How we verify the LOGOS calculations

LOGOS verifies each calculation group by comparing the output of its production code with a reference value that does not come from LOGOS code: a number printed by a primary source (a standard, a technical release, a CIGRE brochure) or computed by an independent, open implementation. The tables right after this text show, row by row, the reference value, the LOGOS value, the difference and the status against the declared tolerance.

How LOGOS verifies a calculation

The verification follows four rules, and each one exists to block a specific kind of false confidence.

1. Primary source, with edition and year. Each group cites the exact source — for example “IAPWS R7-97(2012), Table 15” or “pandapower 3.5.4, calc_sc with lv_tol_percent = 6”. Without the edition, a reference number cannot be reproduced.
2. Independent expected value, never taken from the code itself. If the reference were generated by LOGOS, the comparison would only prove that the code agrees with itself. The reference is therefore always a published number or the result of a different method. For the friction factor, for instance, the reference is the exact Colebrook-White solution through the Lambert W function, 1/sqrt(f) = -a·ln(a·c·W(z)), while LOGOS solves the same equation with a Serghides estimate and Newton steps.
3. Automated tests. The comparison on this page is also a test in the LOGOS repository: if a code change pushes any row out of tolerance, the test fails.
4. The LOGOS column is generated at build time. The numbers in the LOGOS column are not typed. They are computed when the page is generated, by the same modules the calculators run on the server. The status of each row (“Matches” or “Out of tolerance”) is computed too: the page cannot show “Matches” for a result that is out of tolerance.

The references, one by one

IAPWS-IF97. The industrial formulation of the International Association for the Properties of Water and Steam is the worldwide standard for water and steam properties in engineering. Release R7-97 (2012 revision) publishes verification values intended precisely for validating computer programs. LOGOS checks the saturation temperature (Table 36) and the specific volume, enthalpy, isobaric heat capacity and speed of sound of superheated steam (Table 15, Region 2). These properties feed the control valve and safety valve calculators for steam.

pandapower. pandapower is an open-source power-system analysis library that implements IEC 60909-0:2016 short-circuit calculation. LOGOS runs 5 radial-network cases — stiff grid, network transformer and low-voltage bus, one of them with a cable — and compares three-phase Ik″, peak current i_p, two-phase Ik″ and minimum Ik″. The version (3.5.4) and the parameters (case = max/min, ip = True, lv_tol_percent = 6, i.e. c_max = 1.05 and c_min = 0.95 on LV) are declared so anyone can repeat the calculation.

CIGRE TB 880 (2022). CIGRE Technical Brochure 880, “Power cable rating examples for calculation tool verification”, publishes IEC 60287 ampacity reference cases created precisely to verify calculation tools. LOGOS uses the introductory cases with a 132 kV XLPE cable, 630 mm² copper: 0-1 (directly buried trefoil), 0-2 (trefoil of touching HDPE ducts) and 0-3 (PVC duct bank in concrete, with the single-point bonding and eddy-current loss variants).

ASME B36.10M. The standard for welded and seamless wrought steel pipe defines the outside diameter and wall thickness of each NPS and schedule. The reference inside diameter is computed as ID = (OD - 2·t) × 25.4 mm from the inch dimensions and compared with the pipe table used by the hydraulic calculators.

ABNT NBR 5410. The Brazilian low-voltage installation standard tabulates current-carrying capacity (Tables 36 and 37) and the correction factors for ambient temperature (Table 40) and grouping (Table 42). The copper values of Tables 36 and 37 are the same as IEC 60364-5-52 Tables B.52.4 and B.52.5. There is no calculation here: the check confirms that the table transcribed in the code returns exactly the value printed in the standard.

What the verification found and what was fixed

A verification that never finds anything does not prove much. The comparison with pandapower and CIGRE TB 880 found real differences in the LOGOS code. Each one was isolated numerically, fixed, and is now covered by the test.

Short circuit (IEC 60909), comparison with pandapower:

• D51 — transformer K_T correction in the minimum case. IEC 60909-0:2016 (§6.3.3) defines the transformer impedance correction factor with c_max, for the maximum-current calculation. LOGOS also applied K_T to the minimum Ik″, which came out 2% to 4% high. Because the minimum Ik″ is what checks whether the protection trips, the error was on the unsafe side. After the fix, minimum Ik″ dropped by 0.4% to 1.9% in product cases and now matches pandapower.
• D52 — voltage factor in the network impedance. LOGOS used the c factor of the bus under analysis (1.05 on LV) to compute the network impedance, when the correct value is the c of the source voltage level (1.10 and 1.00 at medium voltage). The effect was an Ik″ 0.15% to 0.25% high. Fixed.

Duct-bank cable ampacity (IEC 60287), comparison with CIGRE TB 880:

• D39 — the aluminum sheath resistivity used the value for aluminum conductors; it now uses 2.84×10⁻⁸ Ω·m, the value adopted in TB 880.
• D40 — the insulation thermal resistance (T1) now treats the semiconducting layers separately when the insulation thickness is given.
• D41 — the U, V and Y constants for plastic ducts are now 1.87 / 0.312 / 0.0037, the TB 880 values.
• D42 — eddy-current losses in solid metallic sheaths were always neglected. For a solid-sheath cable with single-point bonding, this made the ampacity 3.1% optimistic in case 0-1 — the only unsafe-side finding in this series. LOGOS now computes these losses per IEC 60287-1-1 §2.3.6.1.
• D43 — the sheath loss factor (λ₁) of an untransposed flat circuit had a term that cancelled out; after the fix, λ₁ of the lagging outer cable went from 1.49 to 1.70, close to the 1.72 of another independent implementation.
• D44 — the oversheath material (PVC, PE/HDPE, polychloroprene) now enters the thermal resistance T3.
• D45 — in an untransposed flat circuit, each cable position now has its own λ₁, and the ampacity is the current at which the hottest cable reaches the maximum temperature.

What this verification does not cover

The verification covers the points in the tables below — it does not prove that every possible result is correct. The known limits are:

• Motor contribution to short circuit. LOGOS adds motor contribution through locked-rotor current; pandapower models the induction motor from nameplate data. The compared cases include no motors.
• Meshed networks. LOGOS short circuit is computed on a radial network, by design; meshed networks and generators as sources are not covered.
• Line-to-ground faults with full zero sequence. LOGOS computes the line-to-ground fault with a factor k0 = Z0/Z1, declared or derived from the grounding resistor or the phase-PE loop; it does not solve the three sequence networks with the transformer vector group, as pandapower does. That is why line-to-ground is not in the comparison.
• General meshed hydraulic networks. The hydraulic calculators solve defined industrial topologies (single line, pumps in series and in parallel, branches, recirculation and the unification loop), not an arbitrary meshed network.
• Hydraulic transients. Water hammer is calculated with the Joukowsky and Michaud formulas. Lines with an irregular profile, several pumps or column separation call for a method-of-characteristics simulation, which LOGOS does not perform.
• Technical responsibility. No result, verified or not, replaces the judgment and the professional responsibility of the engineer who signs the design.