Replication Data and Simulation Code for the Neuro-Cognitive Safety Framework (NCSF): EEG Dataset and Agent-Based Modeling (ABM)

This repository contains the replication data, code, and computational materials supporting the study: “Heidari et al. – Enhancing Safety in Construction: A Framework Predicting Neuro-Cognitive Tipping Points from Environmental Stressors”. It accompanies the Neuro Cognitive Safety Framework (NCSF), which integrates construction safety, cognitive neuroscience, and computational simulation to explore how environmental multi-physics stressors (thermal/acoustic) degrade worker cognitive capacity, laying the groundwork for Neuro-Cognitive Digital Twins (NCDTs). To ensure transparency, traceability, and reproducibility, the repository is structured into four integrated layers, providing all necessary files: 1. Neurophysiological Data Layer EEG Dataset (EEGLAB .set format): Preprocessed, ICA-cleaned EEG data from controlled cognitive workload experiments (n-back, N=16). Isolates components relevant to working memory to provide empirical grounding for the model. 2. Calibration Layer NCSF MATLAB Calibration Code.m & Supplementary MATLAB Calibrated Data.xlsx: Scripts and output datasets used to calibrate cognitive load dynamics, fatigue accumulation, and neural response parameters against the empirical EEG data. 3. Simulation Layer NCSF Agent-Based Model Proof of Concept.py: A Python-based stochastic Agent-Based Model (ABM). Simulates workers as cognitive agents over an 8-hour shift, incorporating probabilistic cognitive transitions to produce safety indicators (e.g., Mean/Peak Risk). 4. Validation & Sensitivity Analysis Layer Supplementary Sensitivity Analysis Data.csv: Data from a 30-iteration Monte Carlo replication with ±30% parameter perturbations, evaluating model robustness and identifying critical tipping points. Traceability & Reproducibility Supplementary Data Traceability Dictionary.xlsx: A structured dictionary linking neuroscience constructs, cognitive mechanisms, and safety engineering concepts. Workflow Pipeline: EEG preprocessing → parameter calibration → ABM simulation → Monte Carlo sensitivity analysis. Keywords Neuro Cognitive Safety Framework (NCSF), Agent-Based Modeling (ABM), Construction Safety, Cognitive Load, EEG, Environmental Stressors, Human Factors, Monte Carlo Simulation, LOD, Digital Twins, Traceability.本仓库包含支撑下述研究的复现数据、代码与计算支撑材料：《Heidari 等人——提升建筑施工安全：基于环境应激源预测神经认知临界点的框架》。 本仓库配套神经认知安全框架（Neuro Cognitive Safety Framework, NCSF），该框架整合建筑施工安全、认知神经科学与计算仿真技术，探究环境多物理场应激源（热/声学）如何削弱作业人员的认知能力，为神经认知数字孪生（Neuro-Cognitive Digital Twins, NCDTs）奠定研究基础。 为保障研究的透明度、可追溯性与可复现性，本仓库划分为四个集成层级，提供全部必要文件： 1. 神经生理学数据层级 脑电图（Electroencephalogram, EEG）数据集（采用 EEGLAB .set 格式）：来自受控认知负荷实验（n-back 任务，样本量 N=16）的预处理后、独立成分分析（Independent Component Analysis, ICA）去伪影脑电数据。该数据集分离出与工作记忆相关的神经成分，为模型提供实证支撑。 2. 校准层级 包含《NCSF MATLAB 校准代码.m》与《补充 MATLAB 校准数据.xlsx》：用于基于实证 EEG 数据校准认知负荷动态、疲劳累积与神经响应参数的脚本与输出数据集。 3. 仿真层级 包含《NCSF 智能体模型概念验证.py》：一款基于 Python 的随机智能体模型（Agent-Based Model, ABM）。该模型将作业人员建模为认知智能体，模拟 8 小时工作班次内的作业流程，纳入概率性认知转换逻辑以生成安全指标（如平均/峰值风险）。 4. 验证与敏感性分析层级 包含《补充敏感性分析数据.csv》：来自 30 次蒙特卡洛复现实验的数据，实验中对模型参数进行 ±30% 的扰动，用于评估模型鲁棒性并识别关键临界点。 可追溯性与可复现性 《补充数据可追溯性词典.xlsx》：一份结构化词典，用于关联神经科学概念、认知机制与安全工程相关理念。 工作流管线：脑电图预处理 → 参数校准 → ABM 仿真 → 蒙特卡洛敏感性分析。 关键词 神经认知安全框架（Neuro Cognitive Safety Framework, NCSF）、智能体建模（Agent-Based Modeling, ABM）、建筑施工安全、认知负荷、脑电图（Electroencephalogram, EEG）、环境应激源、人因工程、蒙特卡洛仿真（Monte Carlo Simulation）、细节层次（Level of Detail, LOD）、数字孪生、可追溯性。