Multi-Fidelity High-Throughput Optimization of Electrical Conductivity in P3HT-CNT Composites

Combining high-throughput experiments with machine learning accelerates materials and process optimization toward user-specified target properties. In this study, a rapid machine learning-driven automated flow mixing setup with a high-throughput drop-casting system is introduced for thin film preparation, followed by fast characterization of proxy optical and target electrical properties that completes one cycle of learning with 160 unique samples in a single day, a > 10 ×  improvement relative to quantified, manual-controlled baseline. Regio-regular poly-3-hexylthiophene is combined with various types of carbon nanotubes, to identify the optimum composition and synthesis conditions to realize electrical conductivities as high as state-of-the-art 1000 S cm⁻¹. The results are subsequently verified and explained using offline high-fidelity experiments. Graph-based model selection strategies with classical regression that optimize among multi-fidelity noisy input-output measurements are introduced. These strategies present a robust machine-learning driven high-throughput experimental scheme that can be effectively applied to understand, optimize, and design new materials and composites.     

Free-running ADC- and FPGA-based signal processing method for brain PET using GAPD arrays

Currently, for most photomultiplier tube (PMT)-based PET systems, constant fraction discriminators (CFD) and time to digital converters (TDC) have been employed to detect gamma ray signal arrival time, whereas anger logic circuits and peak detection analog-to-digital converters (ADCs) have been implemented to acquire position and energy information of detected events. As compared to PMT the Geiger-mode avalanche photodiodes (GAPDs) have a variety of advantages, such as compactness, low bias voltage requirement and MRI compatibility. Furthermore, the individual read-out method using a GAPD array coupled 1:1 with an array scintillator can provide better image uniformity than can be achieved using PMT and anger logic circuits. Recently, a brain PET using 72 GAPD arrays (4×4 array, pixel size: 3 mm×3 mm) coupled 1:1 with LYSO scintillators (4×4 array, pixel size: 3 mm×3 mm×20 mm) has been developed for simultaneous PET/MRI imaging in our laboratory. Eighteen 64:1 position decoder circuits (PDCs) were used to reduce GAPD channel number and three off-the-shelf free-running ADC and field programmable gate array (FPGA) combined data acquisition (DAQ) cards were used for data acquisition and processing. In this study, a free-running ADC- and FPGA-based signal processing method was developed for the detection of gamma ray signal arrival time, energy and position information all together for each GAPD channel. For the method developed herein, three DAQ cards continuously acquired 18 channels of pre-amplified analog gamma ray signals and 108-bit digital addresses from 18 PDCs. In the FPGA, the digitized gamma ray pulses and digital addresses were processed to generate data packages containing pulse arrival time, baseline value, energy value and GAPD channel ID. Finally, these data packages were saved to a 128 Mbyte on-board synchronous dynamic random access memory (SDRAM) and then transferred to a host computer for coincidence sorting and image reconstruction. In order to evaluate the functionality of the developed signal processing method, energy and timing resolutions for brain PET were measured via the placement of a 6 μCi ²²Na point source at the center of the PET scanner. Furthermore the PET image of the hot rod phantom (rod diameter: from 2.5 mm to 6.5 mm) with activity of 1 mCi was simulated, and then image acquisition experiment was performed using the brain PET. Measured average energy resolution for 1152 GAPD channels and system timing resolution were 19.5% (FWHM%) and 2.7 ns (FWHM), respectively. With regard to the acquisition of the hot rod phantom image, rods could be resolved down to a diameter of 2.5 mm, which was similar to simulated results. The experimental results demonstrated that the signal processing method developed herein was successfully implemented for brain PET. This reduced the complexity, cost and developing duration for PET system relative to normal PET electronics, and it will obviously be useful for the development of high-performance investigational PET systems.     

Supply chain management practices as a mediator of the relationship between operations capability and firm performance

The current study uses mediated regression analysis and structural equation modelling to test the proposition that supply chain management practices mediate the relationship between operations capability and firm performance. Operations capability is defined in terms of a firm's new product design and development, total quality management and just-in-time capabilities. Results support the research model and also suggest the existence of a direct relationship between operations capability and performance.     

The effect of oxidation on the corrosion resistance and mechanical properties of a Zr-based metallic glass

The oxidation kinetics of the Zr₆₄Cu₁₆Ni₁₀Al₁₀ bulk metallic glass (BMG) roughly follows a two-stage rate law at both 433 and 593 K in air. An oxide film of 940 nm can be formed by oxidation at 593 K, which is ZrO₂-enriched but Cu-depleted on the outer surface. The oxide film leads to a superior passivity in 0.5 M NaCl and great corrosion resistance improvements in other solutions. The oxidation effect on mechanical properties were characterized by nanoindentation, wedge indentation and compression tests. The Zr-based BMG still keeps the amorphous nature and its good mechanical properties are retained after oxidation.     

Production of Furfural from Lignocellulosic Biomass Using Beta Zeolite and Biomass-Derived Solvent

The production of furfural from the C₅ monosaccharides xylose, arabinose and ribose, as well as from real biomass (corn fiber), was studied using H-Beta zeolite as catalyst in a monophasic system with the biomass-derived solvent, gamma-valerolactone. Due to the combination of Brønsted and Lewis acid sites on this catalyst (Brønsted:Lewis ratio = 1.66), H-Beta acts as a bifunctional catalyst, being able to isomerize (Lewis acid) and dehydrate (Brønsted acid) monosaccharides. The combination of Lewis and Brønsted acid functionality of H-Beta was shown to be effective for the isomerization of xylose and arabinose, followed by dehydration. While no advantages were found in the conversion of xylose, higher furfural yields were achieved from arabinose, using H-Beta, 73 %, compared to sulfuric acid (44 %) and Mordenite (49 %). The furfural yields from corn fiber for H-Beta, H-Mordenite and sulfuric acid were 62, 44, and 55 %, respectively, showing that H-Beta is particularly effective for conversion of this biomass feedstock composed of 45 wt% hemicellulose, of which 66 % is xylose and 33 % arabinose.