Using Recurrent Neural Networks to Reconstruct Temperatures from Simulated Fluorescent Data for use in Bio-Microfluidics.

Many biological systems have a narrow temperature range of operation, meaning high accuracy and spatial distribution level are needed to study these systems. Most temperature sensors cannot meet both the accuracy and spatial distribution required in the microfluidic systems that are often used to study these systems in isolation. This paper introduces a neural network called the Multi-Directional Fluorescent Temperature Long Short-Term Memory Network (MFTLSTM) that can accurately calculate the temperature at every pixel in a fluorescent image to improve upon the standard fitting practice and other machine learning methods use to relate fluorescent data to temperature. This network takes advantage of the nature of heat diffusion in the image to achieve an accuracy of ±0.0199 K RMSE within the temperature range of 298K to 308 K with simulated data. When applied to experimental data from a 3D printed microfluidic device with a temperature range of 290 K to 380 K, it achieved an accuracy of ±0.0684 K RMSE. These results have the potential to allow high temperature resolution in biological systems than is available in many microfluidic devices.

3D-Printed Microfluidic One-Way Valves and Pumps.

New microfluidic lab-on-a-chip capabilities are enabled by broadening the toolkit of devices that can be created using microfabrication processes. For example, complex geometries made possible by 3D printing can be used to approach microfluidic design and application in new or enhanced ways. In this paper, we demonstrate three distinct designs for microfluidic one-way (check) valves that can be fabricated using digital light processing stereolithography (DLP-SLA) with a poly(ethylene glycol) diacrylate (PEGDA) resin, each with an internal volume of 5-10 nL. By mapping flow rate to pressure in both the forward and reverse directions, we compare the different designs and their operating characteristics. We also demonstrate pumps for each one-way valve design comprised of two one-way valves with a membrane valve displacement chamber between them. An advantage of such pumps is that they require a single pneumatic input instead of three as for conventional 3D-printed pumps. We also characterize the achievable flow rate as a function of the pneumatic control signal period. We show that such pumps can be used to create a single-stage diffusion mixer with significantly reduced pneumatic drive complexity.

Effect of Workpiece/Tool Heat Transfer and Friction Coefficients on Accuracy of Simulated Temperatures and Torques in a Friction Stir Welding Plunge.

Friction stir process models are typically validated by tuning heat transfer and friction coefficients until measured temperatures in either the tool or workpiece, but rarely in both, match simulated results. A three-dimensional finite element model for a tool plunge in an AA 6061-T6 is validated for temperature predictions in both the tool and workpiece using a friction coefficient that varies with time. Peak workpiece temperatures were within 1.5% of experimental temperatures and tool temperatures were off by 80 °C. The sensitivity of the predicted temperatures with respect to the workpiece/tool heat transfer coefficient was shown to be high for the tool and low for the workpiece, while the spindle torque was slightly underpredicted in the best case. These results show that workpiece/tool interface properties must be tuned by considering predictions on both sides of the heat generation interface in order to ensure a reliable process simulation.

Demonstration of Neural Networks to Reconstruct Temperatures from Simulated Fluorescent Data Toward Use in Bio-microfluidics.

Biological systems often have a narrow temperature range of operation, which require highly accurate spatially resolved temperature measurements, often near ±0.1 K. However, many temperature sensors cannot meet both accuracy and spatial distribution requirements, often because their accuracy is limited by data fitting and temperature reconstruction models. Machine learning algorithms have the potential to meet this need, but their usage in generating spatial distributions of temperature is severely lacking in the literature. This work presents the first instance of using neural networks to process fluorescent images to map the spatial distribution of temperature. Three standard network architectures were investigated using non-spatially resolved fluorescent thermometry (simply-connected feed-forward network) or during image or pixel identification (U-net and convolutional neural network, CNN). Simulated fluorescent images based on experimental data were generated based on known temperature distributions where Gaussian white noise with a standard deviation of ±0.1 K was added. The poor results from these standard networks motivated the creation of what is termed a moving CNN, with an RMSE error of ±0.23 K, where the elements of the matrix represent the neighboring pixels. Finally, the performance of this MCNN is investigated when trained and applied to three distinctive temperature distributions characteristic within microfluidic devices, where the fluorescent image is simulated at either three or five different wavelengths. The results demonstrate that having a minimum of 10 3.5 data points per temperature and the broadest range of temperatures during training provides temperature predictions nearest to the true temperatures of the images, with a minimum RMSE of ±0.15 K. When compared to traditional curve fitting techniques, this work demonstrates that greater accuracy when spatially mapping temperature from fluorescent images can be achieved when using convolutional neural networks.

3D printing-enabled uniform temperature distributions in microfluidic devices.

Many microfluidic processes rely heavily on precise temperature control. Though internally-contained heaters have been developed using traditional fabrication methods, they are limited in their ability to isothermally heat a precisely defined volume. Advances in 3D printing have led to high resolution printers capable of using bio-compatible materials and achieving geometry resolutions near 20 µm. 3D printing's ability to create arbitrary 3D structures with an arbitrary 3D orientation as opposed to traditional microfluidic fabrication methods enables new three-dimensional heater geometries to be created. As examples, we demonstrate three new 3D heater geometries: a non-planar serpentine channel, a tapered helical channel, and a diamond channel. These new geometries are shown through finite element simulation to isothermally heat microfluidic channels of cross section 200 µm × 200 µm with a 0.1 °C temperature difference along up to 91% of a 10 mm length, compared to designs from the literature that are only able to have that same temperature distance over several µms. Finally, a set of design rules to create isothermal regions in 3D based on the desired temperature, heater pitch, heater gradient, and radial space around a target volume are detailed.

Use of Machine Learning with Temporal Photoluminescence Signals from CdTe Quantum Dots for Temperature Measurement in Microfluidic Devices.

Because of the vital role of temperature in many biological processes studied in microfluidic devices, there is a need to develop improved temperature sensors and data analysis algorithms. The photoluminescence (PL) of nanocrystals (quantum dots) has been successfully used in microfluidic temperature devices, but the accuracy of the reconstructed temperature has been limited to about 1 K over a temperature range of tens of degrees. A machine learning algorithm consisting of a fully-connected network of seven layers with decreasing numbers of nodes was developed and applied to a combination of normalized spectral and time-resolved PL data of CdTe quantum dot emission in a microfluidic device. The data used by the algorithm was collected over two temperature ranges: 10 K to 300 K, and 298 K to 319 K. The accuracy of each neural network was assessed via mean absolute error of a holdout set of data. For the low temperature regime, the accuracy was 7.7 K, or 0.4 K when the holdout set is restricted to temperatures above 100 K. For the high temperature regime, the accuracy was 0.1 K. This method provides demonstrates a potential machine learning approach to accurately sense temperature in microfluidic (and potentially nanofluidic) devices when the data analysis is based on normalized PL data when it is stable over time.

Fluorescent scanning thermal microscope based on a Blu-ray optical head to measure thermal diffusivity of radioactive samples.

Over the life of nuclear fuel, inhomogeneous structures develop, negatively impacting thermal properties. New fuels are under development but require more accurate knowledge of how the properties change to model performance and determine safe operational conditions. Measurement systems capable of microscopic thermal transport measurements and low cost are necessary to measure these properties and integrate into hot cells where electronics are likely to fail during fuel investigation. This project develops a cheaper, smaller, and easily replaceable Fluorescent Scanning Thermal Microscope (FSTM) using the blue laser and focusing circuitry from an Xbox HD-DVD player that incorporates novel fluorescent thermometry methods to determine thermal diffusivity. The FSTM requires minimal sample preparation, does not require access to both sides of the sample, and components can be easily swapped out if damaged, as is likely in irradiated hot cells. Using the optical head from the Xbox for sensing temperature changes, an infrared laser diode provides periodic heating to the sample, and the blue laser induces fluorescence in Rhodamine B deposited on the sample's surface. Thermal properties are fit to modulated temperature models based on the phase delay response at different modulated heating frequencies. With the FSTM method, the thermal diffusivity of a Nordic gold (euro) coin was found to be 21 ± 5 mm2/s. This value is compared to laser flash and thermal conductivity microscope methods, which found the thermal diffusivity to be 30.4 ± 0.1 mm2/s and 19 ± 3 mm2/s. The system shows promise as a feasible property characterization technique with future refinement and testing in progress.

Investigation of synthetic spider silk crystallinity and alignment via electrothermal, pyroelectric, literature XRD, and tensile techniques.

The processes used to create synthetic spider silk greatly affect the properties of the produced fibers. This paper investigates the effect of process variations during artificial spinning on the thermal and mechanical properties of the produced silk. Property values are also compared to the ones of the natural dragline silk of the N. clavipes spider, and to unprocessed (as-spun) synthetic silk. Structural characterization by scanning pyroelectric microscopy is employed to provide insight into the axial orientation of the crystalline regions of the fiber and is supported by XRD data. The results show that stretching and passage through liquid baths induce crystal formation and axial alignment in synthetic fibers, but with different structural organization than natural silks. Furthermore, an increase in thermal diffusivity and elastic modulus is observed with decreasing fiber diameter, trending towards properties of natural fiber. This effect seems to be related to silk fibers being subjected to a radial gradient during production.