Method and device for monitoring LNG pump tower system, computer device and storage medium

Abstract

The present disclosure provides a kind of monitoring method, device, computer equipment and storage medium of LNG pump tower system.The method comprises: obtaining the pressure sensing data of LNG pump tower system, and obtaining the vibration sensing data of LNG pump tower system;From the pressure sensing data, extract pressure feature data, and extract vibration feature data from vibration sensing data;Characteristic data matrix formed by pressure feature data and vibration feature data is input into intelligent identification module, feature extraction is carried out on characteristic data matrix using convolutional neural network in intelligent identification module, target feature data is obtained, and at least two long short-term memory networks in intelligent identification module are used to carry out prediction processing on target feature data, at least two long short-term machine networks respectively corresponding prediction processing result is obtained, and the prediction processing result corresponding to each long short-term memory network is input into the classifier corresponding to each long short-term memory network, to obtain the running state and vibration state of LNG pump tower system.

Description

Technical Field

[0001] This disclosure relates to the field of monitoring technology, specifically to a monitoring method, device, computer equipment, and storage medium for an LNG pump tower system. Background Technology

[0002] Liquefied Natural Gas (LNG) pump towers are key components used in LNG storage and transportation equipment. They are typically suspended inside LNG tanks and are primarily used to perform cryogenic transportation of LNG liquid while also being able to withstand occasional severe conditions such as earthquakes.

[0003] LNG pump towers typically consist of a piping system, a pipe support structure, guide connectors, ladders, and a pump tower base. The piping system includes a pump liquid pipe, an inlet liquid pipe, and a measuring liquid pipe, arranged vertically in a parallel, triangular pattern. The pump liquid pipe transports LNG, the inlet liquid pipe injects LNG into the tank, and the measuring liquid pipe measures parameters such as the LNG level in the tank. The condition of the piping system is crucial to the safety of the LNG pump tower. Therefore, monitoring the condition of the LNG pump tower is necessary to ensure its safe operation during use.

[0004] Current methods for inspecting LNG pump towers typically involve adding pressure sensing modules inside the pipelines. These modules detect changes in pressure levels within the pipelines, and when the pressure is too high or too low, the LNG pump tower's operational status may be compromised. However, current monitoring methods cannot pinpoint the specific type of problem, leading to the need for manual intervention and resulting in delays in problem resolution. Summary of the Invention

[0005] In view of this, the present disclosure provides a monitoring method, device, computer equipment and storage medium for an LNG pump tower system.

[0006] Specifically, this disclosure is achieved through the following technical solution:

[0007] In a first aspect, embodiments of this disclosure provide a monitoring method for a liquefied natural gas (LNG) pump tower system, the method comprising:

[0008] The system acquires pressure sensing data from pressure sensing modules deployed at multiple pipeline locations in the LNG pump tower system, and acquires vibration sensing data from vibration sensing modules deployed at multiple target vibration locations in the LNG pump tower system.

[0009] Pressure feature data is extracted from the pressure sensing data, and vibration feature data is extracted from the vibration sensing data;

[0010] The feature data matrix composed of the pressure feature data and the vibration feature data is input into the intelligent recognition module. The convolutional neural network in the intelligent recognition module is used to extract features from the feature data matrix to obtain target feature data. At least two long short-term memory networks in the intelligent recognition module are used to predict the target feature data to obtain the prediction results corresponding to the at least two long short-term machine networks. The prediction results corresponding to each long short-term memory network are input into the classifier corresponding to each long short-term memory network to obtain the operating status and vibration status of the LNG pump tower system.

[0011] Optionally, the plurality of pipe locations include at least one of the following: pump inlet location, pump outlet location, pipe bend location, and flange connection location; the pressure sensing module includes at least one of the following: a pressure sensing module installed at the pump inlet location for detecting instantaneous pressure changes at the pump inlet.

[0012] A differential pressure sensor, installed at the pump outlet, is used to detect the pressure difference between the pump outlet and the ambient environment.

[0013] And pressure sensing modules installed at the bends and flange connections of the pipeline for detecting the pressure distribution information inside the pipeline.

[0014] Optionally, the target vibration location can be determined using the following method:

[0015] Structural modal analysis was performed on the LNG pump tower system to obtain its modal parameters; and the target vibration location was determined based on the results of the structural modal analysis.

[0016] Optionally, extracting pressure feature data from the pressure sensing data includes:

[0017] The pressure sensing data is bandpass filtered to obtain filtered pressure data;

[0018] Perform a Fast Fourier Transform on the filtered pressure data to obtain the main peak of the pressure spectrum; and,

[0019] Extract the peak-to-peak pressure values ​​of the filtered pressure data within a fixed time window;

[0020] The pressure characteristic data is constructed based on the main peak of the pressure spectrum and the pressure peak-to-peak value.

[0021] Optionally, extracting vibration feature data from the vibration sensing data includes:

[0022] The vibration sensing data is bandpass filtered to obtain filtered vibration data;

[0023] Perform a fast Fourier transform on the filtered vibration data to obtain the main peak of the vibration spectrum;

[0024] Envelope analysis is performed on the filtered vibration data, and envelope energy data is calculated; and,

[0025] Extract the peak-to-peak value of the filtered vibration data within a fixed time window;

[0026] The vibration characteristic data are constructed based on the main peak of the vibration spectrum, the envelope energy data, and the vibration peak-to-peak value.

[0027] Optionally, the pressure feature data includes: pressure feature vectors corresponding to multiple time steps; the vibration feature data includes: feature vectors corresponding to multiple time steps.

[0028] The pressure characteristic data and the vibration characteristic data are used to construct the characteristic data matrix in the following manner:

[0029] For each of the multiple time steps, the pressure feature vector and vibration feature vector corresponding to each time step are concatenated to obtain the concatenated vector corresponding to each time step; and,

[0030] The concatenated vectors corresponding to multiple time steps are superimposed to obtain the feature data matrix; wherein, the rows of the feature data matrix represent the time step dimension and the columns represent the feature dimension.

[0031] Optionally, the step of using the convolutional neural network in the intelligent recognition module to extract features from the feature data matrix to obtain target feature data includes:

[0032] Using a one-dimensional convolution kernel, the feature data is convolved in the time step dimension to obtain a first intermediate feature data matrix; and using a two-dimensional convolution kernel, the first intermediate feature data matrix is ​​convolved in both the time step dimension and the feature dimension to obtain a second intermediate feature data matrix.

[0033] The second intermediate feature data matrix is ​​activated and pooled sequentially to obtain the target feature data.

[0034] Optionally, the at least two long short-term memory networks include: a first long short-term memory network for predicting the operating status of the LNG pump tower system, and a second long short-term memory network for predicting the vibration status of the LNG pump tower system;

[0035] The step of using at least two long short-term memory networks in the intelligent recognition module to predict the target feature data and obtain the prediction results corresponding to the at least two long short-term machine networks includes:

[0036] The target feature data is reshaped or flattened to obtain the target feature data to be processed.

[0037] The target feature data to be processed is input into the first long short-term memory network to obtain the prediction processing result of the first long short-term memory network for predicting the running state of the target feature data to be processed; and...

[0038] The target feature data to be processed is input into the second long short-term memory network to obtain the prediction processing result of the second long short-term memory network for the vibration state prediction of the target feature data to be processed.

[0039] Optionally, extracting pressure feature data from the pressure sensing data and extracting vibration feature data from the vibration sensing data includes:

[0040] The pressure sensing data is denoised and filtered to obtain denoised and filtered pressure sensing data; the pressure feature data is extracted from the denoised and filtered pressure sensing data.

[0041] The vibration sensing data is denoised and filtered to obtain denoised and filtered vibration sensing data; the pressure feature data is extracted from the denoised and filtered pressure sensing data.

[0042] Optionally, the method further includes:

[0043] When the operating status indicates that the LNG pump tower is operating abnormally or malfunctioning, and / or when the LNG pump tower is operating abnormally or malfunctioning under the vibration condition, an alarm process is executed.

[0044] Secondly, embodiments of this disclosure also provide a monitoring device for a liquefied natural gas (LNG) pump tower system, comprising:

[0045] The acquisition module is used to acquire pressure sensing data from pressure sensing modules deployed at multiple pipeline locations in the LNG pump tower system, and to acquire vibration sensing data from vibration sensing modules deployed at multiple target vibration locations in the LNG pump tower system.

[0046] The extraction module is used to extract pressure feature data from the pressure sensing data and vibration feature data from the vibration sensing data.

[0047] The prediction module is used to input the feature data matrix composed of the pressure feature data and the vibration feature data into the intelligent recognition module, use the convolutional neural network in the intelligent recognition module to extract features from the feature data matrix to obtain target feature data, and use at least two long short-term memory networks in the intelligent recognition module to perform prediction processing on the target feature data to obtain the prediction processing results corresponding to the at least two long short-term memory networks respectively, and input the prediction processing results corresponding to each long short-term memory network into the classifier corresponding to each long short-term memory network respectively to obtain the operating status and vibration status of the LNG pump tower system.

[0048] Optionally, the plurality of pipe locations include at least one of the following: pump inlet location, pump outlet location, pipe bend location, and flange connection location; the pressure sensing module includes at least one of the following: a pressure sensing module installed at the pump inlet location for detecting instantaneous pressure changes at the pump inlet.

[0049] A differential pressure sensor, installed at the pump outlet, is used to detect the pressure difference between the pump outlet and the ambient environment.

[0050] And pressure sensing modules installed at the bends and flange connections of the pipeline for detecting the pressure distribution information inside the pipeline.

[0051] Optionally, the target vibration location can be determined using the following method:

[0052] Structural modal analysis was performed on the LNG pump tower system to obtain its modal parameters; and the target vibration location was determined based on the results of the structural modal analysis.

[0053] Optionally, the extraction module 42, when extracting pressure feature data from the pressure sensing data, is used to:

[0054] The pressure sensing data is bandpass filtered to obtain filtered pressure data;

[0055] Perform a Fast Fourier Transform on the filtered pressure data to obtain the main peak of the pressure spectrum; and,

[0056] Extract the peak-to-peak pressure values ​​of the filtered pressure data within a fixed time window;

[0057] The pressure characteristic data is constructed based on the main peak of the pressure spectrum and the pressure peak-to-peak value.

[0058] Optionally, the extraction module, when extracting vibration feature data from the vibration sensing data, is used to:

[0059] The vibration sensing data is bandpass filtered to obtain filtered vibration data;

[0060] Perform a fast Fourier transform on the filtered vibration data to obtain the main peak of the vibration spectrum;

[0061] Envelope analysis is performed on the filtered vibration data, and envelope energy data is calculated; and,

[0062] Extract the peak-to-peak value of the filtered vibration data within a fixed time window;

[0063] The vibration characteristic data are constructed based on the main peak of the vibration spectrum, the envelope energy data, and the vibration peak-to-peak value.

[0064] Optionally, the pressure feature data includes: pressure feature vectors corresponding to multiple time steps; the vibration feature data includes: feature vectors corresponding to multiple time steps.

[0065] The prediction module constructs the feature data matrix from the pressure feature data and the vibration feature data in the following manner:

[0066] For each of the multiple time steps, the pressure feature vector and vibration feature vector corresponding to each time step are concatenated to obtain the concatenated vector corresponding to each time step; and,

[0067] The concatenated vectors corresponding to multiple time steps are superimposed to obtain the feature data matrix; wherein, the rows of the feature data matrix represent the time step dimension and the columns represent the feature dimension.

[0068] Optionally, when the prediction module uses the convolutional neural network in the intelligent recognition module to extract features from the feature data matrix to obtain target feature data, it is used to:

[0069] Using a one-dimensional convolution kernel, the feature data is convolved in the time step dimension to obtain a first intermediate feature data matrix; and using a two-dimensional convolution kernel, the first intermediate feature data matrix is ​​convolved in both the time step dimension and the feature dimension to obtain a second intermediate feature data matrix.

[0070] The second intermediate feature data matrix is ​​activated and pooled sequentially to obtain the target feature data.

[0071] Optionally, the at least two long short-term memory networks include: a first long short-term memory network for predicting the operating status of the LNG pump tower system, and a second long short-term memory network for predicting the vibration status of the LNG pump tower system;

[0072] The prediction module, when using at least two long short-term memory networks in the intelligent recognition module to perform prediction processing on the target feature data and obtain the prediction processing results corresponding to the at least two long short-term machine networks respectively, is used for:

[0073] The target feature data is reshaped or flattened to obtain the target feature data to be processed.

[0074] The target feature data to be processed is input into the first long short-term memory network to obtain the prediction processing result of the first long short-term memory network for predicting the running state of the target feature data to be processed; and...

[0075] The target feature data to be processed is input into the second long short-term memory network to obtain the prediction processing result of the second long short-term memory network for the vibration state prediction of the target feature data to be processed.

[0076] Optionally, the extraction module, when extracting pressure feature data from the pressure sensing data and vibration feature data from the vibration sensing data, is used to:

[0077] The pressure sensing data is denoised and filtered to obtain denoised and filtered pressure sensing data; the pressure feature data is extracted from the denoised and filtered pressure sensing data.

[0078] The vibration sensing data is denoised and filtered to obtain denoised and filtered vibration sensing data; the pressure feature data is extracted from the denoised and filtered pressure sensing data.

[0079] Optionally, it also includes an alarm module, used to perform alarm processing when the operating status indicates that the LNG pump tower is operating abnormally or malfunctioning, and / or when the LNG pump tower is operating abnormally or malfunctioning under the vibration state.

[0080] Thirdly, an optional implementation of this disclosure also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the first aspect above, or any possible implementation of the first aspect.

[0081] Fourthly, an optional implementation of this disclosure also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the first aspect or any possible implementation of the first aspect.

[0082] Fifthly, an optional implementation of this disclosure also provides a computer program product carrying program code, the program code including instructions that can be used to perform steps as described in the first aspect or any possible implementation of the first aspect.

[0083] The LNG pump tower system monitoring method provided in this disclosure involves installing pressure sensing modules at multiple pipeline locations and vibration sensing modules at multiple target vibration locations within the LNG pump tower system. Pressure sensing data is acquired using the pressure sensing modules, and vibration sensing data is acquired using the vibration sensing modules. Pressure feature data is extracted from the pressure sensing data, and vibration feature data is extracted from the vibration sensing data. The pressure feature data and vibration feature data are then combined to form a feature data matrix, which is input into an intelligent recognition module. A convolutional neural network in the intelligent recognition module is used to extract features from the feature data to obtain target feature data. At least two long short-term memory networks in the intelligent recognition module are used to predict the target feature data, obtaining prediction results. Classifiers corresponding to different long short-term memory networks are used to classify the prediction results, thus obtaining the operating status and vibration status of the LNG pump tower system. By combining deep learning with real-time acquired pressure and vibration sensing data from the LNG pump tower system, the specific operating status and vibration status of the LNG pump tower system can be accurately located, enabling targeted and more efficient handling when a fault occurs in the LNG pump tower system.

[0084] To make the above-mentioned objects, features and advantages of this disclosure more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0085] Figure 1 This is a flowchart illustrating an exemplary embodiment of an LNG pump tower system detection method;

[0086] Figure 2 This is a specific example of an intelligent recognition module shown in an exemplary embodiment of the present disclosure;

[0087] Figure 3 This is a schematic diagram of a computer device illustrated in an exemplary embodiment of the present disclosure;

[0088] Figure 4 This is a schematic diagram of a testing device for an LNG pump tower system, as illustrated in an exemplary embodiment of this disclosure. Detailed Implementation

[0089] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this disclosure as detailed in the appended claims.

[0090] The terminology used in this disclosure is for the purpose of describing particular embodiments only and is not intended to be limiting of the disclosure. The singular forms “a,” “the,” and “the” as used in this disclosure and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any and all possible combinations of one or more of the associated listed items.

[0091] It should be understood that although the terms first, second, third, etc., may be used in this disclosure to describe various information, such information should not be limited to these terms. These terms are used only to distinguish information of the same type from one another. For example, without departing from the scope of this disclosure, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to determination."

[0092] This disclosure provides a monitoring method for an LNG pump tower system. Pressure sensing modules are installed at multiple pipeline locations within the LNG pump tower system, and vibration sensing modules are installed at multiple target vibration locations. Pressure sensing data is acquired using the pressure sensing modules, and vibration sensing data is acquired using the vibration sensing modules. Pressure feature data is extracted from the pressure sensing data, and vibration feature data is extracted from the vibration sensing data. The pressure and vibration feature data are then combined to form a feature data matrix, which is input into an intelligent recognition module. A convolutional neural network in the intelligent recognition module extracts features from the feature data to obtain target feature data. At least two long short-term memory networks in the intelligent recognition module perform predictive processing on the target feature data to obtain prediction results. Classifiers corresponding to different long short-term memory networks are used to classify the prediction results, thereby obtaining the operating status and vibration status of the LNG pump tower system. By combining deep learning with real-time acquired pressure and vibration sensing data from the LNG pump tower system, the specific operating status and vibration status of the LNG pump tower system can be accurately located, enabling targeted and more efficient handling when a fault occurs in the LNG pump tower system.

[0093] The shortcomings of the above solutions are the result of the inventor's practical experience and careful research. Therefore, the discovery process of the above problems and the solutions proposed in this disclosure below should be considered as the inventor's contribution to this disclosure.

[0094] To facilitate understanding of this embodiment, a detailed description of the monitoring method for an LNG pump tower system disclosed in this disclosure is provided first. The executor of the monitoring method for the LNG pump tower system provided in this disclosure is generally a computer device with a certain computing capability. This computer device may include, for example, a terminal device, a server, or other processing equipment. The terminal device may be user equipment (UE), a mobile device, a user terminal, a terminal, a vehicle-mounted device, a wearable device, etc. In some possible implementations, the monitoring method for the LNG pump tower system can be implemented by a processor calling computer-readable instructions stored in memory.

[0095] The monitoring method for the LNG pump tower system provided in the embodiments of this disclosure will be described below.

[0096] See Figure 1 The diagram shows a flowchart of a monitoring method for an LNG pump tower system provided in this embodiment of the present disclosure. The method includes steps S101 to S103, wherein:

[0097] S101: Obtain pressure sensing data from pressure sensing modules deployed at multiple pipeline locations in the LNG pump tower system, and obtain vibration sensing data from vibration sensing modules deployed at multiple target vibration locations in the LNG pump tower system.

[0098] S102: Extract pressure feature data from the pressure sensing data, and extract vibration feature data from the vibration sensing data;

[0099] S103: Input the feature data matrix composed of the pressure feature data and the vibration feature data into the intelligent recognition module, use the convolutional neural network in the intelligent recognition module to extract features from the feature data matrix to obtain target feature data, and use at least two long short-term memory networks in the intelligent recognition module to perform prediction processing on the target feature data to obtain the prediction processing results corresponding to the at least two long short-term machine networks respectively, and input the prediction processing results corresponding to each long short-term memory network into the classifier corresponding to each long short-term memory network respectively to obtain the operating status and vibration status of the LNG pump tower system.

[0100] The following provides a detailed explanation of S101 to S103.

[0101] Regarding the above S101:

[0102] In practice, pressure sensing modules are typically deployed inside the pipes of a piping system. Multiple key locations can be identified inside the pipes, and a corresponding pressure sensing module can be deployed at each of these key locations.

[0103] Specifically, the multiple pipe locations may include at least one of the following locations:

[0104] The pump inlet location, pump outlet location, pipe bend location, and flange connection location. Among these:

[0105] a1: For example, a pressure sensing module for detecting instantaneous pressure changes at the pump inlet can be installed at the pump inlet location. Specifically, this pressure sensing module may include, for example, a piezoelectric pulsating pressure sensing module. Piezoelectric pulsating pressure sensing modules have excellent dynamic response characteristics, high sensitivity, small size, wide measurement range, good stability and reliability, and do not require external power supply. Therefore, they can accurately detect instantaneous pressure changes at the pump inlet without affecting the LNG delivery at the pump inlet. Furthermore, this pressure sensing module can also be at least one of the following: strain gauge pressure sensing modules, resistive pressure sensing modules, capacitive pressure sensing modules, resonant pressure sensing modules, etc. Specific embodiments disclosed herein are not limited.

[0106] a2: For example, a differential pressure sensor can be installed at the pump outlet location to detect the pressure difference between the pump outlet and the environment. This differential pressure sensor can be, for example, a micro-electro-mechanical systems (MEMS) differential pressure sensor. MEMS differential pressure sensors are characterized by miniaturization, low power consumption, high accuracy, high stability, fast response, and wide range applicability, thus enabling accurate detection of the pressure difference between the pump outlet and the environment. Alternatively, the differential pressure sensor can also be at least one of a resistive differential pressure sensor, a capacitive differential pressure sensor, or a fiber optic differential pressure sensor; specific embodiments of this disclosure are not limited to these types.

[0107] a3: At pipe bends and flange connections, pressure sensing modules for detecting internal pressure distribution information can be installed, for example. Since the internal pressure of the pipe is usually high, the pressure sensing module can be, for example, a high-pressure resistant static pressure sensing module, which can be at least one of the following: strain gauge high-pressure static pressure sensing module, piezoelectric high-pressure sensing module, capacitive high-pressure differential pressure sensor, resonant high-pressure sensing module, etc. Specific embodiments disclosed herein are not limited to this.

[0108] Multiple target vibration locations in the LNG pump tower system, such as locations within the LNG pump tower system where vibration is more pronounced, can be determined using methods such as the following:

[0109] Structural modal analysis was performed on the LNG pump tower system to obtain its modal parameters; and the target vibration location was determined based on the results of the structural modal analysis.

[0110] Specifically, structural modal analysis of an LNG pump tower system can determine its natural frequencies, mode shapes, and damping ratios. The mode shape describes the vibration pattern of the LNG pump tower system at a specific natural frequency, including the amplitude, direction, and phase relationship of each component. For example, a mode shape might show the largest amplitude in the top pipe and a smaller amplitude in the bottom support structure, indicating that the top pipe is a "vibration-sensitive area" in this mode. In this case, the top pipe can be identified as the target vibration location. Based on structural modal analysis of the LNG pump tower system, multiple vibration-sensitive target vibration locations can be identified within the system. Vibration sensors can then be installed at these locations to acquire vibration sensing data corresponding to each target vibration location.

[0111] The vibration sensor may include, for example, a high-precision triaxial accelerometer. The sensor used has a data acquisition range of 0.1–2000 Hz, a sensitivity of 10 mV / g, and integrates an electronic piezoelectric accelerometer for comprehensive monitoring of the longitudinal, lateral, and torsional vibration components of the structure.

[0112] Regarding S102 above:

[0113] Pressure feature data is extracted from the pressure sensing data. For example, from the pressure sensing data acquired for a specific target time period T, feature data that reflects the internal pressure of the pipeline system of the LNG pump tower during that target time period is extracted.

[0114] Vibration feature data is extracted from the vibration sensing data. For example, from the vibration sensing data acquired for a specific target time period T, feature data that can reflect the vibration status of various parts of the LNG pump tower during that target time period is extracted.

[0115] Specifically, embodiments of this disclosure provide a method for extracting pressure feature data from the pressure sensing data, including:

[0116] The pressure sensing data is bandpass filtered to obtain filtered pressure data;

[0117] Perform a Fast Fourier Transform on the filtered pressure data to obtain the main peak of the pressure spectrum; and,

[0118] Extract the peak-to-peak pressure values ​​of the filtered pressure data within a fixed time window;

[0119] The pressure characteristic data is constructed based on the main peak of the pressure spectrum and the pressure peak-to-peak value.

[0120] In practical implementation, bandpass filtering of pressure-sensing data is a targeted frequency selection process. It accurately filters and retains pressure signals within a specific frequency range relevant to the analysis target, while suppressing interference noise outside this range. Bandpass filtering techniques such as Butterworth, Chebyshev, and Bessel bandpass filtering can be used to apply the bandpass filter to the pressure-sensing data.

[0121] Taking Butterworth bandpass filtering of pressure sensing data as an example, a fifth-order Butterworth bandpass filter can be applied. In an LNG pump tower system, the pump's operating frequency (e.g., 50Hz) and the characteristic frequency of pipeline fluid pulsation (e.g., 100-300Hz) are the key "effective signals" that need to be analyzed. High-frequency components above 300Hz may be irrelevant vibration noise, and low-frequency components below 10Hz may be environmental interference (e.g., foundation settlement). Fifth-order Butterworth bandpass filtering, by setting a specific "passband range" (e.g., 50-300Hz), retains only the frequency components within this range, thoroughly filtering out interference outside the passband and focusing on pressure fluctuation signals relevant to the analysis target. This reduces the proportion of invalid data or data that does not clearly reflect pressure characteristics in the obtained filtered pressure data, thereby highlighting the pressure characteristics included in the filtered pressure data.

[0122] Subsequently, a Fast Fourier Transform (FFT) can be performed on the filtered pressure data to convert it from a time-domain signal to a frequency-domain signal, thereby quantifying the frequency characteristics of the filtered pressure data. The physical meaning of the pressure signal, such as the fundamental frequency of LNG pump tower operation, the harmonic frequency of fluid pulsation, and the characteristic frequency of equipment failure, is often hidden in the frequency characteristics. For example, a normally operating pump may have a main peak (fundamental frequency) at 50Hz, while impeller wear may cause an abnormal peak at 250Hz (5th harmonic). The filtered pressure data after bandpass filtering is essentially a time-domain signal. The core of performing an FFT on it is to convert the time-domain signal into a frequency-domain signal, to convert the time-domain signal into a clear peak in the frequency domain, to establish the correspondence between frequency and physical phenomena, and then to accurately quantify these characteristics through the peak position (frequency) and height (energy) of the amplitude spectrum.

[0123] In addition, the peak-to-peak pressure values ​​within a fixed time window are extracted from the filtered pressure data. The peak-to-peak pressure value refers to the difference between the maximum and minimum pressure values ​​within the fixed time window. The peak-to-peak value directly reflects the maximum pressure fluctuation amplitude within that time period. For example, in an LNG pump tower, a sudden increase in the peak-to-peak pressure value within a certain time window may indicate a momentary impact on the pipeline. Obtaining the peak-to-peak pressure value within a fixed time window allows the target time period corresponding to the pressure sensing data to be divided into multiple local time segments. By calculating the peak-to-peak value of each segment, the temporal distribution characteristics of pressure fluctuations within the target time period corresponding to the pressure sensing data can be reflected. For example, if the peak-to-peak value remains consistently high within the target time period, there may be instability in the LNG pump tower system.

[0124] The main peak of the pressure spectrum and the peak-to-peak value of the pressure spectrum are used to form pressure characteristic data.

[0125] This disclosure also provides a specific method for extracting vibration feature data from the vibration sensing data, including:

[0126] The vibration sensing data is bandpass filtered to obtain filtered vibration data;

[0127] Perform a fast Fourier transform on the filtered vibration data to obtain the main peak of the vibration spectrum;

[0128] Envelope analysis is performed on the filtered vibration data, and envelope energy data is calculated; and,

[0129] Extract the peak-to-peak value of the filtered vibration data within a fixed time window;

[0130] The vibration characteristic data are constructed based on the main peak of the vibration spectrum, the envelope energy data, and the vibration peak-to-peak value.

[0131] In practical implementation, bandpass filtering of vibration sensing data is a targeted frequency screening step that accurately filters and retains vibration signals within a specific frequency range relevant to the analysis target, while suppressing interference noise outside this range. Bandpass filtering of vibration sensing data can be performed using methods such as Butterworth bandpass filtering, Chebyshev bandpass filtering, and Bessel bandpass filtering.

[0132] Taking Butterworth bandpass filtering of vibration sensing data as an example, a fifth-order Butterworth bandpass filter can also be applied to the vibration sensing data. Compared with lower-order filters (such as 2nd-3rd order), the transition between the passband and stopband of the fifth-order filter is steeper (the vibration signal decays faster), which can more thoroughly separate the target signal within the passband from the interference outside the passband, reduce signal aliasing in the "transition band", and enable the filtered vibration signal to better reflect the vibration characteristics of the LNG pump tower system.

[0133] Then, an FFT transformation can be performed on the filtered vibration data to convert the filtered vibration data from a time domain signal to a frequency domain signal, thereby enabling the quantification of the frequency characteristics of the filtered vibration data.

[0134] Simultaneously, envelope analysis can be performed on the filtered vibration data to calculate the envelope energy data. The purpose is to accurately extract periodic impact characteristics from complex vibration signals. In mechanical vibration, many faults generate periodic impact vibrations. These impact signals typically manifest as "low-frequency fault characteristic frequency" modulating "high-frequency natural vibration." The high-frequency natural vibration (acting as the "carrier wave") is triggered by the fault impact; the low-frequency fault characteristic frequency acts as the "modulation signal," controlling the amplitude change of the high-frequency carrier wave. Envelope analysis can demodulate the amplitude of the filtered vibration signal and extract the envelope energy data reflecting the modulation law, i.e., the low-frequency fault characteristic signal.

[0135] Furthermore, the peak-to-peak value of the filtered vibration data within a fixed time window can be extracted. The peak-to-peak value refers to the difference between the maximum and minimum vibration values ​​within the fixed time window. The peak-to-peak value directly reflects the maximum fluctuation amplitude of the vibration during that time period, providing a direct indication of the vibration intensity and stability of the LNG pump tower system. When the LNG pump tower is operating normally, the peak-to-peak value is typically stable within a small range; however, when the LNG pump tower malfunctions or experiences an abnormality, the vibration intensity increases significantly, leading to a larger peak-to-peak value. By monitoring changes in the peak-to-peak value, it is possible to quickly determine whether the LNG pump tower has deviated from its normal operating condition.

[0136] After obtaining the main peak of the vibration spectrum, the envelope energy data, and the vibration peak-to-peak value, vibration characteristic data that can reflect the vibration status of each target location in the LNG pump tower system can be constructed based on the main peak of the vibration spectrum, the envelope energy data, and the vibration peak-to-peak value.

[0137] In another embodiment of this disclosure, extracting pressure feature data from the pressure sensing data and extracting vibration feature data from the vibration sensing data may further include:

[0138] The pressure sensing data is denoised and filtered to obtain denoised and filtered pressure sensing data; the pressure feature data is extracted from the denoised and filtered pressure sensing data.

[0139] The vibration sensing data is denoised and filtered to obtain denoised and filtered vibration sensing data; the pressure feature data is extracted from the denoised and filtered pressure sensing data.

[0140] The purpose of denoising and filtering pressure sensing data and vibration sensing data is to remove various types of interference noise from the original pressure sensing data and vibration sensing data. For example, pressure sensing modules typically contain ground noise from the sensor circuitry, limit drift caused by ambient temperature changes, sensor zero-point drift noise, and impulse noise. By denoising and filtering the pressure sensing module, the interference of these noises on the main pressure signal can be suppressed, ensuring that the overall pressure change trend and main characteristics are preserved in the denoised and filtered pressure sensing data.

[0141] The purpose of denoising and filtering the original vibration sensing data is similar to that of denoising and filtering the pressure sensing water, which is to suppress the interference of noise on the main vibration signal and ensure that the main characteristics of vibration change can be retained in the denoised and filtered vibration sensing data.

[0142] When denoising and filtering pressure sensing data and vibration sensing data, methods such as mean filtering, median filtering, and wavelet denoising can be used, but the embodiments disclosed herein are not limited to these methods.

[0143] Regarding the above S103:

[0144] In this embodiment of the disclosure, the feature data matrix composed of pressure feature data and vibration feature data can reflect both the pressure characteristics and vibration characteristics of the LNG pump tower system within the target time period. The combination of pressure and vibration features can reflect both the operating status and vibration condition of the LNG pump tower.

[0145] Pressure feature data includes pressure feature vectors corresponding to multiple time steps. Vibration feature data includes feature vectors corresponding to multiple time steps.

[0146] The feature data matrix can be constructed by combining the pressure feature data and the vibration feature data in the following manner:

[0147] For each of the multiple time steps, the pressure feature vector and vibration feature vector corresponding to each time step are concatenated to obtain the concatenated vector corresponding to each time step; and the concatenated vectors corresponding to multiple time steps are superimposed to obtain the feature data matrix; wherein, the rows of the feature data matrix represent the time step dimension and the columns represent the feature dimension.

[0148] In order to utilize the aforementioned feature data matrix to obtain the specific operating status and vibration status of the LNG pump tower system during operation, the intelligent recognition module provided in this embodiment includes, for example, a convolutional neural network, at least two long short-term memory (LSTM) networks, and classifiers corresponding to the at least two LSTM networks.

[0149] A convolutional neural network may include, for example, multiple convolutional layers, activation layers, and pooling layers.

[0150] Specifically, the convolutional neural network in the intelligent recognition module is used to extract features from the feature data matrix to obtain target feature data, which may include, for example:

[0151] Using a one-dimensional convolution kernel, the feature data matrix is ​​convolved along the time step dimension to obtain a first intermediate feature data matrix; and,

[0152] Using a two-dimensional convolution kernel, the first intermediate feature data matrix is ​​convolved in both the time step dimension and the feature dimension to obtain the second intermediate feature data matrix.

[0153] The second intermediate feature data matrix is ​​activated and pooled sequentially to obtain the target feature data.

[0154] Here, in the multiple convolutional layers of a convolutional neural network, there is at least one convolutional layer corresponding to a one-dimensional convolutional kernel; each convolutional layer can use at least one one-dimensional convolutional kernel to perform one-dimensional convolution processing on the feature data.

[0155] After one-dimensional convolution processing through at least one convolutional layer, the first intermediate feature data matrix is ​​obtained.

[0156] Among multiple convolutional layers, there can be at least one convolutional layer corresponding to a two-dimensional convolutional kernel; each convolutional layer can use at least one two-dimensional convolutional kernel to perform two-dimensional convolution processing on the first intermediate feature data matrix.

[0157] Here, for each convolutional layer except the last one, the output of the current convolutional layer is the input of the next convolutional layer.

[0158] After multiple convolutional layers process the feature data matrix, a second intermediate feature data matrix is ​​obtained. Then, activation and pooling layers are used sequentially to activate and pool the second intermediate feature data matrix to obtain the target feature data.

[0159] Multiple convolutional layers are used to convolve the input feature data matrix. The convolutional kernel used in each layer can be, for example, 1×3 or 1×5, to extract local features in the time and frequency domains, thereby extracting more specific pressure and vibration features within the target time period from the feature data matrix. This allows for better fusion of pressure and vibration features.

[0160] In this way, the pressure feature vector V_pressure and the vibration feature vector V_vibration are concatenated into a long feature vector. Then, the feature vectors from multiple consecutive time steps (e.g., greater than 30 seconds) are stacked into a two-dimensional pressure + vibration statistical feature data matrix. The rows of this feature data matrix represent time steps, and the columns represent feature dimensions.

[0161] One-dimensional convolutional layers (1D-CNN) are used, with one-dimensional convolutional kernels sliding along the time dimension to extract features in local temporal patterns. Two-dimensional convolutional layers (2D-CNN) are used, with two-dimensional convolutional kernels sliding along the feature data matrix to extract local spatiotemporal patterns. After a series of convolutional, activation, and pooling layers, the convolutional neural network outputs a highly abstract shared feature tensor, which is the target feature data.

[0162] There are at least two Long Short-Term Memory (LSTM) networks, and the number of memory units in each LSM network can be, for example, 64 to 128, to capture the temporal dependence of feature sequences.

[0163] Taking two long short-term memory networks as an example, the two long short-term memory networks include: a first long short-term memory network for predicting the operating status of the LNG pump tower system, and a second long short-term memory network for predicting the vibration status of the LNG pump tower system;

[0164] The two long short-term memory networks have different intrinsic parameters. When the first long short-term memory network performs predictive processing on the input feature data, it mainly captures the time dependence of the feature data from the pressure dimension. The resulting prediction can reflect the pressure change characteristics at different sampling time points within the same target time period, and also reflect the pressure change characteristics between different target time periods. After inputting the prediction results into the classifier corresponding to this long short-term memory network, the operating status of the LNG pump tower system can be obtained.

[0165] When the second long short-term memory network performs predictive processing on the input feature data, it mainly captures the time dependence of the feature data from the vibration dimension. The resulting prediction results can reflect the vibration change characteristics at different sampling time points within the same target time period, and can also reflect the vibration change characteristics between different target time periods. After inputting the prediction results into the classifier corresponding to the long short-term memory network, the vibration state of the LNG pump tower system can be obtained.

[0166] When using at least two long short-term memory networks in the intelligent recognition module to perform prediction processing on the target feature data to obtain the prediction processing results corresponding to the at least two long short-term machine networks respectively, the following method can be used, for example:

[0167] The target feature data is reshaped or flattened to obtain the target feature data to be processed.

[0168] The target feature data to be processed is input into the first long short-term memory network to obtain the prediction processing result of the first long short-term memory network for predicting the running state of the target feature data to be processed; and...

[0169] The target feature data to be processed is input into the second long short-term memory network to obtain the prediction processing result of the second long short-term memory network for the vibration state prediction of the target feature data to be processed.

[0170] The classifier is, for example, a fully connected classification layer based on Softmax, which enables the determination of the state and vibration state of Yunxi Valley in multiple categories.

[0171] For example, multiple operating states include: single pump operation, multiple pump operation, cavitation, normal, performance degradation, blockage, etc. Multiple vibration states include: normal, imbalance, foundation loosening, resonance, leakage fault, etc.

[0172] Each branch of the Long Short-Term Memory network reads the shared features, processes its sequence information, and outputs its final hidden states s_run and s_vib;

[0173] Feed s_run into the Classifier_run classifier, which outputs a probability distribution, such as [0.95, 0.04, 0.01], representing the probability of "different running states".

[0174] Feeding s_vib into the Classifier_vib classifier outputs another probability distribution, such as [0.1, 0.8, 0.1], representing the probability of "different vibration states".

[0175] Finally, based on the two probability distributions, the operating status and vibration status of the LNG pump tower system were obtained.

[0176] In another embodiment of this disclosure, after obtaining the operating status and vibration status of the LNG pump tower system, the following may also be included:

[0177] When the operating status indicates that the LNG pump tower is operating abnormally or malfunctioning, and / or when the LNG pump tower is operating abnormally or malfunctioning under the vibration condition, an alarm process is executed.

[0178] Specifically, for example, the operating status of the LNG pump tower system can be compared with a first preset threshold. Assume the first preset threshold is used to indicate normal operation of the LNG pump tower system. When the operating status and the first preset threshold do not match, the specific abnormality category of the operating status is determined based on the specific value corresponding to the operating status, and then an alarm message is generated based on the abnormality category for alarm processing.

[0179] Similarly, the vibration state of the LNG pump tower system can be compared with a second preset threshold, which indicates that the vibration of the LNG pump tower system is within the normal range. When the vibration state and the second preset threshold do not match, the specific abnormal category of the vibration is determined based on the value corresponding to the vibration state, and then an alarm message is generated and processed according to the abnormal category.

[0180] In another embodiment of this disclosure, the specific identification results of the operating status and vibration status can also be displayed through a visual interface. The visual interface can be implemented in real time, for example, through an industrial touchscreen or a Supervisory Control and Data Acquisition (SCADA) system.

[0181] When the LNG's operating status and vibration status indicate abnormalities or malfunctions, alarms are issued through a visual interface and / or other audible and visual alarm devices. This includes displaying alarm icons or animations on the visual interface, and issuing audible and visual alarms through other devices. Additionally, alarms can be issued via Short Message Service (SMS) or email. The specific alarm methods are not limited in this embodiment.

[0182] See Figure 2As shown in the illustration, this disclosure provides a specific example of an intelligent recognition module, including: a convolutional neural network comprising three convolutional layers: convolutional layer 1, convolutional layer 2, and convolutional layer 3. In this example, activation layers and pooling layers are not shown. It also includes three long short-term memory networks: LSTM1, LSTM2, and LSTM3, each connected to a corresponding classifier. The classifiers connected to the three long short-term memory networks are classifier 1, classifier 2, and classifier 3, respectively.

[0183] In another embodiment of this disclosure, a specific example of monitoring an LNG pump tower system is also provided, which includes:

[0184] Step 1: Obtain pressure data from m pressure sensing modules during the target time period T1. The target time period T1 includes k pressure sampling time points. For pressure sensing module Mi, the corresponding pressure data is represented as: Mi = {p_i1, p_i2, ..., p_ik}.

[0185] Obtain vibration data from N vibration sensors during a target time period T1, where the target time period T1 includes h vibration sampling time points. For vibration sensor Nj, ​​the corresponding vibration data is represented as: Nj={v_j1、v_j2、…、v_jh}.

[0186] Step 2: Preprocess (denoise and filter) the pressure data of M pressure sensing modules in the target time period T1 to obtain preprocessed pressure data; where the preprocessed pressure data corresponding to Mi is represented as Mi'.

[0187] The vibration data of n vibration sensors at the target time period T1 are preprocessed (denoised and filtered) to obtain the preprocessed vibration data; the preprocessed vibration data corresponding to Nj is represented as Nj'.

[0188] Step 3: For the preprocessed pressure data Mi' corresponding to the i-th pressure sensing module among the m pressure sensing modules, perform the following processing: Perform a fifth-order Butterworth bandpass filter on Mi' to obtain filtered pressure data L_Mi', and perform an FFT transform on the filtered pressure data L_Mi' to obtain the main peak F1_i of the pressure spectrum; extract the peak-to-peak value F2_i of the filtered pressure data L_Mi' within a fixed time window.

[0189] For the preprocessed vibration data Nj' corresponding to the j-th vibration sensor among n vibration sensors, the following processing is performed: a fifth-order Butterworth bandpass filter is applied to Nj' to obtain the vibration filter data L_Nj', and an FFT transform is performed on the vibration filter data L_Nj' to obtain the main peak of the spectrum R1_j; Hilbert envelope analysis is performed on the vibration filter data L_Nj' and the envelope energy R2_j is calculated; the peak-to-peak value R3_j of the vibration filter data L_Nj' within a fixed time window is extracted.

[0190] For m pressure sensing modules, the pressure characteristic data corresponding to the target time period T1 are obtained as follows: F1_1, F2_1, F1_2, F2_2, ..., F1_i, F2_i, ..., F1_m, F2_m.

[0191] For n vibration sensors, the vibration characteristic data corresponding to time period T1 are obtained as follows: R1_1, R2_1, R3_1, R1_2, R2_2, R3_2, ..., R1_j, R2_j, R3_j, ..., R1_n, R2_n, R3_n.

[0192] Step 4: Combine the pressure feature data and vibration feature data corresponding to the target time period T1 into a feature data matrix S, and input it into the convolutional neural network.

[0193] The convolutional layers in a convolutional neural network are used to extract features from the aforementioned feature data matrix S, obtaining the target feature data S'.

[0194] The target feature data S' is transmitted to the g-th LSTM_g of at least two LSTMs. The resulting data is then transmitted to Softmax_g to obtain the classification result corresponding to T1.

[0195] Different LSTMs are used for different classifications. Furthermore, the LSTM's processing result for time period T1 depends on the processing of the corresponding historical time period (such as time period T0).

[0196] The LNG pump tower system monitoring method provided in this disclosure involves installing pressure sensing modules at multiple pipeline locations and vibration sensing modules at multiple target vibration locations within the LNG pump tower system. Pressure sensing data is acquired using the pressure sensing modules, and vibration sensing data is acquired using the vibration sensing modules. Pressure feature data is extracted from the pressure sensing data, and vibration feature data is extracted from the vibration sensing data. The pressure feature data and vibration feature data are then combined to form a feature data matrix, which is input into an intelligent recognition module. A convolutional neural network in the intelligent recognition module is used to extract features from the feature data to obtain target feature data. At least two long short-term memory networks in the intelligent recognition module are used to predict the target feature data, obtaining prediction results. Classifiers corresponding to different long short-term memory networks are used to classify the prediction results, thus obtaining the operating status and vibration status of the LNG pump tower system. By combining deep learning with real-time acquired pressure and vibration sensing data from the LNG pump tower system, the specific operating status and vibration status of the LNG pump tower system can be accurately located, enabling targeted and more efficient handling when a fault occurs in the LNG pump tower system.

[0197] Corresponding to the aforementioned embodiments of the monitoring method for LNG pump tower systems, this disclosure also provides embodiments of monitoring devices for LNG pump tower systems.

[0198] The embodiments of the monitoring device for the LNG pump tower system disclosed herein can be applied to computer equipment. The device embodiments can be implemented through software, hardware, or a combination of both. Taking software implementation as an example, as a logical device, it is formed by the processor of the computer equipment loading the corresponding computer program instructions from non-volatile memory into memory for execution. From a hardware perspective, such as... Figure 3 The diagram shown is a hardware structure diagram of the computer equipment housing the monitoring device of the LNG pump tower system disclosed in this invention, except for... Figure 3 In addition to the processor, memory, network interface, and non-volatile memory shown, the computer equipment in which the device is located in the embodiment may also include other hardware depending on the actual function of the monitoring device of the LNG pump tower system, which will not be described in detail here.

[0199] Please refer to Figure 4 The monitoring device for the LNG pump tower system provided in this embodiment includes:

[0200] The acquisition module 41 is used to acquire pressure sensing data acquired by pressure sensing modules at multiple pipeline locations in the LNG pump tower system, and to acquire vibration sensing data acquired by vibration sensing modules at multiple target vibration locations in the LNG pump tower system.

[0201] Extraction module 42 is used to extract pressure feature data from the pressure sensing data and vibration feature data from the vibration sensing data;

[0202] The prediction module 43 is used to input the feature data matrix composed of the pressure feature data and the vibration feature data into the intelligent recognition module, use the convolutional neural network in the intelligent recognition module to extract features from the feature data matrix to obtain target feature data, and use at least two long short-term memory networks in the intelligent recognition module to perform prediction processing on the target feature data to obtain the prediction processing results corresponding to the at least two long short-term memory networks respectively, and input the prediction processing results corresponding to each long short-term memory network into the classifier corresponding to each long short-term memory network respectively to obtain the operating status and vibration status of the LNG pump tower system.

[0203] Optionally, the plurality of pipe locations include at least one of the following: pump inlet location, pump outlet location, pipe bend location, and flange connection location; the pressure sensing module includes at least one of the following: a pressure sensing module installed at the pump inlet location for detecting instantaneous pressure changes at the pump inlet.

[0204] A differential pressure sensor, installed at the pump outlet, is used to detect the pressure difference between the pump outlet and the ambient environment.

[0205] And pressure sensing modules installed at the bends and flange connections of the pipeline for detecting the pressure distribution information inside the pipeline.

[0206] Optionally, the target vibration location can be determined using the following method:

[0207] Structural modal analysis was performed on the LNG pump tower system to obtain its modal parameters; and the target vibration location was determined based on the results of the structural modal analysis.

[0208] Optionally, the extraction module 42, when extracting pressure feature data from the pressure sensing data, is used to:

[0209] The pressure sensing data is bandpass filtered to obtain filtered pressure data;

[0210] Perform a Fast Fourier Transform on the filtered pressure data to obtain the main peak of the pressure spectrum; and,

[0211] Extract the peak-to-peak pressure values ​​of the filtered pressure data within a fixed time window;

[0212] The pressure characteristic data is constructed based on the main peak of the pressure spectrum and the pressure peak-to-peak value.

[0213] Optionally, the extraction module 42, when extracting vibration feature data from the vibration sensing data, is used to:

[0214] The vibration sensing data is bandpass filtered to obtain filtered vibration data;

[0215] Perform a fast Fourier transform on the filtered vibration data to obtain the main peak of the vibration spectrum;

[0216] Envelope analysis is performed on the filtered vibration data, and envelope energy data is calculated; and,

[0217] Extract the peak-to-peak value of the filtered vibration data within a fixed time window;

[0218] The vibration characteristic data are constructed based on the main peak of the vibration spectrum, the envelope energy data, and the vibration peak-to-peak value.

[0219] Optionally, the pressure feature data includes: pressure feature vectors corresponding to multiple time steps; the vibration feature data includes: feature vectors corresponding to multiple time steps.

[0220] The prediction module 43 constructs the feature data matrix from the pressure feature data and the vibration feature data in the following manner:

[0221] For each of the multiple time steps, the pressure feature vector and vibration feature vector corresponding to each time step are concatenated to obtain the concatenated vector corresponding to each time step; and,

[0222] The concatenated vectors corresponding to multiple time steps are superimposed to obtain the feature data matrix; wherein, the rows of the feature data matrix represent the time step dimension and the columns represent the feature dimension.

[0223] Optionally, when the prediction module 43 uses the convolutional neural network in the intelligent recognition module to extract features from the feature data matrix to obtain target feature data, it is used to:

[0224] Using a one-dimensional convolution kernel, the feature data is convolved in the time step dimension to obtain a first intermediate feature data matrix; and using a two-dimensional convolution kernel, the first intermediate feature data matrix is ​​convolved in both the time step dimension and the feature dimension to obtain a second intermediate feature data matrix.

[0225] The second intermediate feature data matrix is ​​activated and pooled sequentially to obtain the target feature data.

[0226] Optionally, the at least two long short-term memory networks include: a first long short-term memory network for predicting the operating status of the LNG pump tower system, and a second long short-term memory network for predicting the vibration status of the LNG pump tower system;

[0227] The prediction module 43, when using at least two long short-term memory networks in the intelligent recognition module to perform prediction processing on the target feature data and obtain the prediction processing results corresponding to the at least two long short-term machine networks respectively, is used for:

[0228] The target feature data is reshaped or flattened to obtain the target feature data to be processed.

[0229] The target feature data to be processed is input into the first long short-term memory network to obtain the prediction processing result of the first long short-term memory network for predicting the running state of the target feature data to be processed; and...

[0230] The target feature data to be processed is input into the second long short-term memory network to obtain the prediction processing result of the second long short-term memory network for the vibration state prediction of the target feature data to be processed.

[0231] Optionally, the extraction module 42, when extracting pressure feature data from the pressure sensing data and vibration feature data from the vibration sensing data, is used to:

[0232] The pressure sensing data is denoised and filtered to obtain denoised and filtered pressure sensing data; the pressure feature data is extracted from the denoised and filtered pressure sensing data.

[0233] The vibration sensing data is denoised and filtered to obtain denoised and filtered vibration sensing data; the pressure feature data is extracted from the denoised and filtered pressure sensing data.

[0234] Optionally, it also includes: an alarm module 44, used to: perform alarm processing when the operating status indicates that the LNG pump tower is operating abnormally or malfunctioning, and / or when the LNG pump tower is operating abnormally or malfunctioning under the vibration state.

[0235] The specific implementation process of the functions and roles of each unit in the above device can be found in the implementation process of the corresponding steps in the above method, and will not be repeated here.

[0236] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this disclosure according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0237] This disclosure also provides a computer-readable storage medium storing a computer program. When executed by a processor, the computer program performs the steps of the LNG pump tower system monitoring method described in the above-described method embodiments. The storage medium can be either volatile or non-volatile computer-readable storage.

[0238] This disclosure also provides a computer program product carrying program code. The program code includes instructions that can be used to execute the steps of the monitoring method for the LNG pump tower system described in the above method embodiments. For details, please refer to the above method embodiments, which will not be repeated here.

[0239] The aforementioned computer program product can be implemented through hardware, software, or a combination thereof. In one optional embodiment, the computer program product is specifically embodied in a computer storage medium; in another optional embodiment, the computer program product is specifically embodied in a software product, such as a software development kit (SDK), etc.

[0240] The computer program or instructions may be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another. For example, the computer program or instructions may be transferred from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless means. The computer-readable storage medium may be any available medium that a computer can access, or a data storage device such as a server or data center that integrates one or more available media. The available medium may be a magnetic medium, such as a floppy disk, hard disk, or magnetic tape; or an optical medium, such as a digital video optical disc; or a semiconductor medium, such as a solid-state drive. The computer-readable storage medium may be a volatile or non-volatile storage medium, or may include both volatile and non-volatile types of storage media.

[0241] The embodiments of the subject matter and functional operation described in this specification can be implemented in the following ways: digital electronic circuits, tangibly embodied computer software or firmware, computer hardware including the structures disclosed in this specification and their structural equivalents, or combinations thereof. Embodiments of the subject matter described in this specification can be implemented as one or more computer programs, i.e., one or more modules of computer program instructions encoded on a tangible, non-transitory program carrier for execution by a data processing apparatus or for controlling the operation of a data processing apparatus. Alternatively or additionally, the program instructions may be encoded on artificially generated propagation signals, such as machine-generated electrical, optical, or electromagnetic signals, which are generated to encode information and transmit it to a suitable receiving device for execution by the data processing apparatus. The computer storage medium may be a machine-readable storage device, a machine-readable storage substrate, a random or serial access memory device, or combinations thereof.

[0242] The processing and logic flow described in this specification can be executed by one or more programmable computers that execute one or more computer programs to perform corresponding functions by operating on input data and generating output. The processing and logic flow can also be executed by dedicated logic circuitry—such as FPGAs (Field-Programmable Gate Arrays) or ASICs (Application-Specific Integrated Circuits), and the device can also be implemented as dedicated logic circuitry.

[0243] Suitable computers for executing computer programs include, for example, general-purpose and / or special-purpose microprocessors, or any other type of central processing unit. Typically, the central processing unit receives instructions and data from read-only memory and / or random access memory. The basic components of a computer include a central processing unit for implementing or executing instructions and one or more memory devices for storing instructions and data. Typically, a computer will also include one or more mass storage devices for storing data, such as disks, magneto-optical disks, or optical disks, or the computer will be operatively coupled to such mass storage devices to receive data from or transfer data to them, or both. However, a computer is not required to have such devices. Furthermore, a computer can be embedded in another device, such as a mobile phone, a personal digital assistant (PDA), a mobile audio or video player, a game console, a global positioning system (GPS) receiver, or a portable storage device such as a universal serial bus (USB) flash drive, to name a few.

[0244] Computer-readable media suitable for storing computer program instructions and data include all forms of non-volatile memory, media, and memory devices, such as semiconductor memory devices (e.g., EPROM, EEPROM, and flash memory devices), magnetic disks (e.g., internal hard disks or removable disks), magneto-optical disks, and CD-ROM and DVD-ROM disks. Processors and memory may be supplemented by or incorporated into dedicated logic circuitry.

[0245] While this specification contains numerous specific implementation details, these should not be construed as limiting the scope of any invention or the scope of the claims, but rather are primarily intended to describe features of specific embodiments of a particular invention. Certain features described in the various embodiments herein may also be implemented in combination in a single embodiment. Conversely, various features described in a single embodiment may also be implemented separately in various embodiments or in any suitable sub-combination. Furthermore, while features may function in certain combinations as described above and even initially claimed in this way, one or more features from a claimed combination may be removed from that combination in some cases, and a claimed combination may refer to a sub-combination or a variation thereof.

[0246] Similarly, although the operations are depicted in a specific order in the accompanying drawings, this should not be construed as requiring these operations to be performed in the specific order shown or sequentially, or requiring all illustrated operations to be performed to achieve the desired result. In some cases, multitasking and parallel processing may be advantageous. Furthermore, the separation of various system modules and components in the above embodiments should not be construed as requiring such separation in all embodiments, and it should be understood that the described program components and systems can generally be integrated together in a single software product or packaged into multiple software products.

[0247] Thus, specific embodiments of the subject matter have been described. Other embodiments are within the scope of the appended claims. In some cases, the actions recited in the claims may be performed in a different order and still achieve the desired result. Furthermore, the processes depicted in the drawings are not necessarily shown in a specific order or sequence to achieve the desired result. In some implementations, multitasking and parallel processing may be advantageous.

[0248] The above description is merely a preferred embodiment of this disclosure and is not intended to limit this disclosure. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A monitoring method for a liquefied natural gas (LNG) pump tower system, characterized in that, The method includes: The system acquires pressure sensing data from pressure sensing modules deployed at multiple pipeline locations in the LNG pump tower system, and acquires vibration sensing data from vibration sensing modules deployed at multiple target vibration locations in the LNG pump tower system. Pressure feature data is extracted from the pressure sensing data, and vibration feature data is extracted from the vibration sensing data; The feature data matrix composed of the pressure feature data and the vibration feature data is input into the intelligent recognition module. The convolutional neural network in the intelligent recognition module is used to extract features from the feature data matrix to obtain target feature data. At least two long short-term memory networks in the intelligent recognition module are used to predict the target feature data to obtain the prediction results corresponding to the at least two long short-term memory networks. The prediction results corresponding to each long short-term memory network are input into the classifier corresponding to each long short-term memory network to obtain the operating status and vibration status of the LNG pump tower system.

2. The method according to claim 1, characterized in that, The plurality of pipe locations include at least one of the following: pump inlet location, pump outlet location, pipe bend location, and flange connection location; the pressure sensing module includes at least one of the following: a pressure sensing module installed at the pump inlet location for detecting instantaneous pressure changes at the pump inlet. A differential pressure sensor, installed at the pump outlet, is used to detect the pressure difference between the pump outlet and the ambient environment. And pressure sensing modules installed at the bends and flange connections of the pipeline for detecting the pressure distribution information inside the pipeline.

3. The method according to claim 1, characterized in that, The target vibration location is determined using the following method: Structural modal analysis was performed on the LNG pump tower system to obtain its modal parameters; and the target vibration location was determined based on the results of the structural modal analysis.

4. The method according to any one of claims 1-3, characterized in that, Extracting pressure feature data from the pressure sensing data includes: The pressure sensing data is bandpass filtered to obtain filtered pressure data; Perform a Fast Fourier Transform on the filtered pressure data to obtain the main peak of the pressure spectrum; and, Extract the peak-to-peak pressure values ​​of the filtered pressure data within a fixed time window; The pressure characteristic data is constructed based on the main peak of the pressure spectrum and the pressure peak-to-peak value.

5. The method according to any one of claims 1-3, characterized in that, The extraction of vibration feature data from the vibration sensing data includes: The vibration sensing data is bandpass filtered to obtain filtered vibration data; Perform a fast Fourier transform on the filtered vibration data to obtain the main peak of the vibration spectrum; Envelope analysis is performed on the filtered vibration data, and envelope energy data is calculated; and, Extract the peak-to-peak value of the filtered vibration data within a fixed time window; The vibration characteristic data are constructed based on the main peak of the vibration spectrum, the envelope energy data, and the vibration peak-to-peak value.

6. The method according to claim 1, characterized in that, The pressure feature data includes: pressure feature vectors corresponding to multiple time steps; the vibration feature data includes: feature vectors corresponding to multiple time steps. The pressure characteristic data and the vibration characteristic data are used to construct the characteristic data matrix in the following manner: For each of the multiple time steps, the pressure feature vector and vibration feature vector corresponding to each time step are concatenated to obtain the concatenated vector corresponding to each time step; and The concatenated vectors corresponding to multiple time steps are superimposed to obtain the feature data matrix; wherein, the rows of the feature data matrix represent the time step dimension and the columns represent the feature dimension.

7. The method according to claim 6, characterized in that, The step of using the convolutional neural network in the intelligent recognition module to extract features from the feature data matrix to obtain target feature data includes: Using a one-dimensional convolution kernel, the feature data matrix is ​​convolved in the time step dimension to obtain a first intermediate feature data matrix; and using a two-dimensional convolution kernel, the first intermediate feature data matrix is ​​convolved in both the time step dimension and the feature dimension to obtain a second intermediate feature data matrix. The second intermediate feature data matrix is ​​activated and pooled sequentially to obtain the target feature data.

8. The method according to claim 1, characterized in that, The at least two long short-term memory networks include: a first long short-term memory network for predicting the operating status of the LNG pump tower system, and a second long short-term memory network for predicting the vibration status of the LNG pump tower system. The step of using at least two long short-term memory networks in the intelligent recognition module to predict the target feature data and obtain the prediction results corresponding to the at least two long short-term machine networks includes: The target feature data is reshaped or flattened to obtain the target feature data to be processed. The target feature data to be processed is input into the first long short-term memory network to obtain the prediction processing result of the first long short-term memory network for predicting the running state of the target feature data to be processed; and... The target feature data to be processed is input into the second long short-term memory network to obtain the prediction processing result of the second long short-term memory network for the vibration state prediction of the target feature data to be processed.

9. The method according to claim 1, characterized in that, The step of extracting pressure feature data from the pressure sensing data and extracting vibration feature data from the vibration sensing data includes: The pressure sensing data is denoised and filtered to obtain denoised and filtered pressure sensing data; the pressure feature data is extracted from the denoised and filtered pressure sensing data. The vibration sensing data is denoised and filtered to obtain denoised and filtered vibration sensing data; the pressure feature data is extracted from the denoised and filtered pressure sensing data.

10. The method according to claim 1, characterized in that, The method further includes: When the operating status indicates that the LNG pump tower is operating abnormally or malfunctioning, and / or when the LNG pump tower is operating abnormally or malfunctioning under the vibration condition, an alarm process is executed.

11. A monitoring device for a liquefied natural gas (LNG) pump tower system, characterized in that, The device includes: The acquisition module is used to acquire pressure sensing data from pressure sensing modules deployed at multiple pipeline locations in the LNG pump tower system, and to acquire vibration sensing data from vibration sensing modules deployed at multiple target vibration locations in the LNG pump tower system. The extraction module is used to extract pressure feature data from the pressure sensing data and vibration feature data from the vibration sensing data. The prediction module is used to input the feature data matrix composed of the pressure feature data and the vibration feature data into the intelligent recognition module, use the convolutional neural network in the intelligent recognition module to extract features from the feature data matrix to obtain target feature data, and use at least two long short-term memory networks in the intelligent recognition module to perform prediction processing on the target feature data to obtain the prediction processing results corresponding to the at least two long short-term memory networks respectively, and input the prediction processing results corresponding to each long short-term memory network into the classifier corresponding to each long short-term memory network respectively to obtain the operating status and vibration status of the LNG pump tower system.

12. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by a processor, it implements the steps of the method according to any one of claims 1-10.

13. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it performs the following steps: The system acquires pressure sensing data from pressure sensing modules deployed at multiple pipeline locations in the LNG pump tower system, and acquires vibration sensing data from vibration sensing modules deployed at multiple target vibration locations in the LNG pump tower system. Pressure feature data is extracted from the pressure sensing data, and vibration feature data is extracted from the vibration sensing data; The feature data matrix composed of the pressure feature data and the vibration feature data is input into the intelligent recognition module. The convolutional neural network in the intelligent recognition module is used to extract features from the feature data matrix to obtain target feature data. At least two long short-term memory networks in the intelligent recognition module are used to predict the target feature data to obtain the prediction results corresponding to the at least two long short-term machine networks. The prediction results corresponding to each long short-term memory network are input into the classifier corresponding to each long short-term memory network to obtain the operating status and vibration status of the LNG pump tower system.

Images