从钻进过程参数采集、数据处理、异常工况识别3方面对煤矿井下钻进工况识别方法进行了分析，提出了由数据采集层、处理层、工况识别层组成的煤矿井下常见钻进工况识别算法架构。其中，数据采集层能对钻进参数进行采集；数据处理层包含异常点的数据清洗、特征参数的提取以及传感器多源信息的融合；工况识别层采用机器学习中的分类算法与优化算法，通过结合2种及以上识别算法形成混合智能工况识别算法，对带有工况分类标记的钻进参数进行数据学习和模型训练，最终实现对钻进工况的智能识别。根据安徽淮南某煤矿井下典型钻进工况现场扭矩、泵压、钻速等钻进参数，构建了基于鲸鱼算法（WOA）优化后的核极限学习机（KELM）识别模型识别典型工况，通过对训练集样本进行学习，采用识别准确率优于SVM、KNN等识别模型的WOA-KELM模型，实现了对典型工况的智能识别。

An analysis was conducted on the identification methods of underground drilling conditions in coal mines from three aspects: parameter collection, data processing, and abnormal condition recognition. A framework for identifying common underground drilling conditions in coal mines was proposed, consisting of a data collection layer, a processing layer, and a condition recognition layer. Among them, the data acquisition layer can collect drilling parameters; the data processing layer includes data cleaning of outlier points, extraction of feature parameters, and fusion of multi-source information from sensors; the working condition recognition layer adopts classification algorithms and optimization algorithms in machine learning, and combines two or more recognition algorithms to form a hybrid intelligent working condition recognition algorithm. It learns data and trains models for drilling parameters with working condition classification labels, ultimately achieves intelligent recognition of drilling working conditions. Based on typical drilling parameters such as torque, pump pressure, and drilling speed in a coal mine in Huainan, Anhui Province, a nuclear extreme learning machine (KELM) recognition model optimized using the whale algorithm (WOA) was constructed to identify typical working conditions. By learning from the training set samples, the WOA-KELM model with higher recognition accuracy than SVM, KNN, and other recognition models was adopted to achieve intelligent recognition of typical working conditions.