基于改进GRU-TCN的磨煤机故障诊断算法

Coal Mill Fault Diagnosis Algorithm Based on Improved GRU-TCN

摘要:

磨煤机系统的监测与诊断对电厂的安全运行至关重要。由于真实故障数据的稀缺性以及故障与正常数据之间的不平衡，传统数据驱动的故障诊断方法在故障识别上表现不佳，有时甚至会产生误判。为了高效地识别磨煤机在不同工况下的典型故障，设计了一种结合了卷积块注意力模块的GRU-TCN融合算法，用于建立磨的故障识别模型，新算法不仅能提升分类准确性，还能实现故障的提前预警。首先，通过调整磨煤机故障生成模型的关键参数，模拟断煤、堵煤和自燃三种典型故障，获取大量不同工况下的故障样本数据。然后，采用新分类算法建立基于典型样本的故障预警模型，旨在提高故障识别的准确性，在故障初期提醒操作人员进行干预，从而避免磨煤机故障进一步扩大

Abstract:

Monitoring and diagnosis of coal mill systems are critical to the safe operation of power plants. Due to the scarcity of real fault data and the imbalance between fault and normal data, traditional data-driven fault diagnosis methods perform poorly in fault identification and sometimes even produce misjudgments. In order to efficiently identify typical faults of coal mills under different operating conditions, this paper designs a GRU-TCN fusion algorithm that combines a convolutional block attention module for building a mill fault identification model, and the new algorithm not only improves the classification accuracy, but also realizes the early warning of faults. First, by adjusting the key parameters of the coal mill fault generation model, three typical faults of coal breakage, coal plugging and spontaneous combustion are simulated to obtain a large amount of fault sample data under different working conditions. Then, the new classification algorithm is used to establish a fault warning model based on typical samples, aiming at improving the accuracy of fault identification and alerting operators to intervene at the early stage of the fault, so as to avoid further expansion of coal mill faults