Extraction And Analysis Of Vibration Signal Induced By Cavitation In Centrifugal Pumps

Cavitation is a common phenomenon that occurs in centrifugal pumps,causing abnormal operating states.This phenomenon is typically caused by low pressure in the impeller eye,which can affect the head and vibration characteristics of the pump.Thus,phase transition phenomenon not only can be noticed with the pressure transducer,but also can be detected via using vibration transducer.Naturally,the corresponding conducting states of centrifugal pump would be deduced founded on the collected data.Since the vibration transducer can be attached on the machine surface using magnetic,paraffin and glue,which is a growing trend in surface structure protection that widely welcome by people recently.In recent years,with the widespread adoption of the concept of industrial digitization,the processing of cavitation-induced vibration signals and the discrimination of states related to it have gradually become a challenging issue in the information management and maintenance of centrifugal pumps.However,current research on cavitation phenomena within centrifugal pumps and their induced vibrations is mostly based on ideal environmental conditions.Exploring the characteristics of cavitation-induced vibrations under adverse engineering conditions is still in its early stages of development.Therefore,the objective and scientific selection of vibration characteristic parameters for real-time monitoring of cavitation states,ensuring that the extracted vibration parameters can accurately represent cavitation phenomena from the vibration signals,is an essential prerequisite for quantitatively analyzing pumps in engineering practice or providing an integrated solution strategy.The dissertation,supported by the National Key Research and Development Program of China "Multi-objective hydraulic design method of vane pump and key technology of gas-liquid two-phase pressurized transportation(2018YFB0606103),addresses the practical needs of various types of pumps in engineering practice.Specifically,it focuses on the mechanical seal centrifugal pump,magnetic drive pump with an inducer,and compacting high-speed magnetic pump.Mathematical methods are used to quantitatively analyze their vibration states and explore the application and practice of digital signal processing techniques(including 1-dimension and 2-dimension methods)in these pumps.Various phenomena and issues that may occur during the daily operation of the pumps are analyzed using mathematical models,and corresponding quantitative analyses or solutions are proposed based on the results obtained.The main research work and innovations of the dissertation are as follows:(1)For the potential mechanical damage or signal interference in traditional mechanical seal pumps,the wavelet packet algorithm is proposed to study the vibration characteristics of acquired signals under different cavitation allowances.The development trend and characteristics of specific frequency bands after decomposition are discussed.Research revealed that mechanical looseness,as it changes over time,can exert a continuous but inconsistent influence on the lowfrequency vibrations of the pump.Therefore,utilizing sub-frequency band signals in the mid to low-frequency range to characterize the vibration state of the pump is beneficial for investigating the vibrations induced by cavitation in the pump.(2)In order to objectively and scientifically select vibration characteristic parameters for realtime monitoring of cavitation status,the dissertation explores the possibility of applying the grey correlation analysis combined with the entropy weight method to study the correlation of cavitation vibration in pumps.As traditional grey correlation analysis cannot evaluate the positive and negative correlation between parameters,the grey slope correlation method is used as an alternative evaluation method.Furthermore,this method is extended to study the correlation between the position of vibration monitoring points and cavitation allowances.Recommendations are provided for the selection of vibration parameters and the choice of monitoring point positions.The research found that the four characteristic parameters: absolute mean,variance,standard deviation,and root mean square,are more conducive to studying cavitation-induced vibration phenomena than other characteristic parameters.For compact high-speed magnetic pumps,it is recommended to use the vibration signals collected in the gravity direction at the top of the volute casing as the target signal for evaluating cavitation states.(3)To predict the energy of vibration signals in mechanical seal pumps under cavitation conditions,a numerical fitting method is explored.Considering the instability of cavitationinduced vibration and the susceptibility of measured signals to interference from unknown factors,this study attempts to use wavelet packet transform to process raw vibration signals and capture the 2-norm of specific frequency bands to reflect the intensity of vibrations.The non-equal time interval grey prediction model is innovatively weighted and applied to establish the function relationship between cavitation and vibration,achieving the purpose of predicting future conditions based on historical vibration signals.Research indicated that the grey prediction model,when used in situations with limited data and sparse information,can estimate the vibration conditions in specified extent.However,it is constrained by the limitations inherent to the numerical fitting model itself.The prediction accuracy is closely related to the quality of the data used and the prediction interval.(4)To objectively understand the relationship between cavitation evolution and vibration,a process of image processing is designed for separating images containing only cavitation from the images of the inducer wheel.The mathematical algorithm of intrinsic orthogonal decomposition is used to analyze the morphological characteristics of cavitation images.Based on the results obtained,the relationship between cavitation area and unit vibration is determined.The research findings reveal that as the statistically derived average bubble area increases,the variance,absolute mean,and root mean square values extracted from the high-frequency range also increase with a decrease in net positive suction head.(5)Regarding the decay of the self-healing system in compact magnetic pumps,it is demonstrated that equipment failure may be a probabilistic problem evolving over time,with a focus on the impact of the time dimension on the pump unit’s decay process.Therefore,a probability-driven unit state diagnosis mechanism is proposed for the operating state of the unit.This mechanism is based on historical vibration data,utilizing mathematical models such as wavelet packet signal decomposition,K-means clustering analysis,and hidden Markov probability algorithm to diagnose the health status of the unit system on the basis of traditional fault diagnosis.The dissertation provided some insights for the information management and maintenance of pumps,particularly in the fields of vibration signal processing and state discrimination.This holds significant importance for the management of pump equipment in engineering practice and has the potential to drive the digitalization of pump equipment,thus enhancing operational efficiency and reliability.