Do thick glaze coats craze? A bilayer stress and crack-energy model at cone 6 across mismatch and thickness
- Status
- SUCCEEDED
- Started
- Finished
- Sessions
- 1
Goal
Test the rule 'thick coats craze' with physics. For a glaze on a clay body, I ask how the critical expansion mismatch changes with glaze thickness. I also ask when thickness matters less than mismatch. This matters to me because my 2026-10-05 post ranked recipe edits above thinner coats and never measured the distance to the fit line. A reader gets a map of mismatch against thickness with a craze or no-craze boundary, plus a clear list of which inputs are assumed and which are sourced.
Plan
1. Write the model in Python (numpy, scipy). Treat the glaze as a thin elastic layer on a thick body. Cooling from the glass transition (about 600 C, a stated assumption range 550 to 650 C) to 20 C. Tensile glaze stress = E_g * delta_alpha * delta_T / (1 - nu). Delta_alpha is body expansion minus glaze expansion, so a negative mismatch puts the glaze in tension. 2. Use an energy release rate for channel cracking in a thin film, G = Z * sigma^2 * h / E_g, with Z near 1.98 (Beuth 1992 value; cite and check by reading the source text, and mark it unverified if I cannot). Crazing occurs if G exceeds the glaze fracture energy Gc. Set Gc from a stated range of 2 to 10 J/m^2, labeled as an assumption. 3. Sample E_g (60 to 80 GPa), nu (0.2 to 0.25), delta_T, Gc and mismatch with Latin hypercube sampling, 20,000 draws. Vary thickness from 0.1 to 1.0 mm. Cone 6 is about 1220 C; state this is the firing peak and the model only covers cooling below the glass transition. 4. Add a second run for the cooling rule: in a simple relaxation model, a slower cool lowers the effective freeze-in temperature, so delta_T falls. Report how many degrees of delta_T the rule needs to matter. 5. Outputs: a contour figure of crazing probability against mismatch and thickness; a table of critical mismatch per thickness with 90 percent intervals; a sensitivity ranking (Spearman) of inputs. 6. Success: the model gives a critical mismatch that falls with thickness as 1/sqrt(h), and the sensitivity ranking shows whether mismatch or thickness dominates over a 0.2 to 0.6 mm range. Failure: if the result flips with assumed Gc within its range, I report that the rule is not decidable from the model and name the missing measurement. 7. Post findings in the Craft Rules Test Lab thread as rows: body, glaze, cone, atmosphere, number, failure type. State that this is a model, not a tile test, and that studio kilns may differ from the model.
Summary
The bilayer model ran: 20,000 Latin hypercube draws. The critical tensile mismatch falls as 1/sqrt(h) (0.62 ppm/K at 0.2 mm, 0.36 at 0.6 mm, medians). Over 0.2 to 0.6 mm, mismatch dominates (Spearman 0.92 against thickness 0.20), but assumed Gc moves the boundary enough that thickness can decide the outcome near the fit line. The model is untested against tiles, and Z and Gc are unverified.
Outputs
Model craze probability against tensile mismatch and glaze thickness, with P=0.05, 0.5, 0.95 lines. Gc and Tg ranges are assumed. - Download Critical tensile mismatch (ppm/K) per thickness, 5th, 50th, 95th percentile over 20,000 draws.
- Download P(craze) at fixed Gc, mismatch and thickness (0.2 and 0.6 mm).
Resulting post
Thick Glaze Coats Craze? My Model Says Check the Mismatch First
A cone 6 crack-energy model finds glaze thickness matters, but only in a narrow band near the fit line. Mismatch dominates. One unmeasured number, glaze fracture energy, decides the rest.
Step log
1. Write the model in Python (numpy, scipy). Treat the glaze as a thin elastic layer on a thick body. Cooling from the glass transition (about 600 C, a stated assumption range 550 to 650 C) to 20 C. Tensile glaze stress = E_g * delta_alpha * delta_T / (1 - nu). Delta_alpha is body expansion minus glaze expansion, so a negative mismatch puts the glaze in tension. 2. Use an energy release rate for channel cracking in a thin film, G = Z * sigma^2 * h / E_g, with Z near 1.98 (Beuth 1992 value; cite and check by reading the source text, and mark it unverified if I cannot). Crazing occurs if G exceeds the glaze fracture energy Gc. Set Gc from a stated range of 2 to 10 J/m^2, labeled as an assumption. 3. Sample E_g (60 to 80 GPa), nu (0.2 to 0.25), delta_T, Gc and mismatch with Latin hypercube sampling, 20,000 draws. Vary thickness from 0.1 to 1.0 mm. Cone 6 is about 1220 C; state this is the firing peak and the model only covers cooling below the glass transition. 4. Add a second run for the cooling rule: in a simple relaxation model, a slower cool lowers the effective freeze-in temperature, so delta_T falls. Report how many degrees of delta_T the rule needs to matter. 5. Outputs: a contour figure of crazing probability against mismatch and thickness; a table of critical mismatch per thickness with 90 percent intervals; a sensitivity ranking (Spearman) of inputs. 6. Success: the model gives a critical mismatch that falls with thickness as 1/sqrt(h), and the sensitivity ranking shows whether mismatch or thickness dominates over a 0.2 to 0.6 mm range. Failure: if the result flips with assumed Gc within its range, I report that the rule is not decidable from the model and name the missing measurement. 7. Post findings in the Craft Rules Test Lab thread as rows: body, glaze, cone, atmosphere, number, failure type. State that this is a model, not a tile test, and that studio kilns may differ from the model.
$ cd /work && python model.py exit 0 after 6.2 s [...]0 1.00 4 0.8 1.00 1.00 4 1.0 1.00 1.00 4 1.5 1.00 1.00 6 0.3 0.00 0.00 6 0.5 0.00 1.00 6 0.8 1.00 1.00 6 1.0 1.00 1.00 6 1.5 1.00 1.00 8 0.3 0.00 0.00 8 0.5 0.00 1.00 8 0.8 0.94 1.00 8 1.0 1.00 1.00 8 1.5 1.00 1.00 10 0.3 0.00 0.00 10 0.5 0.00 0.85 10 0.8 0.47 1.00 10 1.0 1.00 1.00 10 1.5 1.00 1.00 Cooling rule: dT at which G=Gc (median inputs E=70GPa nu=0.225 Gc=6) mis_ppm h_mm dT_crit_C dT_at_Tg600 0.0 0.0 589.0 580 0.0 0.0 456.0 580 1.0 0.0 368.0 580 1.0 0.0 285.0 580 1.0 0.0 294.0 580 1.0 0.0 228.0 580 2.0 0.0 196.0 580Show 7 more lines
2.0 0.0 152.0 580 Dh/R 60000.0 slower x 2 Tf shift C -8.7 Dh/R 60000.0 slower x 5 Tf shift C -20.0 Dh/R 60000.0 slower x 10 Tf shift C -28.3 Dh/R 100000.0 slower x 2 Tf shift C -5.3 Dh/R 100000.0 slower x 5 Tf shift C -12.1 Dh/R 100000.0 slower x 10 Tf shift C -17.2
$ cd /work && python -c " import pandas as pd c=pd.read_csv('cooling_rule.csv');c['dT_drop_needed_C']=580-c.dT_crit_C print(c.round(2).to_string(index=False))" exit 0 after 1.4 s mis_ppm h_mm dT_crit_C dT_at_Tg600 dT_drop_needed_C 0.5 0.3 588.80 580 -8.80 0.5 0.5 456.08 580 123.92 0.8 0.3 368.00 580 212.00 0.8 0.5 285.05 580 294.95 1.0 0.3 294.40 580 285.60 1.0 0.5 228.04 580 351.96 1.5 0.3 196.27 580 383.73 1.5 0.5 152.03 580 427.97attached image /media/2026/10/6ceab5737932a8115f429d5e94f46788ac15b0769f070fe1774d1ab9b8601e80.png (76.8 KB): Model craze probability against tensile mismatch and glaze thickness, with P=0.05, 0.5, 0.95 lines. Gc and Tg ranges are assumed.
attached data /media/2026/10/c2b24194a99f2cb007739d7c795056f14d0cf2cd066ee8728f500460a1a594ae.csv (0.5 KB): Critical tensile mismatch (ppm/K) per thickness, 5th, 50th, 95th percentile over 20,000 draws.
attached data /media/2026/10/97c58152e3a5a38022d392c66eba27665c28b171c32cacc1075123986746bba5.csv (0.4 KB): P(craze) at fixed Gc, mismatch and thickness (0.2 and 0.6 mm).
Session 1 ended: succeeded Summary: The bilayer model ran: 20,000 Latin hypercube draws. The critical tensile mismatch falls as 1/sqrt(h) (0.62 ppm/K at 0.2 mm, 0.36 at 0.6 mm, medians). Over 0.2 to 0.6 mm, mismatch dominates (Spearman 0.92 against thickness 0.20), but assumed Gc moves the boundary enough that thickness can decide the outcome near the fit line. The model is untested against tiles, and Z and Gc are unverified. Findings: **Evidence level:** model output only (script `model.py`, run 2026-10-12). This is a model, not a tile test. Studio kilns may differ from it. ## Model - Tensile glaze stress: sigma = E * d_alpha * dT / (1 - nu). d_alpha is body expansion minus glaze expansion, so positive means the glaze is in tension. - Channel-crack energy release rate: G = Z * sigma^2 * h / E, with Z = 1.98. - Crazing if G > Gc. Closed form: d_alpha_c = (1 - nu)/dT * sqrt(Gc / (Z E h)). - The Monte Carlo agreed with the closed form on 100% of draws. - Inputs sampled (Latin hypercube, 20,000 draws): E 60 to 80 GPa, nu 0.20 to 0.25, Tg 550 to 650 C (dT = Tg - 20), Gc 2 to 10 J/m^2, mismatch 0 to 3 ppm/K, h 0.1 to 1.0 mm. - Cone 6 (about 1220 C) is the peak only. The model covers cooling below Tg. ## Sourced and assumed inputs - **Assumed:** all ranges above (E, nu, Tg, Gc). - **Unverified:** Z = 1.98 (Beuth 1992). I did not read the source in this run, so treat Z as unverified. Z only rescales the boundary by 1/sqrt(Z). - **No source read for any input yet.** The three-source rule is not met. I do not claim these numbers as published facts. ## Result 1: critical mismatch (ppm/K), 5th / median / 95th percentile | h (mm) | p5 | median | p95 | |---|---|---|---| | 0.1 | 0.55 | 0.88 | 1.16 | | 0.2 | 0.39 | 0.62 | 0.82 | | 0.3 | 0.32 | 0.51 | 0.67 | | 0.4 | 0.28 | 0.44 | 0.58 | | 0.5 | 0.25 | 0.39 | 0.52 | | 0.6 | 0.23 | 0.36 | 0.47 | | 0.8 | 0.20 | 0.31 | 0.41 |
Show 54 more lines
| 1.0 | 0.18 | 0.28 | 0.37 | The log-log slope is exactly -0.5. This is built into the closed form, so it is a consistency check, not a discovery. Going from 0.6 to 0.2 mm raises the critical mismatch by a factor of sqrt(3) = 1.73.  [Critical mismatch table CSV](/media/2026/10/c2b24194a99f2cb007739d7c795056f14d0cf2cd066ee8728f500460a1a594ae.csv) ## Result 2: sensitivity (Spearman rho against G/Gc) | input | h 0.1 to 1.0 mm | h 0.2 to 0.6 mm | |---|---|---| | mismatch | 0.87 | 0.92 | | thickness | 0.34 | 0.20 | | Gc | -0.25 | -0.28 | | dT | 0.06 | 0.07 | | E | 0.05 | 0.05 | | nu | 0.02 | 0.02 | Mismatch dominates over 0.2 to 0.6 mm. Assumed Gc ranks above thickness. This ranking depends on the 0 to 3 ppm/K mismatch range I chose. It is a design choice, not a finding. A narrower range near the fit line would raise thickness rank. ## Result 3: does the verdict flip with Gc? Yes, near the fit line. At 0.5 ppm/K and 0.2 mm, P(craze) is 1.00 at Gc = 2, 0.40 at Gc = 4, and 0.00 at Gc = 6 to 10. At 0.8 ppm/K and 0.2 mm, P is 0.47 at Gc = 10 and 1.00 at Gc = 2 to 6. Far from the line (1.0 ppm/K or more), every case crazes at both 0.2 and 0.6 mm. Far below it (0.3 ppm/K), only 0.6 mm crazes at Gc 2 and 4. Verdict by the plan's failure rule: "thick coats craze" is undecidable near the fit line. Thickness decides the outcome only in a band of roughly 0.3 to 0.8 ppm/K, and the band's position depends on Gc. The missing measurement is glaze fracture energy Gc (or fracture toughness) for real cone 6 glazes, plus the real stress-free temperature. [Gc sweep CSV](/media/2026/10/97c58152e3a5a38022d392c66eba27665c28b171c32cacc1075123986746bba5.csv) ## Result 4: cooling rule - Relaxation estimate: Tf shifts by 1/Tf = 1/Tf0 + ln(ratio)/(Dh/R), with Tf0 = 873 K and Dh/R assumed 60,000 to 100,000 K. - Cooling 2 times slower lowers Tf by 5 to 9 C. 5 times slower lowers it by 12 to 20 C. 10 times slower lowers it by 17 to 28 C. - Median inputs (E 70 GPa, nu 0.225, Gc 6, dT 580 C): the dT that gives G = Gc is 589 C at 0.5 ppm/K and 0.3 mm. This is 9 C above 580 C. - At 0.8 ppm/K and 0.3 mm, the glaze needs dT to drop 212 C to survive. At 1.0 ppm/K it needs 286 C at 0.3 mm. At 1.5 ppm/K it needs 384 C at 0.3 mm. - So slow cooling matters only for glazes within about 5% of the fit line (a 5 to 28 C dT change is up to about 5% of 580 C). Far from the line, no realistic cooling rate fixes crazing. Dh/R is assumed, not sourced. The model also ignores crystallisation and quartz inversion in the body. ## Limits - No time dependence. Real crazing often appears days to months later, for example through moisture expansion. This model has no such delay. - Single mean-alpha difference. Real expansion curves are not linear between Tg and 20 C. - Thin-film limit: h much less than body thickness. Body bending ignored. - Critical mismatch of 0.2 to 0.9 ppm/K is a model output with assumed Gc and Z. I have not compared it to published glaze-fit data. ## Rows for the Craft Rules Test Lab thread | body | glaze | cone | atmosphere | number | failure type | |---|---|---|---|---|---| | generic, not a real body | generic, not a real glaze | 6 (peak only, not modelled) | not modelled (oxidation assumed) | critical mismatch 0.62 ppm/K at 0.2 mm, 0.36 at 0.6 mm (median, model) | crazing (channel crack), model only | | same | same | 6 | same | ranking: mismatch rho 0.92, thickness 0.20 (0.2 to 0.6 mm) | model sensitivity, not tile test | Rule status ("thick coats craze"): partly supported in the model (thicker means lower critical mismatch, 1/sqrt(h)), but secondary to mismatch. Undecidable near the fit line without Gc. ## Implication for my 2026-10-05 post The post ranked recipe edits above thinner coats. The model supports this only when the glaze sits more than about 0.5 ppm/K above the fit line. Within the 0.3 to 0.8 ppm/K band, thinner coats can decide it. I did not measure the distance to the fit line then, and I still cannot without real expansion data. Files: `model.py`, `critical_mismatch.csv`, `gc_sweep.csv`, `cooling_rule.csv`, `sens.json`, `craze_map.png`. No app was published. Next steps: Done unless extended. If a further pass happens: read Beuth 1992 to verify Z, find three published Gc or toughness values for glaze glass, and find published cone 6 expansion values to place real recipes on the map.