Measured fabric behaviour makes the digital sample testable

Ancutienė and Sinkevičiūtė investigated virtual fit against tensile, bending and shear properties, together with composition, thickness and area density. Their work distinguishes the material's mechanical description from its visual representation. Its findings belong to the fabrics and software studied, rather than establishing a universal ranking of inputs for current systems.[S05]

GFT proposes treating surface appearance and mechanical behaviour as separate asset components with an explicit link. A photograph, colour description or weave image belongs in the appearance component. Measured response records and their translation into simulator parameters belong in the mechanical component. A supplier's product name can help locate a physical material, but should not serve as its calibration record. The release question should be answerable without opening the presentation render: which specimen supports this model, and which decision was it checked against?

The proposed identity record should connect the asset to the supplier article, receipt reference, roll or lot, finish, face, direction and retained specimen. Where the production process changes the material, record the tested state as well. A raw swatch, prepared printing substrate and finished printed panel should not share an unqualified calibration identity. This is a proposed traceability rule, not a claim that every treatment produces a material change large enough to matter. The purpose is to make that possibility testable rather than erase it through naming.

For an initial service, GFT recommends limiting the library to the materials actually needed by the chosen product family. Admit a provisional material for visual exploration, with its restriction visible to the designer. Require a physical linkage before attempting calibration. Keep any generic asset that it replaces as a separate reference, so the difference between measured input and convenient default remains recoverable. A smaller library with accountable records is a more useful starting point than an extensive catalogue whose apparent completeness obscures its evidence gaps.

The illustrative dictionary in Exhibit 22 sets out what the proposed asset would retain. It defines record fields, rather than claiming measured values or prescribing a laboratory method. A field should name its measurement boundary, its unit and the associated raw record. Where a simulator cannot accept a measured property directly, the mapping belongs in the record too. An unsupported field should be marked unavailable; assigning a plausible default would hide a limitation that the evaluator needs to see.

For this proposed record, mass should state whether it means specimen mass or mass per area. Thickness should retain the measurement conditions rather than appear as an unexplained scalar. A tensile entry should distinguish its raw load and extension observations from a fitted software coefficient. These are data-definition requirements: they stop unlike quantities from being treated as interchangeable merely because an import dialogue accepts a number. The dictionary cannot establish that the selected tests cover every behaviour relevant to the final garment.

The directional finding in Ancutienė and Sinkevičiūtė's diagonal-cut garment study concerned the influence of weft tensile properties. It is a warning against erasing orientation during material transfer, not evidence that weft tension will dominate every product. GFT's proposed import check retains directional identity before investigating which properties matter to the chosen article.[S05]

Under that proposal, mark the principal directions on the physical specimen and carry the convention into the test file. The file should identify whether an angle is measured from warp, weft or another defined reference. The patternmaker should then check how that reference appears on the imported pattern piece. A screen rotation made to arrange pieces conveniently must not be mistaken for a change to the intended fabric grain. The review should compare the recorded direction with the pattern's production intent, rather than infer it from how a weave image happens to look.

GFT recommends a direction-transfer check using a deliberately asymmetric test asset. Give its principal directions visibly different diagnostic settings, inspect the software's reported orientation and then restore the measured asset. This checks the data route; the diagnostic asset must remain excluded from approval work. If an exchange format drops a direction field or substitutes a default, record the loss and correct it through an explicit import step. An unexplained discrepancy should stop the transfer rather than become a convenient adjustment to the garment.

For bias-cut work, the proposed record should state the angle and the scope of evidence behind the model's off-axis response. If a laboratory route tests only principal directions, do not describe an oblique response as independently measured. Keep the distinction between measurement and simulator interpretation visible to the reviewer. Where the product's decision depends on that interpretation, add an appropriate off-axis comparison to the calibration plan. The question is whether the representation supports the intended task, without extending the permissions of the principal-direction tests by assumption.

The same control should apply when an asset is duplicated. GFT proposes retaining the physical identity and direction record in the copy, while changing the asset revision and purpose. A new surface image should not silently create a new mechanical qualification. Conversely, replacing the tested substrate while keeping its old image should not preserve the old permission. The record should reveal whether the designer changed appearance, mechanics or physical origin, because those changes call for different reviews.

This proposed discipline also gives fault isolation a useful order. Check specimen labels, test orientation, import mapping and piece grain before compensating through stiffness edits. Keep an orientation discrepancy as its own failure category. Otherwise the team could obtain an attractive result by changing mechanics to offset a direction error, then carry the incorrect material into another product. The local control is straightforward: preserve the physical reference through each transfer and document any deliberate departure before using the output to make a decision.

For operational use, GFT proposes an import acceptance record signed by the technical role responsible for the pattern. Record the source asset revision, destination application, expected direction convention and result of the diagnostic check. Where pieces intentionally use different grain directions, identify them individually. Keep any approved exception attached to those pieces rather than changing the global fabric record. This allows a later reviewer to distinguish a deliberate product choice from an import defect. It also prevents an exception created for a particular panel from becoming the default behaviour of every subsequent garment that uses the same material.

Youn and co-authors explicitly included particle-distance effects in their investigation of virtual fabric. That makes resolution a variable to control in a comparison, rather than a presentation preference that can remain undocumented. The opened abstract gives no universal particle-distance setting that GFT can prescribe for a new software version or product.[S04]

GFT proposes separating an editing setup from an evaluation setup. The editing setup serves interactive construction work. The evaluation setup is the frozen configuration used for the comparison and its release record. Store the application build, solver mode, resolution, collision settings, layer arrangement, contact assumptions, initial placement and stopping rule with the material revision. A screenshot of the final garment is insufficient to recover those choices. When the application generates a report, retain it with the native scene and record any setting entered outside the report.

The proposed resolution check should hold the material and geometry fixed while changing the evaluation resolution. Compare a defined output rather than asking only whether the image looks better. Record computation time as an operational observation when measurements become available, without treating faster execution as validation. If the selected output continues changing enough to alter the decision, retain a hold on that decision and investigate. A finer setting is a test condition; it should not be described as more accurate solely because it is more demanding to run.

GFT also recommends a repeat-run check at the chosen evaluation setup. Its purpose is to reveal dependence on initialisation, placement or undocumented intervention. The operator should retain failed runs and their cause rather than keeping only the final attractive scene. If manual pins, smoothing or local constraints are needed to complete a run, disclose them and explain whether they have physical counterparts. A result obtained under those constraints should not inherit permission from an unconstrained comparison merely because the underlying fabric filename is unchanged.

Changes to software should trigger a documented impact check under the proposed system. Preserve a reference scene and its prior outputs, then compare the new build before migrating approval work. If the new build changes imported parameters or evaluation behaviour, issue a new qualification record. Keep legacy scenes readable where practical and state any migration limitation. This is a proposed change-control requirement, not a claim that updates necessarily make a model worse. It makes a changed computational experiment visible before its outputs are compared with earlier approvals.

Finally, GFT recommends preserving the meaning of the output itself. A change in camera, projection, scale or image processing should be recorded when the evaluation uses image-derived geometry. Keep a scale reference and a consistent measurement procedure. The technical reviewer should be able to reproduce the reported quantity from the archived output without guessing which silhouette edge was selected. Resolution control cannot rescue an output definition that changes between the physical and virtual observations.