泛研网

公众版 查看说明
全球科研项目数据库2026版

面向大气环境污染气体监测的MOS传感器阵列自确认软测量方法的研究

项目来源

国家自然科学基金(NSFC)

项目主持人

陈寅生

项目受资助机构

哈尔滨理工大学

项目编号

61803128

立项年度

2018

立项时间

未公开

研究期限

未知 / 未知

项目级别

国家级

受资助金额

24.00万元

学科

信息科学-自动化-自动化检测技术与装置

学科代码

F-F03-F0306

基金类别

青年科学基金项目

关键词

故障识别 ; 故障检测 ; 混合气体浓度检测 ; 隔离与重构 ; 自确认软测量模型 ; 污染气体监测 ; 故障识别 ; 故障检测 ; 混合气体浓度检测 ; 隔离与重构 ; 自确认软测量模型 ; 污染气体监测

参与者

赵文杰；王天；张宁；吴雪；龚建铧

参与机构

哈尔滨理工大学

项目标书摘要：本项目面向大气环境污染气体监测的新需求，将自确认传感器技术与软测量技术相结合，提出MOS传感器阵列自确认软测量方法，在保证测量值可靠性的基础上实现污染气体成分与浓度的有效测量。研究基于移动窗稀疏非负矩阵分解的故障检测、隔离与重构方法，准确实现传感器的故障检测、多故障隔离与故障数据重构；研究基于多尺度熵特征提取与多分类相关向量机的传感器故障识别方法，有效实现传感器故障信号的特征提取与故障模式识别，提高MOS气体传感器阵列的自诊断能力。研究基于快速多输出相关向量机回归的软测量方法，准确实现混合污染气体成分与浓度测量，提高测量准确性与实时性。利用以上算法构建自确认软测量模型并固化于监测系统样机中进行实验验证。本项目的研究成果对软测量技术在大气环境监测领域的发展与应用具有重大的理论与实践意义。

Application Abstract: This project facing the new requirements for air pollution monitoring,combined with self-validating sensor technology and soft sensor technology,proposes the self-validating soft sensor method of MOS gas sensor arrays.On the basis of guaranteeing the reliability of measurements,the effective measurement of the pollution gas composition is realized,and the credibility of detection results is improved.A fault detection,isolation and configuration method based on moving window sparse non-negative matrix decomposition is presented to accurately realize the fault detection,multiple faults isolation and faulty data reconfiguration.A sensor fault identification method based on multi-scale entropy feature extraction and multi-classification correlation vector machine is proposed to effectively realize the time-frequency feature extraction and fault pattern recognition of fault sensor signals for improving the self-diagnosis ability of MOS gas sensor array.In the aspect of contaminated gas mixture concentration soft sensor,a soft sensor modeling method based on fast multi-outputs correlation vector machine regression is researched to accurately realize composition and concentration measurement for improving measurements accuracy and real-time performance.The above algorithms are fused into a self-validating soft sensor model,and it is solidified in the prototype of the pollution gas monitoring system for experimental verification.The research results of the project will have great theoretical and practical significance for the application and development of soft sensor technology in the field of atmospheric environment monitoring.

项目受资助省

黑龙江省

项目结题报告(全文)

基于MOS气体传感器阵列的大气环境污染气体监测系统的实用化主要受限于长期监测过程中的异常状态监测与浓度测量的准确性问题。鉴于此，本项目开展MOS气体传感器阵列自确认软测量方法研究以提升系统检测结果的可靠性和准确性。为了解决长期监测过程中模型自适应性较差而导致的故障检测准确率较低的问题，提出了基于移动窗核主成分分析(MWKPCA)与极限学习机(ELM)相结合的故障检测、隔离与重构方法，该方法利用KPCA对非线性信号的处理能力，增强了对故障信号的敏感性，再利用移动窗技术对KPCA模型进行自适应更新，增强方法自适应能力，提高故障检测准确率；利用KPCA的SPE统计量贡献图实现故障传感器隔离；最后利用ELM预测模型对故障传感器信号进行重构，具有较高的故障信号重构精度。针对传感器故障特征提取结果可分性较差的问题，提出了一种基于改进多尺度幅值感知排列熵(AAPE)的故障特征提取方法，利用AAPE对信号幅度变化的敏感特性，提取不同尺度下的AAPE值描述故障特征，提取的特征具有更好的可分性；进一步地，针对实际应用中各种故障训练样本不均衡问题，提出一种基于深度卷积生成对抗网络(DCGAN)的MOS气体传感器故障诊断方法，利用GAN扩充小样本的故障信号数据集，以此来弥补数据不均衡的故障信号样本空间，显著提升样本不均衡条件下的故障识别准确率。为了解决混合气体定性识别和定量分析准确率较低的问题，分别利用模式识别方法和深度学习方法研究混合气体软测量模型，提出的基于SSA-SVM与GA-ADASYN-SVR的软测量方法和基于WOA-LSTM与LSSVM的软测量方法均能够取得较高的测量精度，且对湍流条件下混合气体具有较好的测量效果。设计并研制了大气环境污染气体监测系统样机，通过其获得的实验样本对以上提出的自确认软测量方法进行验证，并证明了方法有效性和可行性，对自确认软测量方法的进一步改进和优化提供硬件基础。项目的研究成果对低成本、便携式的大气污染气体监测系统的研制具有重要的理论与应用价值。

  • 排序方式：
  • 0
  • /

  • 1.Electronic nose and its application in the food industry: a review

    • 关键词：
    • E-nose; Food industry; Gas sensor; Food detection; Informationprocessing;CONVOLUTIONAL NEURAL-NETWORK; SENSORS; DISCRIMINATION; CLASSIFICATION;IDENTIFICATION; ADULTERATION; SPOILAGE; QUALITY; ORIGIN

    Food is closely related to human life. With the development of the times, the human demand for food has changed dramatically. People pay closer attention to the safety, health, composition, brand, origin, and processing method of food, which is precisely inseparable from food testing technology. Currently, there are many food inspection technologies, and the electronic nose (E-nose), as an efficient, fast, non-destructive, and promising technology, has been successfully applied in many aspects of the food industry and has shown promising results. This paper first introduces the basic principle and composition of the E-nose. Then it describes in detail the key elements, including gas sensor selection, sampling method design, data acquisition and information processing. Further summarizes the various typical applications of E-nose technology in the food industry in recent years, including six application directions: freshness assessment, process monitoring, flavor evaluation, authenticity evaluation, quality control, origin traceability and pesticide residue detection. Finally, the critical problems that need to be solved in the current application of E-nose technology in the food industry are discussed, and the potential future research directions in this field are foreseen.

    ...
  • 排序方式：
  • 0
  • /