Method and system for predicting life loss of petroleum transfer hose

Abstract

The present invention relates to the technical field of industrial data prediction and simulation modeling, and particularly to a method and system for predicting the life loss of transfer oil hoses. The method includes: generating loss distribution data of hose materials based on multimodal data fusion and multi-field collaborative simulation technology; realizing resource allocation between regions and optimizing model parameters through a quantum game optimization algorithm and a genetic algorithm, constructing a high-precision regional loss prediction model, and dynamically correcting the prediction model parameters through the difference analysis module of the digital twin system, and iteratively adjusting the input parameters of the simulation environment in combination with a closed-loop feedback optimization process to generate high-precision life prediction data reflecting the real-time operating state; the present invention effectively improves the accuracy and adaptability of life prediction, overcomes the problem of insufficient response ability of the prior art to complex working conditions, and provides reliable technical support for the operation and maintenance and full-life cycle management of hoses.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial data prediction and simulation modeling, and particularly to a method and system for predicting the life loss of transfer petroleum hoses. Background Art

[0002] Transfer petroleum hoses are widely used in the petrochemical industry, and their performance and service life are directly related to the operation safety of equipment and production efficiency. Accurately predicting the life loss of transfer petroleum hoses can not only effectively prevent accidents, ensure the safety of equipment and personnel, but also optimize the maintenance and replacement cycle, avoiding high costs caused by premature replacement or sudden failures. In addition, accurate life prediction can also provide data support for material selection and product design, thereby enhancing the overall durability and reliability of the hose.

[0003] The prior art (Chinese invention patent, publication number: CN118966027B, title: A method for predicting the life loss of composite hoses) adopts a method for predicting the life loss of composite hoses based on the BP neural network, and optimizes the BP neural network by improving the hiking optimization algorithm (HOA). However, this technology has the following main defects:

[0004] The prior art lacks targeted modeling for the complex operating environment of hoses (such as losses under multi-field interactions), fails to fully consider the comprehensive influence of the force field, thermal field and chemical field, and the accuracy of the prediction results for extreme working conditions is relatively low; its prediction model is based on predefined data sets and parameter optimization, and fails to dynamically adjust in combination with the real-time operating state, resulting in weak response ability of the model to environmental changes; although an improved hiking optimization algorithm is adopted, the global adaptation ability of the optimization algorithm to complex loss behaviors is insufficient, especially in scenarios of multi-variable interaction and non-linear loss prediction, local optimal solutions may occur; the prior art does not establish a regional loss analysis framework, making it difficult to accurately capture the local loss distribution and its impact on the global life, resulting in limited prediction accuracy. Summary of the Invention

[0005] Aiming at the many problems existing in the above prior art, the present invention provides a method and system for predicting the life loss of transfer petroleum hoses. The present invention constructs a multi-field collaborative simulation environment, integrates the coupling effects of the force field, thermal field and chemical field to generate loss distribution data; combines quantum game optimization and genetic algorithm to optimize the regional loss prediction model and achieve global consistency; through the difference analysis module of the digital twin system, dynamically adjusts the simulation environment parameters and updates the prediction model to achieve real-time optimization and high-precision life prediction. The present invention improves the accuracy, adaptability and real-time optimization ability of the prediction, providing a reliable basis for the operation and maintenance and design optimization of the hose.

[0006] A method for predicting the life loss of transfer petroleum hoses includes the following steps:

[0007] Collect physical, chemical, and environmental characteristic data during the operation of the oil transfer hose, and process the data to generate fused feature data;

[0008] Based on the fused feature data, construct a multi-field collaborative simulation environment for the oil transfer hose, perform coupled modeling of the force field, thermal field, and chemical field, use the finite element analysis method and multi-feature tensor decomposition algorithm to simulate the multi-field interaction, and generate loss distribution data representing the local loss distribution of the oil transfer hose material;

[0009] Divide the loss distribution data into multiple local regions of the oil transfer hose, and construct a regional loss prediction model. Achieve collaboration and resource allocation between regions through quantum optimization and genetic algorithms, and generate global life prediction data including the life dissipation trend and remaining life prediction;

[0010] Based on the global life prediction data and the real-time collected operation status data, construct a digital twin system, correct the regional loss prediction model and form a closed-loop feedback optimization process, and output optimized data that dynamically reflects the operation status and remaining life prediction of the oil transfer hose.

[0011] Preferably, the processing of the physical, chemical, and environmental characteristic data during the operation of the oil transfer hose includes the following steps: Identify and remove outliers in the operation data of the oil transfer hose through a noise filtering algorithm; Use the dynamic time warping method to perform time offset correction based on the time series of the operation data of the oil transfer hose; Use a feature extraction algorithm to extract the stress vibration frequency, corrosion diffusion rate, and temperature change rate in the operation data of the oil transfer hose respectively, and generate the cleaned dynamic feature data.

[0012] Preferably, the cleaned dynamic feature data is compressed by a feature compression model, and the feature compression model includes a variational autoencoder. The variational autoencoder performs latent space mapping on the cleaned dynamic feature data to remove low-contribution features and generate fused feature data with key dynamic features retained.

[0013] Preferably, the construction of the multi-field collaborative simulation environment includes the following steps:

[0014] Establish a force field model of the oil transfer hose based on a mechanical modeling method to simulate the local stress distribution of the oil transfer hose under different pressure gradients and shear stresses;

[0015] Establish a thermal field model of the oil transfer hose based on heat conduction analysis to calculate the influence of the temperature gradient on the thermal aging rate of the oil transfer hose material;

[0016] Establish a chemical field model of the transfer petroleum hose through diffusion kinetics modeling to predict the diffusion path of corrosive substances in the transfer petroleum hose material;

[0017] Perform multi-field coupling modeling on the above single-field model through numerical calculation methods, where the multi-field coupling is calculated by the following formula:

[0018]

[0019] Wherein, represents the total loss distribution of the transfer petroleum hose material; represents the local loss under the action of the force field; represents the local loss under the action of the thermal field; represents the local loss under the action of the chemical field; are the spatial coordinates respectively.

[0020] Preferably, the multi-field coupling is realized by the multi-feature tensor decomposition algorithm, which specifically includes the following steps:

[0021] Construct a multi-modal tensor of the force field data, thermal field data and chemical field data of the transfer petroleum hose ;

[0022] Use the following formula to decompose the tensor :

[0023]

[0024] Wherein, represents the data point value in the multi-modal tensor of the transfer petroleum hose ; , , respectively represent the weights of the th decomposition component in the dimension; represents the rank of the tensor decomposition;

[0025] Generate coupled loss distribution data based on the decomposed tensor data for calculating the total local loss of the transfer petroleum hose.

[0026] Preferably, the establishment of the regional loss prediction model includes the following steps: divide the loss distribution data into multiple regional data, and each regional data represents the loss information of the corresponding sub-region of the transfer petroleum hose; use the quantum game optimization algorithm to simulate the resource competition and cooperation between regions; combine the genetic algorithm to optimize the parameters of the prediction model for each region to make the prediction results between regions reach global consistency.

[0027] Preferably, the quantum game optimization algorithm is implemented through the following steps: optimizing the resource allocation between regions through quantum state superposition, and simulating the iterative collapse process of the quantum state to determine the optimal allocation; the genetic algorithm optimizes the key parameters of the regional loss prediction model through crossover, mutation, and selection operations, and evaluates the model performance using the following fitness function:

[0028]

[0029] wherein, represents the fitness of the model; represents the actual value of the th sample; represents the predicted value of the th sample; represents the total number of samples.

[0030] Preferably, the digital twin system includes a difference analysis module, which generates a difference analysis result by comparing the operation state data of the transfer petroleum hose collected in real time with the global life prediction data, so as to correct the parameters of the regional loss prediction model and update the state of the digital twin system.

[0031] Preferably, the closed-loop feedback optimization process includes the following steps: dynamically adjusting the input parameters of the multi-field collaborative simulation environment according to the difference analysis result, and recalculating the loss distribution data of the transfer petroleum hose; inputting the updated loss distribution data into the digital twin system to update the real-time operation state and remaining life prediction data of the transfer petroleum hose until the prediction error converges within the set range.

[0032] A system for implementing the life loss prediction method of the transfer petroleum hose includes:

[0033] A multi-modal data acquisition module, which is used to collect physical, chemical, and environmental characteristic data during the operation of the transfer petroleum hose, and process the collected data to generate fused feature data;

[0034] A multi-field collaborative simulation module, which is used to construct a multi-field collaborative simulation environment of the transfer petroleum hose based on the fused feature data, couple and model the force field, thermal field, and chemical field, and simulate the multi-field interaction using the finite element analysis method and the multi-feature tensor decomposition algorithm to generate loss distribution data representing the local loss distribution of the transfer petroleum hose material;

[0035] A regional loss prediction module, which is used to divide the loss distribution data into multiple local regions of the transfer petroleum hose, construct a regional loss prediction model, and use quantum optimization and genetic algorithms to achieve cooperation and resource allocation between regions, and generate global life prediction data including the life dissipation trend and remaining life prediction;

[0036] A digital twin system module is used to correct the regional loss prediction model and form a closed-loop feedback optimization process based on global life prediction data and real-time collected operation status data, and output optimized data that dynamically reflects the operation status and remaining life prediction of the transfer petroleum hose.

[0037] Compared with the prior art, the advantages and beneficial effects of the present invention are as follows:

[0038] By constructing a multi-field collaborative simulation environment, the present invention realizes the accurate modeling and prediction of the coupling effects of the force field, thermal field, and chemical field, overcoming the problem of insufficient adaptability of the prior art to complex working conditions;

[0039] Through the difference analysis and closed-loop feedback optimization process, the present invention realizes the dynamic adjustment and iterative optimization based on the real-time operation status, solving the problem of lack of real-time performance in the prior art;

[0040] By combining quantum game optimization and genetic algorithm, the present invention realizes the global consistency of resource allocation and parameter optimization between regions, improving the global optimization ability of complex loss behaviors;

[0041] By dividing the loss distribution data into regions and constructing a regional prediction model, the present invention captures the local loss distribution characteristics and realizes the collaborative optimization between regions;

[0042] The combination of multi-modal data fusion and digital twin technology in the present invention enables the solution to generate high-precision life prediction data, while improving the adaptability and prediction stability of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 It is a schematic flow chart of the method of the present invention;

[0044] Figure 2 It is a schematic diagram of the construction of the multi-field collaborative simulation environment in the present invention;

[0045] Figure 3 It is a schematic diagram of the establishment of the regional loss prediction model in the present invention;

[0046] Figure 4 It is a structural block diagram of the system of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0047] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are merely exemplary and are not intended to limit the scope of the present disclosure. In the following detailed description, for the sake of explanation, many specific details are set forth in order to provide a comprehensive understanding of the embodiments of the present disclosure.

[0048] As Figure 1As shown, a method for predicting the life loss of an oil transfer hose includes the following steps:

[0049] Collect the physical, chemical, and environmental characteristic data during the operation of the oil transfer hose, and process the data to generate fused feature data;

[0050] Preferably, processing the physical, chemical, and environmental characteristic data during the operation of the collected oil transfer hose includes the following steps: identifying and removing outliers in the operation data of the oil transfer hose through a noise filtering algorithm; performing time offset correction based on the time series of the operation data of the oil transfer hose using the dynamic time warping method; and using a feature extraction algorithm to extract the stress vibration frequency, corrosion diffusion rate, and temperature change rate in the operation data of the oil transfer hose respectively to generate the cleaned dynamic feature data.

[0051] During the operation of the oil transfer hose, data collection involves multi-modal data, including physical characteristic data (such as stress, vibration, etc.), chemical characteristic data (such as the diffusion rate of corrosive substances), and environmental characteristic data (such as temperature, humidity, etc.). These data usually have noise, incompleteness, and time series offset in actual collection. Therefore, a series of data processing steps are required to ensure the accuracy and consistency of the data, so as to generate high-quality feature data for subsequent analysis.

[0052] In principle, the noise filtering algorithm aims to identify and remove outliers in the collected data through statistical methods, signal processing techniques, or model-based methods. Specifically in the present invention, outliers usually originate from sensor device failures, external interference, or environmental changes. For this reason, a noise filtering method based on statistical thresholds and time series analysis is adopted in the solution to detect and remove the outliers in the collected data. For example, by analyzing the time series fluctuation range of the stress signal and combining with the sliding window detection method, the collected points exceeding the normal stress change threshold are determined as noise and removed. For the corrosion diffusion rate data, based on the chemical reaction law model, data that deviates significantly from the actual diffusion path can be removed through smoothing filtering and comparison with historical data.

[0053] The core effect of noise filtering is to significantly improve the accuracy of the data, reduce the interference of outliers on subsequent feature extraction and model training, and ensure that the cleaned data can truly reflect the operation state of the oil transfer hose.

[0054] Example: In a simulation experiment, a set of stress signal data of a hose under pressure was collected, and about 10% of the data points were significantly abnormal due to equipment fluctuations. Through the noise filtering algorithm based on the sliding window, all outliers were successfully removed, and linear interpolation was performed on the missing data points, making the deviation between the processed stress signal and the theoretical calculation result less than 1%.

[0055] Dynamic Time Warping (DTW) is an algorithm commonly used in time series analysis. Its principle is to align two time series through non-linear stretching or compression to eliminate time offsets caused by sampling frequency or sensor delay. In the present invention, due to the complex operating environment of the hose, different characteristic data (such as stress signals and temperature data) may have different sampling frequencies and time delays. Therefore, DTW is needed to perform time alignment on multi-modal data.

[0056] In specific implementation, by constructing a time-amplitude matrix of the stress signal and the temperature change rate signal, calculating the distance cumulative matrix between time series, and finding the optimal alignment path in the matrix. In this way, the multi-modal data is mapped to a unified time axis to eliminate the offset problem between time series. Dynamic Time Warping significantly improves the time synchronization of multi-modal data, enabling physical, chemical, and environmental characteristic data to be comprehensively analyzed at the same time point, thereby enhancing the reliability of subsequent feature extraction and fusion.

[0057] Example: In the experiment, stress signals and temperature signals were collected for 10 seconds, and it was found that there was a sampling delay of about 0.5 seconds between them. Through the Dynamic Time Warping method, the two signals were successfully aligned, reducing the maximum time deviation to 0.01 seconds, meeting the accuracy requirements of engineering applications.

[0058] Feature extraction is a key link in the present invention, which is used to extract dynamic features with predictive value from the cleaned multi-modal data. In the present invention, feature extraction is carried out for physical, chemical, and environmental characteristic data respectively, using different feature extraction algorithms.

[0059] For physical characteristic data, for stress signals, the main vibration frequency and spectral energy distribution are extracted as features. This is usually achieved through Fourier Transform (FFT), where the energy distribution of high-frequency components is used to evaluate the local fatigue state of the hose under high-frequency vibration.

[0060] For chemical characteristic data, for corrosion diffusion rate data, through a feature extraction method based on reaction kinetics, the change rate of the diffusion coefficient and the cumulative length of the corrosion path are calculated as important features characterizing the chemical aging of the hose material.

[0061] For environmental characteristic data, for the temperature change rate, the average change amplitude of the temperature, the fluctuation range of the change rate, and the correlation coefficient with the operating load are extracted through statistical analysis as important parameters describing the change of the hose operating environment.

[0062] The core effect of the feature extraction algorithm is to extract key features that can intuitively reflect the state of the hose from multimodal data, providing a basis for subsequent simulation modeling and life prediction.

[0063] Example: In a test, stress vibration data of a petroleum transfer hose in a complex environment was collected, and the main vibration frequencies were extracted and found to be concentrated in the range of 10 Hz to 30 Hz. The high-frequency components increased significantly after running for 20 hours, indicating that the hose may have local fatigue problems.

[0064] Preferably, the cleaned dynamic feature data is compressed by a feature compression model, and the feature compression model includes a variational autoencoder. The variational autoencoder performs latent space mapping on the cleaned dynamic feature data, removes low-contribution features, and generates fused feature data with retained key dynamic features.

[0065] After the initial cleaning process, dynamic feature data may still have high dimensions and complex features. For example, physical property data may contain multiple frequency components and time series of stress changes, chemical property data may involve multiple change patterns of diffusion rate, and environmental property data may contain the interaction effect of temperature and humidity. Direct use of these data for subsequent modeling and analysis may lead to low computational efficiency and model overfitting. Therefore, it is necessary to use a feature compression model to map it to a low-dimensional space while retaining as much feature information as possible that is critical to life loss prediction.

[0066] The variational autoencoder is a generative deep learning model, which is composed of an encoder and a decoder at its core. The encoder is responsible for mapping the input data to a low-dimensional representation in the latent space, and the decoder restores the original data by reconstructing the low-dimensional representation, thereby ensuring that the representation of the latent space has the greatest information retention. In the present invention, the key functions of the variational autoencoder include the following two points:

[0067] Latent space mapping: By encoding the cleaned dynamic feature data into a low-dimensional representation of the latent space, redundant information is compressed while removing noise or low-contribution features. Feature selection and reconstruction: The feature representation in the latent space is constrained by the reconstruction loss function to ensure that the low-dimensional feature representation retains important information consistent with the original data during reconstruction.

[0068] In the present invention, the variational autoencoder is used to process the cleaned dynamic feature data (including physical properties, chemical properties and environmental characteristics data), specifically including the following steps:

[0069] (1) The cleaned dynamic feature data is input into the encoder part of the variational autoencoder. The data includes the main vibration frequency of the stress signal, the change pattern of the corrosion diffusion rate, the dynamic amplitude of the temperature change, etc.

[0070] (2) The high-dimensional feature data is mapped by an encoder to generate a low-dimensional feature representation in the latent space. During this process, the variational autoencoder imposes a regularization constraint on the representation of the latent space to ensure that its distribution satisfies the Gaussian distribution, so as to retain the global importance of the features.

[0071] (3) The low-dimensional feature representation in the latent space is input into the decoder, and by reconstructing the original dynamic feature data, the reconstruction error is calculated to optimize the model parameters. The goal of model optimization is to minimize the reconstruction error while achieving the maximum compression of the features.

[0072] (4) The optimized latent space representation is used as the final output to form the fused feature data, which is a low-dimensional representation of the high-dimensional dynamic feature data and retains the most important information for life loss prediction in physical, chemical, and environmental characteristics.

[0073] While removing redundant information and low-contribution features, the fused feature data can completely reflect the core information related to life loss in the dynamic feature data, such as key stress change patterns, corrosion diffusion trends, and temperature influence laws. The compressed fused feature data has a lower dimension but a higher information density, which can significantly improve the computational efficiency of subsequent simulation modeling and prediction algorithms and reduce the risk of overfitting. The fused feature data can express the dynamic feature data in a more compact form through the low-dimensional representation of the latent space, thus supporting the operation of subsequent prediction models in resource-constrained environments.

[0074] Example: In an experiment, multimodal data of a petroleum transfer hose under complex operating conditions was collected, including physical property data (stress vibration signals, 200 feature dimensions), chemical property data (corrosion diffusion rate signals, 50 feature dimensions), and environmental property data (temperature change signals, 30 feature dimensions). The variational autoencoder was used to process this data. First, the 280-dimensional high-dimensional data was mapped to a 10-dimensional latent space representation by the encoder, and the reconstruction quality was verified by the decoder. The results show that after feature compression, the fused feature data can accurately reflect the key information of the original data, and its reconstruction error is less than 3%. Further verification found that the model accuracy for life loss prediction using the fused feature data was improved by about 15%, and the computational efficiency was increased by 50%.

[0075] As Figure 2 shown, based on the fused feature data, a multi-field collaborative simulation environment for the petroleum transfer hose is constructed, the force field, thermal field, and chemical field are coupled and modeled, the finite element analysis method and the multi-feature tensor decomposition algorithm are used to simulate the multi-field interaction, and the loss distribution data characterizing the local loss distribution of the petroleum transfer hose material is generated;

[0076] The multi-field collaborative simulation environment is an integrated modeling framework designed for the coupled influence of multiple physical effects on the oil transfer hose under complex working conditions. Its goal is to reproduce the combined effects of the force field (such as internal pressure and external load), thermal field (such as temperature gradient), and chemical field (such as corrosion diffusion) on the hose material through numerical simulation methods. Through this simulation environment, the loss behavior of the hose material under complex conditions can be systematically revealed.

[0077] In the specific implementation, the simulation environment takes the fused feature data as input. The fused feature data contains key dynamic information such as stress vibration frequency, corrosion diffusion rate, and temperature change rate. These data serve as the boundary conditions and initial parameters of the multi-field simulation model and are used to drive the simulation of the force field, thermal field, and chemical field.

[0078] For force field modeling, based on the mechanical modeling method, the stress distribution of the hose under different pressure gradients and shear stresses is calculated through finite element analysis (FEA). The input data includes stress vibration frequency and wall thickness. In the modeling process, the hose is decomposed into multiple finite element units, and the stress distribution of each unit is calculated.

[0079] For thermal field modeling, the heat conduction analysis method is adopted to establish a thermal field model to describe the heat flow transfer and material thermal aging process of the hose under the action of temperature gradient. The input data is the temperature change rate and ambient temperature distribution. By simulating the transfer path and rate of heat flow, the thermal influence distribution of different regions of the hose is generated.

[0080] For chemical field modeling, based on the principle of diffusion kinetics, a chemical field model is constructed to predict the diffusion path and rate of corrosive substances (such as corrosion products generated by chemical reactions). The input data includes corrosion diffusion rate and material corrosion resistance parameters. By simulating the diffusion behavior of corrosive substances in the hose material, the spatial distribution of corrosion influence is generated.

[0081] The multi-field collaborative simulation environment is not only a simulation of single fields (force field, thermal field, chemical field), but more importantly, an integrated modeling of the interactions between multiple fields. The interaction between single fields (such as the influence of the stress field on heat flow transfer, or the promotion of corrosion diffusion by temperature gradient) is reflected through coupled modeling.

[0082] In the present invention, the finite element method is used as the basis for numerical calculation. Through the multi-feature tensor decomposition algorithm, the multi-modal data of the force field, thermal field, and chemical field are decomposed and recombined to extract the correlation features between fields. For example, the coupling of the force field and the thermal field is reflected in the thermal expansion effect of the hose material, while the coupling of the chemical field and the thermal field is manifested as the dynamic characteristic of accelerated corrosion in a high-temperature environment. Through the tensor decomposition algorithm, the core features of these interactions can be accurately captured, and based on this, the spatial distribution of local loss of the hose material is generated.

[0083] The final output of the multi-field collaborative simulation is the loss distribution data, which represents the local loss state of the hose material in the form of spatial coordinates. These data are obtained by coupling the calculation results of multi-field models and mainly include the following information: the stress loss of each local area; the degree of thermal aging caused by the temperature gradient; the depth and range of corrosion diffusion in the material. The loss distribution data is used for subsequent regional loss prediction modeling, which can significantly improve the accuracy of the life prediction model.

[0084] Preferably, the construction of the multi-field collaborative simulation environment includes the following steps:

[0085] Establish a force field model of the oil transfer hose based on the mechanical modeling method to simulate the local stress distribution of the oil transfer hose under different pressure gradients and shear stresses;

[0086] Establish a thermal field model of the oil transfer hose based on heat conduction analysis to calculate the influence of the temperature gradient on the thermal aging rate of the oil transfer hose material;

[0087] Establish a chemical field model of the oil transfer hose through diffusion kinetics modeling to predict the diffusion path of corrosive substances in the oil transfer hose material;

[0088] Perform multi-field coupling modeling on the above single-field models through numerical calculation methods, where the multi-field coupling is calculated by the following formula:

[0089]

[0090] Where, represents the total loss distribution of the oil transfer hose material; represents the local loss under the action of the force field; represents the local loss under the action of the thermal field; represents the local loss under the action of the chemical field; are the spatial coordinates respectively.

[0091] The force field model is established based on the mechanical modeling method to simulate the local stress distribution of the oil transfer hose during operation. The hose will bear internal pressure, shear stress and vibration load during high-pressure oil transportation, and these mechanical factors will cause material fatigue in local areas of the hose.

[0092] In specific implementation, the hose is divided into multiple finite element units through the finite element analysis (FEA) method, and the stress distribution of each unit is calculated. The input parameters include the internal pressure gradient, the magnitude of the shear stress, and the mechanical properties of the hose material (such as elastic modulus and Poisson's ratio). The numerical results of the stress distribution are used to calculate the mechanical loss of the local material of the hose.

[0093] The thermal field model is constructed by means of heat conduction analysis method and is used to simulate the influence of temperature gradient on the thermal aging rate of hose materials. When the hose conveys high-temperature petroleum, it will be affected by the heat conduction of the internal fluid, and at the same time, the external environmental temperature may cause a temperature difference. This kind of thermal gradient will lead to the degradation of material properties and aging.

[0094] In the process of modeling, based on Fourier's law of heat conduction, a conduction model of heat flow is established. The input data includes the temperature of the petroleum fluid, the external environmental temperature, the thermal conductivity and heat capacity parameters of the material. By calculating the heat flux density and temperature distribution, the heat loss distribution of the local area of the hose is generated.

[0095] The chemical field model is established through diffusion kinetics modeling and is used to predict the diffusion path of corrosive substances in the hose material. Petroleum may contain corrosive substances (such as hydrogen sulfide or carbon dioxide). Under certain temperature and pressure conditions, these substances will penetrate into the hose material through diffusion, resulting in internal corrosion of the material.

[0096] Specifically, when implementing, a diffusion model of the chemical field is established based on Fick's law. The input parameters include the diffusion coefficient of the corrosive substance, the anti-corrosion performance parameters of the material, as well as the environmental temperature and pressure. The modeling results can accurately predict the diffusion depth and distribution of the corrosive substance inside the material.

[0097] Multi-field coupling modeling is the key of the present invention. It realizes the collaborative action modeling of the force field, thermal field and chemical field through numerical calculation methods. The purpose of multi-field coupling is to simulate the interaction effects between different physical fields under actual operating conditions, such as:

[0098] The influence of temperature gradient on mechanical properties (such as high temperature causing a decrease in material strength); the accelerating effect of stress concentration in the force field on corrosion diffusion; the promoting effect of temperature rise on the corrosion reaction rate.

[0099] In multi-field coupling, the following formula is used to comprehensively calculate the results of each single-field model:

[0100]

[0101] By mapping the calculation results of each field to a unified spatial coordinate system, three-dimensional total loss distribution data is generated.

[0102] Through multi-field coupling modeling, the solution can comprehensively capture the interaction effects between the force field, thermal field, and chemical field. For example, stress concentration may accelerate corrosion diffusion, while temperature gradients will further promote corrosion reactions. This comprehensive modeling significantly improves the simulation accuracy of complex working conditions. The generated total loss distribution data not only covers the loss results of each single-field model but also accurately reflects the impact of multi-field interactions on the loss of hose materials, providing a reliable basis for regional loss prediction. Through the total loss distribution data, it is possible to more accurately locate the high-risk areas prone to loss in the hose, providing high-quality data for the input of subsequent life prediction models, thereby significantly enhancing the accuracy and reliability of the prediction.

[0103] In an embodiment, in a certain petroleum transfer project, the operating conditions of the hose include an internal pressure gradient of 12 MPa / m, an ambient temperature of 80 °C, and contain corrosive substance hydrogen sulfide (diffusion coefficient of 0.02 mm² / s). Through the multi-field collaborative simulation environment of the present invention, the force field, thermal field, and chemical field are respectively modeled, and multi-field coupling is achieved through numerical calculation to generate the total loss distribution data of the hose.

[0104] The modeling results show that the bending area of the hose has significantly higher losses than the straight pipe section (the total loss is only 3 times that of the straight pipe section) due to the superposition of stress concentration (stress loss reaches 8 MPa) and temperature gradient (thermal loss is 5 MJ / m³), along with the acceleration of corrosion diffusion (corrosion loss depth reaches 1.2 mm). Subsequently, through optimized design (such as reducing the pressure gradient to 10 MPa / m and increasing the anti-corrosion coating thickness at the bend), it is found through re-modeling verification that the total loss in the bending area is reduced to 60% of the original.

[0105] Preferably, the multi-field coupling is achieved through a multi-feature tensor decomposition algorithm, which specifically includes the following steps:

[0106] Construct a multi-modal tensor of the force field data, thermal field data, and chemical field data of the petroleum transfer hose ;

[0107] Use the following formula to decompose the tensor :

[0108]

[0109] Among them, represents the data point value in the multi-modal tensor of the petroleum transfer hose ; , , respectively represent the weights of the th decomposition component in the dimension; Denotes the rank of tensor decomposition, i.e., the number of low-dimensional components decomposed, usually determined by data complexity and computational efficiency;

[0110] Generate coupling loss distribution data based on the decomposed tensor data for calculating the total local loss of the transfer petroleum hose.

[0111] A multimodal tensor is a high-dimensional data structure used to describe the spatial distribution and coupling relationship of multi-field (force field, thermal field, and chemical field) data in the transfer petroleum hose. In the present invention, the multimodal tensor is constructed based on the following logic:

[0112] (1) Input data: force field data (such as local stress magnitude), thermal field data (such as local temperature change), and chemical field data (such as corrosion diffusion depth).

[0113] (2) Spatial dimension: The dimensions of the tensor respectively represent the local positions of the hose in three-dimensional space, and the value range covers the entire spatial range of the transfer petroleum hose.

[0114] (3) Tensor value: Each tensor data point represents the data value corresponding to the combined influence of the force field, thermal field, and chemical field at that position.

[0115] The purpose of tensor construction is to organize the spatial information of multi-field data and the field interaction characteristics in a unified data structure, providing a basis for subsequent tensor decomposition.

[0116] In order to extract the coupling characteristics between the force field, thermal field, and chemical field and reduce the computational complexity of high-dimensional data, the present invention uses a multi-feature tensor decomposition algorithm to decompose the tensor The principle of tensor decomposition is to decompose high-dimensional data into a combined representation of multiple low-dimensional components, thereby revealing the potential patterns of field interaction. The calculation formula is: .

[0117] The implementation steps specifically include:

[0118] (1) Initialize the tensor: Map the force field, thermal field, and chemical field data to the corresponding positions of the tensor to complete the initialization of the tensor.

[0119] (2) Decomposition algorithm: Use a tensor decomposition algorithm based on the minimum error criterion to optimize and iteratively calculate , , the values of to ensure that the tensor decomposition result can reconstruct the original tensor as accurately as possible.

[0120] (3) Extracting coupling features: low-dimensional components in the decomposition results , , It represents the weights of force field, thermal field and chemical field in different spatial dimensions. By analyzing these components, the interaction characteristics between fields can be extracted.

[0121] After completing the tensor decomposition, the sum of the local losses of the oil transfer hose is calculated based on the decomposed tensor data. The coupling loss distribution at each location is calculated by the decomposed components, which can reflect the comprehensive influence of the force field, thermal field and chemical field at that location. The coupling loss distribution data is stored in the form of three-dimensional spatial coordinates, providing high-precision input for subsequent life loss prediction.

[0122] Through the tensor decomposition algorithm, the present invention can extract the interaction characteristics between force fields, thermal fields and chemical fields from high-dimensional multimodal data. For example, it is found that the high temperature environment in a specific area enhances the stress concentration effect, and the stress concentration accelerates the corrosion diffusion. Tensor decomposition maps high-dimensional data into multiple low-dimensional components, significantly reducing the complexity of subsequent calculations, while retaining key information on multi-field interactions, improving modeling efficiency and the compactness of data expression. The coupling loss distribution data generated after tensor decomposition can more accurately describe the degree of loss at each location of the hose, providing a reliable basis for subsequent regional loss prediction and life assessment.

[0123] Embodiment, in a certain experiment, the operation data of the oil transfer hose includes:

[0124] Force field data: local stress distribution range is 5~15 MPa; thermal field data: temperature gradient range is 30~80 ℃; chemical field data: corrosion diffusion rate is 0.05~0.2 mm / day.

[0125] Construct a multimodal tensor based on this data , the tensor size is 100×100×100, representing the multi-field distribution in the hose length, diameter, and time series dimensions.

[0126] Through the multi-feature tensor decomposition algorithm, the tensor is decomposed into ranks The low-dimensional components of the high-temperature region are used to extract the following interaction features: stress concentration effect in the high-temperature region: force field weight in the high-temperature region The significant increase indicates that temperature has an enhancing effect on local stress concentration. The accelerating effect of stress on corrosion diffusion: The chemical field weight in the region with greater stress It increases significantly, indicating that stress concentration has a significant effect on the corrosion diffusion rate.

[0127] The generated coupling loss distribution data shows that the loss in the curved part of the hose is significantly higher than that in the straight section, with the total loss value reaching 4 times that of the straight section.

[0128] Divide the loss distribution data into multiple local regions of the petroleum transfer hose, and construct a regional loss prediction model. Through quantum optimization and genetic algorithms, achieve collaboration and resource allocation among regions, and generate global life prediction data including the life dissipation trend and remaining life prediction.

[0129] Based on the loss distribution data generated by multi-field collaborative simulation, the present invention conducts spatial region division on the petroleum transfer hose. The goal of region division is to refine the loss distribution of the hose to local regions, so as to capture the loss characteristics of different regions and provide an accurate data basis for subsequent modeling. In specific implementation, based on the geometric structure of the hose (such as length, radius, etc.) and the characteristics of the loss distribution, the hose is divided into several independent regions. For example, it can be divided into several segmented regions of equal length according to the length of the pipe section, or divided into high-loss and low-loss regions according to the gradient change of the loss data. Each local region corresponds to a set of loss distribution data, including the comprehensive loss of the force field, thermal field, and chemical field on this region.

[0130] For each local region, the present invention constructs a regional loss prediction model through the loss distribution data. The core function of the model is to predict the life dissipation trend and remaining life of this region based on the local loss data. (1) Model input: The comprehensive data of the force field loss, thermal field loss, and chemical field loss of each region constitutes the input vector of the model. (2) Model output: The life dissipation trend of this region (representing the time evolution law of material loss) and the remaining life (predicting the available time of the material under the current operating conditions). (3) Algorithm selection: Adopt data-driven machine learning methods, such as regression models trained based on historical operation data or prediction models based on physical laws, to ensure high-precision prediction results.

[0131] After completing the loss prediction of each region, the solution introduces quantum optimization and genetic algorithms to optimize the collaboration and resource allocation among regions. Specifically, quantum optimization is used to simulate the resource allocation decision among regions, and genetic algorithms are used to globally optimize the model parameters. The combination of the two improves the regional collaboration efficiency and prediction accuracy.

[0132] The quantum optimization algorithm is based on the quantum superposition state and quantum collapse mechanism, and simulates the resource competition and collaboration decision among regions. For example, by establishing a quantum state space, where each region corresponds to the state value of a qubit, simulate the optimal allocation of resources (such as operating pressure or load). The quantum collapse process will find an optimal solution to determine the allocation ratio of resources for each region.

[0133] The genetic algorithm optimizes the parameters of the regional loss prediction model through crossover, mutation, and selection operations to ensure the global consistency of the prediction results of each region. For example, by adjusting the weight parameters of the regional model, minimize the global error.

[0134] Through the optimized regional loss prediction model and regional cooperation mechanism, global life prediction data is finally generated. The global life prediction data includes: Life dissipation trend: reflecting the time evolution law of the overall loss of the transfer petroleum hose, such as whether the predicted loss rate increases or decreases over time; Remaining life prediction: based on the weighted average of the remaining lives of each region, predicting the overall remaining serviceable time of the hose. The generation of the global life prediction data depends on the collaborative optimization of the regional models and can reflect the overall life state of the hose under complex operating conditions.

[0135] Through the regionalization of the loss distribution data and the construction of the regional loss prediction model, the present invention can accurately capture the loss characteristics of each local area of the transfer petroleum hose and provide high-precision local life prediction results. Through the combination of quantum optimization and genetic algorithms, the solution realizes the synergistic effect between regions, optimizes the resource allocation and improves the reliability and consistency of the global prediction. The generated global life prediction data can comprehensively reflect the life dissipation law and the overall health state of the hose, providing a scientific basis for operation decision-making and maintenance strategies.

[0136] Example: In a certain oil transportation project, the operating length of the transfer petroleum hose is 100 meters. According to the geometric structure and loss distribution data, the hose is divided into 10 local areas (each section is 10 meters long). Through the regional loss prediction model and optimization algorithm, the following operations are completed:

[0137] (1) Regionalization of the loss distribution data: Using the simulation data to analyze the comprehensive losses of the force field, thermal field and chemical field in each region, it is found that the comprehensive losses in the bending parts (the 3rd region and the 8th region) are significantly higher than those in other regions.

[0138] (2) Construction of the regional loss prediction model: A regression model based on historical operating data is established for each region to predict the life dissipation trend and remaining life of each region. For example, the model of the 3rd region predicts its remaining life to be 50 days, while the remaining life of the straight pipe section is 150 days.

[0139] (3) Regional cooperation and optimization: Through the quantum optimization algorithm, the operating pressure resources are preferentially allocated to the regions with shorter lives (such as the 3rd region), and the model parameters are optimized through the genetic algorithm to ensure the consistency of the global prediction results.

[0140] (4) Generation of the global life prediction data: Based on the optimized regional prediction results, the global life prediction data is generated, showing that the overall remaining life of the hose is 80 days and indicating the key high-loss regions.

[0141] Preferably, as Figure 3As shown in the figure, the establishment of the regional loss prediction model includes the following steps: dividing the loss distribution data into multiple regional data, where each regional data represents the loss information of the corresponding sub-region of the petroleum transfer hose; using the quantum game optimization algorithm to simulate the resource competition and cooperation between regions; combining the genetic algorithm to optimize the parameters of the prediction model for each region, so that the prediction results between regions reach global consistency.

[0142] Before establishing the regional loss prediction model, first regionalize the loss distribution data generated by multi-field coupling. The goal of regional division is to refine the loss distribution to each sub-region of the hose, and the data of each region reflects the local loss characteristics.

[0143] Specifically, the following steps are executed: (1) Input data: The coupled loss distribution data of the force field, thermal field and chemical field, stored with three-dimensional space coordinates as the index. (2) Division method: Based on the geometric characteristics of the petroleum transfer hose (such as length and bending radius) and the gradient change of the loss distribution data, the hose is divided into several sub-regions. For example, for a 100-meter hose, it can be divided into 10 equal-length regions every 10 meters, or according to the loss gradient change, the high-loss regions can be separated separately.

[0144] Each sub-region data includes the comprehensive loss value and its time series change, which are used to construct the local prediction model.

[0145] In the multi-region model, the loss characteristics of different regions may affect each other (for example, the high-loss region may cause the cooperative loss of adjacent regions), so it is necessary to model the resource competition and cooperation between regions. For this purpose, the present invention adopts the quantum game optimization algorithm to simulate the resource allocation and cooperation mechanism between regions.

[0146] Quantum game is a method that combines quantum mechanics and game theory, and simulates the decision-making process between multiple game players through the quantum superposition state and quantum collapse mechanism. In the present invention, the resources between regions (such as pressure distribution, operating parameters) are regarded as the resource variables of the game, and each region participates in the competition and cooperation as a game player.

[0147] Specific implementation steps: (1) Quantum state initialization: Assign a quantum bit to each region, and the superposition state of the quantum bit represents the resource state of the region. For example, the resource allocation state of the region can be represented by a probability distribution where and Represent the probabilities of low - resource and high - resource allocations respectively. (2) Modeling resource competition and cooperation: Establish the payoff function of the game. The function takes the resource status and loss data of each region as input and outputs the payoff of each region. For example, for a high - loss region, the payoff function tends to allocate more resources (such as reducing stress or increasing protection measures). (3) Quantum collapse: Find the global optimal solution through the quantum collapse mechanism, that is, the resource allocation plan with the minimum overall loss. Finally, output the resource allocation ratio of each region.

[0148] After completing the resource allocation among regions, optimize the parameters of the loss prediction model for each region to ensure the global consistency of the model output. The present invention uses a genetic algorithm to optimize the model parameters. The genetic algorithm simulates the selection, crossover, and mutation processes of biological evolution and searches for the optimal solution through iterative optimization. In the present invention, the genetic algorithm is used to optimize the parameters (such as weights, learning rates) of the prediction model for each region to minimize the global prediction error. The specific implementation steps are as follows:

[0149] (1) Initialize the population: Generate several groups of initial parameter combinations for the prediction model of each region as the population of the algorithm. (2) Fitness function: Define the fitness function to measure the quality of the parameter combinations. The fitness function aims at the prediction error of the model, and the smaller the error, the higher the fitness. (3) Selection and crossover: Select excellent parameter combinations according to the fitness function and generate new parameter combinations through the crossover operation. (4) Mutation operation: Randomly perturb some parameters to explore the new parameter space. (5) Iterative optimization: After multiple rounds of iteration, finally converge to the optimal parameter combination to ensure that the output of the prediction model for each region has high precision and the global prediction results are consistent.

[0150] Through the combined application of quantum game optimization and genetic algorithm, the present invention not only optimizes the prediction model for each region but also achieves the global consistency of the prediction results among regions. The goal of global consistency is to make the prediction results of all regions be coordinated in time and space, so as to generate the life dissipation trend and remaining life prediction of the overall hose.

[0151] Through the division of regional data and the establishment of regional prediction models, the present invention can capture the loss laws of each local region and achieve high - precision regional life prediction. The quantum game optimization algorithm simulates the resource competition and cooperation mechanism between high - loss regions and low - loss regions and realizes the global optimality of resource allocation. By optimizing the parameters of the regional prediction model through the genetic algorithm, the reliability and consistency of the global life prediction results are ensured, and the cumulative effect of local prediction errors is reduced. The combination of quantum optimization and genetic algorithm not only improves the prediction accuracy but also significantly reduces the iterative optimization time of the global prediction model.

[0152] Embodiment: In a certain oil transportation project, the operation data of the oil transfer hose is divided into 10 local areas (each section is 10 meters long). Based on the area data, the following steps are completed:

[0153] (1) Area data processing: Extract the force field, thermal field, and chemical field loss information of each area, and it is found that the comprehensive loss values of the 3rd area and the 8th area are significantly higher than those of other areas.

[0154] (2) Quantum game optimization: Simulate the allocation of resources (operating pressure), and determine to allocate more resources (such as reducing the operating pressure) to the high-loss areas through the quantum optimization algorithm. Finally, the resource allocation ratio is: the 3rd area obtains 15% of the resources, the 8th area obtains 20% of the resources, and the remaining areas evenly share 65% of the resources.

[0155] (3) Genetic algorithm optimization: Optimize the parameters of the prediction model for each area. Through multiple rounds of iteration, reduce the global error to less than 1%.

[0156] (4) Global life prediction result: Finally, generate global data including the life dissipation trend and the remaining life prediction, showing that the overall remaining life of the hose is 80 days, and pointing out that the remaining life of the key high-loss area is only 50 days.

[0157] Preferably, the quantum game optimization algorithm is implemented through the following steps: Optimize the resource allocation between areas through the superposition of quantum states, and simulate the iterative collapse process of the quantum state to determine the optimal allocation; the genetic algorithm optimizes the key parameters of the area loss prediction model through crossover, mutation, and selection operations, and uses the following fitness function to evaluate the model performance:

[0158]

[0159] where, represents the fitness of the model; represents the actual value of the th sample; represents the predicted value of the th sample; represents the total number of samples.

[0160] The quantum game optimization algorithm is used to simulate the resource competition and cooperation between areas. Through the superposition state and collapse mechanism of quantum mechanics, find the global optimal solution of area resource allocation. In the present invention, resource allocation refers to the dynamic adjustment of operating conditions (such as operating pressure or protective measures), and the goal is to preferentially allocate more resources to high-loss areas to delay local losses and optimize the overall life of the hose. Implementation steps:

[0161] (1)Quantum state initialization: Assign a qubit to each region. The superposition state of each qubit represents the resource state of the region. For example: represents, where and represent the probabilities of the resource shortage state and the resource sufficient state of region respectively.

[0162] (2)Revenue function construction: Define the revenue function to evaluate the resource allocation effect of each region. The revenue function is related to the reduction amplitude of the regional loss and the contribution to the global loss reduction. For example, a high-loss region has a higher revenue after increasing resources.

[0163] (3)Quantum state evolution: Simulate the resource cooperation and competition among regions through the evolution of the quantum superposition state. Before the quantum state collapses, all possible resource allocation schemes are considered simultaneously to find potential global optimal solutions.

[0164] (4)Quantum state collapse: Use the quantum measurement mechanism to collapse the superposition state into a specific allocation state, that is, determine the resource allocation ratio of each region, and ensure that more resources are preferentially allocated to high-loss regions.

[0165] After the resource allocation is completed, for the loss prediction model of each region, the genetic algorithm is used to optimize the model parameters. The genetic algorithm gradually optimizes the weights and learning parameters of the regional model by simulating the processes of natural selection, crossover, and mutation, making the prediction results of each model more accurate and consistent with the global goal. Implementation steps:

[0166] (1)Population initialization: Generate several groups of initial parameter combinations (population) for each region's prediction model, such as weights, learning rates, etc.

[0167] (2)Fitness function definition: Use the fitness function to evaluate the performance of each group of parameter combinations of the model. The fitness function is defined as the average absolute value of the prediction error, and the formula is as follows: .

[0168] (3)Selection and crossover: Select parameter combinations with better performance according to the fitness function, and generate new parameter combinations through the crossover operation. For example, take partial weights of two high-fitness parameter combinations for linear weighting to form a new combination.

[0169] (4)Mutation operation: Randomly adjust some parameters to explore more possible parameter spaces to avoid falling into local optimal solutions.

[0170] (5)Iterative optimization: After multiple rounds of iteration, finally select the parameter combination with the highest fitness function as the optimal parameter of the regional model.

[0171] Through the combination of quantum game optimization and genetic algorithms, the present invention ensures that the resource allocation and model optimization of each region are consistent with the global life prediction goal. The optimized regional model not only improves the local prediction accuracy but also reduces the cumulative effect of local prediction errors through a globally consistent fitness function.

[0172] The quantum game optimization algorithm ensures that high-loss regions are preferentially allocated more resources, such as reducing operating pressure or increasing protective measures, thereby significantly delaying the loss of high-risk regions. The genetic algorithm improves the performance of the regional loss prediction model through parameter optimization, making the local life prediction results more accurate. The global evaluation and optimization of the prediction errors of the regional model through the fitness function ensure the reliability and consistency of the global life prediction results. The combination of quantum optimization and genetic algorithms not only improves the prediction accuracy but also significantly reduces the optimization calculation time.

[0173] Example: In a certain oil transportation project, the oil transfer hose is divided into 10 local regions (each section is 10 meters long), and the loss prediction model for each region is optimized as follows:

[0174] (1) Quantum game optimization: Taking the operating pressure as the optimization resource, the quantum game optimization algorithm is used to determine the resource allocation ratio. Finally, the 3rd region and the 8th region (high-loss regions) obtain 20% and 15% of the resources respectively, and the remaining regions evenly divide the remaining resources.

[0175] (2) Genetic algorithm optimization: Using the historical operation data of the region, the genetic algorithm optimizes the parameters (such as learning rate and weight) of each region model, and finally reduces the prediction error of each region model to less than 1%.

[0176] (3) Fitness evaluation: Through the fitness function the global prediction performance is evaluated. Finally, the fitness value is 0.05, indicating that the deviation between the prediction result and the actual data is extremely small.

[0177] (4) Global life prediction: The optimized regional model generates the global life prediction result, showing that the overall remaining life of the hose is 80 days, and indicating that the remaining life of the high-loss region is 50 days.

[0178] Based on the global life prediction data and the real-time collected operation status data, a digital twin system is constructed to correct the regional loss prediction model and form a closed-loop feedback optimization process, and output the optimized data that dynamically reflects the operation status and remaining life prediction of the oil transfer hose.

[0179] Preferably, the digital twin system includes a difference analysis module. The difference analysis module compares the operation status data of the oil transfer hose collected in real time with the global life prediction data, generates a difference analysis result to correct the parameters of the regional loss prediction model, and updates the status of the digital twin system.

[0180] The core function of the difference analysis module is to generate a quantitative result reflecting the difference between the two by comparing the real-time collected data with the global life prediction data, and feedback the difference to the regional loss prediction model for correction. The main task of difference analysis is to dynamically adjust the prediction ability of the digital twin system to adapt to the changes in the real-time operation status.

[0181] The operation status data collected in real time from the multi-modal sensors of the hose includes physical, chemical, and environmental characteristic data such as stress, temperature, and corrosion diffusion rate. These data represent the current operation status of the hose.

[0182] The global life prediction data generated from the regional loss prediction model is based on historical operation data and multi-field coupling simulation results, including the life dissipation trend and the remaining life prediction.

[0183] The difference analysis module generates a difference analysis result by calculating the residuals between the real-time data and the prediction data. The calculation of the residuals is completed in the form of a time series, which is used to characterize the degree to which the prediction result deviates from the actual operation status.

[0184] The difference analysis module takes the operation status data and the global life prediction data as inputs, and generates a difference analysis result by comparing the corresponding data points in the time series. Implementation steps:

[0185] (1) Data synchronization and alignment: To ensure the comparability of the real-time data and the prediction data, first synchronize the two time series by the Dynamic Time Warping (DTW) method to align the data at different time steps.

[0186] (2) Residual calculation: The difference analysis module calculates the residual at each time step , and the formula is as follows:

[0187]

[0188] Where represents the residual at the th time step; represents the operation status data collected in real time; represents the prediction data.

[0189] (3) Difference evaluation: By statistically analyzing the mean, variance, and trend changes of the residuals, generate the difference analysis results, such as whether there are systematic biases or large local errors in specific regions.

[0190] The difference analysis results are used to dynamically correct the parameters of the regional loss prediction model, enabling the model to adapt to changes in the operating state in real time, thereby improving the accuracy and timeliness of the prediction. Specific correction methods:

[0191] (1) Parameter adjustment: Based on the difference analysis results, adjust the key parameters (such as weights, learning rates) of the regional model. For example, when the residuals show a certain specific trend, the model's response ability to new data can be improved by increasing the learning rate.

[0192] (2) Model retraining: For some regional models with large differences, introduce real-time data to perform local retraining on the model to eliminate significant biases.

[0193] (3) Multi-region consistency optimization: The corrected regional model will evaluate its global consistency through a fitness function to ensure that the collaborative effect of the prediction results between regions is not affected.

[0194] The corrected regional loss prediction model generates new global life prediction data and updates the state of the digital twin system. This state update includes: Real-time operating state: The updated digital twin system accurately reflects the current loss situation of the hose, including the local loss distribution and the global life dissipation trend; Remaining life prediction: Based on the corrected model, output new remaining life prediction results, which are used to guide the operation decision-making and maintenance plan of the hose.

[0195] By using the difference analysis module to dynamically correct the parameters of the regional loss prediction model in real time, the digital twin system can quickly adapt to changes in the hose operating state, significantly improving the system's dynamic response ability. By eliminating the differences between real-time data and predicted data, the global life prediction data generated by the corrected digital twin system is more accurate, providing a more reliable basis for operation and maintenance. The difference analysis module realizes real-time optimization and iterative update, enabling the digital twin system to continuously improve its performance and maintain a high-fidelity mapping of the hose operating state.

[0196] Example, in a certain oil transportation project, the real-time operating data of the transfer oil hose includes:

[0197] Stress data: The stress fluctuation range at the bending part is 10 - 15 MPa; Temperature data: The temperature gradient is 30 - 70 °C; Corrosion diffusion rate: 0.05 - 0.15 mm / day.

[0198] The predicted data generated by the regional loss prediction model shows that the stress fluctuation range at the bending part is 8 - 12 MPa, with a deviation from the actual data. The difference analysis module corrects the model through the following steps:

[0199] (1) Residual calculation: Calculate the stress residual , and the result shows that the average residual at the bending part is 2.5 MPa. (2) Model parameter adjustment: Based on the residual, adjust the learning rate of the regional model from 0.01 to 0.05 to enhance the model's sensitivity to stress changes. (3) Model retraining: Introduce the latest real-time data to retrain the regional model, and the prediction accuracy of the updated model has increased by 20%.

[0200] The state of the updated digital twin system shows that the accuracy of the total loss distribution at the bending part has been improved to 95%, and the global remaining life prediction result is more accurate, showing that the overall remaining life of the hose is 85 days.

[0201] Preferably, the closed-loop feedback optimization process includes the following steps: Dynamically adjust the input parameters of the multi-field collaborative simulation environment according to the difference analysis results, and recalculate the loss distribution data of the petroleum transfer hose; Input the updated loss distribution data into the digital twin system to update the real-time operating state and remaining life prediction data of the petroleum transfer hose until the prediction error converges within the set range.

[0202] The starting point of the closed-loop feedback optimization is the difference analysis result generated by the difference analysis module, which quantitatively reflects the difference between the real-time collected operating state data and the global life prediction data of the current digital twin system. The difference analysis result is provided in the form of time-series residuals or local deviation distributions, which is used to guide the dynamic adjustment of the multi-field collaborative simulation environment.

[0203] The core of the multi-field collaborative simulation environment is to simulate the coupling effects of the force field, thermal field, and chemical field, and its input parameters include pressure gradient, temperature gradient, corrosion diffusion rate, etc. According to the difference analysis results, dynamically adjust the simulation input parameters to ensure that the simulation model can better match the real-time operating state. Specific adjustment methods: (1) Identify the key input parameters causing the deviation from the difference analysis results. For example, if the actual loss data in a certain area is significantly higher than the simulation result, it can be identified that the pressure gradient may be the main source of the deviation. (2) Adjust the identified key parameters. The adjustment amplitude is dynamically set according to the deviation degree of the difference analysis results. For example, increase the pressure gradient to simulate the actual stress conditions, or increase the temperature gradient to match the high-temperature operating state. (3) Re-input the corrected input parameters into the multi-field collaborative simulation environment to generate new loss distribution data. These data reflect the updated description of the hose operating state in the adjusted simulation environment by re-coupling the force field, thermal field, and chemical field.

[0204] The updated loss distribution data contains the results calculated by the simulation environment under the latest input parameters, characterizing the local loss distribution of the hose material under the combined action of the force field, thermal field, and chemical field. These data are input into the digital twin system through closed-loop feedback, replacing the original prediction data and further updating the system state.

[0205] The updated content includes: local loss distribution, reflecting the latest loss conditions in each area of the hose, especially the dynamic changes in high-loss areas. Global life prediction data, the updated global life dissipation trend and remaining life prediction, used to guide operation decisions.

[0206] The closed-loop feedback optimization process gradually reduces the prediction error through multiple rounds of simulation iteration and digital twin system update. When the prediction error converges within a set range (for example, the mean and variance of the residuals are lower than a certain threshold), the closed-loop process ends and the final optimization result is output. Convergence conditions:

[0207] Residual Meet the conditions:

[0208] Among them, represents the mean of the residuals; represents the standard deviation of the residuals; and are the set convergence thresholds.

[0209] By dynamically adjusting the input parameters of the simulation environment, the closed-loop feedback optimization process can quickly respond to changes in the hose operation state, ensuring that the prediction results are consistent with the actual operation state. Through iterative updates, the prediction error of the optimized digital twin system is significantly reduced, and the generated loss distribution data and life prediction results are more accurate. The iterative mechanism of the closed-loop feedback ensures that the prediction results of the system reach a stable state after multiple rounds of updates, unaffected by initial errors or environmental changes. The optimized digital twin system provides real-time guidance for the operation and maintenance of the hose, such as identifying high-risk areas in advance or adjusting operation parameters to extend the service life.

[0210] Example, in a certain oil transfer project, the hose operation environment is complex, the pressure gradient is 10 MPa / m, the temperature gradient is 50 °C, and the corrosion diffusion rate is 0.1 mm / day. By comparing the initial loss distribution data generated by the multi-field collaborative simulation environment with the real-time operation state data, it is found that the loss prediction deviation in some areas is relatively large.

[0211] The implementation of the closed-loop feedback optimization process is as follows:

[0212] (1) Calculate the residuals between the real-time data and the predicted data, and it is found that the loss prediction in the bent part is underestimated, with the mean residual being 1.5 MPa. (2) Based on the results of the difference analysis, dynamically adjust the input parameters of the multi-field collaborative simulation environment, increase the pressure gradient in the bent part from 10 MPa / m to 12 MPa / m, and at the same time increase the temperature gradient to 60 °C. (3) Recalculate the loss distribution data and generate new data reflecting the adjusted environment, showing that the loss value in the bent part is closer to the real-time data. (4) Input the updated loss distribution data into the digital twin system to generate new global life prediction data. After the update, the overall prediction error of the hose is reduced to 0.5 MPa, and the remaining life prediction in the high-loss area is adjusted from 60 days to 50 days.

[0213] After multiple rounds of iteration, the prediction error converges within the set range, the closed-loop process ends, and the final optimization result is generated, providing real-time decision support for the operation and maintenance of the hose.

[0214] As Figure 4 shown, a system for implementing the method for predicting the life loss of the oil transfer hose includes:

[0215] A multi-modal data acquisition module, which is used to collect the physical characteristics, chemical characteristics, and environmental characteristic data during the operation of the oil transfer hose, and process the collected data to generate fused feature data; collect the physical characteristics (such as stress, vibration), chemical characteristics (such as corrosion diffusion rate), and environmental characteristics (such as temperature, humidity) data during the operation of the oil transfer hose. Through data cleaning and feature extraction, generate fused feature data as the core input for subsequent modeling.

[0216] A multi-field collaborative simulation module, which is used to construct a multi-field collaborative simulation environment for the oil transfer hose based on the fused feature data, perform coupled modeling on the force field, thermal field, and chemical field, and use the finite element analysis method and multi-feature tensor decomposition algorithm to simulate the multi-field interaction, generating loss distribution data representing the local loss distribution of the oil transfer hose material; construct a multi-field collaborative simulation environment for the force field, thermal field, and chemical field based on the fused feature data. Use the finite element analysis (FEA) method and multi-feature tensor decomposition algorithm to model the multi-field interaction and generate loss distribution data representing the local loss distribution of the hose material.

[0217] The regional loss prediction module is used to divide the loss distribution data into multiple local areas of the oil transfer hose, construct a regional loss prediction model, and utilize quantum optimization and genetic algorithms to achieve cooperation and resource allocation among regions, generating global life prediction data including life dissipation trends and remaining life predictions; divide the loss distribution data into multiple local areas, construct a regional loss prediction model. Utilize the quantum optimization algorithm to achieve efficient resource allocation among regions, combine the genetic algorithm to optimize the model parameters, and generate global life prediction data, including life dissipation trends and remaining life.

[0218] The digital twin system module is used to correct the regional loss prediction model and form a closed-loop feedback optimization process based on the global life prediction data and the operation status data collected in real time, and output optimization data that dynamically reflects the operation status and remaining life prediction of the oil transfer hose. Based on the global life prediction data and the operation status data collected in real time, construct a digital twin model. Dynamically correct the regional prediction model through difference analysis, combine the closed-loop feedback optimization process to update the simulation parameters, and realize the real-time mapping of the hose operation status and the iterative optimization of life prediction.

[0219] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, various changes and modifications can be made to the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present application shall be included within the scope of the claims of the present application.

Claims

1. A method for predicting the life loss of a petroleum transfer hose, characterized in that: The following steps are involved: Collect the physical, chemical and environmental characteristics data of the oil transfer hose during operation, and process the data to generate fused feature data; Based on the fusion feature data, a multi-field collaborative simulation environment for oil transfer hoses is constructed, and coupling modeling of force field, thermal field and chemical field is performed. The multi-feature tensor decomposition algorithm is used to simulate the interaction of multiple fields, and loss distribution data that characterizes the local loss distribution of oil transfer hose materials is generated. The loss distribution data is divided into multiple local areas of the oil transfer hose, and a regional loss prediction model is constructed. The cooperation and resource allocation between regions are realized through the quantum game optimization algorithm and the genetic algorithm, and the global life prediction data including the life dissipation trend and the remaining life prediction is generated. The establishment of the regional loss prediction model includes the following steps: the loss distribution data is divided into multiple regional data, each regional data represents the loss information of the corresponding sub-area of ​​the oil transfer hose; The quantum game optimization algorithm is used to simulate resource competition and cooperation between regions. The genetic algorithm is used to optimize the parameters of the prediction model of each region so that the prediction results between regions can achieve global consistency. Based on the global life prediction data and real-time collected operating status data, a digital twin system is constructed to correct the regional loss prediction model and form a closed-loop feedback optimization process, which outputs optimization data that dynamically reflects the operating status and remaining life prediction of the oil transfer hose.

2. The method for predicting the life loss of a petroleum transfer hose according to claim 1, characterized in that: The processing of the physical, chemical and environmental characteristics data collected during the operation of the oil transfer hose includes the following steps: identifying and removing abnormal values ​​in the oil transfer hose operation data through a noise filtering algorithm; using a dynamic time alignment method to perform time offset correction based on the time series of the oil transfer hose operation data; using a feature extraction algorithm to extract the stress vibration frequency, corrosion diffusion rate and temperature change rate in the oil transfer hose operation data, and generate dynamic feature data after cleaning.

3. The method for predicting the life loss of a petroleum transfer hose according to claim 2, characterized in that: The cleaned dynamic feature data is compressed by a feature compression model, which includes a variational autoencoder. The variational autoencoder performs latent space mapping on the cleaned dynamic feature data, removes low-contribution features, and generates fused feature data with retained key dynamic features.

4. The method for predicting the life loss of a petroleum transfer hose according to claim 1, characterized in that: The construction of the multi-field collaborative simulation environment includes the following steps: A force field model of the oil transfer hose is established based on the mechanical modeling method to simulate the local stress distribution of the oil transfer hose under different pressure gradients and shear stresses. Based on heat conduction analysis, a thermal field model of oil transfer hose is established to calculate the effect of temperature gradient on the thermal aging rate of oil transfer hose materials; The chemical field model of the oil transfer hose is established through diffusion kinetic modeling to predict the diffusion path of corrosive substances in the oil transfer hose material; The above single-field model is modeled into multi-field coupling by numerical calculation method, where multi-field coupling is calculated by the following formula: in, Represents the total loss distribution of oil transfer hose materials; represents the local loss under the force field; Represents the local loss under the action of thermal field; Represents local losses under the action of chemical fields; are the space coordinates respectively.

5. The method for predicting the life loss of a petroleum transfer hose according to claim 4, characterized in that: The multi-field coupling is achieved by a multi-characteristic tensor decomposition algorithm, which specifically includes the following steps: Constructing a multi-modal tensor of force, thermal and chemical field data for oil hoses ; Use the following formula to transform the tensor To break it down: in, Represents the multi-modal tensor of the oil hose The data point value of , , Respectively represent The decomposition components are Weights on dimensions; represents the rank of tensor decomposition; Coupling loss distribution data is generated based on the decomposed tensor data and used to calculate the sum of local losses in the oil transfer hose.

6. The method for predicting the life loss of a petroleum transfer hose according to claim 1, characterized in that: The quantum game optimization algorithm is implemented by the following steps: optimizing the resource allocation between regions through quantum state superposition, simulating the iterative collapse process of quantum states to determine the optimal allocation; the genetic algorithm optimizes the key parameters of the regional loss prediction model through crossover, mutation and selection operations, and evaluates the model performance using the following fitness function: in, Indicates the fitness of the model; Indicates The actual value of samples; Indicates The predicted value of samples; Represents the total number of samples.

7. The method for predicting the life loss of a petroleum transfer hose according to claim 1, characterized in that: The digital twin system includes a difference analysis module, which generates a difference analysis result by comparing the real-time collected operating status data of the oil transfer hose with the global life prediction data to correct the parameters of the regional loss prediction model and update the status of the digital twin system.

8. The method for predicting the life loss of a petroleum transfer hose according to claim 1, characterized in that: The closed-loop feedback optimization process includes the following steps: dynamically adjusting the input parameters of the multi-field collaborative simulation environment according to the difference analysis results, and recalculating the loss distribution data of the oil transfer hose; inputting the updated loss distribution data into the digital twin system, and updating the real-time operating status and remaining life prediction data of the oil transfer hose until the prediction error converges within the set range.

9. A system for implementing the method for predicting the life loss of a petroleum transfer hose according to any one of claims 1 to 8, characterized in that: include: A multimodal data acquisition module is used to collect data on the physical, chemical and environmental characteristics of the oil transfer hose during operation, and process the collected data to generate fused feature data; The multi-field collaborative simulation module is used to build a multi-field collaborative simulation environment for oil transfer hoses based on fusion feature data, conduct coupled modeling of force field, thermal field and chemical field, and use a multi-feature tensor decomposition algorithm to simulate multi-field interactions, generating loss distribution data that characterizes the local loss distribution of oil transfer hose materials; The regional loss prediction module is used to divide the loss distribution data into multiple local areas of the oil transfer hose, and to build a regional loss prediction model, and to use the quantum game optimization algorithm and the genetic algorithm to achieve inter-regional collaboration and resource allocation, and to generate global life prediction data including life dissipation trend and remaining life prediction; the establishment of the regional loss prediction model includes the following steps: dividing the loss distribution data into multiple regional data, each regional data represents the loss information of the corresponding sub-region of the oil transfer hose; using the quantum game optimization algorithm to simulate the resource competition and collaboration between regions; combining the genetic algorithm to optimize the parameters of the prediction model of each region, so that the prediction results between regions reach global consistency; The digital twin system module is used to correct the regional loss prediction model and form a closed-loop feedback optimization process based on the global life prediction data and real-time collected operating status data, and output optimization data that dynamically reflects the operating status and remaining life prediction of the oil transfer hose.

Images