A gas pipeline inspection and maintenance method based on augmented reality technology

Abstract

This invention provides a gas pipeline inspection and maintenance method based on augmented reality technology. This involves connecting an intelligent inspection system and an edge computing platform through a wearable augmented reality device, using a multimodal sensor network to scan the environment and generate a three-dimensional model. Real-time monitoring of pipeline status is combined with quantum image processing technology to identify fault points, and data is processed using a distributed Turing learning algorithm. Fault points are annotated and identified using a quantum causal inference model, displaying personalized repair recommendations. Real-time operational guidance is provided during the maintenance process, with data automatically recorded and uploaded to the cloud via the Internet of Things. Big data and machine learning are used for fault prediction and maintenance optimization, and expert guidance is provided through remote communication.

Description

Technical Field

[0001] The present invention relates to the field of augmented reality technology, and in particular to a gas pipeline inspection and maintenance method based on augmented reality technology. Background Art

[0002] With the continuous development of modern industry and the acceleration of urbanization, the safety and reliability of gas pipelines, as an important part of urban infrastructure, have attracted much attention. Traditional gas pipeline inspection and maintenance methods mainly rely on manual operations and regular inspections, which have many shortcomings.

[0003] First, manual inspections are inefficient and lack accuracy. Traditional inspections rely on manual visual inspections of pipelines and simple measurement tools. Limited by human experience and tool limitations, these methods struggle to comprehensively and accurately identify potential fault points. Furthermore, manual inspections are time-consuming and often fail to detect and address sudden faults in a timely manner, posing safety risks.

[0004] Secondly, traditional maintenance methods rely on manual labor, which limits efficiency and accuracy. Due to the complexity of the maintenance process, manual labor makes it difficult to ensure the accuracy of each step. Especially in complex maintenance scenarios, the experience and skills of the maintenance personnel become crucial factors in determining the quality of the repair. This method is not only time-consuming and labor-intensive, but also prone to operational errors, resulting in inconsistent maintenance quality.

[0005] Third, existing inspection and maintenance methods struggle to achieve real-time data collection and analysis. Traditional methods rely primarily on paper records and simple spreadsheets for data management, which is incapable of real-time data collection, transmission, and analysis. Data lags and inaccuracies hinder accurate assessment of gas pipeline status and preventive maintenance.

[0006] To address these issues, a number of new technologies and methods have emerged in recent years, such as drone and robotic inspections. While these methods have improved inspection efficiency and accuracy to a certain extent, they still have some shortcomings. For example, drone inspections are limited by flight altitude and environmental conditions, making it difficult to fully cover all pipelines. Robotic inspections are limited by their mobility and environmental adaptability, limiting their application in complex environments. Furthermore, these methods still rely on traditional technologies for data processing and analysis, making it difficult to achieve real-time analysis and processing of large-scale, multi-source data, and unable to effectively support preventive maintenance and intelligent decision-making. Therefore, finding a gas pipeline inspection and maintenance method based on augmented reality technology is an urgent issue for those skilled in the art. Summary of the Invention

[0007] One purpose of the present invention is to propose a gas pipeline inspection and maintenance method based on augmented reality technology. One purpose of the present invention is to propose a gas pipeline inspection and maintenance method based on augmented reality technology. By integrating multimodal sensor networks, quantum image processing technology, distributed Turing learning algorithm and Internet of Things technology, comprehensive monitoring and intelligent analysis of the gas pipeline status are achieved, and it has the advantages of high efficiency, accuracy, safety and intelligence.

[0008] According to an embodiment of the present invention, a gas pipeline inspection and maintenance method based on augmented reality technology includes the following steps:

[0009] S1. Wear an augmented reality device, connect it to the intelligent inspection system and edge computing platform, use the multimodal sensor network to perform an initial environment and equipment status scan, and generate 3D environment model data;

[0010] S2. Use augmented reality devices to view real-time gas pipeline status information, including pipeline layout, direction, and connection points. Combined with 3D environmental model data and quantum image processing technology, potential fault points can be identified and located in real time.

[0011] S3 uses augmented reality technology to overlay and display pipeline status data in real time, combines historical data for multi-dimensional anomaly detection and trend analysis, uses augmented reality annotation and virtual assistance technology to mark fault points, and performs data processing and analysis using a distributed Turing learning algorithm.

[0012] S4. Based on the data processing and analysis results, automatically or manually mark the fault points found during the inspection, use the fault identification algorithm of the quantum causal reasoning model, and display personalized maintenance suggestions and steps;

[0013] S5. During the maintenance process, augmented reality devices provide real-time operation guidance, including specific maintenance steps, required tools and materials;

[0014] S6, automatically record inspection and maintenance data and upload it to the cloud in real time through IoT technology;

[0015] S7. Utilize big data analytics and machine learning algorithms to intelligently analyze uploaded maintenance data, combine environmental models and real-time sensor data to predict fault trends and provide multi-level warnings to optimize maintenance plans.

[0016] S8. Through remote communication technology, on-site inspection and maintenance personnel communicate with remote experts in real time, and experts provide professional guidance through a shared augmented reality perspective.

[0017] Optionally, the S1 includes the following steps:

[0018] S11. Wearing an augmented reality device, where the augmented reality device includes a display unit, a camera unit, and a data processing unit;

[0019] S12. Connecting the augmented reality device to the intelligent inspection system and the edge computing platform via wireless or wired connection;

[0020] S13, intelligent inspection system and edge computing platform: consists of a multimodal sensor network, a data processing unit and communication module, a data preprocessing unit, a real-time computing unit and a storage unit;

[0021] S14. Deploy a multimodal sensor network. The multimodal sensors include temperature sensors, pressure sensors, flow sensors, and gas composition sensors. The multimodal sensors are connected to the intelligent inspection system via a wireless network or a wired network.

[0022] S15, performing an initial environment scan, using the camera unit and multimodal sensor network of the augmented reality device to obtain environmental images and device status data, and performing fusion processing through the data processing unit to generate an initial environment model;

[0023] S16. Based on the initial environment model, applying a three-dimensional reconstruction algorithm to generate dynamic three-dimensional environment model data;

[0024] S17. Mark the pipeline layout, direction and connection points in the three-dimensional environment model data and synchronize them with the augmented reality device in real time.

[0025] Optionally, the S2 includes the following steps:

[0026] S21. Viewing gas pipeline status information in real time through a display unit of an augmented reality device, including pipeline layout, direction, and connection points, wherein the display unit includes a high-definition display screen and a high-resolution camera;

[0027] S22. Acquire multimodal sensor data, including temperature sensor data, pressure sensor data, flow sensor data, and gas composition sensor data, through the intelligent inspection system, and transmit the multimodal sensor data to the edge computing platform through the wireless communication module;

[0028] S23, the edge computing platform preprocesses the multimodal sensor data, including data cleaning, filtering and preliminary anomaly detection, to generate a preliminary preprocessed data set D sensor :

[0029] Data cleaning: Multimodal sensor data is D = {d1, d2, ..., d n}, by removing noise and outliers, we get the cleaned dataset D clean ;

[0030] Dclean ={d i ∣|d i -μ|≤3σ};

[0031] Among them, d i is the i-th multimodal sensor data point, μ is the mean of the multimodal sensor data, and σ is the standard deviation of the multimodal sensor data;

[0032] Filtering: Use Kalman filtering to smooth the data:

[0033]

[0034] P k|k-1 =AP k-1|k-1 A T +Q;

[0035] K k =P k|k-1 H T (HP k|k-1 H T +R) -1 ;

[0036]

[0037] P k|k =(IK k H)P k|k-1 ;

[0038] in, is the predicted state, is the updated state, P is the covariance matrix, K is the Kalman gain, A is the state transfer matrix, B is the control input matrix, u k is the control input, Q is the process noise covariance, H is the observation matrix, R is the observation noise covariance, z k is the observed value;

[0039] Preliminary anomaly detection: Detect and mark outliers through statistical analysis methods and generate preprocessed dataset D sensor ;

[0040] S24, using the camera unit of the augmented reality device to obtain the pipeline environment image data I, combined with the multimodal sensor data D sensor , image enhancement and analysis are performed using quantum image processing technology, the quantum image processing technology including:

[0041] The quantum Fourier transform converts image data into the frequency domain, enhancing image details:

[0042]

[0043] Among them, I(x,y) is the original image data, I QFT (u, v) is the frequency domain image data after quantum Fourier transform, M and N are the width and height of the image, and u and v are the coordinates in the frequency domain;

[0044] The Quantum Edge Detection algorithm identifies edges and contours in images:

[0045]

[0046] Among them, E(x,y) is the image after edge detection, and are the gradients of the image in the x and y directions respectively;

[0047] S25. Fusion preprocessing dataset D sensor and quantum image processing results I QFT , through the improved convolutional neural network to process and analyze data, generate the fault point identification model M fault :

[0048] The training process of the improved convolutional neural network involves optimizing the following objective function:

[0049]

[0050] Where N is the number of samples, Indicates the category with the highest probability of selecting e natural constant, σ activation function, max k,l Indicates taking the maximum value within the pooling window in the pooling operation, The mth preprocessed data, w mj Convolution kernel weight, b j Convolutional layer bias, w jk Fully connected layer weight, b k Fully connected layer bias, y i The true label of the i-th sample;

[0051] The specific convolutional neural network model structure is as follows:

[0052] Input layer: accepts preprocessed data D sensor and quantum image processing results I QFT ;

[0053] Convolutional layer: Apply convolution kernel W to extract features:

[0054]

[0055] in, is the output of the jth neuron in the lth layer, σ is the activation function, M j is the receptive field of the jth neuron, is the convolution kernel weight, is the bias term;

[0056] Pooling layer: downsamples the output of the convolutional layer:

[0057]

[0058] Among them, P j is the pooling window;

[0059] Fully connected layer: Connect the output of the pooling layer to the fully connected layer to identify the fault point:

[0060] z=W fc h+b fc ;

[0061] Among them, z is the output of the fully connected layer, W fc is the weight of the fully connected layer, b fc is the bias term;

[0062] S26, through the fault point identification model M fault Real-time identification and location of potential fault points P fault , including the location coordinates (x, y, z) of the fault point and the fault type:

[0063] P fault ={(x i ,y i ,z i ,T i )|i=1,2,…,m};

[0064] Among them, (x i ,y i ,z i ) is the location coordinate of the i-th fault point, T i is the fault type of the i-th fault point;

[0065] S27, the fault point information P fault It is superimposed on the three-dimensional environmental model data of the pipeline and presented to the inspection personnel in real time through the display unit of the augmented reality device, and generates real-time alarms and detailed fault information for the inspection personnel to further process.

[0066] Optionally, S3 includes the following steps:

[0067] S31. Use augmented reality devices to overlay and display pipeline status data in real time, including multimodal sensor data such as temperature, pressure, flow, and gas composition. sensor , the data is transmitted from the multimodal sensor network to the augmented reality device via the wireless communication module;

[0068] S32. Combined with historical data Dhistorical Perform multi-dimensional anomaly detection and trend analysis using historical data including past multi-modal sensor data records and known fault point information:

[0069] D historical ={(t i ,d i )|i=1,2,…,n};

[0070] Among them, t i Indicates the timestamp, d i Represents multimodal sensor data corresponding to time;

[0071] S33, real-time data D sensor and historical data D historical Perform multi-dimensional anomaly detection using a dimensionality reduction method based on manifold learning:

[0072] Z = φ(X);

[0073] Where X is the normalized data matrix, φ is the manifold learning mapping function, and Z is the data matrix after dimensionality reduction. The specific steps include:

[0074] Data standardization:

[0075]

[0076] Where D is the original data matrix, μ is the data mean, and σ is the data standard deviation;

[0077] Manifold Learning Mapping:

[0078] Use a locally linear embedding algorithm to map high-dimensional data to a low-dimensional manifold:

[0079]

[0080] Where W is the weight matrix, I is the identity matrix, represents pseudo-inverse;

[0081] S34. Use quantum Bayesian networks to perform trend analysis and predict future fault trends. Quantum Bayesian networks combine quantum computing and Bayesian reasoning to construct quantum states for fault prediction:

[0082]

[0083] Among them, |ψ> is the quantum state, α i is the amplitude of the state, |x i > is the base state, N is the number of samples in the multimodal sensor dataset;

[0084] S35. Use augmented reality annotation and virtual assistance technology to mark detected fault points, and use augmented reality equipment to display the fault point location information and detailed fault description, including fault type, severity, and recommended repair measures, in real time;

[0085] S36. Use the distributed Turing learning algorithm to process and analyze the fused data to generate a fault point identification and prediction model M. fault , the specific steps include:

[0086] S361. Define the input data matrix X and label vector Y:

[0087] X=[D sensor ,D historical ];

[0088] Y=[y1,y2,…,y N ];

[0089] Among them, y i is the fault label of the i-th sample;

[0090] S362, initializing model parameters θ, including weight matrix and bias vector;

[0091] S363, parameter update is performed using the distributed Turing learning algorithm, and each distributed node i calculates the local gradient:

[0092]

[0093] Among them, m i is the number of samples on the i-th node, is the predicted value of the jth sample, X ij is the input feature of the jth sample;

[0094] S364, global parameter update uses a distributed synchronization mechanism, each node passes the local gradient to the central node for aggregation:

[0095]

[0096] Among them, θ t is the model parameter of the tth iteration, η is the learning rate, and N is the number of distributed nodes;

[0097] S365, repeat steps S363 and S364 until convergence;

[0098] S37, according to the fault point identification and prediction model M fault , identify and locate potential fault points in real time, and superimpose the fault point information on the three-dimensional environmental model data of the pipeline through augmented reality equipment, and provide it to inspection personnel for further processing.

[0099] Optionally, the S4 includes the following steps:

[0100] S41, automatically or manually marking the fault points found during the inspection, wherein the marking includes extracting fault features from multimodal sensor data and image data to generate a preliminary fault point set P fault :

[0101] P fault ={(x i ,y i ,z i ,t i )|i=1,2,…,m};

[0102] Among them, (x i ,y i ,z i ) is the three-dimensional coordinate of the fault point, t i The timestamp of the failure point;

[0103] S42, the initial fault point set P fault Perform multimodal sensor data fusion and use extended Kalman filter for data fusion and state estimation:

[0104] The prediction steps of the extended Kalman filter include:

[0105]

[0106]

[0107] in, is the predicted state, P k|k-1 is the prediction covariance matrix, f is the nonlinear state transfer function, F k is the Jacobian matrix of the state transfer matrix, Q k is the process noise covariance matrix;

[0108] The update steps of the extended Kalman filter include:

[0109]

[0110]

[0111] P k|k =(IK k H k )P k|k-1 ;

[0112] Among them, K k is the Kalman gain, H k is the Jacobian matrix of the observation matrix, R kis the observation noise covariance matrix, h is the nonlinear observation function, z k is the observed value;

[0113] S43. Use a quantum causal reasoning model to identify faults on the fused data. The quantum causal reasoning model includes the following steps:

[0114] Construct a quantum state space where each state represents a specific configuration of multimodal sensor data and define the initial quantum state |ψ0>:

[0115]

[0116] Among them, α i is the amplitude of the state, satisfying |x i > is the base state of multimodal sensor data;

[0117] Define the quantum causality operator O causal Representing causal relationships between multimodal sensor data:

[0118]

[0119] Among them, β ij Represents data x i and x j The weight of the causal relationship between

[0120] Using quantum gate operations to construct a causal network, the specific steps include:

[0121] Use Hadamard gate H to initialize the quantum state:

[0122]

[0123] Using CNOT gate U CNOT Constructing causal relationships:

[0124] U CNOT (|x i >|x j >)=|x i >|x i ⊕x j >

[0125] Among them, ⊕ represents modulo 2 addition;

[0126] Solve the causal influence through quantum computing and apply the quantum state |ψ0> to the causal relationship operator O causal On the quantum state |ψ causal >:

[0127] |ψ causal >=O causal |>

[0128] Calculate the observation operator O for the quantum state |ψ causal > the expected value of the fault point, and obtain the causal impact measure I c :

[0129]

[0130] Among them, I c is the causal influence measure, O is the observation operator, is the complex conjugate of the amplitude of the state, <x i ||x j > represents the causal relationship ground state |x i > and |x j >causal relationship between them;

[0131] Reconstructing quantum states using quantum state tomography:

[0132]

[0133] Among them, ρ is the density matrix, which represents the statistical description of the quantum state, ρ ij is the probability amplitude of the quantum state;

[0134] S44、Calculate the quantum state |ψ causal >normalization constant:

[0135]

[0136] S45. Based on the fault identification results, personalized maintenance suggestions and steps are generated, including the required tools, materials, and specific operation instructions, and displayed to the maintenance personnel through the augmented reality device.

[0137] The beneficial effects of the present invention are:

[0138] (1) This paper proposes a gas pipeline inspection and maintenance method based on augmented reality technology and quantum image processing technology. This method uses a multimodal sensor network to perform an initial environmental scan and combines quantum image processing technology to identify and locate potential fault points in real time. Data processing and analysis using a distributed Turing learning algorithm effectively improves the accuracy of fault point identification and location, particularly in complex environments.

[0139] (2) The present invention uses augmented reality devices to provide real-time operational guidance, including specific repair steps, required tools, and materials. By combining gesture recognition and voice interaction technology, the system enables real-time interaction and feedback with the user, dynamically adjusting the guidance content. This intelligent, real-time guidance not only improves the efficiency and accuracy of repairs but also reduces errors associated with manual operation.

[0140] (3) This invention enables the automatic recording of inspection and maintenance data and its real-time upload via IoT technology. It utilizes big data analytics and machine learning algorithms to intelligently analyze the uploaded data, combining environmental models with real-time sensor data to predict fault trends and provide multi-level early warnings. These measures optimize maintenance plans and ensure the long-term safe operation of gas pipelines.

[0141] (4) This invention uses remote communication technology to enable real-time communication between on-site inspection and maintenance personnel and remote experts. Experts provide professional guidance through a shared augmented reality perspective and, combined with virtual reality technology, simulate complex maintenance scenarios to achieve collaborative work and training. This remote assistance mechanism improves the efficiency of on-site problem solving and enhances the ability to handle complex maintenance tasks. BRIEF DESCRIPTION OF THE DRAWINGS

[0142] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0143] Figure 1 This is a flow chart of a gas pipeline inspection and maintenance method based on augmented reality technology proposed by the present invention;

[0144] Figure 2 Schematic diagram of the three-dimensional environment model data and quantum image processing technology proposed in this invention;

[0145] Figure 3 Schematic diagram of data processing and analysis for the distributed Turing learning algorithm proposed in this invention. DETAILED DESCRIPTION

[0146] The present invention will now be described in further detail with reference to the accompanying drawings, which are simplified schematic diagrams that illustrate the basic structure of the present invention in a schematic manner.

[0147] refer to Figure 1-3 , a gas pipeline inspection and maintenance method based on augmented reality technology, comprising the following steps:

[0148] S1. Wear an augmented reality device, connect it to the intelligent inspection system and edge computing platform, use the multimodal sensor network to perform an initial environment and equipment status scan, and generate 3D environment model data;

[0149] S2. Use augmented reality devices to view real-time gas pipeline status information, including pipeline layout, direction, and connection points. Combined with 3D environmental model data and quantum image processing technology, potential fault points can be identified and located in real time.

[0150] S3 uses augmented reality technology to overlay and display pipeline status data in real time, combines historical data for multi-dimensional anomaly detection and trend analysis, uses augmented reality annotation and virtual assistance technology to mark fault points, and performs data processing and analysis using a distributed Turing learning algorithm.

[0151] S4. Based on the data processing and analysis results, automatically or manually mark the fault points found during the inspection, use the fault identification algorithm of the quantum causal reasoning model, and display personalized maintenance suggestions and steps;

[0152] S5. During the maintenance process, augmented reality devices provide real-time operation guidance, including specific maintenance steps, required tools and materials;

[0153] S6, automatically record inspection and maintenance data and upload it to the cloud in real time through IoT technology;

[0154] S7. Utilize big data analytics and machine learning algorithms to intelligently analyze uploaded maintenance data, combine environmental models and real-time sensor data to predict fault trends and provide multi-level warnings to optimize maintenance plans.

[0155] S8. Through remote communication technology, on-site inspection and maintenance personnel communicate with remote experts in real time, and experts provide professional guidance through a shared augmented reality perspective.

[0156] In this embodiment, S1 includes the following steps:

[0157] S11. Wearing an augmented reality device, where the augmented reality device includes a display unit, a camera unit, and a data processing unit;

[0158] S12. Connecting the augmented reality device to the intelligent inspection system and the edge computing platform via wireless or wired connection;

[0159] S13, intelligent inspection system and edge computing platform: consists of a multimodal sensor network, a data processing unit and communication module, a data preprocessing unit, a real-time computing unit and a storage unit;

[0160] S14. Deploy a multimodal sensor network. The multimodal sensors include temperature sensors, pressure sensors, flow sensors, and gas composition sensors. The multimodal sensors are connected to the intelligent inspection system via a wireless network or a wired network.

[0161] S15, performing an initial environment scan, using the camera unit and multimodal sensor network of the augmented reality device to obtain environmental images and device status data, and performing fusion processing through the data processing unit to generate an initial environment model;

[0162] S16. Based on the initial environment model, applying a three-dimensional reconstruction algorithm to generate dynamic three-dimensional environment model data;

[0163] S17. Mark the pipeline layout, direction and connection points in the three-dimensional environment model data and synchronize them with the augmented reality device in real time.

[0164] In this embodiment, S2 includes the following steps:

[0165] S21. Viewing gas pipeline status information in real time through a display unit of an augmented reality device, including pipeline layout, direction, and connection points, wherein the display unit includes a high-definition display screen and a high-resolution camera;

[0166] S22. Acquire multimodal sensor data, including temperature sensor data, pressure sensor data, flow sensor data, and gas composition sensor data, through the intelligent inspection system, and transmit the multimodal sensor data to the edge computing platform through the wireless communication module;

[0167] S23, the edge computing platform preprocesses the multimodal sensor data, including data cleaning, filtering and preliminary anomaly detection, to generate a preliminary preprocessed data set D sensor :

[0168] Data cleaning: Multimodal sensor data is D = {d1, d2, ..., d n}, by removing noise and outliers, we get the cleaned dataset D clean ;

[0169] D clean ={d i ∣|d i -μ|≤3σ};

[0170] Among them, d i is the i-th multimodal sensor data point, μ is the mean of the multimodal sensor data, and σ is the standard deviation of the multimodal sensor data;

[0171] Filtering: Use Kalman filtering to smooth the data:

[0172]

[0173] P k|k-1 =AP k-1|k-1 A T +Q;

[0174] K k =P k|k-1 H T (HP k|k-1 H T +R) -1 ;

[0175]

[0176] P k|k =(IK k H)P k|k-1 ;

[0177] in, is the predicted state, is the updated state, P is the covariance matrix, K is the Kalman gain, A is the state transfer matrix, B is the control input matrix, u k is the control input, Q is the process noise covariance, H is the observation matrix, R is the observation noise covariance, z k is the observed value;

[0178] Preliminary anomaly detection: Detect and mark outliers through statistical analysis methods and generate preprocessed dataset D sensor ;

[0179] S24, using the camera unit of the augmented reality device to obtain the pipeline environment image data I, combined with the multimodal sensor data D sensor , image enhancement and analysis are performed using quantum image processing technology, the quantum image processing technology including:

[0180] The quantum Fourier transform converts image data into the frequency domain, enhancing image details:

[0181]

[0182] Among them, I(x,y) is the original image data, I QFT (u, v) is the frequency domain image data after quantum Fourier transform, M and N are the width and height of the image, and u and v are the coordinates in the frequency domain;

[0183] The Quantum Edge Detection algorithm identifies edges and contours in images:

[0184]

[0185] Among them, E(x,y) is the image after edge detection, and are the gradients of the image in the x and y directions respectively;

[0186] S25. Fusion preprocessing dataset D sensor and quantum image processing results I QFT , through the improved convolutional neural network to process and analyze data, generate the fault point identification model M fault :

[0187] The training process of the improved convolutional neural network involves optimizing the following objective function:

[0188]

[0189] Where N is the number of samples, Indicates the category with the highest probability of selecting e natural constant, σ activation function, max k,l Indicates taking the maximum value within the pooling window in the pooling operation, The mth preprocessed data, w mj Convolution kernel weight, b j Convolutional layer bias, w jk Fully connected layer weight, b k Fully connected layer bias, y i The true label of the i-th sample;

[0190] The specific convolutional neural network model structure is as follows:

[0191] Input layer: accepts preprocessed data D sensor and quantum image processing results I QFT ;

[0192] Convolutional layer: Apply convolution kernel W to extract features:

[0193]

[0194] in, is the output of the jth neuron in the lth layer, σ is the activation function, M j is the receptive field of the jth neuron, is the convolution kernel weight, is the bias term;

[0195] Pooling layer: downsamples the output of the convolutional layer:

[0196]

[0197] Among them, P j is the pooling window;

[0198] Fully connected layer: Connect the output of the pooling layer to the fully connected layer to identify the fault point:

[0199] z=W fc h+b fc ;

[0200] Among them, z is the output of the fully connected layer, W fc is the weight of the fully connected layer, b fc is the bias term;

[0201] S26, through the fault point identification model M fault Real-time identification and location of potential fault points P fault, including the location coordinates (x, y, z) of the fault point and the fault type:

[0202] P fault ={(x i ,y i ,z i ,T i )|i=1,2,…,m};

[0203] Among them, (x i ,y i ,z i ) is the location coordinate of the i-th fault point, T i is the fault type of the i-th fault point;

[0204] S27, the fault point information P fault It is superimposed on the three-dimensional environmental model data of the pipeline and presented to the inspection personnel in real time through the display unit of the augmented reality device, and generates real-time alarms and detailed fault information for the inspection personnel to further process.

[0205] In this embodiment, S3 includes the following steps:

[0206] S31. Use augmented reality devices to overlay and display pipeline status data in real time, including multimodal sensor data such as temperature, pressure, flow, and gas composition. sensor , the data is transmitted from the multimodal sensor network to the augmented reality device via the wireless communication module;

[0207] S32. Combined with historical data D historical Perform multi-dimensional anomaly detection and trend analysis using historical data including past multi-modal sensor data records and known fault point information:

[0208] D historical ={(t i ,d i )|i=1,2,…,n};

[0209] Among them, t i Indicates the timestamp, d i Represents multimodal sensor data corresponding to time;

[0210] S33, real-time data D sensor and historical data D historical Perform multi-dimensional anomaly detection using a dimensionality reduction method based on manifold learning:

[0211] Z = φ(X);

[0212] Where X is the normalized data matrix, φ is the manifold learning mapping function, and Z is the data matrix after dimensionality reduction. The specific steps include:

[0213] Data standardization:

[0214]

[0215] Where D is the original data matrix, μ is the data mean, and σ is the data standard deviation;

[0216] Manifold Learning Mapping:

[0217] Use a locally linear embedding algorithm to map high-dimensional data to a low-dimensional manifold:

[0218]

[0219] Where W is the weight matrix, I is the identity matrix, represents pseudo-inverse;

[0220] S34. Use quantum Bayesian networks to perform trend analysis and predict future fault trends. Quantum Bayesian networks combine quantum computing and Bayesian reasoning to construct quantum states for fault prediction:

[0221]

[0222] Among them, |ψ> is the quantum state, α i is the amplitude of the state, |x i > is the base state, N is the number of samples in the multimodal sensor dataset;

[0223] S35. Use augmented reality annotation and virtual assistance technology to mark detected fault points, and use augmented reality equipment to display the fault point location information and detailed fault description, including fault type, severity, and recommended repair measures, in real time;

[0224] S36. Use the distributed Turing learning algorithm to process and analyze the fused data to generate a fault point identification and prediction model M. fault , the specific steps include:

[0225] S361. Define the input data matrix X and label vector Y:

[0226] X=[D sensor ,D historical ];

[0227] Y=[y1,y2,…,y N ];

[0228] Among them, y i is the fault label of the i-th sample;

[0229] S362, initializing model parameters θ, including weight matrix and bias vector;

[0230] S363, parameter update is performed using the distributed Turing learning algorithm, and each distributed node i calculates the local gradient:

[0231]

[0232] Among them, m i is the number of samples on the i-th node, is the predicted value of the jth sample, X ij is the input feature of the jth sample;

[0233] S364, global parameter update uses a distributed synchronization mechanism, each node passes the local gradient to the central node for aggregation:

[0234]

[0235] Among them, θ t is the model parameter of the tth iteration, η is the learning rate, and N is the number of distributed nodes;

[0236] S365, repeat steps S363 and S364 until convergence;

[0237] S37, according to the fault point identification and prediction model M fault , identify and locate potential fault points in real time, and superimpose the fault point information on the three-dimensional environmental model data of the pipeline through augmented reality equipment, and provide it to inspection personnel for further processing.

[0238] In this embodiment, S4 includes the following steps:

[0239] S41, automatically or manually marking the fault points found during the inspection, wherein the marking includes extracting fault features from multimodal sensor data and image data to generate a preliminary fault point set P fault :

[0240] P fault ={(x i ,y i ,z i ,t i )|i=1,2,…,m};

[0241] Among them, (x i ,y i ,z i ) is the three-dimensional coordinate of the fault point, t i The timestamp of the failure point;

[0242] S42, the initial fault point set P fault Perform multimodal sensor data fusion and use extended Kalman filter for data fusion and state estimation:

[0243] The prediction steps of the extended Kalman filter include:

[0244]

[0245]

[0246] in, is the predicted state, P k|k-1 is the prediction covariance matrix, f is the nonlinear state transfer function, F k is the Jacobian matrix of the state transfer matrix, Q k is the process noise covariance matrix;

[0247] The update steps of the extended Kalman filter include:

[0248]

[0249]

[0250] P k|k =(IK k H k )P k|k-1 ;

[0251] Among them, K k is the Kalman gain, H k is the Jacobian matrix of the observation matrix, R k is the observation noise covariance matrix, h is the nonlinear observation function, z k is the observed value;

[0252] S43. Use a quantum causal reasoning model to identify faults on the fused data. The quantum causal reasoning model includes the following steps:

[0253] Construct a quantum state space where each state represents a specific configuration of multimodal sensor data and define the initial quantum state |ψ0>:

[0254]

[0255] Among them, α i is the amplitude of the state, satisfying |x i > is the base state of multimodal sensor data;

[0256] Define the quantum causality operator O causal Representing causal relationships between multimodal sensor data:

[0257]

[0258] Among them, β ij Represents data xi and x j The weight of the causal relationship between

[0259] Using quantum gate operations to construct a causal network, the specific steps include:

[0260] Use Hadamard gate H to initialize the quantum state:

[0261]

[0262] Using CNOT gate U CNOT Constructing causal relationships:

[0263] U CNOT (|x i >|x j >)=|x i >|x i ⊕x j >

[0264] Among them, ⊕ represents modulo 2 addition;

[0265] Solve the causal influence through quantum computing and apply the quantum state |ψ0> to the causal relationship operator O causal On the quantum state |ψ causal >:

[0266] |ψ causal >=O causal |>

[0267] Calculate the observation operator O for the quantum state |ψ causal > the expected value of the fault point, and obtain the causal impact measure I c :

[0268]

[0269] Among them, I c is the causal influence measure, O is the observation operator, is the complex conjugate of the amplitude of the state, <x i ||x j > represents the causal relationship ground state |x i > and |x j >causal relationship between them;

[0270] Reconstructing quantum states using quantum state tomography:

[0271]

[0272] Among them, ρ is the density matrix, which represents the statistical description of the quantum state, ρ ij is the probability amplitude of the quantum state;

[0273] S44、Calculate the quantum state |ψ causal >normalization constant:

[0274]

[0275] S45. Based on the fault identification results, personalized maintenance suggestions and steps are generated, including the required tools, materials, and specific operation instructions, and displayed to the maintenance personnel through the augmented reality device.

[0276] Example 1:

[0277] The safe operation of gas pipelines is crucial for a large city gas company. Due to the complex urban pipeline network, inspection and maintenance tasks pose significant challenges. Traditional manual inspection and maintenance methods suffer from low efficiency, poor accuracy, and numerous safety hazards. To address these issues, the company decided to adopt a gas pipeline inspection and maintenance method based on augmented reality technology.

[0278] In practice, the company equips each inspector with an augmented reality (AR) device, which includes a display unit, camera unit, and data processing unit. The inspector first wears the AR device and connects it to the intelligent inspection system and edge computing platform. A multimodal sensor network performs an initial scan of the environment and equipment status, generating a three-dimensional environmental model. The AR device provides real-time information on the gas pipeline status, including its layout, route, and connection points.

[0279] During inspections, AR devices combine 3D environmental model data with quantum image processing technology to identify and locate potential fault points in real time. Quantum image processing technology, using methods such as quantum Fourier transform and quantum edge detection, enhances image detail and accurately identifies tiny cracks and corrosion points in the pipeline. Once a fault point is identified, the system overlays and displays pipeline status data in real time, combining it with historical data for multi-dimensional anomaly detection and trend analysis. Fault points are marked using augmented reality annotation and virtual assistance technologies, and data processing and analysis are performed using a distributed Turing learning algorithm.

[0280] When a fault point is discovered, the system automatically or manually marks it during the inspection, uses a fault identification algorithm based on a quantum causal inference model, and displays personalized repair suggestions and steps. During the repair process, AR devices provide real-time operational guidance, including specific repair steps, required tools, and materials. Inspectors can complete repair operations quickly and safely based on the AR device's guidance.

[0281] All inspection and maintenance data is automatically recorded and uploaded to the cloud in real time via IoT technology. Leveraging big data analytics and machine learning algorithms, this uploaded data is intelligently analyzed, combining environmental models with real-time sensor data to predict fault trends and provide multi-level warnings, optimizing maintenance plans. Furthermore, inspectors and maintenance personnel can communicate in real time with remote experts via remote communication technology. Remote experts, through a shared AR perspective, provide professional guidance and resolve complex issues.

[0282] Concrete data proves our beneficial effects

[0283] In one application, the company conducted a month-long inspection and maintenance test on a section of gas pipeline in the city center. The test results are as follows:

[0284] Comparison of meter inspection and maintenance efficiency

[0285]

[0286] Comparison of table fault identification and repair accuracy

[0287]

[0288] Table data processing and transmission comparison

[0289]

[0290] These data fully demonstrate the superiority of this invention in practical applications. By using augmented reality and quantum image processing technologies, combined with distributed Turing learning algorithms and Internet of Things technologies, it not only significantly improves the efficiency and accuracy of inspections and maintenance, but also significantly enhances the safety and reliability of gas pipelines.

[0291] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solution and inventive concept of the present invention, should be covered by the scope of protection of the present invention.

Claims

1. A gas pipeline inspection and maintenance method based on augmented reality technology, comprising the following steps: S1. Wear an augmented reality device, connect it to the intelligent inspection system and edge computing platform, use the multimodal sensor network to perform an initial environment and equipment status scan, and generate 3D environment model data; S2. Use augmented reality devices to view gas pipeline status information in real time, obtain multimodal sensor data, and perform preprocessing. Simultaneously, obtain environmental image data of the gas pipeline. Combined with 3D environmental model data and quantum image processing technology, an improved convolutional neural network is used to generate a fault point identification model to identify and locate potential fault points in real time. S3. Use augmented reality technology to overlay and display multimodal sensor data in real time, combine it with historical multimodal sensor data for multi-dimensional anomaly detection and trend analysis, use augmented reality annotation and virtual assistance technology to mark fault points, and generate fault point identification and prediction models through distributed Turing learning algorithms to identify and predict potential fault points in real time; S4. Automatically or manually label the identified fault points, generate a preliminary set of fault points, collect multimodal sensor data from the preliminary fault points, perform data fusion, use the quantum causal inference model to perform final fault identification on the fused data, and generate and display personalized maintenance suggestions and steps. S5. During the maintenance process, augmented reality devices provide real-time operation guidance, including specific maintenance steps, required tools and materials; S6, automatically record inspection and maintenance data and upload it to the cloud in real time through IoT technology; S7. Utilize big data analytics and machine learning algorithms to intelligently analyze uploaded maintenance data, combine environmental models and real-time sensor data to predict fault trends and provide multi-level warnings to optimize maintenance plans. S8. Through remote communication technology, on-site inspection and maintenance personnel can communicate with remote experts in real time, and experts can provide professional guidance through a shared augmented reality perspective; Said S1 comprises the following steps: S11. Wearing an augmented reality device, where the augmented reality device includes a display unit, a camera unit, and a data processing unit; S12. Connecting the augmented reality device to the intelligent inspection system and the edge computing platform via wireless or wired connection; S13, intelligent inspection system and edge computing platform: consists of a multimodal sensor network, a data processing unit and communication module, a data preprocessing unit, a real-time computing unit and a storage unit; S14. Deploy a multimodal sensor network. The multimodal sensors include temperature sensors, pressure sensors, flow sensors, and gas composition sensors. The multimodal sensors are connected to the intelligent inspection system via a wireless network or a wired network. S15, performing an initial environment scan, using the camera unit and multimodal sensor network of the augmented reality device to obtain environmental images and device status data, and performing fusion processing through the data processing unit to generate an initial environment model; S16. Based on the initial environment model, applying a three-dimensional reconstruction algorithm to generate dynamic three-dimensional environment model data; S17. Mark the pipeline layout, direction, and connection points in the 3D environment model data, and synchronize them with the augmented reality device in real time; The S2 comprises the following steps: S21. Viewing gas pipeline status information in real time through a display unit of an augmented reality device, including pipeline layout, direction, and connection points, wherein the display unit includes a high-definition display screen and a high-resolution camera; S22. Acquire multimodal sensor data, including temperature sensor data, pressure sensor data, flow sensor data, and gas composition sensor data, through the intelligent inspection system, and transmit the multimodal sensor data to the edge computing platform through the wireless communication module; S23, the edge computing platform preprocesses the multimodal sensor data, including data cleaning, filtering and preliminary anomaly detection, to generate a preliminary preprocessed data set D sensor : Data cleaning: Multimodal sensor data is D = {d1, d2, ..., d n }, by removing noise and outliers, we get the cleaned dataset D clean ; D clean ={d i ∣|d i -μ|≤3σ}; Among them, d i is the i-th multimodal sensor data point, μ is the mean of the multimodal sensor data, and σ is the standard deviation of the multimodal sensor data; Filtering: Use Kalman filtering to smooth the data: P k|k-1 =AP k-1|k-1 From T +Q; K k =P k|k-1 H T (HP k|k-1 H T +R) -1 ; P k|k =(I-K k H)P k|k-1 ; in, is the predicted state, is the updated state, P is the covariance matrix, K is the Kalman gain, A is the state transfer matrix, B is the control input matrix, u k is the control input, Q is the process noise covariance, H is the observation matrix, R is the observation noise covariance, z k is the observed value; Preliminary anomaly detection: Detect and mark outliers through statistical analysis methods and generate preprocessed dataset D sensor ; S24. Using the camera unit of the augmented reality device to obtain environmental image data I of the pipeline, and performing image enhancement and analysis using quantum image processing technology, the quantum image processing technology including: The quantum Fourier transform converts image data into the frequency domain, enhancing image details: Among them, I(x,y) is the original image data, I QFT (u, v) is the frequency domain image data after quantum Fourier transform, M and N are the width and height of the image, and u and v are the coordinates in the frequency domain; The Quantum Edge Detection algorithm identifies edges and contours in images: Among them, E(x,y) is the image after edge detection, and are the gradients of the image in the x and y directions respectively; S25. Fusion preprocessing dataset D sensor and quantum image processing results I QFT , through the improved convolutional neural network to process and analyze data, generate the fault point identification model M fault : The training process of the improved convolutional neural network involves optimizing the following objective function: Where N is the number of samples, Indicates the category with the highest probability of selecting e natural constant, σ activation function, max k,l Indicates taking the maximum value within the pooling window in the pooling operation, D sensor,m The mth preprocessed data, w mj Convolution kernel weight, b j Convolutional layer bias, w jk Fully connected layer weight, b k Fully connected layer bias, y i The true label of the i-th sample; The specific convolutional neural network model structure is as follows: Input layer: accepts preprocessed data D sensor and quantum image processing results I QFT ; Convolutional layer: Apply convolution kernel W to extract features: in, is the output of the jth neuron in the lth layer, σ is the activation function, M j is the receptive field of the jth neuron, is the convolution kernel weight, is the bias term; Pooling layer: downsamples the output of the convolutional layer: Among them, P j is the pooling window; Fully connected layer: Connect the output of the pooling layer to the fully connected layer to identify the fault point: z=W fc h+b fc ; Among them, z is the output of the fully connected layer, W fc is the weight of the fully connected layer, b fc is the bias term; S26, through the fault point identification model M fault Real-time identification and location of potential fault points P fault , including the location coordinates (x, y, z) of the fault point and the fault type: P fault ={(x i ,y i ,z i ,T i )∣i=1,2,…,m}; Among them, (x i ,y i ,z i ) is the location coordinate of the i-th fault point, T i is the fault type of the i-th fault point; S27, the fault point information P fault It is superimposed on the three-dimensional environmental model data of the pipeline and presented to the inspection personnel in real time through the display unit of the augmented reality device, and generates real-time alarms and detailed fault information for the inspection personnel to further process.

2. The gas pipeline inspection and maintenance method based on augmented reality technology according to claim 1 is characterized in that: The S3 includes the following steps: S31. Using an augmented reality device to overlay and display multimodal sensor data in real time, the multimodal sensor data including temperature, pressure, flow rate, and gas composition, transmitted from the multimodal sensor network to the augmented reality device via a wireless communication module; S32. Combined with historical data D historical Perform multi-dimensional anomaly detection and trend analysis using historical data including past multi-modal sensor data records and known fault point information: D historical ={(t i ,d i )∣i=1,2,…,n}; Among them, t i Indicates the timestamp, d i Represents multimodal sensor data corresponding to time; S33, real-time data D sensor and historical data D historical Perform multi-dimensional anomaly detection using a dimensionality reduction method based on manifold learning: Z = φ(X); Where X is the normalized data matrix, φ is the manifold learning mapping function, and Z is the data matrix after dimensionality reduction. The specific steps include: Data standardization: Where D is the original data matrix, μ is the data mean, and σ is the data standard deviation; Manifold Learning Mapping: Use a locally linear embedding algorithm to map high-dimensional data to a low-dimensional manifold: Where W is the weight matrix, I is the identity matrix, represents pseudo-inverse; S34. Use quantum Bayesian networks to perform trend analysis and predict future fault trends. Quantum Bayesian networks combine quantum computing and Bayesian reasoning to construct quantum states for fault prediction: Among them, |ψ> is the quantum state, α i is the amplitude of the state, |x i > is the base state, N is the number of samples in the multimodal sensor dataset; S35. Use augmented reality annotation and virtual assistance technology to mark detected fault points, and use augmented reality equipment to display the fault point location information and detailed fault description, including fault type, severity, and recommended repair measures, in real time; S36, using the distributed Turing learning algorithm to process and analyze the fused data, and generate the fault point identification and prediction model M' fault , the specific steps include: S361. Define the input data matrix X and label vector Y: X=[D sensor ,D historical ]; Y=[y1,y2,…,y N ]; Among them, y i is the fault label of the i-th sample; S362, initializing model parameters θ, including weight matrix and bias vector; S363, parameter update is performed using the distributed Turing learning algorithm, and each distributed node i calculates the local gradient: Among them, m i is the number of samples on the i-th node, is the predicted value of the jth sample, X ij is the input feature of the jth sample; S364, global parameter update uses a distributed synchronization mechanism, each node passes the local gradient to the central node for aggregation: Among them, θ t is the model parameter of the tth iteration, η is the learning rate, and N is the number of distributed nodes; S365, repeat steps S363 and S364 until convergence; S37, according to the fault point identification and prediction model M' fault , identify and locate potential fault points in real time, and superimpose the fault point information on the three-dimensional environmental model data of the pipeline through augmented reality equipment, and provide it to inspection personnel for further processing.

3. The gas pipeline inspection and maintenance method based on augmented reality technology according to claim 1 is characterized in that: The S4 comprises the following steps: S41, automatically or manually marking the fault points found during the inspection, wherein the marking includes extracting fault features from multimodal sensor data and image data to generate a preliminary fault point set P fault : P fault ={(x i ,y i ,z i ,t i )∣i=1,2,…,m}; Among them, (x i ,y i ,z i ) is the three-dimensional coordinate of the fault point, t i The timestamp of the failure point; S42, the initial fault point set P fault Perform multimodal sensor data fusion and use extended Kalman filter for data fusion and state estimation: The prediction steps of the extended Kalman filter include: in, is the predicted state, P k|k-1 is the prediction covariance matrix, f is the nonlinear state transfer function, F k is the Jacobian matrix of the state transfer matrix, Q k is the process noise covariance matrix; The update steps of the extended Kalman filter include: P k|k =(I-K k H k )P k|k-1 ; Among them, K k is the Kalman gain, H k is the Jacobian matrix of the observation matrix, R k is the observation noise covariance matrix, h is the nonlinear observation function, z k is the observed value; S43. Use a quantum causal reasoning model to identify faults on the fused data. The quantum causal reasoning model includes the following steps: Construct a quantum state space where each state represents a specific configuration of multimodal sensor data and define the initial quantum state |ψ0>: Among them, α i is the amplitude of the state, satisfying |x i > is the base state of multimodal sensor data; Define the quantum causality operator O causal Representing causal relationships between multimodal sensor data: Among them, β ij Represents data x i and x j The weight of the causal relationship between Using quantum gate operations to construct a causal network, the specific steps include: Use Hadamard gate H to initialize the quantum state: Using CNOT gate U CNOT Constructing causal relationships: in, represents modulo 2 addition; Solve the causal influence through quantum computing and apply the quantum state |ψ0> to the causal relationship operator O causal On the quantum state |ψ causal >: |ψ causal >=O causal |>; Calculate the observation operator O for the quantum state |ψ causal > the expected value of the fault point, and obtain the causal impact measure I c : Among them, I c is the causal influence measure, O is the observation operator, is the complex conjugate of the amplitude of the state, <x i ||x j > represents the causal relationship ground state |x i > and |x j >causal relationship between them; Reconstructing quantum states using quantum state tomography: Among them, ρ is the density matrix, which represents the statistical description of the quantum state, ρ ij is the probability amplitude of the quantum state; S44、Calculate the quantum state |ψ causal >normalization constant: S45. Based on the fault identification results, personalized maintenance suggestions and steps are generated, including the required tools, materials, and specific operation instructions, and displayed to the maintenance personnel through the augmented reality device.

Images