Junling Fang

Consistency verification of drilling reports is critical for engineering quality control because a single data item may be distributed across reports with different formats, units, and page structures. Existing retrieval-augmented generation methods remain sensitive to retrieval and parsing errors in such documents, whereas ultra-large models impose substantial local computing and memory costs. This study proposes a lightweight tool-augmented framework based on dual-teacher distillation and sparse mixture-of-experts (MoE) modeling. Qwen3-235B-A22B serves as the primary teacher and Qwen3-30B-A3B as the assistant teacher. Their tool-use and task-planning capabilities are transferred to a sparse MoE student upgraded from a Qwen3-1.7B dense backbone through trajectory pruning, sample decomposition, and token-level Kullback–Leibler (KL) distillation. The student adopts an eight-expert Top-2 routing architecture. Experiments on 1000 drilling reports containing 30,127 verification instances show an F1 score of 58.0 ± 0.5%, with file-level, location-level, and exact-match accuracies of 66.5%, 55.2%, and 45.0%, respectively. The model contains 9.1B total parameters and 2.8B activated parameters, and reaches a latency of 12.1 ms per forward pass and a memory footprint of 18.4 GB under bfloat16 (BF16) precision. The reported F1 score characterizes the end-to-end verification task rather than an autonomous safety decision capability. The framework is intended to support evidence localization, anomaly prioritization, and expert review in local deployment settings.

This paper presents an integrated approach of discrete fracture network modelling for a naturally fractured buried-hill carbonate reservoir in the Jingbei Oilfield by using a 3D seismic survey, conventional well logs, and core data. The ant tracking attribute, extracted from 3D seismic data, is used to detect the faults and large-scale fractures. Fracture density and dip angle are evaluated by observing drilling cores of seven wells. The fracture density distribution in spatiality was predicted in four steps; firstly, the ant tracking attribute was extracted as a geophysical log; then an artificial neural network model was built by relating the fracture density with logs, e.g., acoustic, gamma ray, compensated neutron, density, and ant tracking; then 3D distribution models of acoustic, gamma ray, compensated neutron and density were generated by using a Gaussian random function simulation; and, finally, the fracture density distribution in 3D was predicted by using the generated artificial neural network model. Then, different methods were used to build the discrete fracture network model for different types of fractures of which large-scale fractures were modelled deterministically and small-scale fractures were modelled stochastically. The results show that the workflow presented in this study is effective for building discrete fracture network models for naturally fractured reservoirs.