compare the performance of nanometric and micrometric aluminum in various explosives. No enhancement of performance was found in plastic-bonded explosives. In fact, a reduction of the detonation velocity was found in plastic-bonded explosives that are based on an energetic binder system. No increase of the detonation velocity was found in mixtures of aluminum and either Composition B or Ammonium Nitrate Fuel Oil, but a small increase in the heat of detonation was measured. The mixture of TNT and nano-aluminum demonstrated higher detonation velocities and heats of detonation. The increase was higher at small charge diameters. Nanometric aluminum was shown to reduce the critical diameter of such mixtures, and it is concluded that the nano-aluminum reacts faster than regular micron-size particles in TNT/Al compositions.

Backscatter and depolarization lidar measurements from clouds and precipitation are reported as functions of the elevation angle of the pointing lidar direction. We recorded the data by scanning the lidar beam (Nd:YAG) at a constant angular speed of ~3.5 degrees /s while operating at a repetition rate of 10 Hz. We show that in rain there is an evident and at times spectacular dependence on the elevation angle. That dependence appears to be sensitive to raindrop size. We have developed a three-dimensional polarization-dependent ray-tracing algorithm to calculate the backscatter and the depolarization ratio by large nonspherical droplets. We have applied it to raindrop shapes derived from existing static and dynamic (oscillating) models. We show that many of the observed complex backscatter and depolarization features can be interpreted to a good extent by geometrical optics. These results suggest that there is a definite need for more extensive calculations of the scattering phase matrix elements for large deformed raindrops as functions of the direction of illumination. Obvious applications are retrieval of information on the liquid-solid phase of precipitation and on the size and the vibration state of raindrops.