Nitrogen and helium leakage detection data analysis and processing system based on multi-sensor fusion

Abstract

The invention relates to the technical field of data processing, and discloses a nitrogen and helium leakage detection data analysis and processing system based on multi-sensor fusion. Comprising a detection point configuration module, a data acquisition module, a leakage judgment module, a position determination module, a leakage position determination module and a model construction module, wherein the detection point configuration module is in an annular dispersion form by taking a detection point as a center and is used for performing leakage configuration on the detection point; the data acquisition module is used for acquiring multimode sensing data of the detection point; generating a visual pipeline leakage model; according to the method, the dimension of the leakage state identification data is expanded, the limitation problem caused by single-dimension data is avoided, the specific leakage positions of nitrogen and helium can be accurately positioned from the gas pipeline, the interference influence caused by outward diffusion of the leaked nitrogen and helium is avoided, and the reliability of the leakage state identification data is improved. Therefore, the comprehensive analysis effect of the actual leakage conditions of the nitrogen and the helium on the gas pipeline can be realized.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and more specifically, to a nitrogen and helium leak detection data analysis and processing system based on multi-sensor fusion. Background Technology

[0002] In industrial production, the manufacture and production of certain special types of gas detection equipment require the use of a mixture of nitrogen and helium. When delivering nitrogen and helium into the production equipment, they need to be mixed. To ensure the safety of the mixed gas pipeline when delivering the nitrogen and helium mixture, it is necessary to test the sealing performance of the gas pipeline in order to detect leaks in the gas pipeline in a timely and accurate manner.

[0003] Reference patent application CN118228601A discloses a helium storage leak early warning system and method. By combining acoustic fingerprint technology, molecular vibration spectroscopy analysis and wireless sensor network, it achieves comprehensive monitoring of helium leaks. It uses a high-sensitivity microphone array to capture acoustic features related to helium leaks, a laser spectrometer to detect changes in the vibrational spectrum of helium molecules, and a wireless sensor network to continuously monitor the concentration and distribution of helium in the environment. By integrating acoustic, spectral and chemical sensing technologies, it achieves comprehensive and multi-angle detection of helium leaks, greatly improving the detection coverage and accuracy. Existing leak detection and analysis systems typically use spaced sensors to collect multimodal data when detecting and analyzing gas leaks in gas pipelines. Because the sensors are loosely spaced, the multimodal data at the detection points lacks temporal synchronization, leading to misalignment of data at the same detection point on the timeline. This increases the probability of delayed response or misjudgment in nitrogen and helium leak detection. Furthermore, the outward diffusion of leaked gas mixtures from the pipeline can interfere with the accurate location of the leak, resulting in low accuracy in subsequent leak location and reducing the effectiveness of nitrogen and helium leak detection and analysis results.

[0004] In view of this, the present invention proposes a nitrogen and helium leak detection data analysis and processing system based on multi-sensor fusion to solve the above problems. Summary of the Invention

[0005] To overcome the aforementioned shortcomings of the prior art and to achieve the above objectives, the present invention provides the following technical solution: a nitrogen and helium leak detection data analysis and processing system based on multi-sensor fusion, comprising: The detection point configuration module is used to identify the gas pipeline to be detected, mark the detection sections with detection points in the gas pipeline, and configure the detection points in a ring-shaped dispersion pattern with the detection points as the center for leak detection. The data acquisition module is used to collect the balanced flow velocity and interval length of the detection section, calculate the detection cycle, and collect multi-mode sensor data of the detection point within the detection cycle. The leak detection module is used to identify the leak status of the detection point through the pipeline leak identification model. The leak status includes leaked and non-leaking, and determines whether to perform leak location operation. The location determination module is used to identify the target point from the detection points, mark the target section corresponding to the target point from the detection section, and collect the leakage characteristics of the target section, including the diffusion distance and leakage direction, and determine the leakage location within the target section. The model building module is used to build a basic model with leak nodes and import multi-mode sensor data into the information bits of the leak nodes to generate a visualized pipeline leak model.

[0006] Preferably, when marking the detection section, the two ends of the gas pipeline are marked, and the two ends are designated as the first and last section points. A equally spaced section points are marked within the gas pipeline, and the gas pipeline between two adjacent section points is recorded as the detection interval, thus obtaining... Each detection interval is marked individually. The midpoint of each detection interval is identified, and the location of the midpoint is recorded as the detection point. There are 1 testing site.

[0007] Preferably, the leak configuration steps for the detection points are as follows: respectively Centered on a detection point, and using a distance of one calibration length from the detection point as a standard, four points are marked in a ring at equal angles on the gas pipeline; The positions of the four points on the gas pipeline are continuously adjusted until the line connecting two non-adjacent points is parallel to the axis of the gas pipeline, thus obtaining four configuration points. An oxygen sensor and a flow rate sensor are deployed at two points on a line parallel to the axial line of the gas pipeline, respectively. A temperature sensor is deployed at one of the remaining points, and a vibration sensor and an ultrasonic sensor are deployed at the other point.

[0008] Preferably, the steps for determining the testing cycle are as follows: At the same time, the flow rate sensors detect the gas flow rate in each pipeline one by one. Real-time flow rate at each detection point, and The equilibrium flow velocity is calculated by averaging the sum of the real-time flow velocities. The length of the detection interval is measured and recorded as the interval length. The interval length is then compared with the equilibrium flow velocity to calculate the diffusion time. The diffusion time is reduced by one-fifth of the diffusion time, and the reduced diffusion time is recorded as the detection cycle.

[0009] Preferably, the multi-mode sensing data includes temperature gradient values, oxygen concentration values, vibration frequency values, and sound wave frequency values; The steps for acquiring temperature gradient values ​​are as follows: At time T1, temperature data is collected via a temperature sensor. The ambient temperature at each detection point was obtained. One initial temperature value; After one detection cycle, data is collected at time T2. The ambient temperature at each detection point was obtained. One termination temperature value; Will The initial temperature value and After subtracting each termination temperature value, the result is calculated. A temperature gradient value.

[0010] Preferably, the training steps for the pipeline leak detection model are as follows: Multiple sets of multimode sensing data and the corresponding leakage status are collected in advance; Multimodal sensing data is converted into corresponding feature vectors, leakage status is converted into labels corresponding to multimodal sensing data, non-leaking is converted to 0, and leaked is converted to 1; The feature vectors are used as input to the machine learning model, and the leakage state corresponding to each set of feature vectors is used as the output of the machine learning model, with the leakage state as the prediction target. The machine learning model is trained with the goal of minimizing the sum of prediction errors of all training data until the sum of prediction errors converges, at which point training stops and a pipeline leak identification model is obtained.

[0011] Preferably, the determination steps for performing leak location operations are as follows: When the output of the pipeline leak identification model is 0, the leak status of the detection point is no leak; When the output of the pipeline leak identification model is 1, the leak status of the detection point is "leaked". The number of detection points that are in a leaking state is counted and recorded as abnormal values; When the abnormal value is 0, it is determined that no leak location operation will be performed; When the abnormal value is greater than 0, a leak location operation is performed.

[0012] Preferably, the marking steps for the target segment are as follows: When the abnormal value is 1, the detection point is recorded as an abnormal point, and the detection section where the abnormal point is located is recorded as the target section; When the abnormal value is greater than 1, the detection point is recorded as an abnormal point, resulting in D abnormal points. The detection segment where the abnormal point is located is recorded as the target segment, resulting in D target segments.

[0013] The preferred method for collecting diffusion distance and leakage direction is as follows: The vibration frequency between the first calibration frequency and the second calibration frequency is recorded as the abnormal frequency. The time when the vibration sensor first detected the abnormal frequency is found by querying the timestamp and recorded as the first moment. A 1% drop in oxygen concentration is recorded as an abnormal phenomenon. After the first moment, the time when the oxygen concentration sensor first detected the abnormal phenomenon is retrieved by the timestamp and recorded as the second moment. The duration between the first moment and the second moment is recorded as the propagation duration. The room temperature and room pressure of D target sections at the first moment were detected. The diffusion coefficient of nitrogen under the room temperature and room pressure conditions was found by looking up a table. The diffusion distance was calculated by multiplying the propagation time by the diffusion coefficient of nitrogen. The previous detection segment of the target segment is recorded as the upper segment, and the real-time flow velocity of the detection points in the upper segment and the target segment at the first moment is queried to obtain the first flow velocity value and the second flow velocity value. When the first flow velocity value is greater than or equal to the second flow velocity value, the leakage direction is upstream; when the first flow velocity value is less than the second flow velocity value, the leakage direction is downstream.

[0014] Preferably, the steps for generating the pipeline leakage model are as follows: The point cloud data of all points in the gas pipeline is retrieved from the design drawings, and the point cloud data is summarized and modeled using 3D modeling technology to construct a basic model. Following the order from upstream to downstream, mark the points corresponding to the D leakage locations on the basic model in sequence to obtain D leakage nodes; Message frames are created one by one on each of the D leak nodes, and four independent information bits are divided within each message frame. The four information bits are then numbered in ascending order. Following the ascending order of numbering, the temperature gradient values, oxygen concentration values, vibration frequency values, and sound wave frequency values ​​of the D target intervals are sequentially imported into the four information bits of the corresponding leakage node, thereby converting the basic model into a pipeline leakage model.

[0015] The technical effects and advantages of this invention's multi-sensor fusion-based nitrogen and helium leak detection data analysis and processing system are as follows: (1): The present invention can ensure that the hardware devices corresponding to each detection point are in a relatively clustered state, and ensure that the data collected by the hardware at each detection point are synchronized on the timeline as much as possible. This means that oxygen concentration and gas flow rate can be detected at the first moment of nitrogen and helium leakage, ensuring that the multimodal data at the same detection point are strictly aligned, thus solving the problem of response delay.

[0016] (2): This invention can represent the environmental changes caused by nitrogen and helium leaks from multiple dimensions such as changes in environmental concentration gradient, changes in environmental temperature field and characteristics of acoustic interference. It can also combine artificial intelligence technology to identify whether there is a leak. This not only expands the dimensions of the identification data and avoids the limitations caused by single-dimensional data, but also improves the accuracy and applicability of leak status identification under complex working conditions with multiple sources of interference.

[0017] (3): This invention can accurately locate the specific location of nitrogen and helium leaks in gas pipelines, avoiding the interference caused by the outward diffusion of leaked nitrogen and helium. It can also visualize the specific location and data of the leak in the gas pipeline in three-dimensional space, thus achieving a comprehensive analysis of the actual situation of nitrogen and helium leaks in gas pipelines, which facilitates targeted treatment of gas pipeline leaks in the future. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the nitrogen and helium leak detection data analysis and processing system based on multi-sensor fusion provided in Embodiment 1 of the present invention. Figure 2 This is a flowchart illustrating the nitrogen and helium leak detection data analysis and processing method based on multi-sensor fusion provided in Embodiment 2 of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Example 1: Please refer to Figure 1 As shown in this embodiment, the nitrogen and helium leak detection data analysis and processing system based on multi-sensor fusion includes: The detection point configuration module identifies the gas pipeline to be inspected, marks the inspection sections with detection points in the gas pipeline, and configures the detection points for leaks based on leak detection criteria. The gas pipeline to be tested is a mixed gas pipeline that transports a mixture of nitrogen and helium from their respective storage cylinders to the reaction equipment. It is also a mixed gas pipeline that needs to be tested for nitrogen and helium leaks. The gas pipeline between the nitrogen cylinder, the helium cylinder and the reaction equipment is not a single main pipe, but a combination structure of a main pipe and two branch pipes. One end of the main pipe is connected to the reaction equipment, and one end of each of the two branch pipes is connected to the nitrogen cylinder and the helium cylinder, respectively. The other end of each branch pipe is connected to the main pipe. In this embodiment, in order to accurately and comprehensively detect whether nitrogen and helium leaks during transportation, it is necessary to identify the gas pipeline to be tested in the gas transportation pipeline. Therefore, the gas pipeline to be tested needs to be able to transport nitrogen and helium simultaneously. In summary, the gas pipeline to be tested is the pipeline used for the mixed transport of nitrogen and helium between the reaction equipment and the nitrogen and helium transport pipelines.

[0021] After the gas pipeline is determined, its length is relatively long, resulting in a relatively wide range of potential gas leak locations. To improve the accuracy of gas leak detection, the gas pipeline needs to be divided into segments, breaking down the long pipeline into continuous sections. At the same time, the precise locations for leak detection are marked in each section and designated as detection points. These detection points can then serve as the corresponding locations for subsequent data collection. Specifically, when marking the detection sections, firstly, the two ends of the gas pipeline are marked, serving as the first and last section points. A equally spaced section points are then marked within the gas pipeline. Finally, the gas pipeline between any two adjacent section points is recorded as the detection interval, thus obtaining... Each detection interval is then marked individually. The midpoint of each detection interval is identified, and the location of the midpoint is recorded as the detection point. There are 1 testing site.

[0022] After the detection points are marked, these detection points can be used as data collection locations for subsequent determination of whether a gas pipeline has a gas leak. In order to collect the required data at the detection points, corresponding leak configuration operations must be performed at the detection points to ensure that each detection point on the gas pipeline has the hardware configuration to meet the subsequent data collection requirements. When configuring leak detection points, the necessary hardware equipment for each detection point must be accurately and reasonably configured within the constraints of leak detection criteria.

[0023] The leakage detection criterion is: to form a ring-shaped distribution with the detection point as the center; by distributing the equipment in a ring, it is possible to ensure that the hardware devices corresponding to each detection point are in a relatively clustered state, thereby increasing the degree of clustering of hardware devices and avoiding the phenomenon that the hardware devices configured at the same detection point are too far apart, and ensuring that the data collected by the hardware at each detection point is kept synchronized on the timeline as much as possible.

[0024] The steps for configuring leak detection points are as follows: respectively Centered on a single detection point, four points are marked in a ring at equal angles, with a standard distance of one calibration length from each detection point. The calibration length is a numerical representation of the distance between each point and the center, ensuring that each point maintains a certain distance from the center and providing the necessary space for the subsequent installation and deployment of sensors at these points. The calibration length is obtained by averaging the minimum distances from the center collected from a large number of historical data on the deployment of different types of sensors in a ring-shaped, dispersed structure. Furthermore, marking the four points in a ring at equal angles also achieves synchronization between sensor node design and timestamps, ensuring strict alignment of multimodal data at the same detection point and minimizing response delay issues. Continuously adjust the positions of the four points on the gas pipeline until the line connecting two non-adjacent points is parallel to the axis of the gas pipeline, then stop adjusting to obtain four configuration points. An oxygen sensor and a flow rate sensor are deployed at two points on a line parallel to the axial line of the gas pipeline, respectively. A temperature sensor is deployed at one of the remaining points, and a vibration sensor and an ultrasonic sensor are deployed at the other point.

[0025] It should be noted that the oxygen sensor, flow rate sensor, vibration sensor, temperature sensor, and ultrasonic sensor are all standard products on the market that can meet the detection needs of nitrogen and helium transportation. By deploying the oxygen sensor and flow rate sensor along the axial line of the gas pipeline, the oxygen sensor and flow rate sensor can be parallel to the flow direction of the mixed gas in the gas pipeline. Therefore, when nitrogen and helium leak, the leaked mixed gas will flow to both sides, and the oxygen concentration and gas flow rate can be detected immediately, improving the accuracy of the detection data.

[0026] The data acquisition module collects the balanced flow velocity and interval length of the detection section, calculates the diffusion time of the detection section, determines the detection cycle based on the diffusion time, and collects multi-mode sensor data of the detection point within the detection cycle. After marking the detection points and configuring them for leaks, the detection points are now fully capable of detecting data related to nitrogen and helium leaks. Therefore, it is necessary to set the data acquisition interval for each sensor at the detection point, i.e., the detection cycle, so that the detection cycle serves as the time span between two consecutive data acquisitions by the sensors at the detection point.

[0027] The detection cycle is not set arbitrarily; it needs to be combined with the flow characteristics of the mixed gas in the gas pipeline to calculate the diffusion time corresponding to each detection section (i.e., two adjacent detection points).

[0028] Since the flow characteristics of the mixed gas are different at different times and under different conditions, in order to meet the time requirements of all detection sections, it is necessary to formulate the detection cycle based on the diffusion time, so that the detection cycle is the time corresponding to two data collections in the detection section. The steps for determining the testing cycle are as follows: At the same time, the flow rate sensors detect the gas flow rate in each pipeline one by one. Real-time flow rate at each detection point, and The equilibrium flow velocity is calculated by averaging the sum of the real-time flow velocities. The formula for calculating the equilibrium flow velocity is: ; In the formula, To balance the flow rate, For the first Real-time flow rate at each detection point; The length of the detection interval is measured and recorded as the interval length. The interval length is then compared with the equilibrium flow velocity to calculate the diffusion time. The formula for calculating diffusion time is: ; In the formula, For diffusion duration, The interval length; The diffusion time is reduced by one-fifth of the diffusion time, and the reduced diffusion time is recorded as the detection cycle.

[0029] It should be clarified that by reducing the expansion time based on the diffusion time, the detection cycle is made shorter than the diffusion time. In this case, using the detection cycle as the basis for two consecutive data acquisitions ensures a shorter interval between adjacent data acquisitions and a higher data acquisition frequency. This allows for adaptive data acquisition and processing of gas flow states in gas pipelines where the gas flow rate is greater than the equilibrium flow rate, further improving the accuracy of data acquisition.

[0030] After the detection cycle is determined, the multi-mode sensing data of each detection point is collected by various sensors configured at the detection points at intervals of one detection cycle. This allows the multi-mode sensing data to be used as diverse data for analyzing and judging whether nitrogen and helium leaks have occurred at the detection points. Specifically, the multi-mode sensing data includes temperature gradient values, oxygen concentration values, vibration frequency values, and sound wave frequency values.

[0031] The temperature gradient value is used to represent the range of change in ambient temperature at the location of the detection point in the detection section; the larger the temperature gradient value, the greater the range of change in ambient temperature at the location of the detection point, and the higher the probability of nitrogen and helium leakage at that detection point. The steps for acquiring temperature gradient values ​​are as follows: At time T1, temperature data is collected via a temperature sensor. The ambient temperature at each detection point was obtained. One initial temperature value; After one detection cycle, data is collected at time T2. The ambient temperature at each detection point was obtained. One termination temperature value; Will The initial temperature value and After subtracting each termination temperature value, the result is calculated. One temperature gradient value; The formula for calculating the temperature gradient is: ; In the formula, For the first Temperature gradient values ​​at each detection point For the first The initial temperature value at each detection point For the first The termination temperature value at each detection point.

[0032] The oxygen concentration value is a real-time numerical representation of the ambient oxygen concentration at the location of the detection point within the detection section; the oxygen concentration value is obtained through... The data was obtained after detection by oxygen sensors at each detection point; It should be noted that when a nitrogen leak occurs at the detection point, the leaked nitrogen will flow directly into the ambient air. At this time, the nitrogen concentration in the air in a certain area will increase, and correspondingly, the oxygen concentration in the air will decrease, thus causing a change in the oxygen concentration data detected by the oxygen sensor.

[0033] The vibration frequency value is used to represent the pipe vibration frequency generated by the friction between the airflow and the gas pipeline at the location of the detection point in the detection section; the vibration frequency value is obtained by... Vibration sensors at each detection point are used to acquire data.

[0034] The acoustic frequency value is used to represent the frequency of airflow turbulence generated by air friction at the location of the detection point in the detection section; the acoustic frequency value is obtained by... Data is obtained after detection by ultrasonic sensors at each detection point.

[0035] The leak detection module takes multi-mode sensor data as input, identifies the leak status of the detection point through the pipeline leak identification model, and determines whether to perform a leak location operation. After acquiring the multi-mode sensing data, this multi-mode sensing data can then be used to... The multimodal sensor data of each detection point within a detection cycle is represented, so that the multimodal sensor data can be used as the basis for judging whether nitrogen and helium leaks at the detection points of the gas pipeline, thereby judging the leakage status of the detection points. Leakage status refers to the identification conclusion of whether nitrogen and helium leaks occur at the detection point within a detection cycle, which can directly indicate whether a gas leak has occurred at the detection point. Specifically, leakage status includes leaked and non-leaking. Among them, leaked means that no nitrogen and helium leaks have occurred at the detection point within a detection cycle, and non-leaking means that nitrogen and helium leaks have occurred at the detection point within a detection cycle. Leaked and non-leaking are identified by collecting a large number of historical leaked and non-leaking corresponding temperature gradient values, oxygen concentration values, vibration frequency values, and sound wave frequency values.

[0036] In this embodiment, when identifying the leakage status of a detection point using multi-mode sensor data, it is necessary to combine machine learning models in artificial intelligence technology to perform artificial intelligence identification. After repeated and optimized training with a large amount of historical multi-mode sensor data and corresponding leakage status, the machine learning model can obtain a pipeline leakage identification model that can identify the leakage status of a detection point based on multi-mode sensor data.

[0037] The training steps for the pipeline leak detection model are as follows: Multiple sets of multimode sensing data and the corresponding leakage status are collected in advance; The multi-mode sensing data is converted into a set of corresponding feature vectors, the leakage state is converted into a label corresponding to the multi-mode sensing data, and the label is numerically converted. For example, non-leaking is converted into 0 and leaking is converted into 1. Multiple sets of feature vectors and multiple sets of labels constitute multiple sets of training data. These multiple sets of training data are divided into training set and validation set, with 70% of the training data serving as training set and 30% serving as validation set. The feature vectors are used as input to the machine learning model, and the leakage state corresponding to each set of feature vectors is used as the output of the machine learning model. The leakage state is used as the prediction target. The machine learning model is trained using the training set and validated using the validation set. The machine learning model is trained with the goal of minimizing the sum of prediction errors of all training data until the sum of prediction errors converges, at which point training stops and a pipeline leak identification model is obtained.

[0038] For example, the pipeline leak detection model can be either a CNN neural network model or AlexNet; The formula for calculating prediction error is: ; In the formula, For prediction error, This represents the group number of the feature vector; For the first The predicted state corresponding to the group of feature vectors. For the first The actual state corresponding to the training data set.

[0039] Collected By importing multi-mode sensor data from each detection point into the pipeline leak identification model, the leak status corresponding to each detection point can be identified. Specifically, when the output of the pipeline leak identification model is 0, it means that no nitrogen or helium leaks have occurred at the detection point within a detection cycle, and the leakage status of the detection point is no leak. When the output of the pipeline leak identification model is 1, it indicates that nitrogen and helium leaks have occurred at the detection point within a detection cycle, and the leak status of the detection point is "leaked".

[0040] In the identification After identifying the leakage status at each detection point, preliminary analysis results can be provided for the nitrogen and helium leak detection based on the identified leakage status, and a decision can be made on whether to perform leak location operations based on the preliminary analysis results. Leak location operations are performed in the event of nitrogen or helium leaks in gas pipelines to precisely locate the leak point. Specifically, the steps for determining whether to perform a leak location operation are as follows: The number of detection points that are in a leaking state is counted and recorded as abnormal values; When the abnormal value is 0, it means that there is no nitrogen or helium leakage at any of the detection points on the gas pipeline. Therefore, it is not necessary to accurately locate the gas leak point, and the leak location operation is not performed. When the abnormal value is greater than 0, it indicates that there is a nitrogen and helium leak at the detection point on the gas pipeline. In this case, it is necessary to accurately locate the gas leak point and perform leak location operation.

[0041] It should be noted that the number of outliers is not unique; it can be as few as 0 or as many as 10. There are several detection points, and the location relationship between the detection points where nitrogen and helium leaks occur is not limited; they can be adjacent or non-adjacent.

[0042] In this embodiment, when the abnormal value is 0, it means that there is no nitrogen or helium leakage in the gas pipeline within a detection cycle. Then, after another detection cycle, the above steps of multi-mode sensor data acquisition and leakage location judgment are repeated.

[0043] The location determination module, when performing a leak point location operation, identifies the target point from the detection points, marks the target segment corresponding to the target point, collects the leakage characteristics of the target segment, and determines the leak location within the target segment; After determining that a leak location operation has been performed, the detection points that are in a leaking state may be the actual locations where gas leaks have occurred, or they may be other locations affected by leaked nitrogen and helium. In order to accurately identify the detection points where nitrogen and helium leaks have actually occurred from the detection points, it is necessary to further analyze and process the detection points in a leaking state. The detection points where nitrogen and helium leaks have actually occurred are recorded as target points, and the detection section where the target points are located is recorded as target sections. The steps for marking the target segment are as follows: When the abnormal value is 1, it means that there is a real nitrogen and helium leak in the detection section where only one detection point is located in the gas pipeline. The detection point is recorded as the abnormal point, and the detection section where the abnormal point is located is recorded as the target section. If the abnormal value is greater than 1, it indicates that there is a real nitrogen and helium leak in the detection section where more than one detection point is located in the gas pipeline. Then, the detection point is recorded as an abnormal point, resulting in D abnormal points. The detection section where the abnormal point is located is recorded as the target section, resulting in D target sections.

[0044] After marking the target section, it is only possible to determine that the exact location of the nitrogen and helium leak is within the detection section, but it is not possible to directly and accurately locate the leak location. Therefore, it is necessary to accurately locate the specific leak location where the nitrogen and helium leak actually occurred within the target section. When determining the location of a leak, it is necessary to collect and analyze the leak characteristics within the target section. The leak characteristics can provide a detailed representation of the actual situation when nitrogen and helium leaks occur in the target section during the detection period, and serve as the data basis for determining the precise location of the leak.

[0045] Leakage characteristics include diffusion distance and leakage direction; among them, diffusion distance is a numerical representation of the distance between the leak location and the target point, and leakage duration is a specific representation of the left and right direction between the leak location and the target point. For example, the leakage direction includes upstream and downstream directions; when the leakage direction is upstream, it means that the leakage location is located at the target point and close to the gas pipeline of the nitrogen and helium cylinders; when the leakage direction is downstream, it means that the leakage location is located at the target point and close to the gas pipeline of the reaction equipment.

[0046] The methods for collecting diffusion distance and leakage direction are as follows: Vibration frequencies between the first and second calibration frequencies are recorded as abnormal frequencies. The moment when the vibration sensor first detected the abnormal frequency is retrieved using a timestamp and recorded as the first moment. The first and second calibration frequencies are the minimum and maximum frequencies of the mechanical vibration generated by the friction between nitrogen gas and the gas pipeline wall during a nitrogen leak, thus providing a numerical range limit for the identification of abnormal frequencies. The first and second calibration frequencies are obtained by collecting a large number of historical minimum and maximum frequencies of the mechanical vibration generated by the friction between nitrogen gas and the gas pipeline wall during nitrogen leaks and averaging them respectively. A 1% drop in oxygen concentration is recorded as an abnormal phenomenon. After the first moment, the time when the oxygen concentration sensor first detected the abnormal phenomenon is retrieved by the timestamp and recorded as the second moment. The duration between the first moment and the second moment is recorded as the propagation duration. The room temperature and pressure of D target sections at the first moment were detected. The diffusion coefficient of nitrogen under the conditions of room temperature and pressure was obtained by looking up a table. The diffusion coefficient is used to measure the speed at which nitrogen diffuses and flows in all directions under different temperature and pressure conditions. Different temperatures and pressures will affect the value of the diffusion coefficient. The table lookup method refers to consulting the experimentally determined gas diffusion coefficient data table in scientific or engineering handbooks. For example, the data table can be a gas and liquid diffusion coefficient handbook, Rerry chemical engineering handbook, etc. The diffusion distance is calculated by multiplying the propagation time by the nitrogen diffusion coefficient. The formula for calculating the diffusion distance is: ; In the formula, For the first The diffusion distance of each target segment, =1,2,...,D, For the first The propagation duration of each target segment, For the first Nitrogen diffusion coefficient for each target section; The previous detection segment of the target segment is recorded as the upper segment, and the real-time flow velocity of the detection points in the upper segment and the target segment at the first moment is queried to obtain the first flow velocity value and the second flow velocity value. When the first flow velocity value is greater than or equal to the second flow velocity value, it means that the real-time flow velocity of the upper section exceeds or reaches the real-time flow velocity of the target section. At this time, the leakage direction of nitrogen and helium is closer to the nitrogen cylinder and helium cylinder, and the leakage direction is upstream. When the first flow velocity value is less than the second flow velocity value, it means that the real-time flow velocity of the upper section is lower than the real-time flow velocity of the target section. At this time, the leakage direction of nitrogen and helium is closer to the reaction equipment, and the leakage direction is downstream.

[0047] Once the leak characteristics are obtained, the precise leak locations of nitrogen and helium in the gas pipeline can be determined by the diffusion distance and leak direction. Specifically, when determining the leak location, if the leak direction of the target section is upstream, the leak location of the target section is the position in the target section that is upstream of the detection point and one diffusion distance away from the detection point. When the leakage direction of the target section is downstream, the leakage location of the target section is a position in the target section that is downstream of the detection point and one diffusion distance away from the detection point.

[0048] The model building module constructs a basic model with leak nodes and imports multi-mode sensor data into the leak nodes to generate a visualized pipeline leak model. After the leak location is determined, the position of each leak point on the gas pipeline can be accurately located. At this point, it is necessary to integrate and summarize the relevant data of this gas pipeline leak to facilitate direct query and understanding of gas pipeline leak information in the future. When integrating and summarizing relevant data, in order to understand the specific information of the gas pipeline leak location intuitively and visually, it is necessary to construct a basic model of the gas pipeline using 3D modeling technology, so as to represent the gas pipeline leak in 3D space.

[0049] After constructing the pipeline model, the basic model cannot fully represent the relevant leakage information of the leakage location. Therefore, it is necessary to establish leakage nodes in the basic model so that the position of the leakage node on the basic model can be relatively consistent with the position of the leakage location on the gas pipeline, and also provide the point limit for importing the multi-mode sensing data of each leakage location. The pipeline model with the leakage node imported into the multi-mode sensing data is called the pipeline leakage model. The steps for generating a pipeline leakage model are as follows: The point cloud data of all points in the gas pipeline is retrieved from the design drawings, and the point cloud data is summarized and modeled using 3D modeling technology to construct a basic model. Following the order from upstream to downstream, mark the points corresponding to the D leakage locations on the basic model in sequence to obtain D leakage nodes; Information frames are created one by one on D leakage nodes, and four independent information bits are divided within each information frame. The four information bits are numbered in ascending order. The information frame is a blank data group used to hold multi-mode sensing data. The information bit is the smallest unit that makes up the information frame and provides a unique import location for each multi-mode sensing data. Following the ascending order of numbering, the temperature gradient values, oxygen concentration values, vibration frequency values, and sound wave frequency values ​​of the D target intervals are sequentially imported into the four information bits of the corresponding leakage node, thereby converting the basic model into a pipeline leakage model.

[0050] Once the pipeline leakage model is constructed, it can visualize all the relevant information of the leakage locations on the gas pipeline in three-dimensional space, thus providing the analysis and processing personnel with intuitive and accurate visualization information, which facilitates the analysis and processing personnel to carry out targeted treatment of the gas pipeline leakage.

[0051] In this embodiment, by configuring the detection points in a ring-shaped, dispersed manner around the detection points, it is possible to ensure that the hardware devices corresponding to each detection point are in a relatively clustered state. This ensures that the data collected by the hardware at each detection point is kept synchronized on the timeline as much as possible. This allows for the detection of oxygen concentration and gas flow rate at the first moment of nitrogen and helium leakage, ensuring strict alignment of multimodal data at the same detection point and solving the problem of response delay.

[0052] By collecting multi-mode sensor data from detection points and combining it with a pipeline leak identification model to identify the leak status of the detection points, the environmental changes caused by nitrogen and helium leaks can be represented from multiple dimensions such as changes in environmental concentration gradient, changes in environmental temperature field, and acoustic interference characteristics. Furthermore, artificial intelligence technology is used to identify and judge whether a leak has occurred. This not only expands the dimensions of the identification data and avoids the limitations caused by single-dimensional data, but also improves the accuracy and applicability of leak status identification under complex working conditions with multi-source interference.

[0053] By collecting leakage characteristics to determine the leakage location, the exact location of nitrogen and helium leaks in gas pipelines can be precisely pinpointed. This avoids the interference caused by the outward diffusion of leaking nitrogen and helium. Combined with the construction of a pipeline leakage model with leakage nodes, the specific location and data of the leak in the gas pipeline can be visualized and displayed intuitively in three-dimensional space. This enables a comprehensive analysis of the actual situation of nitrogen and helium leaks in gas pipelines, facilitating subsequent targeted handling of gas pipeline leaks.

[0054] Example 2: Please refer to Figure 2 As shown, the parts not described in detail in this embodiment are described in Embodiment 1. A method for analyzing and processing nitrogen and helium leak detection data based on multi-sensor fusion is provided, implemented through a nitrogen and helium leak detection data analysis and processing system based on multi-sensor fusion, including: S01: Identify the gas pipeline to be inspected, mark the inspection sections with inspection points in the gas pipeline, and configure the inspection points in a ring-shaped dispersion pattern with the inspection points as the center for leak detection. S02: Collect the balanced flow velocity and interval length of the detection section, calculate the detection cycle, and collect multi-mode sensing data of the detection point within the detection cycle; S03: Identify the leakage status of the detection point through the pipeline leakage identification model and determine whether to perform the leakage location operation; if the leakage location operation is performed, proceed to S04; if the leakage location operation is not performed, repeat S02-S03. S04: Identify the target point from the detection points, mark the target section corresponding to the target point from the detection section, collect the leakage characteristics of the target section, and determine the leakage location within the target section; S05: Construct a basic model with leaking nodes, and import the multi-mode sensor data into the information bits of the leaking nodes to generate a visualized pipeline leak model.

[0055] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A nitrogen and helium leak detection data analysis and processing system based on multi-sensor fusion, characterized in that, include: The detection point configuration module is used to identify the gas pipeline to be detected, mark the detection sections with detection points in the gas pipeline, and configure the detection points in a ring-shaped dispersion pattern with the detection points as the center for leak detection. The data acquisition module is used to collect the balanced flow velocity and interval length of the detection section, calculate the detection cycle, and collect multi-mode sensor data of the detection point within the detection cycle. The leak detection module is used to identify the leak status of the detection point through the pipeline leak identification model. The leak status includes leaked and non-leaking, and determines whether to perform leak location operation. The location determination module is used to identify the target point from the detection points, mark the target section corresponding to the target point from the detection section, and collect the leakage characteristics of the target section, including the diffusion distance and leakage direction, and determine the leakage location within the target section. The model building module is used to build a basic model with leak nodes and import multi-mode sensor data into the information bits of the leak nodes to generate a visualized pipeline leak model.

2. The nitrogen and helium leak detection data analysis and processing system based on multi-sensor fusion according to claim 1, characterized in that, When marking the detection section, the two ends of the gas pipeline are marked, and these two ends are designated as the first and last section points. A equally spaced section points are marked within the gas pipeline. The gas pipeline between any two adjacent section points is recorded as the detection interval. Each detection interval is marked individually. The midpoint of each detection interval is identified, and the location of the midpoint is recorded as the detection point. There are 1 testing site.

3. The nitrogen and helium leak detection data analysis and processing system based on multi-sensor fusion according to claim 2, characterized in that, The steps for configuring leak detection points are as follows: respectively Centered on a detection point, and using a distance of one calibration length from the detection point as a standard, four points are marked in a ring at equal angles on the gas pipeline; The positions of the four points on the gas pipeline are continuously adjusted until the line connecting two non-adjacent points is parallel to the axis of the gas pipeline, thus obtaining four configuration points. An oxygen sensor and a flow rate sensor are deployed at two points on a line parallel to the axial line of the gas pipeline, respectively. A temperature sensor is deployed at one of the remaining points, and a vibration sensor and an ultrasonic sensor are deployed at the other point.

4. The nitrogen and helium leak detection data analysis and processing system based on multi-sensor fusion according to claim 3, characterized in that, The steps for determining the testing cycle are as follows: At the same time, the flow rate sensors detect the gas flow rate in each pipeline one by one. Real-time flow rate at each detection point, and The equilibrium flow velocity is calculated by averaging the sum of the real-time flow velocities. The length of the detection interval is measured and recorded as the interval length. The interval length is then compared with the equilibrium flow velocity to calculate the diffusion time. The diffusion time is reduced by one-fifth of the diffusion time, and the reduced diffusion time is recorded as the detection cycle.

5. The nitrogen and helium leak detection data analysis and processing system based on multi-sensor fusion according to claim 4, characterized in that, Multimodal sensing data includes temperature gradient values, oxygen concentration values, vibration frequency values, and sound frequency values; The steps for acquiring temperature gradient values ​​are as follows: At time T1, temperature data is collected via a temperature sensor. The ambient temperature at each detection point was obtained. One initial temperature value; After one detection cycle, data is collected at time T2. The ambient temperature at each detection point was obtained. One termination temperature value; Will The initial temperature value and After subtracting each termination temperature value, the result is calculated. A temperature gradient value.

6. The nitrogen and helium leak detection data analysis and processing system based on multi-sensor fusion according to claim 5, characterized in that, The training steps for the pipeline leak detection model are as follows: Multiple sets of multimode sensing data and the corresponding leakage status are collected in advance; Multimodal sensing data is converted into corresponding feature vectors, leakage status is converted into labels corresponding to multimodal sensing data, non-leaking is converted to 0, and leaked is converted to 1; The feature vectors are used as input to the machine learning model, and the leakage state corresponding to each set of feature vectors is used as the output of the machine learning model, with the leakage state as the prediction target. The machine learning model is trained with the goal of minimizing the sum of prediction errors of all training data until the sum of prediction errors converges, at which point training stops and a pipeline leak identification model is obtained.

7. The nitrogen and helium leak detection data analysis and processing system based on multi-sensor fusion according to claim 6, characterized in that, The steps for determining whether to perform a leak location operation are as follows: When the output of the pipeline leak identification model is 0, the leak status of the detection point is no leak; When the output of the pipeline leak identification model is 1, the leak status of the detection point is "leaked". The number of detection points that are in a leaking state is counted and recorded as abnormal values; When the abnormal value is 0, it is determined that no leak location operation will be performed; When the abnormal value is greater than 0, a leak location operation is performed.

8. The nitrogen and helium leak detection data analysis and processing system based on multi-sensor fusion according to claim 7, characterized in that, The steps for marking the target segment are as follows: When the abnormal value is 1, the detection point is recorded as an abnormal point, and the detection section where the abnormal point is located is recorded as the target section; When the abnormal value is greater than 1, the detection point is recorded as an abnormal point, resulting in D abnormal points. The detection segment where the abnormal point is located is recorded as the target segment, resulting in D target segments.

9. The nitrogen and helium leak detection data analysis and processing system based on multi-sensor fusion according to claim 8, characterized in that, The methods for collecting diffusion distance and leakage direction are as follows: The vibration frequency between the first calibration frequency and the second calibration frequency is recorded as the abnormal frequency. The time when the vibration sensor first detected the abnormal frequency is found by querying the timestamp and recorded as the first moment. A 1% drop in oxygen concentration is recorded as an abnormal phenomenon. After the first moment, the time when the oxygen concentration sensor first detected the abnormal phenomenon is retrieved by the timestamp and recorded as the second moment. The duration between the first moment and the second moment is recorded as the propagation duration. The room temperature and room pressure of D target sections at the first moment were detected. The diffusion coefficient of nitrogen under the room temperature and room pressure conditions was found by looking up a table. The diffusion distance was calculated by multiplying the propagation time by the diffusion coefficient of nitrogen. The previous detection segment of the target segment is recorded as the upper segment, and the real-time flow velocity of the detection points in the upper segment and the target segment at the first moment is queried to obtain the first flow velocity value and the second flow velocity value. When the first flow velocity value is greater than or equal to the second flow velocity value, the leakage direction is upstream; when the first flow velocity value is less than the second flow velocity value, the leakage direction is downstream.

10. The nitrogen and helium leak detection data analysis and processing system based on multi-sensor fusion according to claim 9, characterized in that, The steps for generating a pipeline leakage model are as follows: The point cloud data of all points in the gas pipeline is retrieved from the design drawings, and the point cloud data is summarized and modeled using 3D modeling technology to construct a basic model. Following the order from upstream to downstream, mark the points corresponding to the D leakage locations on the basic model in sequence to obtain D leakage nodes; Message frames are created one by one on each of the D leak nodes, and four independent information bits are divided within each message frame. The four information bits are then numbered in ascending order. Following the ascending order of numbering, the temperature gradient values, oxygen concentration values, vibration frequency values, and sound wave frequency values ​​of the D target intervals are sequentially imported into the four information bits of the corresponding leakage node, thereby converting the basic model into a pipeline leakage model.

Images