High-pressure water pump real-time data analysis method based on edge calculation

Abstract

The invention relates to the technical field of data processing, in particular to a high-pressure water pump real-time data analysis method based on edge computing, which allows dynamic adjustment of operating parameters of a high-pressure water pump through real-time communication between edge computing equipment and the high-pressure water pump, distributes data processing tasks to the edge equipment, and improves data processing efficiency. The data transmission time is shortened, the processing efficiency is improved, it is ensured that data can be analyzed and fed back in time, the processing efficiency is improved, and the working efficiency of the high-pressure water pump is optimized.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a real-time data analysis method for a high-pressure water pump based on edge computing. Background Art

[0002] At present, high-pressure water pumps are specially designed to transport high-pressure liquids. They can raise the liquid to a higher pressure level. Such pumps are widely used in various industrial and civil fields, such as water treatment, fire protection systems, agricultural irrigation, mining, oil and gas extraction, construction, and chemical industries.

[0003] At present, real-time data analysis technology for high-pressure water pumps usually relies on traditional monitoring systems and data processing methods. These methods mainly include the following aspects: 1. Fixed frequency data acquisition: The sensor collects data at a preset fixed frequency, maintaining the same sampling rate regardless of the working status of the high-pressure water pump.

[0004] 2. Centralized data processing: All collected data is transmitted via the network to a central server or data center for centralized processing and analysis.

[0005] 3. Rule-based fault detection: Use simple threshold judgments or experience-based rules to detect faults, such as triggering an alarm when a parameter exceeds a preset safety range.

[0006] 4. Regular maintenance and inspection: Rely on manual inspection and maintenance of equipment regularly to detect potential problems.

[0007] However, the above-mentioned methods in the prior art have the following problems: 1. Low data processing efficiency The fixed frequency data collection method uses a fixed sampling frequency and cannot be adjusted dynamically according to actual needs. This results in a large amount of redundant data being collected under normal operating conditions, while key information may not be obtained in a timely manner under abnormal or fault conditions.

[0008] Centralized data processing: All data needs to be transmitted to a remote data center for processing, which increases the pressure on network bandwidth, prolongs data processing time, and reduces the system's response speed.

[0009] Delay problem: The entire process from data collection and transmission to analysis and decision-making is long, making it difficult to achieve instant response, which affects the rapid adjustment of the operating status of the high-pressure water pump.

[0010] 2. Insufficient fault detection capabilities Rule-based methods: Traditional fault detection methods mainly rely on simple threshold judgments or rules based on experience. They are not accurate enough for the analysis of complex signals (such as vibration signals) and are prone to false positives or negative positives.

[0011] Lack of advanced analysis methods: Existing technologies may not fully utilize advanced signal processing technologies such as frequency domain feature analysis and wavelet transform, resulting in low sensitivity and accuracy of fault diagnosis.

[0012] Weak early warning capability: Due to the lack of effective recognition of weak fault signals, existing methods can often only detect faults after they become obvious, and early warning cannot be achieved.

[0013] 3. Inadequate use of resources Data redundancy: Fixed-frequency data collection results in a large amount of non-critical data being stored and transmitted, wasting storage space and network resources.

[0014] High energy consumption: Frequent data transmission and unnecessary data processing increase the energy consumption of the system.

[0015] Low degree of automation: The existing system relies on manual intervention for parameter adjustment and maintenance, which not only increases the workload of operators but also reduces the continuity and stability of the system.

[0016] Therefore, a real-time data analysis method for high-pressure water pumps based on edge computing is needed to solve the above problems. Summary of the invention

[0017] The present invention provides a real-time data analysis method for a high-pressure water pump based on edge computing. In the present invention, real-time communication between an edge computing device and a high-pressure water pump is used to allow dynamic adjustment of the operating parameters of the high-pressure water pump, and to assign data processing tasks to edge devices, thereby reducing data transmission time and improving processing efficiency, thereby ensuring that data can be analyzed and fed back in a timely manner, improving processing efficiency, and optimizing the working efficiency of the high-pressure water pump.

[0018] The technical solution adopted by the present invention to solve the above technical problems is: a real-time data analysis method for a high-pressure water pump based on edge computing, characterized in that it includes the following steps: Step S1: deploying at least one edge computing device within a radius of 5 meters around the high-pressure water pump, wherein the edge computing device is interconnected with multiple sensors arranged on the high-pressure water pump; Step S2: dynamically adjusting the data acquisition frequency of each sensor according to the working status and historical data of the high-pressure water pump; Step S3: extracting the time domain characteristic data of the water pressure mean, variance, flow rate change rate, temperature peak value and the frequency domain characteristic data of the vibration signal from the raw data collected by the sensor; Step S4: deploying a lightweight deep learning model that supports online learning on the edge device; Step S5: analyzing the extracted frequency domain feature data to identify the fault signal of the high-pressure water pump; Step S6: compress the analysis result data obtained in step S7 on the edge computing device and transmit it to the data analysis center; Step S7: decompressing the compressed analysis result data at the data analysis center, and using a real-time data analysis algorithm to respond immediately to the operating status of the high-pressure water pump based on the decompressed data; Step S8: According to the result of the immediate response, the corresponding instructions are sent to the high-pressure water pump through the communication protocol between the edge computing device and the high-pressure water pump to realize dynamic adjustment of the operating parameters of the high-pressure water pump.

[0019] Furthermore, the sensor comprises: Pressure sensor: installed at the water inlet and outlet of the high-pressure water pump to monitor the water pressure at the water inlet and outlet of the high-pressure water pump; Flow sensor: installed on the outlet pipe of the high-pressure water pump to monitor the actual flow output of the high-pressure water pump; Temperature sensor: installed at the water inlet and outlet of the high-pressure water pump, as well as the bearing seat and motor housing of the high-pressure water pump, to monitor the water inlet temperature, water outlet temperature and the temperature of the high-pressure water pump bearing and motor; Vibration sensor: installed on the casing, bearing seat and motor housing of the high-pressure water pump, used to monitor the overall vibration of the high-pressure water pump and the vibration of the high-pressure water pump bearings and motor.

[0020] Furthermore, in step S2, dynamically adjusting the data acquisition frequency of each sensor according to the working state and historical data of the high-pressure water pump includes: Step S2-1: extracting historical operation data of the high-pressure water pump from the data storage system, monitoring the output data of the sensor in real time on the edge computing device to evaluate the current operation status of the high-pressure water pump, and analyzing the current load condition of the high-pressure water pump according to the real-time operation status and historical operation data of the high-pressure water pump; Step S2-2: setting a preliminary data acquisition frequency strategy based on the working state of the high-pressure water pump, reducing the acquisition frequency when the high-pressure water pump is in a normal working state, and increasing the acquisition frequency when the high-pressure water pump is in an abnormal state or a fault state; Step S2-3: During the operation of the high-pressure water pump, the data collection frequency of each sensor is automatically adjusted according to the real-time data analysis results and status evaluation; Step S2-4: Communicate with each sensor through the edge computing device and issue the adjusted acquisition frequency instruction in real time.

[0021] Furthermore, in step S3, extracting the time domain characteristic data of the water pressure mean, variance, flow rate change rate, temperature peak value and the frequency domain characteristic data of the vibration signal from the raw data collected by the sensor includes: Step S3-1: Calculation of mean water pressure: Obtain continuous pressure readings within a certain period of time from the pressure sensor, and calculate the average value of all readings within this period of time as the average water pressure; Step S3-2: Variance calculation: Calculate the variance based on the pressure reading obtained in step S3-1; Step S3-2: Determine the flow rate change rate: Obtain flow data from the flow meter, calculate the change in flow between two adjacent time points, and divide it by the time interval to obtain the flow change rate; Step S3-3: Temperature peak identification: Collect temperature data from temperature sensors, analyze temperature change curves over time, and identify temperature peaks; Step S3-4: Obtain a vibration signal through a vibration sensor.

[0022] Furthermore, the raw data in step S3 refers to data directly collected from multiple sensors, including pressure values ​​at the pump inlet and outlet, flow values ​​at the pump outlet, temperature values ​​of the pump inlet and outlet as well as the high-pressure water pump bearings and motor, overall vibration conditions of the pump housing and the high-pressure water pump, and vibration signals of the high-pressure water pump bearings and motor.

[0023] Furthermore, the variance refers to the variance in the data sequence obtained from the pressure sensor, and is a statistic used to measure the fluctuation of water pressure readings over a period of time.

[0024] Furthermore, in step S5, the extracted time domain feature data and frequency domain feature data are analyzed to identify the fault signal of the high pressure water pump, including: Step S5-1: Time domain feature extraction: segment the time domain signal of the time domain feature data according to a certain time window to perform local feature analysis; Step S5-2: frequency domain feature extraction: applying fast Fourier transform to the time domain signal in each time period to convert the time domain signal into a frequency domain signal; Step S5-3: extracting main frequency components from the frequency domain signal of the frequency domain feature data and the frequency domain signal obtained in step S5-2, and calculating the frequency domain features; Step S5-4: wavelet transform: perform wavelet transform on the time domain signal; Step S5-5: Signal reconstruction: Reconstruct the time domain signal according to the result of wavelet transform, and extract the time domain characteristic signals in different frequency ranges; Step S5-6: Feature fusion: splicing the time domain signal with all frequency domain signals and wavelet transform results to form a comprehensive feature vector; Step S5-7: Feature selection: Use an automatic feature selection algorithm to extract the most representative feature vector and reduce the feature dimension; Step S5-8: Fault signal identification: Use the historical fault data of the high-pressure water pump to train a lightweight deep learning model, input the extracted feature vectors into the trained model, infer the operating status of the water pump in real time, and identify potential fault signals.

[0025] Furthermore, in step S7, the compressed analysis result data is decompressed in the data analysis center, and based on the decompressed data, a real-time data analysis algorithm is used to immediately respond to the operating status of the high-pressure water pump, including: Step S7-1: receiving compressed analysis result data: the data analysis center receives the compressed data packet from the data acquisition system or storage unit, decompresses the received compressed data using a preset decompression algorithm, and stores the decompressed data in the memory for subsequent analysis; Step S7-2: Applying a real-time data analysis algorithm: selecting a real-time data analysis algorithm to evaluate the operating status of the high-pressure water pump; Step S7-3: According to the results of the algorithm analysis, the current operating status of the high-pressure water pump is evaluated, including normal operation, potential failures and emergency shutdown, and an operating status report is generated; Step S7-4: Implement an immediate response mechanism based on the operating status report. If a fault is detected in the high-pressure water pump, trigger the alarm system to notify relevant personnel.

[0026] Furthermore, the real-time data analysis algorithm in step S7-2 adopts an analysis method based on the combination of deep learning and anomaly detection, including: Step S7-2-1: To capture dynamic patterns in the time series, a sliding window technique is used to segment continuous data points into multiple overlapping or non-overlapping time periods; Step S7-2-2: using an unsupervised learning method to train a convolutional autoencoder to learn a compact representation of the vibration signal under normal working conditions; Step S7-2-3: After the training is completed, the encoder is used to extract the feature vector, and the decoder is used to reconstruct the signal to facilitate the subsequent calculation of the reconstruction error; Step S7-2-4: extracting the feature vector from the convolutional autoencoder as the input of the LSTM network, and training the LSTM network to predict the feature vector of the next time step of the high-pressure water pump; Step S7-2-5: Compare the difference between the actually observed feature vector and the feature vector predicted by LSTM. If the difference between the two exceeds a preset threshold, it is considered that an abnormality has occurred in the high-pressure water pump at the current moment.

[0027] Furthermore, the convolutional autoencoder in step S7-2-2 consists of an encoder and a decoder, the encoder maps the input to a low-dimensional latent space, and the decoder restores it back to the original dimension.

[0028] The advantages of the present invention are: 1. In the present invention, real-time communication between the edge computing device and the high-pressure water pump allows dynamic adjustment of the operating parameters of the high-pressure water pump, assigns data processing tasks to the edge device, reduces data transmission time, and improves processing efficiency, so as to ensure that data can be analyzed and fed back in a timely manner, improve processing efficiency, and optimize the working efficiency of the high-pressure water pump.

[0029] 2. The present invention can analyze the data collected from the sensor, especially through the frequency domain feature analysis of complex information such as vibration signals, and can accurately capture and identify the fault signal of the high-pressure water pump. This method is more sensitive and accurate than the traditional method, which helps to achieve early warning of faults, thereby significantly improving system reliability and maintenance efficiency.

[0030] 3. The present invention dynamically adjusts the data acquisition frequency of the sensor according to the working state of the high-pressure water pump. This intelligent regulation not only ensures the timely acquisition of key data, but also avoids unnecessary data redundancy, effectively improving resource utilization efficiency. In addition, preliminary data analysis and feature extraction are performed on the edge computing device, which reduces the amount of data that needs to be transmitted to the remote data center and improves the response speed and efficiency of the overall system. DETAILED DESCRIPTION

[0031] The technical solution of the present invention is described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0032] Embodiment 1: The present invention provides a high-pressure water pump real-time data analysis method based on edge computing, characterized in that it includes the following steps: Step S1: deploy at least one edge computing device within a radius of 5 meters around the high-pressure water pump. The edge computing device is interconnected with multiple sensors arranged on the high-pressure water pump, where the sensors include: Pressure sensor: installed at the water inlet and outlet of the high-pressure water pump to monitor the water pressure at the water inlet and outlet of the high-pressure water pump; Flow sensor: installed on the outlet pipe of the high-pressure water pump to monitor the actual flow output of the high-pressure water pump; Temperature sensor: installed at the water inlet and outlet of the high-pressure water pump, as well as the bearing seat and motor housing of the high-pressure water pump, to monitor the water inlet temperature, water outlet temperature and the temperature of the high-pressure water pump bearing and motor; Vibration sensor: installed on the casing, bearing seat and motor housing of the high-pressure water pump, used to monitor the overall vibration of the high-pressure water pump and the vibration of the high-pressure water pump bearings and motor.

[0033] The edge computing device is deployed with a lightweight deep learning model that supports online learning. The deep learning model is based on a long short-term memory network with an attention mechanism, and can continuously update model parameters to adapt to changing operating conditions. Advanced analysis of vibration signals: using a combination of frequency domain and time domain analysis methods, combined with wavelet transform and Fourier transform, to identify fault signals of high-pressure water pumps. Remaining service life prediction: based on a deep reinforcement learning model, combined with current operating status and historical data, to predict the remaining service life of key components.

[0034] Step S2: dynamically adjust the data collection frequency of each sensor according to the working status and historical data of the high-pressure water pump, specifically including: According to the working status and historical data of the high-pressure water pump, the data collection frequency of each sensor is dynamically adjusted, including: Step S2-1: extracting historical operation data of the high-pressure water pump from the data storage system, monitoring the output data of the sensor in real time on the edge computing device to evaluate the current operation status of the high-pressure water pump, and analyzing the current load condition of the high-pressure water pump according to the real-time operation status and historical operation data of the high-pressure water pump; Step S2-2: setting a preliminary data acquisition frequency strategy based on the working state of the high-pressure water pump, reducing the acquisition frequency when the high-pressure water pump is in a normal working state, and increasing the acquisition frequency when the high-pressure water pump is in an abnormal state or a fault state; Step S2-3: During the operation of the high-pressure water pump, the data collection frequency of each sensor is automatically adjusted according to the real-time data analysis results and status evaluation; Step S2-4: Communicate with each sensor through the edge computing device and issue the adjusted acquisition frequency instruction in real time.

[0035] By implementing the above steps, the edge computing device can adaptively adjust the data collection frequency of the sensor according to the real-time status and historical data of the high-pressure water pump, optimize the data collection process, and improve the sensitivity and accuracy of fault detection. In this way, the edge computing device can play a more effective role in the operation monitoring of the high-pressure water pump and improve the overall performance and reliability of the system.

[0036] The edge computing device used in the present invention refers to a small computing device with sufficient computing power and storage space, which can perform data processing, analysis and control tasks close to the data source. The device has the following characteristics: Computing power: Sufficient processing power to run machine learning models, perform data preprocessing, and perform real-time analytics.

[0037] Storage space: Able to store necessary data and model parameters.

[0038] Low power consumption: Designed for low power consumption to meet the needs of long-term operation.

[0039] Network connection: Support multiple network connection methods (such as 5G, Wi-Fi, Ethernet, etc.) to ensure stable communication with the cloud and other devices.

[0040] Operating system: Pre-installed lightweight operating system, such as Linux distribution or operating system designed for IoT devices.

[0041] Containerized environment: Supports container technologies (such as Docker) to facilitate rapid deployment and update of applications.

[0042] Security: Built-in security features including data encryption, authentication, and access control.

[0043] The following are examples of specific edge computing devices that can be deployed: 1. Industrial-grade edge computing gateway Model examples: Advantech UNO-2000 series or Siemens SIMATIC IOT2000 series devices are used. These devices are equipped with multi-core processors, support real-time data processing and analysis, are rugged and suitable for industrial environments, and can provide a variety of sensor interfaces (such as RS485, Modbus, CAN bus) and network interfaces (such as Ethernet, Wi-Fi).

[0044] Scalability: Supports modular expansion, and can add more computing resources or storage space as needed.

[0045] 2. Embedded computing platform Model examples: NVIDIA Jetson Nano or Raspberry Pi 4. The VIDIA Jetson Nano is equipped with a GPU and is suitable for running deep learning models. The Raspberry Pi 4 provides sufficient CPU performance and is compact and portable: its small size makes it easy to install in a small space, and its relatively low price makes it suitable for large-scale deployment.

[0046] 3. Dedicated edge computing devices Model examples: Huawei Atlas 200 AI acceleration module or Intel NUC. For example, Huawei Atlas 200 integrates an AI accelerator, which is suitable for complex machine learning tasks; while Intel NUC is equipped with a high-performance processor and graphics card to support complex data processing and analysis.

[0047] Deployment example: Specifically, the present invention uses one of the industrial edge computing gateways as the edge computing device. The following are specific deployment instructions: Select Advantech UNO-2000 series industrial edge computing gateways. Location determination: Install the device in a protective box within a radius of 5 meters around the high-pressure water pump to ensure that it is not affected by environmental factors. Sensor connection: Pressure sensor: Connect to the pressure sensor at the water inlet and outlet through the RS485 interface. Flow sensor: Connect to the flow sensor on the outlet pipe through the Modbus interface. Temperature sensor: Connect to the temperature sensor on the water inlet, water outlet, bearing seat and motor housing through the analog input interface. Vibration sensor: Connect to the vibration sensor on the pump housing, bearing seat and motor housing through the I2C interface. Network configuration: Configure 5G or Ethernet connection to ensure secure and reliable communication with the cloud server. Software operating system: Pre-install Ubuntu Core or lightweight Linux distribution. Containerized environment: Use Docker to deploy data acquisition, preprocessing, model reasoning and control execution applications. Security settings: Enable data encryption, authentication and access control mechanisms to ensure the security of the system. By deploying edge computing devices as described above, it can be ensured that the edge computing devices can efficiently support real-time data analysis and intelligent control of high-pressure water pumps.

[0048] Step S3: extracting the time domain characteristic data of the water pressure mean, variance, flow rate change, temperature peak value and the frequency domain characteristic data of the vibration signal from the raw data collected by the sensor, wherein the raw data refers to the data directly collected from multiple sensors, including the pressure value of the pump inlet and outlet, the flow value of the pump outlet, the temperature value of the pump inlet and outlet and the high-pressure water pump bearing, the motor, the overall vibration of the pump housing and the high-pressure water pump, and the vibration signals of the high-pressure water pump bearing and the motor. Specifically, this step includes: Step S3-1: Calculation of mean water pressure: Obtain continuous pressure readings within a certain period of time from the pressure sensor, and calculate the average value of all readings within this period of time as the average water pressure; Step S3-2: Variance calculation: According to the pressure readings obtained in step S3-1, the variance is calculated, which reflects the pressure fluctuation. Specifically, the variance refers to the variance in the data sequence obtained from the pressure sensor, which is a statistic used to measure the fluctuation of the water pressure readings over a period of time. The specific calculation formula is: , where x i is the i-th pressure reading in the time period, μ is the average of all water pressure readings, and n is the total number of readings. The larger the variance, the greater the pressure change during this period, which means that the working state of the water pump is not stable enough. Therefore, the variance can be used as an important feature to monitor the operating status of the high-pressure water pump and predict faults.

[0049] Step S3-2: Determine the flow rate change rate: Obtain flow data from the flow meter, calculate the change in flow between two adjacent time points, and divide it by the time interval to obtain the flow change rate; Step S3-3: Temperature peak identification: Collect temperature data from temperature sensors, analyze temperature change curves over time, and identify temperature peaks; Step S3-4: Obtain a vibration signal through a vibration sensor.

[0050] After completing the above steps, a set of feature data representing the operating status of the high-pressure water pump is obtained, and the extracted feature data is stored for use in the deep learning model training in the next step S4.

[0051] Step S4: deploying a lightweight deep learning model that supports online learning on the edge device; Step S5: Analyze the extracted frequency domain feature data to identify the fault signal of the high-pressure water pump, specifically including: Step S5-1: Time domain feature extraction: segment the time domain signal of the time domain feature data according to a certain time window to perform local feature analysis; Step S5-2: frequency domain feature extraction: applying fast Fourier transform to the time domain signal in each time period to convert the time domain signal into a frequency domain signal; Step S5-3: extracting main frequency components from the frequency domain signal of the frequency domain feature data and the frequency domain signal obtained in step S5-2, and calculating the frequency domain features; Step S5-4: wavelet transform: perform wavelet transform on the time domain signal; Step S5-5: Signal reconstruction: Reconstruct the time domain signal according to the result of wavelet transform, and extract the time domain characteristic signals in different frequency ranges; Step S5-6: Feature fusion: splicing the time domain signal with all frequency domain signals and wavelet transform results to form a comprehensive feature vector; Step S5-7: Feature selection: Use an automatic feature selection algorithm (such as LASSO regression or random forest) to extract the most representative feature vector and reduce the feature dimension; Step S5-8: Fault signal identification: Use the historical fault data of the high-pressure water pump to train a lightweight deep learning model, input the extracted feature vectors into the trained model, infer the operating status of the water pump in real time, and identify potential fault signals.

[0052] Through the above steps, we can effectively combine frequency domain and time domain analysis techniques, and use wavelet transform and Fourier transform to identify weak fault signals that may occur during the operation of high-pressure water pumps. This not only helps to monitor the status of equipment in real time, but also enables timely maintenance measures to avoid major faults. Step S6: compressing the analysis result data obtained in step S7 on the edge computing device and transmitting the data to the data analysis center; Step S7: decompress the compressed analysis result data at the data analysis center, and use a real-time data analysis algorithm to respond to the operating status of the high-pressure water pump based on the decompressed data, specifically including: Step S7-1: receiving compressed analysis result data: the data analysis center receives the compressed data packet from the data acquisition system or storage unit, decompresses the received compressed data using a preset decompression algorithm, and stores the decompressed data in the memory for subsequent analysis; Step S7-2: Applying a real-time data analysis algorithm: Selecting a real-time data analysis algorithm to evaluate the operating status of the high-pressure water pump. Specifically, the real-time data analysis algorithm adopts an analysis method based on a combination of deep learning and anomaly detection, including: Step S7-2-1: To capture dynamic patterns in the time series, a sliding window technique is used to segment continuous data points into multiple overlapping or non-overlapping time periods; Step S7-2-2: Using unsupervised learning, a convolutional autoencoder is trained to learn a compact representation of the vibration signal under normal working conditions. The convolutional autoencoder consists of an encoder and a decoder. The encoder maps the input to a low-dimensional latent space, and the decoder restores it back to the original dimension. Step S7-2-3: After the training is completed, the encoder is used to extract the feature vector, and the decoder is used to reconstruct the signal to facilitate the subsequent calculation of the reconstruction error; Step S7-2-4: extracting the feature vector from the convolutional autoencoder as the input of the LSTM network, and training the LSTM network to predict the feature vector of the next time step of the high-pressure water pump; Step S7-2-5: Compare the difference between the actually observed feature vector and the feature vector predicted by LSTM. If the difference between the two exceeds a preset threshold, it is considered that an abnormality has occurred in the high-pressure water pump at the current moment; Step S7-3: According to the results of the algorithm analysis, the current operating status of the high-pressure water pump is evaluated, including normal operation, potential failures and emergency shutdown, and an operating status report is generated; Step S7-4: Implement an immediate response mechanism based on the operating status report. If a fault is detected in the high-pressure water pump, trigger the alarm system to notify relevant personnel.

[0053] Step S8: According to the result of the immediate response, the corresponding instructions are sent to the high-pressure water pump through the communication protocol between the edge computing device and the high-pressure water pump to realize dynamic adjustment of the operating parameters of the high-pressure water pump.

[0054] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A real-time data analysis method for a high-pressure water pump based on edge computing, characterized in that: The following steps are involved: Step S1: deploying at least one edge computing device within a radius of 5 meters around the high-pressure water pump, wherein the edge computing device is interconnected with multiple sensors arranged on the high-pressure water pump; Step S2: dynamically adjusting the data acquisition frequency of each sensor according to the working status and historical data of the high-pressure water pump; Step S3: extracting the time domain characteristic data of the water pressure mean, variance, flow rate change rate, temperature peak value and the frequency domain characteristic data of the vibration signal from the raw data collected by the sensor; Step S4: deploying a lightweight deep learning model that supports online learning on the edge device; Step S5: analyzing the extracted frequency domain feature data to identify the fault signal of the high-pressure water pump; Step S6: compressing the analysis result data obtained in step S7 on the edge computing device and transmitting the data to the data analysis center; Step S7: decompressing the compressed analysis result data at the data analysis center, and using a real-time data analysis algorithm to respond immediately to the operating status of the high-pressure water pump based on the decompressed data; Step S8: According to the result of the immediate response, the corresponding instructions are sent to the high-pressure water pump through the communication protocol between the edge computing device and the high-pressure water pump to realize dynamic adjustment of the operating parameters of the high-pressure water pump.

2. According to the edge computing-based real-time data analysis method for high-pressure water pumps according to claim 1, it is characterized in that: The sensor comprises: Pressure sensor: installed at the water inlet and outlet of the high-pressure water pump to monitor the water pressure at the water inlet and outlet of the high-pressure water pump; Flow sensor: installed on the outlet pipe of the high-pressure water pump to monitor the actual flow output of the high-pressure water pump; Temperature sensor: installed at the water inlet and outlet of the high-pressure water pump, as well as the bearing seat and motor housing of the high-pressure water pump, to monitor the water inlet temperature, water outlet temperature and the temperature of the high-pressure water pump bearing and motor; Vibration sensor: installed on the casing, bearing seat and motor housing of the high-pressure water pump, used to monitor the overall vibration of the high-pressure water pump and the vibration of the high-pressure water pump bearings and motor.

3. According to a method for real-time data analysis of a high-pressure water pump based on edge computing according to claim 1, it is characterized in that: In step S2, dynamically adjusting the data acquisition frequency of each sensor according to the working state and historical data of the high-pressure water pump includes: Step S2-1: extracting historical operation data of the high-pressure water pump from the data storage system, monitoring the output data of the sensor in real time on the edge computing device to evaluate the current operation status of the high-pressure water pump, and analyzing the current load condition of the high-pressure water pump according to the real-time operation status and historical operation data of the high-pressure water pump; Step S2-2: setting a preliminary data acquisition frequency strategy based on the working state of the high-pressure water pump, reducing the acquisition frequency when the high-pressure water pump is in a normal working state, and increasing the acquisition frequency when the high-pressure water pump is in an abnormal state or a fault state; Step S2-3: During the operation of the high-pressure water pump, the data collection frequency of each sensor is automatically adjusted according to the real-time data analysis results and status evaluation; Step S2-4: Communicate with each sensor through the edge computing device and issue the adjusted acquisition frequency instruction in real time.

4. According to claim 1, a high-pressure water pump real-time data analysis method based on edge computing is characterized in that: The step S3 extracts the time domain characteristic data of the water pressure mean, variance, flow rate change rate, temperature peak value and the frequency domain characteristic data of the vibration signal from the raw data collected by the sensor, including: Step S3-1: Calculation of mean water pressure: Obtain continuous pressure readings within a certain period of time from the pressure sensor, and calculate the average value of all readings within this period of time as the average water pressure; Step S3-2: Variance calculation: Calculate the variance based on the pressure reading obtained in step S3-1; Step S3-2: Determine the flow rate change rate: Obtain flow data from the flow meter, calculate the change in flow between two adjacent time points, and divide it by the time interval to obtain the flow change rate; Step S3-3: Temperature peak identification: Collect temperature data from temperature sensors, analyze temperature change curves over time, and identify temperature peaks; Step S3-4: Obtain a vibration signal through a vibration sensor.

5. According to claim 4, a high-pressure water pump real-time data analysis method based on edge computing is characterized in that: The raw data in step S3 refers to data directly collected from multiple sensors, including pressure values ​​at the pump inlet and outlet, flow values ​​at the pump outlet, temperature values ​​of the pump inlet and outlet as well as the high-pressure water pump bearings and motor, overall vibration conditions of the pump housing and the high-pressure water pump, and vibration signals of the high-pressure water pump bearings and motor.

6. A high-pressure water pump real-time data analysis method based on edge computing according to claim 4, characterized in that: The variance refers to the variance in the data sequence obtained from the pressure sensor, and is a statistic used to measure the fluctuation of the water pressure readings over a period of time.

7. A high-pressure water pump real-time data analysis method based on edge computing according to claim 1, characterized in that: The step S5 analyzes the extracted time domain feature data and frequency domain feature data to identify the fault signal of the high pressure water pump, including: Step S5-1: Time domain feature extraction: segment the time domain signal of the time domain feature data according to a certain time window to perform local feature analysis; Step S5-2: frequency domain feature extraction: applying fast Fourier transform to the time domain signal in each time period to convert the time domain signal into a frequency domain signal; Step S5-3: extracting main frequency components from the frequency domain signal of the frequency domain feature data and the frequency domain signal obtained in step S5-2, and calculating the frequency domain features; Step S5-4: wavelet transform: perform wavelet transform on the time domain signal; Step S5-5: Signal reconstruction: Reconstruct the time domain signal according to the result of wavelet transform, and extract the time domain characteristic signals in different frequency ranges; Step S5-6: Feature fusion: splicing the time domain signal with all frequency domain signals and wavelet transform results to form a comprehensive feature vector; Step S5-7: Feature selection: Use an automatic feature selection algorithm to extract the most representative feature vector and reduce the feature dimension; Step S5-8: Fault signal identification: Use the historical fault data of the high-pressure water pump to train a lightweight deep learning model, input the extracted feature vectors into the trained model, infer the operating status of the water pump in real time, and identify potential fault signals.

8. The real-time data analysis method for a high-pressure water pump based on edge computing according to claim 1 is characterized in that: In step S7, the compressed analysis result data is decompressed in the data analysis center, and based on the decompressed data, a real-time data analysis algorithm is used to immediately respond to the operating status of the high-pressure water pump, including: Step S7-1: receiving compressed analysis result data: the data analysis center receives the compressed data packet from the data acquisition system or storage unit, decompresses the received compressed data using a preset decompression algorithm, and stores the decompressed data in the memory for subsequent analysis; Step S7-2: Applying a real-time data analysis algorithm: selecting a real-time data analysis algorithm to evaluate the operating status of the high-pressure water pump; Step S7-3: According to the results of the algorithm analysis, the current operating status of the high-pressure water pump is evaluated, including normal operation, potential failures and emergency shutdown, and an operating status report is generated; Step S7-4: Implement an immediate response mechanism based on the operating status report. If a fault is detected in the high-pressure water pump, trigger the alarm system to notify relevant personnel.

9. The real-time data analysis method for a high-pressure water pump based on edge computing according to claim 8 is characterized in that: The real-time data analysis algorithm in step S7-2 adopts an analysis method based on the combination of deep learning and anomaly detection, including: Step S7-2-1: To capture dynamic patterns in the time series, a sliding window technique is used to segment continuous data points into multiple overlapping or non-overlapping time periods; Step S7-2-2: using an unsupervised learning method to train a convolutional autoencoder to learn a compact representation of the vibration signal under normal working conditions; Step S7-2-3: After the training is completed, the encoder is used to extract the feature vector, and the decoder is used to reconstruct the signal to facilitate the subsequent calculation of the reconstruction error; Step S7-2-4: extracting the feature vector from the convolutional autoencoder as the input of the LSTM network, and training the LSTM network to predict the feature vector of the next time step of the high-pressure water pump; Step S7-2-5: Compare the difference between the actually observed feature vector and the feature vector predicted by LSTM. If the difference between the two exceeds a preset threshold, it is considered that an abnormality has occurred in the high-pressure water pump at the current moment.

10. A high-pressure water pump real-time data analysis method based on edge computing according to claim 9, characterized in that: The convolutional autoencoder in step S7-2-2 consists of an encoder and a decoder. The encoder maps the input to a low-dimensional latent space, and the decoder restores it back to the original dimension.