Research On Checking And Fault Diagnosis Of Check Valve In Reciprocating High Pressure Diaphragm Pump

Large reciprocating high-pressure diaphragm pump, just like human’s heart, is the core power equipment of a long distance, high-lift, high-concentration slurry transportation pipeline, the working condition for which is important to the production efficiency of the enterprise. As one of the key components of the pump, check valve is required to have good performance as quick-opening, quick-closing, sealing status and pressure resistance. Therefore, it is more prone to failure than the other parts of the pump. In addition, the operation status of the check is associated closely with several factors including the particle size distribution of minerals, rheological properties of the slurry, delivery pressure, inherent material properties and installation of the pump, for which are responsible for the unexpectedness, concurrency, multi-sources, non-stationary and nonlinearity characteristics of check valve failures. Moreover, the difficulty of the check valve condition monitoring and fault diagnosis has been greatly increased. Therefore, from check-valve vibration signal analysis, selection of effective feature extraction and fault diagnosis method is the core research contents of running state monitoring and fault diagnosis for check valve, has important theoretical research value and economic significance. This thesis focuses on condition monitoring and fault diagnosis of the check valve and the main contents are organized as follows:1. Check valve fault detection method which is based on Local Mean Decomposition(LMD) and Envelope demodulation. Check valve fault vibration signal usually appears as complex AM-FM signal, which makes it possible to extract the fault characteristic frequency of the check valve by the envelope demodulation method. However, the check valve is affected by environmental noise, coupling conditions and other incentives. As a result, its vibration signal is a complex nonlinear signal. Thus, applying the envelope demodulation method directly to the signal would not result in a desirable outcome. Therefore, the fault detection method based on LMD and envelope demodulation is proposed. Firstly, the signal is decomposed into a series of pure AM-FM signal named Production Function(PF) by using LMD, then PF components are processed via envelope demodulation to complete the check valve fault detection.2. Check valve fault diagnosis method which is based on Multi-domain mixing features and ELM. Aiming to solve problems such as single domain characteristics cannot fully describe the running state of the check valve, issues of the Support vector machines and BP neural network model having more optimizable parameters and running slow in speed. Combining the advantages of multi-domain mixing characteristics and ELM, a fault diagnosis method for check valve based on multi-domain mixing features and ELM is proposed. The mixing feature set is constructed by extracting features from the vibration signal of check valve in time domain, frequency domain, wavelet domain and TK (Teager Kaiser, TK) domain. Correspondingly, the KPCA is then introduced to eliminate the feature redundancy. Finally, and ELM fault diagnosis model is established based on the mixing features to complete check valve fault diagnosis.3. Check valve fault diagnosis method that is based on wavelet packet energy entropy and F-KELM. Throughout the discussion of several issues of classification performance influenced by the nature of nonlinear vibration signal, imbalanced sample distribution and the number of neurons in hidden layer of ELM, a fault diagnosis method based on the wavelet packet energy entropy and F-KELM is put forward by combining the wavelet packet energy entropy and kernel function with fuzzy membership function. Comparing the rolling bearing experiment results and the check valve experiment results, it suggests that the proposed method have solved the above problems better and improved the model classification performance and its generalization ability.4. Check valve fault diagnosis method which is based on MKL-CS-ELM. Attempting to resolve the following problems:a single kernel classifier is not capable of fully interpreting the signal, the assumption of the cost of diagnosis can be debatable and imbalanced sample distribution has great influence on the classification performance, the multi-kernel function and cost sensitive learning mechanism are introduced to build the fault diagnosis model. Comparing the experiment results of rolling bearings with the results of check valves fault diagnosis in the binary classification and multi-classification experiments, the proposed method not only can provide a similar or even better performance with the MKL-CS-SVM, but also can retain the advantages of ELM. Meanwhile, the indicators of robustness is introduced to evaluate different cost-sensitive processing methods, and provide a criterion for choosing cost-sensitive processing method accordingly.5. The development and evaluation of the check valve condition monitoring and fault diagnosis system. C# and Matlab are used to programme and complete the system. The diaphragm pumps of Yunnan Dahongshan iron ore concentrate slurries are used as the research subjects. The vibration signals of the check valve are collected in the entire life cycle to carry out the system test.In this thesis, the large reciprocating high-pressure diaphragm pump check valve for the slurry transportation pipeline is the main object of study. A variety of methods of status monitoring and fault diagnosis have been explored. The research results should enrich the study content of status monitoring and fault diagnosis and promote the application and development of fault diagnosis of reciprocating machinery.