| Cavitation is something that needs to be avoided in the operation of hydraulic machinery.Its production will not only lead to a decrease in performance and efficiency of the unit,but also induce vibration and noise of the unit.When it is in a cavitation state for a long time,it will also cause corrosion damage to the flow components and affect the stable operation of the unit.The cavitation identification method can prevent the hydraulic machinery from operating in a cavitation state for a long time,and can help carry out mechanical repair and maintenance in advance to avoid the occurrence of unit failure.Therefore,how to accurately and quickly identify and predict cavitation has become the focus of current hydraulic machinery research.This paper takes the single-vane centrifugal pump as the research object,through the method of combining numerical simulation and experiment,and explores the hydrodynamic noise evolution law of the centrifugal pump under different cavitation development degrees.A cavitation recognition based on support vector machine method is proposed,and developed a centrifugal pump cavitation recognition system.The main research content and results of this article are as follows:(1)The SST k-ωturbulence model was selected,and the cavitation characteristics of the single-vane centrifugal pump under the five flow conditions of 0.6Q_d,0.8Q_d,1.0Q_d,1.2Q_d,1.4Q_d were numerically simulated.Under the rated flow condition 1.0Q_d,with the continuous decrease of the cavitation number,the degree of cavitation was intensified,and the cavitation in the impeller channel continued to develop.When the initial critical cavitation number was reached,the head of the centrifugal pump began to decrease sharply,and the distribution position of the cavitation gradually developed from the leading edge of the impeller blade suction surface to the trailing edge.The length of the cavity increased continuously,and finally developed to the vicinity of the pressure surface of the impeller blade.When the cavitation number of the single-blade centrifugal pump wasσ=0.064,the rated operating condition 1.0Q_d was the critical initial cavitation point.The volume distribution of the five different flows of cavitation was compared and analyzed.Under the small flow condition 0.6Q_d,the degree of cavitation was relatively slight,and there were only a few cavitation distributions.Under the large flow condition 1.4Q_d,the cavitation was very serious,and the cavitation even developed to the impeller outlet,and had a huge impact on the flow field.(2)Cavitation tests were carried out on the single-vane centrifugal pump,and the cavitation characteristic curves under five flow conditions were analyzed.The influence of different flow rates on the cavitation characteristics was discussed.The research results showed that as the cavitation number decreased,the degree of cavitation continued to increase.In the initial stage of cavitation,the number of cavitation was large,and the generation of cavitation in the impeller made the surface of the blade smooth,and the head of the single-blade centrifugal pump rose slightly.With the further deepening of cavitation,the head continued to decrease,and the cavitation in the flow channel was continuously produced and collapsed,which ultimately affected the normal operation of the fluid in the flow channel.Comparing the cavitation curves under five flow rates,as the flow rate increased,the pressure at the pump inlet became lower,resulting in a continuous increase in the number of critical primary cavitation,and the degree of cavitation became more obvious.(3)The noise test of single-vane centrifugal pumps under normal operating conditions,initial cavitation conditions and severe cavitation conditions was carried out.The noise signals of the pump inlet and outlet were monitored through hydrophones,and the time domain,frequency domain,and time-frequency domain were carried out to analysis respectively.The research results showed that under normal operating conditions,the amplitude of the noise signal at the inlet and outlet of the centrifugal pump was constantly changing.With the intensification of cavitation,the peak of the noise at the pump inlet first rose up and then fell down,and the main frequency was49Hz.The noise energy was mainly concentrated in the frequency conversion and the first several frequency multiplications.In the pump outlet section,the influence of cavitation was small,and the amplitude of the noise signal changed little.In the initial cavitation stage,the low-frequency signal at the pump inlet increased,and the main frequency became a low-frequency signal less than 20Hz.The signal amplitude greater than the double frequency was very small and can be basically ignored.At the pump outlet,the main frequency was still the motor rotation frequency,but the amplitude was small,and the amplitude of the high-order frequency doubling basically disappeared.In the severe cavitation stage,the main frequency of the noise signal at the inlet of the pump was restored to the motor frequency,and the low-frequency amplitude was reduced and no longer dominated.While the noise signal amplitude at the pump outlet increased with a little degree,and restored to the main frequency of the motor 49Hz.(4)Fourier transform and support vector machine(SVM)were used as the classifier to classify and recognize the three different degrees of cavitation of the single-blade centrifugal pump,and introduced the Gaussian kernel function to map the data from low-dimensional space to high-dimensional space,and select the most excellent penalty factor C and parameter g to improve the accuracy of support vector machine classification and recognition.The research results showed that when the time domain feature values were used as training samples for SVM training,the recognition accuracy of skewness was higher,and the recognition rate of standard deviation was the lowest.When the frequency domain feature value was used for training,the accuracy of empty talk recognition was improved,and the highest can reach more than 99%,but the accuracy was still low in the case of small traffic.When the time-frequency domain was selected as the training sample,the accuracy of cavitation recognition was the highest due to more information.Among them,the accuracy of cavitation recognition was the highest under rated conditions 1.0Q_d,which can reach more than 99.5%.The accuracy of cavitation recognition under small flow conditions 0.6Q_d was lower than that under other flow conditions,but it was also above 90%. |
