Font Size: a A A

Machinery Fault Diagnosis, Adaptive Alarm

Posted on:2007-10-02Degree:MasterType:Thesis
Country:ChinaCandidate:S Q ZhangFull Text:PDF
GTID:2192360185471947Subject:Mechanical and electrical engineering
Abstract/Summary:PDF Full Text Request
As one of the kernel techniques in the system of machinery condition monitor and faults diagnosis, the alert technique has an extreme important effect in the process of generalizing and promoting the development of the Predictive Maintenance System. Whereas the alert technique has still stayed in the step of static alert, namely, the setting up of warning threshold values has nothing to do with the actual conditions of machinery. The alert technique in machinery condition monitor was lucubrated in this paper. The main research contents as fellows:1. Some important methods about the current alert technique of machinery condition monitor were summarized at the beginning, and then textual main research contents and the significance of the research were put forward on this foundation.2. Whereas the principle of setting up of warning threshold values in the adaptive alarm technique based on probabilistic neural network, namely, the smooth factor is too smaller to be as the boundary of normal range, and the distributing of the historical data was supposed normal school, the principle of setting up of warning threshold values was improved. So the warning threshold values not only could be adjusting adaptive, but also met the actual conditions of machinery better. So we could offer reliable evidence for machinery condition evaluation.3. Narrow-band spectrum alarm technique is one of the most common uses alarm technique, and there are two methods to obtain the alarm parameters current, one by the orthogonal of trigonometric functions, another by FFT. The former supposes that the signals are periodic and the latter supposes that the signals are stationary, but in fact, the signals are non-periodic and non-stationary. The technique of wavelet pack was applied to the adaptive alarm technique in order to solve the problem mentioned forward, and experiment results show that this method is feasible and effective.4. Usually two sensors fixed orthogonally are used to detect vibration signals in the application of rotary machinery condition monitor, nevertheless, there are great different in both energy and structure of the signals detected. So different warning results are given out based on the signals of different direction, whereas there exist indeterminacy in warning results brought by the alarm technique based on single channel, the vector spectrum improved was applied to the adaptive alarm technique, and experiment results show that this method is feasible and effective.5. Though there are many parameters which can reflect faults of machinery, different parameters have different sensibilities for the same fault, at the same time, there are relativities among the parameters. As a result, the relativities won't be ensured if using single parameter and there will be information redundancies if using all of the parameters obtained form vibration signals, and it bring prodigious...
Keywords/Search Tags:Probabilistic Neural-Network, adaptive alarm, fault diagnosis, Machinery condition monitoring, wavelet pack, vector spectrum, Principal Component Analysis (PCA)
PDF Full Text Request
Related items