Understanding the influence of soot reactivity on layer break-up and particle structure detachment and transport during diesel particulate filter regeneration is essential for optimizing aftertreatment performance. Seven carbon blacks with differing reactivity were investigated as reactive particle model systems. The oxidation behavior was characterized by temperature-programmed oxidation, revealing a wide reactivity range, with the peak oxidation temperature spanning from 817 K for a propane-soot reference to 894-976 K for the carbon blacks. This paper examines the regeneration of a model filter channel in situ with high temporal and spatial resolution. The filter is loaded with 10 mg carbon black particles and then regenerated. The regeneration of the filter is analyzed by varying the particle system under constant regeneration conditions at a gas temperature of 823 K and a channel inlet gas velocity of 60 m/s. In addition, the layer height and temperature are varied for a selected carbon black, and a more reactive hydrocarbon mixture was added to the particle layer of the selected carbon black. In selected experiments high-speed imaging of the model filter channel enabled direct observation of layer break-up and particles detaching from the filters surface. Image-based analysis enables the quantification of the black surface area reduction and isolated particle structures. All carbon blacks showed a reaction of the carbonaceous particles, with little layer break-up and formation of isolated structures, as well as minimal detachment events (0-10 events per experiment). Introducing more reactive hydrocarbons to the particle layer markedly increased fragmentation and particle relocations to more than 500 events.

The interaction of liquid droplets with fibers is a key phenomenon in many technical applications, especially in coalescence filtration. While previous studies have addressed droplet detachment under various gas flow conditions and fiber surface properties, detailed experiments capturing the full dynamics of crossflowinduced detachment at high temporal resolution, particularly for oleophobic fibers, remain limited. In this study, we investigate the motion and detachment of individual oil droplets on a horizontally mounted, oleophobic stainless steel fiber exposed to transverse air flow at filtration-relevant velocities as low as 1.3 m s-1, representative of locally elevated velocities within fibrous filter media. Using high-speed imaging at up to 42345 fps, we resolve the droplet dynamics, including oscillation, deformation, and necking on millisecond time scales. The individual droplets of a standard compressor oil with volume-equivalent sphere radii between 276 mu m-749 mu m were placed on a single oleophobic fiber with a diameter of 80 mu m. The experiments are conducted in a viscous-inertial regime characterized by moderate Weber numbers (We approximate to 3-5) and relatively high Ohnesorge numbers (Oh approximate to 0.5-1). The results reveal high-frequency oscillations up to 160 Hz for small droplets (rd = 276 mu m) and show that droplet detachment proceeds via ligament elongation and pinch-off, forming only a few satellite droplets. For oleophobic fibers, the droplets adopt a clamshell-shaped configuration in axial view, whose blow-off dynamics are resolved here for the first time at high temporal resolution.

A prominent way to lower the required power for baghouse filter operation is the application of energy efficient filter media. Standards for filter testing of cleanable filter media (e.g. DIN ISO 11057) set a framework under which the differential pressure behavior of filter samples can be characterized - however there is no reference value to evaluate the application of filter media during operation of the test procedure based on energy demand. Hoflinger and Laminger proposed a suitable evaluation in the past where fan power and the energy demand for jet-pulse cleaning have to be considered. This publication further discusses the proposed energy evaluation within testing standards that set a framework consisting of Delta p-controlled filter regeneration during the measurement stage of a filter test. Different filter medium samples were investigated (spunbond, needle-felt and membrane). The impact of cycle time on the power for filter operation within Delta p-controlled regeneration strategies was investigated by inter-and extrapolation of the differential pressure behavior. For low cycle times, the power required for filter regeneration dominates the total power. For longer cycle times and thicker dustcakes, the contribution of the fan power to the total power increases while the power required for filter regeneration decreases and stable levels for the total power are reached yielding robust comparative values for different filter media. Furthermore, data from a single filter test can be extrapolated to identify suitable cycle times and estimate the potential of the filter medium considering operation at lower power and further mitigation potential for particle emissions.

Pulse-jet cleaning of surface filters at reduced absolute pressure is not yet well quantified. This work investigates pulse-jet regeneration at pabs=900, 100, 25, and 10hPa using a flexible spunbond (SB) medium and a rigid sintered granular medium (SG). Regeneration pulses were first characterized on clean media while varying valve opening time, tank pressure, absolute pressure and filter face velocity. The valve opening time primarily controlled pulse duration, whereas variation of tank pressure mainly affected the maximum differential pressure. At constant tank pressure and valve opening time, decreasing absolute pressure increased the maximum differential pressure and delayed the occurrence of this maximum. For comparable cleaning strengths across all absolute pressures, tank pressure was adjusted at each absolute pressure to approach a predefined target value of the maximum differential pressure. Filter cakes were formed using Mikhart MU 08 at similar mass loadings of approximately 50gm-2. High-speed imaging synchronized with the differential pressure signal reveals distinct medium dependent detachment dynamics. For SB, cake detachment starts close to flow reversal, whereas for SG detachment is delayed. Reduced absolute pressure shifts the timing and prolongs the duration of the cake detachment phase. Cleaning efficiencies decrease with decreasing absolute pressure for comparable pulse strengths, with a stronger sensitivity for SG than for SB.

The influence of pleat geometry on the operating behavior of cleanable surface filters was investigated using harmonized regeneration conditions and complementary modeling approaches. An adapted DIN ISO 11057 procedure was applied to isolate geometric effects for differently aged filter media by maintaining consistent regeneration intensity, raw gas concentration, and filter media face velocity across all geometries. The results demonstrate that under these harmonized experimental conditions pleat geometry influences the operating behavior from the first filtration cycle. The resulting differences in cycle time, residual differential pressure, and clean gas particle number concentration persist for the duration of the experiments. While aging leads to increased residual differential pressure and reduced particle emissions, similar trends across all geometries indicate comparable macroscopic aging behavior. No significant loss of effective filter media surface area was observed for the different geometries under the investigated operating conditions using a coarse, free-flowing test dust. A geometric model describing dust deposition on pleat sidewalls and in the pleat bottom suggests that the observed differential pressure development can be explained by the dust deposition inside the pleat in combination with the pleat geometry. In summary, the results show that increasing the pleat ratio leads to a shorter cycle time under the investigated harmonized conditions, and the geometry influence was evident from the very first cycle.

The dynamic behavior of a magnetizable single fiber exposed to a spatially homogeneous alternating magnetic field was investigated experimentally and numerically with the purpose of laying the foundations for subsequent contactless particle removal in gas-phase filtration. A Helmholtz coil arrangement was used to generate a well-defined AC magnetic field, which was characterized by Hall probe measurements and finite element simulations using COMSOL Multiphysics®. A ferritic stainless steel fiber clamped at one end was excited by the aligning magnetic torque induced by the alternating magnetic field with sinusoidal time dependence. High-speed imaging combined with automated MATLAB-based motion analysis enabled the determination of deflection, velocity, and acceleration. The steady-state response exhibits harmonic oscillation at the driving frequency (35–55 Hz) and can be accurately described by the analytical solution of a driven damped harmonic oscillator. A pronounced resonance peak was observed near the theoretically predicted natural frequency (f0,theo. = 49.2 Hz). The transient response reveals a superposition of the natural and driving frequencies, leading to beat phenomena for off-resonant excitation. The experimentally determined frequency response agrees well with the analytical model and confirms the interpretation of the distributed magnetic torque as an effective modal driving force. The results provide a consistent physical framework for magnetically induced oscillations of slender structures and form the basis for future studies on particle detachment from particle-laden magnetizable fibers.

Some gas particle separation processes operate at low absolute pressure, where surface filter media are widely used. Yet filter cake pressure drop behavior at these conditions remains underexplored. To improve understanding, we determined the filter cake pressure drop as a function of absolute pressure for filter cakes formed on two different surface media at a single filter face velocity. The filter cake pressure drop decreases as absolute pressure is reduced. We adapt a semi-empirical model validated for determination of clean surface media pressure drop to filter cakes and introduce a mean empirical correction factor to reduce the experimental effort. The approach captures the trends accurately, with the simplification limited to homogeneous filter cakes.

Current industrial dust emission limits and emission characterization in filter testing standards for cleanable surface filters are based on gravimetric (non-size-resolved) measurements. However, submicron and nanoscale particles often dominate emissions in terms of particle number concentration. This study provides a comprehensive analysis of the fractional separation efficiency and its evolution during dust cake formation over the full emission-relevant size range. Experiments were conducted in a filter test rig for cleanable filter media (VDI 3926 / DIN ISO 11057). Because particle penetration decreases rapidly as the dust cake develops, a suitable experimental methodology was developed. Fractional separation efficiency was measured using submicron test aerosols (dried NaCl and high-purity water) that do not contribute to cake formation or enhance filtration characteristics. Between these measurements, a separate phase using coarse test dust (Pural SB) enabled controlled cake formation, while concentration decay on the clean-gas side was monitored. Aerosol measurements covered a wide range from approximately 10 nm to 10 µm using a combination of light-scattering aerosol spectrometer, condensation particle counter, and scanning mobility particle sizer. The filter medium itself had a most penetrating particle size (MPPS) of 305 nm with an efficiency of 53.3% and an upper separation limit near 4 µm. Even small increases in differential pressure (e.g. from 100 Pa to 200 Pa) shifted the emission-relevant range completely into the submicron particle size range. At 500 Pa, efficiency reached 99.8% at a MPPS of 126 nm. The results demonstrate the particle-size dependent filtration kinetics and the mitigation potential of surface filters considering potential (ultra-) fine particle emissions.

The testing, characterization and evaluation of cleanable filter media for gas-cleaning applications are standardized in different standards and VDI guidelines. These procedures characterize flat filter media coupons and rely on specific test setups, methodology and experimental parameters such as the tank pressure for regeneration and filter media face velocity. However, if the geometry of the filter media deviates from the standardized flat coupon, characterization cannot be conducted in the test rig according to these standards. A common modification is pleating/folding the filter media, which allows for an increasing filter media surface area in a smaller installation space. Typically, pleatable filter media are evaluated using flat filter media coupons, with the results subsequently being applied to characterize pleated filter media. Direct characterization of pleated filter media in the pulse-jet-cleaned test-rig can prevent inaccurate conclusions transferred from flat to pleated geometries. To address this limitation, a novel filter holder was developed to enable the direct testing of pleated filter media in a standardized, pulse-jet-cleaned test-rig with harmonized experimental parameters for both media geometries. This allows the same experimental parameters and regeneration efficiency. The operating behavior was evaluated based on cycle time and residual differential pressure. The results show differences between the flat and pleated geometry under identical test conditions. These findings highlight the need for direct characterization of pleated filter media to accurately determine their performance.

The design of surface filtration systems operating at low absolute pressure remains underexplored, due to the lack of detailed studies on filter cake formation at such conditions. This study investigates both formation and flow resistance of filter cakes at varying absolute pressures after characterization of aerosol properties in the feed flow at reduced pressures. At constant operating conditions at the dosing unit, the particle number concentration decreases with decreasing absolute pressure due to gas expansion and dilution effects. In contrast, the number based particle size distribution remains largely unaffected across the investigated pressure range. At these conditions, filter cake formation at very low absolute pressures is significantly more time-consuming than at ambient conditions, two methods for cake formation were compared at constant filter face velocity: (A) cake formation and pressure drop measurement at the target system pressure and (B) cake formation at normal temperature and pressure followed by stepwise system pressure reduction and subsequent pressure drop measurement at a fixed flow velocity. The resulting cake resistances were compared using a specific filter cake constant CFC. The comparison revealed good agreement between both methods at each absolute pressure level, indicating that no structural changes occur in the filter cake due to absolute pressure variation. These findings demonstrate that method B is a valid approach to substantially reduce experimental time while maintaining structural comparability of the filter cakes. The results contribute to a deeper understanding of filtration behavior at vacuum conditions and provide a methodological basis for future investigations and process development in low pressure filtration systems.

Current industrial dust emission limits and emission characterization in filter testing standards for cleanable surface filters are based on gravimetric (non-size-resolved) measurements. However, submicron and nanoscale particles often dominate emissions in terms of particle number concentration. This study provides a comprehensive analysis of the fractional separation efficiency and its evolution during dust cake formation over the full emission-relevant size range. Experiments were conducted in a filter test rig for cleanable filter media (VDI 3926 / DIN ISO 11057). Because particle penetration decreases rapidly as the dust cake is formed, a suitable experimental methodology was developed. Fractional separation efficiency was measured using submicron test aerosols (dried NaCl and high-purity water) that do not contribute to cake formation or enhance filtration characteristics. Between these measurements, a separate phase using coarse test dust (Pural SB) enabled controlled cake formation, while concentration decay on the clean-gas side was monitored. Aerosol measurements covered a wide range from approximately 10 nm to 10 µm using a combination of light-scattering aerosol spectrometer, condensation particle counter, and scanning mobility particle sizer. The filter medium itself had a most penetrating particle size (MPPS) of 305 nm with an efficiency of 53.3% and an upper separation limit near 4 µm. Even small increases in differential pressure (e.g. from 100 Pa to 200 Pa) shifted the emission-relevant range completely into the submicron particle size range. At 500 Pa, efficiency reached 99.8% at a MPPS of 126 nm. The results demonstrate the particle-size dependent filtration kinetics and the mitigation potential of surface filters considering potential (ultra-) fine particle emissions.

Coalescence filters are highly effective for separating sub-micron oil mist particles from gas streams. In the past, the development of these filters has focused on understanding the propagation of oil within the filter and investigating the operational parameters and material properties of the coalescing material. This review examines studies on oil mist filtration, with a particular emphasis on drainage and entrainment processes. The analysis of single fiber–droplet interactions provides essential insights into droplet motion and detachment dynamics within fiber–liquid systems, considering forces such as gravity, drag, surface tension, and inertia, as well as factors like wettability and fiber orientation. The drainage behavior is regarded through internal and rear-side drainage mechanisms in porous material. Internal drainage mechanisms are examined by analyzing material properties and operational settings to understand the underlying processes. This includes examining internal pathways and flow dynamics within porous structures or composite materials. Rear-side drainage mechanisms are assessed by evaluating drainage on larger surfaces of porous material, focusing on the interaction between the liquid and the material system. Key factors include the thickness and distribution of the liquid film, the drainage rate influenced by saturation levels, and the impact of support structures and drainage layers. Entrainment mechanisms are considered based on factors influencing how droplets become entrained, emphasizing three primary mechanisms: gas flow over liquid surfaces, droplet detachment from fibers, and bubble bursting. This includes describing the forces and conditions affecting droplet and bubble dynamics. The entrainment in oil mist filtration investigations is characterized as well as the methods for measuring a wide range of entrained droplet sizes, including in-situ and offline techniques. This comprehensive review aims to compare various examination methods and their findings while identifying unresolved research questions in the field of oil mist filtration.

Combustion-related particulate emissions are a challenge to air quality and regulatory compliance. In modern combustion engines, wall-flow particulate filters effectively capture soot particles, whereby periodic high-temperature (02) regeneration or passive (NO2) regeneration is necessary to reduce the pressure drop. During regeneration, the soot layer breaks up, and small particle structures can detach and be transported further downstream towards the end of the filter channel. A Python-based image analysis workflow is presented for detecting and verifying particle structure detachments in high-speed video recordings of the filter regeneration. The method consists of two integrated modules using OpenCV and NumPy. In the first step, background subtraction (MOG2) and morphological operations are applied to identify candidate structures across video frames. The second step checks the particle structures detected in the first step, isolates a region of interest around the potential detachment and analyzes it using thresholding and pixel-wise difference mapping to confirm or reject the detachment event. Both modules allow parameters to be set and generate visual outputs for verification. The method was validated using a 796,000 frames dataset in which a model filter channel with carbon black loading was regenerated and six small detachment events (xeq ≈ 100 - 300 µm) were detected.• A Python-based method for detection of particle structure detachments in high‑speed videos of model filter regeneration.• Semi-automated two-step detection and verification of detachments.• Validated on 796 000 frames, reliably finding detachment events while reducing manual review time.

Atmospheric composition is an essential part of weather, climate and Earth system modeling. However, modeling atmospheric composition is a computationally expensive and time-consuming task that requires a significant amount of energy. As models scale to finer spatial and temporal resolutions, maintaining real-time performance becomes increasingly challenging. To address this, optimization and acceleration techniques are essential. One promising approach is the use of deep neural networks, which have demonstrated the capability to efficiently approximate complex systems with high accuracy. Predictions using these neural networks are notably faster compared to traditional methods, significantly reducing the computational burden. In this study, we present the development of a surrogate model designed to emulate ISORROPIA, a traditional model used for calculating the concentrations of chemical compounds in the ICON-ART (ICOsahedral Nonhydrostatic model with Aerosol and Reactive Trace gases) model. Specifically, ISORROPIA is an aerosol thermodynamic equilibrium model used by ART that requires substantial computational resources, occupying a significant portion of the overall calculation time, making it particularly well-suited for emulation. The methodology involved generating a comprehensive dataset using the traditional model, which served as the training data for the neural network. This dataset encompassed a wide range of chemical concentrations and conditions, ensuring the neural network could effectively learn the underlying patterns and relationships for real-life scenarios. A simple feedforward architecture was used and fine-tuned with the primary goal of maintaining a low approximation error while also striving to achieve the lowest possible inference timing. After training, the new neural network model was compared to ISORROPIA on ICON-ART simulation data. The results demonstrated that the neural network model successfully achieved the desired outcomes, maintaining low approximation error across the globe and efficient inference timing.

In the past 5 years, data‐driven prediction models and Machine Learning (ML) techniques have revolutionized weather forecasting. Meteorological services around the world are now developing ML components to enhance (or even replace) their numerical weather prediction systems. This shift creates new challenges and opportunities for universities and research centers, calling for a much closer cooperation of meteorology with mathematics and computer sciences, updates of teaching curricula, and new research infrastructures and strategies. To address these challenges, an interdisciplinary team of scientists from the Karlsruhe Institute of Technology (KIT) and the German Meteorological Service (DWD) created the TEstbed for Exploring Machine LEarning in Atmospheric Prediction (TEEMLEAP). Implemented on KIT's supercomputer HoreKa, the TEEMLEAP testbed simulates the entire operational weather forecasting chain using ERA5 reanalysis data as pseudo‐observations and DWD's Basic Cycling environment for conducting assimilation‐prediction‐cycling experiments. Moreover, first steps are taken toward the integration of new data‐driven components like FourCastNet and ML‐based post‐processing methods. The TEEMLEAP testbed allows systematic investigation of a wide range of issues related to weather forecasting such as optimizing the observational system, uncertainty quantification, and developing hybrid systems that integrate ML with physics‐based models. This document outlines the testbed's setup, demonstrates its functionality with a pilot experiment, and discusses examples of potential applications. Future plans include creating educational modules and developing a higher‐resolution regional version of the testbed that could be used for assimilating field campaign observations.

In mist filtration, fiber-based coalescers are an established form of filtering droplets contained in mist. The filtration process can be divided into different process steps, describing the impact of the droplets on fibers, the formation of fluid structures and the liquid transport. In order to investigate mechanisms inside depth filters on a microscopic level, investigations are often reduced to single fibers. In this work, the coalescence and transport mechanisms of axially distributed water droplets on a vertical fiber, subjected to gravity, are reported. This is done with the latest high-speed camera technology commercially available. Automated tracking of droplets is used for a frame-by-frame investigation of droplet position, size, and oscillation. Coalescence mechanisms describe the process of fluid formation. The first observed coalescence mechanism is identified by the coalescence of droplets contained in mist with an adhering droplet at the fiber. The second coalescence mechanism describes the coalescence of two closely spaced sessile droplets on a fiber. As a result, the newly formed droplet oscillates and can begin to drain. Furthermore, the coalescence process of a draining and sessile droplet is reported. Both the draining droplet and the second droplet involved in the coalescence event can be subject to oscillation. The given temporal high-resolution information about the droplet position and deformation improves the understanding of droplet coalescence on fibers and by that also its influence on transport processes and therefore droplet drainage. Observed transport mechanisms which take part in the drainage process are gravitational draining, droplet bouncing, and droplet sweeping.

HYPOTHESIS:Adhesion between particles and a filter fiber is an important process of the filtration as it dictates the process of separation and in the following the detachment process of particles during filter regeneration. In addition to the shear stress that a new polymeric stretchable filter fiber implements into the particulate structure, the elongation of the substrate (fiber) is also expected to cause a structural change in the surface of the polymer. Thus, the changed contact area and surface energy could affect the adhesion force between particles and fibers. EXPERIMENTS:Systematic measurements of adhesion forces between a single particle and the stretchable substrate were performed using Atomic Force Microscope (AFM). The substrate surface characteristics (roughness) was changed directly beneath the modified measurement head using piezo-motors to achieve stepless elongation state. Polystyrene particles and particles made of Spheriglass were applied. FINDINGS:In the experiments, a reduced adhesion force between the particles and the filter fiber was found for a new high range of substrate roughness and peak-to-peak distance, in which the Rabinovich model has not been used before [1]. Further, the influence of high and low energy surface particulate material was evaluated to understand the detachment process in the new real adaptive filter and in DEM-simulation.

Additive manufacturing processes are gaining increasing attention in various gas cleaning applications since new process optimizations can be achieved (e.g. higher particle separation efficiency, lower process energy consumption and less entrainment of secondary aerosol into the clean gas). When processing metals by grinding, brushing and polishing, submicron metal particles as well as airborne liquid droplets (e.g. oil droplets) may occur. These contaminants in the process air are undesired and need to be separated from the gas stream. Hence, new technologies are aiming to separate these submicron particles at a high particle separation efficiency and a comparatively low pressure drop. By utilizing additive manufacturing processes in gas cleaning applications, novel 3D printed nozzle geometries for innovative wet separators can be applied. Using those nozzles as pre-separator, the service life of downstream filter elements can be extended. Regarding the separation of airborne oil droplets from a gas stream, highly efficient coalescence filters are required. However, entrainment of secondary aerosols into the clean gas may ruin the overall filtration efficiency. Therefore, novel 3D printed support structures on the filter downstream side are developed in order to reduce this secondary aerosol, by applying e.g. drainage channels on the filter rear side to accelerate the drainage of oil. In this contribution the implementation of additive manufactured parts in gas cleaning applications, exemplary for an innovative 3D printed wet-scrubber nozzle and novel 3D printed support structures for oil mist filters, are presented.

Particulate filters are used as a standard component in the exhaust gas aftertreatment of vehicles. The reactive (soot) and inert (ash) particles generated during engine operation are deposited in wall-flow filter. The resulting particle layer increases the differential pressure of the filter, which is why it is regenerated regularly. During regeneration of the filter, the reactive particles oxidize, and the inert particles remain in the filter. The oxidation of the soot particles results in a layer break-up, and the resuspension of particle structures can occur. The layer break-up over the entire length of an inlet channel and the resuspension of particle structures have not yet been observed, which is why the fundamental processes in a particulate filter have not yet been fully clarified. In these investigations, the regeneration of a single wall-flow filter channel is observed in situ with high temporal and spatial resolution. For this purpose, the filter is loaded with soot particles and regenerated subsequently. The regeneration of the filter is analyzed in relation to the process parameters of temperature, layer thickness, and flow velocity. Before the visual layer break-up, the pressure drop decreases and declines to a constant value before resuspension of particle structures are detected. As the temperature is increased, the regeneration time is reduced. With a thicker particle layer, the particle structures formed during layer break-up become larger, the location of resuspension shifts to the posterior channel region, and the number of resuspensions increases. A higher flow velocity causes more particle structures to be resuspended and transported to the channel end.

Nonwoven coalescence filter media are often used in real-world applications to separate droplet aerosols from a gas stream. While mechanisms like droplet deposition, liquid transport and the evolution of the pressure drop are well understood, information on separated liquid structures formed during filtration is scarce. However, this information is important to understand the initial state of the coalescence filtration process, more specifically, the transition from individual deposited droplets to larger oil structures and subsequently the effect of these oil structures on the separation efficiency. In this work, deposited oil structures in the micrometer region of an oleophilic filter medium of different oil loading stages (saturations) are presented for the first time by utilizing micro-computed tomography (mu-CT) and several post-processing steps. The local (determined via mu-CT) and global (determined gravimetrically) porosities as well as local and global saturations are compared and evaluated. While local (Phi(local) = 0.9575) and global (Phi(global) = 0.9588) porosities are in good accordance, significant deviations in the saturation are observed. The local saturation fluctuates significantly at the highest investigated oil loading stage. Especially at high global saturations, local areas are completely saturated with oil. Other areas are weakly saturated, which are particularly relevant for the gas flow to pass through the filter media. Furthermore, a novel routine to distinguish between different coalesced oil objects (e.g. oil droplets or oil sails) is developed and presented. It is shown, that the majority of the deposited oil (> 80%) is stored in oil sails between adjacent fibers. Oil droplets surrounding fibers are found to be growing in size with increasing oil loading time. In addition, the volume fraction of the deposited oil in the direction of filter thickness shows, that most of the oil is deposited on the upstream side at the first 80% of the filter.