Research On Fault Diagnosis Technology For Mass Sensor

The rapid development of economy brings negative issue is that the growing pollution of atmospheric particulate matter concentration, which endangers people’s health and safety. The effective governance of the concentration of atmospheric particulate matter pollution becomes a key factor to protect the health of people and the harmonious development of society. Micro-oscillation balance monitor can effectively monitor the pollution of concentration atmospheric particulate matter, and its most important interior unit is the mass sensor. Mass sensor protects the accuracy of output results of the concentration of atmospheric particulate matter pollution, which provides a more scientific basis for environmental management. The paper makes deep research and study on fault diagnosis technology of mass sensor.1).the meaning of mass sensor fault diagnosis and development status and trends of domestic and foreign are introduced on the basis of reviewing pretty documents.2).The four operating status of mass sensor are simulated on the micro-oscillation balance bench, including normal, unbalance, looseness and misalignment. So the vibration signal of mass sensors in four different states can be acquired, which supplies available experimental data for further research on wavelet packet transform.3).The paper makes signal feature extraction based on wavelet packet decomposition, and makes wavelet packet feature extraction with the data of mass sensor in four different states collected from the micro oscillating balance bench. In summary, it is more simple and practical to use the method of "frequency aliquot".4).The fault diagnosis of combining the wavelet analysis with neural network is proposed. Firstly, the wavelet packet and BP neural network theory are combined, and make feature vector extraction of vibration signal of mass sensor under four states using wavelet packet, then the extracted feature vector is took as the input sample of BP neural network which will be trained and tested, so the BP neural network has the fault identification function of mass sensor. This method gets good engineering application value.Fault recognition method based on wavelet packet and BP neural network combined with four kinds of state on mass the sensor was more than 90%, this method has better fault recognition ability on the mass sensor...