Method and equipment for predicting residual life of oil and gas drilling equipment
By combining image segmentation models and Kalman filters, a linear degradation model was established, which solved the error and limitation problems in the life prediction of oil and gas drilling equipment, and achieved more accurate remaining life prediction and equipment health management.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CHINA UNIV OF PETROLEUM (BEIJING)
- Filing Date
- 2025-11-28
- Publication Date
- 2026-05-05
AI Technical Summary
Existing methods for predicting the lifespan of oil and gas drilling equipment rely on indirect wear effect detection, which has large errors and limited application, making it difficult to accurately predict the remaining lifespan of equipment in complex operating environments.
An image segmentation model is used to extract the wear area of components. A linear degradation model is established by combining a Kalman filter and a simulated Wiener process. The degradation rate state estimate is obtained by iterative updating through the expectation-maximization algorithm, and the remaining life of the components is predicted.
It improves the accuracy and versatility of life prediction, avoids the subjectivity of parameter settings, and can more accurately reflect the health status of equipment, supporting the effective maintenance of oil and gas drilling equipment.
Smart Images
Figure CN121980730A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of oil and gas drilling technology, and in particular to a method and equipment for predicting the remaining life of oil and gas drilling equipment. Background Technology
[0002] Oil and gas drilling equipment is crucial for oil and gas resource exploration and development, closely linked to development and production, internal and external gathering and transportation, and terminal distribution. Unplanned failures of critical components leading to unexpected equipment downtime can cause incalculable losses. Therefore, it is essential to apply effective health management techniques to critical components, predict the remaining lifespan of oil and gas drilling equipment, and ensure that the relationship between maintenance activity scheduling and corresponding maintenance resource management is fully considered, ultimately achieving maintenance at the lowest possible cost.
[0003] Currently, many data-driven health management methods have been developed, most of which rely on degradation data detected by indirect wear effects, such as structural noise, preload, and motor current signals. However, detection methods based on indirect wear effects inevitably introduce errors due to the introduction of other physical quantities, affecting the accuracy of predictions, and have significant limitations in application to the complex operating environment of oil and gas drilling.
[0004] To address the aforementioned shortcomings, this application proposes a method and equipment for predicting the remaining lifespan of oil and gas drilling equipment. Summary of the Invention
[0005] This application provides a method and equipment for predicting the remaining life of oil and gas drilling equipment, in order to improve the accuracy of life prediction.
[0006] In a first aspect, this application provides a method for predicting the remaining life of oil and gas drilling equipment, comprising:
[0007] Collect image data of components over a historical period and obtain the wear area of the image data based on an improved image segmentation model;
[0008] The degradation characteristics of components over a historical period are obtained based on the wear area;
[0009] A linear degradation model is established based on degradation characteristics. The linear degradation model includes a system equation based on degradation rate and a state equation based on degradation characteristics.
[0010] The optimal estimate of the degradation rate state estimation function is obtained by solving the linear degradation model using a Kalman filter that is iteratively updated based on the expectation-maximization algorithm.
[0011] Based on the optimal estimate of the degradation characteristics and degradation rate state estimation function, the probability density function of the remaining life of the component is obtained by simulating the Wiener process. The probability density function is used to predict the remaining life of the component.
[0012] In one possible design, obtaining the optimal estimate of the degradation rate state estimation function by solving the linear degradation model using a Kalman filter iteratively updated based on the expectation-maximization algorithm includes:
[0013] The linear degradation model is solved by a Kalman filter to obtain the iterative solution of the function parameters in the degradation rate state estimation function; wherein, the function parameters include the expectation of the posterior estimate, the variance of the posterior estimate, and the Kalman gain;
[0014] The initialization parameters of the Kalman filter are updated based on the iterative solution of the function parameters using the expectation-maximization algorithm combined with the RTS smoothing algorithm.
[0015] The steps of solving the linear degenerate model using the Kalman filter are returned based on the updated initialization parameters until the expectation-maximization algorithm meets the convergence condition or reaches the number of iterations. Then, the iterative solution of the function parameters is taken as the optimal estimate of the function parameters.
[0016] The optimal estimate of the degradation rate state estimation function is obtained based on the optimal estimate of the function parameters.
[0017] In one possible design, the initialization parameters of the Kalman filter include the expectation of the initial degraded state of the state estimation function of the degradation rate, the variance of the initial degraded state, the system process noise, and the diffusion coefficient.
[0018] The step of updating the initial parameters of the Kalman filter using the expected maximum algorithm combined with the RTS smoothing algorithm based on the iterative solution of the function parameters includes:
[0019] The first expected value, the second expected value, and the third expected value are obtained by using the RTS smoothing algorithm based on the iterative solution of the state estimation function of the degradation rate; wherein, the first expected value is the expected value of the degradation rate at the current time in the i-th iteration; the second expected value is the expected value of the square of the degradation rate at the current time in the i-th iteration; and the third expected value is the expected value of the product of the degradation rate at the current time in the i-th iteration and the degradation rate at the previous time.
[0020] The first, second, and third expected values are used as the expectations of the E-step of the expectation-maximization algorithm. The updated expectations of the initial degraded state, the variance of the initial degraded state, the system process noise, and the diffusion coefficient are obtained through the M-step operation of the expectation-maximization algorithm.
[0021] In one possible design, the system equation based on the degradation rate is:
[0022]
[0023] in, For components The degradation rate at any given time; This is the drift coefficient;
[0024] The state equation based on the degradation characteristics is:
[0025]
[0026] Among them, y k The degradation characteristics at time k; The diffusion coefficient is denoted as . To comply with It follows a normal distribution.
[0027] In one possible design, obtaining the degradation characteristics of the component over a historical period based on the wear area includes:
[0028] The degradation characteristics are obtained based on the wear area and the historical time using a degradation characteristic formula; the degradation characteristic formula is:
[0029]
[0030] Among them, y k For the tth k The degradation characteristics of time, The wear area of the components at any given time. It is a constant. These are the characteristic coefficients.
[0031] In one possible design, the optimal estimation of the remaining life probability density function of the component based on the degradation characteristics and the degradation rate state estimation function is obtained by simulating a Wiener process, including:
[0032] The probability density function of the remaining lifetime is obtained according to the following formula:
[0033]
[0034] in, Let be the probability density function of the remaining lifespan of the component. Here, represents the damaged area data of the component, and w represents the failure threshold of the component. The diffusion coefficient is... This is the optimal estimate of the degradation rate state estimation function. Let k be the time length up to time k.
[0035] In one possible design, obtaining the wear area of the image data according to the improved image segmentation model includes: identifying the initial wear area of the image data according to the improved image segmentation model; dividing the initial wear area into an initial internal wear area and an initial edge wear area; processing the initial internal wear area according to a preset loss function to obtain an internal wear area; constructing an edge loss function according to a preset structural similarity function, and dividing the initial edge wear area into multiple edge wear image blocks; processing the multiple edge wear image blocks according to the edge loss function to obtain an edge wear area; and integrating the internal wear area and the edge wear area into the wear area of the image data.
[0036] In one possible design, the improvement process of the image segmentation model includes: acquiring an image sample dataset of the components, and performing image enhancement processing on multiple image samples in the image sample dataset to obtain multiple enhanced image samples; dividing the multiple enhanced image samples to obtain a training set and a test set; training a preset image segmentation model based on the training set to obtain an original image segmentation model; testing the original image segmentation model based on the test set to complete the testing of the original image segmentation model; extracting the backbone network and the original decoder of the original image segmentation model; extracting the feature extractor from the backbone network and configuring the feature extractor to obtain a configured feature extractor; establishing a feature compensation path process after each transposed convolutional layer in the original decoder according to a preset residual connection mechanism; and improving the original image segmentation model based on the configured feature extractor and the feature compensation path process to obtain an improved image segmentation model.
[0037] In one possible design, processing the plurality of edge-wearing image patches according to the edge loss function to obtain a formula for calculating the edge-wearing area includes:
[0038]
[0039] In the formula, L is the loss value of the edge wear area; k is the number of image patches; A value of 0 indicates that the image does not contain the edge of the component, and the calculation result of the image position is not included in the loss and middle; and Let represent the local mean of the i-th image patch in the predicted mask and the real mask, respectively; and Let represent the standard deviations of the i-th image patch in the predicted mask and the true mask, respectively; This represents the covariance between the predicted mask and the true mask for the i-th image patch; and It is a numerical stability constant used to prevent the denominator from being zero.
[0040] Secondly, this application provides a device for predicting the remaining life of oil and gas drilling equipment, comprising:
[0041] The segmentation module is used to collect image data of the parts over a historical period and obtain the wear area of the image data based on the improved image segmentation model;
[0042] The feature module is used to obtain the degradation characteristics of parts over a historical period based on the wear area;
[0043] The model module is used to establish a linear degradation model based on degradation characteristics. The linear degradation model includes a system equation based on the degradation rate and a state equation based on the degradation characteristics.
[0044] The iterative module is used to solve the linear degradation model by iteratively updating the Kalman filter based on the expectation-maximization algorithm to obtain the optimal solution of the degradation rate state estimation function and the optimal estimate of the degradation rate state estimation function.
[0045] The prediction module is used to obtain the probability density function of the remaining life of the component by simulating the Wiener process, based on the optimal solution of the degradation rate state estimation function and the degradation characteristics. The probability density function is used to predict the remaining life of the component.
[0046] Thirdly, this application provides an electronic device, including: a memory and a processor;
[0047] The memory stores the instructions that the computer executes;
[0048] The processor executes computer execution instructions stored in memory, causing the processor to perform the first aspect and / or various possible implementations of the first aspect as described above.
[0049] This application provides a method and apparatus for predicting the remaining life of oil and gas drilling equipment. It constructs degradation features by extracting the damaged area of wear defects on the surface of equipment components, thereby establishing a linear degradation model that accurately reflects the health status of equipment components. The optimal estimate of the degradation rate state estimation function is obtained by solving the linear degradation model using a Kalman filter updated iteratively based on the expectation-maximization algorithm, avoiding subjectivity in parameter settings and improving the accuracy of life prediction. Attached Figure Description
[0050] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.
[0051] Figure 1A flowchart illustrating a method for predicting the remaining life of oil and gas drilling equipment provided in this application embodiment. Figure 1 ;
[0052] Figure 2 A flowchart illustrating a method for predicting the remaining life of oil and gas drilling equipment provided in this application embodiment. Figure 2 ;
[0053] Figure 3 A flowchart illustrating a method for predicting the remaining life of oil and gas drilling equipment provided in this application embodiment. Figure 3 ;
[0054] Figure 4 A flowchart illustrating a method for predicting the remaining life of oil and gas drilling equipment provided in this application embodiment. Figure 4 ;
[0055] Figure 5 This is a schematic diagram illustrating the enhancement effect of defect image data provided in an embodiment of this application;
[0056] Figure 6 A schematic diagram of the structure of the improved image segmentation model provided in the embodiments of this application;
[0057] Figure 7 A graph showing the variation of training loss values for the improved image segmentation model provided in this application embodiment;
[0058] Figure 8 Comparison of segmentation performance of the improved image segmentation model provided in the embodiments of this application;
[0059] Figure 9 The pitting image segmented according to the trained improved image segmentation model provided in the embodiments of this application;
[0060] Figure 10 A comparison diagram of the original defect area data provided for the embodiments of this application and the degradation feature data obtained according to the embodiments of this application;
[0061] Figure 11 The probability density function distribution diagram for predicting the remaining life of a ball screw provided in the embodiments of this application;
[0062] Figure 12 A graph showing the predicted remaining life of the ball screw and its 30% accuracy range, provided in an embodiment of this application.
[0063] Figure 13 A schematic diagram of a device for predicting the remaining life of oil and gas drilling equipment, provided as an embodiment of this application;
[0064] Figure 14 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.
[0065] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation
[0066] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of devices and methods consistent with some aspects of this application as detailed in the appended claims, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without inventive effort are within the scope of protection of this invention.
[0067] It should be noted that "at the time of..." in the embodiments of this application can be either at the instant when a certain situation occurs, or for a period of time after the occurrence of a certain situation. The embodiments of this application do not make specific limitations on this.
[0068] Existing technology explanation:
[0069] The UNet network model is a convolutional neural network architecture that adds skip connections to the encoder-decoder structure, enabling the network to better capture and preserve spatial information. It is primarily used for medical image segmentation tasks. Due to its excellent performance and flexibility, UNet has been widely applied to various image segmentation tasks.
[0070] The Kalman filter is a recursive filter used to estimate the state of a system in the presence of noise. It has wide applications in many fields, including navigation, control, signal processing, and computer vision. The core idea of the Kalman filter is to recursively estimate the system state by combining a dynamic model of the system with observational data, while minimizing the estimation error.
[0071] The Rauch-Tung-Striebel Smoother (RTS) smoothing algorithm is a post-processing method used to improve Kalman filter results. By utilizing future observations to re-estimate past system states, it can significantly improve the accuracy and stability of state estimation.
[0072] The Expectation-Maximization Algorithm (EMA) is an iterative algorithm used to maximize the likelihood function or posterior probability in the presence of latent variables. The core idea of EMA is to progressively optimize model parameters by alternating between two steps—the E-step (expectation step) and the M-step (maximization step).
[0073] Brownian motion: Botanist Brown observed tiny particles floating on the surface of a calm liquid under a microscope and discovered that they were constantly undergoing chaotic motion. This phenomenon was later called Brownian motion.
[0074] The Wiener process is an important independent increment process and a mathematical model of Brownian motion. It describes a continuous-time stochastic process with the following characteristics: (1) The current value of the process contains all the information needed to make its future predictions. (2) The Wiener process has independent increments. The probability distribution of its change over any given time interval is independent of the probability of its change over any other time interval. (3) Its change over any finite time interval follows a normal distribution, and its variance increases linearly with the length of the time interval.
[0075] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will be described below with reference to the accompanying drawings.
[0076] Figure 1 A flowchart illustrating a method for predicting the remaining life of oil and gas drilling equipment provided in this application embodiment. Figure 1 ,like Figure 1 As shown, the method includes:
[0077] S101. Collect image data of the components over a historical period and obtain the wear area of the image data based on the improved image segmentation model.
[0078] Specifically, step S101 includes:
[0079] S1011. Identify the initial wear area of the image data based on the improved image segmentation model.
[0080] Specifically, the improvement process of the image segmentation model includes:
[0081] S10111. Obtain the image sample dataset of the parts and perform image enhancement processing on multiple image samples in the image sample dataset to obtain multiple enhanced image samples.
[0082] Specifically, pixel size normalization, contrast enhancement, tone transformation, grayscale conversion, Gaussian noise reduction, and image flipping are performed on multiple image samples in the image sample dataset to obtain multiple enhanced image samples.
[0083] S10112. Divide the multiple enhanced image samples into training and test sets.
[0084] In addition, multiple enhanced image samples can be divided into training set, validation set and test set.
[0085] The training set is used to train the image segmentation model, the validation set is used to validate the image segmentation model, and the test set is used to test the image segmentation model.
[0086] S10113. Train the preset image segmentation model based on the training set to obtain the original image segmentation model.
[0087] S10114. Test the original image segmentation model using the test set to complete the test of the original image segmentation model.
[0088] S10115. Extract the backbone network and original decoder of the original image segmentation model.
[0089] S10116. Extract the feature extractor from the backbone network and configure the feature extractor to obtain the configured feature extractor.
[0090] For example, the configured feature extractor is a lightweight VGG16 feature extractor.
[0091] S10117. After each transposed convolutional layer in the original decoder, a feature compensation path process is established according to the preset residual connection mechanism.
[0092] Specifically, a Dropout layer is added during each upsampling process of the original decoder, and a feature compensation path process is established after each transposed convolutional layer using a residual connection mechanism.
[0093] S10118. Improve the original image segmentation model based on the configured feature extractor and feature compensation path process to obtain an improved image segmentation model.
[0094] S1012. Divide the initial wear area into the initial internal wear area and the initial edge wear area.
[0095] S1013. Process the initial internal wear area according to the preset loss function to obtain the internal wear area.
[0096] S1014. Construct an edge loss function based on a preset structural similarity function, and divide the initial edge wear area into multiple edge wear image blocks.
[0097] Specifically, the initial edge wear area is divided into M×M edge wear image blocks using a local window calculation paradigm.
[0098] S1015. Process multiple edge wear image blocks according to the edge loss function to obtain the edge wear area.
[0099] In this embodiment, multiple edge-wearing image patches are processed according to an edge loss function to obtain a formula for calculating the edge wear area, including:
[0100]
[0101] In the formula, L is the loss value of the edge wear area; k is the number of image patches; A value of 0 indicates that the image does not contain the edge of the component, and the calculation result of the image position is not included in the loss and middle; and Let represent the local mean of the i-th image patch in the predicted mask and the real mask, respectively; and Let represent the standard deviations of the i-th image patch in the predicted mask and the true mask, respectively; This represents the covariance between the predicted mask and the true mask for the i-th image patch; and It is a numerical stability constant used to prevent the denominator from being zero.
[0102] Specifically, when the pixel categories of two mask positions corresponding to a certain image patch are completely identical... A value of 0 indicates that the image patch does not contain edge regions, and the calculation result of the image patch position is not included in the loss sum; when the pixel categories of the two mask positions corresponding to the image patch are not completely consistent, The value will be between (0, 0.5], indicating that the image patch contains edge regions. EC Loss uses this method to detect defect edge regions.
[0103] S1016. Integrate the internal wear area and the edge wear area into the wear area of the image data.
[0104] S102. Obtain the degradation characteristics of the parts over a historical period based on the wear area.
[0105] Specifically, for each historical time period, a polynomial is constructed based on the corresponding wear area and historical time to realize the transformation from nonlinear features to linear features, thereby obtaining the degradation features corresponding to that historical time period.
[0106] S103. Establish a linear degradation model based on degradation characteristics. The linear degradation model includes a system equation based on degradation rate and a state equation based on degradation characteristics.
[0107] Specifically, assuming the component undergoes a linear degradation process, we construct system equations based on the degradation rate and state equations based on the degradation characteristics.
[0108] S104. The optimal estimate of the degradation rate state estimation function is obtained by solving the linear degradation model using a Kalman filter that is iteratively updated based on the expectation-maximization algorithm.
[0109] Specifically, by combining the RTS smoothing algorithm with the expectation-maximization algorithm to estimate the initial parameters of the model, and performing iterative calculations with the Kalman filter, the optimal initial parameters of the model and the adaptive estimation of the degradation state are achieved; thus, the optimal estimate of the degradation rate state estimation function is obtained.
[0110] S105. Based on the optimal estimate of the degradation characteristics and degradation rate state estimation function, the probability density function of the remaining life of the component is obtained by simulating the Wiener process. The probability density function is used to predict the remaining life of the component.
[0111] Specifically, based on the optimal estimation of degradation characteristics and degradation rate state estimation function, the probability density function of the remaining life of the component is obtained by fitting the Wiener process to simulate the degradation trend. The probability density function is used to predict the remaining life of the component.
[0112] This application provides a method for predicting the remaining life of oil and gas drilling equipment. Combining machine vision technology, it extracts the damaged area of wear defects on the surface of equipment components using an improved U-Net semantic segmentation network. Based on the defect evolution law, degradation features are constructed using the damaged area to accurately reflect the health status of equipment components. An optimal estimate of the degradation rate state estimation function is obtained by solving a linear degradation model using a Kalman filter updated iteratively based on the expectation-maximization algorithm, avoiding subjectivity in parameter settings. The probability density function of the remaining life of the components is obtained by simulating a Wiener process, and the remaining life of the components is predicted. This application improves the accuracy of life prediction.
[0113] To address the issues of small sample size, low contrast, and weak edge features in wear defect image data of oil and gas drilling equipment used in practical applications, the original UNet network was improved, effectively enhancing the model's edge segmentation performance and enabling accurate quantitative calculation of the damaged area.
[0114] Figure 2 A flowchart illustrating a method for predicting the remaining life of oil and gas drilling equipment provided in this application embodiment. Figure 2 ; Figure 3A flowchart illustrating a method for predicting the remaining life of oil and gas drilling equipment provided in this application embodiment. Figure 3 ,like Figure 2 and Figure 3 As shown, the method includes:
[0115] S210. Collect image data of the components over a historical period and obtain the wear area of the image data based on the improved image segmentation model.
[0116] Specifically, the camera equipment is fixed at a preset location to collect image data of the evolution process of wear defects on the surface of parts within the preset area. The camera takes images at a fixed angle and records the shooting time. .
[0117] Specifically, the improved UNet segmentation network model, which has been trained, is used as the preset image segmentation model to segment image data collected at different times in history to obtain wear images; the area of the wear images is calculated to obtain the wear area at different times.
[0118] S220. Obtain degradation characteristics using degradation characteristic formula based on wear area and historical time.
[0119] Specifically, due to manufacturing errors in various equipment components, the load distribution is uneven in the initial stage of use. These errors are gradually eliminated during use, causing the wear rate to decrease over time. Therefore, the combined wear damage area... With time Polynomial features are constructed as degenerate data features to achieve the transformation of data into approximately linear features. The formula for the degenerate features is:
[0120] (1)
[0121] Among them, y k For the tth k The degradation characteristics of time, The wear area of the components at any given time. It is a constant. The characteristic coefficients, For the middle The characteristic coefficients, the values of which depend on the specific working conditions.
[0122] Specifically, since the initial wear area is too small, the semantic segmentation network may have defects and misjudgments. Therefore, a threshold is set to remove data with too small an area value and use data with an area value greater than the threshold as degenerate features.
[0123] Specifically, the damage area obtained by S201 is expressed as... ,in, The damaged area of a specific wear defect at a given time. Then, according to the degradation characteristic formula, it can be obtained... All Degenerate Feature Datasets at Any Time .
[0124] S230. Establish a linear degradation model based on degradation characteristics. The linear degradation model includes a system equation based on degradation rate and a state equation based on degradation characteristics.
[0125] The system equation based on the degradation rate is:
[0126] (2)
[0127] in, For components The degradation rate at any given time; This is the drift coefficient.
[0128] The state equation based on the degradation characteristics is:
[0129] (3)
[0130] Among them, y k The degradation characteristics at time k; The diffusion coefficient is denoted as . To comply with It follows a normal distribution.
[0131] S240. Obtain the iterative solution of the function parameters in the degradation rate state estimation function by solving the linear degradation model through a Kalman filter; wherein, the function parameters include the expectation of the posterior estimate, the variance of the posterior estimate, and the Kalman gain.
[0132] As one implementation method, S240 specifically includes:
[0133] S241, Initialize parameters The initial covariance matrix.
[0134] S242, Estimation System state at any given time:
[0135] (4)
[0136] (5)
[0137] This step involves predicting the state of the next time step. Q is the process noise covariance matrix.
[0138] S243. Update the posterior state estimate of the system:
[0139] (6)
[0140] (7)
[0141] (8)
[0142] in, The expectation of the posterior estimate; Expectation of prior estimates; The variance of the posterior estimate; The variance of the prior estimate; Kalman gain at time t.
[0143] The Kalman filter performs forward iterations through two steps (S242 and S243) to recursively estimate the state of the system while minimizing the estimation error, thereby obtaining an iterative solution to the degradation rate state estimation function.
[0144] Specifically, the function parameters include .
[0145] S250. Based on the iterative solution of the state estimation function of the degradation rate, the first expected value, the second expected value, and the third expected value are obtained through the RTS smoothing algorithm; wherein, the first expected value is the expected value of the degradation rate at the current time in the i-th iteration; the second expected value is the expected value of the square of the degradation rate at the current time in the i-th iteration; and the third expected value is the expected value of the product of the degradation rate at the current time in the i-th iteration and the degradation rate at the previous time.
[0146] As one implementation method, S250 includes the following steps:
[0147] S251, Backward Iterative RTS Smoothing Algorithm:
[0148] (9)
[0149] (10)
[0150] (11)
[0151] S252. Initialize smooth covariance:
[0152] (12)
[0153] S253, Backward Iterative Calculation of Smoothing Covariance:
[0154] (13)
[0155] S254. Calculate the expected value for each condition:
[0156] (14)
[0157] (15)
[0158] (16)
[0159] in, Smooth gain at time step; At this moment The estimated smooth expectation; The estimated smoothed variance; Smoothed covariance at time points; first expected value In the i-th iteration The expected value; the second expected value In the i-th iteration The expected value; the third expected value The expected value.
[0160] S260. The first expected value, the second expected value, and the third expected value are used as the expected value of the E-step of the expected maximum algorithm. The updated values of the initialization parameters of the Kalman filter are obtained through the M-step operation of the expected maximum algorithm. The initialization parameters of the Kalman filter are updated according to the updated values. The initialization parameters of the Kalman filter include the expected value of the initial degenerate state, the variance of the initial degenerate state, the system process noise, and the diffusion coefficient.
[0161] Specifically, the M-step operation includes:
[0162] (17)
[0163] (18)
[0164] (19)
[0165] (20)
[0166] In the formula: .in, Let be the expectation of the initial degenerate state; where, The estimated values of the parameters are used to initialize the i-th iteration, which are the expectation of the initial degenerate state, the variance of the initial degenerate state, and the system process noise, respectively.
[0167] S270. Determine whether the convergence condition is met. If the convergence condition is met, execute S280. If the convergence condition is not met, execute S240.
[0168] Specifically, the convergence of the first expected value, the second expected value, and the third expected value is calculated. If the first expected value, the second expected value, and the third expected value converge, or if the number of iterations reaches the prediction number, the convergence condition is determined to be met; otherwise, the convergence condition is determined not to be met.
[0169] S280, repeat S240, and use the iterative solution of the obtained function parameters as the optimal estimate of the function parameters.
[0170] Specifically, once the convergence condition is met, S240 is repeated to obtain the optimal estimate of the function parameters based on the updated Kalman filter.
[0171] S290. Obtain the optimal estimate of the degradation rate state estimation function based on the optimal estimate of the function parameters.
[0172] Specifically, substitute the expected value and variance of the posterior estimate in the optimal estimation of function parameters into... Obtain the optimal estimate of the degradation rate state estimation function.
[0173] S2100. Based on the optimal solution of the state estimation function of degradation characteristics and degradation rate, the probability density function of the remaining life of the component is obtained by simulating the Wiener process. The probability density function is used to predict the remaining life of the component.
[0174] Specifically, by fitting the Wiener process, the probability density function of the remaining lifetime is obtained as follows:
[0175] (twenty one)
[0176] in, Let be the probability density function of the remaining lifespan of the component. Here, represents the damaged area data of the component, and w represents the failure threshold of the component. The diffusion coefficient is... This is the optimal estimate of the degradation rate state estimation function. Let k be the time length up to time k.
[0177] The method for predicting the remaining life of oil and gas drilling equipment provided in this application has the following technical effects:
[0178] This method uses pitting area as a health status indicator based on image data. Compared with existing health management methods that rely on indirect wear effect data, it has smaller errors and higher accuracy, sensitivity and identification ability for system health status. It can effectively support the prediction of the remaining life of oil and gas drilling equipment components.
[0179] Image-based semantic segmentation technology can intuitively show the evolution process of pitting defects. The method in this application is combined with machine vision to obtain degradation features, making the degradation features more consistent with the evolution law of defects, and the measurement of data and the evaluation of remaining lifetime have better interpretability.
[0180] Existing methods that rely on indirect wear effect data are often limited by specific equipment, operating conditions, and environments, and have significant limitations when applied to oil and gas drilling environments. In contrast, this method uses image detection technology based on direct wear effects, which avoids the shortcomings of existing methods and has greater versatility.
[0181] The EM expectation estimation algorithm can adaptively estimate the optimal model parameters. By fitting the Wiener process, the prediction is more scientific and accurate, avoiding the subjectivity of manual parameter setting.
[0182] Figure 4 A flowchart illustrating a method for predicting the remaining life of oil and gas drilling equipment provided in this application embodiment. Figure 4 ,like Figure 4 As shown, according to the method provided in the embodiments of this application, the life of a ball screw is predicted. The method includes:
[0183] S410: Collect image data of the ball screw at historical moments, and obtain the pitting area of the image data based on the trained UNet segmentation network model.
[0184] Specifically, the camera system is mounted on the ball screw nut. When the spindle rotates, the camera system moves linearly with the nut to observe the raceway surface. It automatically triggers a surface recording every certain period of time, obtaining a total of 394 labeled data images.
[0185] Specifically, the collected image dataset was divided into training, validation, and test sets in a 7:1.5:1.5 ratio, and all image samples were cropped to a uniform 448×1088 pixel size. Image enhancement processing was applied to the divided training set images, including contrast enhancement, tone transformation, grayscale conversion, Gaussian noise reduction, and image flipping. The application effects are shown below. Figure 5 As shown, where Figure 5 This is a schematic diagram illustrating the enhancement effect on defective image data.
[0186] Furthermore, the network structure is optimized based on the original UNet model, such as... Figure 6 This is a schematic diagram of the structure of the improved image segmentation model.
[0187] Specifically, in the backbone network of the original U-Net network model, a lightweight VGG16 is used as a feature extractor to reduce overfitting.
[0188] Specifically, a Dropout layer with a dropout rate of 0.3 is added to each upsampling process of the original UNet decoder, and a feature compensation path is established after each transposed convolutional layer using a residual connection mechanism to avoid excessive information loss during feature fusion.
[0189] Specifically, an additional loss function is added to the network. An edge loss function is constructed based on the structural similarity function, employing a local window calculation paradigm. The input image is divided into 8×8 image blocks. A soft constraint mechanism is built by measuring the structural similarity of defect edge regions to reduce the interference of small edge noise on the model. The formula for the edge loss function can be expressed as:
[0190]
[0191] Where k represents the total number of image blocks, and and represent the local mean of the i-th image block in the predicted mask and the ground truth mask, respectively. and represent the standard deviation of the i-th image block in the predicted mask and the ground truth mask, respectively. represents the covariance of the i-th image block between the predicted mask and the ground truth mask, and and are numerical stability constants used to prevent the denominator from being zero. When the pixel categories of the two mask positions corresponding to a certain image block are completely consistent, the value is 0, indicating that the image block does not contain edge regions, and the calculation result of the image block position is not included in the loss sum; when the pixel categories of the two mask positions corresponding to an image block are not completely consistent, the value will be between (0, 0.5], indicating that the image block contains edge regions. EC Loss uses this method to detect defect edge regions.
[0192] Furthermore, the pitting area of the image data is obtained based on the trained improved UNet segmentation network model.
[0193] in, Figure 7 The graph shows the variation of the training loss value for the improved image segmentation model. Figure 8 A comparison of the segmentation results of the improved image segmentation model.
[0194] Specifically, the improved UNet segmentation network model, after training, performs image segmentation to obtain pitting images, such as... Figure 9 The image shows a pitted image segmented based on an improved image segmentation model that has been trained.
[0195] Furthermore, the area data of the pitting image is extracted as the pitting area.
[0196] Specifically, the extracted pitting area data is represented as follows: ,in The area value of pitting erosion at a specific time.
[0197] S420, through the combined pitting area A polynomial feature is constructed, and the pitting area data of the nonlinear feature is transformed into an approximate linear feature through the degradation feature formula to obtain the degradation feature.
[0198] Specifically, the degradation characteristic formula can be expressed as:
[0199] (twenty two)
[0200] Referring to formula (1), under this working condition, the constant is... The value is 1 / 10, and the characteristic coefficient is... The value is 3 / 5.
[0201] Based on the above degradation feature formula, the degradation feature data obtained after feature extraction is represented as follows: .
[0202] like Figure 10 This is a comparison chart of the original defect area data and the degradation feature data obtained according to the embodiments of this application. Figure 10 The comparison shows that the features used in this method have stronger linearity.
[0203] S430. Establish a linear degradation model based on degradation characteristics. The linear degradation model includes a system equation based on degradation rate and a state equation based on degradation characteristics.
[0204] Specifically, a linear degradation model is established based on formulas (2) and (3).
[0205] S440. Obtain the iterative solution of the degradation rate state estimation function by solving the linear degradation model using a Kalman filter; update the initialization parameters of the Kalman filter using the expectation-maximization algorithm combined with the RTS smoothing algorithm based on the iterative solution of the degradation rate state estimation function; return the steps of solving the linear degradation model using the Kalman filter based on the updated initialization parameters, until the expectation-maximization algorithm meets the convergence condition or reaches the number of iterations, and use the iterative solution of the degradation rate state estimation function as the optimal estimate of the degradation rate state estimation function.
[0206] Specifically, the implementation steps are similar to those in S240~S280, and will not be repeated here.
[0207] S450. By fitting the Wiener process, the probability density function of the remaining life of the ball screw is obtained as follows:
[0208] (twenty three)
[0209] in, Let be the probability density function of the remaining lifespan of the component. Here, represents the damaged area data of the component, and w represents the failure threshold of the component. The diffusion coefficient is... This is the optimal estimate of the degradation rate state estimation function. Let k be the time length up to time k.
[0210] in, Figure 11 The probability density function distribution diagram for predicting the remaining life of a ball screw; Figure 12 This is a graph showing the predicted remaining life of the ball screw and its 30% accuracy range.
[0211] The method for predicting the remaining life of oil and gas drilling equipment provided in this application embodiment is similar in principle and technical effect to the above method, and will not be described in detail here.
[0212] Figure 13 A schematic diagram of a device for predicting the remaining life of oil and gas drilling equipment provided in this application is shown below. Figure 13 As shown, the prediction device 1000 provided in this application embodiment includes:
[0213] The segmentation module 1010 is used to collect image data of the parts over a historical period and obtain the wear area of the image data based on the improved image segmentation model.
[0214] Feature module 1020 is used to obtain the degradation characteristics of parts over a historical period based on the wear area.
[0215] Model module 1030 is used to establish a linear degradation model based on degradation characteristics. The linear degradation model includes a system equation based on degradation rate and a state equation based on degradation characteristics.
[0216] Iteration module 1040 is used to obtain the optimal estimate of the degradation rate state estimation function by solving the linear degradation model through a Kalman filter that is iteratively updated based on the expectation-maximization algorithm.
[0217] The prediction module 1050 is used to obtain the probability density function of the remaining life of the component by simulating the Wiener process based on the optimal estimate of the degradation characteristics and degradation rate state estimation function. The probability density function is used to predict the remaining life of the component.
[0218] In one possible implementation, the iteration module 1040 is specifically used for:
[0219] The linear degradation model is solved by using a Kalman filter to obtain the iterative solution of the function parameters in the degradation rate state estimation function; where the function parameters include the expectation of the posterior estimate, the variance of the posterior estimate, and the Kalman gain.
[0220] The initial parameters of the Kalman filter are updated based on the iterative solution of the function parameters using the expectation-maximization algorithm combined with the RTS smoothing algorithm.
[0221] Based on the updated initialization parameters, return the steps for solving the linear degenerate model using the Kalman filter until the expectation-maximization algorithm meets the convergence condition or reaches the required number of iterations. Then, use the iterative solution of the function parameters as the optimal estimate of the function parameters.
[0222] The optimal estimate of the degradation rate state estimation function is obtained based on the optimal estimate of the function parameters.
[0223] In one possible implementation, the initialization parameters of the Kalman filter include the expectation of the initial degradation rate of the degradation state, the variance of the initial degradation state, the system process noise, and the diffusion coefficient; the iteration module 1040 is further used for:
[0224] The first expected value, the second expected value, and the third expected value are obtained by using the RTS smoothing algorithm based on the iterative solution of the state estimation function of the degradation rate; wherein, the first expected value is the expected value of the degradation rate at the current time in the i-th iteration; the second expected value is the expected value of the square of the degradation rate at the current time in the i-th iteration; and the third expected value is the expected value of the product of the degradation rate at the current time in the i-th iteration and the degradation rate at the previous time.
[0225] The first, second, and third expected values are used as the expectations of the E-step of the expectation-maximization algorithm. The updated expectations of the initial degraded state, the variance of the initial degraded state, the system process noise, and the diffusion coefficient are obtained through the M-step operation of the expectation-maximization algorithm.
[0226] In one possible implementation, the system equation based on the degradation rate is:
[0227]
[0228] in, For components The degradation rate at any given time; This is the drift coefficient;
[0229] The state equation based on the degradation characteristics is:
[0230]
[0231] Among them, y k The degradation characteristics at time k; The diffusion coefficient is denoted as . To comply with It follows a normal distribution.
[0232] In one possible implementation, feature module 1020 is specifically used for:
[0233] Degradation characteristics are obtained based on the wear area and historical time using a degradation characteristic formula; the degradation characteristic formula is:
[0234]
[0235] Among them, y k For the tth k The degradation characteristics of time, The wear area of the components at any given time. It is a constant. These are the characteristic coefficients.
[0236] In one possible implementation, the prediction module 1050 is further used for:
[0237] The probability density function of remaining lifetime can be obtained using the following formula:
[0238]
[0239] in, Let be the probability density function of the remaining lifespan of the component. Here, represents the damaged area data of the component, and w represents the failure threshold of the component. The diffusion coefficient is... This is the optimal estimate of the degradation rate state estimation function. Let k be the time length up to time k.
[0240] In one possible implementation, obtaining the wear area of the image data according to the improved image segmentation model includes: identifying the initial wear area of the image data according to the improved image segmentation model; dividing the initial wear area into an initial internal wear area and an initial edge wear area; processing the initial internal wear area according to a preset loss function to obtain an internal wear area; constructing an edge loss function according to a preset structural similarity function, and dividing the initial edge wear area into multiple edge wear image blocks; processing the multiple edge wear image blocks according to the edge loss function to obtain an edge wear area; and integrating the internal wear area and the edge wear area into the wear area of the image data.
[0241] In one possible implementation, the improvement process of the image segmentation model includes: acquiring an image sample dataset of the components, and performing image enhancement processing on multiple image samples in the image sample dataset to obtain multiple enhanced image samples; dividing the multiple enhanced image samples to obtain a training set and a test set; training a preset image segmentation model based on the training set to obtain an original image segmentation model; testing the original image segmentation model based on the test set to complete the testing of the original image segmentation model; extracting the backbone network and the original decoder of the original image segmentation model; extracting the feature extractor in the backbone network and configuring the feature extractor to obtain a configured feature extractor; establishing a feature compensation path process after each transposed convolutional layer in the original decoder according to a preset residual connection mechanism; and improving the original image segmentation model based on the configured feature extractor and the feature compensation path process to obtain an improved image segmentation model.
[0242] In one possible implementation, the plurality of edge-wearing image patches are processed according to the edge loss function to obtain a formula for calculating the edge-wearing area, including:
[0243]
[0244] In the formula, L is the loss value of the edge wear area; k is the number of image patches; A value of 0 indicates that the image does not contain the edge of the component, and the calculation result of the image position is not included in the loss and middle; and Let represent the local mean of the i-th image patch in the predicted mask and the real mask, respectively; and Let represent the standard deviations of the i-th image patch in the predicted mask and the true mask, respectively; This represents the covariance between the predicted mask and the true mask for the i-th image patch; and It is a numerical stability constant used to prevent the denominator from being zero.
[0245] This embodiment provides a device for predicting the remaining life of oil and gas drilling equipment. It can execute the method provided in the above-described method embodiment. Its implementation principle and technical effect are similar, and will not be described in detail here.
[0246] Figure 14 This is a schematic diagram of the structure of an electronic device provided in this application. Figure 14As shown, the electronic device 1100 provided in this embodiment includes at least one processor 1101 and a memory 1102. Optionally, the device 110 also includes a communication component 1103. The processor 1101, memory 1102, and communication component 1103 are connected via a bus 1104.
[0247] In a specific implementation, at least one processor 1101 executes computer execution instructions stored in memory 1102, causing at least one processor 1101 to perform the above-described method.
[0248] The specific implementation process of processor 1101 can be found in the above method embodiments, and its implementation principle and technical effect are similar. It will not be repeated here.
[0249] In the above embodiments, it should be understood that the processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor.
[0250] The memory may include random access memory (RAM) and may also include non-volatile memory (NVM), such as at least one disk storage device.
[0251] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of illustration, the buses shown in the accompanying drawings are not limited to a single bus or a single type of bus.
[0252] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.
[0253] This application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the above-described method.
[0254] The aforementioned readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The readable storage medium can be any available medium accessible to a general-purpose or special-purpose computer.
[0255] An exemplary readable storage medium is coupled to a processor, enabling the processor to read information from and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can reside in an Application Specific Integrated Circuit (ASIC). Alternatively, the processor and the readable storage medium can exist as discrete components in the device.
[0256] The division of units is merely a logical functional division; in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.
[0257] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.
[0258] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.
[0259] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0260] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.
[0261] Finally, it should be noted that other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This invention is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein, and is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of the invention is limited only by the appended claims.
Claims
1. A method for predicting the remaining lifespan of oil and gas drilling equipment, characterized in that, include: Collect image data of the components over a historical period, and obtain the wear area of the image data based on an improved image segmentation model; The degradation characteristics of the component over a historical period are obtained based on the wear area. A linear degradation model is established based on the degradation characteristics, and the linear degradation model includes a system equation based on the degradation rate and a state equation based on the degradation characteristics; The optimal estimate of the degradation rate state estimation function is obtained by solving the linear degradation model using a Kalman filter that is iteratively updated based on the expectation-maximization algorithm. Based on the degradation characteristics and the optimal estimate of the degradation rate state estimation function, the probability density function of the remaining life of the component is obtained by simulating the Wiener process. The probability density function is used to predict the remaining life of the component.
2. The prediction method according to claim 1, characterized in that, The step of obtaining the optimal estimate of the degradation rate state estimation function by solving the linear degradation model using a Kalman filter that is iteratively updated based on the expectation-maximization algorithm includes: The linear degradation model is solved by a Kalman filter to obtain the iterative solution of the function parameters in the degradation rate state estimation function; wherein, the function parameters include the expectation of the posterior estimate, the variance of the posterior estimate, and the Kalman gain; The initialization parameters of the Kalman filter are updated based on the iterative solution of the function parameters using the expectation-maximization algorithm combined with the RTS smoothing algorithm. The steps of solving the linear degenerate model using the Kalman filter are returned based on the updated initialization parameters until the expectation-maximization algorithm meets the convergence condition or reaches the number of iterations. Then, the iterative solution of the function parameters is taken as the optimal estimate of the function parameters. The optimal estimate of the degradation rate state estimation function is obtained based on the optimal estimate of the function parameters.
3. The prediction method according to claim 2, characterized in that, The initialization parameters of the Kalman filter include the expectation of the initial degraded state of the state estimation function of the degradation rate, the variance of the initial degraded state, the system process noise, and the diffusion coefficient; The step of updating the initial parameters of the Kalman filter using the expected maximum algorithm combined with the RTS smoothing algorithm based on the iterative solution of the function parameters includes: The first expected value, the second expected value, and the third expected value are obtained by using the RTS smoothing algorithm based on the iterative solution of the state estimation function of the degradation rate; wherein, the first expected value is the expected value of the degradation rate at the current time in the i-th iteration; the second expected value is the expected value of the square of the degradation rate at the current time in the i-th iteration; and the third expected value is the expected value of the product of the degradation rate at the current time in the i-th iteration and the degradation rate at the previous time. The first, second, and third expected values are used as the expectations of the E-step of the expectation-maximization algorithm. The updated expectations of the initial degraded state, the variance of the initial degraded state, the system process noise, and the diffusion coefficient are obtained through the M-step operation of the expectation-maximization algorithm.
4. The prediction method according to claim 1, characterized in that, The system equation based on the degradation rate is as follows: in, For components The degradation rate at any given time; This is the drift coefficient; The state equation based on the degradation characteristics is: Among them, y k The degradation characteristics at time k; The diffusion coefficient is denoted as . To comply with It follows a normal distribution.
5. The prediction method according to claim 1, characterized in that, The method of obtaining the degradation characteristics of components over a historical period based on the wear area includes: The degradation characteristics are obtained based on the wear area and the historical time using a degradation characteristic formula; the degradation characteristic formula is: Among them, y k For the tth k The degradation characteristics of time, The wear area of the components at any given time. It is a constant. These are the characteristic coefficients.
6. The prediction method according to any one of claims 1-5, characterized in that, The process of obtaining the probability density function of the remaining life of the component through a Wiener process simulation, based on the optimal estimate of the degradation characteristics and the degradation rate state estimation function, includes: The probability density function of the remaining lifetime is obtained according to the following formula: in, Let be the probability density function of the remaining lifespan of the component. Here, represents the damaged area data of the component, and w represents the failure threshold of the component. The diffusion coefficient is... This is the optimal estimate of the degradation rate state estimation function. Let k be the time length up to time k.
7. The prediction method according to claim 1, characterized in that, The step of obtaining the wear area of the image data according to the improved image segmentation model includes: The initial wear area of the image data is identified based on the improved image segmentation model; The initial wear area is divided into an initial internal wear area and an initial edge wear area; The initial internal wear area is processed according to a preset loss function to obtain the internal wear area; An edge loss function is constructed based on a preset structural similarity function, and the initial edge wear area is divided into multiple edge wear image blocks; The edge wear image blocks are processed according to the edge loss function to obtain the edge wear area; The internal wear area and the edge wear area are integrated into the wear area of the image data.
8. The prediction method according to claim 1, characterized in that, The improvement process of the image segmentation model includes: Obtain an image sample dataset of the component, and perform image enhancement processing on multiple image samples in the image sample dataset to obtain multiple enhanced image samples; The multiple enhanced image samples are divided to obtain a training set and a test set; The preset image segmentation model is trained based on the training set to obtain the original image segmentation model; The original image segmentation model is tested according to the test set to complete the test of the original image segmentation model; Extract the backbone network and original decoder of the original image segmentation model; Extract the feature extractors from the backbone network and configure the feature extractors to obtain the configured feature extractors; After each transposed convolutional layer in the original decoder, a feature compensation path process is established according to a preset residual connection mechanism; The original image segmentation model is improved based on the configured feature extractor and the feature compensation path process to obtain an improved image segmentation model.
9. The prediction method according to claim 7, characterized in that, The step of processing the plurality of edge-wearing image patches according to the edge loss function to obtain the calculation formula for the edge-wearing area includes: In the formula, L is the loss value of the edge wear area; k is the number of image patches; A value of 0 indicates that the image does not contain the edge of the component, and the calculation result of the image position is not included in the loss and middle; and Let represent the local mean of the i-th image patch in the predicted mask and the real mask, respectively; and Let represent the standard deviations of the i-th image patch in the predicted mask and the true mask, respectively; This represents the covariance between the predicted mask and the true mask for the i-th image patch; and It is a numerical stability constant used to prevent the denominator from being zero.
10. A device for predicting the remaining life of oil and gas drilling equipment, characterized in that, include: The segmentation module is used to collect image data of the parts over a historical period and obtain the wear area of the image data based on the improved image segmentation model; The feature module is used to obtain the degradation characteristics of parts over a historical period based on the wear area; The model module is used to establish a linear degradation model based on degradation characteristics. The linear degradation model includes a system equation based on the degradation rate and a state equation based on the degradation characteristics. The iterative module is used to solve the linear degradation model by iteratively updating the Kalman filter based on the expectation-maximization algorithm to obtain the optimal solution of the degradation rate state estimation function and the optimal estimate of the degradation rate state estimation function. The prediction module is used to obtain the probability density function of the remaining life of the component by simulating the Wiener process, based on the optimal solution of the degradation rate state estimation function and the degradation characteristics. The probability density function is used to predict the remaining life of the component.
11. An electronic device, characterized in that, include: Memory, processor; The memory stores the instructions that the computer executes; The processor executes computer execution instructions stored in memory, causing the processor to perform the method as claimed in any one of claims 1-9.