AI-based oil line leakage warning method and system
Through the AI-based oil circuit leakage warning method, the use of temperature, strain and vibration sensors for data processing and neural network model training, the problems of low oil circuit leakage monitoring efficiency and poor early warning capability in the existing technology are solved, and more efficient Monitoring and early warning improve the reliability of the oil circuit.
Patent Information
- Application Number
- CN202411291074.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-14
- Publication Date
- 2024-12-06
- Estimated Expiration
- 2044-09-14
AI Technical Summary
The existing oil circuit leakage monitoring and early warning methods have low monitoring efficiency and poor early warning capabilities, which affect production.
Using the AI-based oil circuit leakage warning method, by obtaining the detailed information of the target oil circuit, a monitoring and sensing network composed of temperature sensors, strain sensors and vibration sensors are arranged along the oil circuit path, and the original sensing data is obtained for adaptive filtering and denoising. and temperature correction, build a neural network model for supervision and training, generate an oil circuit early warning model, and output high-risk locations, early warning confidence and critical time.
It improves monitoring efficiency and early warning capabilities, improves oil circuit reliability, and reduces equipment downtime and maintenance costs.
Smart Images

Figure CN118793955B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of fluid pipeline monitoring, and in particular to an AI-based oil line leakage warning method and system. Background Art
[0002] The oil circuit system plays a vital role in industrial production, transportation, and mechanical equipment. In the shell-and-tube oil cooler used to reduce oil temperature and ensure the normal operation of the equipment, due to stress and temperature changes, oil circuit leakage is prone to occur in long-term operation, which in turn affects the oil circuit operation efficiency. Existing oil circuit leakage monitoring and early warning methods are mostly based on coolant oil content detection, which has low monitoring efficiency, poor early warning capabilities, and technical problems that affect production. Summary of the invention
[0003] The present invention provides an AI-based oil circuit plugging and early warning method and system to solve the technical problems of low monitoring efficiency, poor early warning capability, and production impact in the prior art, and to achieve the technical effect of improving monitoring efficiency and early warning capability and enhancing oil circuit reliability.
[0004] In a first aspect, the present invention provides an AI-based oil line leakage warning method, wherein the method comprises:
[0005] Acquire a target oil circuit, and deploy a monitoring and sensing network along the path direction of the target oil circuit, wherein the monitoring and sensing network includes a temperature sensor, a strain sensor, and a vibration sensor; based on a preset acquisition window, activate the monitoring and sensing network to acquire original sensor data, wherein the original sensor data includes an original strain signal, an original vibration signal, and an associated temperature series; perform adaptive filtering and denoising on the original sensor data to acquire denoised sensor data, and perform temperature correction on the denoised sensor data based on the temperature series to acquire standard monitoring data; acquire oil circuit leakage samples, wherein the oil circuit leakage samples include sample monitoring data, sample operation time, and sample leakage results; construct a neural network model, and perform supervised training using the oil circuit leakage samples as training data to acquire an oil circuit early warning model; input the standard monitoring data and the accumulated operation time of the target oil circuit into the oil circuit early warning model, and output the oil circuit early warning result, wherein the oil circuit early warning result includes a high-risk position, early warning confidence, and critical time.
[0006] In a second aspect, the present invention further provides an AI-based oil line leakage warning system, wherein the system comprises:
[0007] A sensing deployment module is used to acquire a target oil circuit and deploy a monitoring sensing network along the path direction of the target oil circuit, wherein the monitoring sensing network includes a temperature sensor, a strain sensor and a vibration sensor.
[0008] A sensor acquisition module, wherein the sensor acquisition module is used to activate the monitoring perception network to obtain raw sensor data based on a preset acquisition window, wherein the raw sensor data includes raw strain signals, raw vibration signals and associated temperature series.
[0009] The data quality correction module is used to perform adaptive filtering and denoising on the original sensor data to obtain denoised sensor data, and perform temperature correction on the denoised sensor data based on the temperature series to obtain standard monitoring data.
[0010] The sample acquisition module is used to acquire oil circuit leakage samples, and the oil circuit leakage samples include sample monitoring data, sample operation time and sample leakage results.
[0011] The early warning model construction module is used to construct a neural network model and perform supervised training using the oil circuit leakage sample as training data to obtain an oil circuit early warning model.
[0012] The oil circuit warning module is used to input the standard monitoring data and the accumulated operation time of the target oil circuit into the oil circuit warning model, and output the oil circuit warning result, which includes the high-risk position, warning confidence and critical time.
[0013] The present invention discloses an AI-based oil circuit plugging and early warning method and system, comprising: obtaining detailed information of a target oil circuit, and laying a monitoring perception network composed of temperature sensors, strain sensors and vibration sensors along the oil circuit path. Based on a preset acquisition window, the monitoring perception network is activated to obtain raw sensor data, including raw strain signals, raw vibration signals and temperature sequences. The raw sensor data is subjected to adaptive filtering and denoising processing to remove noise interference and obtain denoised sensor data. Then, the denoised sensor data is temperature corrected based on the temperature sequence to obtain standard monitoring data. Oil circuit leakage sample data is obtained, including sample monitoring data, sample operation time and sample leakage results. Using the oil circuit leakage sample data, a neural network model is constructed and trained to generate an oil circuit early warning model. The standard monitoring data and the cumulative operation time of the target oil circuit are input into the oil circuit early warning model, and the oil circuit early warning results are output, and the results include high-risk positions, early warning confidence and critical time. The AI-based oil circuit plugging and early warning method and system disclosed in the present invention solve the technical problems of low monitoring efficiency, poor early warning capability and production impact, and achieve the technical effects of improving monitoring efficiency and early warning capability and improving oil circuit reliability. BRIEF DESCRIPTION OF THE DRAWINGS
[0014] Figure 1 It is a flow chart of the oil line leakage plugging and early warning method based on AI of the present invention;
[0015] Figure 2 It is a structural schematic diagram of the oil line leakage warning system based on AI of the present invention.
[0016] Explanation of the reference numerals: perception layout module 11, sensor acquisition module 12, data quality correction module 13, sample acquisition module 14, early warning model construction module 15, oil circuit early warning module 16. DETAILED DESCRIPTION
[0017] The technical solution provided in the embodiments of the present invention is to solve the technical problems of low monitoring efficiency, poor early warning capability and production impact in the prior art. The overall idea adopted is as follows:
[0018] First, the detailed information of the target oil circuit is obtained, and a monitoring and sensing network is deployed along the path of the oil circuit. This network includes temperature sensors, strain sensors, and vibration sensors.
[0019] Then, based on the preset acquisition window, the monitoring perception network is activated to obtain the original sensing data, which includes the original strain signal, the original vibration signal and the temperature sequence related thereto.
[0020] Then, the original sensor data is subjected to adaptive filtering and denoising to remove noise interference, thereby obtaining denoised sensor data. Then, the denoised sensor data is temperature corrected based on the temperature series to obtain standard monitoring data.
[0021] Next, sample data of oil line leakage are obtained, including sample monitoring data, sample operation time and sample leakage results.
[0022] Furthermore, these oil circuit leakage sample data are used to build a neural network model and perform supervised training to obtain the oil circuit early warning model.
[0023] Finally, the standard monitoring data and the cumulative operation time of the target oil circuit are input into the oil circuit early warning model to output the oil circuit early warning results. These early warning results include high-risk locations, early warning confidence and critical time.
[0024] The above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods of the specification to better understand the above technical solution. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments of the present invention. It should be understood that the present invention is not limited to the example embodiments used only to explain the present invention. 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. In addition, it should be noted that, for the convenience of description, only the parts related to the present invention are shown in the drawings, rather than all of them.
[0025] Embodiment 1
[0026] Figure 1 The figure is a flow chart of the oil circuit leakage warning method based on AI of the present invention, wherein the method comprises:
[0027] A target oil circuit is acquired, and a monitoring and sensing network is deployed along the path direction of the target oil circuit, wherein the monitoring and sensing network includes a temperature sensor, a strain sensor, and a vibration sensor.
[0028] Specifically, the target oil circuit is the working oil cooling pipeline in the shell-and-tube oil cooler, including a tube bundle composed of heat exchange tubes. By obtaining a detailed drawing of the oil circuit, including all pipes, elbows, joints and other detailed information, the shape and path direction of the target oil circuit can be obtained, and then the key points can be identified, and corresponding monitoring sensors can be deployed to form a monitoring perception network.
[0029] Specifically, the monitoring and sensing network includes multiple temperature sensors, strain sensors and vibration sensors, which are used to monitor the temperature, strain and vibration distribution of the target oil circuit along the path direction.
[0030] Specifically, strain monitoring can reflect the stress of the oil circuit under different working conditions. Excessive stress and strain may cause structural damage or failure. At the same time, the oil circuit in long-term operation will produce fatigue due to cyclic loads. Monitoring strain helps to predict fatigue life and avoid safety problems caused by fatigue failure.
[0031] Specifically, temperature changes will cause thermal stress, which will in turn affect the strain and structural integrity of the oil circuit. At the same time, temperature changes will also affect the monitoring accuracy of strain sensors and vibration sensors. The temperature of the target pipeline along the path direction may vary, and thermal stress and ambient temperature analysis can be performed through temperature monitoring.
[0032] Specifically, vibrations in the oil circuit can cause structural fatigue and damage, which in turn can cause changes in the vibration characteristics of the heat exchange tube material. By monitoring the vibration frequency and amplitude, potential structural problems can be discovered and located.
[0033] In some embodiments, a target oil circuit is obtained, and a monitoring and sensing network is deployed along the path direction of the target oil circuit, including:
[0034] The target oil circuit is analyzed to obtain a target oil route map. The target oil route map is adaptively segmented to obtain multiple monitoring segments. According to the multiple monitoring segments, the temperature sensor, the strain sensor and the vibration sensor are arranged along the target oil circuit with the cross section where the midpoint of the segment is located as the reference plane to form the monitoring perception network.
[0035] Specifically, first, based on the shape of the target oil circuit and the flow direction of the oil in the oil circuit, a detailed structure diagram and path diagram of the target oil circuit are determined, including all pipes, elbows, joints and support points.
[0036] Specifically, based on the oil route map, adaptive segmentation is performed. The segmentation basis of adaptive segmentation includes: dividing the oil circuit into different monitoring segments according to the geometric shape of the oil circuit (such as straight pipe sections, elbows, joints, etc.). Identifying stress concentration points, such as elbows and joints, as separate monitoring segments or improving segmentation accuracy. According to the temperature changes in the oil circuit, the density of monitoring segments in areas with significant temperature changes is increased. Based on the vibration characteristics of the oil circuit, such as support points and fixed points are segmented. Through adaptive segmentation, different segment lengths are used in different areas to more accurately monitor key areas.
[0037] Specifically, temperature sensors are arranged around the outer wall of the oil circuit on the midpoint cross section of each monitoring segment, and the arrangement methods include pasting, clamping, etc., wherein the strain sensor includes a fiber grating strain sensor or a resistance strain gauge, etc. Preferably, a fiber grating strain sensor is used and encapsulated on the outer wall of the target pipeline by pasting.
[0038] Specifically, multiple temperature sensors, strain sensors and vibration sensors are mapped by coding to determine the monitoring point of each sensor. Exemplarily, a hierarchical coding format is defined, including segment number, position number, sensor type, sensor unique identification code, etc., and based on the above coding format, a coding mapping table is constructed to record the code, specific location and related information of each sensor.
[0039] By analyzing the target oil circuit, adaptively dividing it into segments, and placing temperature sensors, strain sensors, and vibration sensors on the midpoint cross-section of the segment, the formed monitoring perception network can accurately capture the temperature, strain, and vibration changes along the path of the oil circuit, providing comprehensive data support for subsequent early warning analysis.
[0040] In some implementations, the temperature sensor, the strain sensor, and the vibration sensor are arranged along the target oil circuit, further comprising:
[0041] Based on the multiple monitoring segments, extract the first monitoring segment. Take the cross section where the midpoint of the first segment of the first monitoring segment is located as the reference plane, and arrange the first main strain sensor and the first auxiliary strain sensor with an angle of 90 degrees along the outer wall of the target oil circuit. Arrange the first temperature sensor at the equal point between the first main strain sensor and the first auxiliary strain sensor. Arrange the first vibration sensor in alignment with the first temperature sensor. Traverse the multiple monitoring segments and arrange multiple temperature sensors, multiple strain sensors, and multiple vibration sensors.
[0042] Specifically, in order to ensure the accuracy of strain monitoring, the midpoint position of the monitoring segment is determined, and the cross-section where the midpoint is located is used as the reference plane. On the reference plane, two strain sensors are arranged along the outer wall of the target oil circuit, namely the main strain sensor and the auxiliary strain sensor. The main strain sensor and the auxiliary strain sensor are both arranged along the axial direction of the target oil circuit, and the angle between the two along the circumferential direction of the outer wall of the target oil circuit is 90 degrees to capture strain information in different directions.
[0043] Optionally, during the actual monitoring process, the principal stress and the auxiliary stress corresponding to the principal strain sensor and the auxiliary strain sensor are calculated respectively, and the square root of the sum of the principal stress and the auxiliary stress is taken as the oil circuit stress at the point.
[0044] Specifically, temperature sensors are arranged at equally divided points along the circumference of the outer wall of the target oil circuit between the main strain sensor and the auxiliary strain sensor to ensure the accuracy and representativeness of the temperature monitoring data. Then, vibration sensors are arranged at positions on the outer wall of the target oil circuit opposite to the temperature sensors to reduce the interference of sensor arrangement on vibration monitoring.
[0045] Based on a preset acquisition window, the monitoring sensing network is activated to obtain raw sensor data, wherein the raw sensor data includes raw strain signals, raw vibration signals and associated temperature series.
[0046] Specifically, a time window for data collection is set so that the sensor is activated and data is collected within a specified time period. The setting of the collection window is determined based on the monitoring requirements of the target scene, including the collection frequency and collection duration. The collection frequency determines the accuracy of the acquired raw sensor data, and the collection duration determines the representativeness and universality of the acquired raw sensor data. In other words, the raw sensor data acquired under a longer collection duration can better reflect the characteristic data of the target oil circuit operation.
[0047] Specifically, within a preset acquisition window, the monitoring and sensing network deployed on the target oil circuit is activated, and all sensors are synchronously activated to start collecting data at the same time point so that the data have time consistency. Exemplarily, an external clock source such as a crystal oscillator is used as an acquisition clock source to achieve synchronous acquisition of sensor data.
[0048] Specifically, the original sensing data includes: the original strain signal, which is the oil circuit surface strain change data recorded by the strain sensor. The original vibration signal is the oil circuit vibration condition data recorded by the vibration sensor. The associated temperature series refers to the oil circuit temperature data at multiple monitoring points recorded by the temperature sensor, which serves as the reference data for strain and vibration signals.
[0049] Adaptively filter and denoise the original sensor data to obtain denoised sensor data, and perform temperature correction on the denoised sensor data based on the temperature series to obtain standard monitoring data.
[0050] Specifically, since the original sensor data will be interfered by various noises during the acquisition process, it is necessary to denoise it through adaptive filtering technology to extract the real signal data. Among them, adaptive filtering is a method of dynamically adjusting filter parameters, which can automatically adjust the filter according to the characteristics of the input signal to achieve the best denoising effect. Optionally, the adaptive filtering method includes wavelet threshold filtering, Kalman filtering, empirical mode decomposition, etc.
[0051] Furthermore, since temperature changes will affect the output of the sensor, it is necessary to perform temperature correction on the denoised sensor data to obtain standardized monitoring data. Specifically, each sensor has its own specific temperature sensitivity curve, which reflects the effect of temperature on the measurement results can be eliminated through the correction curve at different temperatures.
[0052] In some embodiments, performing adaptive filtering and denoising on the raw sensor data includes:
[0053] Adaptively decompose the noise mode of the original sensor data to obtain an eigenmode component set. Evaluate the correlation between multiple eigenmode components in the eigenmode component set and the original sensor data respectively, and calculate and obtain multiple correlation metrics. According to a preset correlation threshold, perform threshold screening on the multiple correlation metrics to determine a related sensor data set. Filter and reduce noise on the related sensor data set based on wavelet transform to obtain filtered sensor data. Update the filtered sensor data to the original sensor data for data reconstruction to obtain the denoised sensor data.
[0054] Specifically, first, based on the adaptive noise modal classification method, the original sensor data is decomposed into a number of intrinsic mode components (IMFs), wherein each intrinsic mode component (IMF) represents a signal with different frequency components. Exemplarily, the adaptive noise modal classification method includes empirical mode decomposition (EMD) and adaptive noise empirical mode decomposition (CEEMDAN).
[0055] Specifically, based on the Pearson correlation coefficient, the correlation between each intrinsic mode component (IMF) and the original sensing data is calculated, and the obtained multiple correlation coefficients are stored as multiple correlation metrics. Then, according to a preset correlation threshold, the intrinsic mode components (IMF) with a high correlation with the original signal are screened out to form a related sensing data set. The correlation threshold is determined based on experiments or empirical values.
[0056] Specifically, the intrinsic mode components (IMFs) in the relevant sensor data set can be considered to contain a high content of noise, so wavelet transform is used to filter and reduce noise on the relevant sensor data set to further remove noise and retain useful signal components.
[0057] Exemplarily, wavelet threshold filtering is performed on the relevant sensor data set to obtain filtered sensor data. First, based on the data type in the relevant sensor data set, a suitable wavelet basis function is configured and selected, wherein the Molet wavelet is used as the wavelet basis function for vibration data, and the Daubechies series wavelet is used as the wavelet basis function for strain data. Then, according to the preset number of decomposition layers, the relevant sensor data set is subjected to multi-scale decomposition based on wavelet transform to obtain wavelet coefficients. Furthermore, according to the threshold function and the wavelet coefficient, the wavelet coefficient (corresponding to noise) less than the threshold is set to zero, thereby achieving the purpose of noise reduction. Finally, the wavelet coefficients after filtering and noise reduction are converted back to the original domain through inverse wavelet transform to obtain filtered sensor data.
[0058] In some embodiments, performing temperature correction on the denoised sensor data based on the temperature series includes:
[0059] Based on the temperature sensor series, the oil circuit temperature sequence is obtained. The temperature sensitivity curves of the strain sensor and the vibration sensor are obtained, and based on the temperature sensitivity curves and the oil circuit temperature sequence, the denoised sensor data is temperature corrected to obtain standard monitoring data. Specifically, first, the temperature sensitivity curve of each sensor is determined based on the experiment, and the curve characterizes the corresponding changes in the output signal of the sensor at different temperatures. Exemplarily, the linear temperature sensitivity curve is expressed as:
[0060] ;
[0061] in, is the signal after temperature correction. is the original signal after denoising. is the temperature sensitivity coefficient. is the current temperature. is the reference temperature.
[0062] Furthermore, based on the temperature series recorded by the temperature sensor, the denoised strain data and vibration data are corrected according to the current temperature and the temperature sensitivity curve. The temperature-corrected sensor data can more accurately reflect the actual working conditions and eliminate the influence of temperature changes on the sensor output signal, thereby obtaining standard monitoring data.
[0063] An oil circuit leakage sample is obtained, wherein the oil circuit leakage sample includes sample monitoring data, sample operation time and sample leakage result.
[0064] In some embodiments, obtaining an oil line leakage sample includes:
[0065] Analyze the target oil circuit and obtain the material property data of the target oil circuit. Based on the material property data, interact with big data to obtain homologous sample data, which includes homologous sample monitoring data, homologous sample operation time and homologous sample leakage results. Based on the target oil circuit early warning requirements, redefine the leakage of the homologous sample data, update the homologous sample leakage results, and obtain the oil circuit leakage sample.
[0066] Specifically, first, the target oil circuit is analyzed in detail to obtain its material property data, including material type, oil circuit structure, oil circuit size specifications, etc. Then, based on the existing big data resources, the data of other oil circuits with the same or similar material properties as the target oil circuit are queried to obtain homologous sample data.
[0067] Specifically, the homologous sample data includes homologous sample monitoring data, homologous sample operation time and homologous sample leakage results. Among them, the homologous sample monitoring data refers to the temperature monitoring data, strain monitoring data and vibration monitoring data of the homologous oil circuit under specified working conditions. The homologous sample operation time refers to the operation time of the homologous oil circuit under the specified working conditions, which is related to the fatigue accumulation degree of the homologous oil circuit. The homologous sample leakage results are the leakage events that occurred in the homologous oil circuit and their detailed information, including leakage time, leakage amount, leakage location, etc.
[0068] Specifically, the oil circuits in different application scenarios have different leakage control needs and leakage classification rules, and the determination of the leakage level based on the leakage classification rules will affect the judgment and warning of the leakage. Then, the leakage of the homologous sample data is redefined according to the warning requirements of the target oil circuit. Exemplarily, first, understand the leakage classification rules of the target oil circuit, including the judgment criteria of the size, location, frequency of occurrence and other factors of the leakage. Then, according to the leakage classification rules of the target oil circuit, the homologous sample data is re-evaluated according to the size, location, frequency of occurrence and other factors of the leakage, and the new leakage level and leakage result are determined. Finally, the leakage result of the homologous sample is updated based on the new leakage level and leakage result, that is, the leakage label of the homologous sample is modified to reflect the warning requirements of the target oil circuit.
[0069] Through the above steps, it is possible to obtain oil circuit leakage samples that better meet the target oil circuit early warning requirements, thereby improving the accuracy and reliability of the oil circuit leakage early warning.
[0070] A neural network model is constructed, and supervised training is performed using the oil circuit leakage samples as training data to obtain an oil circuit early warning model.
[0071] Exemplarily, an oil circuit early warning model is constructed and trained based on a long short-term memory network LSTM. First, sample monitoring data and sample operation time are used as input sample data, and sample leakage results are used as output sample data for data standardization. Then, the continuous monitoring data is divided into a time series of fixed length to meet the LSTM input requirements, and the leakage amount in the sample leakage results is encoded to obtain a training data set. Next, an oil circuit early warning model is constructed based on the LSTM neural network model, wherein the oil circuit early warning model includes: an input layer for receiving processed time series data, which usually includes temperature, strain, vibration monitoring data and operation time. Multiple LSTM layers are used to process long-term dependencies of time series data. A fully connected layer is used to map the output of the LSTM layer to the final prediction result. Output layer: divided into parallel classification layers (for predicting binary leakage situations) or regression layers (for predicting leakage amounts).
[0072] Furthermore, the training data set is divided into a training set and a validation set, and the training data is used for model training. Through the mean square error (MSE) and cross entropy loss functions, optimizers such as Adam and RMSprop are used to adjust the weights and biases to reduce the value of the loss function. Then, the model performance is evaluated on the validation set, the overfitting phenomenon during the training process is monitored, and hyperparameters are adjusted when necessary. If the model performance (such as accuracy, F1 score, recall rate, etc.) meets the target oil circuit requirements, the oil circuit warning model is output.
[0073] By training the model, oil line leakage can be predicted more accurately, reducing unexpected downtime and maintenance costs. In addition, the predictive information provided by the early warning model can help formulate more scientific maintenance plans and extend the service life of equipment.
[0074] The standard monitoring data and the accumulated operation time of the target oil circuit are input into the oil circuit early warning model, and an oil circuit early warning result is output. The oil circuit early warning result includes a high-risk position, a warning confidence and a critical time.
[0075] Specifically, input the standard monitoring data and the cumulative operating time of the target oil circuit into the oil circuit early warning model, perform oil circuit early warning, and obtain the oil circuit early warning results. Among them, the high-risk position refers to the monitoring segment position where the leakage confidence is greater than the confidence threshold. The warning confidence is the confidence level of the model in its prediction, expressed in the form of probability. The critical time is the time when the leakage is expected to occur. The critical time is jointly affected by the warning confidence, the cumulative operating time of the target oil circuit, and the operating conditions of the target oil circuit, and is updated in real time.
[0076] Optionally, based on the above-mentioned high position, the corresponding multiple monitoring segments perform endpoint monitoring, including increasing the collection frequency of the monitoring sensors, increasing the number of monitoring sensors, expanding the monitoring of adjacent segments of the multiple monitoring segments corresponding to the high position, etc.
[0077] In some embodiments, after outputting the oil circuit warning result, the method further includes:
[0078] Based on the high-risk location and the warning confidence, a multi-path plugging warning is performed. According to the critical time, combined with the operation plan of the target oil circuit, the plugging solution and the plugging time node are matched and obtained to perform forward-looking plugging maintenance.
[0079] Specifically, multi-path leak plugging warning based on high-risk locations and warning confidence levels refers to leak plugging warnings through a variety of warning channels, including visual display of high-risk locations and warning confidence levels, automatic triggering of sound and / or visual alarms when the warning confidence level reaches a set threshold, setting alarm thresholds of different levels, performing different alarm processing for different confidence levels, setting up SMS and email notification systems, and sending warning information to maintenance personnel, including high-risk locations, warning confidence levels, and processing suggestions.
[0080] Specifically, obtain the operation plan of the target oil circuit, including operation schedule, downtime, etc. Combined with the operation plan, select the most suitable plugging solution, and arrange the plugging maintenance time node in the oil circuit operation plan to ensure that maintenance is carried out during non-critical periods to reduce the impact on production.
[0081] The above method steps can more comprehensively deal with potential leakage risks and improve the safety of the system through multi-path plugging and early warning. According to the matching of critical time and operation plan, a forward-looking maintenance plan is formulated to help reduce equipment downtime and maintenance costs, while improving maintenance efficiency and ensuring equipment reliability and stability.
[0082] In summary, the AI-based oil line leakage warning method provided by the present invention has the following technical effects:
[0083] By obtaining detailed information of the target oil circuit, a monitoring perception network consisting of temperature sensors, strain sensors and vibration sensors is arranged along the oil circuit path. Based on the preset acquisition window, the monitoring perception network is activated to obtain the original sensor data, including the original strain signal, the original vibration signal and the temperature sequence. The original sensor data is subjected to adaptive filtering and denoising to remove noise interference and obtain denoised sensor data. Then, the denoised sensor data is temperature corrected based on the temperature sequence to obtain standard monitoring data. The oil circuit leakage sample data is obtained, including sample monitoring data, sample operation time and sample leakage results. Using the oil circuit leakage sample data, a neural network model is constructed and trained to generate an oil circuit early warning model. The standard monitoring data and the cumulative operation time of the target oil circuit are input into the oil circuit early warning model, and the oil circuit early warning results are output, including high-risk locations, warning confidence and critical time. Thus, the technical effect of improving monitoring efficiency and early warning capability and improving oil circuit reliability is achieved.
[0084] Embodiment 2
[0085] Figure 2 Schematic diagram of the structure of the AI-based oil leakage warning system of the present invention. Figure 1 The flowchart of the AI-based oil line leakage warning method of the present invention can be shown as follows: Figure 2 The structure shown is implemented.
[0086] Based on the same concept as the AI-based oil line leakage warning method in the above embodiment, the AI-based oil line leakage warning system provided by the present invention includes:
[0087] The sensing deployment module 11 is used to acquire the target oil circuit and deploy a monitoring sensing network along the path direction of the target oil circuit, wherein the monitoring sensing network includes a temperature sensor, a strain sensor and a vibration sensor.
[0088] The sensor acquisition module 12 is used to activate the monitoring perception network to obtain raw sensor data based on a preset acquisition window, wherein the raw sensor data includes raw strain signals, raw vibration signals and associated temperature series.
[0089] The data quality correction module 13 is used to perform adaptive filtering and denoising on the original sensor data to obtain denoised sensor data, and perform temperature correction on the denoised sensor data based on the temperature series to obtain standard monitoring data.
[0090] The sample acquisition module 14 is used to acquire oil circuit leakage samples, and the oil circuit leakage samples include sample monitoring data, sample operation time and sample leakage results.
[0091] The early warning model building module 15 is used to build a neural network model and perform supervised training using the oil circuit leakage samples as training data to obtain an oil circuit early warning model.
[0092] The oil circuit warning module 16 is used to input the standard monitoring data and the accumulated operation time of the target oil circuit into the oil circuit warning model, and output the oil circuit warning result, which includes the high-risk position, warning confidence and critical time.
[0093] Among them, the sensing layout module 11 includes:
[0094] The line map acquisition unit is used to analyze the target oil circuit and acquire the target oil route map.
[0095] The adaptive segmentation unit is used to adaptively segment the target oil roadmap to obtain multiple monitoring segments.
[0096] The sensor arrangement unit is used to arrange the temperature sensor, the strain sensor and the vibration sensor along the target oil circuit according to the plurality of monitoring segments and with the cross section where the midpoint of the segment is located as a reference plane to form the monitoring sensing network.
[0097] Wherein, the data quality correction module 13 includes:
[0098] The modal decomposition unit is used to adaptively decompose the noise modal of the original sensor data to obtain an eigenmodal component set.
[0099] The correlation calculation unit is used to respectively evaluate the correlation between a plurality of intrinsic mode components in the intrinsic mode component set and the original sensing data, and calculate and obtain a plurality of correlation metrics.
[0100] The correlation selection unit is used to perform threshold screening on the multiple correlation metrics according to a preset correlation threshold to determine the relevant sensor data set.
[0101] The wavelet filtering unit is used to filter and reduce noise on the relevant sensor data set based on wavelet transform to obtain filtered sensor data.
[0102] A data reconstruction unit is used to update the filtered sensor data to the original sensor data for data reconstruction to obtain the denoised sensor data.
[0103] The data quality correction module 13 further includes:
[0104] The temperature calling unit is used to obtain the oil circuit temperature sequence based on the temperature sensor series.
[0105] The temperature correction unit is used to obtain the temperature sensitivity curves of the strain sensor and the vibration sensor, and based on the temperature sensitivity curves and the oil circuit temperature sequence, perform temperature correction on the denoised sensor data to obtain standard monitoring data.
[0106] The sample acquisition module 14 includes:
[0107] The material analysis unit is used to analyze the target oil circuit and obtain material characteristic data of the target oil circuit.
[0108] The sample collection unit is used to obtain homologous sample data according to the material characteristic data and the interactive big data, and the homologous sample data includes homologous sample monitoring data, homologous sample operation time and homologous sample leakage results.
[0109] The sample adaptation and updating unit is used to redefine the leakage of the homologous sample data based on the target oil circuit early warning requirement, update the leakage result of the homologous sample, and obtain the oil circuit leakage sample.
[0110] Furthermore, the system also includes an early warning response unit, which is used to:
[0111] Based on the high-risk location and the warning confidence, a multi-path plugging warning is performed. According to the critical time, combined with the operation plan of the target oil circuit, the plugging solution and the plugging time node are matched and obtained to perform forward-looking plugging maintenance.
[0112] In some embodiments, the sensor arrangement unit includes:
[0113] The segment extraction unit is used to extract the first monitoring segment based on the multiple monitoring segments.
[0114] The strain sensor arrangement unit is used to arrange a first main strain sensor and a first auxiliary strain sensor with an included angle of 90 degrees along the outer wall of the target oil circuit, taking the cross section where the midpoint of the first segment of the first monitoring segment is located as a reference plane.
[0115] The temperature sensing arrangement unit is used to arrange a first temperature sensor at a point where the first main strain sensor and the first auxiliary strain sensor are equally divided.
[0116] The vibration sensor arrangement unit is used to arrange the first vibration sensor in accordance with the first temperature sensor.
[0117] The multi-segment layout unit is used to traverse the plurality of monitoring segments and layout the plurality of temperature sensors, the plurality of strain sensors and the plurality of vibration sensors.
[0118] It should be understood that the embodiments mentioned in this specification focus on their differences from other embodiments. The specific embodiments in the aforementioned embodiment one are also applicable to the AI-based oil circuit leakage warning system described in embodiment two. For the sake of brevity of the specification, they will not be further elaborated here.
[0119] It should be understood that the embodiments disclosed in the present invention and the above description can enable those skilled in the art to use the present invention to implement the present invention. At the same time, the present invention is not limited to the above-mentioned embodiments. It should be understood that those skilled in the art can still modify the technical solutions recorded in the above-mentioned embodiments, or replace some of the technical features therein by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention.
Claims
1. AI-based oil line leakage warning method, characterized in that: The method comprises: Acquire a target oil circuit, and deploy a monitoring and sensing network along the path direction of the target oil circuit, wherein the monitoring and sensing network includes a temperature sensor, a strain sensor, and a vibration sensor; Based on a preset acquisition window, the monitoring sensing network is activated to obtain raw sensor data, wherein the raw sensor data includes raw strain signals, raw vibration signals and associated temperature series; Adaptively filtering and denoising the original sensor data to obtain denoised sensor data, and performing temperature correction on the denoised sensor data based on the temperature series to obtain standard monitoring data; Obtaining an oil circuit leakage sample, wherein the oil circuit leakage sample includes sample monitoring data, sample operation time and sample leakage result; Constructing a neural network model, and performing supervised training using the oil circuit leakage sample as training data to obtain an oil circuit early warning model; Input the standard monitoring data and the cumulative operation time of the target oil circuit to the oil circuit early warning model, and output the oil circuit early warning result, wherein the oil circuit early warning result includes a high-risk position, a warning confidence and a critical time; Among them, obtaining oil line leakage samples includes: Analyze the target oil circuit and obtain the material characteristic data of the target oil circuit; According to the material characteristic data and the interactive big data, homologous sample data is obtained, wherein the homologous sample data includes homologous sample monitoring data, homologous sample operation time and homologous sample leakage results; Based on the target oil circuit early warning requirement, redefine the leakage of the homologous sample data, update the leakage result of the homologous sample, and obtain the oil circuit leakage sample; Wherein, after outputting the oil circuit warning result, the method further includes: Based on the high-risk location and the warning confidence, a multi-path leakage plugging warning is performed; According to the critical time, combined with the operation plan of the target oil circuit, the plugging scheme and the plugging time node are matched and obtained to perform forward-looking plugging maintenance; Among them, the target oil circuit is obtained, and a monitoring and sensing network is deployed along the path direction of the target oil circuit, including: Analyze the target oil circuit and obtain the target oil route map; Adaptively segmenting the target oil roadmap to obtain multiple monitoring segments; According to the plurality of monitoring segments, the temperature sensor, the strain sensor and the vibration sensor are arranged along the target oil circuit with the cross section where the midpoint of the segment is located as the reference plane to form the monitoring sensing network; Wherein, the temperature sensor, the strain sensor and the vibration sensor are arranged along the target oil circuit, and further comprises: Based on the plurality of monitoring segments, extracting a first monitoring segment; Taking the cross section where the midpoint of the first segment of the first monitoring segment is located as a reference plane, a first main strain sensor and a first auxiliary strain sensor are arranged with an angle of 90 degrees along the outer wall of the target oil circuit; Arrange a first temperature sensor at a point where the first main strain sensor and the first auxiliary strain sensor are equally divided; Arrange a first vibration sensor in alignment with the first temperature sensor; Traverse the plurality of monitoring sections and deploy a plurality of the temperature sensors, a plurality of the strain sensors, and a plurality of the vibration sensors.
2. The AI-based oil line leak warning method according to claim 1, characterized in that: Adaptively filtering and denoising the raw sensor data, comprising: Adaptively decomposing the noise mode of the original sensor data to obtain an eigenmode component set; Respectively evaluating the correlations between a plurality of eigenmode components in the eigenmode component set and the original sensing data, and calculating and obtaining a plurality of correlation metrics; According to a preset correlation threshold, threshold screening is performed on the multiple correlation metrics to determine a relevant sensor data set; Filtering and denoising the relevant sensor data set based on wavelet transform to obtain filtered sensor data; The filtered sensor data is updated to the original sensor data for data reconstruction to obtain the denoised sensor data.
3. The AI-based oil line leak warning method according to claim 2, characterized in that: Performing temperature correction on the denoised sensor data based on the temperature series includes: Based on the temperature sensor series, obtaining an oil circuit temperature sequence; The temperature sensitivity curves of the strain sensor and the vibration sensor are obtained, and based on the temperature sensitivity curves and the oil circuit temperature sequence, the denoised sensor data is temperature corrected to obtain standard monitoring data.
4. AI-based oil line leakage warning system, characterized by: For executing the method according to any one of claims 1 to 3, the system comprises: A sensing layout module, the sensing layout module is used to obtain a target oil circuit and layout a monitoring sensing network along the path direction of the target oil circuit, wherein the monitoring sensing network includes a temperature sensor, a strain sensor and a vibration sensor; A sensor acquisition module, the sensor acquisition module is used to activate the monitoring perception network to obtain raw sensor data based on a preset acquisition window, wherein the raw sensor data includes raw strain signals, raw vibration signals and associated temperature series; A data quality correction module, the data quality correction module is used to perform adaptive filtering and denoising on the original sensor data to obtain denoised sensor data, and perform temperature correction on the denoised sensor data based on the temperature series to obtain standard monitoring data; A sample acquisition module, the sample acquisition module is used to acquire oil circuit leakage samples, the oil circuit leakage samples include sample monitoring data, sample operation time and sample leakage results; An early warning model building module, the early warning model building module is used to build a neural network model, and use the oil circuit leakage sample as training data for supervised training to obtain an oil circuit early warning model; The oil circuit warning module is used to input the standard monitoring data and the accumulated operation time of the target oil circuit into the oil circuit warning model, and output the oil circuit warning result, which includes the high-risk position, warning confidence and critical time.
