基于液压支架工作阻力数据进行工作面来压判识需解决2个问题：一是如何从海量的工作阻力数据中提取循环末阻力数据，二是如何有效利用提取出的循环末阻力数据对工作面是否来压实现有效判断。现有的循环末阻力提取方法大多依赖固定规则和经验值参数，在复杂工作面环境下准确性低且适应性差。针对该问题，提出一种基于割煤循环智能检测的工作面来压判识方法。将割煤循环检测转化为二分类问题，使用支持向量机分类器对割煤循环结束时刻进行智能检测，以自动判别割煤循环的结束时刻；在获取所有割煤循环结束时刻的基础上，提取各支架循环末阻力数据；通过数据融合生成能够反映工作面整体压力状态的单序列数据，并基于来压判定公式进行工作面来压判识。基于不连沟煤矿某工作面的液压支架工作阻力数据进行实验，结果表明，该方法割煤循环检测的精确率、召回率、F₁分数分别为85.91%，81.84%，83.83%，来压判识的精确率、召回率、F₁分数分别为79.43%，78.76%，79.09%，均优于滑动窗口极值法和阈值法，在识别循环末阻力和工作面来压判识方面具有显著优势。

The method for identifying face pressure based on hydraulic support working resistance data needs to address two issues: first, how to extract the cycle-end resistance data from large volumes of working resistance data, and second, how to effectively utilize the extracted cycle-end resistance data to determine whether face pressure is occurring. Most existing methods for extracting cycle-end resistance rely on fixed rules and empirical parameter values, which have low accuracy and poor adaptability in complex working face environments. To address this issue, an intelligent detection method for face pressure identification based on coal cutting cycles was proposed. Coal cutting cycle detection was transformed into a binary classification problem, using a support vector machine (SVM) classifier to intelligently detect the end time of coal cutting cycles, automatically identifying the end of each coal cutting cycle. After obtaining the end times of all coal cutting cycles, the cycle-end resistance data for each support was extracted. Data fusion was performed to generate a single sequence of data that reflects the overall pressure state of the working face. Face pressure identification was then made based on a pressure judgment formula. Experiments were conducted on hydraulic support working resistance data from a working face in a non-contiguous coal mine. The results showed that the proposed method had precision, recall, and F₁ scores of 85.91%, 81.84%, and 83.83%, respectively, for coal cutting cycle detection, and precision, recall, and F₁ scores of 79.43%, 78.76%, and 79.09%, respectively, for face pressure identification These results are superior to the sliding window extreme value method and threshold method, demonstrating significant advantages in cycle-end resistance identification and face pressure judgment.