Pressure Drop In Channel Research Articles

Progress in reflood thermal hydraulics studies in the past 40 years

Accurate simulation of reflood transients of a reactor core is one of the top priorities for nuclear reactor thermal–hydraulic and safety analysis. An in-depth understanding of the two-phase flow and heat transfer process during reflooding can provide extremely useful information on the design and analysis of the various safety systems as well as on the development of numerical analysis tools. In the current paper, the status of the reflood transients in the past several decades is overviewed and discussed in detail, in the hope to provide a roadmap for future relevant research. Following an introduction of the current topic, the phenomenological aspects related to reflood are discussed first. Then, studies on the two-phase flow thermal–hydraulic non-equilibrium are reviewed and evaluated critically. The flow channel pressure drop as well as the spacer grid effect constitute important parts of reflood thermal–hydraulic and safety analysis, and various studies on these topics are discussed. Another important consideration during reflood transients is the liquid droplet behavior since it affects the rod bundle thermal response. In addition, the current paper also identifies the several limitations and the need for future research work on reflood thermal hydraulics.

Two-phase pressure drop and void fraction in a cross-corrugated plate heat exchanger channel: Impact of flow direction and gas-liquid distribution

Two-phase pressure drop is studied in a transparent cross-corrugated channel for uniform and non-uniform gas–liquid distribution. Both uniform and non-uniform distribution are created by injecting air and water through different numbers of nozzles directly into the channel. The frictional pressure drop clearly indicates the impact of maldistribution.The single channel is pivot-mounted to allow horizontal, vertical upward and downward flow. Two-phase frictional pressure drop can be directly measured only for horizontal flow, neglecting acceleration pressure drop. In vertical flow, a void fraction model is necessary to deduct gravitational pressure drop from the measured pressure difference.The two-phase friction factors of all flow directions are compared to the correlation of single-phase flow. In this manner, homogeneous flow is distinguished from slip. Homogenous flow is confirmed for homogeneous void fractions up to 0.7 at uniform and up to 0.25 at non-uniform gas injection. The single-phase correlation is valid for homogeneous flow when inserting two-phase variables. The experimental two-phase multiplier is compared to correlations from literature.The pressure measurement data are used to determine the actual void fraction. The results are compared to published void fraction models. The homogeneous model fits best for uniform two-phase distribution. For maldistribution, the correlations of Margat et al. (1997) and of Rouhani and Axelsson (1970) predict the experimental void fractions reasonably well, except for very small flow rates.Analysis of experimental results is facilitated by previously published research on flow pattern visualization and modelling in the same test channel. The cross-corrugated geometry generates a swirling motion inducing centrifugal forces often much larger than buoyancy, which thus loses impact on flow pattern, pressure drop and void fraction. The ratio of buoyancy to centrifugal force is expressed by the corrugation Froude number. Slip independent of buoyancy is identified by introducing the hydrostatic correction factor to evaluate friction factors deviating with flow direction.

Optimal Design and Performance Analysis of Radial MR Valve with Single Excitation Coil

The main issue addressed in this paper involves the magnetorheological (MR) valve increasing the pressure drop by changing the internal structure, which leads to the increase of dimension sizes and the easy blocking of the internal channel. Optimizing the design of the traditional radial MR valve without changing the internal structure and whole dimension size is indispensable. Firstly, a radial MR valve with single excitation coil was proposed. The mathematical models of the field-dependent pressure drop and viscosity pressure drop in fluid flow channels were deduced, and the calculation formula of pressure drop was also established. Then, ANSYS software was used to simulate and analyze the distributions of the magnetic flux lines and magnetic flux densities of the proposed radial MR valve. Subsequently, the radial MR valve was simulated and analyzed by using the ANSYS first-order and zero-order simulation tools. In addition, the experimental test bench of the proposed MR valve was setup, the static performance of pressure drop was tested, and the change of pressure drop of the optimal radial MR valve under different loads was studied, furthermore, the response time with current of the initial and optimal radial MR valve were also investigated. Finally, the dynamic performances of the optimal radial MR valve controlled cylinder system under different currents, frequencies and amplitudes were tested, respectively. The experimental results indicate that the total pressure drop of the initial valve is 1.842 MPa when the applied current is 1.8 A, and the total pressure drop of the optimal valve is 2.58 MPa, the increase is 40.07%. Meanwhile, the maximum damping force of the optimal radial MR valve controlled cylinder system can reach about 3.6 kN at the current of 1.25 A, which shows a better optimization effect of the optimal radial MR valve.

Numerical study on the impact of interface contact resistance on the performance of a PEMFC with serpentine flow field

The performance of a proton exchange membrane fuel cell (PEMFC) strongly relies on the flow-field design. In this study, a three-dimensional, multiphase, and non-isothermal PEMFC model with single serpentine flow field has been developed to investigate the effect of interface contact resistance (ICR) between the bipolar plate (BP) and gas diffusion layer (GDL) on the flow field dimension. In the cases of 25 different combinations of channel and rib width, polarization curves and power density curves of the PEMFC without ICR and with ICR of 7.418 mΩ.cm2 as well as 21 mΩ.2 are investigated. The results show that ICR has little effect on channel pressure drop but on liquid water accumulation. The value of ICR affects both the ohmic loss stage and concentration loss stage of the polarization curves. The power density of the PEMFC without ICR at the operating voltage 0.5V decreases as the rib width increases, while that with ICR increases as the rib width increases. Furthermore, the influence of invalid active area in BPs caused by flow field configuration on the cell performance should not be neglected when the ICR value is lower than 10 mΩ.2.

Effect of Inlet Plenum on Pressure Drop and Velocity in Fractal Micro Channels

The computational analysis for micro channel flow in a branched network was investigated by three dimensional CFD approach. The effect of the change of Inlet Plenum (IP) size at a constant Aspect Ratio (AR) of the outermost channel on pressure drop in a fractal branched micro channel was performed. The properties are compared along a particular path and it was observed that the pressure drop along a bifurcated path has considerably less effect when compared to that of the outer most straight branched channel for a constant aspect ratio model. Pressure does not change significantly if we change the IP radius even when all other parameters are constant. Velocity in the inner channel after a straight run has reduced significantly even for same AR and Reynolds Number (Re). This leads to the conclusion that the IP size affects the velocity after the bifurcation.

Performance investigation and enhancement of membrane-contactor microchannel absorber towards compact absorption cooling

Solar absorption cooling is a promising approach to address the energy and environmental issues. Large-size and low-efficiency are two big challenges for wider applications of solar absorption cooling, while membrane-based microchannel absorbers are potential for effective enhancement in compactness and efficiency. To investigate the local heat and mass transfer characteristics of membrane-based microchannel absorbers, a two-dimension coupling model is developed and validated. The influences of key parameters and entrance effect on the absorption rate are analyzed in detail. Results show that the average absorption rate increases by 40% as the solution channel thickness decreases from 1 mm to 0.15 mm, but the pressure drop in the solution channel increases nearly exponentially. Meanwhile, as solution inlet velocity increases from 0.002 m/s to 0.02 m/s, the average absorption rate increases by 56%. Considering both pressure drop and absorption rate, the recommended solution inlet velocity and channel thickness are approximately 0.004 m/s and 0.5 mm, respectively. In addition, an enhancement strategy (channel baffles) is proposed to improve the performance of microchannel absorbers. The average absorption rate of the baffled module increases to 0.0053 kg/(m2•s), which is 20% higher than that without baffles. As for the compactness, the volumetric cooling capacity with baffles is 230% higher than that of a conventional falling-film absorber. Given a similar volumetric cooling capacity, the pressure drop of the baffled absorber is approximately 50% lower as compared to the non-baffle absorber. This study can facilitate the in-depth analysis and optimal design of membrane-based microchannel absorbers towards compact solar absorption cooling systems.

Comparative assessment of the effects of 3D printed feed spacers on process performance in MD systems

In this study, process enhancements achieved by the use of 3D printed feed spacers based on triply periodic minimal surfaces (TPMS) were comparatively assessed using the two most common membrane distillation (MD) configurations: air gap (AGMD) and direct contact (DCMD). The MD performance was assessed based on the impact of spacer design (TPMS vs. commercial spacer) on flux and feed channel pressure drop, and their consequent impact on the levelized cost of water (COW). The studied spacer architectures led to a minimal improvement (≤17%) in AGMD flux, much lower than that achieved in DCMD (≤57%). Consequently, for a waste heat-operated MD process, the spacer-induced channel pressure drop became the most influential cost bottleneck on COW. Generally, the contribution of the pumping cost to the total operating cost was found to be greater for DCMD than AGMD. Thus, the COW of a waste heat operated DCMD is more sensitive to a decrease in spacer-induced channel pressure drop than in an AGMD. For an MD process operated with additional heat cost, the flux improvement achieved using TPMS spacers reduces the thermal energy component of the operating cost. Ultimately, this predominantly contributes to a lower COW.

Thermal analysis of a high-efficiency internally-cooled strut injector for scramjet engine

Adding fuel strut in scramjet combustor is an effective solution to improve fuel-air mixing and combustion efficiency. However, strut injector suffers from extreme aerodynamic heat, thereby needing high-efficiency cooling method to ensure its thermal safety. This study presented a numerical demonstration of an internally-cooled strut injector based on coiled and miniature-sized cooling configuration that could operate under low coolant mass flow rate of 10 g/s, while more than three times of coolant was needed for another two control cases of previous designs. The heat transfer coefficient and heatflux on the leading edge of present-designed strut were the maximum of the three struts. The non-uniform distribution of pressure in the manifold caused the deviations in the mass flow and pressure drop in the subsequent paired channels. Friction head loss and local resistance loss was separately responsible for 72.3% and 27.7% of total pressure drop. However, as the mass flow increased, the expense of pressure drop outstripped the merit of heat transfer improvement.

Simulation study on the influence of electrostatic adsorption grid on the pressure and particle concentration of flow field

Electrostatic adsorption dust removal can effectively improve the dust removal effect of inhalable particles. The grid layout in the flue gas affects the dust removal efficiency and channel pressure drop, and increases the power loss of the fan. In this paper, according to the velocity of flue gas and the concentration of particulate matter in flue gas, combined with the principle of electrostatic adsorption, a three-dimensional module including the main components of adsorption and dust removal was established. The distribution of pressure drop in the flue gas channel caused by the adsorption components with grid layout was simulated and analyzed, which provided a reference for the reasonable layout of dust adsorption components. The simulation results show that the smaller the component diameter, the smaller the pressure difference caused by the same number of components, and the smaller the longitudinal spacing of components.

Study on Unsteady Cavitation Flow and Pressure Pulsation Characteristics in the Regulating Valve

A combined numerical‐experiment investigation on the unsteady cavitation flow and pressure fluctuation characteristics in the regulating valves is conducted in this paper. The cavitation flow in the regulating valve is an unsteady and periodic flow which could be divided into fixed and travelling cavitation bubbles. The fixed cavitation bubbles are formed in the gap in the initial stage and then fell off and formed the travelling cavitation bubbles because of the re‐entrant jet. The travelling cavitation bubbles move downstream, oscillate, and break up into several smaller bubbles. Changes in the length/radius ratio (L/R0) of the valve spool is an important factor affecting unsteady cavitation flow and pressure pulsation characteristics in the regulating valve. The length of travelling cavitation bubbles increases firstly and then decreases with increasing time. With the increase of the length/radius ratio (L/R0), the oscillation period of cavitation bubbles also increases. In the initial stage of cavitation bubbles, the velocity distribution inside the regulating valve is relatively stable, and no re‐entrant jet could be found although L/R0 is different. In the collapse stage of cavitation bubbles, the velocity distribution becomes extremely unstable because the collapsing cavitation bubbles affect the pressure drop and velocity field in the flow channel. Furthermore, the amplitude of pressure pulsation increases gradually, and the peak time of the pressure pulsation is gradually delayed while increasing the length/diameter ratio.

Influence of design anode and cathode channel on (PEMFC) fuel cell performance

Abstract This paper is focussed on investigating into some key aspects relating to design of an integrated proton exchange membrane fuel cell. Analysis of all currently utilised storage mediums was undertaken. High and low inlet operating conditions were studied by determining the distributions of over potentials, velocity distribution, and pressure. At entrance and exit, a triangle channel flow-field was predicted to have better polarization performance and regular velocity distribution, whereas for rectangular that was performance differences among four serpentine flow-fields for hydrogen and air. The pressure in the triangle channels were monitored to have periodically similar plots and showed important low pressure drop. Considering low pressure drop of rectangular channel has mid effect between triangles and rectangular for the same dimension that depends on the hydraulic diameter, these flow-fields would be beneficial for the larger inlet channel area. The study found that the triangular channels produced a better performance than their square flow path counterparts.

Cost of fouling in full-scale reverse osmosis and nanofiltration installations in the Netherlands

The economic impact of fouling in spiral wound membranes is not yet well explored. There has been an established assumption that the cost of fouling in membrane processes is significant, but this hypothesis has not been thoroughly evaluated. We conducted an economic analysis on seven full-scale installations, four nanofiltration (NF) and three reverse osmosis (RO), to estimate the cost of fouling in industrial plants. The cost of fouling was calculated in detail, including costs of increase in feed channel pressure drop, water permeability reduction, early membrane replacement, and extensive cleaning-in-place (CIP). The estimated cost of fouling was expressed as a fraction of operational expenses (OPEX) for each plant and the major cost factors in fouling and CIP costs were identified.The selected NF plants were fed with anoxic ground water, while the feed water to RO plants was either surface water or municipal wastewater effluent. All the NF plants produce drinking water, while the RO plants produce demineralized water for industrial applications. We found that the cost of fouling in the RO plants was around 24% of OPEX, while the fouling related costs in NF cases was only around 11% due to the low biofouling potential of the anoxic ground water. The major factor in the cost of fouling is the early membrane replacement cost, followed by additional energy and with only a minor contribution from the cleaning costs. The down-time cost (caused by the interruption of water production during a CIP event) can be the major CIP cost factor for the plants with frequent cleaning events, while the cost of chemicals dominates in the plants with non-frequent CIP. In case of manual cleaning-in-place, the cost of fouling is increased by around 2% for the RO plants with frequent CIP. The manual execution of CIP cleaning is an attention point to reconsider, as the reviewed plants hold an automated CIP cleaning, providing membrane productivity advantages.

Enhanced hydraulic cleanability of biofilms developed under a low phosphorus concentration in reverse osmosis membrane systems

A critical problem in seawater reverse osmosis (RO) filtration processes is biofilm accumulation, which reduces system performance and increases energy requirements. As a result, membrane systems need to be periodically cleaned by combining chemical and physical protocols. Nutrient limitation in the feed water is a strategy to control biofilm formation, lengthening stable membrane system performance. However, the cleanability of biofilms developed under various feed water nutrient conditions is not well understood.This study analyzes the removal efficiency of biofilms grown in membrane fouling simulators (MFSs) supplied with water varying in phosphorus concentrations (3 and 6 μg P·L-1 and with constant biodegradable carbon concentration) by applying hydraulic cleaning after a defined 140% increase in the feed channel pressure drop, through increasing the cross-flow velocity from 0.18 m s-1 to 0.35 m s-1 for 1 h. The two phosphorus concentrations (3 and 6 μg P·L-1) simulate the RO feed water without and with the addition of a phosphorus-based antiscalant, respectively, and were chosen based on measurements at a full-scale seawater RO desalination plant. Biomass quantification parameters performed after membrane autopsies such as total cell count, adenosine triphosphate, total organic carbon, and extracellular polymeric substances were used along with feed channel pressure drop measurements to evaluate biofilm removal efficiency. The outlet water during hydraulic cleaning (1 h) was collected and characterized as well. Optical coherence tomography images were taken before and after hydraulic cleaning for visualization of biofilm morphology.Biofilms grown at 3 μg P·L-1 had an enhanced hydraulic cleanability compared to biofilms grown at 6 μg P·L-1. The higher detachment for biofilms grown at a lower phosphorus concentration was explained by more soluble polymers in the EPS, resulting in a lower biofilm cohesive and adhesive strength. This study confirms that manipulating the feed water nutrient composition can engineer a biofilm that is easier to remove, shifting research focus towards biofilm engineering and more sustainable cleaning strategies.

Reducing the Impacts of Biofouling in RO Membrane Systems through In Situ Low Fluence Irradiation Employing UVC-LEDs.

Biofouling is a major concern for numerous reverse osmosis membrane systems. UV pretreatment of the feed stream showed promising results but is still not an established technology as it does not maintain a residual effect. By conducting accelerated biofouling experiments in this study, it was investigated whether low fluence UV in situ treatment of the feed using UVC light-emitting diodes (UVC-LEDs) has a lasting effect on the biofilm. The application of UVC-LEDs for biofouling control is a novel hybrid technology that has not been investigated, yet. It could be shown that a low fluence of 2 mJ∙cm−2 delays biofilm formation by more than 15% in lab-scale experiments. In addition, biofilms at the same feed channel pressure drop exhibited a more than 40% reduced hydraulic resistance. The delay is probably linked to the inactivation of cells in the feed stream, modified adsorption properties or an induced cell cycle arrest. The altered hydraulic resistance might be caused by a change in the microbial community, as well as reduced adenosine triphosphate levels per cells, possibly impacting quorum sensing and extracellular polymeric substances production. Due to the observed biofilm attributes, low fluence UV-LED in situ treatment of the feed stream seems to be a promising technology for biofouling control.

Thermodynamic analysis of absorption cooling system with LiBr-Al2O3/water nanofluid using solar energy

Together with the developing nanotechnology, nanofluids and nanoparticles are used as working fluid in energy applications. It is foreseen that nanoparticles have high heat conduction coefficient and it will increase system performance by using as a working fluid in energy systems. Many studies in the literature show that nanofluids increase the heat transfer rate by improving heat transfer. In this study, a performance analysis of an absorption cooling system using solar energy was performed as numerically. LiBr-Al2O3-water nanofluid has been used in the cooling system as working fluid. The thermodynamic values and calculations used in the analyses were performed with Engineering Equation Solver program. Heat load necessary for the generator is provided with a flat plate solar collector. For different operation condition, the variation of COP values was determined depend on Al2O3-water nanoparticle concentration ratio. When the Al2O3-water nanoparticle concentrations are changed as 0%, 0.5% and 0.1%, it was determined that the COP values increased. Nanoparticles added to the refrigerant at certain concentration values affects the COP values positively of cooling systems. Maximum COP value is 0.86 for 85 oC generator temperature and 0.1% Al2O3-water nanoparticle concentration. The lowest COP value was obtained for the 75 oC generator temperature. When the Al2O3-water nanoparticle concentration was increased together with the generator temperature, COP values also increased. When the nanoparticle concentration of the working fluid increases, the viscosity of the nanofluid can be increases. Due to, increased viscosity increases the pressure drop in the flow channel and the pump power required for the flow. Thus, minimum viscosity with maximum thermal conductivity optimisation in applications is very important.

Applications of Artificial Intelligence for Analysis of Two-Phase Flow in PEM Fuel Cell Flow Fields

Many electrochemical devices rely upon flow-field channels to transport reactants and byproducts to and from critical electrode sites. Often, this results in multiphase flow, which, if not managed properly, can reduce system performance and lifespan. In the case of PEM fuel cells, two-phase flow occurs when liquid water forms as a byproduct of the cathode oxygen reduction reaction. Over the past two-decades much effort has been invested to understand the nature of this stochastic transport phenomena and how to optimize flow-field design to mitigate its effects. This work probes the use of artificial intelligence to guide our understanding of complex relationships between liquid phase behavior and pressure drop in these reactant channels [1]. Using a supervised decision tree (DT) machine learning algorithm, the liquid distributions in ex-situ experimental reactant channels are correlated to the induced pressure drop for a range of air flow rates; this process is diagrammed in Fig. 1. Liquid phase data collection is accomplished by capturing 2D images via a CCD camera of liquid water distributions in a transparent flow channel simulating PEM fuel cell conditions. In this setup, liquid water is injected through the bottom of the GDL in the same manner that water must travel from the catalyst layer through the GDL to reach the channel an operating cell. Pressure transducers measure pressure drop along the length of the channel and are synchronized with the image collection. The images are processed for noise reduction and feature extraction and then used with pressure data to train a regression DT. The models produced using the DT show the potential to improve predictions over earlier methods that relied solely upon liquid area [2]. Overall, the DT models predicted pressure drop using liquid distributions with 88.9% to 98.54% accuracy and were able to capture complex behavior such as pressure changes due to the breakup of liquid slugs. In general, the methods may be valuable for attaining deeper understanding of the impact variables such as geometry, hydrophobicity, and system operating conditions have on two-phase flow pressure drop trends. Additionally, these models may serve as a validation tool for numerical efforts aimed at flow-field design. A reliable model that can accurately predict the liquid-gas two-phase flow pressure drop may be used to estimate the amount and potentially the distribution of the water content within the flow channel by inputting the pressure drop measured during the operation of the cell.Figure 1: Overview of the regression DT training and testing processes.[1] Santamaria AD, Mortazavi M, Chauhan V, Benner J. Two-Phase Flow Characterization in PEM Fuel Cells Using Machine Learning. In Meeting s 2019 May 1 (No. 30, pp. 1538-1538). The Electrochemical Society.[2] Hussaini IS, Wang CY. Visualization and quantification of cathode channel flooding in PEM fuel cells. Journal of Power Sources. 2009 Feb 15;187(2):444-51. Figure 1

Newly Developed Pediatric Membrane Oxygenator that Suppresses Excessive Pressure Drop in Cardiopulmonary Bypass and Extracorporeal Membrane Oxygenation (ECMO).

This article developes a pediatric membrane oxygenator that is compact, high performance, and highly safe. This novel experimental approach, which imaging the inside of a membrane oxygenator during fluid perfusion using high-power X-ray CT, identifies air and blood retention in the local part of a membrane oxygenator. The cause of excessive pressure drop in a membrane oxygenator, which has been the most serious dysfunction in cardiovascular surgery and extracorporeal membrane oxygenation (ECMO), is the local retention of blood and air inside the oxygenator. Our designed blood flow channel for a membrane oxygenator has a circular channel and minimizes the boundary between laminated parts. The pressure drop in the blood flow channel is reduced, and the maximum gas transfer rates are increased by using this pediatric membrane oxygenator, as compared with the conventional oxygenator. Furthermore, it would be possible to reduce the incidents, which have occurred clinically, due to excessive pressure drop in the blood flow channel of the membrane oxygenator. The membrane oxygenator is said to be the “last stronghold” for patients with COVID-19 receiving ECMO treatment. Accordingly, the specification of our prototype is promising for low weight and pediatric patients.

Performance Comparison of Spiral-Wound and Plate-and-Frame Forward Osmosis Membrane Module.

We compared two representative forward osmosis (FO) modules—spiral-wound (SW) and plate-and-frame (PF)—to provide practical information for the selection of FO element for a large-scale FO process. The FO operating performance of commercially available SW FO and PF FO was explored under different membrane area and flow rate conditions. The performance trend as a function of the membrane was obtained by adjusting the number of serially connected elements. Although SW FO and PF FO elements exhibited comparable feed pressure drops, SW FO demonstrated a significantly higher draw channel pressure drop than PF FO. Furthermore, the significant draw pressure drop in SW FO increased the draw inlet pressure, consequently limiting the number of serially connected elements. For example, the maximum number of serially connected elements for the normal operation was three elements for SW FO (45.9 m2) but nine elements for PF FO (63 m2) when the flow rate of 10 LMP was applied for feed and draw streams. Additionally, a footprint analysis indicated that SW FO module exhibited a slightly larger footprint than PF FO. Under investigated conditions, PF FO exhibited relatively better performance than SW FO. Therefore, this pilot-scale FO study highlighted the need to reduce the flow resistance of SW FO draw channel to take advantage of the high packing density of the SW element.

Optimization of anode performance in the ethanol electrochemical reforming for clean hydrogen production

Carbon-based fuel electrochemical reforming is considered as a promising hydrogen production method. Ethanol is one of the most appropriate carbon-based fuels. In this work, anode performance, especially the flow, ethanol electro-oxidization and energy consumption in the ethanol electrochemical reforming is numerically studied and experimental verified. Take the straight serpentine channel with square cross-section as a base structure in the electrochemical cell (EC), the effects of channel geometry and operating parameters are analyzed. Another five different configurations of flow channels, as well as another three different cross-sections are designed and explored. Results indicate that at the same cross-section area, the wider channel provides the higher effective area for proton transfer, and thereby improves the electrode reactions. The appropriate decrease of inlet velocity or increase of input voltage promotes the anode reaction and reduces the pressure drop in channel, while the operating temperature has the opposite effects on ethanol conversion and pressure drop. The arc channel is found optimal considering its highest ethanol conversion, although its pressure drop is a bit higher. The sector cross-section with uniform flow field distribution is found most favorable for the straight serpentine channel considering the ethanol electro-oxidization. These findings will favor the improvement of EC.

Two-phase flow characterization in PEM fuel cells using machine learning

This work presents a novel application of machine learning to identify the two-phase flow pressure drop in a flow channel of a proton exchange membrane (PEM) fuel cell. Liquid water management and flow-field design remain critical areas of research for many technologies including PEM fuel cells which are focused on in this work. Liquid water buildup in reactant flow channels can lead to parasitic pressure drop and performance degradation. To correlate liquid water distribution with a two-phase flow pressure drop, this work trains various machine learning models using images of water slugs in a flow channel to predict the pressure drop range. An ex-situ experimental setup was designed consisting of a single transparent PEM fuel cell channel with gas flow supplied by compressed air and liquid water production simulated using a digital injector feeding the gas diffusion layer side. A CCD camera monitored the liquid distribution from above while liquid–gas two-phase flow pressure drop was measured across the channel. Images were post-processed and used as input data to three machine learning models: Logistic Regression, Support Vector Machine and Artificial Neural Networks (ANN) to classify the images into three pressure classes: (i) pressure drop less than 15 Pa, (ii) pressure drop between 15 and 30 Pa, and (iii) pressure drop greater than 30 Pa. The performance comparison of machine learning models is reported using the confusion matrices and classification accuracy. ANN performed best for this application and resulted in 95% accuracy on both train and test datasets. This approach can be utilized to predict the pressure drop values in the flow channels of PEM fuel cells based on liquid water content distribution along the channel.