摘要
为预测性维护磨煤机运行中磨辊的磨损程度,本文采用小波信号处理技术对磨煤机单耗信号进行趋势分解,从重构的低频单耗信号初步反映磨辊的磨损状况,进一步地,为解决重构后的低频磨单耗信号可能受到磨煤机工况影响而在一定范围内波动,且基于该方法所建立预报模型的固定阈值可能带来的误报和漏报问题,采用主成分分析法来筛选与磨辊磨损密切相关的磨煤机变量,并基于正常运行数据建立了重构单耗信号的自适应神经网络基准模型。仿真结果显示,采用自适应阈值后,相比采用固定阈值,模型准确度从91.3%上升到了97.4%,漏报率从9.4%下降到了1.5%,平均检测时间从413 s下降到230 s。说明采用自适应神经网络方法能够有效地表征磨辊的磨损状态,可以有效地对重构信号进行监测和预警,增加报警的准确性,减少漏报率和检测时间,从而为磨煤机磨辊的预测性维护提供了一种更可靠和高效的方法。
In order to predict the wear degree of the grinding roller in the operation of the coal mill,this paper uses the wavelet signal processing technology to decompose the unit consumption signal of the coal mill.The low-frequency unit consumption signal reconstructed by this method can initially reflect the wear condition of the grinding roller.Furthermore,in order to solve the problem that the reconstructed low-frequency unit consumption signal may fluctuate within a certain range due to the influence of the working conditions of the coal mill,and based on the fixed threshold of the prediction model established by this method,the principal component analysis method is used to screen the coal mill variables closely related to the wear of the grinding roller.Based on the normal operation data,an adaptive neural network benchmark model for reconstructing the unit consumption signal is established.The simulation results show that after using the adaptive threshold,compared with the fixed threshold,the accuracy of the model increases from 91.3%to 97.4%,the false negative rate decreases from 9.4%to 1.5%,and the average detection time decreases from 413 s to 230 s.It shows that the adaptive neural network method can effectively characterize the wear state of the grinding roller,effectively monitor and warn the reconstructed signal,increase the accuracy of the alarm,and reduce the false negative rate and detection time,thus providing a more reliable and efficient method for the predictive maintenance of the grinding roller of the coal mill.
作者
张军亮
白伟
张鹏伟
焦明明
艾宇
王亮
ZHANG Junliang;BAI Wei;ZHANG Pengwei;JIAO Mingming;AI Yu;WANG Liang(National Energy Jinjie Energy Co.Ltd.,Shenmu 719319,China)
出处
《电力科技与环保》
2024年第1期77-86,共10页
Electric Power Technology and Environmental Protection
基金
国家自然科学基金青年基金(52206009)。
关键词
磨辊磨损
多尺度分析
小波分解
趋势分量
roll wear
multi-scale analysis
wavelet transform
trend component