Compare choke correlations against a measured rate
A measured 720 bbl/d through a 32/64-in bean at 550 psia and a GLR of 900 scf/STB: the rate the Gilbert, Ros, Baxendell, Achong and Pilehvari correlations predict, each one's error against the measurement, the factor that calibrates Gilbert to it, and the Gilbert choke performance curves, rate against bean size, for wellhead pressures from 200 to 1,000 psia.
- UC-SF-004
- UC-SF-003
Engineering use cases UC-SF-004 (Compare Choke Correlation Predictions) and UC-SF-003 (Generate Choke Performance Curves).
Every Gilbert-type correlation has the form p1 = A q RB / SC (p1 in psi, q in bbl/d, R in scf/STB, S in 64ths of an inch); only the constants differ, because each author fitted a different field's data: Gilbert (1954) 10.0, 0.546, 1.89; Ros (1960) 17.40, 0.500, 2.00; Baxendell (1957) 9.56, 0.546, 1.93; Achong (1961) 3.82, 0.650, 1.88; Pilehvari (1981) 46.67, 0.313, 2.11 (Guo et al. 2007, Table 5.1). The rate is linear in the wellhead pressure and grows roughly with the square of the bean.
What an engineer learns from it: at this test every correlation reads high. Gilbert is closest at 937 bbl/d, 30 % over; Ros, Baxendell and Achong are 50 to 62 % over, and Pilehvari predicts almost three times the test. No correlation is right out of the box: the field's own factor, 0.769 on Gilbert here, scales every later prediction onto the well. The performance curves are the operator's chart: at 600 psia a 32/64 bean passes about 1,090 bbl/d at a GLR of 800 scf/STB, and doubling the pressure doubles the rate.
Units: psia, bbl/d, scf/STB, 64ths of an inch. The correlations hold for critical flow only; the original fits used gauge pressure, which this entry, like the use case, does not correct for (about 3 % at 550 psi).
Source
using System;
using System.Linq;
using Team76.Engineering.SF.Choke.MultiPhase;
public class S : ICalcScript
{
public ScriptResult Run(ICalcContext ctx)
{
// The use case's well test: 550 psia upstream of a 32/64-in bean at a GLR of 900 scf/STB, measured at 720 bbl/d.
const double pwh = 550; // psia
const double choke = 32.0 / 64; // in
const double glr = 900; // scf/STB
const double measured = 720; // bbl/d
(string Name, Func<double, double, double, double> Rate)[] correlations =
[
("Gilbert (1954)", GilbertChoke.Rate),
("Ros (1960)", RosChoke.Rate),
("Baxendell (1957)", BaxendellChoke.Rate),
("Achong (1961)", AchongChoke.Rate),
("Pilehvari (1981)", PilehvariChoke.Rate),
];
var rates = correlations.Select(c => (c.Name, Rate: c.Rate(pwh, choke, glr))).ToList(); // bbl/d
var rows = rates.Select(r => new object[] { r.Name, Math.Round(r.Rate), Math.Round(100 * (r.Rate - measured) / measured, 1) }).ToList();
var best = rates.OrderBy(r => Math.Abs(r.Rate - measured)).First();
// The Gilbert rate the field's correction factor brings onto the test: every later prediction is scaled by it.
double calibration = measured / GilbertChoke.Rate(pwh, choke, glr);
// UC-SF-003: Gilbert performance curves, the rate through each bean at wellhead pressures from 200 to 1,000 psia
// and a GLR of 800 scf/STB. A Gilbert-type rate is linear in the wellhead pressure.
double[] beans = [16, 20, 24, 28, 32, 36, 40, 48, 56, 64];
double[] pressures = [200, 400, 600, 800, 1000];
var curves = pressures.Select(p => new ScriptSeries($"{(int)p} psia", beans,
beans.Select(b => Math.Round(GilbertChoke.Rate(p, b / 64, 800))).ToArray())).ToList();
return ctx.Result()
.Set("Correlations against the test", new ScriptTable(["Correlation", "Rate (bbl/d)", "Error (%)"], rows))
.Set("Measured rate (bbl/d)", measured)
.Set("Closest correlation", best.Name)
.Set("Gilbert calibration factor", Math.Round(calibration, 3))
.Set("Choke performance curves", new ScriptChart(curves, title: "Gilbert rate by bean, GLR 800 scf/STB", xLabel: "Choke (64ths in)", yLabel: "Liquid rate (bbl/d)"));
}
}