白酒电子鼻信号去漂移方法及空载条件漂移现象初探

发布时间：2018-03-06 21:36

本文选题：电子鼻　切入点：漂移现象　出处：《河南科技大学》2017年硕士论文　论文类型：学位论文

【摘要】：电子鼻被认为是检测挥发性样品最具发展潜力的一种分析仪器。近年来,电子鼻广泛应用于食品检测领域。然而,传感器漂移现象使电子鼻在实际应用中不能长期稳健地检测食品等样品。因此,消除或降低漂移影响对电子鼻实现长期稳健检测十分重要。电子鼻在对6种中国白酒样品的鉴别过程中,漂移现象十分明显,由先验数据构建的识别模型不能有效和准确地预测后期样本数据。因此,对6种白酒的长期监控,如何减少漂移现象的影响,实现电鼻子长期稳健检测是必须研究的工作。本文从信号变换的角度出发,提出了一种基于小波包变换的递归漂移校正方法,用于长期稳健检测6种白酒样品。该方法主要包括构建小波包分解系数的校正函数和确定构建检测模型的样本测试时间窗口(SMTW)。借助于所构建的小波包分解系数的校正函数,6种白酒样品的电子鼻漂移信号被很好地校正。通过比较不同SMTW下所构建的检测模型之间的差异,确定了SMTW为6个月。通过递归校正方法处理SMTW中的样本数据之后,建立了递归Fisher判别分析(FDA)鉴别模型,并且通过递归FDA模型可以准确地鉴别SMTW之后的2个月的样本数据。因此,将确定的6个月的SMTW向前移动,FDA模型也被不断地递归修正,并且能实时对随后2个月内的样本数据进行正确地鉴别,最低鉴别正确率达96.5%。因此,采用上述方法实现了电子鼻对6种白酒样品的长期稳健鉴别。最后,考虑到漂移是气敏传感器的一种固有行为,论文从无样本(空载)条件下初步探讨了电子鼻漂移现象,给出了空载条件下去漂移的可行性方法,以期为空载条件去漂移奠定部分基础。
[Abstract]:The electronic nose is considered as a kind of analysis instrument to detect volatile samples with the most potential for development. In recent years, the electronic nose is widely used in the field of food detection. However, sensor drift phenomenon in the practical application of the electronic nose can not be long-term and steady detection of food samples. Therefore, eliminate or long-term robust detection is very important for the realization of electronic nose to reduce the drift effect. The electronic nose in the identification process of 6 kinds of Chinese liquor samples, the drift phenomenon is very obvious, can not identify the model constructed by the prior data forecast late sample data effectively and accurately. Therefore, the long-term monitoring of the 6 kinds of liquor, how to reduce the effect of drift phenomenon, to achieve long-term stable detection is the electric nose we must study work. From the point of view of signal transformation, proposes a recursive drift correction method based on wavelet packet transform, for the long-term stable detection of 6 kinds of liquor Sample. This method mainly includes the construction of wavelet packet decomposition coefficient correction function and determine the sample test time window construction detection model (SMTW). With the help of the correction function decomposition coefficients in wavelet packets constructed by the electronic nose signal drift 6 kinds of liquor samples is well corrected. The difference between the detection model by comparing different SMTW the construction of the SMTW is determined for 6 months. After treatment by recursive correction method of sample data in SMTW, establishes the recursive Fisher discriminant analysis (FDA) identification model, and through the recursive FDA model can sample data of 2 months after the accurate identification of SMTW. Therefore, the 6 months the SMTW moves forward, FDA model has been modified and can real-time recursive, then the sample data of 2 months were correctly identified, the minimum identification accuracy reaches 96.5%. therefore, using the above method to achieve The long-term stable identification of 6 kinds of liquor samples by electronic nose. Finally, taking into account the shift is an inherent behavior of gas sensors, the paper from the sample (no load) under the conditions discussed electronic nose drift phenomenon, given the feasibility method of no-load condition. The drift, in order to lay the foundation for the part of no-load condition to drift.

【学位授予单位】：河南科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TS262.3;O657

【参考文献】