A gas tank piston tilt detection and early warning system based on CAN bridge

Abstract

The present invention provides a CAN bridge-based gas tank piston tilt detection and warning system, which relates to the technical field of tilt detection. The CAN bridge-based gas tank piston tilt detection and warning system includes: a point cloud data acquisition unit, a point cloud data transmission unit, a tilt angle analysis unit, and a tilt warning judgment unit. The present invention can quickly and accurately obtain the geometric shape and position information of the piston. The CAN network has high reliability and real-time performance, ensuring stable data transmission. By processing and analyzing laser point cloud data in a cloud server, the tilt angle of the gas tank cabinet piston can be accurately calculated. Based on the tilt angle of the gas tank cabinet piston, analysis is performed. When the warning mechanism is triggered, a warning signal can be promptly issued to the operator so that timely measures can be taken to address the problem, thereby improving the accuracy and safety of gas tank piston tilt detection.

Description

Technical Field

[0001] The present invention relates to the technical field of inclination detection, and in particular to a gas cabinet piston inclination detection and early warning system based on a CAN network bridge. Background Art

[0002] Gasholders in steel plants play a crucial role in environmental protection and resource utilization. Their ability to instantly handle gas flow helps recover transient gas emissions caused by uneven production. Simply put, gasholders can effectively handle frequent, short-term gas fluctuations that boiler rooms struggle to accommodate. During production, gas production and demand don't always perfectly match, leading to gas surpluses and shortages. When there's excess gas, it can be stored in gasholders for future use. Conversely, when gas is insufficient or pipeline pressure is low, gas can be withdrawn from gasholders and replenished in the pipeline network, ensuring continuous production. In this way, gasholders serve as a "surplus to compensate for shortfalls," reducing gas emissions and effectively conserving resources.

[0003] Gas holders play a key role in maintaining the balance of the gas pipeline network. They control the storage and release of gas by moving the piston up and down. During the operation of the gas holder, the tilt angle of the piston changes dynamically. Precise control of the piston tilt is very important to ensure the stability and safe operation of the gas holder. At present, traditional piston tilt detection methods may not be able to accurately detect the tilt of the gas holder piston, and thus cannot accurately issue early warnings, which may cause equipment damage or more serious safety accidents. At the same time, in traditional tilt detection, the transmission and processing of data may be inefficient, which may affect the real-time performance of the piston tilt detection and have a negative impact, which in turn makes it impossible for staff to respond to and handle abnormal situations in a timely manner.

[0004] Currently, no effective solutions have been proposed for the problems in related technologies. Summary of the Invention

[0005] In view of this, the present invention provides a gas tank piston tilt detection and warning system based on a CAN bridge to solve the problems of low detection accuracy and poor real-time performance in the above-mentioned traditional piston tilt detection method.

[0006] In order to solve the above problems, the specific technical solutions adopted by the present invention are as follows:

[0007] A gas tank piston tilt detection and warning system based on a CAN network bridge, the gas tank piston tilt detection and warning system based on a CAN network bridge comprising: a point cloud data acquisition unit, a point cloud data transmission unit, a tilt angle analysis unit, and a tilt warning judgment unit, wherein the point cloud data acquisition unit, the point cloud data transmission unit, the tilt angle analysis unit, and the tilt warning judgment unit are sequentially connected;

[0008] A point cloud data acquisition unit is used to acquire laser point cloud data of the gas tank cabinet piston through a laser radar, where the laser point cloud data includes initial point cloud data and dynamic point cloud data;

[0009] Point cloud data transmission unit, used to transmit the acquired laser point cloud data to the cloud server based on the preset CAN bridge;

[0010] The tilt angle analysis unit is used to obtain the laser point cloud data in the cloud server, analyze the posture of the gas cabinet piston, and determine the tilt angle of the gas cabinet piston;

[0011] The tilt warning judgment unit is used to perform tilt warning analysis on the gas cabinet piston according to the tilt angle of the gas cabinet piston, and send a warning signal to the operator after the gas cabinet piston triggers the warning mechanism.

[0012] Preferably, the tilt angle analysis unit includes: a preprocessing module, a normal vector extraction module, a center line extraction module and a tilt angle calculation module, and the preprocessing module, the normal vector extraction module, the center line extraction module and the tilt angle calculation module are connected in sequence;

[0013] A preprocessing module, used to preprocess the initial point cloud data and dynamic point cloud data respectively;

[0014] Normal vector extraction module, used to extract normal vectors from pre-processed initial point cloud data and dynamic point cloud data respectively;

[0015] The centerline extraction module is used to extract the centerlines of the initial point cloud data and the dynamic point cloud data respectively based on the extracted normal vector using the piston centerline extraction algorithm;

[0016] The tilt angle calculation module is used to calculate the piston tilt angle based on the center line of the initial point cloud data and the dynamic point cloud data.

[0017] Preferably, preprocessing the initial point cloud data and the dynamic point cloud data respectively includes:

[0018] The initial point cloud data and the dynamic point cloud data are respectively cropped by a preset straight-through filter to remove the background point cloud data in the initial point cloud data and the dynamic point cloud data;

[0019] Perform uniform downsampling on the cropped initial point cloud data and dynamic point cloud data respectively;

[0020] The statistical filtering algorithm is used to remove noise from the initial point cloud data and dynamic point cloud data after downsampling.

[0021] Preferably, extracting normal vectors from the pre-processed initial point cloud data and the dynamic point cloud data respectively includes:

[0022] Determine the optimal neighborhood size of each point in the preprocessed initial point cloud data and the dynamic point cloud data respectively;

[0023] Perform preliminary normal vector estimation for each point in the initial point cloud data and dynamic point cloud data based on the optimal neighborhood size of each point;

[0024] The estimated preliminary normal vector is checked for orientation consistency. If the orientation is inconsistent, the normal vector direction is flipped to obtain the final normal vector.

[0025] Preferably, determining the optimal neighborhood size of each point in the pre-processed initial point cloud data and the dynamic point cloud data comprises the following steps:

[0026] Step 1: Define the maximum and minimum values ​​of the neighborhood search radius and the step change value, and initialize the neighborhood search radius to the minimum value;

[0027] Step 2: Calculate the covariance matrix of each point in the preprocessed initial point cloud data and the dynamic point cloud data using the PCA algorithm, and perform eigenvalue decomposition on the calculated covariance matrix to obtain the eigenvalue of each point;

[0028] Step 3: Based on the obtained eigenvalues, calculate the information entropy value of the neighborhood using the information entropy value calculation method;

[0029] Step 4: Update the current search radius value of the neighborhood by taking the sum of the search radius value and the step change value of the current neighborhood;

[0030] Step 5: If the updated neighborhood search radius value is less than or equal to the defined maximum radius value, repeat steps 2 to 5; otherwise, execute step 6.

[0031] Step 6: Compare the information entropy values ​​corresponding to each search radius, and select the search radius corresponding to the minimum information entropy value as the optimal neighborhood size.

[0032] Preferably, the calculation formula of the covariance matrix is:

[0033] ;

[0034] Where, Grepresents the covariance matrix;

[0035] p m Represents the first m points;

[0036] Represents the centroid of the neighborhood of any point in the point cloud data;

[0037] n Represents the number of points in the neighborhood of any point in the point cloud data;

[0038] T Represents the transpose of a matrix.

[0039] Preferably, performing preliminary normal vector estimation for each point based on the optimal neighborhood size of each point includes:

[0040] Construct a local neighborhood for each point in the initial point cloud data and the dynamic point cloud data based on the optimal neighborhood size of each point;

[0041] According to the local neighborhood of each point, the average position of each point in the local neighborhood is calculated and used as the center of gravity of the local neighborhood;

[0042] For each point in the local neighborhood, calculate the difference between its coordinates and the coordinates of the center of gravity, and construct a difference matrix based on the difference;

[0043] Perform singular value decomposition on the difference matrix to obtain singular values ​​and singular vectors, and select the singular vector corresponding to the smallest singular value as the preliminary normal vector.

[0044] Preferably, the estimated preliminary normal vector is checked for orientation consistency. If the orientation is inconsistent, the normal vector direction is flipped, and the final normal vector is obtained, including:

[0045] Select a reference point and its corresponding preliminary normal vector from the estimated preliminary normal vector;

[0046] For each remaining point, calculate the dot product of its preliminary normal vector and the reference point normal vector;

[0047] If the dot product is positive, the directions of the two normal vectors are the same; if the dot product is negative, the directions of the two normal vectors are opposite;

[0048] Flip the normal vectors whose dot products are negative to get the final normal vector with the same direction.

[0049] Preferably, based on the extracted normal vector, extracting the center lines of the initial point cloud data and the dynamic point cloud data respectively by using a piston center line extraction algorithm includes:

[0050] Randomly select several points from the final normal vectors of the initial point cloud data and the dynamic point cloud data as sample sets;

[0051] According to the selected sample set, the initial linear model is fitted by the least squares method;

[0052] Calculate the distances from the unselected points in the final normal vectors of the initial point cloud data and the dynamic point cloud data to the linear model respectively, and use them as error values;

[0053] If the error values ​​of all unselected points are less than the preset threshold, they will be regarded as internal points;

[0054] The initial linear model is updated based on the obtained interior points until the preset conditions are met, and the optimal linear model is obtained and used as the center line.

[0055] Preferably, calculating the piston tilt angle according to the center lines of the initial point cloud data and the dynamic point cloud data includes:

[0056] Select two points on the center lines of the initial point cloud data and the dynamic point cloud data respectively, and obtain the coordinate values ​​of the two points on each center line;

[0057] Calculate the direction vector of the center line of the initial point cloud data and the dynamic point cloud data according to the coordinate values ​​of two points on each center line;

[0058] The angle between the center lines of the initial point cloud data and the dynamic point cloud data is calculated using the angle calculation method between vectors, and is used as the inclination angle of the piston.

[0059] The beneficial effects of the present invention are:

[0060] 1. The present invention uses a laser radar to obtain laser point cloud data of the gas cabinet piston, which can quickly and accurately obtain the geometric shape and position information of the piston. The laser point cloud data is transmitted to the cloud server based on a preset CAN bridge. The CAN network has high reliability and real-time performance, which can ensure stable data transmission. By processing and analyzing the laser point cloud data in the cloud server, the inclination angle of the gas cabinet piston can be accurately calculated. After analysis is performed based on the inclination angle of the gas cabinet piston, after the early warning mechanism is triggered, an early warning signal can be promptly issued to the operator so that timely measures can be taken to deal with it, thereby improving the accuracy and safety of gas cabinet piston tilt detection.

[0061] 2. The present invention can significantly reduce the amount of data by cropping the background point cloud data and performing downsampling processing, thereby reducing the complexity and computational requirements of subsequent processing and analysis. By removing noise from the data through a statistical filtering algorithm, the quality of the data can be improved, thereby improving the quality of the final result.

[0062] 3. The present invention can better capture and represent the geometric shape and structural characteristics of the point cloud by extracting the normal vector of each point, thereby improving the data representation ability. The use of the optimal neighborhood size for normal vector estimation can take into account the local scale and density changes of the point cloud data, thereby improving the robustness of the processing. Through the calculation of the information entropy value, the selection of the neighborhood can be further optimized, making the normal vector estimation more accurate and stable. Through reasonable neighborhood search and normal vector estimation strategies, the processing efficiency can be effectively improved and unnecessary computing and storage requirements can be reduced.

[0063] 4. The present invention can improve measurement accuracy by extracting the center line through the normal vector, because the normal vector can well indicate the surface features in the point cloud data, and this process can be performed automatically, reducing human measurement errors and workload. For dynamic point cloud data, the center line and tilt angle can be updated in real time, which is very important for real-time monitoring and control. The robustness of the model is increased through least squares fitting and iterative updating. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative work. In the drawings:

[0065] Figure 1 The present invention is a block diagram of a CAN bridge-based gas tank piston tilt detection and warning system according to an embodiment of the present invention.

[0066] In the picture:

[0067] 1. Point cloud data acquisition unit; 2. Point cloud data transmission unit; 3. Tilt angle analysis unit; 301. Preprocessing module; 302. Normal vector extraction module; 303. Centerline extraction module; 304. Tilt angle calculation module; 4. Tilt warning judgment unit. DETAILED DESCRIPTION

[0068] In order to enable those skilled in the art to better understand the technical solutions in this application, the following will clearly and completely describe the technical solutions in the embodiments of this application in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments of this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of this application.

[0069] According to an embodiment of the present invention, a gas tank piston tilt detection and warning system based on a CAN network bridge is provided.

[0070] The present invention will now be further described with reference to the accompanying drawings and specific embodiments. Figure 1 As shown, according to an embodiment of the present invention, a gas tank piston tilt detection and warning system based on a CAN network bridge comprises: a point cloud data acquisition unit 1, a point cloud data transmission unit 2, a tilt angle analysis unit 3, and a tilt warning judgment unit 4, wherein the point cloud data acquisition unit 1, the point cloud data transmission unit 2, the tilt angle analysis unit 3, and the tilt warning judgment unit 4 are connected in sequence;

[0071] Point cloud data acquisition unit 1, used to acquire laser point cloud data of the gas cabinet piston through laser radar, the laser point cloud data including initial point cloud data and dynamic point cloud data;

[0072] Specifically, lidar is a remote sensing technology that can measure the distance to the surface of an object by emitting laser pulses and receiving reflected signals. Through lidar, laser point cloud data of the gas tank cabinet piston can be obtained.

[0073] The initial point cloud data represents the state of the gas cabinet piston before displacement, and the dynamic point cloud data represents continuous scanning during the movement of the gas cabinet piston. Each scan generates a new point cloud, which records the dynamic changes of the piston. All dynamic scanning results can be combined into a dynamic point cloud data sequence.

[0074] Point cloud data transmission unit 2, used to transmit the acquired laser point cloud data to the cloud server based on a preset CAN bridge;

[0075] It should be noted that a CAN bridge is a device used to connect two or more Controller Area Network (CAN) systems and forward information between them. CAN is a communication protocol widely used in automotive and industrial automation. It allows multiple microcontrollers and devices to communicate over a simple two-wire bus without a host computer.

[0076] During transmission, first, the acquired laser point cloud data is serialized and encapsulated into CAN messages, and the encapsulated CAN messages are sent to the preset CAN bridge. The CAN bridge converts the received CAN messages into data packets suitable for transmission over the network and sends them to the cloud server through the preset network connection. On the cloud server side, the received data packets need to be parsed and converted into original laser point cloud data.

[0077] The tilt angle analysis unit 3 is used to obtain the laser point cloud data in the cloud server, analyze the posture of the gas cabinet piston, and determine the tilt angle of the gas cabinet piston;

[0078] As a preferred embodiment, the tilt angle analysis unit 3 includes: a pre-processing module 301, a normal vector extraction module 302, a center line extraction module 303 and a tilt angle calculation module 304, and the pre-processing module 301, the normal vector extraction module 302, the center line extraction module 303 and the tilt angle calculation module 304 are connected in sequence;

[0079] A preprocessing module 301 is used to preprocess the initial point cloud data and the dynamic point cloud data respectively;

[0080] Specifically, the preprocessing of the initial point cloud data and the dynamic point cloud data includes:

[0081] The initial point cloud data and the dynamic point cloud data are respectively cropped by a preset straight-through filter to remove the background point cloud data in the initial point cloud data and the dynamic point cloud data;

[0082] Specifically, a straight-through filter is a method commonly used in point cloud data processing. It can crop the point cloud on a specific coordinate axis according to a set threshold. For the initial point cloud data and dynamic point cloud data, appropriate thresholds are set respectively to remove the background point clouds in these data.

[0083] Perform uniform downsampling on the cropped initial point cloud data and dynamic point cloud data respectively;

[0084] Specifically, downsampling is the process of reducing the amount of point cloud data. It can simplify the data by reducing the number of points while trying to maintain the shape and characteristics of the point cloud. Uniform downsampling methods include voxel grid downsampling and random sampling. Voxel grid downsampling creates a three-dimensional grid (voxel) in the point cloud and merges all points into a representative point in each voxel, thereby reducing the number of points. Random sampling randomly selects a part of the points in the point cloud.

[0085] The statistical filtering algorithm is used to remove noise from the initial point cloud data and dynamic point cloud data after downsampling.

[0086] Specifically, performing noise removal on the downsampled initial point cloud data and dynamic point cloud data using a statistical filtering algorithm includes the following steps:

[0087] Determine the filtering parameters, including neighborhood size and standard deviation multiple;

[0088] For each point, the average distance of all points in its neighborhood is calculated and compared with the overall average distance of the neighborhood. If the average distance of a point is greater than the overall average distance plus the set standard deviation multiple, it will be regarded as an outlier (noise) and removed from the dataset.

[0089] A normal vector extraction module 302 is used to extract normal vectors from the pre-processed initial point cloud data and the dynamic point cloud data respectively;

[0090] Specifically, extracting normal vectors from the pre-processed initial point cloud data and dynamic point cloud data includes:

[0091] Determine the optimal neighborhood size of each point in the preprocessed initial point cloud data and the dynamic point cloud data respectively;

[0092] Specifically, determining the optimal neighborhood size of each point in the preprocessed initial point cloud data and the dynamic point cloud data includes the following steps:

[0093] Step 1: Define the maximum and minimum values ​​of the neighborhood search radius and the step change value, and initialize the neighborhood search radius to the minimum value;

[0094] It should be noted that the maximum and minimum values ​​determine the radius range of the search neighborhood, and the step change value determines how the neighborhood search radius gradually increases.

[0095] Step 2: Calculate the covariance matrix of each point in the preprocessed initial point cloud data and the dynamic point cloud data using the PCA algorithm, and perform eigenvalue decomposition on the calculated covariance matrix to obtain the eigenvalue of each point;

[0096] It should be noted that the eigenvalue decomposition is obtained by solving the characteristic equation. Solving this equation will obtain the eigenvalues ​​of the covariance matrix. This is usually done using a linear algebra library in mathematical software or programming languages. For each eigenvalue, the corresponding eigenvector is found by solving a system of linear equations.

[0097] Specifically, the calculation formula of the covariance matrix is:

[0098] ;

[0099] Where, G represents the covariance matrix;

[0100] p m Represents the first m points;

[0101] Represents the centroid of the neighborhood of any point in the point cloud data;

[0102] n Represents the number of points in the neighborhood of any point in the point cloud data;

[0103] T Represents the transpose of a matrix.

[0104] Step 3: Based on the obtained eigenvalues, calculate the information entropy value of the neighborhood using the information entropy value calculation method;

[0105] It should be noted that to calculate the information entropy value of the neighborhood, the eigenvalues ​​need to be normalized, each normalized eigenvalue is multiplied by the negative of its natural logarithm, and then the results of all eigenvalues ​​are summed to obtain the information entropy value.

[0106] Step 4: Update the current search radius value of the neighborhood by taking the sum of the search radius value and the step change value of the current neighborhood;

[0107] Step 5: If the updated neighborhood search radius value is less than or equal to the defined maximum radius value, repeat steps 2 to 5; otherwise, execute step 6.

[0108] Step 6: Compare the information entropy values ​​corresponding to each search radius, and select the search radius corresponding to the minimum information entropy value as the optimal neighborhood size.

[0109] Perform preliminary normal vector estimation for each point in the initial point cloud data and dynamic point cloud data based on the optimal neighborhood size of each point;

[0110] Specifically, the preliminary normal vector estimation of each point based on the optimal neighborhood size of each point includes:

[0111] Construct a local neighborhood for each point in the initial point cloud data and the dynamic point cloud data based on the optimal neighborhood size of each point;

[0112] It should be noted that based on the optimal neighborhood size of each point, a local neighborhood can be constructed through a spatial search algorithm.

[0113] According to the local neighborhood of each point, the average position of each point in the local neighborhood is calculated and used as the center of gravity of the local neighborhood;

[0114] It should be noted that for each point, its local neighborhood is first determined. This local neighborhood includes all points around the point, and its range is determined by the optimal neighborhood size determined previously. Then the sum of the coordinate values ​​of all points in the local neighborhood is calculated. The average value of the sum of these coordinates is obtained by dividing the sum of each coordinate by the number of points in the neighborhood, and then used as the center of gravity of the neighborhood.

[0115] For each point in the local neighborhood, calculate the difference between its coordinates and the coordinates of the center of gravity, and construct a difference matrix based on the difference;

[0116] Perform singular value decomposition on the difference matrix to obtain singular values ​​and singular vectors, and select the singular vector corresponding to the smallest singular value as the preliminary normal vector.

[0117] It should be noted that singular value decomposition (SVD) is a powerful linear algebra tool that can be used to process and analyze matrix data. In point cloud data processing, SVD is used to analyze the difference matrix to obtain the preliminary normal vector of each point. The result of singular value decomposition of the difference matrix can obtain all singular values ​​and corresponding singular vectors. The singular values ​​represent the degree of change or dispersion of the data in the direction of the corresponding singular vectors, while the singular vectors represent these directions.

[0118] The estimated preliminary normal vector is checked for orientation consistency. If the orientation is inconsistent, the normal vector direction is flipped to obtain the final normal vector.

[0119] Specifically, the estimated preliminary normal vector is checked for orientation consistency. If the orientation is inconsistent, the normal vector direction is flipped, and the final normal vector is obtained, including:

[0120] Select a reference point and its corresponding preliminary normal vector from the estimated preliminary normal vector;

[0121] For each remaining point, calculate the dot product of its preliminary normal vector and the reference point normal vector;

[0122] If the dot product is positive, the directions of the two normal vectors are the same; if the dot product is negative, the directions of the two normal vectors are opposite;

[0123] It should be noted that if the dot product is positive, it means that the directions of the two vectors are roughly the same, because the angle between them in space is less than 90 degrees. If the dot product is negative, it means that the directions of the two vectors are opposite, that is, the angle between them in space is greater than 90 degrees.

[0124] Flip the normal vectors whose dot products are negative to get the final normal vector with the same direction.

[0125] It should be noted that for those normal vectors that are in the opposite direction of the reference normal vector (that is, normal vectors whose dot product is negative), flip them. Flipping a vector means multiplying all its component values ​​by -1.

[0126] A centerline extraction module 303 is used to extract the centerlines of the initial point cloud data and the dynamic point cloud data respectively using a piston centerline extraction algorithm based on the extracted normal vector;

[0127] Specifically, based on the extracted normal vector, the center lines of the initial point cloud data and the dynamic point cloud data are extracted using the piston center line extraction algorithm, including:

[0128] Randomly select several points from the final normal vectors of the initial point cloud data and the dynamic point cloud data as sample sets;

[0129] It should be noted that two points are selected on the center line of the piston, two from the center line of the initial point cloud data and two from the center line of the dynamic point cloud data, and the three-dimensional coordinates of these four points are recorded.

[0130] According to the selected sample set, the initial linear model is fitted by the least squares method;

[0131] It should be noted that first, a data matrix is ​​constructed based on the selected sample set. Assuming there are A samples, each with B features, then the data matrix will be an A×B matrix, where each row represents a sample and each column represents a feature. Next, a target vector is constructed to represent the target value of the fit. The length of this vector should be the same as the number of samples. Then, the least squares method is used to find the optimal linear model. Mathematically, the goal of the least squares method is to find a parameter vector that minimizes the sum of the squares of the difference between the product of the data matrix and this parameter vector and the target vector.

[0132] Calculate the distances from the unselected points in the final normal vectors of the initial point cloud data and the dynamic point cloud data to the linear model respectively, and use them as error values;

[0133] It should be noted that calculating the distance from the unselected points to the linear model can be achieved according to the Euclidean distance calculation method. The Euclidean distance is a commonly used distance calculation method, which can be used in spaces of any dimension.

[0134] If the error values ​​of all unselected points are less than the preset threshold, they will be regarded as internal points;

[0135] The initial linear model is updated based on the obtained interior points until the preset conditions are met, and the optimal linear model is obtained and used as the center line.

[0136] It should be noted that using least squares or other fitting methods to refit the model based on the inlier data is more accurate than a model based on all points because inliers are closer to the true model. Using this new model, the distance from all unselected points (including inliers and outliers from the previous round) to the model is recalculated. This distance becomes the new error value. Using the same threshold, the inliers and outliers are redefined based on the new error value. All points with an error value less than the threshold are considered inliers; otherwise, they are outliers. The algorithm then checks whether a predefined termination condition is met. This condition could be the maximum number of iterations or the improvement in the model being less than a threshold. If the termination condition is met, the iteration ends, and the current model is considered the optimal model. Otherwise, the iteration continues. At the end of the iteration, the current model is considered the optimal model and is used as the centerline.

[0137] The tilt angle calculation module 304 is used to calculate the tilt angle of the piston based on the center lines of the initial point cloud data and the dynamic point cloud data.

[0138] Specifically, the calculation of the piston tilt angle based on the center line of the initial point cloud data and the dynamic point cloud data includes:

[0139] Select two points on the center lines of the initial point cloud data and the dynamic point cloud data respectively, and obtain the coordinate values ​​of the two points on each center line;

[0140] Calculate the direction vector of the center line of the initial point cloud data and the dynamic point cloud data according to the coordinate values ​​of two points on each center line;

[0141] It should be noted that the direction vectors of the center lines of the initial point cloud data and the dynamic point cloud data are calculated based on the coordinate values ​​of two points on each center line. The direction vector can be calculated by the coordinate difference between the two points.

[0142] The angle between the center lines of the initial point cloud data and the dynamic point cloud data is calculated using the angle calculation method between vectors, and is used as the inclination angle of the piston.

[0143] It should be noted that the dot product formula of vectors is usually used to calculate the angle between two vectors.

[0144] The tilt warning judgment unit 4 is used to perform tilt warning analysis on the gas cabinet piston according to the tilt angle of the gas cabinet piston, and send a warning signal to the operator after the gas cabinet piston triggers the warning mechanism.

[0145] Specifically, first, it is necessary to determine the safe tilt angle threshold of the piston. This threshold is determined based on the equipment's design parameters, operating safety standards and historical data. When the monitored tilt angle exceeds this threshold, the system will trigger an early warning mechanism.

[0146] Once the tilt angle of the piston exceeds the safety threshold, the preset early warning system should be activated immediately. The early warning mechanism may include automatically adjusting the control system to correct the tilt, issuing visual and sound alarms, etc. In addition to the automatic early warning mechanism, there should also be a communication system to transmit the early warning information to the operator, which can be achieved through e-mail, text messages, automatic telephone calls or the alarm system in the control room.

[0147] In summary, with the help of the above technical solution of the present invention, the present invention obtains the laser point cloud data of the gas cabinet piston through the laser radar, which can quickly and accurately obtain the geometric shape and position information of the piston, and adopts a preset CAN bridge to transmit the laser point cloud data to the cloud server. The CAN network has high reliability and real-time performance, which can ensure stable data transmission. By processing and analyzing the laser point cloud data in the cloud server, the inclination angle of the gas cabinet piston can be accurately calculated, and the inclination angle of the gas cabinet piston can be analyzed according to the inclination angle of the gas cabinet piston. After the early warning mechanism is triggered, an early warning signal can be sent to the operator in time so that timely measures can be taken to deal with it, thereby improving the accuracy and safety of gas cabinet piston tilt detection; the present invention can significantly reduce the amount of data by cropping the background point cloud data and performing downsampling processing, thereby reducing the complexity and calculation requirements of subsequent processing and analysis, and can improve the quality of the data by removing noise from the data through a statistical filtering algorithm, thereby being able to Improve the quality of the final result; the present invention can better capture and represent the geometric shape and structural characteristics of the point cloud by extracting the normal vector of each point, thereby improving the characterization ability of the data, and adopting the optimal neighborhood size for normal vector estimation, which can take into account the local scale and density changes of the point cloud data, thereby improving the robustness of the processing, and through the calculation of the information entropy value, the selection of the neighborhood can be further optimized, making the normal vector estimation more accurate and stable, and through reasonable neighborhood search and normal vector estimation strategies, the processing efficiency can be effectively improved and unnecessary calculation and storage requirements can be reduced; the present invention can improve the measurement accuracy by extracting the center line through the normal vector, because the normal vector can well indicate the surface features in the point cloud data, and this process can be automatically executed, reducing human measurement errors and workload, and for dynamic point cloud data, it can provide real-time updates of the center line and tilt angle, which is very important for real-time monitoring and control, and increase the robustness of the model through least squares fitting and iterative updates.

[0148] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, optical storage, etc.) containing computer-usable program code.

[0149] The specific embodiments described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A gas tank piston tilt detection and warning system based on CAN bridge, characterized in that: include: A point cloud data acquisition unit (1) is used to acquire laser point cloud data of a gas tank cabinet piston through a laser radar, wherein the laser point cloud data includes initial point cloud data and dynamic point cloud data; A point cloud data transmission unit (2), configured to transmit the acquired laser point cloud data to a cloud server based on a preset CAN bridge; The tilt angle analysis unit (3) is used to obtain the laser point cloud data in the cloud server, analyze the posture of the gas cabinet piston, and determine the tilt angle of the gas cabinet piston; A tilt warning judgment unit (4) is used to perform tilt warning analysis on the gas cabinet piston according to the tilt angle of the gas cabinet piston, and send a warning signal to the operator after the gas cabinet piston triggers the warning mechanism; The tilt angle analysis unit (3) comprises: The normal vector extraction module (302) is used to determine the optimal neighborhood size of each point in the pre-processed initial point cloud data and the dynamic point cloud data respectively; construct a local neighborhood for each point in the initial point cloud data and the dynamic point cloud data based on the optimal neighborhood size of each point; calculate the average position of each point in the local neighborhood according to the local neighborhood of each point and use it as the center of gravity of the local neighborhood; for each point in the local neighborhood, calculate the difference between its coordinate values ​​and the coordinate value of the center of gravity, and construct a difference matrix based on the difference; perform singular value decomposition on the difference matrix to obtain singular values ​​and singular vectors, and select the singular vector corresponding to the smallest singular value as the preliminary normal vector; perform a direction consistency check on the estimated preliminary normal vector, and if the direction is inconsistent, flip the direction of the normal vector to obtain the final normal vector.

2. The gas tank piston tilt detection and warning system based on CAN bridge according to claim 1 is characterized in that: The tilt angle analysis unit (3) further comprises: A preprocessing module (301), configured to preprocess the initial point cloud data and the dynamic point cloud data respectively; A centerline extraction module (303) is used to extract the centerlines of the initial point cloud data and the dynamic point cloud data respectively using a piston centerline extraction algorithm based on the extracted normal vector; The tilt angle calculation module (304) is used to calculate the tilt angle of the piston based on the center lines of the initial point cloud data and the dynamic point cloud data.

3. The CAN bridge-based gas tank piston tilt detection and warning system according to claim 2 is characterized in that: The pre-processing of the initial point cloud data and the dynamic point cloud data comprises: The initial point cloud data and the dynamic point cloud data are respectively cropped by a preset straight-through filter to remove the background point cloud data in the initial point cloud data and the dynamic point cloud data; Perform uniform downsampling on the cropped initial point cloud data and dynamic point cloud data respectively; The statistical filtering algorithm is used to remove noise from the initial point cloud data and dynamic point cloud data after downsampling.

4. The CAN bridge-based gas tank piston tilt detection and warning system according to claim 3 is characterized in that: Determining the optimal neighborhood size of each point in the pre-processed initial point cloud data and the dynamic point cloud data comprises the following steps: Step 1: Define the maximum and minimum values ​​of the neighborhood search radius and the step change value, and initialize the neighborhood search radius to the minimum value; Step 2: Calculate the covariance matrix of each point in the preprocessed initial point cloud data and the dynamic point cloud data using the PCA algorithm, and perform eigenvalue decomposition on the calculated covariance matrix to obtain the eigenvalue of each point; Step 3: Based on the obtained eigenvalues, calculate the information entropy value of the neighborhood using the information entropy value calculation method; Step 4: Update the current search radius value of the neighborhood by taking the sum of the search radius value and the step change value of the current neighborhood; Step 5: If the updated neighborhood search radius value is less than or equal to the defined maximum radius value, repeat steps 2 to 5; otherwise, execute step 6. Step 6: Compare the information entropy values ​​corresponding to each search radius, and select the search radius corresponding to the minimum information entropy value as the optimal neighborhood size.

5. The CAN bridge-based gas tank piston tilt detection and warning system according to claim 4 is characterized in that: The calculation formula of the covariance matrix is: Where G represents the covariance matrix; p m Represents the mth point in the neighborhood of any point in the point cloud data; Represents the centroid of the neighborhood of any point in the point cloud data; n represents the number of points in the neighborhood of any point in the point cloud data; T represents the transpose of the matrix.

6. The CAN bridge-based gas tank piston tilt detection and warning system according to claim 5 is characterized in that: The process of performing a direction consistency check on the estimated preliminary normal vector and flipping the normal vector direction if the direction is inconsistent, and obtaining the final normal vector includes: Select a reference point and its corresponding preliminary normal vector from the estimated preliminary normal vector; For each remaining point, calculate the dot product of its preliminary normal vector and the reference point normal vector; If the dot product is positive, the directions of the two normal vectors are the same; if the dot product is negative, the directions of the two normal vectors are opposite; Flip the normal vectors whose dot products are negative to get the final normal vector with the same direction.

7. The CAN bridge-based gas tank piston tilt detection and warning system according to claim 6 is characterized in that: The extracting of the center lines of the initial point cloud data and the dynamic point cloud data respectively by using a piston center line extraction algorithm based on the extracted normal vector includes: Randomly select several points from the final normal vectors of the initial point cloud data and the dynamic point cloud data as sample sets; According to the selected sample set, the initial linear model is fitted by the least squares method; Calculate the distances from the unselected points in the final normal vectors of the initial point cloud data and the dynamic point cloud data to the linear model respectively, and use them as error values; If the error values ​​of all unselected points are less than the preset threshold, they will be regarded as internal points; The initial linear model is updated based on the obtained interior points until the preset conditions are met, and the optimal linear model is obtained and used as the center line.

8. The CAN bridge-based gas tank piston tilt detection and warning system according to claim 2 is characterized in that: Calculating the piston tilt angle according to the center lines of the initial point cloud data and the dynamic point cloud data includes: Select two points on the center lines of the initial point cloud data and the dynamic point cloud data respectively, and obtain the coordinate values ​​of the two points on each center line; Calculate the direction vector of the center line of the initial point cloud data and the dynamic point cloud data according to the coordinate values ​​of two points on each center line; The angle between the center lines of the initial point cloud data and the dynamic point cloud data is calculated using the angle calculation method between vectors, and is used as the inclination angle of the piston.

Images