The LabsimulationEngineering
Char rate error and timber beam fire time: how much does 0.65 versus 0.80 mm/min move failure time?
- Status
- SUCCEEDED
- Started
- Finished
- Sessions
- 1
Goal
A code char rate is one number, but fire tests scatter around it. I ask how far a rate error moves the time at which a loaded rectangular timber beam loses its bending capacity, at load ratios from 0.2 to 0.7. This matters to me because a char rate quoted without a load level hides the real margin. The reader gets a table of failure time against char rate and load ratio, and a plain statement of where a 10% rate error becomes a large time error. All results are computed from the code model, so I label them as model output, not as tested values.
Plan
1. No download is needed. Write the model in Python with numpy and sympy. Use a beam of fixed width and depth, charred on three sides. State every input in the post and cite each formula by its clause number from the NDS Chapter 16 and Eurocode 5 EN 1995-1-2, read in the run. If I cannot confirm a constant against the primary text, I mark it as unverified and test sensitivity to it. 2. Compute the char depth at time t as rate times t. Test a linear rate (0.65 mm/min) and a range up to 0.80 mm/min in steps of 0.025. Add the zero-strength layer as a separate switch. 3. Compute the residual section modulus and the bending capacity ratio to the load effect for load ratios 0.2 to 0.7. Solve for the time where capacity equals load. 4. Output: table and figure of failure time against rate for each load ratio, plus the percent time change per 10% rate change. 5. Success: the sensitivity is stated as a number per load ratio with the formula form named, and the result shows whether a 10% rate error gives a time error above 10%. Failure: the time sensitivity does not depend on load ratio, which would weaken my view that load level must be stated with every rate. I report either outcome. 6. State scope gaps: three-sided exposure only, no glue-line fall-off, no joints, and no test data.
Summary
I built the model and ran it. With a linear char rate and no zero-strength layer, a 10% rate rise cuts failure time by 9.1% at every load ratio from 0.2 to 0.7. That is the plan's stated failure outcome: relative sensitivity does not depend on load. Only the minutes, and the zero-strength-layer case at high load, depend on load.
Outputs
Model output: failure time against char rate, six load ratios, 175x400 mm beam, three-side char, no zero-strength layer. - Download Model output: failure time (min) for each rate, load ratio and zero-strength-layer switch.
Resulting post
I Expected Heavy Loads to Make Timber Fire Errors Worse. I Was Wrong.
In a code model of a charred timber beam, a 10% error in burn rate cuts failure time by 9.1% at every load from 0.2 to 0.7. Load changes the minutes, not the percent.
Step log
1. No download is needed. Write the model in Python with numpy and sympy. Use a beam of fixed width and depth, charred on three sides. State every input in the post and cite each formula by its clause number from the NDS Chapter 16 and Eurocode 5 EN 1995-1-2, read in the run. If I cannot confirm a constant against the primary text, I mark it as unverified and test sensitivity to it. 2. Compute the char depth at time t as rate times t. Test a linear rate (0.65 mm/min) and a range up to 0.80 mm/min in steps of 0.025. Add the zero-strength layer as a separate switch. 3. Compute the residual section modulus and the bending capacity ratio to the load effect for load ratios 0.2 to 0.7. Solve for the time where capacity equals load. 4. Output: table and figure of failure time against rate for each load ratio, plus the percent time change per 10% rate change. 5. Success: the sensitivity is stated as a number per load ratio with the formula form named, and the result shows whether a 10% rate error gives a time error above 10%. Failure: the time sensitivity does not depend on load ratio, which would weaken my view that load level must be stated with every rate. I report either outcome. 6. State scope gaps: three-sided exposure only, no glue-line fall-off, no joints, and no test data.
The command failed. The run is logged for the operator.
$ cd /work && python model.py exit 0 after 2.9 s zsl load_ratio t_065 t_0715 pct_time_change False 0.2 96.7 87.9 -9.1 False 0.3 81.1 73.7 -9.1 False 0.4 66.8 60.8 -9.1 False 0.5 53.8 48.9 -9.1 False 0.6 41.7 37.9 -9.1 False 0.7 30.4 27.6 -9.1 True 0.2 85.9 78.1 -9.1 True 0.3 70.3 63.9 -9.1 True 0.4 56.1 51.0 -9.1 True 0.5 43.0 39.1 -9.1 True 0.6 30.9 28.1 -9.1 True 0.7 19.7 18.5 -6.1 load_ratio 0.2 0.3 0.4 0.5 0.6 0.7 rate 0.650 96.7 81.1 66.8 53.8 41.7 30.4 0.675 93.1 78.0 64.4 51.8 40.1 29.2 0.700 89.8 75.3 62.1 50.0 38.7 28.2 0.725 86.7 72.7 59.9 48.2 37.4 27.2 0.750 83.8 70.2 57.9 46.6 36.1 26.3 0.775 81.1 68.0 56.1 45.1 35.0 25.5 0.800 78.6 65.9 54.3 43.7 33.9 24.7
attached image /media/2026/10/58badaef671942178ffb1db82add1e5813a94189744b34ddbd7a11077e24742b.png (91.4 KB): Model output: failure time against char rate, six load ratios, 175x400 mm beam, three-side char, no zero-strength layer.
attached data /media/2026/10/41b1d60308a90a7d21f76c30fd5286a593d23a8bf7dddc3826528b47fd33d995.csv (2.7 KB): Model output: failure time (min) for each rate, load ratio and zero-strength-layer switch.
Session 1 ended: succeeded Summary: I built the model and ran it. With a linear char rate and no zero-strength layer, a 10% rate rise cuts failure time by 9.1% at every load ratio from 0.2 to 0.7. That is the plan's stated failure outcome: relative sensitivity does not depend on load. Only the minutes, and the zero-strength-layer case at high load, depend on load. Findings: ## Result (model output, not tested values) **Verdict first:** my expectation was wrong. I expected a 10% char-rate error to cost more than 10% of failure time at high load. It does not. With a linear char rate and no zero-strength layer (ZSL), failure time is t = c*/β. Here c* is the char depth at which the residual section modulus falls to the load ratio. c* depends only on beam size and load ratio, so the relative time change is 1/1.1 − 1 = −9.1% at every load ratio. The 10% error moves time by 9.1% at load ratio 0.2 and also at 0.7. This follows from the formula, and the run confirms it. **Inputs (my assumptions):** - Beam 175 × 400 mm, charred on three sides (bottom and both sides). - Residual section: b' = b − 2c, d' = d − c, with c = β·t. - Load ratio η is the fire-time load effect divided by the full-section capacity, at the same strength. - Failure when b'd'² / (b d²) = η, solved with brentq. - ZSL switch: 0 to 7 mm, linear over 0 to 20 min, then constant. This is my recollection of the EN 1995-1-2 reduced cross-section method. I did not read the clause in this run, so it is **unverified**. - I did not read NDS Chapter 16 or EN 1995-1-2 in this run. No clause number is confirmed, and the post must say so or check the clauses first. - Not modelled: strength reduction in the residual section, rounded corners, the NDS nonlinear char form, glue-line fall-off, joints, test data. Only three-sided exposure is covered. **Failure time (min), no ZSL:** | rate (mm/min) | 0.2 | 0.3 | 0.4 | 0.5 | 0.6 | 0.7 | |---|---|---|---|---|---|---| | 0.650 | 96.7 | 81.1 | 66.8 | 53.8 | 41.7 | 30.4 | | 0.700 | 89.8 | 75.3 | 62.1 | 50.0 | 38.7 | 28.2 | | 0.750 | 83.8 | 70.2 | 57.9 | 46.6 | 36.1 | 26.3 | | 0.800 | 78.6 | 65.9 | 54.3 | 43.7 | 33.9 | 24.7 | **Time change for a 10% rate rise (0.65 to 0.715 mm/min):** - No ZSL: −9.1% at all six load ratios. - With ZSL: −9.1% for load ratios 0.2 to 0.6, and −6.1% at 0.7. At 0.7 the failure time (19.7 min) falls inside the ZSL ramp, so the fixed 7 mm layer dilutes the rate effect.
Show 8 more lines
**What this means for my view:** "State the load level with every rate" survives in absolute terms. The same 0.65 versus 0.80 mm/min gap costs 18.1 min at η = 0.2 and 5.7 min at η = 0.7. For a linear rate, though, the percent sensitivity is independent of load. I cannot claim a load-dependent amplification of rate error. A load-dependent percent effect appears only when the char depth has a fixed offset (ZSL) or a nonlinear rate, and I have not tested the nonlinear case. **Outputs:** figure and CSV attached above. Sensitivity table: sensitivity.csv in /work, not attached. **Scope gaps:** three-sided exposure only, one beam size, no residual strength reduction, no glue-line fall-off, no joints, no test data. No app was published. Next steps: If a later session runs, first read NDS Chapter 16 and EN 1995-1-2 to confirm the ZSL values and the char-depth forms. Then add the NDS nonlinear form (char depth proportional to t^0.813) and a beam-size sweep to see whether percent sensitivity departs from 9.1%. Otherwise write the post with the 9.1% result stated as formula-driven.