Fluidmapper: measuring flow inside opaque vessels with RPT-AI

What Fluidmapper measures and why it matters

Fluidmapper is a Montreal company founded by Jocelyn Doucet, Ph.D., a chemical engineer trained at Polytechnique Montréal [S6]. It measures how fluid moves inside industrial vessels with radioactive particle tracking (RPT) and AI reconstruction, which it calls RPT-AI [S1], and delivers the result as flow data built from a customer's CAD files and operating conditions [S13].

The problem is easy to state. Inside a stirred tank or reactor the flow is three-dimensional and turbulent, and in most industrial equipment it cannot be seen [S1]. The walls are steel, the fluid is an opaque slurry, or the liquid is full of gas bubbles. Mixing, heat transfer, gas dispersion and scale-up all depend on that flow.

Engineers have three ways to learn about it, each with a weak point. Simulation (computational fluid dynamics, CFD) is only as good as its turbulence, rheology and multiphase models and the experiments that validate them [S9]. Optical measurement needs a transparent system, and most experimental techniques return two-dimensional structures, limited coverage and no quantified uncertainty [S7]. Pilot campaigns are long and expensive [S10]. Validation data is scarce. Most available experimental datasets are simplified, two-dimensional or proprietary, so new solvers are often benchmarked against other simulations [S11], and AI models trained on CFD output inherit its assumptions [S12].

Fluidmapper's answer is to measure instead of model: three-dimensional velocity fields from real experiments, with confidence intervals attached [S8]. It is plain about the boundary too: the measurements complement CFD, and are best used to validate and refine simulation models, not to replace early design work [S3].

The science: how RPT-AI works

One tracer, tracked continuously

RPT follows a single particle instead of imaging the whole fluid, a method developed for multiphase reactors in research groups that include Professor Jamal Chaouki's at Polytechnique Montréal [K1, K2, R1]. The tracer is made to match the density of the phase under study, so it moves the way that phase moves, and it is activated in a nuclear reactor so that it emits gamma rays [S1]. Gamma rays pass through metal and opaque material, so scintillation detectors placed around the equipment can locate the particle with no optical access [S1]. In the published test rig the tracer was a 3.5 mm scandium-46 particle coated to match the fluid's density, and the detectors were sodium iodide crystals, nine around the tank and one under it [K12]. Its Stokes number of 0.006 means, in the authors' reading, that it follows the fluid with high fidelity [K12].

A measurement runs in four stages [S1]:

1. Geometry. The key parts of the vessel (shell, inserts, baffles, nozzles) are 3D printed from the customer's CAD files. The customer supplies fluid properties and operating conditions.
2. Calibration. A robotic arm moves the tracer through the whole geometry. Millions of measurements at known positions train the reconstruction model, and validation measurements map precision and error in three dimensions. This takes about two days.
3. Measurement. The tracer moves freely for about 24 hours, with its position recorded every 10 milliseconds. If the geometry is unchanged, new experiments can start at once, up to 28 from a single calibration [S8].
4. Analysis. The trajectory becomes velocity fields, pumping rates, mixing dynamics and near-wall velocity distributions.

Run time depends on how fast the tracer visits the whole flow. The method rests on ergodicity: given enough time, one particle samples every region, so observations over time describe space. Fluidmapper watches coverage and statistical convergence instead of a fixed clock [S7]. The published runs lasted 4 to 24 hours [K12].

What the AI reconstruction adds

Classic RPT converts detector counts to a position with a physical model of how gamma rays travel from tracer to detector, with parameters fitted for each setup [K11, P5]. It is calibrated with the tracer held still, but in an experiment the tracer moves, and the mismatch is called dynamic bias [K12]. The model also assumes a uniform attenuation coefficient, a poor fit for gas bubbles, metal walls and internal objects [K12].

The AI approach removes the model. It was published in 2025 by Mirakhori et al. with the founder and Professor Chaouki, who is the corresponding author [K12]. A collaborative robot carries the tracer through the volume while the detectors record counts at each known position, pairing radiation levels with true coordinates. A neural network, with five hidden layers in the published rig, learns to map counts to x, y and z [K12]. It needs no assumptions about nuclear parameters, and the authors found that Latin hypercube sampling of the calibration volume gave lower error than a regular grid [K12].

The company is specific about what AI does not do. It does not generate synthetic flow fields or invent trajectories. The radiation is measured, the position is reconstructed, and the flow field is derived from the measured trajectory. The company compares the AI step to image reconstruction in medical imaging [S7], and its performance page separates what is measured from what is reconstructed, with uncertainty propagated through the reconstruction [S2]. If the tracer leaves the calibrated volume, the system flags it instead of returning unreliable numbers [S7].

What comes out

The primary output is a Lagrangian time series: position (x, y, z) against time, with the expected standard deviation of each position [S5, S7]. From it come average velocity fields, pumping rates, mixing dynamics and near-wall velocity distributions [S1], plus occurrence maps of where the tracer spent its time [S7]. Near-wall velocity serves as a proxy for heat transfer, not a direct measurement of it [C4]. Velocity fields ship as VTU files that open in ParaView next to simulation output [S5].

Process measurements taken during the run can be synchronized to the flow data: pressure, temperature, dissolved oxygen, pH, torque, power, gas flow rate, bubble size distribution and liquid level [S8]. That lets an engineer tie a local flow structure to a global result such as oxygen transfer (kLa) and look for dead zones, short-circuiting and recirculation [S10].

How accurate, and on what basis

The site states velocity uncertainty of 2 to 5 percent across the flow field, more than 8 million points for validation and calibration, and a position repeatability of 0.03 mm [S2].

The peer-reviewed record is more specific, and more modest:

• In a controlled test (2025), a robot drove the tracer along a spiral path in air. With a Latin hypercube training set the mean absolute error was 0.4 mm in x, 0.4 mm in y and 0.9 mm in z, and the mean Euclidean distance error was 1.2 mm; with a grid training set it was 2.8 mm [K12].
• A 2023 finite element reconstruction reached 0.49 and 0.76 mm mean absolute error in x and y in a two-dimensional experiment, and 3.37 ms per position against a 10 ms sampling interval. Its 0.08 mm figure comes from a simulated test without experimental noise [K11].
• In the 2025 free-flowing test in a stirred glycerol tank, RPT and CFD agreed on the double recirculation loops. Peak axial velocity in the loops differed by about 2 percent at one radial position away from the impeller. Near the impeller the differences grew: the radial peak was about 16 percent lower in RPT, the axial peak about 29 percent higher, and the tangential peak about 44 percent different around the blades. The authors point to rapid acceleration, 90-degree changes of direction and shaft vibration, and they excluded cells the tracer visited fewer than 400 times [K12]. These are differences between a measurement and a simulation. The paper does not say which is closer to the truth, and neither does this page.

Limits and scope

Fluidmapper publishes its limits. It is built for stirred tanks and mixing vessels from lab to pilot scale, up to about 1 m in diameter, with opaque and multiphase fluids at ambient to moderate temperatures [S3]. It is not yet validated above 1 m. Fluids hotter than 80 °C need specialized tracers still in development. Tracer and detectors need direct access to the vessel.

Other limits come from the physics. A tracer must follow the phase of interest, and very small particles, droplets and bubbles are hard, because a smaller tracer holds less radioactive material and needs longer activation [S7]. Fluidmapper does not measure rheology like a rheometer; its flow fields let effective rheology be inferred or calibrated [S7]. Work is done at the company's facility, not on customer sites, under a Canadian Nuclear Safety Commission permit [S7]. RPT in general has been applied to industrial vessels several meters across, and the company treats larger sizes as an engineering question of detectors and tracer strength, but its own validated range is the one above [S3, S7].

History: where RPT comes from

RPT comes out of university research on multiphase reactors. Fluidmapper's founder trained at Polytechnique Montréal under Professor Jamal Chaouki [S6], a professor in the Department of Chemical Engineering there [R1]. Professor Chaouki's group developed RPT techniques for multiphase reactors: his own research summary lists the development of RPT techniques for three-dimensional flow fields in multiphase reactors among his main contributions [R1], and Fluidmapper's About page names him as one of three professors who are among the field's most influential figures in advancing RPT for industry over four decades [S6]. He is a co-author with the founder on the 2008 Powder Technology paper and on the 2023 and 2025 papers below [K7, K11, K12].

The record starts with detector physics. Beam et al. (1978) gave a Monte Carlo method for the efficiency of cylindrical sodium iodide detectors that see a point source anywhere in space, and RPT reconstruction models still build on it [P5, K11]. In 1994 Larachi, Kennedy and Chaouki described a gamma-ray detection system for three-dimensional particle tracking in multiphase reactors [K1]. A year later the same authors used RPT to map solids flow fields in three dimensions [K2]. In 1997 Professor Chaouki, with two co-authors, published a review of noninvasive tomographic and velocimetric methods for multiphase flows [K3], and edited a book on the subject whose chapter on RPT sets out its principles and applications [K4]. The same year Godfroy et al. applied neural networks to on-line flow visualization [K5], so learning algorithms in this field are older than the current platform.

Over the next two decades papers with Professor Chaouki as an author widened the method. Hamidipour et al. used RPT to monitor particle-wall contact in a gas fluidized bed [K6]; the founder, with two co-authors, extended RPT to systems with irregular moving boundaries [K7] and proposed a mixing measure derived from Lagrangian tracking [P4]; Dubé et al. worked out how to position detectors [K8]; Rasouli et al. introduced a technique that tracks two tracers at once, with accuracy and precision under 5 mm in their tests [K9]; and Bashiri et al. investigated turbulent flow in stirred tanks with a non-intrusive particle tracking technique [K10]. Other groups worked on reconstruction and calibration. Devanathan et al. mapped bubble-column flow with computer-automated RPT in 1990 [P7]. Chen et al. compared RPT with computed tomography and PIV in a bubble column in 1999 [P9]. Yadav et al. proposed a genetic-algorithm reconstruction in 2017 [P10] and machine-learning reconstructions in 2020, the latter performing better than a model-based method around internal tubes [P11]. Khane et al. proposed a hybrid dynamic calibration to reduce the dynamic bias of static calibration in 2017 [P12].

The line then runs through the founder's own work. Mirakhori et al. introduced finite element position reconstruction in 2023 [K11] and the AI-enhanced method, with a collaborative robot and a neural network, in 2025 [K12]. Professor Chaouki is the corresponding author of the 2025 paper, which is the one Fluidmapper's own pages cite [S11].

For most of this history RPT stayed in research labs. Fluidmapper's account is that it was too slow, too complex and too hard to access for industrial users, and that universities are rarely set up for industrial turnaround or guaranteed schedules [S7]. The company does not claim a new measurement principle. It says the new part is turning a specialist academic technique into a service companies can order like any other [S7].

Key publications

1. [K1] F. Larachi, G. Kennedy, J. Chaouki, "A γ-ray detection system for 3-D particle tracking in multiphase reactors," Nuclear Instruments and Methods in Physics Research A 338(2-3), 568-576, 1994. DOI 10.1016/0168-9002(94)91343-9. A detector system for following a single particle in three dimensions inside multiphase reactors.
2. [K2] F. Larachi, J. Chaouki, G. Kennedy, "3-D mapping of solids flow fields in multiphase reactors with RPT," AIChE Journal 41(2), 439-443, 1995. DOI 10.1002/aic.690410226. RPT used to map solids flow in three dimensions.
3. [K3] J. Chaouki, F. Larachi, M. P. Duduković, "Noninvasive tomographic and velocimetric monitoring of multiphase flows," Industrial & Engineering Chemistry Research 36(11), 4476-4503, 1997. DOI 10.1021/ie970210t. A review of noninvasive tomography and velocimetry techniques for multiphase systems.
4. [K4] J. Chaouki, F. Larachi, M. P. Duduković (eds.), Non-Invasive Monitoring of Multiphase Flows, Elsevier, 1997. ISBN 978-0-444-82521-6. DOI 10.1016/B978-0-444-82521-6.X5000-1. Chapter 11, F. Larachi, J. Chaouki, G. Kennedy, M. P. Duduković, "Radioactive particle tracking in multiphase reactors," pp. 335-406. DOI 10.1016/B978-044482521-6/50012-7.
5. [K5] L. Godfroy, F. Larachi, G. Kennedy, B. Grandjean, J. Chaouki, "On-line flow visualization in multiphase reactors using neural networks," Applied Radiation and Isotopes 48(2), 225-235, 1997. DOI 10.1016/S0969-8043(96)00183-2. Neural networks applied to on-line flow visualization.
6. [K6] M. Hamidipour, N. Mostoufi, R. Sotudeh-Gharebagh, J. Chaouki, "Monitoring the particle-wall contact in a gas fluidized bed by RPT," Powder Technology 153(2), 119-126, 2005. DOI 10.1016/j.powtec.2005.02.002. RPT used in a fluidized bed.
7. [K7] J. Doucet, F. Bertrand, J. Chaouki, "An extended radioactive particle tracking method for systems with irregular moving boundaries," Powder Technology 181(2), 195-204, 2008. DOI 10.1016/j.powtec.2006.12.019. RPT extended to irregular moving boundaries.
8. [K8] O. Dubé, D. Dubé, J. Chaouki, F. Bertrand, "Optimization of detector positioning in the radioactive particle tracking technique," Applied Radiation and Isotopes 89, 109-124, 2014. DOI 10.1016/j.apradiso.2014.02.019. Detector placement treated as an optimization problem.
9. [K9] M. Rasouli, F. Bertrand, J. Chaouki, "A multiple radioactive particle tracking technique to investigate particulate flows," AIChE Journal 61(2), 384-394, 2015. DOI 10.1002/aic.14644. Tracks two tracers at once; accuracy and precision under 5 mm in the authors' tests.
10. [K10] H. Bashiri, E. Alizadeh, F. Bertrand, J. Chaouki, "Investigation of turbulent fluid flows in stirred tanks using a non-intrusive particle tracking technique," Chemical Engineering Science 140, 233-251, 2016. DOI 10.1016/j.ces.2015.10.005. Particle tracking applied to turbulence in stirred tanks.
11. [K11] G. Mirakhori, A. Collard-Daigneault, A. Alphonius, J. Doucet, B. Blais, J. Chaouki, "An improved position reconstruction method for radioactive particle tracking," Nuclear Instruments and Methods in Physics Research A 1055, 168504, 2023. DOI 10.1016/j.nima.2023.168504. Finite element position reconstruction.
12. [K12] G. Mirakhori, J. Doucet, S. Chidami, B. Blais, J. Chaouki, "AI-enhanced radioactive particle tracking: A practical methodology for accelerating industrial process development," Chemical Engineering Science 318, 122173, 2025. DOI 10.1016/j.ces.2025.122173. Robot-generated calibration and a neural network for position reconstruction.

Why it is effective, and how it differs from the alternatives

It is effective for four reasons. Gamma rays pass through steel and opaque fluid, so the method does not need a view of the flow [S1, K12]. A phase-matched tracer follows the fluid, so trajectories are real [S1, K12]. Detectors sample every 10 ms, so fast dynamics are captured [S1]. And a robot-generated calibration with millions of known positions replaces fitted physical models and their assumptions [S1, K12]. The 2025 study adds a practical point: in its laminar test the CFD run took 12 hours on a cluster, while sampling the tank (4 hours) and running the experiment (10 hours) took a similar time, and at higher Reynolds numbers CFD can take days while experiment time stays roughly unchanged [K12].

Positron emission particle tracking is a related tracer method; the sources behind this page do not compare it with RPT, so it is not ranked here.

Against PIV, the company's position is that the two solve different problems: PIV is excellent where there is optical access, and Fluidmapper was built for opaque, large or multiphase systems where optics become difficult or impossible [S7]. Against CFD, CFD is a predictive tool built on models and Fluidmapper is a measurement platform built on observed physics. For simple, well-understood systems CFD can do very well. For turbulent, multiphase or poorly characterized ones its accuracy becomes uncertain and calls for experimental validation [S7]. On the data itself, every reconstruction carries quantified uncertainty, and each customer receives the confidence interval chart and spatial confidence intervals [S2, S7].

The four verticals

The company names four buyer groups [S8]. Its own staging puts industrial customers with specific process questions first, then CFD and engineering firms that need validation, and AI and digital twin companies that need large volumes of measured data over the longer term [S7]. Each group below follows the company's audience flyer and page.

AI developers

Who they are. Teams building surrogate models, neural operators, engineering foundation models and digital twins for fluid systems [S12, S13]. Some are simulation companies adding neural networks to speed up solvers. The company's point is that such models are still trained on simulated data and inherit its assumptions, so faster solving does not fix data quality [S7].

The problem. A model trained on solver output learns the solver. Experimental velocity fields are scarce, and those that exist tend to be simplified, two-dimensional or proprietary [S12]. Without experimental ground truth, confidence drops as models move toward complex or safety-critical processes [S12].

What measured data gives them. Three-dimensional velocity fields from real vessels, each reconstruction carrying its uncertainty. Synchronized process variables (dissolved oxygen, kLa, torque, power, gas holdup, temperature, pressure) let a model learn how local flow structure relates to global performance [S12]. One calibration supports up to 28 experiments [S8], and the site cites 33 experiments in 35 days [S12]. The flyer adds that measured data can replace some GPU-heavy CFD runs used to generate training data [C1].

What a deliverable looks like. Large labeled time series, labeled VTU files in ML-ready formats, and correlated process datasets [S12]. The free sample package shows the shape: raw tracer time series plus a trained Keras model per operating condition, taking coordinates in and returning velocity components and uncertainty out [S5]. A physics-informed surrogate model is an optional deliverable [S8].

Keep in view. A licensable dataset library is a stated ambition in the FAQ, not a catalogue on the shelf [S7].

Simulation software developers

Who they are. Groups that build CFD solvers and the turbulence, multiphase and rheology models inside them [S11].

The problem. Their customers ask whether the model has been experimentally validated [S11]. That is hard to answer. Simulation is only as reliable as its assumptions about turbulence, rheology, gas holdup, solids, drag laws and bubble size, three-dimensional validation data is scarce, and traditional validation campaigns are slow, expensive and built for one application [S11].

What measured data gives them. Independent three-dimensional velocity fields with confidence intervals, to test turbulence, multiphase and transport models and to benchmark solver features before release [S11]. For rheology the route is indirect: Fluidmapper does not measure it, but measured flow together with torque, power and other process data can be used to infer or calibrate effective rheology inside operating equipment [S7]. The third use is citable benchmarks, reusable and publishable validation datasets [S11].

The published comparison shows what validation looks like: in the 2025 study, simulation and measurement agreed closely away from the impeller and differed by tens of percent near the blades, with reasons given [K12].

What a deliverable looks like. Tracer time series for every condition, a VTU file of average velocities, confidence intervals and process data [S11]. The VTU file loads like CFD output, in ParaView or a team's own post-processor, with no conversion step [C2, S5].

Equipment manufacturers

Who they are. Makers of agitators, impellers, spargers, mixers and other process equipment, including equipment for demanding multiphase service such as mining autoclaves [S9].

The problem. New equipment takes many design iterations. Optics cannot penetrate opaque slurries or stainless steel, so manufacturers lean on CFD, whose accuracy rests on its models and their validation [S9]. That limits how many projects a team can run and raises the technical and commercial risk of any performance guarantee [S9].

What measured data gives them. Flow measured in the customer's actual fluid and operating conditions, not a transparent mock-up [S9]. Designs can be compared head to head, guarantees backed with experimental evidence, and underperforming installed equipment diagnosed [S9, C3]. The site's stirred-tank studies show the method. A Rushton turbine reached a mixing rate of 0.783 per second at 5.2 W, against 0.310 per second at 3.5 W for a propeller at the same speed, so the better mixing came with higher power draw. Two Rushton turbines at the spacing the site labels H/3 mixed at 1.176 per second at 4.7 W, against 0.390 per second at 4.9 W at spacing H/2, while wall velocity and pumping fell [S4].

What a deliverable looks like. Tracer time series for all conditions, a VTU file of average velocities, confidence intervals, and process data such as kLa, temperature, pressure and gas flow [S9].

Engineering firms and process operators

Who they are. Process engineering firms, reactor design groups and operators of mixing-intensive plants: chemical, biotech and fermentation, food and beverage, mining and minerals, water treatment, and nuclear and energy [S7, S10].

The problem. Scale-up, troubleshooting and optimization need a real picture of the flow, and getting one is rarely practical. Custom visualization rigs are capital-intensive and specific to one project, and most techniques are limited by vessel size, opaque fluids or steel walls and give two-dimensional or local data [S10]. Teams fall back on CFD, which is harder to trust as solids loading, gas holdup and changing rheology come into play [S10].

What measured data gives them. De-risked scale-up on the actual equipment geometry, dead zones, short-circuiting and recirculation located, operating conditions compared to find a good process window, and retrofits evaluated before capital is spent [S10]. For fermenters and bioreactors, flow structure can be tied to measured kLa, dissolved oxygen and power [S7, S10]. The company says its datasets can support qualification and validation packages but that it does not certify equipment or processes [S7]. The site's fifth case study ran 30 configurations in 34 days including calibration, and found different configurations best for mixing, for wall heat transfer potential and for axial pumping [S4].

What a deliverable looks like. The same core files, plus calibration of instruments and the documentation authorities require [S10]. The limits for this group are the scope limits above [S3, S7].

The sample data package

Fluidmapper offers a free sample package so a team can test the data before buying any. It is a complete measured package for a two-Rushton turbine system at 300, 400 and 500 rpm: three cases, one geometry, identical folder structure [S5]. Each case includes:

• Geometry and mesh. The STL geometry and the VTU mesh used for reconstruction.
• Raw time series. Reconstructed Lagrangian tracer trajectories: time and x, y, z in millimetres, with expected standard deviation.
• Trained Keras model. Coordinates in, local velocity components and uncertainty out, one model per operating condition, with scalers.
• Engineering results. Position and velocity distributions, mixing index, pumping and wall velocity profiles, as data and plots.
• ParaView-ready velocity field. The final three-dimensional field for each case, to compare against simulation output [S5].

Access is through a short form, for evaluation use only and no redistribution [S5]. The site calls it the same structure and quality as a paid campaign [S5]. For a solver team it tests the VTU workflow; for an AI team, the model format.

Common Ground's role

Common Ground, the advisory firm that publishes this wiki, worked with Fluidmapper on how the science reaches buyers. It did not design the measurement method; the science and every figure above belong to Fluidmapper and the published papers.

• Brand system. Positioning, palette, type and logo system, built from the science.
• Website. The design and build of every page of fluidmapper.com, including the platform, performance, scope and case study pages and the four audience pages.
• Sales materials. The sales deck and pitch, and one flyer for each of the four verticals.
• Proposal system. A statement-of-work format delivered as an editable file in the client's own brand fonts, so the founder can produce proposals directly.
• Go-to-market research. A ranking of buyer segments and research into the companies within them.
• SEO and GEO preparation. Structured data, question-led FAQ content and plain-text summaries written for AI assistants (llms.txt and llms-full.txt), so search engines and AI tools can describe the company accurately.

The website carries the credit line "Created by Common Ground" [S8].

Sources

Site pages (fluidmapper.com, read 4 October 2026): [S1] /platform · [S2] /performance · [S3] /scope · [S4] /case-studies · [S5] /datasets · [S6] /about · [S7] /faq · [S8] home page · [S9] /for/oem · [S10] /for/engineering · [S11] /for/simulation · [S12] /for/ai · [S13] /llms.txt

Fluidmapper audience materials (client's own, not public pages): [C1] AI Developers flyer v.6 · [C2] Software Developers flyer v.3 · [C3] OEM flyer v.2 · [C4] Engineering flyer v.3

Professional record: [R1] J. Chaouki, curriculum vitae, January 2019 (professional positions and research summary only), https://pearl.polymtl.ca/wp-content/uploads/2021/03/CV-Chaouki-Jamal.pdf

Key publications [K1] to [K12]: see the list in the History section.

Other peer-reviewed papers: [P4] J. Doucet, F. Bertrand, J. Chaouki, "A measure of mixing from Lagrangian tracking and its application to granular and fluid flow systems," Chemical Engineering Research and Design 86(12), 1313-1321, 2008. DOI 10.1016/j.cherd.2008.09.003 [P5] G. Beam, L. Wielopolski, R. Gardner, K. Verghese, "Monte Carlo calculation of efficiencies of right-circular cylindrical NaI detectors for arbitrarily located point sources," Nuclear Instruments and Methods 154, 501-508, 1978. [P7] N. Devanathan, M. P. Duduković, B. A. Toseland, "Flow mapping in bubble columns using CARPT," Chemical Engineering Science 45(8), 2285-2291, 1990. DOI 10.1016/0009-2509(90)80107-P [P9] J. Chen, A. Kemoun, M. H. Al-Dahhan, M. P. Duduković, D. J. Lee, L.-S. Fan, "Comparative hydrodynamics study in a bubble column using computer-automated radioactive particle tracking (CARPT)/computed tomography (CT) and particle image velocimetry (PIV)," Chemical Engineering Science 54(13-14), 2199-2207, 1999. [P10] A. Yadav, M. Ramteke, H. J. Pant, S. Roy, "Monte Carlo real coded genetic algorithm (MC-RGA) for radioactive particle tracking (RPT) experimentation," AIChE Journal 63(7), 2850-2863, 2017. [P11] A. Yadav, T. K. Gaurav, H. J. Pant, S. Roy, "Machine learning based position-rendering algorithms for radioactive particle tracking experimentation," AIChE Journal 66(6), e16954, 2020. [P12] V. Khane, M. H. Al-Dahhan, "Hybrid dynamic radioactive particle tracking (RPT) calibration technique for multiphase flow systems," Measurement Science and Technology 28(5), 055904, 2017.

External links

• Official website: Fluidmapper
• Public data room: Fluidmapper data room

Public record

• Fluidmapper performance page
• Fluidmapper case studies

Stories from this work

At a glance

Project type	Brand creation, go-to-market strategy and fractional CMO
Industry	Industrial technology and measurement (deep tech)
Ran through	Common Ground
Dates	February 2026 to present
Location	Montreal, Canada
Official website	Fluidmapper
Public data room	Fluidmapper data room
Related pages	Fluidmapper, Common Ground
Credit	Created by Common Ground