Fault Diagnosis Based On Multi-symptom And The Fusion Of Multiple Intelligent Methods

With the continuous development of science and technology, rotating machineries and equipment becomes more and more complicated integrated and intelligent. Thus, the machinery structure grows more and more complicated, and the connection between the different parts is also becoming more and more closely. All of these lead to the greater risk of failure and more serious failure consequences. So, the exact condition monitoring and fault diagnosis play an important role in ensuring the reliable running of machinery, and the researches on REB fault diagnosis are very necessary to be developed and continuously improved.The most common and main faults of rotating machinery are shafting vibration faults, and the shafting vibration signals are the most direct embodiments of the machinery running state. Thus, the analysis and identification of shafting vibration signals is the most critical fault diagnosis method. In addition, the shaft orbit synthesized by shaft vibration signals reflects various fault information of rotating machinery, and its shape directly reflects the running state of the shaft. So, the shape identification of shaft orbits is another important fault diagnosis method for rotating machinery. For the first method, it gets the critical information which could reveal the inherent relationship between signals and the running status from the shaft signals, and based on this, it could realize the fault diagnosis of rotating machinery. For shaft orbit identification, it completes the fault diagnosis based on the mapping relationship between the shaft orbit shape and the running status.(1) For the difficult in the description and identification of rotating machinery shaft signals, this paper presents a new time-frequency analysis method based on further study on the theoretical system of empirical mode decomposition and the development feature of rotating machinery shafting fault. This method effectively restrains the endpoint effect in empirical mode decomposition, and could prepare an effective data base for the feature extraction of shaft signals. So, the time-frequency analysis method presented in this paper could improve the effectiveness of the shaft faults description and identification.(2) There is a large number of redundant information in the signal features extracted through empirical mode decomposition, and it serious influences the precision and efficiency of the rotating machinery fault diagnosis. In order to overcome this problem, a hierarchical feature selection method based on classification tree (FSBCS) is put forward in this paper. In each branch node, HFSMCT selects the optimal feature through the filter evaluation criterion and heuristic search strategy. HFSMCT is easy to design and could complete the calculation with great rapidity. Moreover, it could remove the redundant information to the hilt, but retains the most obvious difference between samples. Therefore, HFSMCT improves the precision and speed of the algorithm at the same time, and it is a very useful feature selection method for rotating machinery.(3) Feature selection methods ignore the different description ability of each feature about different samples, as well as its different influence on the correct recognition of different samples. And this limits the improvement of diagnostic accuracy. In order to overcome this problem, dependent feature vector (DFV) is proposed in this paper to denote the fault symptom attributes. DFV is a self-adapting sample representation method. And it uses a unique feature selection technique which selects the most effective features for each sample according to its own characteristics. It also has a unique structure:The LF of all samples are the same, the effective DF is decided by the specific LF value of each sample, and all feature items in the invalid DF have the same appropriate value (DV). Therefore, DFV realizes the most concise and effective sample representation. DFV could reduce the difference among the samples of the same class. Then, through selecting appropriate value for the invalid DF feature items, DFV could significantly enlarge the difference among the different classes. Therefore, DFV could effectively enhance the compactness of the samples in the same class and the separability among different classes, and it is very excellent in distinguish capacity. Moreover, DFV removes the redundant information to the hilt for each sample, greatly reduces the complexity and the elapsed time of feature extraction and fault identification, and improves the efficiency of fault diagnosis.(4) DFV has innovative advantage in sample description, but it could not effectively describe the mixing modes. Through an intensive study on the generation mechanism and corresponding feature extraction method of DFV, this paper found that the key reason caused this defect is the boundary processing method based on two-valued logic. Thus, fuzzy DFV (FDFV) is proposed to overcome this problem. In FDFV, fuzzy logic is used to optimize the boundary processing method in DFV generation mechanism and the corresponding feature extraction. On the basis of the novel and unique structure of DFV, FDFV greatly enhances the universality and robustness, and provides a simple and efficient fault description method for the fault diagnosis with mixing modes.(5) Shaft orbit recognition is an important approach for the vibrational state judgment of rotating machinery. Extracting the features of shaft orbit images is not an easy task, and the traditional feature extraction methods are not perfect in comprehensiveness, accuracy and stability. In order to overcome these problems, comprehensive geometric characteristics (CGC) is presented and discussed in this paper. CGC imitates human eyes to extract the most important information of image structure, boundary and region, and realizes the shape characterization comprehensively and accurately through the full integration of effective information. Based on CGC, a shaft orbit identification method based on imitating human eyes is proposed for the fault diagnosis of rotating machinery. In this new method, CGC is the "human eye", and it could descript the shaft orbit shape like human eyes; the intelligent classification methods are "human brain", and they realize the automatically recognition of shaft orbit. The shaft orbit identification method based on imitating human eyes is a simple, efficient and accurate fault diagnosis method for rotating machinery.