A test method and system for water medium in a high-flow vertical centrifugal pump

Abstract

This application relates to the technical field of centrifugal pumps, and discloses a method and system for testing water media in a large-flow vertical centrifugal pump. The method includes the following steps: preparing before water cutoff; setting a time window and sampling frequency, and dividing the time window into two segments; after water cutoff, collecting bearing temperature, mechanical seal temperature, and motor winding temperature at the sampling frequency, and simultaneously recording water temperature, ambient temperature, vibration intensity, and noise data; fusing the data obtained from the first segment of a time window into input features based on window sliding; inputting the input features into a pre-trained LSTM model to obtain the predicted temperature rise trend, and calculating the deviation between the actual temperature rise and the predicted temperature rise. If the deviation is greater than a preset threshold, it indicates an abnormal temperature rise trend; if the deviation is less than the preset threshold, it indicates a normal temperature rise trend. This application has the advantage of fully utilizing historical data, extracting deep time series features, and improving the accuracy of temperature rise prediction.

Description

Technical Field

[0001] This application relates to the technical field of centrifugal pumps, and in particular to a method and system for testing water media in a high-flow-rate vertical centrifugal pump. Background Technology

[0002] In applications such as nuclear industry, petrochemicals, marine propulsion, and large-scale water-cooling systems, high-flow vertical centrifugal pumps are key equipment, responsible for the stable delivery of coolant at high flow rates over extended periods. Their core structure includes an impeller, volute, pump shaft, mechanical seal, and bearings. The high-speed rotation of the impeller draws liquid in from the inlet and discharges it tangentially under centrifugal force, thus achieving high-flow-rate, stable-pressure fluid delivery. Compared to horizontal centrifugal pumps, vertical centrifugal pumps have a more compact design, reducing floor space and facilitating integration with large cooling systems. Especially in high-temperature, high-pressure, or closed systems, such as the secondary cooling system of a nuclear reactor, the reliability of high-flow vertical centrifugal pumps directly impacts the safety and operational stability of the entire system. Therefore, in practical applications, the pump's cooling performance, sealing reliability, and bearing durability become key evaluation indicators, and the water shortage test is an important means of conducting extreme tests on these indicators.

[0003] Water shortage testing is a specialized extreme condition test for centrifugal pumps. Its purpose is to simulate a sudden failure of the cooling system to evaluate the pump's resilience and temperature rise characteristics when the cooling water supply is interrupted. During normal operation, components of a high-flow vertical centrifugal pump, such as bearings and mechanical seals, rely on cooling water for heat exchange, ensuring that the temperature remains within a safe range. However, in the event of a water shortage, internal friction, shearing, and high-speed rotation cause a rapid temperature rise. If heat dissipation is inadequate, this can lead to bearing overheating, lubrication system failure, deterioration of sealing materials, and even pump body damage. Water shortage testing quantifies the temperature rise rate of these components under extreme conditions and analyzes their failure modes, providing important insights for optimizing pump structural design, improving the cooling system, and enhancing overall reliability.

[0004] Existing water cut-off testing methods still have many shortcomings in terms of testing and data analysis, mainly in the following aspects. First, traditional water cut-off tests usually use a fixed-time sampling method, recording data through temperature sensors, vibration sensors, and noise measurement equipment. However, the analysis of temperature rise trends mainly relies on human experience, making it difficult to predict temperature change trends in real time. Since bearing temperature rise is affected by many factors, such as pump speed, ambient temperature, and lubrication condition, manual analysis is prone to misjudgment or delayed response, making it difficult to accurately predict when the pump will reach a dangerous temperature. Second, existing tests lack systematic research on the impact of microbubbles. The presence of microbubbles in the pump cavity may cause local lubricating oil film rupture, thereby increasing friction and leading to abnormal temperature rise. However, most current test methods only focus on overall temperature rise changes and fail to combine high-frequency noise signals to analyze the impact mechanism of microbubbles, making it difficult to accurately determine the cause of abnormalities under complex operating conditions. Summary of the Invention

[0005] In order to make full use of historical data, extract deep time series features, and improve the accuracy of temperature rise prediction, this application provides a test method and system for water medium of a high-flow vertical centrifugal pump.

[0006] Firstly, this application provides a method for testing water media in a large-flow vertical centrifugal pump, employing the following technical solution:

[0007] A method for testing a high-flow-rate vertical centrifugal pump with water as the medium includes the following steps:

[0008] S1. Prepare for shutting off the water supply to the high-flow vertical centrifugal pump;

[0009] S2. Set the time window and sampling frequency, and divide the time window into two segments. The first segment of the time window is used to acquire data to calculate the predicted temperature rise trend, and the second segment of the time window is used to acquire data to obtain the actual temperature rise trend corresponding to the first segment.

[0010] S3. After the water supply is cut off, the bearing temperature, mechanical seal temperature and motor winding temperature are collected at the sampling frequency, and the water temperature, ambient temperature, vibration intensity and noise data are recorded simultaneously.

[0011] S4. Based on window sliding, the bearing temperature, mechanical seal temperature, motor winding temperature, water temperature, ambient temperature, vibration intensity and noise data obtained in the first part of a time window are fused into input features;

[0012] S5. Input the input features into the pre-trained LSTM model to obtain the predicted temperature rise trend, and calculate the deviation between the actual temperature rise and the predicted temperature rise. If the deviation is greater than the preset threshold, it indicates that the temperature rise trend is abnormal. If the deviation is less than the preset threshold, it indicates that the temperature rise trend is normal.

[0013] Optionally, S1 includes the following steps:

[0014] S101. Maintain the water temperature in the test circuit at 85±10℃;

[0015] S102. The pump chamber liquid level is maintained at 990±200mm below the mounting plane of the large flange on the outer casing;

[0016] S103. The air pressure covering the pump chamber is maintained at 0.33±0.05MPa;

[0017] S104. Adjust the inverter frequency to 20Hz and start the pump;

[0018] S105. Adjust the outlet valve until the main circuit flow meter reaches 4888±200 m³ / h, then stop adjusting the valve;

[0019] S106. After the pump set is running stably, gradually adjust the frequency of the frequency converter to 50Hz so that the pump reaches the rated flow rate of 12220±367m³ / h.

[0020] S107. After stable operation, close the main valve of the cooling water system.

[0021] Optionally, step S106 includes the following steps:

[0022] After the pump set is running stably, the frequency of the frequency converter is adjusted sequentially to 25Hz, 30Hz, 35Hz, 40Hz, 45Hz, and 50Hz to increase the speed.

[0023] After the pump set is running stably, observe whether the main circuit flow rate is 12220±367m³ / h. If it exceeds this range, fine-tune the outlet valve to bring the main circuit flow rate within the range of 12220±367m³ / h.

[0024] Optionally, S1 includes the following steps:

[0025] S111. Maintain the water temperature in the test circuit at 85±10℃;

[0026] S112. The pump chamber liquid level is maintained at 990±200mm below the mounting plane of the large flange on the outer casing;

[0027] S113. The air pressure covering the pump chamber is maintained at 0.33±0.05MPa;

[0028] S114. Adjust the inverter frequency to 20Hz and start the pump;

[0029] S115. Adjust the outlet valve until the main circuit flow meter reaches 4888±200 m³ / h, then stop adjusting the valve;

[0030] S116. After the pump set is running stably, gradually adjust the frequency of the frequency converter to 12.5Hz so that the pump reaches the rated flow rate of 3055±92m³ / h.

[0031] S117. After stable operation, close the main valve of the cooling water system.

[0032] Optionally, S116 includes the following steps:

[0033] After the pump set is running stably, adjust the frequency of the frequency converter to 12.5Hz;

[0034] After the pump set is running stably, observe whether the main circuit flow rate is 3055±92m³ / h. If it exceeds this range, fine-tune the outlet valve to bring the main circuit flow rate within the range of 3055±92m³ / h.

[0035] Optionally, the pre-training step of the LSTM model includes:

[0036] S501. Perform data acquisition and data preprocessing;

[0037] S502. Set a time window of seventy seconds and generate a training set, a validation set, and a test set; wherein, the training set uses the data of the first sixty seconds as input and the temperature rise trend of the last ten seconds as output;

[0038] S503. Construct an LSTM network;

[0039] S504. Input the training set into the LSTM network and calculate the loss function;

[0040] S505. Calculate the gradient of the loss with respect to the weights of the LSTM layer, and use the Adam optimizer to perform gradient descent to update the weights of the LSTM layer;

[0041] S506. After the model training is complete, use the test set and calculate the test set error, and adjust the model complexity based on the test set error.

[0042] Optionally, the structure of the LSTM network in S503 includes:

[0043] The input layer is used to receive the input features corresponding to sixty time steps;

[0044] The first LSTM layer contains sixty-four units and is used to extract time series features and output information for all time steps.

[0045] The second LSTM layer contains thirty-two units and is used to extract deep time series information, outputting only the information of the last time step;

[0046] A fully connected layer is used to map the features extracted by the LSTM to a numerical value to predict the temperature rise trend over the next ten seconds.

[0047] Optionally, S501 includes the following steps:

[0048] S5011. Acquire training data, wherein the training data comes from normal water cut-off test data without microbubbles under different ambient temperatures and different test conditions, abnormal water cut-off test data with microbubbles, and some test data of artificially introduced microbubbles.

[0049] S5012. Detect outliers and remove them;

[0050] S5013. Interpolation to complete missing values;

[0051] S5014. Normalize the data.

[0052] Secondly, this application provides a high-flow-rate vertical centrifugal pump water medium testing system, which adopts the following technical solution:

[0053] A method for testing a large-flow vertical centrifugal pump with water medium includes a processor, wherein the processor runs a program for testing a large-flow vertical centrifugal pump with water medium as described in any one of the above-mentioned methods.

[0054] Thirdly, this application provides a storage medium, which adopts the following technical solution:

[0055] A storage medium storing a program for a high-flow-rate vertical centrifugal pump water medium test method as described in any one of the above.

[0056] In summary, this application includes at least one of the following beneficial technical effects: This solution proposes an intelligent diagnostic method based on deep learning (LSTM), which constructs a time series prediction model to accurately predict the temperature rise trend and combines it with an anomaly detection algorithm to determine whether the temperature rise exceeds the safety threshold. The advantage of this method lies in fully utilizing historical data, extracting deep time series features, improving the accuracy of temperature rise prediction, and combining high-frequency noise analysis to analyze the impact of microbubbles on temperature rise, thus significantly improving the intelligence level of the water cut-off test. Attached Figure Description

[0057] Figure 1 This is a flowchart of a water medium testing method for a high-flow vertical centrifugal pump according to a certain embodiment of this application. Detailed Implementation

[0058] The embodiments of this application are described in detail below, and examples of the embodiments are shown in the accompanying drawings.

[0059] In the description of this specification, the references to "certain embodiments," "one embodiment," "some embodiments," "illustrative embodiment," "example," "specific example," or "some examples" refer to specific features, structures, materials, or characteristics described in connection with the described embodiment or example, which are included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0060] This application discloses a test method for a high-flow-rate vertical centrifugal pump with water medium, referring to... Figure 1 This includes the following steps S1-S5.

[0061] S1. Prepare for shutting off the water supply to the high-flow vertical centrifugal pump.

[0062] High-flow vertical centrifugal pumps are specifically designed for transporting liquids at high flow rates. Their impellers rotate vertically, relying on centrifugal force to lift the liquid to a high-pressure state. They are widely used in the nuclear industry, petrochemical industry, and large-scale hydraulic systems. In these applications, the stability and reliability of the pump are crucial, especially in the event of a cooling water supply interruption. The pump's ability to withstand short-term extreme operating conditions is a key factor affecting system safety. Therefore, a water shortage test is typically conducted, which involves artificially shutting down the cooling system and observing the pump's temperature rise, vibration intensity, and noise changes under different operating conditions to verify its design rationality and operational reliability.

[0063] The purpose of the water shortage test is to simulate potential cooling failure scenarios and assess the pump's resilience. Without this test, the pump's ultimate capacity may be unpredictable in the event of a sudden cooling failure during actual use, potentially leading to serious problems such as bearing overheating, mechanical seal failure, or even pump body damage. Furthermore, cooling failures are often sudden; without prior knowledge of the pump's response characteristics, it's impossible to develop appropriate countermeasures, which could result in uncontrollable system malfunctions.

[0064] Typically, the results of a water shortage test mainly include the pump's temperature rise, vibration changes, and noise level. Temperature rise is influenced by various factors, including initial water temperature, ambient temperature, bearing lubrication condition, and the sealing of the test circuit. The impact of microbubbles is particularly significant; residual bubbles in the test circuit reduce the liquid's heat transfer efficiency, causing the bearing temperature rise rate to be initially low, followed by a sharp increase. Furthermore, the collapse of microbubbles causes localized pressure fluctuations, leading to a temporary increase in vibration and generating brief high-frequency noise. If the influence of microbubbles is not fully considered, test data may show abnormal deviations, leading to misjudgments of the equipment's health status. Therefore, thorough venting is necessary before the test, and intelligent data analysis methods should be used to detect the impact of microbubbles on the measurement data.

[0065] The experiment began with preparations before the water cutoff, ensuring standardized test conditions and maintaining water temperature, pump chamber level, and air pressure within set ranges to eliminate the influence of environmental variables. Subsequently, the pump was started and its speed gradually adjusted to reach the rated flow rate. After stable operation, the cooling system was shut off, officially commencing the water cutoff test phase.

[0066] Optionally, in one embodiment, S1 includes the following steps S101-S107.

[0067] S101. Maintain the water temperature in the test circuit at 85±10℃.

[0068] Before the test begins, the water temperature in the test circuit needs to be maintained at 85±10℃. This setting is based on the physical properties of water. Water temperature affects its viscosity and heat transfer coefficient, which in turn affects the pump's flow resistance and heat dissipation capacity. If the temperature is too low, the water viscosity increases, which may lead to insufficient bearing lubrication and increased friction loss; if the temperature is too high, the water's heat dissipation capacity decreases, which may lead to an excessively high rate of temperature rise in the pump. Therefore, maintaining an appropriate water temperature ensures that the pump's thermal conductivity characteristics meet the engineering design expectations and reduces errors caused by abnormal temperature rise. Furthermore, water temperature control involves the heat exchange system, and a PID controller is typically used to adjust the heating element to maintain the water temperature within the target range.

[0069]

[0070] Where e represents the deviation between the current water temperature and the target water temperature. These are PID control parameters to ensure stable operation of the temperature within the allowable error range.

[0071] S102. The pump chamber liquid level is maintained at 990±200mm below the mounting plane of the large flange on the outer casing.

[0072] To ensure the pump operates under sufficient static pressure, it's crucial to prevent cavitation caused by excessively low liquid levels. Cavitation occurs when liquid bubbles form in low-pressure areas; these bubbles burst in high-pressure areas, releasing significant energy that can damage the impeller surface and even affect pump vibration and efficiency. Conversely, excessively high liquid levels can lead to abnormal pump inlet pressure, impacting fluid dynamics. Therefore, it's necessary to install a liquid level sensor to monitor the liquid level in real-time, and combine this with a solenoid valve-controlled water level replenishment system to ensure the liquid level remains within the set range, minimizing experimental errors caused by level fluctuations.

[0073] S103. The air pressure covering the pump chamber is maintained at 0.33±0.05Mpa.

[0074] In addition to level control, the air pressure inside the pump chamber is set at 0.33 ± 0.05 MPa. This is to ensure that the pump's internal environment operates under controlled conditions, preventing external gases from entering the pump chamber and affecting fluid characteristics. Typically, a pressure sensor is used to detect the air pressure inside the pump chamber, and it is dynamically adjusted through an airtight system to ensure that the air pressure is maintained within the set range. Appropriate pump chamber pressure can optimize the pump's sealing performance, reduce leakage, and improve operational stability. For example, in high-pressure environments, mechanical seals may withstand greater pressure, affecting their frictional properties and increasing the risk of seal failure.

[0075] S104. Adjust the frequency of the frequency converter to 20Hz and start the pump.

[0076] After setting the parameters for the test environment, the pump needs to be started and its operating status adjusted. First, adjust the inverter frequency to 20Hz and start the pump. This allows the pump to gradually enter a stable operating state at a lower speed, reducing mechanical shock caused by instantaneous high load startup. Under inverter control, the pump's acceleration process can be smoothly transitioned, thus avoiding the influence of transient changes during startup on measurement data.

[0077] S105. Adjust the outlet valve until the main circuit flow meter reaches 4888±200m³ / h, then stop adjusting the valve.

[0078] The outlet valve is gradually adjusted to achieve a main circuit flow rate of 4888±200 m³ / h. The purpose of this step is to establish a controllable liquid flow state, enabling the pump to operate at the set flow rate. The valve adjustment process is typically completed by an automatic flow control system, in which a flow sensor monitors the flow rate in real time, and the system dynamically adjusts the valve opening based on feedback signals to ensure that the flow rate remains stable within the set range.

[0079] S106. After the pump set is running stably, gradually adjust the frequency of the frequency converter to 50Hz so that the pump reaches the rated flow rate of 12220±367m³ / h.

[0080] Specifically, in this embodiment, S106 includes the following steps ab.

[0081] a. After the pump set is running stably, adjust the frequency of the frequency converter to 25Hz, 30Hz, 35Hz, 40Hz, 45Hz and 50Hz in sequence to increase the speed.

[0082] b. After the pump set is running stably, observe whether the main circuit flow rate is 12220±367m³ / h. If it exceeds this range, fine-tune the outlet valve to bring the main circuit flow rate within the range of 12220±367m³ / h.

[0083] After completing low-speed operation and flow rate adjustment, the pump unit needs to enter the rated flow operation state, which involves gradually adjusting the inverter frequency to 50Hz to achieve the rated flow rate of 12220±367m³ / h. This process is not completed all at once, but rather through incremental adjustment to reduce flow field disturbances caused by sudden changes in speed. During frequency conversion adjustment, the frequency can be increased in the following order: 25Hz→30Hz→35Hz→40Hz→45Hz→50Hz. After each increase, observe whether the pump flow rate is stable to ensure that the pump can maintain stability in each transition state.

[0084] S107. After stable operation, close the main valve of the cooling water system.

[0085] After the pump has stabilized, the main valve of the cooling water system is closed to artificially create an extreme condition of cooling failure, thereby assessing the pump's tolerance in the absence of external cooling and its temperature rise trend. This step marks the official start of the water shortage test. At this point, all test parameters (bearing temperature, mechanical seal temperature, motor winding temperature, water temperature, ambient temperature, vibration intensity, noise, etc.) will enter a free variation phase. The pump's operating state will depend entirely on its own heat transfer and heat dissipation capabilities, no longer relying on the effect of external cooling media.

[0086] Optionally, in another embodiment, S1 includes the steps S111-S117.

[0087] S111. Maintain the water temperature in the test circuit at 85±10℃.

[0088] S112. The pump chamber liquid level is maintained at 990±200mm below the mounting plane of the large flange on the outer casing.

[0089] S113. The air pressure covering the pump chamber is maintained at 0.33±0.05Mpa.

[0090] S114. Adjust the frequency of the frequency converter to 20Hz and start the pump.

[0091] S115. Adjust the outlet valve until the main circuit flow meter reaches 4888±200m³ / h, then stop adjusting the valve.

[0092] S116. After the pump set is running stably, gradually adjust the frequency of the frequency converter to 12.5Hz so that the pump reaches the rated flow rate of 3055±92m³ / h.

[0093] Specifically, in this embodiment, S116 includes the following steps, including steps ab.

[0094] a. After the pump set is running stably, adjust the frequency of the frequency converter to 12.5Hz.

[0095] b. After the pump set is running stably, observe whether the main circuit flow rate is 3055±92m³ / h. If it exceeds this range, fine-tune the outlet valve to bring the main circuit flow rate within the range of 3055±92m³ / h.

[0096] After the pump set is running stably, the frequency of the inverter is gradually adjusted to 12.5Hz to enable the pump to reach the rated flow rate of 3055±92m³ / h, ensuring stable operation of the pump at low speed and providing a reliable test environment for long-term water cut-off test (72 hours).

[0097] S117. After stable operation, close the main valve of the cooling water system.

[0098] After the pump has stabilized, the main valve of the cooling water system is closed to artificially create an extreme condition of cooling failure, thereby assessing the pump's tolerance in the absence of external cooling and its temperature rise trend. This step marks the official start of the water shortage test. At this point, all test parameters (bearing temperature, mechanical seal temperature, motor winding temperature, water temperature, ambient temperature, vibration intensity, noise, etc.) will enter a free variation phase. The pump's operating state will depend entirely on its own heat transfer and heat dissipation capabilities, no longer relying on the effect of external cooling media.

[0099] S2. Set the time window and sampling frequency, and divide the time window into two segments. The first segment of the time window is used to acquire data to calculate the predicted temperature rise trend, and the second segment of the time window is used to acquire data to obtain the actual temperature rise trend corresponding to the first segment.

[0100] In the water-cutoff test, the pump's temperature change is a time series problem, meaning the current temperature depends not only on the pump's instantaneous state but also on factors such as past temperatures, vibrations, and ambient temperatures. Therefore, a sliding time window is needed to construct the sample, enabling the model to predict future temperature trends based on historical data. After setting the time window W, the data will be split into two parts:

[0101]

[0102] in: (First half of the time window): Represents the data from the past 60 seconds, used as input to the model.

[0103] (Later part of the time window): Represents the temperature rise trend in the next 10 seconds, and is used as the true value to calculate the prediction error.

[0104] The selection of the time window needs to balance the sufficiency of historical information with the timeliness of prediction. If the window is too short, the model will be unable to capture long-term trends; if the window is too long, the computational complexity will increase, and the contribution of remote historical information may diminish. Typically, the window size is optimized through experiments, such as using 60 seconds of historical data to predict a 10-second future trend.

[0105] In this scheme, the acquisition frequency of data such as temperature rise and vibration is set to 1Hz, that is, once per second, to ensure that the data is dense enough to capture subtle changes in temperature rise.

[0106]

[0107] Each data point contains multiple measurements, such as bearing temperature, mechanical seal temperature, motor winding temperature, vibration, noise, and ambient temperature. To reduce noise interference, the data needs to be preprocessed before entering the time window, such as by moving average filtering.

[0108] Optionally, a dynamic window mechanism can also be used in this scheme. This is because the temperature rise is usually relatively gradual in the minute before the water supply is cut off, while the temperature rise rate increases sharply in the 10-20 seconds after the water supply is cut off. Therefore, .

[0109] S3. After the water supply is cut off, the bearing temperature, mechanical seal temperature and motor winding temperature are collected at the sampling frequency, and the water temperature, ambient temperature, vibration intensity and noise data are recorded simultaneously.

[0110] After a water outage, the pump's temperature changes are no longer controlled by the cooling water, but are entirely determined by internal heat conduction, lubrication conditions, and ambient environmental conditions. To capture this dynamic change, data must be sampled at a high frequency (typically 1 Hz) to ensure a sufficiently dense data point and to prevent any minute temperature fluctuations from being missed. The data acquisition system typically employs high-precision temperature sensors, such as thermocouples or platinum resistance temperature detectors (RTDs), to transmit the measurement data to the central control system in real time.

[0111] For vibration data acquisition, accelerometers are typically used. Their output signals can be analyzed using Fourier transform to determine the spectral characteristics and detect the presence of abnormal high-frequency components.

[0112] Recording noise data is highly relevant for detecting microbubbles and lubrication anomalies. For example, after water supply is cut off, the noise level will briefly increase if microbubbles are present or the lubricating oil deteriorates. If a high-frequency noise peak occurs within 1 to 2 minutes after water supply is cut off, it may be a precursor to cavitation or bearing lubrication failure.

[0113] Furthermore, data from all sensors must be sampled and stored strictly according to the same time step. Data asynchrony can lead to misalignment during time-series feature extraction, affecting the learning performance of the LSTM model. Therefore, timestamp synchronization mechanisms are typically used in data acquisition systems.

[0114] S4. Based on window sliding, the bearing temperature, mechanical seal temperature, motor winding temperature, water temperature, ambient temperature, vibration intensity and noise data obtained in the first part of a time window are fused into input features.

[0115] In the water shortage test, the pump's temperature rise is a dynamic evolution process, influenced not only by the current state but also by past temperature trends, environmental conditions, and mechanical operating status. Therefore, analyzing data from a single moment is insufficient; a sliding time window is needed to construct a time series input, enabling the model to learn the long-term dependence of the temperature rise. The time window is set as follows:

[0116]

[0117] Each time window contains data from the past 60 seconds, which is used as input features to predict the temperature rise trend in the next 10 seconds.

[0118] The key to feature fusion lies in data synchronization and feature transformation, ensuring that all variables are aligned in the time dimension while eliminating the interference of noisy data on the model. In implementation, data standardization is first required to avoid the impact of magnitude differences in different physical quantities on model training. Min-Max normalization is used to map all features to the [0,1] interval:

[0119]

[0120] Where X represents the original data. These are the minimum and maximum values ​​of the dataset, respectively. This ensures that features such as temperature, vibration, and noise are on the same numerical scale, preventing certain features from having a dominant influence on model training.

[0121] Furthermore, to improve the robustness of the prediction model, it is necessary to construct statistical features, such as calculating the mean, variance, maximum and minimum temperatures over the past 60 seconds, to enhance the feature representation capability of the time series.

[0122]

[0123] The ultimate goal of data fusion is to construct a structured input matrix for LSTM training and prediction:

[0124]

[0125] Each row represents a measurement at a specific moment, and each column represents a feature variable. This input format conforms to the requirements of LSTM, enabling it to process time series data and learn dependencies across time steps.

[0126] S5. Input the input features into the pre-trained LSTM model to obtain the predicted temperature rise trend, and calculate the deviation between the actual temperature rise and the predicted temperature rise. If the deviation is greater than the preset threshold, it indicates that the temperature rise trend is abnormal. If the deviation is less than the preset threshold, it indicates that the temperature rise trend is normal.

[0127] This step feeds the fused input features into a pre-trained LSTM model to obtain the predicted temperature rise trend and calculates the deviation between the actual and predicted temperature rise to determine if the temperature rise trend is abnormal. The data matrix generated in the previous step is input into the input layer of the LSTM model. After two LSTM layers, the fully connected layer (DenseLayer) outputs the predicted temperature rise value for the next 10 seconds.

[0128]

[0129] The predicted temperature rise trend is compared with the actual temperature rise data, and the error is calculated:

[0130]

[0131] Error calculation uses mean squared error to measure the deviation between the predicted value and the actual value:

[0132]

[0133] if If the temperature rise exceeds a preset threshold, it is considered an abnormal trend, which may indicate pump lubrication failure, bearing overheating, or microbubble interference. For example, the normal temperature rise deviation should be less than 3°C. If so, an alarm may need to be issued.

[0134] Specifically, in one embodiment, the pre-training steps of the LSTM model include S501-S506.

[0135] S501. Perform data acquisition and data preprocessing.

[0136] Specifically, in one embodiment, S501 includes the following steps S5011-S5014.

[0137] S5011. Acquire training data, wherein the training data comes from normal water cut-off test data without microbubbles under different ambient temperatures and different test conditions, abnormal water cut-off test data with microbubbles, and some test data with artificially introduced microbubbles.

[0138] The S5011 is responsible for acquiring training data, which comes from water cut-off tests under multiple different operating conditions, including normal water cut-off test data (without microbubbles), abnormal water cut-off test data (with microbubbles), and some test data with artificially introduced microbubbles. The purpose of this data is to cover the pump's operating state under various possible water cut-off conditions, ensuring that the LSTM model can learn temperature rise patterns under different conditions. The construction of the training data relies on a sensor data acquisition system. The acquired variables include bearing temperature, mechanical seal temperature, motor winding temperature, water temperature, ambient temperature, vibration intensity, and noise data, recorded at a fixed frequency (e.g., 1Hz) to form a continuous time series. For microbubble data acquisition, a high-sensitivity noise sensor is used because microbubbles may burst in the early stages of water cut-off, generating high-frequency noise, and the temporal characteristics of this noise can reflect the impact of microbubbles on the pump's operating state. After data acquisition, all data is timestamped to ensure consistency of input features.

[0139] S5012. Detect outliers and remove them.

[0140] The S5012 is responsible for detecting and removing outliers to ensure the stability of training data. Outliers may originate from sensor malfunctions, environmental interference, or abnormal fluctuations during the experiment.

[0141] S5013. Interpolation to complete missing values.

[0142] During sensor data acquisition, some data may be lost due to signal interference or equipment malfunction. Common methods for filling missing values ​​include linear interpolation and time series interpolation.

[0143] S5014. Normalize the data.

[0144] This step ensures that the numerical ranges of all variables are consistent, preventing scale differences between different physical quantities from affecting model training. Common normalization methods include Min-Max normalization and Z-score standardization.

[0145] S502. Set a time window of seventy seconds and generate a training set, a validation set, and a test set; wherein, the training set uses the data of the first sixty seconds as input and the temperature rise trend of the last ten seconds as output.

[0146] The dataset is divided according to the standard procedure of training set, validation set, and test set, usually in an 8:1:1 ratio:

[0147] Training set (80%): Used for LSTM training, enabling the model to learn the basic patterns of temperature rise trends.

[0148] Validation set (10%): Used during training to monitor the model's generalization ability and prevent overfitting.

[0149] Test set (10%): Used to finally evaluate the model's predictive ability and ensure that it can maintain high accuracy on unseen data.

[0150] Training data is typically organized using a sliding window method, which allows new training samples to be generated at each time step.

[0151] S503. Construct an LSTM network.

[0152] Specifically, in one embodiment, the structure of the LSTM network in S503 includes: an input layer, a first LSTM layer, a second LSTM layer, and a fully connected layer.

[0153] The input layer receives input features corresponding to sixty time steps; the first LSTM layer contains sixty-four units, which are used to extract time series features and output information for all time steps; the second LSTM layer contains thirty-two units, which are used to extract deep time series information and output information for only the last time step; the fully connected layer is used to map the features extracted by the LSTM to a numerical value to predict the temperature rise trend in the next ten seconds.

[0154] The input layer is responsible for receiving sensor data from the past 60 seconds. The data format is as follows:

[0155]

[0156] Each row represents sensor data at a specific time step, and each column represents a feature variable, such as bearing temperature, mechanical seal temperature, or motor winding temperature. The function of this input layer is to format the data into a time-series structure that can be processed by an LSTM, enabling the model to establish data correlations across multiple time steps.

[0157] The input data enters the first LSTM layer, which contains 64 LSTM units and is capable of extracting complex temporal features in the time dimension. LSTM uses memory gates and gating mechanisms to control the flow of information, overcoming the gradient vanishing problem that traditional recurrent neural networks are prone to when processing long-running sequences. The first LSTM layer uses `return_sequences=True`, meaning it outputs a 64-dimensional feature vector at each time step, forming a complete time-series feature matrix.

[0158]

[0159] in, This represents the i-th eigenvalue calculated by the LSTM at time step t. The function of this layer is to establish patterns within the time series, such as periodic changes in temperature rise and abrupt changes.

[0160] The data then enters the second LSTM layer, which contains 32 LSTM units to further extract deeper features from the time series. Unlike the first LSTM layer, this layer has `return_sequences=False`, meaning it only outputs information from the last time step.

[0161]

[0162] This indicates that the second LSTM layer has already highly compressed the information of the entire time series, retaining only the most representative feature values ​​as the final output. The main function of this layer is to further filter and refine the time series features with the most predictive value.

[0163] Finally, the time-series features extracted by LSTM are fed into a fully connected layer. This layer maps the 32-dimensional features calculated by LSTM to the final temperature rise prediction value.

[0164]

[0165] The number of neurons in the fully connected layer determines the shape of the predicted output. For this task, only one value needs to be output, namely the predicted temperature rise trend for the next 10 seconds. Therefore, the number of neurons in the fully connected layer is 1. The calculation method for this layer can be expressed as follows:

[0166]

[0167] Where W is the weight matrix of the fully connected layer, b is the bias term, and the final result is... This is the predicted trend of rising temperatures.

[0168] S504. Input the training set into the LSTM network and calculate the loss function.

[0169] Preprocessed time-series data is fed into an LSTM model. Each sample contains sensor data from the past 60 seconds as input, and the target output is the temperature rise trend for the next 10 seconds. The forward propagation process of the LSTM involves a series of matrix operations, with each time step including input gate, forget gate, output gate, and cell state update.

[0170] In this step, a loss function is calculated to measure the difference between the LSTM prediction and the true value, and the weight parameters of the LSTM layer are adjusted based on this error to minimize the prediction error. In this embodiment, the loss function is used as the mean squared error.

[0171] S505. Calculate the gradient of the loss with respect to the weights of the LSTM layer, and use the Adam optimizer to perform gradient descent to update the weights of the LSTM layer.

[0172] After calculating the loss, LSTM uses backpropagation to calculate the gradient of the loss with respect to the network parameters, and then updates the weights using the Adam optimizer.

[0173] S506. After the model training is complete, use the test set and calculate the test set error, and adjust the model complexity based on the test set error.

[0174] This step is responsible for model testing and optimization. It involves evaluating the predictive performance of the LSTM using unseen test data and adjusting the model's complexity based on the test errors to ensure high predictive accuracy under various water shortage test conditions. During testing, the MSE on the test set is calculated and compared with the training error.

[0175]

[0176] If the test error is much larger than the training error, it indicates that the model may be overfitting and the LSTM structure needs to be adjusted, such as reducing the number of LSTM layers, increasing regularization, or adopting an early stopping strategy.

[0177] After the test is completed, the final model is saved and used to predict the temperature rise trend in real time during water outage tests. In practical applications, the model receives new sensor data and generates temperature rise predictions for the next 10 seconds. This mechanism enables the LSTM model to detect abnormal temperature rise trends in real time during water outage tests and issue early warnings to prevent equipment damage.

[0178] This application also discloses a high-flow-rate vertical centrifugal pump water medium testing system, including a processor, wherein the processor runs a program for the high-flow-rate vertical centrifugal pump water medium testing method described in any one of the above embodiments.

[0179] This application also discloses a storage medium storing a program for a high-flow-rate vertical centrifugal pump water medium test method as described in any one of the above embodiments.

[0180] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. A method of testing a large flow vertical centrifugal pump water medium, characterized by, Comprising the following steps: S1. Perform the water cut preparation of the large-flow vertical centrifugal pump; S2. Set the time window and sampling frequency, and divide the time window into two segments, wherein the front segment of the time window is used to obtain data to calculate the predicted temperature rise trend, and the rear segment of the time window is used to obtain data to obtain the actual temperature rise trend corresponding to the front segment; S3. Collect the bearing temperature, mechanical seal temperature and motor winding temperature according to the sampling frequency after the water is cut off, and synchronously record the water temperature, environmental temperature, vibration intensity and noise data; S4. Based on window sliding, the bearing temperature, mechanical seal temperature, motor winding temperature, water temperature, environmental temperature, vibration intensity and noise data obtained in the front segment of a time window are fused into input features; S5. The input features are input into a pre-trained LSTM model to obtain a predicted temperature rise trend, and the deviation between the actual temperature rise and the predicted temperature rise is calculated, if the deviation is greater than a preset threshold, it indicates that the temperature rise trend is abnormal, and if the deviation is less than the preset threshold, it indicates that the temperature rise trend is normal; The S1 comprises the following steps: S101. Keep the test circuit water temperature at 85±10℃; S102. Keep the pump cavity liquid level below the large flange installation plane of the outer shell by 990±200mm; S103. Keep the pump cavity covering air pressure at 0.33±0.05MPa; S104. Adjust the frequency converter frequency to 20Hz, and start the pump; S105. Adjust the outlet valve, and the main circuit flow reaches 4888±200m³ / h, stop adjusting the valve; S106. After the pump set runs stably, gradually adjust the frequency converter frequency to 50Hz to make the pump reach the rated flow of 12220±367m³ / h; S107. After stable operation, close the cooling water system total valve; Or, the S1 comprises the following steps: S111. Keep the test circuit water temperature at 85±10℃; S112. Keep the pump cavity liquid level below the large flange installation plane of the outer shell by 990±200mm; S113. Keep the pump cavity covering air pressure at 0.33±0.05MPa; S114. Adjust the frequency converter frequency to 20Hz, and start the pump; S115. Adjust the outlet valve, and the main circuit flow reaches 4888±200m³ / h, stop adjusting the valve; S116. After the pump set runs stably, gradually adjust the frequency converter frequency to 12.5Hz to make the pump reach the rated flow of 3055±92m³ / h; S117. After stable operation, close the cooling water system total valve.

2. The water medium test method for large flow vertical centrifugal pump according to claim 1, characterized by, The S106 comprises the following steps: After the pump set runs stably, adjust the frequency converter frequency to 25Hz, 30Hz, 35Hz, 40Hz, 45Hz and 50Hz in sequence to increase the speed; After the pump set runs stably, observe whether the main circuit flow is 12220±367m³ / h, if it exceeds, slightly adjust the outlet valve to make the main circuit flow reach the range of 12220±367m³ / h.

3. The water medium test method for large flow vertical centrifugal pumps according to claim 1, characterized by, The S116 comprises the following steps: After the pump set runs stably, adjust the frequency converter frequency to 12.5Hz; After the pump set is stable, observe whether the main circuit flow is 3055±92 m³ / h, if not, adjust the outlet valve to make the main circuit flow reach 3055±92 m³ / h.

4. The water medium test method for a large flow vertical centrifugal pump according to any one of claims 1 to 3, characterized by, The pre-training step of the LSTM model comprises: S501. Data acquisition and data preprocessing are performed; S502. Set seventy seconds as a time window, and generate a training set, a validation set and a test set; wherein the training set takes the data of the previous sixty seconds as input, and the temperature rise trend of the following ten seconds as output; S503. Construct an LSTM network; S504. Input the training set into the LSTM network, and calculate the loss function; S505. Calculate the gradient of the loss to the LSTM layer weight, and use the Adam optimizer to perform gradient descent to update the weight of the LSTM layer; S506. After the model training is completed, use the test set and calculate the test set error, and adjust the model complexity based on the test set error.

5. The water medium test method for large flow vertical centrifugal pump according to claim 4, characterized by, The structure of the LSTM network in S503 comprises: An input layer for receiving input features corresponding to sixty time steps; A first layer LSTM containing sixty-four units for extracting time series features and outputting information of all time steps; A second layer LSTM containing thirty-two units for extracting deep time series information and outputting information of only the last time step; A fully connected layer for mapping the features extracted by the LSTM to a numerical value to predict the temperature rise trend in the next ten seconds.

6. The water medium test method for large flow vertical centrifugal pump according to claim 5, characterized by, S501 comprises the following steps: S5011. Obtain training data, wherein the training data comes from normal water break experiment data without microbubbles, abnormal water break experiment data with microbubbles and test data with part of microbubbles introduced artificially under different environmental temperatures and different test conditions; S5012. Detect abnormal values and eliminate abnormal points; S5013. Interpolate to complete missing values; S5014. Normalize the data.

7. A large flow rate vertical centrifugal pump water medium test system characterized by, A processor, wherein the processor runs a program of the water medium test method of the large-flow vertical centrifugal pump according to any one of claims 1-6.

8. A storage medium, characterized by A storage device storing a program of the water medium test method of the large-flow vertical centrifugal pump according to any one of claims 1-6.

Images