An Agency-specific Project Authoring Advisor: An LLM-based RAG System

Advanced Engineering Informatics (under review).

This study builds the first AEC-agency project-authoring advisor by coupling a vectorized technical document database with retrieval-augmented generation framework and prompts optimized through an adversarial, multi-dimensional LLM-assisted evaluation method, yielding a tool helps engineers on project authoring and designing.

Zhong, Y., & El-Diraby, T. (2025)

Domain-specific language models pre-trained on construction management systems corpora

Automation in Construction, 160, 105316.

Builds the first CMS domain corpus and pre-trains BERT/RoBERTa variants, then fine-tunes them on text classification and NER tasks. Achieves 5.9 % and 8.5 % F1-score gains over general PLMs, proving the value of domain adaptation for construction-focused NLP.

Zhong, Y., & Goodfellow, S. D. (2024)

Generative project question answering system: triangulating three approaches for project authoring

Automation in Construction, 160, 105316.

Develops a retrieval-augmented GPT-4–based QA advisor for AEC technical docs, using a vector database and prompt engineering to cut hallucinations and boost relevance. Expert trials show it outperforms Google and doc-search in accuracy and response speed.

Zhong, Y., & El-Diraby, T. (2024)

Shoreline Recognition Using Machine Learning Techniques

IOP Conference Series: Earth and Environmental Science (Vol. 1101, No. 2, p. 022025).

Applies supervised ML classifiers to satellite imagery for automated shoreline extraction, comparing SVM, RF, and CNN approaches. Demonstrates > 90 % boundary accuracy across coastal scenes, enabling scalable, high-precision shoreline monitoring.

Zhong, Y., & El-Diraby, T. (2022)

Monitoring tilting angle of the slope surface to predict loess fall landslide: an on-site evidence from Heifangtai loess fall landslide in Gansu Province, China

Landslides, 1-11.

Deploys wireless tilt sensors on a Gansu loess slope and tracks real-time angle changes to predict landslide onset. Correlates tilt acceleration with failure stages, validating tiltmeter networks as cost-effective early-warning tools.

Wang, H., Zhong, P., Xiu, D., Zhong, Y., Peng, D., & Xu, Q. (2022)

A novel wireless underground transceiver for landslide internal parameter monitoring based on magnetic induction

International Journal of Circuit Theory and Applications, 49(6), 1549-1558.

Introduces the MI-S125-III magnetic-induction transceiver that wirelessly transmits tilt-sensor data up to 5.3 m through soil. Field tests in a physical landslide model confirm reliable angle monitoring, paving the way for buried, real-time early-warning networks.

Wang, H., Zhuo, T., Zhong, P., Wei, C., Zou, D., & Zhong, Y. (2021)

A preliminary study of environmental monitoring using embedded censors in the soil

Japanese Geotechnical Society Special Publication, 9(5), 164-168.

Embeds fiber-optic strain, temperature, and moisture sensors into soil tanks to link strain signals with environmental changes (evapotranspiration). Demonstrates feasibility of in-ground WSNs for green infrastructure, offering continuous, real-time soil‐condition monitoring.

Mukai, K., Zhong, Y., Hubbard, P., & Soga, K. (2021)

Research on crack monitoring at the trailing edge of landslides based on image processing

Landslides, 17(4), 985-1007.

Proposes an image-processing workflow (preprocessing + Otsu binarization + Canny edges) and a custom interval median comparison algorithm to extract azimuth and displacement of trailing-edge cracks. Validated on 3D, gravel, soil models and a real landslide site, demonstrating accurate early-warning indicators.

Wang, H., Nie, D., Tuo, X., & Zhong, Y. (2020)