PIGPROX electromagnetic ILI product materials (2026-08)

PIGPROX electromagnetic in-line-inspection product materials

This auxiliary release publishes the project-owner-supplied 12-page English brochure for the PIGPROX eddy-current in-line-inspection system.

• Original PDF: PIGPROX-English-Eddy-Current-ILI-Brochure.pdf (52,563,788 bytes)
• SHA-256: 5ce81f911a344d9f8bc9807a3b174cba74ab2a3f836710b816536fe3d30705e8
• Browser preview, MEM ultralight MFL detector poster, provenance and rights notice: docs/product_materials/
• Related peer-reviewed research: docs/RELATED_PUBLICATIONS.md

These are product-presentation materials, not P110/406 experimental data or independent stress-performance validation. Product materials retain their separate rights notice and are not covered by the repository's Apache-2.0 or CC BY 4.0 licences unless separately stated.

v0.4.1 — Zenodo archival release

v0.4.1 — Zenodo archival release

Release date: 2026-08-29

Purpose

This maintenance release triggers the repository's newly enabled Zenodo
integration and makes the archive metadata explicit. It does not change the
experimental data, processing algorithms, frozen features or scientific
conclusions.

Metadata changes

• set the software version to 0.4.1 in the Python package, citation metadata
  and Zenodo metadata;
• identify the Zenodo upload as open-access software in English;
• retain Bin Gao, University of Electronic Science and Technology of China
  (UESTC), as the creator;
• retain Apache-2.0 for software and the separate CC BY 4.0 data terms described
  in DATA_LICENSE.md.

Frozen data boundary

• P110 magnetic-and-strain data remain frozen in release v0.2.0;
• 406-mm MEM, remanence and ETP data remain frozen in release v0.3.0;
• no raw measurement, strain reference, feature value or validation result is
  rewritten by this release.

The evidence boundary is unchanged: the repository supports traceable
same-pipe relative stress studies and quality control; it does not claim a
field-universal absolute-stress model in MPa.

v0.4.0 — open licensing, citation and community release

v0.4.0 — open licensing, citation and community release

Release date: 2026-08-29

What changed

• software code is now licensed under Apache License 2.0;
• experimental data, public tables, result figures and field photographs are licensed under CC BY 4.0 unless stated otherwise;
• CITATION.cff and .zenodo.json identify Bin Gao, UESTC, as the author/creator;
• the English and Chinese landing pages now expose real-data, licence, citation and reproduction entry points;
• GitHub Topics, Discussions and structured reproduction/data/method issue forms are enabled;
• CI uses the Node 24 generations of the official checkout and Python setup actions.

Data-version boundary

This platform release does not replace or silently rewrite the frozen real-data assets:

• P110 EXP2 remains frozen as data release v0.2.0;
• 406 calibration and blind repeats remain frozen as data release v0.3.0.

The scientific claim limits of those releases remain unchanged. Version v0.4.0 adds reuse rights, citation metadata and community infrastructure; it does not claim independent-pipe absolute stress validation.

v0.3.0 — 406 multimodal public dataset

v0.3.0 — 406 mm multimodal public case

This release adds the real 406 mm four-point-bending / in-line pull-test campaign while preserving the modality and truth boundaries of the existing P110 EXP2 release.

Added

• A fully documented case_studies/406_multimodal package with MATLAB source, derived CSV/JSON tables, 13 real blind-repeat figures plus one frozen calibration figure, and an independent fail-fast validator.
• The 406 calibration partition: zero load S0 plus six loaded states S1–S6, shared MEM/remanence magnetic CSVs and strain-gauge records.
• The 406 blind-repeat partition: C1 and C2 complete, C3 partial S0–S2, for 17 paired magnetic/ETP packets.
• Public field photographs with provenance, colour normalization, metadata stripping and an explicit background-person privacy note.
• Real calibration and repeatability figures on the repository front page.
• A versioned data Release index, per-file and per-asset SHA-256 manifests, channel contracts and a bilingual dataset card.

Sensor contracts

• The 406 magnetic CSV contract has 1307 named numeric fields. The first 1280 fields are 320 × (X,Y,Z,T).
• There are 32 physical columns with 10 sensing positions each. One-based odd physical columns are remanence and even physical columns are MEM; each group therefore contains 160 sensing channels.
• MEM and remanence share the same magnetic CSV stream. They must not be counted as two raw files.
• ETP is an independent 67-field stream with 20 (Amplitude, Phase) channel pairs plus acquisition fields.

Frozen evidence boundary

• MAG-F1-DW-Q90-v1: primary same-pipe relative-order/change evidence. Strict C1↔C2 repeatability after the pre-declared C2/S2 rejection: Spearman ρ 1.000, Lin CCC 0.996, normalized RMSE 3.35%.
• MEM-F4-ZSD-v1: unsigned, same-cycle-S0 auxiliary evidence. Strict C1↔C2 CCC 0.980, normalized RMSE 7.51%.
• ETP candidates: research/QC only because target specificity and negative-control gates do not pass.
• Fusion: fail-closed modality/QC eligibility gating only; no numeric fusion value. Explicit stagewise cross-modality conflict scoring is not implemented in v0.3.0.
• Blind absolute MPa: disabled. The blind-repeat partition has no contemporaneous strain/MPa truth.

The calibration MPa column is derived as 206 GPa × median bending strain; it is not a load-cell measurement and must not be copied into the blind-repeat labels.

Deliberately excluded

• blind test/data/未处理数据, which duplicates or precedes the curated clipped magnetic source tree.
• Acquisition screenshots, old single-cycle conclusions, deprecated reports, large MAT/FIG runtime products and local absolute paths.
• The old continuous ETP recording as a stage-by-stage calibration source; it represents one strain state and cannot be aligned to S0–S6.

Compatibility

The generic Python data platform remains backward-compatible with the P110 v0.2.0 package. The 406 raw-data reader is supplied as a dedicated MATLAB case because its 1307-field fragmented magnetic contract and independent complex ETP contract differ materially from the P110 three-axis CSV layout.

Download and integrity

This Release contains 13 attachments: 8 raw-data ZIP parts plus DATASET_CARD.md, channel_schema.csv,
aw_file_manifest.csv,
elease_index.json, and SHA256SUMS.txt. Verify every downloaded file against SHA256SUMS.txt before use. Extract ZIPs while preserving their canonical
aw/... paths.

v0.2.0 — P110 EXP2 real magnetic + strain dataset

P110 数据模态边界（重要）

当前 P110 EXP2 数据集只包含三轴磁传感数据和应变片数据，不包含独立采集的剩磁信号数据，也不包含涡流 ETP 数据。探头或文件夹名称中涉及“剩磁/磁化节”的文字仅为实验硬件或磁化工况命名；对应 CSV 仍是三轴磁传感器输出，不能解释为另一套剩磁测量模态。

内容

• 可追溯管道应力实验数据处理平台；
• 匿名、确定性 P110 类磁信号阅读者 Demo 及独立应变标签；
• P110 真实磁传感+应变一级案例：冻结派生表、真实结果图、Python/MATLAB 重绘程序、完整脱敏配置和失败闭锁校验；
• Python/MATLAB 数据读取接口；
• 数据质量、血缘、盲态隔离与版本指纹验证；
• 经明确授权公开的 P110 EXP2 全量数据附件：原始快照、标准化磁信号、应变、特征、目录和校验信息。

全量附件校验

• 文件：P110_EXP2_full_release_1.0.0+97b0dca62768.zip
• 大小：169,639,531 字节
• SHA-256：38cc098c9a590f05efea429073e7c17c0b2c5f8b0232a7625e50957e05fa6a7c

解压后以 dataset_index.json 为唯一程序入口，不要通过文件名重新猜测工况。新增实验应生成新配置、独立标签和新数据集版本，禁止覆盖本 Release。

全量包的 40 项 PASS 只表示文件完整性与结构一致性校验通过，不表示应力模型性能通过。案例目录内的 10 行表是经审查的真实派生证据，但目前尚未包含逐行扫描 ID、源哈希和方法版本，因此不得声称它已实现从全部原始数据独立重建。

工程边界

当前磁特征适合应力相关性、相对分级和应力集中筛查。未经新管道标定、独立盲测和不确定度验证时，不得直接解释为绝对应力 MPa。