测控技术2024，Vol.43Issue(6)：75-81,7.DOI:10.19708/j.ckjs.2024.06.012
基于核密度估计的故障诊断信号非均匀量化方法
Inhomogeneous Quantization Method of Fault Diagnosis Signal Based on Kernel Density Estimation
摘要
Abstract
In some fault diagnosis scenarios of industrial Internet of things,due to the lack of telecommunication network coverage,the collected signals are wirelessly and reliably backhauled by long range radio(LoRa)tech-nology,but its lower transmission rate can limit the accuracy of fault diagnosis.An inhomogeneous quantization scheme for time-domain signals under the conditions of fixed sampling frequency and quantization resolution is proposed to address the technical limitations of the narrow bandwidth of LoRa.Firstly,the probability density function(PDF)is fitted to generate the sensed signal amplitude by establishing a nonparametric fitting model based on kernel density estimation(KDE),focusing on the suitable type of kernel function and bandwidth de-termination criteria in bandwidth-constrained scenarios.Next,using PDF as input and minimizing the quantiza-tion noise as the objective function,the optimal set of inhomogeneous quantization level values is output through nonlinear programming.It is characterized by using smaller quantization intervals for the time-domain ampli-tude with the highest frequency of occurrence to minimize the quantization noise.Finally,taking axial fan con-dition detection as an example,the experimental results show that the base loosening and bearing faults have a greater impact on the quantization level.With the increase of quantization resolution,the KDE quantization gradually converges to the uniform quantization,and the advantage over Gaussian quantization is gradually re-duced.Therefore,the proposed KDE quantization scheme is suitable for inhomogeneous quantization under nar-row bandwidth conditions,which can improve channel utilization and achieve a compromise between transmis-sion bandwidth and quantization noise.关键词
故障诊断/非均匀量化/LoRa/核密度估计/非线性规划Key words
fault diagnosis/inhomogeneous quantization/LoRa/kernel density estimation/nonlinear program-ming分类
信息技术与安全科学引用本文复制引用
张少波,崔英英,陈攀,朱许彬,张博..基于核密度估计的故障诊断信号非均匀量化方法[J].测控技术,2024,43(6):75-81,7.基金项目
陕西省地方标准制修订项目(SDBXM67-2020) （SDBXM67-2020）
国家重点研发计划项目(2019YFB1600100) （2019YFB1600100）
