Intelligent monitoring method and system applied to high and medium pressure valves

Abstract

The invention discloses an intelligent monitoring method and system applied to a high-pressure and medium-pressure valve, and the method comprises the steps: collecting data through an embedded multi-sensor network according to the real-time operation state and environment pressure data of the high-pressure and medium-pressure valve, and carrying out the comprehensive processing of the collected data through a self-adaptive data fusion algorithm, so as to obtain the comprehensive working characteristics of the valve; based on the working characteristics of the valve, identifying a potential fault mode of the high and medium pressure valve in real time by adopting an anomaly detection algorithm based on machine learning; and according to the potential fault mode, a corresponding maintenance and repair plan is generated based on a decision tree algorithm, and the maintenance and repair plan is fed back to an operator through a mobile terminal application, so that intelligent monitoring of the high and medium pressure valve is realized. According to the embodiment of the invention, real-time monitoring and fault early warning of the valve can be realized, optimization of maintenance and repair plans is facilitated, and the intelligent level of valve management is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of high and medium pressure valves, and in particular to an intelligent monitoring method and system applied to high and medium pressure valves. Background Art

[0002] In modern industrial production, high and medium pressure valves are important control components and are widely used in the fields of oil, natural gas, chemical industry, electricity, water supply, etc. They play a vital role in the circulation and control of the medium, ensuring the safety and efficiency of the production process. Therefore, the safety, stability and reliability of valves have become important issues that cannot be ignored in industrial production.

[0003] Traditionally, the monitoring and maintenance of high and medium pressure valves rely on manual inspections and regular maintenance. This method is not only highly dependent on the operator's experience, but also due to the frequency and accuracy limitations of manual inspections, it is often impossible to detect potential valve failures in a timely manner, leading to possible safety accidents and economic losses. In addition, regular maintenance plans often fail to target specific operating conditions and actual needs, which may result in over-maintenance or under-maintenance, further affecting production efficiency and equipment life. Summary of the invention

[0004] The purpose of the present invention is to provide an intelligent monitoring method and system for high and medium pressure valves to address the deficiencies in the prior art. The method and system can utilize embedded multi-sensor networks, adaptive data fusion algorithms, and abnormality detection methods based on machine learning to not only achieve real-time monitoring and fault warning of valves, but also help optimize maintenance and overhaul plans and improve the intelligence level of valve management.

[0005] An embodiment of the present application provides an intelligent monitoring method for high and medium pressure valves, the method comprising: According to the real-time operating status and environmental pressure data of the high and medium pressure valves, data is collected through an embedded multi-sensor network, and the collected data is comprehensively processed using an adaptive data fusion algorithm to obtain comprehensive valve operating characteristics, wherein the multi-sensor network includes at least a pressure sensor, a temperature sensor, and a vibration sensor; Based on the valve operating characteristics, a machine learning-based anomaly detection algorithm is used to identify potential failure modes of the high and medium pressure valves in real time; According to the potential failure mode, a corresponding maintenance and overhaul plan is generated based on a decision tree algorithm, and the maintenance and overhaul plan is fed back to the operator through a mobile application to achieve intelligent monitoring of high and medium pressure valves.

[0006] Optionally, according to the real-time operating status and environmental pressure data of the high and medium pressure valves, data is collected through an embedded multi-sensor network, and the collected data is comprehensively processed using an adaptive data fusion algorithm to obtain comprehensive valve operating characteristics, including: For each sensor data stream, continuous wavelet transform is applied to extract features of different scales, which represent the change details and trends of the signal to capture multi-level information of the data; A perception game model is constructed, and each sensor is regarded as a participant in the game. Each sensor participates in the game according to the contribution and reliability of the data it collects, competing for its importance in the fusion process. Through iterative calculation, a Nash equilibrium point is reached, at which the data weight distribution of each sensor reaches the optimal state. The different scale features extracted from different sensors are input into the fuzzy logic system, and a fusion feature is output as a comprehensive valve working feature, wherein the fuzzy rule set is adjusted in real time according to the weights assigned by the perception game model to adapt to different operating states and environments, and the fuzzy reasoning process is executed. By matching the fuzzy set with the fuzzy rules, the different scale features of the sensor are converted into a comprehensive fusion feature to reflect the valve working state.

[0007] Optionally, based on the valve operating characteristics, using an abnormality detection algorithm based on machine learning to identify potential failure modes of the high and medium pressure valves in real time includes: Inputting the valve operation feature into a deep autoencoder network to learn its low-dimensional representation, and using the bottleneck layer output of the autoencoder as a feature embedding, the embedding represents the core features of the data and can distinguish between normal and abnormal modes, wherein the autoencoder captures the potential structural features of the data by reconstructing errors; Based on the feature embedding, anomaly detection is performed using an isolation forest algorithm, wherein the isolation forest constructs a decision tree by randomly selecting features and segmentation points to identify abnormal data points that are different from normal patterns; The causal relationship between the abnormal detection results of the isolation forest algorithm and the operating conditions of the high and medium pressure valves is analyzed by using a causal reasoning network to identify potential failure modes. The causal reasoning network reveals the driving factors behind the abnormal phenomena through a causal graph model.

[0008] Optionally, generating a corresponding maintenance and repair plan based on a decision tree algorithm according to the potential failure mode includes: The potential failure mode is represented as a binary vector using one-hot encoding, and combined with the numerical features of the collected data to form a complete feature vector as the input of the decision tree model; Starting from the root node of the decision tree, the tree structure is traversed layer by layer. For each node, the feature vectors are compared using the feature partitioning conditions of the node. If a feature value in the feature vector meets a preset condition, the traversal continues downward along the branch that meets the preset condition. If a feature value does not meet a preset condition, the traversal continues downward along another branch until a leaf node is reached. After path tracing, a leaf node is eventually reached, which represents a specific maintenance and overhaul recommendation, and a complete maintenance and overhaul plan is generated based on each leaf node reached by the tracking.

[0009] Another embodiment of the present application provides an intelligent monitoring system for high and medium pressure valves, the system comprising: An acquisition module, for acquiring data through an embedded multi-sensor network according to the real-time operating status and environmental pressure data of the high and medium pressure valves, and comprehensively processing the acquired data using an adaptive data fusion algorithm to obtain comprehensive valve operating characteristics, wherein the multi-sensor network includes at least a pressure sensor, a temperature sensor, and a vibration sensor; An identification module, for identifying potential failure modes of the high and medium pressure valves in real time based on the valve operating characteristics and using an abnormality detection algorithm based on machine learning; The monitoring module is used to generate corresponding maintenance and overhaul plans based on the decision tree algorithm according to the potential failure mode, and feed back the maintenance and overhaul plans to the operator through the mobile application to realize intelligent monitoring of high and medium pressure valves.

[0010] Yet another embodiment of the present application provides a storage medium, wherein the storage medium stores a computer program, wherein the computer program is configured to execute any of the above methods when running.

[0011] Yet another embodiment of the present application provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute any of the methods described above.

[0012] Compared with the prior art, the present invention provides an intelligent monitoring method for high and medium pressure valves. According to the real-time operating status and environmental pressure data of the high and medium pressure valves, data is collected through an embedded multi-sensor network, and the collected data is comprehensively processed using an adaptive data fusion algorithm to obtain comprehensive valve operating characteristics; based on the valve operating characteristics, an abnormality detection algorithm based on machine learning is used to identify the potential failure modes of the high and medium pressure valves in real time; according to the potential failure modes, a corresponding maintenance and overhaul plan is generated based on a decision tree algorithm, and the maintenance and overhaul plan is fed back to the operator through a mobile application to realize intelligent monitoring of the high and medium pressure valves, thereby being able to utilize an embedded multi-sensor network, an adaptive data fusion algorithm and an abnormality detection method based on machine learning, which can not only realize real-time monitoring and fault warning of valves, but also help to optimize maintenance and overhaul plans and improve the intelligence level of valve management. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 A hardware structure block diagram of a computer terminal for an intelligent monitoring method for high and medium pressure valves provided in an embodiment of the present invention; Figure 2 A schematic flow chart of an intelligent monitoring method for high and medium pressure valves provided in an embodiment of the present invention; Figure 3 A schematic diagram of the structure of an intelligent monitoring system for high and medium pressure valves provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0014] The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, but should not be construed as limiting the present invention.

[0015] The embodiment of the present invention first provides an intelligent monitoring method for high and medium pressure valves. The method can be applied to electronic devices, such as computer terminals, specifically ordinary computers, etc.

[0016] The following describes it in detail by taking running on a computer terminal as an example. Figure 1 The hardware structure block diagram of a computer terminal for an intelligent monitoring method for high and medium pressure valves provided by an embodiment of the present invention is shown in FIG. Figure 1 As shown, the computer device includes a processor, a memory, and a network interface connected via a system bus, wherein the memory may include a non-volatile storage medium and an internal memory.

[0017] The non-volatile storage medium can store an operating system and a computer program. The computer program includes program instructions, and when the program instructions are executed, the processor can execute any intelligent monitoring method applied to high and medium pressure valves.

[0018] The processor is used to provide computing and control capabilities and support the operation of the entire computer equipment.

[0019] The internal memory provides an environment for the operation of the computer program in the non-volatile storage medium. When the computer program is executed by the processor, the processor can execute any intelligent monitoring method applied to high and medium pressure valves.

[0020] The network interface is used for network communication, such as sending assigned tasks, etc. Those skilled in the art will understand that Figure 1 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0021] It should be understood that the processor may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0022] See also Figure 2 The embodiment of the present invention provides an intelligent monitoring method for high and medium pressure valves, which may include the following steps: S201, according to the real-time operating status and environmental pressure data of the high and medium pressure valves, collect data through an embedded multi-sensor network, and use an adaptive data fusion algorithm to comprehensively process the collected data to obtain comprehensive valve operating characteristics, wherein the multi-sensor network includes at least a pressure sensor, a temperature sensor, and a vibration sensor; In this step, the system collects real-time operating status and environmental pressure data of high and medium pressure valves through an embedded multi-sensor network. The multi-sensor network includes pressure sensors, temperature sensors, and vibration sensors, which work together to provide comprehensive environmental and operating condition data. The collected data is comprehensively processed by an adaptive data fusion algorithm to obtain the comprehensive working characteristics of the valve. The adaptive data fusion algorithm can dynamically adjust the data processing strategy to adapt to different operating conditions and environmental changes, thereby ensuring the accuracy and reliability of the data.

[0023] The significance of this step is that through the collaborative work of the multi-sensor network, the operating status and environmental conditions of the valve can be monitored in real time. Through the application of adaptive data fusion algorithm, the system can effectively integrate data from different sensors to generate a comprehensive valve operating feature. This feature provides reliable basic data support for subsequent fault detection and maintenance plans, ensuring the safe and efficient operation of the valve in complex environments.

[0024] Specifically, for each sensor data stream, continuous wavelet transform is applied to extract features of different scales, which represent the change details and trends of the signal to capture multi-level information of the data; In this step, the system applies continuous wavelet transform (CWT) to the data stream of each sensor to extract the multi-scale features of the signal. Wavelet transform is a signal processing technique that can analyze signals in both time and frequency domains to capture their changing details and trends. In this way, the system can identify multi-level information in the data and provide rich features for subsequent data fusion and analysis.

[0025] The significance of applying continuous wavelet transform is that it can extract more detailed and comprehensive feature information from sensor data. These features not only reflect the instantaneous changes of the signal, but also reveal its long-term trend, providing a more accurate and reliable basis for subsequent fault detection and diagnosis.

[0026] In the specific implementation, suppose we have a data stream from a pressure sensor. First, the system preprocesses the data stream to remove noise and outliers. Then, a continuous wavelet transform is applied to decompose the data into components of different scales. Each component represents the change of the signal in a different frequency range. By analyzing these components, the system can identify detailed changes (such as short-term fluctuations) and trend changes (such as long-term pressure changes) in the pressure signal. For example, when an abnormal fluctuation in a certain frequency range is detected, the system can infer possible mechanical vibration or external interference.

[0027] For example, suppose we use a pressure sensor in an industrial environment to monitor the working status of a valve. The pressure sensor collects data at a high frequency, generating a time series. In order to extract useful information from it, we first perform simple preprocessing on the data, such as removing noise and outliers. Next, we apply continuous wavelet transform (CWT) to analyze the sequence data. By selecting a suitable wavelet basis (such as Morlet wavelet), the original signal can be decomposed into wavelet coefficients of different frequency components. These coefficients reflect the changes of the signal at different scales. For example, in a normal operation cycle, the wavelet transform may show a stable trend in the low-frequency region, while it may show abnormal oscillations in some high-frequency regions. These oscillations may indicate potential faults or external interference. By analyzing these features at different scales, we can understand the multi-level information of the signal and provide a basis for subsequent data fusion and analysis.

[0028] A perception game model is constructed, and each sensor is regarded as a participant in the game. Each sensor participates in the game according to the contribution and reliability of the data it collects, competing for its importance in the fusion process. Through iterative calculation, a Nash equilibrium point is reached, at which the data weight distribution of each sensor reaches the optimal state. In this step, the system constructs a perception game model, treating each sensor as a participant in the game. Each sensor participates in the game based on the contribution and reliability of its data, competing for importance in the data fusion process. Through iterative calculations, the system eventually reaches a Nash equilibrium point, at which the data weight distribution of each sensor reaches the optimal state.

[0029] The application of the perception game model enables the system to dynamically adjust the data weight of each sensor, ensuring that the most reliable and most contributing data is given a higher weight during the data fusion process. This dynamic adjustment mechanism improves the accuracy and robustness of data fusion, ensuring that the system can obtain the best comprehensive characteristics under different environments and operating conditions.

[0030] In the specific implementation, suppose we have pressure, temperature, and vibration sensors. First, the system evaluates the contribution and reliability of each sensor data. For example, the pressure sensor may contribute more when detecting valve sealing, while the temperature sensor is more important when detecting thermal expansion. Next, the system regards these sensors as participants in the game and constructs a perceptual game model. Through iterative calculations, the system adjusts the weight of each sensor until it reaches a Nash equilibrium point. At this point, the pressure sensor may receive a higher weight because its data is more reliable and important in the current environment.

[0031] For example, in a valve monitoring system, suppose we have three sensors: pressure sensor, temperature sensor, and vibration sensor. We need to determine the weight of each sensor in the data fusion process. First, we establish an initial contribution and reliability index for each sensor. For example, a pressure sensor may have a higher contribution in a high-pressure environment, while a temperature sensor may be more reliable when the ambient temperature changes greatly. Next, we build a perceptual game model in which each sensor participates in the game according to its index. Through multiple rounds of iterative calculations, the sensor will adjust its behavior according to the strategies of other sensors to gain more weight in the game. Finally, the game reaches a Nash equilibrium and the weight distribution of each sensor reaches the optimal state. For example, during the iteration process, the vibration sensor may have a lower weight because its data does not provide significant fault indications in this operation, while the pressure sensor has a higher weight because it provides important status information. This dynamic weight distribution mechanism ensures that the system optimizes the use of the data of each sensor under different conditions.

[0032] The different scale features extracted from different sensors are input into the fuzzy logic system, and a fusion feature is output as a comprehensive valve working feature, wherein the fuzzy rule set is adjusted in real time according to the weights assigned by the perception game model to adapt to different operating states and environments, and the fuzzy reasoning process is executed. By matching the fuzzy set with the fuzzy rules, the different scale features of the sensor are converted into a comprehensive fusion feature to reflect the valve working state.

[0033] In this step, the system inputs the multi-scale features extracted from different sensors into the fuzzy logic system. The fuzzy logic system performs a fuzzy reasoning process by matching fuzzy sets and fuzzy rules to transform these features into a comprehensive fusion feature. The fuzzy rule set is adjusted in real time according to the weights assigned by the perception game model to adapt to different operating states and environments.

[0034] The application of fuzzy logic system enables the system to handle uncertainty and ambiguity, generating a comprehensive feature that reflects the working status of the valve. By adjusting the fuzzy rule set in real time, the system can adapt to different operating conditions and environmental changes, improving the accuracy and flexibility of fault detection and diagnosis.

[0035] In the specific implementation, suppose we have extracted multi-scale features from pressure, temperature, and vibration sensors. The system first inputs these features into a fuzzy logic system. The fuzzy logic system contains a set of fuzzy rules, such as "If the pressure is high and the temperature is high, the valve may be overheated." These rules are adjusted according to the weights assigned by the perception game model to ensure that the most important features have the greatest impact on the final decision. Through the fuzzy reasoning process, the system transforms features at different scales into a comprehensive fusion feature that reflects the overall working status of the valve. For example, when the pressure and temperature are detected to be increased at the same time, the system may output a high-risk fusion feature, indicating that further inspection is needed.

[0036] For example, in this step, we use a fuzzy logic system to integrate features from different sensors. In the specific implementation, it is assumed that we have extracted multi-scale features from pressure, temperature and vibration sensors, and obtained their respective weights based on the perception game model. When we design a fuzzy logic system, we need to define fuzzy sets and fuzzy rules. For example, the definition of fuzzy sets includes "low pressure", "medium pressure", "high pressure", "low temperature", "high temperature", etc. A fuzzy rule may be like "if the pressure is high and the temperature is high, the risk level is high". According to the weights given by the game model, we adjust the influence of these rules in real time. After the extracted features are input into the fuzzy logic system, the system performs the reasoning process according to the fuzzy rule set. Finally, a comprehensive fusion feature is output. For example, at a certain moment, the pressure and temperature features show abnormally high values. The system obtains a high-risk fusion feature value by applying the rules "high pressure" and "high temperature", indicating that the valve needs to be inspected and maintained. This process ensures that the comprehensive features can accurately reflect the current working status of the valve and provide a reliable basis for decision-making.

[0037] S202, based on the valve operating characteristics, using an abnormality detection algorithm based on machine learning to identify potential failure modes of the high and medium pressure valves in real time; In this step, the system uses machine learning anomaly detection algorithms to identify potential failure modes of high and medium pressure valves in real time based on valve operating characteristics. Specifically, the system first obtains the comprehensive operating characteristics of the valve from the multi-sensor network, which reflect the current operating status of the valve. The system then uses machine learning algorithms to analyze these characteristics and identify abnormal modes that are different from normal operating modes. In this way, the system is able to detect potential problems before failures occur, thereby improving the safety and reliability of the valve.

[0038] The significance of this step is that through the application of machine learning algorithms, the system can automatically identify potential failure modes of valves, reducing reliance on manual monitoring. Real-time detection of abnormal patterns helps to detect problems in advance and avoid downtime and losses caused by failures. By identifying and handling potential failures in a timely manner, the system can extend the service life of the valve and improve the overall operating efficiency and safety of the equipment.

[0039] Specifically, the valve operation feature can be input into a deep autoencoder network to learn its low-dimensional representation, and the bottleneck layer output of the autoencoder is used as a feature embedding, which represents the core features of the data and can distinguish between normal and abnormal modes, wherein the autoencoder captures the potential structural features of the data by reconstructing the error; In this step, the system inputs the operating characteristics of the valve into a deep autoencoder network. An autoencoder is an unsupervised learning model that can learn low-dimensional representations of data by compressing and reconstructing it. Specifically, the input features are compressed to the bottleneck layer through the encoder part, and the output of the bottleneck layer is the feature embedding, which represents the core characteristics of the data. By reconstructing the error, the autoencoder can capture the underlying structural characteristics of the data and help distinguish normal from abnormal patterns.

[0040] The significance of using deep autoencoders is that they can effectively extract the core features of the data, reduce the dimensionality of the data, and retain its important information. By learning low-dimensional representations, the system can more easily identify abnormal patterns, because abnormal data usually produces large errors when reconstructed. In this way, the system can detect potential faults more accurately and improve the efficiency and accuracy of anomaly detection.

[0041] In the specific implementation, suppose we have a high-dimensional data set containing pressure, temperature, and vibration features. First, the system inputs these features into a pre-trained deep autoencoder. The structure of the autoencoder includes an encoder and a decoder. The encoder compresses the input data to the bottleneck layer, and the output of the bottleneck layer is the low-dimensional feature embedding. The decoder attempts to reconstruct the original input from the output of the bottleneck layer. By minimizing the reconstruction error, the autoencoder can learn the potential structural characteristics of the data. Normal data can usually be reconstructed well, while abnormal data will produce a large reconstruction error. By analyzing the output of the bottleneck layer, the system can identify which data points may be abnormal, thus providing a basis for subsequent anomaly detection.

[0042] For example, in an industrial site, we monitor the operating status of a high- and medium-pressure valve and collect data through pressure sensors, temperature sensors, and vibration sensors. Assuming that each sensor collects 800 data points per second for a week, we have a large amount of high-dimensional data. In order to solve the problem of high dimensionality of the data, we decided to use a deep autoencoder to extract a low-dimensional representation of the data.

[0043] First, we organize the data for a whole day into a high-dimensional matrix containing pressure, temperature, and vibration features. Then, this matrix is ​​fed into a deep autoencoder. The autoencoder consists of a multi-layer neural network, where the encoder part gradually compresses the input data to the bottleneck layer. For example, the encoder can have three layers, the first layer compresses from 2400 dimensions to 1200 dimensions, the second layer compresses from 1200 dimensions to 600 dimensions, and the final bottleneck layer compresses the data to 100 dimensions.

[0044] The output of the bottleneck layer represents low-dimensional feature embeddings, which are learned by the trained autoencoder and can effectively retain the core information of the data. Next, the decoder attempts to reconstruct the original input data. By minimizing the error between the input data and the reconstructed data, the network can capture the underlying structural characteristics of the data. By comparing the reconstruction error, we can distinguish between normal data and abnormal data. Usually, abnormal data will show a higher reconstruction error because it cannot be reconstructed well.

[0045] Based on the feature embedding, anomaly detection is performed using an isolation forest algorithm, wherein the isolation forest constructs a decision tree by randomly selecting features and segmentation points to identify abnormal data points that are different from normal patterns; In this step, the system uses the Isolation Forest algorithm to perform anomaly detection on feature embeddings. Isolation Forest is a tree-based unsupervised learning algorithm that builds multiple decision trees by randomly selecting features and split points. The path length of each data point in these trees is used to evaluate its degree of anomaly. Typically, abnormal data points will have shorter paths in the tree because they are more easily isolated.

[0046] The application of the Isolation Forest algorithm enables the system to efficiently identify abnormal data points. Due to its randomness and tree structure, the Isolation Forest can quickly process large data sets and has high detection accuracy for abnormal data points. By identifying abnormal data that is different from the normal pattern, the system can promptly warn of potential failures, reducing equipment downtime and maintenance costs.

[0047] In the specific implementation, it is assumed that we have obtained low-dimensional feature embeddings from the autoencoder. The system inputs these embeddings into the isolation forest algorithm. The isolation forest consists of multiple randomly constructed decision trees, each of which splits the data set by randomly selecting features and split points. For each data point, the system calculates its average path length in all trees. Usually, normal data points will have a longer path in the tree, while abnormal data points will have a shorter path. By setting a threshold, the system is able to identify abnormal data points with shorter path lengths. These abnormal points may correspond to potential failure modes of the valve, and the system can further analyze these points to determine the specific cause of the failure.

[0048] For example, after obtaining the low-dimensional feature embedding extracted from the autoencoder, we use the isolation forest algorithm for anomaly detection. The isolation forest consists of multiple randomly constructed binary trees, each of which splits the data by randomly selecting features and cutting points.

[0049] Assume that we have low-dimensional feature embeddings of 5000 data points from the bottleneck layer. To implement Isolation Forest, we build 100 decision trees. In each tree, the data is partitioned by randomly selecting features and split points. For each data point, we calculate its average path length in all trees.

[0050] In actual operation, we found that the path length of a data point in multiple trees is shorter, which means that the point is easier to be isolated than other data points. We determine a threshold based on the distribution of these path lengths, and data points exceeding this threshold are marked as abnormal. For example, if it is determined that the valve has abnormal vibration for a period of time, the abnormal data point has significant isolation in the isolation forest, prompting the operator to further inspect and analyze it.

[0051] The causal relationship between the abnormal detection results of the isolation forest algorithm and the operating conditions of the high and medium pressure valves is analyzed by using a causal reasoning network to identify potential failure modes. The causal reasoning network reveals the driving factors behind the abnormal phenomena through a causal graph model.

[0052] In this step, the system uses a causal inference network to analyze the causal relationship between the abnormal results detected by the isolation forest algorithm and the valve operating conditions. The causal inference network uses a causal graph model to reveal the potential causal relationships in the data and help identify the driving factors behind the abnormal phenomena. Through this analysis, the system can gain a deeper understanding of the causes of abnormal patterns and thus identify potential failure modes.

[0053] The application of causal reasoning networks enables the system to go beyond simple correlation analysis and dig deeper into the root causes of abnormal phenomena. By identifying the causal relationship between abnormal data and operating conditions, the system can more accurately identify failure modes and provide a basis for formulating effective maintenance strategies. This causal analysis helps improve the accuracy and reliability of fault diagnosis.

[0054] In the specific implementation, assume that the isolation forest algorithm detects some abnormal data points. The system combines these abnormal points with the operating conditions of the valve (such as pressure, temperature, vibration, etc.) to build a causal graph model. The causal graph model represents variables and their causal relationships through nodes and directed edges. The system uses a causal reasoning network to analyze these relationships and identify which operating conditions may be the driving factors of the abnormal phenomenon. For example, the system may find that the probability of abnormal data points increases significantly when high pressure and high temperature occur at the same time, indicating that these conditions may lead to potential failure of the valve. Through this causal analysis, the system is able to identify specific failure modes and provide guidance for subsequent maintenance and overhaul.

[0055] For example, after identifying abnormal data points, we need to figure out the root causes of these anomalies, so the causal inference network is introduced. The causal inference network analyzes the causal relationship between variables by constructing a causal graph model.

[0056] First, we build a causal graph, where nodes represent different sensor readings and operating conditions, such as pressure levels, temperature changes, vibration intensity, etc. Directed edges represent causal relationships between variables. For example, an edge in the graph might represent the potential impact of a temperature change on pressure.

[0057] Next, we use machine learning techniques and domain knowledge to fill in the relationships and conditional probabilities between variables in the causal graph. At this point, we can input the abnormal data points marked by the isolation forest into the causal reasoning network to infer which operating conditions or combinations may have caused the abnormal phenomenon. For example, through causal reasoning, the analysis found that high pressure and high temperature were the main factors of the vibration anomaly. This information will be fed back to the maintenance team to help them focus on areas that are more likely to cause problems for detailed inspection and preventive maintenance. This method is more in-depth than simple correlation analysis because it reveals the causal drivers behind the anomaly, which helps to quickly and accurately formulate fault repair plans.

[0058] S203, generating a corresponding maintenance and overhaul plan based on the decision tree algorithm according to the potential failure mode, and feeding back the maintenance and overhaul plan to the operator through a mobile application to realize intelligent monitoring of high and medium pressure valves.

[0059] In the intelligent monitoring system, based on the potential failure modes identified, the system uses a decision tree algorithm to generate corresponding maintenance and overhaul plans. The decision tree algorithm automatically generates a series of maintenance steps and overhaul suggestions by analyzing the characteristics of the failure mode. These suggestions are integrated into a detailed plan and fed back to the operator through a mobile application. Operators can view these plans in real time on mobile devices to ensure that timely actions are taken to prevent further development of the failure. In this way, the system can not only identify the failure, but also provide specific solutions to help operators respond quickly.

[0060] The significance of this step is to closely integrate fault detection with actual maintenance operations. By automatically generating maintenance and overhaul plans, the system can significantly improve the efficiency and accuracy of fault handling. Operators no longer need to rely on experience or manual analysis to develop maintenance plans, but can rely on the scientific suggestions provided by the system. This automated process reduces the possibility of human error and ensures the timeliness and effectiveness of maintenance work, thereby improving the overall reliability and safety of the equipment.

[0061] Specifically, according to the potential failure mode, a corresponding maintenance and repair plan is generated based on a decision tree algorithm. The potential failure mode can be represented as a binary vector using one-hot encoding, and combined with the numerical features of the collected data to form a complete feature vector as the input of the decision tree model; In this step, the system first one-hot encodes the identified potential fault modes and converts them into binary vectors. One-hot encoding is a technique that converts categorical data into binary form, where each fault mode is represented as a unique binary vector. Next, the system combines these binary vectors with the numerical features collected by the sensors to form a complete feature vector. This feature vector contains the encoded information of the fault mode and the real-time data features, which serves as the input to the decision tree model.

[0062] By encoding the failure mode as a binary vector, the system can handle different failure types in a structured way. The complete feature vector formed by combining the numerical features ensures that the decision tree model can fully understand the current equipment status. This approach not only improves the input quality of the model, but also enhances the model's ability to identify different failure modes, ensuring that the generated maintenance plan is more targeted and effective.

[0063] Assume that the system detects three potential failure modes: A, B, and C. Through one-hot encoding, mode A is represented as [1, 0,0], mode B is [0, 1, 0], and mode C is [0, 0, 1]. Assume that the data collected by the sensor includes pressure, temperature, and vibration, which are [100, 75, 0.5] respectively. For failure mode A, the complete feature vector is [1, 0, 0, 100, 75, 0.5]. This feature vector is input into the decision tree model, which analyzes these input features and identifies the most suitable maintenance and repair strategy.

[0064] For example, in an industrial valve monitoring system, the system detects three potential failure modes: A, B, and C. Use the one-hot encoding technique to process these categorical information. Failure mode A is encoded as [1, 0, 0], mode B is encoded as [0, 1,0], and mode C is encoded as [0, 0, 1]. Now assume that the data collected by the sensor includes the following numerical features: pressure is 105psi, temperature is 320 Kelvin, and vibration amplitude is 0.4 m / s². Assume that the current identified failure mode is B. Combining the one-hot encoding and sensor data, the complete feature vector formed is [0, 1, 0, 105, 320, 0.4]. This feature vector is used as the input of the decision tree model. This feature vector not only carries a clear failure mode, but also combines the multi-dimensional data collected during the monitoring process, providing rich information for subsequent decision analysis.

[0065] Starting from the root node of the decision tree, the tree structure is traversed layer by layer. For each node, the feature vectors are compared using the feature partitioning conditions of the node. If a feature value in the feature vector meets a preset condition, the traversal continues downward along the branch that meets the preset condition. If a feature value does not meet a preset condition, the traversal continues downward along another branch until a leaf node is reached. In this step, the system starts from the root node of the decision tree and traverses the tree structure layer by layer. Each node contains a feature partitioning condition, which is used to compare the input feature vector. The system determines whether the value in the feature vector meets the node partitioning condition. If it does, it continues to traverse down along the branch corresponding to the condition; if it does not, it continues along another branch. This process continues until a leaf node is reached.

[0066] By traversing the decision tree layer by layer, the system can effectively analyze the information in the feature vector and gradually narrow down the possible fault range. The division conditions of each node help the system make effective decisions in the complex feature space, and when it finally reaches the leaf node, the system can provide specific maintenance and repair suggestions. This method ensures the accuracy and efficiency of fault analysis and helps operators quickly identify and solve problems.

[0067] Suppose the root node of the decision tree is divided according to the pressure feature, and the condition is "pressure > 90". If the pressure value in the feature vector is 100, the condition is met, and the traversal continues down along the "yes" branch. The next node may be divided according to the temperature feature, and the condition is "temperature > 70". The temperature in the feature vector is 75, and the condition is still met, and the "yes" branch continues. Finally, the leaf node is reached, which may mean "check the valve tightness". Through this layer-by-layer traversal, the system can provide accurate maintenance recommendations based on the specific value of the feature vector.

[0068] Taking a simple decision tree structure as an example, suppose the root node's partitioning condition is "pressure > 100 psi", and the pressure value represented by the current feature vector is 105 psi, which meets the condition. Therefore, the system continues to traverse along the "yes" branch. The next node may be partitioned based on temperature, and the condition is set to "temperature > 300 Kelvin". The temperature value in the current feature vector is 320 Kelvin, which also meets this condition, so the system continues to traverse down along the "yes" branch. Suppose the next node is based on vibration amplitude, and the condition is "vibration amplitude < 0.5 m / s²". The current vibration amplitude is 0.4 m / s², which meets the condition. The system traverses along the "yes" branch until it reaches the leaf node. Finally, the leaf node reached by the system provides specific maintenance recommendations. This process is like screening layer by layer, applying simple judgment conditions in turn to narrow down the possible solutions to the most appropriate options, ensuring that the final recommendation is based on the most directly related path of the feature vector.

[0069] After path tracing, a leaf node is eventually reached, which represents a specific maintenance and overhaul recommendation. A complete maintenance and overhaul plan is generated based on each leaf node reached by the tracking.

[0070] During the path tracing process, the system eventually reaches a leaf node. Each leaf node represents a specific maintenance and overhaul suggestion. The system integrates all suggestions based on the leaf nodes reached by the path tracing to generate a complete maintenance and overhaul plan. This plan lists in detail the steps and precautions that need to be performed to ensure that operators can perform maintenance according to the suggestions.

[0071] By generating a complete maintenance and overhaul plan, the system can provide operators with clear operational guidance. This plan not only improves the efficiency of maintenance work, but also reduces errors caused by uncertainty. Operators can perform maintenance tasks step by step according to the suggestions in the plan to ensure the normal operation and safety of the equipment.

[0072] Assume that during the path tracing process, the system reaches three leaf nodes, which provide the following recommendations: "Check valve tightness", "Replace temperature sensor", and "Calibrate vibration sensor". The system integrates these recommendations into a maintenance plan, detailing the execution order and precautions of each step. For example, the plan may recommend first checking the valve tightness to ensure that there is no leakage; then replacing the temperature sensor to improve measurement accuracy; and finally calibrating the vibration sensor to ensure that it is working properly. In this way, operators can follow the plan step by step to ensure the reliability and safety of the equipment.

[0073] Through the above path tracing, assuming that the system eventually reaches a specific leaf node, the output of the leaf node is "perform valve sealing inspection" and "record temperature abnormality data". These outputs are not just individual maintenance suggestions, but responses to comprehensive information. Based on these suggestions, the operator will first perform a valve sealing inspection to determine whether there is a leak or other related problems. Next, the possible impact of temperature abnormalities will be recorded, which may involve data recording or further analysis so that there will be more comprehensive reference data to adjust the system or detection process in the future. Based on the suggestions of these leaf nodes, the system generates a detailed maintenance and overhaul plan. The plan may include: the specific steps of the inspection procedure, the tools and materials required, the estimated completion time, and any possible safety precautions. Through comprehensive suggestions, operators can solve problems quickly and effectively, minimize downtime, and improve the availability and safety of equipment.

[0074] It can be seen that according to the real-time operating status and environmental pressure data of the high and medium pressure valves, data is collected through an embedded multi-sensor network, and the collected data is comprehensively processed using an adaptive data fusion algorithm to obtain comprehensive valve working characteristics; based on the valve working characteristics, an abnormality detection algorithm based on machine learning is used to identify the potential failure modes of the high and medium pressure valves in real time; according to the potential failure modes, a corresponding maintenance and overhaul plan is generated based on a decision tree algorithm, and the maintenance and overhaul plan is fed back to the operator through a mobile application to realize intelligent monitoring of the high and medium pressure valves, so that the embedded multi-sensor network, adaptive data fusion algorithm and abnormality detection method based on machine learning can be used to not only realize real-time monitoring and fault warning of valves, but also help optimize maintenance and overhaul plans and improve the intelligence level of valve management.

[0075] Another embodiment of the present invention provides an intelligent monitoring system for high and medium pressure valves. Figure 3 , the system may include: The acquisition module 301 is used to acquire data through an embedded multi-sensor network according to the real-time operation status and environmental pressure data of the high and medium pressure valves, and to comprehensively process the acquired data using an adaptive data fusion algorithm to obtain comprehensive valve operation characteristics, wherein the multi-sensor network includes at least a pressure sensor, a temperature sensor, and a vibration sensor; An identification module 302 is used to identify potential failure modes of the high and medium pressure valves in real time based on the valve operating characteristics and using an abnormality detection algorithm based on machine learning; The monitoring module 303 is used to generate a corresponding maintenance and overhaul plan based on the decision tree algorithm according to the potential failure mode, and feed back the maintenance and overhaul plan to the operator through the mobile application to realize intelligent monitoring of high and medium pressure valves.

[0076] It can be seen that according to the real-time operating status and environmental pressure data of the high and medium pressure valves, data is collected through an embedded multi-sensor network, and the collected data is comprehensively processed using an adaptive data fusion algorithm to obtain comprehensive valve working characteristics; based on the valve working characteristics, an abnormality detection algorithm based on machine learning is used to identify the potential failure modes of the high and medium pressure valves in real time; according to the potential failure modes, a corresponding maintenance and overhaul plan is generated based on a decision tree algorithm, and the maintenance and overhaul plan is fed back to the operator through a mobile application to realize intelligent monitoring of the high and medium pressure valves, so that the embedded multi-sensor network, adaptive data fusion algorithm and abnormality detection method based on machine learning can be used to not only realize real-time monitoring and fault warning of valves, but also help optimize maintenance and overhaul plans and improve the intelligence level of valve management.

[0077] An embodiment of the present invention further provides a storage medium, in which a computer program is stored, wherein the computer program is configured to execute the steps of any of the above method embodiments when running.

[0078] Specifically, in this embodiment, the above storage medium may be configured to store a computer program for performing the following steps: S201, according to the real-time operating status and environmental pressure data of the high and medium pressure valves, collect data through an embedded multi-sensor network, and use an adaptive data fusion algorithm to comprehensively process the collected data to obtain comprehensive valve operating characteristics, wherein the multi-sensor network includes at least a pressure sensor, a temperature sensor, and a vibration sensor; S202, based on the valve operating characteristics, using an abnormality detection algorithm based on machine learning to identify potential failure modes of the high and medium pressure valves in real time; S203, generating a corresponding maintenance and overhaul plan based on the decision tree algorithm according to the potential failure mode, and feeding back the maintenance and overhaul plan to the operator through a mobile application to realize intelligent monitoring of high and medium pressure valves.

[0079] It can be seen that according to the real-time operating status and environmental pressure data of the high and medium pressure valves, data is collected through an embedded multi-sensor network, and the collected data is comprehensively processed using an adaptive data fusion algorithm to obtain comprehensive valve working characteristics; based on the valve working characteristics, an abnormality detection algorithm based on machine learning is used to identify the potential failure modes of the high and medium pressure valves in real time; according to the potential failure modes, a corresponding maintenance and overhaul plan is generated based on a decision tree algorithm, and the maintenance and overhaul plan is fed back to the operator through a mobile application to realize intelligent monitoring of the high and medium pressure valves, so that the embedded multi-sensor network, adaptive data fusion algorithm and abnormality detection method based on machine learning can be used to not only realize real-time monitoring and fault warning of valves, but also help optimize maintenance and overhaul plans and improve the intelligence level of valve management.

[0080] An embodiment of the present invention further provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute the steps in any one of the above method embodiments.

[0081] Specifically, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor, and the input / output device is connected to the processor.

[0082] Specifically, in this embodiment, the processor may be configured to perform the following steps through a computer program: S201, according to the real-time operating status and environmental pressure data of the high and medium pressure valves, collect data through an embedded multi-sensor network, and use an adaptive data fusion algorithm to comprehensively process the collected data to obtain comprehensive valve operating characteristics, wherein the multi-sensor network includes at least a pressure sensor, a temperature sensor, and a vibration sensor; S202, based on the valve operating characteristics, using an abnormality detection algorithm based on machine learning to identify potential failure modes of the high and medium pressure valves in real time; S203, generating a corresponding maintenance and overhaul plan based on the decision tree algorithm according to the potential failure mode, and feeding back the maintenance and overhaul plan to the operator through a mobile application to realize intelligent monitoring of high and medium pressure valves.

[0083] It can be seen that according to the real-time operating status and environmental pressure data of the high and medium pressure valves, data is collected through an embedded multi-sensor network, and the collected data is comprehensively processed using an adaptive data fusion algorithm to obtain comprehensive valve working characteristics; based on the valve working characteristics, an abnormality detection algorithm based on machine learning is used to identify the potential failure modes of the high and medium pressure valves in real time; according to the potential failure modes, a corresponding maintenance and overhaul plan is generated based on a decision tree algorithm, and the maintenance and overhaul plan is fed back to the operator through a mobile application to realize intelligent monitoring of the high and medium pressure valves, so that the embedded multi-sensor network, adaptive data fusion algorithm and abnormality detection method based on machine learning can be used to not only realize real-time monitoring and fault warning of valves, but also help optimize maintenance and overhaul plans and improve the intelligence level of valve management.

[0084] The above describes in detail the structure, features and effects of the present invention based on the embodiments shown in the drawings. The above is only a preferred embodiment of the present invention, but the present invention is not limited to the scope of implementation shown in the drawings. Any changes made according to the concept of the present invention, or modifications to equivalent embodiments with equivalent changes, which still do not exceed the spirit covered by the description and drawings, should be within the protection scope of the present invention.

Claims

1. An intelligent monitoring method for high and medium pressure valves, characterized in that: The method comprises: According to the real-time operating status and environmental pressure data of the high and medium pressure valves, data is collected through an embedded multi-sensor network, and the collected data is comprehensively processed using an adaptive data fusion algorithm to obtain comprehensive valve operating characteristics, wherein the multi-sensor network includes at least a pressure sensor, a temperature sensor, and a vibration sensor; Based on the valve operating characteristics, a machine learning-based anomaly detection algorithm is used to identify potential failure modes of the high and medium pressure valves in real time; According to the potential failure mode, a corresponding maintenance and overhaul plan is generated based on a decision tree algorithm, and the maintenance and overhaul plan is fed back to the operator through a mobile application to achieve intelligent monitoring of high and medium pressure valves.

2. The method according to claim 1, characterized in that According to the real-time operating status and environmental pressure data of the high and medium pressure valves, data is collected through an embedded multi-sensor network, and the collected data is comprehensively processed using an adaptive data fusion algorithm to obtain comprehensive valve operating characteristics, including: For each sensor data stream, continuous wavelet transform is applied to extract features of different scales, which represent the change details and trends of the signal to capture multi-level information of the data; A perception game model is constructed, and each sensor is regarded as a participant in the game. Each sensor participates in the game according to the contribution and reliability of the data it collects, competing for its importance in the fusion process. Through iterative calculation, a Nash equilibrium point is reached, at which the data weight distribution of each sensor reaches the optimal state. The different scale features extracted from different sensors are input into the fuzzy logic system, and a fusion feature is output as a comprehensive valve working feature, wherein the fuzzy rule set is adjusted in real time according to the weights assigned by the perception game model to adapt to different operating states and environments, and the fuzzy reasoning process is executed. By matching the fuzzy set with the fuzzy rules, the different scale features of the sensor are converted into a comprehensive fusion feature to reflect the valve working state.

3. The method according to claim 2, characterized in that Based on the valve working characteristics, the abnormality detection algorithm based on machine learning is used to identify the potential failure mode of the high and medium pressure valve in real time, including: Inputting the valve operation feature into a deep autoencoder network to learn its low-dimensional representation, and using the bottleneck layer output of the autoencoder as a feature embedding, the embedding represents the core features of the data and can distinguish between normal and abnormal modes, wherein the autoencoder captures the potential structural features of the data by reconstructing errors; Based on the feature embedding, anomaly detection is performed using an isolation forest algorithm, wherein the isolation forest constructs a decision tree by randomly selecting features and segmentation points to identify abnormal data points that are different from normal patterns; The causal relationship between the abnormal detection results of the isolation forest algorithm and the operating conditions of the high and medium pressure valves is analyzed by using a causal reasoning network to identify potential failure modes. The causal reasoning network reveals the driving factors behind the abnormal phenomena through a causal graph model.

4. The method according to claim 3, characterized in that The generating of a corresponding maintenance and repair plan based on the decision tree algorithm according to the potential failure mode includes: The potential failure mode is represented as a binary vector using one-hot encoding, and combined with the numerical features of the collected data to form a complete feature vector as the input of the decision tree model; Starting from the root node of the decision tree, the tree structure is traversed layer by layer. For each node, the feature vectors are compared using the feature partitioning conditions of the node. If a feature value in the feature vector meets a preset condition, the traversal continues downward along the branch that meets the preset condition. If a feature value does not meet a preset condition, the traversal continues downward along another branch until a leaf node is reached. After path tracing, a leaf node is eventually reached, which represents a specific maintenance and overhaul recommendation, and a complete maintenance and overhaul plan is generated based on each leaf node reached by the tracking.

5. An intelligent monitoring system for high and medium pressure valves, characterized in that: The system comprises: An acquisition module, for acquiring data through an embedded multi-sensor network according to the real-time operating status and environmental pressure data of the high and medium pressure valves, and comprehensively processing the acquired data using an adaptive data fusion algorithm to obtain comprehensive valve operating characteristics, wherein the multi-sensor network includes at least a pressure sensor, a temperature sensor, and a vibration sensor; An identification module, for identifying potential failure modes of the high and medium pressure valves in real time based on the valve operating characteristics and using an abnormality detection algorithm based on machine learning; The monitoring module is used to generate corresponding maintenance and overhaul plans based on the decision tree algorithm according to the potential failure mode, and feed back the maintenance and overhaul plans to the operator through the mobile application to realize intelligent monitoring of high and medium pressure valves.

6. The system according to claim 5, characterized in that The acquisition module is specifically used for: For each sensor data stream, continuous wavelet transform is applied to extract features of different scales, which represent the change details and trends of the signal to capture multi-level information of the data; A perception game model is constructed, and each sensor is regarded as a participant in the game. Each sensor participates in the game according to the contribution and reliability of the data it collects, competing for its importance in the fusion process. Through iterative calculation, a Nash equilibrium point is reached, at which the data weight distribution of each sensor reaches the optimal state. The different scale features extracted from different sensors are input into the fuzzy logic system, and a fusion feature is output as a comprehensive valve working feature, wherein the fuzzy rule set is adjusted in real time according to the weights assigned by the perception game model to adapt to different operating states and environments, and the fuzzy reasoning process is executed. By matching the fuzzy set with the fuzzy rules, the different scale features of the sensor are converted into a comprehensive fusion feature to reflect the valve working state.

7. The system according to claim 6, characterized in that The identification module is specifically used for: Inputting the valve operation feature into a deep autoencoder network to learn its low-dimensional representation, and using the bottleneck layer output of the autoencoder as a feature embedding, the embedding represents the core features of the data and can distinguish between normal and abnormal modes, wherein the autoencoder captures the potential structural features of the data by reconstructing errors; Based on the feature embedding, anomaly detection is performed using an isolation forest algorithm, wherein the isolation forest constructs a decision tree by randomly selecting features and segmentation points to identify abnormal data points that are different from normal patterns; The causal relationship between the abnormal detection results of the isolation forest algorithm and the operating conditions of the high and medium pressure valves is analyzed by using a causal reasoning network to identify potential failure modes. The causal reasoning network reveals the driving factors behind the abnormal phenomena through a causal graph model.

8. The system according to claim 7, characterized in that The monitoring module is specifically used for: The potential failure mode is represented as a binary vector using one-hot encoding, and combined with the numerical features of the collected data to form a complete feature vector as the input of the decision tree model; Starting from the root node of the decision tree, the tree structure is traversed layer by layer. For each node, the feature vectors are compared using the feature partitioning conditions of the node. If a feature value in the feature vector meets a preset condition, the traversal continues downward along the branch that meets the preset condition. If a feature value does not meet a preset condition, the traversal continues downward along another branch until a leaf node is reached. After path tracing, a leaf node is eventually reached, which represents a specific maintenance and overhaul recommendation, and a complete maintenance and overhaul plan is generated based on each leaf node reached by the tracking.

9. A storage medium, characterized in that: The storage medium stores a computer program, wherein the computer program is configured to execute the method according to any one of claims 1 to 4 when executed.

10. An electronic device comprising a memory and a processor, characterized in that: A computer program is stored in the memory, and the processor is configured to run the computer program to perform the method according to any one of claims 1 to 4.

Images