PhD researcher · AI for ScienceHarbin Institute of Technology

Peng JI季鹏

AI for Science: from impact signals to spacecraft protection.

I am a PhD student at Harbin Institute of Technology, studying damage and residual protection prediction from acoustic-emission signals. My work spans physics-guided machine learning, simulation automation, and research agents for spacecraft protection and sensing design.

Open to AI4S research internships and collaborations.

Get in touch GitHub Google Scholar

Scientific machine learning Simulation automation Research agents

Selected projects

From scientific questions to working models and tools.

Scientific machine learning

Whipple RPP

From acoustic signals to residual protection

Predicting residual protection fields in damaged Whipple shields from acoustic-emission signals, with supervision defined by ballistic-limit equations (BLEs).

Whipple RPP overview linking damaged-shield measurements, ballistic-limit supervision, acoustic-emission spectra and comparisons of four surrogate models. — Study overview: acoustic-emission spectra and BLE-defined targets connect damage sensing to residual-protection prediction.Paper source View full-size figure

My work. I built the simulation and training pipeline, compared deep and traditional surrogates, and evaluated BLE-guided regularization and robustness.

Outcome. Within the paper's test conditions, U-Net led on full-field and minimum-RPP prediction; VQ-SM better preserved weak regions. BLE-guided gains varied by condition.

Public code Paper

First-author study · Aerospace Science and Technology, 2026

Damage assessment

AE-PDSM

From acoustic signals to damage morphology

Predicting perforation state and two-dimensional pitting-damage morphology from acoustic-emission signals.

AE-PDSM graphical abstract showing the pitting-damage assessment problem, LS-DYNA data generation, perforation classification, morphology reconstruction and a crater-density comparison. — Graphical abstract: simulation data and a surrogate model connect acoustic-emission signals to perforation state and pitting-damage morphology.Paper source View full-size figure

My work. I built the FE–SPH data pipeline, surrogate model and cross-validation workflow, and conducted the real-impact proof of concept.

Outcome. The model predicted perforation state and reconstructed damage morphology in simulation tests, with a separate proof of concept using real impact signals.

Public code Paper

First-author study · International Journal of Mechanical Sciences, 2026

Research tooling

elsevier-figure-style

Reusable figure styles and automated checks

An independent, unofficial toolkit and Agent skill for applying published research-figure requirements.

My work. I independently developed Python/R styles, Python source checks, export-metadata validation and Agent visual QA.

Outcome. Released with reusable styles, checks, examples, tests and a software DOI.

Repository Software DOI

Open-source research tool · MIT license

Research in progress

Auditable research agents

Screening and checking spacecraft designs

Screening and checking spacecraft protection and sensing candidates under mass, penetration-risk and geometric-coverage constraints.

My work. I independently lead the design and implementation of a Designer–Runner–Reviewer workflow, connecting structured state, deterministic tools and independent review.

Outcome. The workflow tracks evidence and supports execution-record replay; development is ongoing.

Code is not yet public.

Publications

Journal articles · first and co-first author work listed first

2026First author

Surrogate model-based online quantitative prediction of pitting damage in spacecraft

Peng Ji, Runqiang Chi, Wuxiong Cao, Jiaxin Gao, Houyang Wang
International Journal of Mechanical Sciences · 315, 111416

Paper Public code
2026First author

Ballistic limit equation-constrained and data-driven residual protective performance assessment of Whipple shields with pitting damage

Peng Ji, Runqiang Chi, Wuxiong Cao
Aerospace Science and Technology · 179, 113473

Paper Public code
2025Co-first author · Listed second

Localization of hypervelocity impact-induced damage in thermal protection systems of aerospace vehicle using a weighted-KNN algorithm

Feng Du, Peng Ji, Wuxiong Cao, Jiaxin Gao, Yibo Lu, Jianxia Gao, Xianpeng Zhou, Runqiang Chi, Diqi Hu, Baojun Pang
Acta Astronautica · 237, 326–338

Paper
2025

Quantitative characterisation of debris cloud-induced pitting damage in AL-Whipple shields of spacecraft using infrared thermography

Jiaxin Gao, Runqiang Chi, Han Feng, Wuxiong Cao, Peng Ji, Xianpeng Zhou, Yibo Lu, Diqi Hu, Shengpeng Chen, Baojun Pang
Nondestructive Testing and Evaluation · 40(9), 4136–4163

Paper
2026

An ameliorated GWO-KNN method for localization of hypervelocity impact on the complex spacecraft structure

Wuxiong Cao, Jianxia Gao, Jiaxin Gao, Runqiang Chi, Peng Ji, Feng Du, Diqi Hu, Baojun Pang
Advances in Space Research · 77(12), 12221–12240

Paper

Open-source contributions

Contributions to the scientific Python ecosystem

scikit-fem

Finite elements

Fixed periodic high-order element consistency; added separate-array mesh I/O and round-trip tests; repaired documentation links.

#1223 #1222 #1221

Author-submitted PRs merged upstream.

PyDyna

LS-DYNA automation

Fixed Windows paths, solver discovery, container arguments and memory-unit handling; added tests and run documentation.

#1336 #1337

My original contributions were integrated by maintainers and merged through these two mirror PRs.

PyVista

3D visualization

Fixed singleton-array squeeze to preserve views, dataset association and write-back notifications, with regression tests.

#9276

Author-submitted PR merged upstream.

Zarr

Scientific array storage

Fixed v2 uint8 aliases and canonical serialization; tested that read-only access leaves metadata unchanged.

#4381

Author-submitted PR merged upstream.

Trimesh

Geometry processing

Fixed SVG rotation units, direction and composed transforms, with regression tests.

#2602

Author-submitted PR merged upstream.

Education

2023.08–2028.08

PhD · Integrated master's–PhD program

Harbin Institute of Technology

School of Astronautics · Aeronautical and Astronautical Science and Technology. Expected graduation: August 2028. Research on quantitative identification of hypervelocity-impact damage and performance assessment.

2019.09–2023.06

BEng · Flight Vehicle Design and Engineering

Harbin Institute of Technology

School of Astronautics. Undergraduate thesis on neural-network methods for low-Earth-orbit satellite prediction.

Let’s connect

Open to AI4S research internships and collaborations.

jipeng@stu.hit.edu.cn

Google Scholar GitHub