| Centrifugal pumps are broadly used as mechanical equipment. To use a centrifugal pump efficiently, we must specify a suitable environment for this equipment by consulting its performance under different circumstances. So, it is a very important issue to benchmark the pump's performance on different circumstances. On benchmarking the cooling pump's performance of internal-combustion engines, a national standard has been developed: operates the pump on a specified rotate speed, changes the runoff of the pump, when the system is stable, and measures the following parameters at the same time: runoff, pressure of entrance and exit, torque, rotate speed, temperature of the medium. Based on this standard, a data-collecting system for centrifugal pumps has been designed using Labwindows/CVI. This system consists of: pump test bed, sensors, signal processing circuit, data-collecting card, computer processing. Under the control of the computer, experiments, such as characteristic experiments, vapor corrosion experiments, airproof experiments, are carried out on the test bed. Sensors are responsible for obtaining data from the pump under testing. These signals are processed by the signal processing circuit and then sent to the data-collecting card. The computer gathers data from the data-collecting card, processes them, displays and prints the performance curves and generates benchmark reports.The research work of this dissertation is as following:1. A data-collecting system is implemented based on the idea of virtual instrument. To gather and transform the signals from the pumps, we use the PCI-MIO-16E-4 data-collecting card from NI in this system.2. In a benchmark system, large inaccuracies can be resulted from inaccurate measuring equipments, inadequate measuring method, interference and so on. These inaccuracies are not acceptable when high precision is required. To analyze, estimate and amend these inaccuracies, inspection and filtering processes are needed for the experiment results. Wavelet analysis is a burgeoning subject in recent years, which owns not only profoundly theoretical significance, also widely practical value. Meanwhile wavelet analysis is applied to deal with the signal data that is practical without considering the system model and the statistic character of signals. In this dissertation, wavelet multi-resolution analysis to filtering the outliers is discussed to avoid the deficiency of incompletely filtering the outliers using direct multi-resolution analysis. Firstly, Extending the signal boundary is done by this means, then, adopt the multi-resolution analysis to decompose and recompose the signal message. When the evaluation of the signal is obtaining, a serial of new information is constructed and according to the quality of the new information some signals are judged and eliminated. In this dissertation, adopting the multi-resolution analysis, the master-control computer checks out the data of torque, flux and power by sending out the instructions to slave system. After that, the signals are three-layers decomposed recomposed by selecting the symlets wavelet as the wavelet basis of the multi-resolution analysis. In addition, signal re-composition is defined as the method of compulsory threshold on decomposing the high frequency coefficient of wavelet, which means that all high coefficients is set to zero, the low frequency coefficient of the deepest layer of signal is decomposing by wavelet analysis in which confidence is 0.05, degree of freedom is 1 in distribution, and the threshold is 3.84.3. Theoretically, functions like exists on a certain section . In fact, we can only get the values on a series of discrete . For other x, is not available. That is to say, we only have a table , , for . From this table, we cannot make out the value for a specified x. To facilitate the research work, a function curve is often made from this table. When generating the function curve, if we use the values directly, all the measuring inaccuracy will be reserved. So, it is de...
|
PDF Full Text Request