Neutron tomography of particulate filters: a non-destructive investigation tool for applied and industrial research

Abstract

This research describes the development and implementation of high-fidelity neutron imaging and the associated analysis of the images. This advanced capability allows the non-destructive, non-invasive imaging of particulate filters (PFs) and how the deposition of particulate and catalytic washcoat occurs within the filter. The majority of the efforts described here were performed at the High Flux Isotope Reactor (HFIR) CG-1D neutron imaging beamline at Oak Ridge National Laboratory; the current spatial resolution is approximately 50 μm. The sample holder is equipped with a high-precision rotation stage that allows 3D imaging (i.e., computed tomography) of the sample when combined with computerized reconstruction tools. What enables the neutron-based image is the ability of some elements to absorb or scatter neutrons where other elements allow the neutron to pass through them with negligible interaction. Of particular interest in this study is the scattering of neutrons by hydrogen-containing molecules, such as hydrocarbons (HCs) and/or water, which are adsorbed to the surface of soot, ash and catalytic washcoat. Even so, the interactions with this adsorbed water/HC is low and computational techniques were required to enhance the contrast, primarily a modified simultaneous iterative reconstruction technique (SIRT). This effort describes the following systems: particulate randomly distributed in a PF, ash deposition in PFs, a catalyzed washcoat layer in a PF, and three particulate loadings in a SiC PF.

Introduction

To capture the biologically and environmentally harmful particulate that has been a staple of diesel powered vehicles, stringent emissions regulations were enacted in 2007. This led to the implementation of particulate filters (PFs or DPFs for diesel vehicles) on every on-road diesel vehicle. Soot and other particulates are deposited in the PF as the gases flow through the porous wall and out the exhaust channels. Unfortunately, with the implementation of emissions control devices, there is a fuel penalty and lost efficiency. When the PFs trap soot, they must periodically go to high temperatures to regenerate or oxidize the soot, which requires additional fuel injection—greater than 4% fuel penalty has been reported solely for PFs [1]. While this is currently the case, it is recognized that the capture and regeneration process is not well understood and thus not optimized for fuel economy. Specifically, it is not well known how soot is distributed in the PFs during loading, how this distribution changes during regeneration and how much fuel is required to adequately regenerate the PFs.

An additional issue associated with PF operation is the accumulation of non-regenerable “ash” components [2], [3], [4], [5], [6], [7]. These particulates are primarily associated with lube oil consumption and the metal-containing additives that form metal oxides/sulfates/carbonates during combustion. They can also be traced back to matting material that is employed to hold and cushion the PFs and upstream catalysts within their metallic cases. Over the lifetime of the vehicle this ash particulate accumulates within the PF and disrupts the soot accumulation profiles. Understanding how this accumulation proceeds with vehicle lifetime is important for both PF sizing and regeneration timing, specifically since it is necessary to avoid high concentrations of soot in a section of the PF, as this could lead to very high local temperatures and possibly PF mechanical failure or cracking.

Typical methods of studying this behavior employ destructive measurements that limit sequential studies of the devices while they are still functional. Additionally, the mechanical action of the cutting process can disturb the position and density of the particulate layers and aggravate the materials being investigated. A non-destructive method for analyzing low-density depositions and faults in PFs would accelerate research efforts and improve the quality of PF investigation. With these concerns in mind recent efforts have turned to non-destructive techniques based on X-ray [8], [9], [10], [11], tera-hertz [12], and neutron tomography [13], [14], [15], [16]. Each of these techniques has their advantages and disadvantages, and all heavily rely on sophisticated computerized reconstructions of the recorded radiographs to achieve high-quality representative 3D data sets. This effort builds on our earlier study [13] that first describes our interest in using neutrons to image PFs and continues the investigation that was started by others [17], [18], [19]. Our previous collaborative study [13] relied on neutron imaging at the ANTARES neutron imaging facility of the Forschungs-Neutronenquelle Heinz Maier-Leibnitz (FRM II) at the Technische Universität München (TUM), while this effort describes the further characterization of these images and the utilization of this technique at the Oak Ridge National Laboratory (ORNL).

Neutrons are very sensitive to light elements such as hydrogen (H) atoms and can penetrate through thick layers of metals. These two properties ensure neutrons are well suited to probe relevant transportation technologies such as diesel particulate filters, exhaust gas recirculators (EGRs), and fuel injectors [20], [21], [22], [23]. Moreover, investigations of engines in operation have permitted real-time observations of oil movement in the engine [24]. Neutron imaging is based on the measurement of a beam attenuated by a sample. The attenuation is caused by absorption and scattering of neutrons within the sample. A two-dimensional position-sensitive detector placed behind the sample can measure the transmitted neutron flux, as illustrated in Fig. 1.

The neutron beam (of wavelength λ) attenuation caused by a homogeneous uniformly thick sample composed of a single isotope is given by

$$I(\mathit{\lambda })=I_0(\mathit{\lambda })e^{-\mathit{\mu }(\mathit{\lambda })\Delta x}$$

where I₀ and I are, respectively, the incident and transmitted beam intensities, μ is the attenuation coefficient, and Δx is the thickness of the sample. The attenuation coefficient μ is given by

$$\mu (\lambda )={\sigma }_t(\lambda )\frac{\rho N_A}{M}$$

where σ_t(λ) is the material's total neutron cross section, ρ is its density, N_A is Avogadro's number, and M is its molar mass. Neutron computed tomography measures attenuation of a neutron beam in three dimensions by rotating a sample to record attenuation for multiple beam paths through an object. Computational reconstruction permits analysis of the sample in three dimensions.

In the presented research study, neutron sensitivity to hydrogen is used to localize in 3D the particulate matter deposit in 25 mm diameter×76 mm tall wall-flow particulate filters composed of cordierite (Mg₂Al₄Si₅O₁₈) or silicon carbide (SiC) material, with and without a catalyzed washcoat. The PFs, washcoats, and the particulate (both carbonaceous soot and metal sulfate containing ash) are all hygroscopic, and the amount of ambient water they adsorb is relative to the accessible surface area. This water is the major contributor in neutron contrast in the data. Thus it is possible to distinguish carbonaceous particulate/soot from cordierite due to the higher surface area of soot. A similar argument holds true for ash and washcoats. Neutron computed tomography (nCT) was performed to identify exact distribution of deposits in the filters and to permit comparison with conventional destructive laboratory investigations.