基于大数据挖掘的立式击弦机磁力自复位水平测试研究  

Study on the self-reset level of vertical string machine based on big data mining

作　　者：陈海霞[1] CHEN Haixia(Shang Luo vocational and technical college,Shangluo,Shaanxi 726000,China)

机构地区：[1]商洛职业技术学院,陕西商洛726000

出　　处：《自动化与仪器仪表》2023年第7期244-248,252,共6页Automation & Instrumentation

基　　金：校级《高水平专业群建设视角下高职音乐课程教学改革研究——以商洛职业技术学院为例》(JYKT202316)。

摘　　要：针对传统立式击弦机微弱信号检测率低,导致磁力自复位水平不高的问题,提出一种基于WT-LSTM的微弱信号检测方法。首先,基于长短时神经网络构建基于LSTM的微弱信号检测模型,以检测立式击弦机的原始信号;然后引入小波分解,得到基于WT-LSTM的微弱信号检测模型,通过小波分解提取信号分量的近似系数,去除噪声分量;最后将数据传输至LSTM中进行新特征学习。结果表明,在-13 dB~0 dB的信噪比下,提出的WT-LSTM微弱信号检测方法的检测准确率均保持在85%及以上,其ROC曲线和AUC值明显高于传统的LSTM检测方法和RBF检测方法,虚警概率和漏警概率低于前两种方法。在信噪比为-13 dB时,本方法的检测准确率最高可达99.87%,比另外两种方法分别高出了8.8%和25.8%。由此说明,本方法可实现噪声抑制,提升微弱信号检测准确率,进一步增强立式击弦机磁力自复位水平。A weak signal detection method based on WT-LSTM is proposed.First,construct a weak signal detection model based on LSTM based on the neural network to detect the original signal of the vertical string machine,then introduce wavelet decomposition,obtain the weak signal detection model based on WT-LSTM,extract the approximate coefficient of the noise component;finally the data is transmitted to the LSTM for new feature learning.The results show that under the-13dB-0dB,the detection accuracy of the proposed WT-LSTM weak signal detection method remains 85%or above,the ROC curve and AUC values are significantly higher than the traditional LSTM detection method and RBF detection method,and the false alarm probability is lower than the previous two methods.At SNR-13dB,the highest detection accuracy of this method is 99.87%,which is 8.8%and 25.8%higher than the other two methods,respectively.This shows that this method can realize noise suppression,improve the detection accuracy of weak signal,and further enhance the magnetic self-reset level of the vertical string strike machine.

关 键 词：立式击弦机 微弱信号检测 LSTM 小波分解 磁力自复位 

分 类 号：TP392[自动化与计算机技术—计算机应用技术]