Structural health monitoring is widely utilized in outdoor environments,especially under harsh conditions,which can introduce noise into the monitoring system.Therefore,designing an effective denoising strategy to enh...Structural health monitoring is widely utilized in outdoor environments,especially under harsh conditions,which can introduce noise into the monitoring system.Therefore,designing an effective denoising strategy to enhance the performance of guided wave damage detection in noisy environments is crucial.This paper introduces a local temporal principal component analysis(PCA)reconstruction approach for denoising guided waves prior to implementing unsupervised damage detection,achieved through novel autoencoder-based reconstruction.Experimental results demonstrate that the proposed denoising method significantly enhances damage detection performance when guided waves are contaminated by noise,with SNR values ranging from 10 to-5 dB.Following the implementation of the proposed denoising approach,the AUC score can elevate from 0.65 to 0.96 when dealing with guided waves corrputed by noise at a level of-5 dB.Additionally,the paper provides guidance on selecting the appropriate number of components used in the denoising PCA reconstruction,aiding in the optimization of the damage detection in noisy conditions.展开更多
In order to raise the efficiency,automatization and intelligentization of condition monitoring and fault diagnosis for complex equipment systems,rough set theory is used to the field. A feature reduction algorithm bas...In order to raise the efficiency,automatization and intelligentization of condition monitoring and fault diagnosis for complex equipment systems,rough set theory is used to the field. A feature reduction algorithm based on rough set theory is adopted to extract condition information in monitoring and diagnosis for an engine,so that the technology condition monitoring parameters are optimized. The decision tables for each fault source are built and the diagnosis rules rooting in rough set reduction is applied to carry through intelligent fault diagnosis. The cases studied show that rough set method in condition monitoring and fault diagnosis can lighten the work burden in feature selection and afford advantages for autonomic learning and decision during diagnosis.展开更多
It is an urgent project to realize online and overall condition monitoring and timely fault diagnosis for large-scale mobile and complex equipment. Moreover, most of the existing large-scale complex equipment has quit...It is an urgent project to realize online and overall condition monitoring and timely fault diagnosis for large-scale mobile and complex equipment. Moreover, most of the existing large-scale complex equipment has quite insufficient accessibility of examination, although it still has quite a long service life. The decentralized and overall condition monitoring, as a new concept, is proposed from the point of view of the whole system. A set of complex equipment is divided into several parts in terms of concrete equipment. Every part is processed via one detecting unit, and the main detecting unit is connected with other units. The management work and communications with the remote monitoring center have been taken on by it. Consequently, the difficulty of realizing a condition monitoring system and the complexity of processing information is reduced greatly. Furthermore, excellent maintainability of the condition monitoring system is obtained because of the modularization design. Through an application example, the design and realization of the decentralized and overall condition monitoring system is introduced specifically. Some advanced technologies, such as, micro control unit （MCU）, advanced RISC machines （ARM）, and control area network （CAN）, have been adopted in the system. The system＇s applicability for the existing large-scale mobile and complex equipment is tested.展开更多
Based on the physical meaning of sensitivity,a new finite element(FE) model updating method was proposed. In this method,a three-dimensional FE model of the Nanjing Yangtze River Bridge(NYRB) with ANSYS program was es...Based on the physical meaning of sensitivity,a new finite element(FE) model updating method was proposed. In this method,a three-dimensional FE model of the Nanjing Yangtze River Bridge(NYRB) with ANSYS program was established and updated by modifying some design parameters. To further validate the updated FE model,the analytical stress-time histories responses of main members induced by a moving train were compared with the measured ones. The results show that the relative error of maximum stress is 2.49% and the minimum relative coefficient of analytical stress-time histories responses is 0.793. The updated model has a good agreement between the calculated data and the tested data,and provides a current baseline FE model for long-term health monitoring and condition assessment of the NYRB. At the same time,the model is validated by stress-time histories responses to be feasible and practical for railway steel bridge model updating.展开更多
The Mahalanobis distance features proposed by P.C.Mahalanobis, an Indian statistician, can be used in an automatic on-line cutting tool condition monitoring process based on digital image processing. In this paper, a ...The Mahalanobis distance features proposed by P.C.Mahalanobis, an Indian statistician, can be used in an automatic on-line cutting tool condition monitoring process based on digital image processing. In this paper, a new method of obtaining Mahalanobis distance features from a tool image is proposed. The key of calculating Mahalanobis distance is appropriately dividing the object into several component sets. Firstly, a technique is proposed that can automatically divide the component groups for calculating Mahalanobis distance based on the gray level of wearing or breakage regions in a tool image. The wearing region can be divided into high gray level component group and the tool-blade into low one. Then, the relation between Mahalanobis distance features of component groups and tool conditions is investigated. The results indicate that the high brightness region on the flank surface of the turning tool will change with its abrasion change and if the tool is heavily abraded, the area of high brightness will increase apparently. The Mahalanobis distance features of high gray level component group are related with wearing state of tool and low gray level component group correlated with breakage of tool. The experimental results show that the abrasion of the tool’s flank surface affected the Mahalanobis distances of high brightness component of the tool and the pixels of high brightness component set. Compared with the changes of them, we found that the Mahalanobis distance of high brightness component of the tool was more sensitive to the abrasion of cutting tool than the area of high brightness component set of the tool. Here we found that the relative changing rate of the area of high brightness component set was not quite obvious and it was ranging from 2% to 15%, while the relative changing rate of the Mahalanobis distance in table 1 ranges from 13.9% to 47%. It is 3 times higher than the changing rate of the area.展开更多
One of the most important features of the modern ma ch ining system in an "unmanned" factory is to change tools that have been subjec ted to wear and damage. An integrated tool condition monitoring system co...One of the most important features of the modern ma ch ining system in an "unmanned" factory is to change tools that have been subjec ted to wear and damage. An integrated tool condition monitoring system composed of multi-sensors, signal processing devices and intelligent decision making pla ns is a necessary requirement for automatic manufacturing processes. An intellig ent tool wear monitoring system will be introduced in this paper. The system is equipped with power consumption, vibration, AE and cutting force sensors, signal transformation and collection apparatus and a microcomputer. Tool condition monitoring is a pattern recognition process in which the characte ristics of the tool to be monitored are compared with those of the standard mode ls. The tool wear classification process is composed of the following parts: fea ture extraction; determination of the fuzzy membership functions of the features ; calculation of the fuzzy similarity; learning and tool wear classification. Fe atures extracted from the time domain and frequency domain for the future patter n recognition are as follows. Power consumption signal: mean value; AE-RMS sign al: mean value, skew and kutorsis; Cutting force, AE and vibration signal: mean value, standard deviation and the mean power in 10 frequency ranges. These signa l features can reflect the tool wear states comprehensively. The fuzzy approachi ng degree and the fuzzy distance between corresponding features of different obj ects are combined to describe the closeness of two fuzzy sets more accurately. A unique fuzzy driven neural network based pattern recognition algorithm has bee n developed from this research. The combination of Artificial Neural Networks (A NNs) and fuzzy logic system integrates the strong learning and classification ab ility of the former and the superb flexibility of the latter to express the dist ribution characteristics of signal features with vague boundaries. This methodol ogy indirectly solves the automatic weight assignment problem of the conventiona l fuzzy pattern recognition system and let it have greater representative power, higher training speed and be more robust. The introduction of the two-dimensio nal weighted approaching degree can make the pattern recognition process more re liable. The fuzzy driven neural network can effectively fuse multi-sensor i nformation and successfully recognize the tool wear states. Armed with the advan ced pattern recognition methodology, the established intelligent tool condition monitoring system has the advantages of being suitable for different machini ng conditions, robust to noise and tolerant to faults. Cooperated with the contr ol system of the machine tool, the optimized machining processed can be achieved .展开更多
Vibration monitoring and vibration severity evaluation of armored vehicle transmission are realized by additional sensors. An algorithm of vibration severity in frequency domain is presented. The algorithm has powerfu...Vibration monitoring and vibration severity evaluation of armored vehicle transmission are realized by additional sensors. An algorithm of vibration severity in frequency domain is presented. The algorithm has powerful applicability for signal type and flexible selectivity for frequency range,and avoids the processing of signal conversion used calculus and filtering compared to the algorithm of vibration severity in time domain. An applied example is given in company with attentive proceedings and measures for improving evaluation effect.展开更多
随着高压电缆应用规模的日益扩大及逐步进入老化周期,对其智能化运维成为保障城市电网可靠性的紧迫需求。传统的高压电缆运维模式存在隐蔽缺陷识别难、多故障耦合溯源难及有效案例数据稀少等方面的瓶颈。人工智能(artificial intelligen...随着高压电缆应用规模的日益扩大及逐步进入老化周期,对其智能化运维成为保障城市电网可靠性的紧迫需求。传统的高压电缆运维模式存在隐蔽缺陷识别难、多故障耦合溯源难及有效案例数据稀少等方面的瓶颈。人工智能(artificial intelligence,AI)技术,凭借其在处理复杂、高维、非线性数据上的优势,为电缆运维从“经验驱动”向“数据智能驱动”转型提供了契机。为此系统性阐述了人工智能技术在高压电缆系统状态监测中的典型应用现状及前景。首先阐述了高压电缆系统状态监测中主要监测物理量的特点及AI应用前景。其次,梳理了机器学习与深度学习等主流AI技术的特点,并分析其在电缆状态监测中的应用进展。随后,进一步深入剖析了AI在局部放电监测、电缆逸出气体分析(evolved gas analysis,EGA)及金属护套环流监测等三大核心领域的典型应用进展,并针对EGA领域数据稀缺的挑战,探讨了基于数据增强、迁移学习和小样本学习的模型优化策略。最后,探讨了AI在多模态数据融合及诊断评估的技术路径。通过这些研究表明,AI技术通过知识自进化机制与多模态数据融合,可能推动高压电缆系统从传统运维模式向智能化的预测性维护模式转变。展开更多
为了提高利用监控和数据采集(supervisory control and data acquisition,SCADA)多变量长时间序列预测齿轮箱油温的精度,解决不同风电机组因处不同运行环境导致的数据分布不一致的问题,提出了一种基于多分支时间序列预测与迁移学习相结...为了提高利用监控和数据采集(supervisory control and data acquisition,SCADA)多变量长时间序列预测齿轮箱油温的精度,解决不同风电机组因处不同运行环境导致的数据分布不一致的问题,提出了一种基于多分支时间序列预测与迁移学习相结合的齿轮箱状态监测方法。首先,利用极致梯度提升(extreme gradient boosting,XGBoost)算法筛选输入参数组成原始序列,对其进行分解得到季节与趋势序列。其次,提出季节、趋势序列特征提取模块获取季节及趋势特征的序列,将其与经过Informer模型处理后的特征序列进行融合后输入进多层感知机映射成最终的预测值,以构建提出的多分支时间序列预测网络(multi-branch time series prediction network,MBFN)。最后,利用迁移学习并结合一分类向量支持机(one-class support vector machine,OCSVM)模型及滑动窗口构建齿轮箱的健康指数,完成齿轮箱状态监测。实验结果表明,所提出模型的MBFN显著提高了油温预测精度,优于常规时间序列预测模型,所使用的迁移策略能以较少数据适应不同数据的分布,进而实现对齿轮箱的状态监测,并且所提出的模型可以提前18.9 d发出齿轮箱故障预警。展开更多
基金National Science Foundation of Zhejiang under Contract(LY23E010001)。
文摘Structural health monitoring is widely utilized in outdoor environments,especially under harsh conditions,which can introduce noise into the monitoring system.Therefore,designing an effective denoising strategy to enhance the performance of guided wave damage detection in noisy environments is crucial.This paper introduces a local temporal principal component analysis(PCA)reconstruction approach for denoising guided waves prior to implementing unsupervised damage detection,achieved through novel autoencoder-based reconstruction.Experimental results demonstrate that the proposed denoising method significantly enhances damage detection performance when guided waves are contaminated by noise,with SNR values ranging from 10 to-5 dB.Following the implementation of the proposed denoising approach,the AUC score can elevate from 0.65 to 0.96 when dealing with guided waves corrputed by noise at a level of-5 dB.Additionally,the paper provides guidance on selecting the appropriate number of components used in the denoising PCA reconstruction,aiding in the optimization of the damage detection in noisy conditions.
文摘In order to raise the efficiency,automatization and intelligentization of condition monitoring and fault diagnosis for complex equipment systems,rough set theory is used to the field. A feature reduction algorithm based on rough set theory is adopted to extract condition information in monitoring and diagnosis for an engine,so that the technology condition monitoring parameters are optimized. The decision tables for each fault source are built and the diagnosis rules rooting in rough set reduction is applied to carry through intelligent fault diagnosis. The cases studied show that rough set method in condition monitoring and fault diagnosis can lighten the work burden in feature selection and afford advantages for autonomic learning and decision during diagnosis.
基金This project was supported by the Hebei Provincial Nature Science Foundation (E20070011048).
文摘It is an urgent project to realize online and overall condition monitoring and timely fault diagnosis for large-scale mobile and complex equipment. Moreover, most of the existing large-scale complex equipment has quite insufficient accessibility of examination, although it still has quite a long service life. The decentralized and overall condition monitoring, as a new concept, is proposed from the point of view of the whole system. A set of complex equipment is divided into several parts in terms of concrete equipment. Every part is processed via one detecting unit, and the main detecting unit is connected with other units. The management work and communications with the remote monitoring center have been taken on by it. Consequently, the difficulty of realizing a condition monitoring system and the complexity of processing information is reduced greatly. Furthermore, excellent maintainability of the condition monitoring system is obtained because of the modularization design. Through an application example, the design and realization of the decentralized and overall condition monitoring system is introduced specifically. Some advanced technologies, such as, micro control unit （MCU）, advanced RISC machines （ARM）, and control area network （CAN）, have been adopted in the system. The system＇s applicability for the existing large-scale mobile and complex equipment is tested.
基金Project(2001G025) supported by the Foundation of the Science and Technology Section of Ministry of Railway of ChinaProject(2006FJ4233) supported by Hunan Postdoctoral Scientific Program of ChinaProject(2006) supported by the Postdoctoral Foundation of Central South University,China
文摘Based on the physical meaning of sensitivity,a new finite element(FE) model updating method was proposed. In this method,a three-dimensional FE model of the Nanjing Yangtze River Bridge(NYRB) with ANSYS program was established and updated by modifying some design parameters. To further validate the updated FE model,the analytical stress-time histories responses of main members induced by a moving train were compared with the measured ones. The results show that the relative error of maximum stress is 2.49% and the minimum relative coefficient of analytical stress-time histories responses is 0.793. The updated model has a good agreement between the calculated data and the tested data,and provides a current baseline FE model for long-term health monitoring and condition assessment of the NYRB. At the same time,the model is validated by stress-time histories responses to be feasible and practical for railway steel bridge model updating.
文摘The Mahalanobis distance features proposed by P.C.Mahalanobis, an Indian statistician, can be used in an automatic on-line cutting tool condition monitoring process based on digital image processing. In this paper, a new method of obtaining Mahalanobis distance features from a tool image is proposed. The key of calculating Mahalanobis distance is appropriately dividing the object into several component sets. Firstly, a technique is proposed that can automatically divide the component groups for calculating Mahalanobis distance based on the gray level of wearing or breakage regions in a tool image. The wearing region can be divided into high gray level component group and the tool-blade into low one. Then, the relation between Mahalanobis distance features of component groups and tool conditions is investigated. The results indicate that the high brightness region on the flank surface of the turning tool will change with its abrasion change and if the tool is heavily abraded, the area of high brightness will increase apparently. The Mahalanobis distance features of high gray level component group are related with wearing state of tool and low gray level component group correlated with breakage of tool. The experimental results show that the abrasion of the tool’s flank surface affected the Mahalanobis distances of high brightness component of the tool and the pixels of high brightness component set. Compared with the changes of them, we found that the Mahalanobis distance of high brightness component of the tool was more sensitive to the abrasion of cutting tool than the area of high brightness component set of the tool. Here we found that the relative changing rate of the area of high brightness component set was not quite obvious and it was ranging from 2% to 15%, while the relative changing rate of the Mahalanobis distance in table 1 ranges from 13.9% to 47%. It is 3 times higher than the changing rate of the area.
文摘One of the most important features of the modern ma ch ining system in an "unmanned" factory is to change tools that have been subjec ted to wear and damage. An integrated tool condition monitoring system composed of multi-sensors, signal processing devices and intelligent decision making pla ns is a necessary requirement for automatic manufacturing processes. An intellig ent tool wear monitoring system will be introduced in this paper. The system is equipped with power consumption, vibration, AE and cutting force sensors, signal transformation and collection apparatus and a microcomputer. Tool condition monitoring is a pattern recognition process in which the characte ristics of the tool to be monitored are compared with those of the standard mode ls. The tool wear classification process is composed of the following parts: fea ture extraction; determination of the fuzzy membership functions of the features ; calculation of the fuzzy similarity; learning and tool wear classification. Fe atures extracted from the time domain and frequency domain for the future patter n recognition are as follows. Power consumption signal: mean value; AE-RMS sign al: mean value, skew and kutorsis; Cutting force, AE and vibration signal: mean value, standard deviation and the mean power in 10 frequency ranges. These signa l features can reflect the tool wear states comprehensively. The fuzzy approachi ng degree and the fuzzy distance between corresponding features of different obj ects are combined to describe the closeness of two fuzzy sets more accurately. A unique fuzzy driven neural network based pattern recognition algorithm has bee n developed from this research. The combination of Artificial Neural Networks (A NNs) and fuzzy logic system integrates the strong learning and classification ab ility of the former and the superb flexibility of the latter to express the dist ribution characteristics of signal features with vague boundaries. This methodol ogy indirectly solves the automatic weight assignment problem of the conventiona l fuzzy pattern recognition system and let it have greater representative power, higher training speed and be more robust. The introduction of the two-dimensio nal weighted approaching degree can make the pattern recognition process more re liable. The fuzzy driven neural network can effectively fuse multi-sensor i nformation and successfully recognize the tool wear states. Armed with the advan ced pattern recognition methodology, the established intelligent tool condition monitoring system has the advantages of being suitable for different machini ng conditions, robust to noise and tolerant to faults. Cooperated with the contr ol system of the machine tool, the optimized machining processed can be achieved .
基金Sponsored by National Defense Science and Technology Key Lab Foundation of China (51457120104JB3505)
文摘Vibration monitoring and vibration severity evaluation of armored vehicle transmission are realized by additional sensors. An algorithm of vibration severity in frequency domain is presented. The algorithm has powerful applicability for signal type and flexible selectivity for frequency range,and avoids the processing of signal conversion used calculus and filtering compared to the algorithm of vibration severity in time domain. An applied example is given in company with attentive proceedings and measures for improving evaluation effect.
文摘随着高压电缆应用规模的日益扩大及逐步进入老化周期,对其智能化运维成为保障城市电网可靠性的紧迫需求。传统的高压电缆运维模式存在隐蔽缺陷识别难、多故障耦合溯源难及有效案例数据稀少等方面的瓶颈。人工智能(artificial intelligence,AI)技术,凭借其在处理复杂、高维、非线性数据上的优势,为电缆运维从“经验驱动”向“数据智能驱动”转型提供了契机。为此系统性阐述了人工智能技术在高压电缆系统状态监测中的典型应用现状及前景。首先阐述了高压电缆系统状态监测中主要监测物理量的特点及AI应用前景。其次,梳理了机器学习与深度学习等主流AI技术的特点,并分析其在电缆状态监测中的应用进展。随后,进一步深入剖析了AI在局部放电监测、电缆逸出气体分析(evolved gas analysis,EGA)及金属护套环流监测等三大核心领域的典型应用进展,并针对EGA领域数据稀缺的挑战,探讨了基于数据增强、迁移学习和小样本学习的模型优化策略。最后,探讨了AI在多模态数据融合及诊断评估的技术路径。通过这些研究表明,AI技术通过知识自进化机制与多模态数据融合,可能推动高压电缆系统从传统运维模式向智能化的预测性维护模式转变。
文摘为了提高利用监控和数据采集(supervisory control and data acquisition,SCADA)多变量长时间序列预测齿轮箱油温的精度,解决不同风电机组因处不同运行环境导致的数据分布不一致的问题,提出了一种基于多分支时间序列预测与迁移学习相结合的齿轮箱状态监测方法。首先,利用极致梯度提升(extreme gradient boosting,XGBoost)算法筛选输入参数组成原始序列,对其进行分解得到季节与趋势序列。其次,提出季节、趋势序列特征提取模块获取季节及趋势特征的序列,将其与经过Informer模型处理后的特征序列进行融合后输入进多层感知机映射成最终的预测值,以构建提出的多分支时间序列预测网络(multi-branch time series prediction network,MBFN)。最后,利用迁移学习并结合一分类向量支持机(one-class support vector machine,OCSVM)模型及滑动窗口构建齿轮箱的健康指数,完成齿轮箱状态监测。实验结果表明,所提出模型的MBFN显著提高了油温预测精度,优于常规时间序列预测模型,所使用的迁移策略能以较少数据适应不同数据的分布,进而实现对齿轮箱的状态监测,并且所提出的模型可以提前18.9 d发出齿轮箱故障预警。
