Report APPL-PGV-01 rev B is issued in full on request. The abstract is published online at advikprotofab.com/publications, and the complete report, ready to print, is sent by Advik Protofab Private Limited to anyone who asks for it.
Technical report · Fixed-pitch marine propellers
Propeller Geometry Validation
How closely a parametric propeller model reproduces four manufactured propellers: their blade geometry, boss, mass and inertia, measured against the figures on the makers' own drawings.
1Abstract
Four fixed-pitch propellers from three manufacturers, 5.75 to 8.5 m in diameter, were used as references. For each one the maker's drawing states the radial geometry, the boss, the cast weight and a polar moment of inertia. The model was checked in two ways: rebuilt from the full drawing data, and generated from the one-page particulars alone.
The report sets out where the two approaches agree with the drawings, where they do not, and why. It quantifies the variation between drawings, between classification rule minima and the thicknesses makers actually cast, and it states the resulting uncertainty when a propeller is generated from its particulars sheet. The largest remaining errors are shown to be matters of design practice rather than of modelling.
2Headline findings
| Figure | Quantity | Finding |
|---|---|---|
| +0.22% | Mass, from full drawing data | Built from a complete drawing, the reference propeller's mass is within 0.22% of the stamped weight and its expanded area within 0.06% of the stated area. Neither figure is used as a calibration target. |
| −0.7% | Section area by polygon | Integrating a section as the polygon through its tabulated ordinates under-reads its area by 0.70 to 0.75% at every radius. The fair curve through the same ordinates is the section that gets cast. |
| ±7% | Boss volume from a derived rule | The three drawn bosses share one shape, a single taper, larger forward, about 0.18 D long. A rule built from that shape predicts each boss within 7% under cross-validation when the shaft diameter is known. A barrel-and-cone approximation overestimates the same bosses by 14% and 67%. |
| −2.0% | Independent case | The independent validation propeller, none of whose blade data enters the model, generated from its particulars sheet only, is within 2.0% of its stamped weight and 5.5% of its inertia. The baseline method gives −8.2% and −16.7%. |
| 1.02 to 1.13 | Design margin over the class minimum | Makers design blade thickness 2 to 13% above the classification minimum, varying by maker and by radius. That spread is not predictable from any rule, and it is the largest residual error source when a propeller is generated from its particulars. |
3Terms used above
The report defines the terms used above as follows, so that a figure is read as the report means it.
- Baseline
- The conventional parametric approximations, namely a barrel-and-cone boss and a generic blade outline with thickness a fixed fraction of diameter, carried through for comparison.
- Proposed
- The method of this report.
- In-sample
- Constants derived with the case included.
- Cross-validated
- Each case predicted from constants derived from the other cases only, that is leave-one-out.
4Evidence base
The four references are four and five bladed fixed-pitch propellers for large merchant ships, keyless-fitted on a 1:20 taper with a hydraulic nut, in nickel-aluminium and manganese-aluminium bronze. Two share a manufacturer and the other two come from one each. They are identified in the report by letter only, and no manufacturer, vessel or owner is named. One of the four is held back as the independent validation case: none of its blade data enters any table or constant in the report, so the figure quoted for it is a prediction and not a fit. The other three supply the radial distributions, and results for them are in-sample.
5Fitness for use
| Application | Status | Basis |
|---|---|---|
| Mass and inertia for shafting analysis, from full drawing data | Ready | Within 0.2 to 1.2% on mass and 0 to 1% on inertia across two independent drawings. |
| General-arrangement and presentation geometry | Ready | Tip, edges, boss and fillet follow the drawings. |
| Concept design from particulars | Usable within a stated tolerance | RMS 4.3% on mass and 6.5% on inertia over four propellers; 2.0% and 5.5% on the independent case. |
| Solid model for CAD and finite-element pre-processing | Beta | A valid single solid, accepted by two kernels, with the fillet approximated and the edges cut square. |
6Further work identified
- Comparison against a measured casting or a maker's CAD surface.
- Further drawn references, particularly five and six bladed and highly skewed designs.
- The classification rule's own definition of the section coefficient.
- Skew and rake distributions derived from drawings rather than from the present simple laws.
- A true blend in place of the approximated root fillet.
7Contents of the full report
- Summary
- Evidence base
- Method
- Blade sections and their integration
- Tip closure
- Boss
- Root fillet
- Blade thickness from the class rule
- Radial distributions
- Whole-propeller validation
- Solid model integrity
- Observed variation and uncertainty
- Guidance for designers and owners
- Fitness for use
- Appendix: formulas, nomenclature and revision record
8Revision record
| Revision | Date | Description |
|---|---|---|
| A | 2026-10-05 | First issue. |
| B current | 2026-10-05 | Editorial. Comparative terms standardised as baseline, proposed, in-sample and cross-validated, and defined in the method section. The solid model section restated as requirements. No change to data, results or conclusions. |
Obtaining the full report
Report APPL-PGV-01 rev B carries the method, the derivations, the complete reference tables and the uncertainty budget. It is issued on request.
Please state who is asking, the organisation, and the use intended. Where the subject matter requires it, a non-disclosure agreement is executed before the document is sent.
APPL-PGV-01 rev B. © 2026 Advik Protofab Private Limited. This abstract is published for information. It is not licensed for reproduction, redistribution or inclusion in another document, and it is kept out of web archives and search-engine caches. Text selection and printing are disabled on this page as a deterrent, not as a security measure. Please request the report rather than copying from this page; attribution and the correct revision matter to anyone who later relies on the figures.