What flow visualisation can teach us about reactor design What? Flow visualisation can teach us about reactor design? Hugh Stitt [1] & Peter Jackson [2] [1] [2]

Outline In research –Laboratory experiments, –Model development Scale up –Role of flow visualisation –Measurement density Flow visualisation in the field –Reactors behaving badly –Knowledge vs. information vs. data –Implementation

Stirred Tank Tomography in 4D at Medium Scale 3 m 3 demonstration scale mixing tank with 8 planes of electrical sensors R Mann et al., Chem Eng Sci, 52, (1997) –Sensor readings reconstructred to give resistivity map

Stirred Tank Tomography in 4D Video frame and tomogram showing tracer distribution after 3 secs R Mann et al., Chem Eng Sci, 52, (1997)

Stirred Tank Tomography in 4D Video frame and tomogram showing tracer distribution after 3 secs R Mann et al., Chem Eng Sci, 52, (1997) This is great – good picture!! – But gives little quantitative information on mixing UNLESS we have a model to compare it with

Getting High Quality Information on Stirred Tanks Needs a Lagrangian experimental approach –Velocimetry – or particle Tracking Positron Emission Particle Tracking (PEPT) Computer Automated Radioactive Particle Tracking (CARPT )

Lagrangian Measurements on a Stirred Tank Loop circulation patterns are severely averaged Actual fluid motion is far more random –Direction & velocity Velocity Trajectory Fishwick, Winterbottom & Stitt

Lagrangian Measurements on a Stirred Tank Fishwick, Winterbottom & Stitt CARPT on 8" dia vessel PEPT on 4" vessel Rammohan, Kemoun & Dudukovic

Radioactive Velocimetry on a Rushton Turbine Agitated Baffled Vessel Time-averaged velocity plots

Radioactive Velocimetry on a Rushton Turbine Agitated Baffled Vessel Time-averaged velocity plots Strength of these spatial velocity data – they can be compared directly to simulations Great pictures!! – But they give little quantitative information on mixing UNLESS we have a model to compare it with the data

Stirred Tank Experimental vs Simulation Velocity Vectors Both give recirculation loop centres at –Upper loop : 0.575, –Lower loop : 0.225, Rammohan, Dudukovic & Ranade: IECRes 42, 2589 (2003)

Stirred Tank Experimental vs Simulation Turbulent Kinetic Energy Rammohan, Dudukovic, Ranade: IECRes. 42, 2589 (2003) Model quality reduced for derived value

Optical techniques not appropriate –Need penetrative methods; eg.  -rays –Flow visualisation in highly dispersed multiphase operation Understanding of instantaneous effects Valuable data for comparison to time averaged models CREL Velocimetry in Multiphase Bubble Column Operation

Gas Sparging in a Stirred Tank Radioactive Techniques allow interrogation at high hold up of dispersed phases Effect of gas sparging on liquid velocities –PEPT data Gas hold up patterns in a sparged stirred tank –  -CT data No gas Gas sparged Fishwick, Winterbottom & StittRammohan & Dudukovic

Tomography & Velocimetry in Multiphase Flow Reactors Modelling of multiphase reactors is subject to many uncertainties –Multiphase flow regime: bubbly, unstable –Coalescence - redispersion Population balance: bubble class models –Momentum transfer –CFD “Closures” Require validation of models against detailed experimental data

Tomography on a Bubble Column Electrical Resistance Tomography Computer Tomography (  -ray) Williams, Wang et al, Leeds Univ, UKAPCI / CREL data Temporal resolution – but uncertain spatial precision Time averaged – good spatial resolution Both have been done on columns  18" diameter

MRI – TBR Trickle-Pulse Flow Transition Trickle regime 1.4 mm/s Pulsing regime L = 13.3 mm/s Transition regime 4.6 mm/s Gas flow: mm/s Resolution: 0.7×1.4 mm Acquired at 50 f.p.s. All presented on the same intensity scale Lim, Sederman, Gladden, Stitt, Chem Eng Sci, in press Flow transition is a local phenomenon. Specific information on pulsing, its origin and the bed structures that promote it

Flow Visualisation in the Laboratory Range of techniques available for use with multiphase systems –  -ray, X-ray, Electrical, MRI Varying cost, spatial and temporal resolution Important role in building models and fundamental understanding –Specific information on flow regimes –Model discrimination and validation Next question Next question –How do we exploit these techniques in scale up and design ?

“The bench scale results were so good that we by-passed the pilot plant”

Design and Scale up Role of Flow Visualisation Experimental tomography and velocimetry have a clear role in reactor design and development –Quantitative information for model validation –Qualitative role in understanding flow behaviour and phase interactions –Quantitative evaluation of changes in mixing / hydrodynamics behaviour with changes in scale

Low Cost Radial Flow Packed Bed Proof of Concept High pressure processes Ammonia synthesis –Low  P at a premium Radial flow benefits –High cost engineering retrofits available –But a very cost sensitive industry Can radial flow be induced by directed packing? Header Space Feed Feed distributor Large dia. inert packing Smaller dia. catalyst Exit collector (porous wall) Exit flow

Low Cost Radial Flow Packed Bed Flow Modelling Radial flow patterns predicted using CFD Process gas conditions and flow –Based on assumptions of global packed bed permeabilities But are these predictions correct and realistic ? –Use Electrical Resistance Tomography Bolton, Hooper, Mann & Stitt:, Chem Eng Sci, 59, (2004)

Low Cost Radial Flow Packed Bed Experimental Validation with ERT Electrical Resistance tomography –4D resolution Low spatial resolution Use 36" diameter vessel –Packed aspect ratio 1:1 –Annular configuration, 2 particle diameters Central collector –8 planes of 32 electrodes Injection of concentrated brine tracer and monitor conductivity –Reconstruct conductivity maps Bolton, Hooper, Mann & Stitt, Chem Eng Sci, 54, 1989 (2004)

Radial Flow Packed Bed ERT Flow Pattern ERT provides demonstration of overall axial / radial flow profile Bolton, Hooper, Mann & Stitt, Chem Eng Sci, 54, 1989 (2004) Reconstructed conductivity maps at single horizontal plane for 8 different times

Low Cost Radial Flow Packed Bed Quantitative Validation Velocity mapping from ER tomography CFD simulation of experiment Bolton, Hooper, Mann & Stitt, Chem Eng Sci, 54, 1989 (2004) Qualitatively reproduces main features –Quantitation is less conclusive

What? Flow visualisation can teach us about larger scale operation? Scale up –Use measurement system and measurement density appropriate to validation of design concept and models Does not need same precision as lab scale. Objective different – Justification of scale up protocol – Testing of models at increased scale – NOT fundamental understanding and derivation of models per se But what about manufacturing scale ?

Tracking of fluid movement –within and between oil and gas reservoir wells during drilling and production. Examination of transfer pipelines to and from processing facilities –for slugging effects, phase flow rates, solids build-up or blockage, pigging operation monitoring. It’s only one dimensional and single pass but ……….. it is an invaluable technique Priority list : 1) Is there a blockage? 2) If yes, then where is it? 3) Then characterise the blockage Tomography & Velocimetry in the Field Large Scale Particle Tracking : An old technology

Reactors Behaving Badly Stirred Tank Reactors Liquid level below top impeller Impeller damage makes good mixing impossible

Pelleted catalysts –Shallow bed (4") –Large dia (8´) Reactor operating at reduced conversion Observation (through spy glass) indicates “dark patches” Reactor Behaving Badly Catalytic Oxidation Reactor

“Field” Particle Tracking Technology? What are the objectives ? –Detailed diagnosis of flow patterns with high spatial resolution ? But how high a spatial resolution is required? Customer requirement –Measuring the degree of mixing with sufficient resolution to establish: overall quality of mixing and any severe maloperation at minimum cost –Do mixing and flow patterns adversely affect production and profits? –————————————————————— ————————— ——————————————————————

Modality for “Field” Operation Key requirements for field and research use are not the same

Modality for “Field” Operation Currently - only  -ray systems meet all the requirements for field use

BUT : Cost vs. Information is exponential The 80 : 20 Rule 80% of the information is only 20% of the cost –And that 80% is normally sufficient to make an educated decision or diagnosis Corollary : the remaining 20% of information requires an additional 80% of the total effort Cost vs number of data points may be linear

“Field” Tomography Technology? What information are we trying to obtain ? –And at what level ? High levels of information cost money & time Diagnosis of good, adequate or poor operation can often be done with little measurement and information –Provided you know what information or data to measure ……. & how to interpret it Detailed measurement will only be done in the field where it is essential –Where it adds value Hence - if an operator can get enough information to understand what he critically needs to know by a 1D,  1m measurement –Then he won’t pay for more!!!

Reactors Behaving Badly Steam Reformer Not too good Not good at all

Reactors Behaving Badly Steam Reformer Tube wall temperature surveys can be used routinely to identify zones of misbehaviour –Use Gold Cup Pyrometry Zone of hot tubes –Operator needs to trim burners to avoid premature tube failure And the resulting cost penalty But here we’re lucky. We have observation windows to look through

Dignostics and Tomography at Scale A Case Study Pilot plant slurry bubble column reactor, –18” diameter, heat exchange tube internals,

Base line scan - Densitometry Two successive sets of scans - Data are nearly identical showing good reproducibility

Field Measurements on a Slurry Bubble Column Reactor –18” diameter, heat exchange tube internals High number of detectors / scans required to achieve spatial resolution –Very long time (thus high cost) to collect statistically significant data set Internals effect “lines of sight” Very complex reconstruction Calibration during operation? Questionable value proposition –Consider an alternative approach

Gas Inlet Slurry outlet Gas Outlet Detector 2 Detector 1 Tracer Study - Application Example 1 Slurry Bubble Column Open Tracer Studies –For axial mixing and entrainment measurements Inject gas tracer at gas inlet. –Responses from detectors 1 & 2 gives mean residence time, Axial mixing information –Use third detector at slurry outlet to measure gas carryover

Tracer Study - Application Example 2 Slurry Bubble Column Open tracer studies with ring detectors –Investigate phase distribution and mixing –Tracers Catalyst particles – doped with Mn 56 2 O 3 “Liquid follower” : – powdered Mn 56 2 O 3 Open gas tracer : Ar 41 Gas Inlet Slurry outlet Gas Outlet –Use of more than one ring allows measurement of rise velocities

Particle Tracer Studies on a SBCR Install several rings of collimated detectors Use pulse injection of active particle tracers –“Liquid” –Catalyst - Pilot plant operated by Air Products - Tracking particles prepared by JM - Data measurement by JM-Tracerco - Data interpretation by CREL,

Particle Tracer Studies on a SBCR Catalyst and “liquid follower” particles show almost identical behaviour –Assumption of pseudo-homogeneous slurry phase is valid

Particle Tracer Studies on a SBCR Pulse injection of multiple particles and ring detectors used in lieu of single Lagrangian trace or tomography –Simpler to install, calibrate and use Ring detector responses compared to model predictions –In general - good comparability –Demonstrates model validity OR....If we have a model that predicts behaviour then we can assess any deviation from that ideal using simpler (tracing) techniques

Pelleted catalysts –Shallow bed (4“) –Large dia (8´) Reactor operating at reduced conversion Observation through (spy glass) indicates “dark patches” Modelling –Local extinction of catalyst and stable “cold channels” with steep thermal gradients With very high mass flow Reactor Behaving Badly Catalytic Oxidation Reactor Hot (active) catalyst) Dark patches

Catalytic Oxidation Reactor CFD modelling of gas distribution system and head space indicated no problem If modelling is correct (catalyst extinction and cold flow channels) ….. –Would expect massive mal-distribution of gas flow Significantly higher flow though cold zones Hot (active) catalyst) Dark patches

Evaluation of Flow (mal)Distribution Through a Packed Bed Reactor Flow distribution study using –Open 85 Kr tracer –Ring of detectors just above catalyst bed Detectors were not colliimated

Reactor Flow Distribution using Tracer Typical test trace Inlet detector response Ring detector responses – showing significant differences

Reactor Flow Distribution Using Tracer Flow distribution by Segment High response at locations of persistent dark patches - Consistent with model Unexpected area of low flow Repeat runs, and detectors at bottom of catalyst bed all gave similar results

Flow Visualisation in the field High measurement density not appropriate –Financial considerations Information rich data, with few measurements feasible based on –Selecting appropriate measurements Not necessarily the same as in the lab – Open tracers, chordal scans, ……… –A priori knowledge of what results represent poor / bad behaviour –Availability of models to interpret data and relate to lab-based understanding Validation of model scalability

But sometimes we need a “map” Development of Tomography for Field Use A portable  -ray tomographic toolkit For process diagnostic application on steel vessels –Robust & portable. Accurate, repeatable & quick to analyse, Non-intrusive and non-invasive, Easy to install & remove. Economic Experimental & Methods –Steel vessel, thin walled, 40 cm diameter Source : 137Cs : 662 keV –Use of Phantoms Steel bar, tube and plate, Hollow polystyrene block –Ab initio reconstructions From calculated line densities Darwood et al., WCIPT3, Sept 2003

Densitometry : Results for Dual Phantoms Experimental 20 x 8 grid Theoretical 40 x 4 grid Ghost images on both experimental and theoretical reconstructions –Grid scanning not able to discriminate multiple features at low numbers of scans Steel Pipe Steel Plate

Fan-beam Tomograms of Phantoms Drilled polystyrene block 32 nodes x 6 scans Pipe & plate dual phantom 32 nodes x 6 scans Tomograms show good representation –Note absence of ghost images on tomogram of dual phantom

 -ray Computed Tomography Scanning Imaging of Process Vessels & Reactors “Fan beam” arrangement of sensors Use multiple source positions 6.2 m dia. packed column –32 source locations –6 scans per position

Tomography of Commercial Units Tomography can be done on commercial units with reduced number of scans –Scale limited by  -ray attenuation Particle tracking also feasible but issues on tracer retrieval 6.2 m dia fractionation column 1m dia FCC Riser

What? Flow Visualisation Can Teach us about Reactor Design and Operation ? Research –Building fundamental understanding Model building, discrimination and validation Requires high density of measurements Scale up and Design –Objective to test the model at the larger scale Lower measurement density probably adequate Manufacturing scale –Objective is diagnostic Good operation or not: is it a financial burden? Even lower (single point?) measurement may suffice

What? Flow visualisation can teach us about reactor design and operation