Cavitation state detection and health diagnosis system for refrigerant pump under severe working condition

Abstract

The invention discloses a cavitation state detection and health diagnosis system for a refrigerant pump under a severe working condition. The cavitation state detection and health diagnosis system comprises the refrigerant pump; a refrigerant tank; the cavitation state simulation test stand and the data processing and control prediction terminal are used for receiving the performance parameter data sent by the cavitation state simulation test stand, performing Fourier transform and wavelet transform on the performance parameter data, performing deep learning by taking a result as an input end of an LSTM model, extracting feature vectors of different frequency bands, and outputting the feature vectors to the data processing and control prediction terminal; and as a cavitation fault index standard, accurate prediction of the cavitation state is realized. According to the method, multiple data can be collected in real time, signals in different time frequency domains are extracted to serve as input data of a neural network, the cavitation intensity and position are subjected to refined analysis in combination with a CFD hydrodynamic model, and therefore predictive maintenance of cavitation faults is achieved; equipment damage caused by excessive cavitation of the refrigerant pump is effectively prevented, and performance reduction and various risks are further prevented.

Description

Technical Field

[0001] The present invention relates to the technical field of testing systems, and in particular to a cavitation state detection and health diagnosis system for a refrigerant pump under severe working conditions. Background Art

[0002] In fluids, cavitation refers to the phenomenon of vapor or gas cavities (i.e., bubbles) formed inside a liquid or on a liquid-solid interface when the local pressure drops below the saturated vapor pressure. These bubbles may develop and aggregate as the fluid flows, and quickly collapse when the pressure is restored. Vibration and pulsation data monitoring can effectively measure cavitation indirectly, and the full cavitation model in CFD is one of the most effective means of analyzing refrigerant pump cavitation, providing important theoretical guidance for experiments and monitoring.

[0003] During the operation of the refrigerant pump, especially under harsh working conditions, the cavitation phenomenon is difficult to predict due to the volatility and low viscosity of the refrigerant medium. The data noise is very serious. It is difficult to filter out effective information with ordinary data processing alone, which brings lag and inaccuracy to maintenance. Cavitation not only affects the operating efficiency of the refrigerant pump, but also may damage the equipment and increase maintenance costs. Summary of the invention

[0004] The purpose of the present invention is to provide a technical solution for cavitation status detection and health diagnosis system under severe working conditions of refrigerant pumps in response to the shortcomings of the prior art. It can not only collect multiple data of refrigerant pumps under different working conditions in real time, and extract more effective information from signals in different time and frequency domains, thereby serving as input data for neural networks, but also combine with the CFD fluid dynamics model to make a detailed analysis of cavitation intensity and position, thereby realizing predictive maintenance of cavitation failures, and effectively preventing equipment damage caused by excessive cavitation of refrigerant pumps, which in turn brings about performance degradation and various risks.

[0005] In order to solve the above technical problems, the present invention adopts the following technical solutions:

[0006] The cavitation state detection and health diagnosis system of a refrigerant pump under severe working conditions is characterized by: comprising a refrigerant pump;

[0007] A refrigerant tank;

[0008] A cavitation state simulation test bench, the cavitation state simulation test bench includes an inlet pressure flow detection unit, an outlet pressure flow detection unit and a thermal management unit, the refrigerant pump is connected to the refrigerant tank through the inlet pressure flow detection unit and the outlet pressure flow detection unit, and the thermal management unit is connected to the refrigerant tank, which is used to collect the performance parameter data of the refrigerant pump in real time and realize the transportation and recovery of the refrigerant;

[0009] A data processing and control prediction terminal is used to receive the performance parameter data sent by the cavitation state simulation test bench, and perform Fourier transform and wavelet transform on the performance parameter data. The results are used as the input of the LSTM model for deep learning, and the feature vectors of different frequency bands are extracted as the standard of cavitation fault indicators to achieve accurate prediction of the cavitation state.

[0010] The system can not only collect multiple data of refrigerant pumps under different working conditions in real time, and extract more effective information from signals in different time and frequency domains as input data for the neural network, but also combine with the CFD fluid dynamics model to make a detailed analysis of the cavitation intensity and location, thereby realizing predictive maintenance of cavitation failures and effectively preventing equipment damage caused by excessive cavitation of the refrigerant pump, which in turn leads to performance degradation and various risks.

[0011] Furthermore, when Fourier transform and wavelet transform are performed on the performance parameter data, low-pass filtering is first used to preprocess the data, and then Fourier transform is used to transform the vibration and pulsation signals. At the same time, the original time series data is used as input, and it is decomposed into components of different frequencies and scales through wavelet transform. The wavelet coefficients of each component are extracted as the input of the subsequent LSTM model, and the frequency components are identified through the spectrum diagram. The expression is:

[0012]

[0013] Where j is the imaginary unit, e -j2πkn / N is the rotation factor.

[0014] Furthermore, the processed data is used as the model input layer X1, X2, ... X in the LSTM time series n , and label the corresponding data set according to the valve tightness for learning. The expression of the LSTM layer is:

[0015]

[0016] Among them, f t is the forgetting coefficient; i t is the input coefficient; is the input data; C t is the updated cell state; o t is the output coefficient; h t is the output data; σ is the sigmoid activation function; W f , W i , W c and W o are the forget gate weight, input gate weight, input data weight and output gate weight respectively; b f 、b i 、bc and b o They are forget gate bias, input gate bias, input data bias and output gate bias respectively.

[0017] Furthermore, when collecting the performance parameter data of the refrigerant pump, the rotor area and the overall pump area are divided, the internal flow characteristics of the refrigerant pump are analyzed through the full cavitation model of CFD, and the collected data are weighted. The full cavitation model is obtained based on the RP equation, and the expression of the RP equation is:

[0018]

[0019] Among them, r is the bubble radius, υ is the fluid kinematic viscosity, and σ is the fluid surface tension.

[0020] Furthermore, the inlet pressure flow detection unit includes an inlet flow sensor, a first valve, an inlet pressure sensor and a first vibration sensor. The inlet flow sensor is used to monitor the inlet flow pulsation curve of the refrigerant pump under different working conditions. The first valve is used to control the medium flow at the inlet of the refrigerant pump. The inlet pressure sensor is used to obtain the pressure pulsation in the inlet area of ​​the refrigerant pump. The first vibration sensor is used to obtain vibration data of the area near the refrigerant pump rotor.

[0021] Furthermore, the outlet pressure flow detection unit includes an outlet flow sensor, a second valve, an outlet pressure sensor and a second vibration sensor. The outlet flow sensor is used to monitor the outlet flow pulsation curve of the refrigerant pump under different working conditions. The second valve is used to control the medium flow at the outlet of the refrigerant pump and adjust the valve tightness to simulate cavitation. The outlet pressure sensor is used to obtain the pressure pulsation in the outlet area of ​​the refrigerant pump. The second vibration sensor is used to obtain vibration data of the area near the outlet of the refrigerant pump.

[0022] Furthermore, a third valve and a hexagonal screw are connected to one side of the refrigerant tank close to the outlet pressure flow detection unit. The third valve is used to adjust the inlet and outlet pressure difference of the refrigerant pump, and the hexagonal screw is used to adjust the tightness of the third valve.

[0023] Furthermore, the thermal management unit includes a temperature sensor, a temperature control module, a blower cooler and at least one heat exchanger. The temperature sensor is used to obtain the real-time temperature in the working state and send the temperature data to the temperature control module. The temperature control module is used to receive the temperature data and adjust the temperature in the pipeline in the working state. The heat exchanger and the blower cooler are used to adjust the temperature.

[0024] Furthermore, the refrigerant pump is connected to a frequency converter speed control module for controlling the speed of the motor of the refrigerant pump.

[0025] The present invention has the following beneficial effects due to the adoption of the above technical solution:

[0026] 1. The present invention can not only collect multiple data of the refrigerant pump under different working conditions in real time, and extract more effective information from signals in different time and frequency domains as input data of the neural network, but also combine it with the CFD fluid dynamics model to make a detailed analysis of the cavitation intensity and position, thereby realizing predictive maintenance of cavitation failures, and effectively preventing equipment damage caused by excessive cavitation of the refrigerant pump, which in turn leads to performance degradation and various risks.

[0027] 2. The present invention not only visualizes the performance of the refrigerant pump, but also dynamically monitors the performance of the refrigerant pump, thereby greatly reducing the difficulty of predictive maintenance of the equipment and reducing maintenance costs.

[0028] 3. The present invention fully solves the problem of cavitation phenomenon being difficult to diagnose by integrating Fourier transform and wavelet transform. Fourier transform mainly analyzes the frequency domain, which can effectively judge the influence of different cavitation degrees, clearly reveal the frequency components of the signal, and has a good analytical effect on the high-frequency signal of cavitation. On this basis, the innovative fusion of wavelet transform makes up for the information loss caused by the loss of time domain in Fourier transform, and avoids the problems of local low-frequency analysis defects and misjudgment.

[0029] 4. The present invention uses the results of Fourier transform and wavelet transform as the input layer of the LSTM model. LSTM solves the gradient vanishing and gradient exploding problems of traditional RNN when processing long sequence data. At the same time, it can capture the dependency between vibration and pulsation phenomena and cavitation based on the historical data of long-term testing of the refrigerant pump, and shows good robustness when processing noisy data and irregular sequences, so that the final cavitation fault diagnosis result has a higher credibility. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] The present invention will be further described below in conjunction with the accompanying drawings:

[0031] Figure 1 It is a structural schematic diagram of the cavitation state detection and health diagnosis system of the refrigerant pump under severe working conditions of the present invention;

[0032] Figure 2 is a schematic diagram of the structure of the thermal management unit in the present invention;

[0033] Figure 3 It is a flow chart of the refrigerant pump cavitation fault diagnosis and data collection and analysis in the present invention;

[0034] Figure 4 It is the pressure difference cavitation cloud diagram and cavitation characteristic curve diagram in the present invention.

[0035] In the figure: 1-refrigerant pump; 2-frequency converter speed control module; 3-outlet pressure sensor; 4-hexagonal screw; 5-second valve; 6-outlet flow sensor; 7-third valve; 8-data processing and control prediction terminal; 9-refrigerant tank; 10-temperature sensor; 11-temperature control module; 12-heat exchanger; 13-inlet flow sensor; 14-first valve; 15-inlet pressure sensor; 16-first vibration sensor; 17-blast cooler; 18-second vibration sensor. DETAILED DESCRIPTION

[0036] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0037] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only embodiments of a part of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0038] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions.

[0039] like Figures 1 to 4 As shown, the cavitation state detection and health diagnosis system of the refrigerant pump under severe working conditions of the present invention comprises a refrigerant pump 1, a refrigerant tank 9, a cavitation state simulation test bench and a data processing and control prediction terminal 8. The refrigerant pump 1 adopts a fully enclosed refrigerant pump 1.

[0040] The cavitation state simulation test bench includes an inlet pressure flow detection unit, an outlet pressure flow detection unit and a thermal management unit. The refrigerant pump 1 is connected to the refrigerant tank 9 through the inlet pressure flow detection unit and the outlet pressure flow detection unit.

[0041] The inlet pressure flow detection unit includes an inlet flow sensor 13, a first valve 14, an inlet pressure sensor 15 and a first vibration sensor 16. The inlet flow sensor 13 is used to monitor the inlet flow pulsation curve of the refrigerant pump 1 under different working conditions. The first valve 14 is used to control the medium flow at the inlet of the refrigerant pump 1. The inlet pressure sensor 15 is used to obtain the pressure pulsation in the inlet area of ​​the refrigerant pump 1. The first vibration sensor 16 is used to obtain the vibration data of the area near the rotor of the refrigerant pump 1.

[0042] The outlet pressure flow detection unit includes an outlet flow sensor 6, a second valve 5, an outlet pressure sensor 3 and a second vibration sensor 18. The outlet flow sensor 6 is used to monitor the outlet flow pulsation curve of the refrigerant pump 1 under different working conditions. The second valve 5 is used to control the medium flow at the outlet of the refrigerant pump 1 and adjust the valve tightness to simulate cavitation. The outlet pressure sensor 3 is used to obtain the pressure pulsation in the outlet area of ​​the refrigerant pump 1. The second vibration sensor 18 is used to obtain vibration data of the area near the outlet of the refrigerant pump 1.

[0043] The third valve 7 and the hexagonal screw 4 are connected to the side of the refrigerant tank 9 close to the outlet pressure flow detection unit. The third valve 7 is used to adjust the inlet and outlet pressure difference of the refrigerant pump 1, and the hexagonal screw 4 is used to adjust the tightness of the third valve 7. The larger the opening of the third valve 7, the greater the pressure difference. Different valve tightness represents the simulation conditions under different cavitation severity, and the parameters under each valve tightness will be measured continuously for a period of time.

[0044] The refrigerant pump 1 is connected to a frequency converter speed control module 2 for controlling the speed of the motor of the refrigerant pump 1 .

[0045] The thermal management unit is connected to the refrigerant tank 9 and is used to collect the performance parameter data of the refrigerant pump 1 in real time and realize the transportation and recovery of the refrigerant. The thermal management unit is intended to prevent the system efficiency from decreasing or cavitation due to excessively high or low temperature by accurately controlling the temperature of the refrigerant.

[0046] The thermal management unit includes a temperature sensor 10 , a temperature control module 11 , a blast cooler 17 and at least one heat exchanger 12 .

[0047] The temperature sensor 10 is used to obtain the real-time temperature of the working state and send the temperature data to the temperature control module 11. The temperature sensor 10 monitors the temperature of the refrigerant in real time and feeds back the temperature signal to the temperature control module 11. It is installed on the flow path of the refrigerant, such as the inlet and outlet of the cooling water circulation.

[0048] The temperature control module 11 is used to receive temperature data and adjust the temperature in the pipeline during the working state. The temperature control module 11 receives the signal of the temperature sensor 10, automatically adjusts according to the preset temperature range, and controls the working state of the blast cooler 17 and the heat exchanger 12. The PID control method can be used to achieve high-precision temperature control by adjusting the proportional coefficient, integral time and differential time, and manual control can also be performed.

[0049] The heat exchanger 12 and the blast cooler 17 are used to adjust the temperature. The blast cooler 17 uses a blower to direct air toward the condensation pipeline to cool the refrigerant by physical heat dissipation. The heat exchanger 12 transfers heat to the refrigerant by contacting other heat sources to increase or decrease the temperature. Three are installed at the same time to improve the heat exchange efficiency. The spiral heat exchanger 12 is mainly used.

[0050] After the system is started, the temperature sensor 10 starts to monitor the temperature of the refrigerant in real time and transmits a signal to the temperature control module 11. The temperature control module 11 determines whether the current temperature of the refrigerant needs to be adjusted according to a preset temperature range.

[0051] If the temperature of the refrigerant is too high, the temperature control module 11 will control the air blast cooler 17 to start, and cool the refrigerant by forced convection. If the temperature of the refrigerant is too high, the temperature control module 11 will control the air blast cooler 17 to start, and cool the refrigerant by forced convection. If the temperature of the refrigerant is too high, the temperature control module 11 will control the air blast cooler 17 to start, and cool the refrigerant by forced convection. After the system runs stably, the temperature sensor 10 continues to monitor the temperature of the refrigerant in real time, and the temperature control module 11 makes fine adjustments according to the feedback signal to ensure that the system is always in the best working state.

[0052] As the temperature increases, the difference between the external pressure and the saturated vapor pressure decreases, resulting in a decrease in the cavitation number. The decrease in the cavitation number means that the gas nucleus is more likely to develop into bubbles, thereby increasing the possibility of cavitation. The increase in temperature increases the saturated vapor pressure of the liquid, which is conducive to the development of more gas nuclei into bubbles. At the same time, the bubbles expand as the surrounding temperature increases, and the space they occupy becomes larger and larger. The increase in temperature will cause the density of the liquid to decrease, the dynamic viscosity to decrease, and the surface tension to decrease. These changes have an important impact on the development and collapse of cavitation. For example, the decrease in dynamic viscosity weakens the damping effect on the development and collapse of cavitation, which increases the maximum radius of cavitation development and decreases the minimum radius of collapse. At the same time, the cold medium is very volatile at high temperatures, so it is necessary to continuously control the temperature to control cavitation and evaporation.

[0053] The data processing and control prediction terminal 8 is used to receive the performance parameter data sent by the cavitation state simulation test bench, and perform Fourier transform and wavelet transform on the performance parameter data, and use the results as the input of the LSTM model for deep learning, extract the feature vectors of different frequency bands as the standard of cavitation fault indicators, and realize accurate prediction of the cavitation state.

[0054] When Fourier transform and wavelet transform are performed on the performance parameter data, low-pass filtering is first used to preprocess the data, and then Fourier transform is used to transform the vibration and pulsation signals. At the same time, the original time series data is used as input, and it is decomposed into components of different frequencies and scales through wavelet transform. The wavelet coefficients of each component are extracted as the input of the subsequent LSTM model, and the frequency components are identified through the spectrum diagram. The expression is:

[0055]

[0056] Where j is the imaginary unit, e -j2πkn / N is the rotation factor.

[0057] Fourier transform can convert vibration signals from time domain to frequency domain, thus clearly showing the frequency components in the signal. Through Fourier transform, the individual frequency components in the vibration signal, as well as their amplitude and phase information, can be accurately identified. Cavitation faults usually cause specific characteristic frequencies to appear in the vibration signal.

[0058] The processed data is used as X1, X2, ... Xn in the LSTM time series model input layer, and the corresponding data sets are labeled according to the valve tightness for learning.

[0059] In the output layer, 0 corresponds to no obvious cavitation, 1 corresponds to slight cavitation, 2 corresponds to moderate cavitation, and 3 corresponds to high cavitation.

[0060] The LSTM layer consists of multiple neural units. The internal structure of each neural unit consists of a forget gate, an input gate, a memory cell, and an output gate. The expression is:

[0061]

[0062] Among them, f t is the forgetting coefficient; i t is the input coefficient; is the input data; C t is the updated cell state; o t is the output coefficient; h t is the output data; σ is the sigmoid activation function; W f , W i , W c and W o are the forget gate weight, input gate weight, input data weight and output gate weight respectively; b f 、b i 、b c and b o They are forget gate bias, input gate bias, input data bias and output gate bias respectively.

[0063] The cavitation location is mainly divided into the rotor area and the overall area, and the cavitation occurrence in the rotor area is monitored. In order to better understand the mechanism and characteristics of cavitation, the internal flow characteristics of the refrigerant twin-rotor pump are analyzed through the CFD full cavitation model. The cavitation cloud diagram shows (such as Figure 4 As shown in the figure, the greater the pressure difference, the greater the pressure pulsation, the smaller the flow rate, and the overall cavitation of the refrigerant pump 1 shows a downward trend, but the cavitation volume fraction at the rotor meshing point shows a trend of convergence, indicating that the intensity and main occurrence location of cavitation will change due to different pressure differences, and damage will also occur at the corresponding location of the equipment. At the same time, the occurrence of cavitation will reduce the equipment flow rate and increase the pulsation rate.

[0064] When collecting the performance parameter data of the refrigerant pump 1, the rotor area and the overall pump area are divided, and the internal flow characteristics of the refrigerant pump 1 are analyzed through the CFD full cavitation model, and the vibration, flow and pulsation data are weighted. The full cavitation model is derived based on the Rayleigh-Plesset equation, which is referred to as the RP equation. The expression of the RP equation considering surface tension and viscosity is:

[0065]

[0066] Among them, r is the bubble radius, υ is the fluid kinematic viscosity, and σ is the fluid surface tension.

[0067] After the LSTM model has trained the data set, it will use part of the data for testing and comparison to verify the accuracy of the model. When the model meets expectations, it will be saved and cavitation prediction steps will be performed. By randomly setting the valve tightness to simulate the unknown equipment operation, the computer collects data, analyzes the data and inputs it into the model. The model will judge and output the cavitation intensity, cavitation position and valve tightness. If severe cavitation occurs, the host computer will send a text message to the mobile phone to implement the early warning function. At the same time, the real-time cavitation data of refrigerant pump 1 will be visualized on the cloud screen.

[0068] The system can not only collect multiple data of the refrigerant pump 1 under different working conditions in real time, and extract more effective information from signals in different time and frequency domains as input data for the neural network, but also combine it with the CFD fluid dynamics model to make a detailed analysis of the cavitation intensity and location, thereby realizing predictive maintenance of cavitation failures and effectively preventing equipment damage to the refrigerant pump 1 caused by excessive cavitation, which in turn leads to performance degradation and various risks.

[0069] When the present invention is actually working:

[0070] 1. Open the first valve and the second valve at the same time. At this time, the refrigerant pump system is in a circulation state. Then, the refrigerant pump rotor is started and operated through the variable frequency speed regulation module. At this time, the pressure difference between the inlet and outlet of the equipment is 0. Then, the pressure difference is controlled by adjusting the third valve to test the data under the required head.

[0071] 2. By adjusting the tightness of the second valve, the outlet pressure is lower than the normal operating 0.8MPa. The rotor area is more likely to produce local low pressure, which is lower than the R134 saturated vapor pressure of 0.665MPa, resulting in obvious cavitation. Then, 4 different valve tightnesses are selected, corresponding to cavitation levels 0 to 3, as labels for subsequent fault learning. Each label runs for 2 hours to test data transformation.

[0072] 3. Through the data processing and control prediction terminal, all sensor data are collected and sent to the cloud, and Fourier transform and wavelet transform are performed on them respectively. Wavelet transform decomposes vibration and pulsation data into components of different frequencies and time scales. This is achieved by selecting the appropriate wavelet basis function and the number of decomposition layers.

[0073] 4. After receiving the preprocessed data and the corresponding empty labels, the LSTM model normalizes or standardizes the extracted features to ensure that all features are on the same scale. The model is continuously trained and iterated, and the dimensions of the LSTM input layer are set according to the dimensions of the feature vector. The loss and performance indicators during the training process are monitored, and the model structure and hyperparameters are adjusted.

[0074] 5. After the model training is completed and saved, the refrigerant pump performance data with random valve tightness is input into the LSTM model through Fourier transform and wavelet transform. The model evaluates the cavitation intensity by outputting 0, 1, 2, 3, and determines the key location of cavitation by outputting A and B. The data is visualized on the cloud server.

[0075] The above are only specific embodiments of the present invention, but the technical features of the present invention are not limited thereto. Any simple changes, equivalent replacements or modifications made based on the present invention to achieve basically the same technical effects are all included in the protection scope of the present invention.

Claims

1. Cavitation state detection and health diagnosis system for refrigerant pumps under severe working conditions, characterized by: including a refrigerant pump; A refrigerant tank; A cavitation state simulation test bench, the cavitation state simulation test bench comprises an inlet pressure flow detection unit, an outlet pressure flow detection unit and a thermal management unit, the refrigerant pump is connected to the refrigerant tank through the inlet pressure flow detection unit and the outlet pressure flow detection unit, the thermal management unit is connected to the refrigerant tank, and is used to collect performance parameter data of the refrigerant pump in real time, and realize the transportation and recovery of the refrigerant; and a data processing and control prediction terminal, which is used to receive the performance parameter data sent by the cavitation state simulation test bench, and perform Fourier transform and wavelet transform on the performance parameter data, and use the results as the input of the LSTM model for deep learning, extract feature vectors of different frequency bands as the standard of cavitation fault indicators, and realize accurate prediction of the cavitation state.

2. The cavitation state detection and health diagnosis system for refrigerant pumps under severe working conditions according to claim 1, characterized in that: When Fourier transform and wavelet transform are performed on the performance parameter data, low-pass filtering is first used to preprocess the data, and then Fourier transform is used to transform the vibration and pulsation signals. At the same time, the original time series data is used as input, and it is decomposed into components of different frequencies and scales through wavelet transform. The wavelet coefficients of each component are extracted as the input of the subsequent LSTM model, and the frequency components are identified through the spectrum diagram. The expression is: Where j is the imaginary unit, e -j2vkn / N is the rotation factor.

3. The cavitation state detection and health diagnosis system for refrigerant pumps under severe working conditions according to claim 2, characterized in that: The processed data is used as the model input layer X1, X2, ... X in the LSTM time series n , and label the corresponding data set according to the valve tightness for learning. The expression of the LSTM layer is: Among them, f t is the forgetting coefficient; i t is the input coefficient; is the input data; C t is the updated cell state; o t is the output coefficient; h t is the output data; σ is the sigmoid activation function; W f , W i , W c and W o are the forget gate weight, input gate weight, input data weight and output gate weight respectively; b f , b i , b c and b o They are forget gate bias, input gate bias, input data bias and output gate bias respectively.

4. The cavitation state detection and health diagnosis system for refrigerant pumps under severe working conditions according to claim 1, characterized in that: When collecting the performance parameter data of the refrigerant pump, the rotor area and the overall pump area are divided, the internal flow characteristics of the refrigerant pump are analyzed by the full cavitation model of CFD, and the collected data are weighted. The full cavitation model is obtained based on the RP equation, and the expression of the RP equation is: Among them, r is the bubble radius, υ is the fluid kinematic viscosity, and σ is the fluid surface tension.

5. The cavitation state detection and health diagnosis system for refrigerant pumps under severe working conditions according to claim 1, characterized in that: The inlet pressure flow detection unit includes an inlet flow sensor, a first valve, an inlet pressure sensor and a first vibration sensor. The inlet flow sensor is used to monitor the inlet flow pulsation curve of the refrigerant pump under different working conditions. The first valve is used to control the flow of medium at the inlet of the refrigerant pump. The inlet pressure sensor is used to obtain the pressure pulsation in the inlet area of ​​the refrigerant pump. The first vibration sensor is used to obtain vibration data of the area near the refrigerant pump rotor.

6. The cavitation state detection and health diagnosis system for refrigerant pumps under severe working conditions according to claim 1, characterized in that: The outlet pressure flow detection unit includes an outlet flow sensor, a second valve, an outlet pressure sensor and a second vibration sensor. The outlet flow sensor is used to monitor the outlet flow pulsation curve of the refrigerant pump under different working conditions. The second valve is used to control the medium flow at the outlet of the refrigerant pump and adjust the valve tightness to simulate cavitation. The outlet pressure sensor is used to obtain the pressure pulsation in the outlet area of ​​the refrigerant pump. The second vibration sensor is used to obtain vibration data of the area near the outlet of the refrigerant pump.

7. The cavitation state detection and health diagnosis system for refrigerant pumps under severe working conditions according to claim 1, characterized in that: A third valve and a hexagonal screw are connected to a side of the refrigerant tank close to the outlet pressure flow detection unit. The third valve is used to adjust the inlet and outlet pressure difference of the refrigerant pump, and the hexagonal screw is used to adjust the tightness of the third valve.

8. The cavitation state detection and health diagnosis system for refrigerant pumps under severe working conditions according to claim 1, characterized in that: The thermal management unit includes a temperature sensor, a temperature control module, a blast cooler and at least one heat exchanger. The temperature sensor is used to obtain the real-time temperature in the working state and send the temperature data to the temperature control module. The temperature control module is used to receive the temperature data and adjust the temperature in the pipeline in the working state. The heat exchanger and the blast cooler are used to adjust the temperature.

9. The cavitation state detection and health diagnosis system for refrigerant pumps under severe working conditions according to claim 1, characterized in that: The refrigerant pump is connected to a frequency converter speed regulation module for regulating the speed of the motor of the refrigerant pump.

Images