Zhengpeng Liu ZPL-03

👋 Hi, I'm Zhengpeng Liu

AI for Science · Industrial CAE Software · Composite Mechanics

M.Eng. Aerospace Engineering, Xi'an Jiaotong University — GPA 3.89/4.0
B.Eng. Engineering Mechanics, Wuhan University of Science and Technology — GPA 3.77/4.0
Intern Beijing Shenzhou Aerospace Software Technology Co., Ltd. — Industrial CAE R&D

Bridging computational solid mechanics and machine learning to build next-generation intelligent simulation tools.

AI for Science (AI4S): Physics-informed & data-driven models for materials and structures
Composite Materials Mechanics: Failure criteria, constitutive modeling, damage evolution, multi-scale analysis
Intelligent CAE: ML-accelerated solvers, surrogate modeling, reduced-order models (ROM)
Industrial Software R&D: CAD/CAE kernel development, solver architecture, FreeCAD secondary development

ABAQUS Secondary Development: Parametric modeling (Python scripting / GUI plugins), UMAT/VUMAT/USDFLD subroutine authoring, cohesive element and PBC applying
Structural analysis: buckling, progressive damage, thermo-mechanical coupling

Architectures: DNN · CNN (ResNet) · ICNN (Input Convex NN) · GAN · DeepONet · PINNs
Applications: Failure envelope prediction, surrogate constitutive modeling, operator learning for PDEs, deep transfer learning (DTL)
Frameworks: PyTorch · TensorFlow · Scikit-learn
LLM/Agent: Ollama · ChromaDB · LangChain · Claude Code (agentic workflows)

Python — ML pipelines, ABAQUS scripting, GUI (PyQt6 / PyVista), scientific computing
C++ / Fortran — Solver kernels, constitutive model implementation (Eigen, pybind11)
Parallel Computing: OpenMP · MPI · CUDA (cuSPARSE)
Dev environment: VS2022 · WSL2 (Ubuntu) · Intel oneAPI · Anaconda

Project	Description	Stack
CSDM	Multi-agent AI system for composite stiffened panel structural design; 6-agent pipeline from NL input → LLM candidate gen + RAG retrieval + surrogate screening → ABAQUS FEM validation → report	Python · LLM · RAG · ABAQUS
MDD-TFC	Model-data-driven transverse failure criterion for UD-CFRP; ResNet-101 + Tsai-Wu + ICNN maps RVE microstructure to full failure envelope via only 4 load cases; validated via few-shot transfer learning (TWS Q1, 1st author)	PyTorch · ABAQUS
FE-NN-KT	Multi-scale homogenization framework: offline RVE sampling → surrogate constitutive model training → online FE-NN macro solver; coarse-to-fine transfer reduces data cost 3.3×, macro analysis 20× faster than DNS	PyTorch · ABAQUS · Q4 FEM
Composite Intelligent Design Platform	Hierarchical design platform (microstructure → laminate → stiffened panel → fuselage); sub-second inference with per-level AI models; supports user fine-tuning	PyQt6 · TF · PyTorch
CAE FEM Software	Lightweight FEM solver: PyQt6 + PyVista frontend, C++/Eigen compute core, pybind11 bridge, Gmsh meshing, real-time stress/displacement rendering	C++ · PyQt6 · pybind11 · Gmsh

Papers (2 SCI papers, total IF 15.5):

📝 Liu Z, et al. A model and data-driven transverse failure criterion for UD-CFRP composites. Thin-Walled Structures, 2026. (Q1 Top, 1st author)
📝 Wang Q, Li Z, Liu Z, et al. Interpretable prediction of ultimate strength of composite pressure hulls using symbolic regression. Composites Part A, 2026. (Q2 Top, 3rd author)

Software Copyrights (4×, all 1st inventor):

Awards:

Custom ABAQUS plugins & batch simulation pipelines (parametric modeling, pre/post-processing)
UMAT/VUMAT subroutine development for composite constitutive models
ML surrogate models for mechanical property prediction & inverse design
Multi-agent AI systems for simulation automation (LLM + RAG + FEM in the loop)
DeepONet / PINN implementation for engineering PDE problems
CAE solver prototyping with C++/Python/Fortran hybrid architecture