▸ processconvert
Process Design

RTD Tracer Test Calculator

A pulse or step tracer test gives a response curve C(t) — the tracer concentration leaving a vessel against time. From that curve you can read the mean residence time (the centre of mass of the response) and how spread out the distribution is (the variance), which together say a lot about whether a tank behaves like an ideal well-mixed vessel, an ideal plug-flow vessel, or something non-ideal with short-circuiting or dead zones. This calculator takes a small table of time and response values, integrates the curve with the trapezoidal rule, and returns the area, the mean residence time, the variance, the standard deviation, and the dimensionless variance. It is deliberately simple and transparent: it is a preliminary RTD summary, not a full RTD model, and it does not fit a CSTR, plug-flow, tanks-in-series, or dispersion model. Noisy plant data usually needs baseline correction, a mass-balance check, and specialist interpretation before the numbers mean anything.

TypeInteractive engineering calculator

Calculator

Time (min)
Response C(t)

Enter time and response in any consistent units; blank rows are ignored. Capture the curve back to baseline for a meaningful mean and variance.

Result
Points used5
Area under response40 response·min
Mean residence time20 min
Variance50 min²
Standard deviation7.07107 min
Dimensionless variance0.125

Preliminary RTD summary only. Assumes representative, conservative tracer data integrated by the trapezoidal rule. Not a full RTD model and not a CSTR/plug-flow/tanks-in-series/dispersion fit. A truncated or noisy tail biases the mean and variance. Does not replace test planning, data validation, mass-balance checks, process expertise, or qualified engineering review.

Formulas

Area under response
A = ∫ C(t) dt
Mean residence time
t_mean = ∫ t·C(t) dt / ∫ C(t) dt
Variance
σ² = ∫ (t − t_mean)²·C(t) dt / ∫ C(t) dt
Standard deviation
σ = √(σ²)
Dimensionless variance
σ_θ² = σ² / t_mean²

Diagram

t_mean = ∫ t·C dt / ∫ C dttC(t)t_meant_mean = ∫ t·C dt / ∫ C dt

Worked example

A symmetric tracer response is sampled every 10 minutes: at t = 0, 10, 20, 30, 40 min the response reads 0, 1, 2, 1, 0. Estimate the area, mean residence time, and spread.

  1. 01Area (trapezoidal): A = 10·(0+1)/2 + 10·(1+2)/2 + 10·(2+1)/2 + 10·(1+0)/2 = 5 + 15 + 15 + 5 = 40 response·min
  2. 02∫ t·C dt = 50 + 250 + 350 + 150 = 800 → t_mean = 800 / 40 = 20 min
  3. 03∫ (t−20)²·C dt = 500 + 500 + 500 + 500 = 2000 → σ² = 2000 / 40 = 50 min²
  4. 04σ = √50 = 7.07 min; dimensionless variance σ_θ² = 50 / 20² = 0.125
Result

Area = 40 response·min, mean residence time = 20 min, variance = 50 min², standard deviation ≈ 7.07 min, dimensionless variance = 0.125. (The symmetric triangular response gives a second-central-moment variance of 50 min², not 100 — the spread of this dataset is one sampling interval.)

FAQ

How is the mean residence time from a tracer test different from V/Q?
τ = V/Q is the nominal residence time from the vessel volume and flow. The tracer mean residence time is measured from the actual response curve as its centre of mass, ∫t·C dt / ∫C dt. If the two disagree, the vessel is not behaving ideally — dead volume makes the measured mean shorter than nominal, and recycle or a long tail can push it the other way. Comparing them is the whole point of running a tracer test.
What does the variance tell me?
The variance (and its square root, the standard deviation) measures how spread out the residence-time distribution is around the mean. A very small variance points toward plug flow; a large variance toward strong mixing or non-ideal behaviour. The dimensionless variance σ²/t_mean² is the form most often compared against ideal-reactor models, but this tool reports the summary statistics only — it does not fit a model.
Why does my tail change the answer so much?
Both the mean and the variance weight the long-time data heavily (by t and by t²). If the response is cut off before it returns to baseline, the integrals are truncated and the mean and variance are under-estimated. Capture the full curve back to baseline, and use baseline correction if there is a constant offset, before trusting the numbers.
Can I use this to prove short-circuiting or a dead zone?
Not on its own. A measured mean shorter than nominal, or an oddly shaped curve, is a flag — but confirming short-circuiting, dead volume, channelling, or recycle needs the curve shape, a tracer mass balance, the vessel geometry, and engineering judgement. This calculator gives the preliminary summary statistics that start that diagnosis, not the diagnosis itself.
Does this fit a tanks-in-series or dispersion model?
No. It computes the area, mean, variance, standard deviation, and dimensionless variance by trapezoidal integration and stops there. It does not back out an equivalent number of tanks N, a dispersion number, or a CSTR/PFR fit. Those are separate modelling steps that need validated data and a chosen model.

Related conversions