标题： 基于可解释人工智能的故障力学广义时间序列分析 显示英文标题

标题： Generalised Time-Series Analysis of Fault Mechanics Using Explainable AI

摘要： 理解断层如何萌生和扩展是地震科学和灾害评估中的关键问题。 本研究通过将时延神经网络（TDNN）应用于多参数声发射（AE）数据，考察了在三轴压力范围为5至40 MPa条件下Alzo花岗岩中的断层发育情况。 TDNN整合了由波形衍生的属性（包括峰值延迟和散射衰减）以及基于事件的度量指标（如时间分布、古登堡-里克特b值和空间分形维数），以表征从分布式微裂纹到局部化断层的转变。 通过遗传算法优化后，TDNN动态加权这些参数，从而能够准确地表征断层生长阶段。 我们的结果描绘了断层演化的三个不同阶段：由弹性波散射变化指示的随机微裂纹萌生，由声发射空间和震级分布演变反映的断层萌发，以及由峰值延迟指数增长和b值偏移标志的断层合并。 该模型预测了不同压力和失效机制下（从轴向劈裂到剪切局部化）应力降的时间和幅度，通过可解释的人工智能模型提供了对断层力学的更深入见解。 显示英文摘要

摘要： Understanding how faults nucleate and grow is a critical problem in earthquake science and hazard assessment. This study examines fault development in Alzo granite under triaxial pressures ranging from 5 to 40 MPa by applying a Time Delay Neural Network (TDNN) to multi-parameter acoustic emission (AE) data. The TDNN integrates waveform-derived attributes, including peak delay and scattering attenuation, with occurrence-based metrics such as time distributions, Gutenberg-Richter b-values, and spatial fractal dimensions, to characterize the transition from distributed microcracking to localised faulting. Optimised via genetic algorithms, the TDNN dynamically weights these parameters, enabling accurate characterisation of fault growth stages. Our results delineate three distinct phases of fault evolution: nucleation of random microcracks indicated by changes in elastic wave scattering, initiation of fault growth reflected in evolving AE spatial and magnitude distributions, and fault coalescence marked by exponential increases in peak delay and b-value shifts. The model predicts the timing and magnitude of stress drops across varying pressures and failure mechanisms, from axial splitting to shear localisation, providing deeper insights into fault mechanics through explainable AI models.