Constraints of order unity

We have seen some failures on main recently where regression tests are reporting slightly different constraint violations depending on the computer system running the tests; I believe this to be either caused by (non-Python) software dependencies or the hardware environment.

We recently merged #4702 which I thought might address this, however the failures persist. I think it is worth checking whether moving all constraints to order unity will ~~fix~~ hide this problem with our solver. Even if it does not stop the failures, it is still good-practice and will be of no detriment to us.

@ukaea/process-model-review before I start working on this change I would welcome any input

For example, at the moment we provide a table of constraints in physical units:

                                                             Physical constraint             Physical constraint bound    Constraint residue          Normalised residue
-----------------------------------------------------------  --------------------------  --  ---------------------------  --------------------------  --------------------
Plasma injected power (Pₐᵤₓ) upper limit                     90.63680742703416 MW        <=  200.0 MW                     -109.36319257296584 MW      -0.546816
Pₛₑₚ > Pₗₕ consistency                                       157.08916305378975 MW       >=  86.84755802547093 MW         -70.24160502831882 MW       -0.808792
Net electric power lower limit                               400.0283540398513 MW        >=  400.0 MW                     -0.028354039851308244 MW    -7.08851e-05
⟨β⟩ upper limit                                              0.040870763942806566        <=  0.05068216915813131          -0.009811405215324746       -0.193587
TF peak toroidal field upper limit                           10.723705319967207 T        <=  14.0 T                       -3.276294680032793 T        -0.234021

Is it important to the modellers (and users of PROCESS) that the physical units be retained where possible?

The current format for reporting results in physical units is very useful - at a glance information in the table should be retained.

Agree with Graeme, units should be kept.

Navigation Menu

Repository navigation

Constraints of order unity #4704

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Constraints of order unity #4704

Description

Activity

timothy-nunn commented on Oct 21, 2026

grmtrkngtn commented on Oct 21, 2026

ajpearcey commented on Oct 21, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions