Weibull statistics-based internal water ballast load optimization method and system

Abstract

The application discloses a kind of inner water pressure load optimization method and system based on Weibull statistics, belong to pressure test technical field, method includes: collecting the inner water pressure load monitoring data of each small porcelain piece area, the feature extraction of porcelain piece image data is carried out, and the porcelain piece feature data is obtained, and the distribution parameter of Weibull statistics is fitted, and the porcelain piece inner water pressure load screening interval is deduced back, by multiple linear regression algorithm and random forest algorithm, relevant model is constructed, and the porcelain piece inner water pressure load screening interval is calibrated, and then the porcelain piece optimal inner water pressure load is screened out, realize the optimization based on the inner water pressure load.This application accurately depicts porcelain piece inner water pressure intensity distribution law by fusing Weibull statistics, multidimensional data acquisition and multiple intelligent algorithms, adapts to different environmental conditions, effectively avoids the problem of missing detection or damage qualified piece of traditional method, significantly improves the intelligence and accuracy of inner water pressure load optimization.

Description

Technical Field

[0001] This invention belongs to the field of pressure testing technology, specifically relating to a method and system for optimizing internal water pressure load based on Weibull statistics. Background Technology

[0002] In extremely cold environments with large temperature differences, disc suspension porcelain insulators frequently fail due to internal insulation failure, seriously threatening the safety of ultra-high voltage power transmission projects. As a brittle material, porcelain components have randomly distributed internal defects, resulting in significant dispersion in their failure strength. Traditional internal hydrostatic tests rely on experience to set fixed loads, which can easily lead to problems such as underestimating defects due to excessively low loads or damaging qualified components due to excessively high loads. Low-temperature environments amplify the stress concentration effect of minute defects, increasing the risk of failure of potentially defective components under temperature-induced stress. Similarly, the response mechanisms of internal defects in materials also differ in high-temperature, high-humidity, or high-temperature cycling environments.

[0003] Traditional testing methods struggle to accurately quantify loads under varying environmental conditions, leading to blind testing and insufficient reliability. Furthermore, traditional internal water pressure load optimization primarily addresses loads generated by the liquid pressure within a closed structure, relying heavily on single-device monitoring. This inherent risk makes it difficult to effectively optimize internal water pressure loads on closed structures under different environments, resulting in high costs and low optimization accuracy. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for optimizing internal water pressure load based on Weibull statistics. By integrating Weibull statistics, multi-dimensional data acquisition and various intelligent algorithms, it accurately depicts the distribution law of internal water pressure intensity in ceramic parts, adapts to different environmental conditions, effectively avoids the problem of missed detection or damage to qualified parts in traditional methods, and significantly improves the intelligence and accuracy of internal water pressure load optimization.

[0005] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows:

[0006] An optimization method for internal water pressure load based on Weibull statistics includes the following steps:

[0007] S1. Divide the surface of the target ceramic part into several small ceramic parts regions with the same area, and collect the internal water pressure load monitoring data of each small ceramic part region, including internal water pressure strength test data, ceramic part image data, ceramic part physical property data and test environment data.

[0008] S2. Extract features from the porcelain image data to obtain porcelain feature data;

[0009] S3. Referring to the test data of internal water pressure resistance and the characteristic data of ceramic parts, the maximum likelihood estimation method is used to fit the distribution parameters of Weibull statistics to obtain the Weibull statistical distribution model.

[0010] S4. Set the failure probability threshold for ceramic parts, and deduce the water pressure load screening range inside the ceramic parts based on the Weibull statistical distribution model.

[0011] S5. Construct a regression model for the physical properties of ceramic parts using physical property data and a multiple linear regression algorithm, and output the regression coefficients for the physical properties of ceramic parts.

[0012] S6. Using the random forest algorithm, a calibration model for the internal water pressure load of ceramic parts is constructed based on the regression coefficients of the physical properties of the ceramic parts and the test environment data. The internal water pressure load calibration coefficient is output to calibrate the internal water pressure load screening range of the ceramic parts.

[0013] S7. Based on the calibrated internal water pressure load screening range of the ceramic parts, the optimal internal water pressure load of the ceramic parts is selected through particle swarm optimization algorithm.

[0014] S8. Using the calibrated internal water pressure load screening range and the optimal internal water pressure load of the ceramic part as references, optimize the internal water pressure load of the ceramic part and generate the final output report of the internal water pressure load of the ceramic part.

[0015] Preferably, in S1, the process of collecting monitoring data on the internal water pressure load of each small ceramic component area includes:

[0016] Assign numbers to each small ceramic component area, deploy different types of data acquisition equipment, and collect internal water pressure load monitoring data for each small ceramic component area;

[0017] The data from the internal water pressure resistance test include the burst water pressure value of each small ceramic part area in each test ceramic part, the pressure drop value when the ceramic part bursts, the inner diameter of the main body, the outer diameter of the main body, the effective height of the main body, the effective length of the main body, the wall thickness, the inner diameter of the flange, the flange thickness, the diameter of the flange sealing surface, the diameter of the center circle of the bolt hole, the arc transition radius between the main body and the flange, the arc transition radius between the main body and the end cap, the radius of curvature of the end cap, and the thickness of the end cap.

[0018] The ceramic image data consists of real-time surface images of each experimental ceramic piece.

[0019] The physical property data of the ceramic parts include the internal water pressure load, hardness, temperature, humidity, compressive strength and elastic modulus of each small ceramic part area in each ceramic part;

[0020] The test environment data includes the real-time steady-state low temperature value and its holding time, the real-time steady-state high temperature value and its holding time, the number of steady-state high and low temperature difference cycles, the real-time relative humidity, the real-time concentration of corrosive media, the cumulative contact time with corrosive media, the real-time external vibration frequency, the real-time external vibration acceleration, the real-time solar radiation intensity, and the cumulative ultraviolet radiation dose.

[0021] Image enhancement and denoising were performed on the ceramic image data. Data cleaning and standardization were performed on the water pressure strength test data, ceramic physical property data, and test environment data. The pre-processed internal water pressure load monitoring data were integrated to generate the Weibull statistical data set.

[0022] Preferably, in S2, the process of acquiring the ceramic feature data includes:

[0023] The ceramic feature data is the crack length of each small ceramic region in each experimental ceramic piece. Feature extraction is performed on the real-time surface image of each experimental ceramic piece. Based on the semantic segmentation technology of deep learning, the small ceramic regions in each experimental ceramic piece are identified in the real-time surface image of each experimental ceramic piece.

[0024] A U-shaped convolutional neural network was used to segment the crack regions in each small ceramic region of each identified test ceramic piece, and the cross-entropy loss function was used to optimize the crack region segmentation results.

[0025] The skeleton of the crack region segmentation result is extracted. Combined with the Zhang-Suen thinning algorithm, the crack region segmentation result is reduced to a skeleton line with a width of one pixel, while preserving the topological structure of the crack. The skeleton line with a width of one pixel is traversed at the pixel level, and the distance between adjacent skeleton pixels is calculated. The distances of all adjacent pixels are accumulated to obtain the total pixel length of the crack in each small ceramic part region of each test ceramic piece. According to the image scale, the total pixel length of the crack in each small ceramic part region of each test ceramic piece is converted into the actual physical length to obtain the crack length of each small ceramic part region of each test ceramic piece.

[0026] Preferably, in S3, the process of using the maximum likelihood estimation method to fit the distribution parameters of the Weibull statistics and obtain the Weibull statistical distribution model includes:

[0027] The distribution parameters of Weibull statistics are the shape parameters, scale parameters, and location parameters of the Weibull statistical distribution model.

[0028] The maximum likelihood estimation method is adopted. Based on the probability density function of the Weibull statistical distribution, the probability density function expression of the Weibull statistical distribution of each experimental porcelain piece is obtained, the likelihood function is fitted, and the natural logarithm of the likelihood function is taken to obtain the log-likelihood function.

[0029] By maximizing the log-likelihood function, the estimated values ​​of the shape, scale, and location parameters of the Weibull statistical distribution model are obtained. The estimated values ​​of the shape, scale, and location parameters of the Weibull statistical distribution model are then substituted into the Weibull statistical distribution model to obtain the Weibull statistical distribution model.

[0030] The goodness-of-fit test method was used to verify the fit between the internal water pressure strength test data, ceramic component characteristic data and the Weibull statistical distribution model. The shape parameters, scale parameters and position parameters of the Weibull statistical distribution model were further optimized to obtain the final Weibull statistical distribution model.

[0031] Preferably, in S4, the process of deducing the water pressure load screening range inside the ceramic part based on the Weibull statistical distribution model includes:

[0032] The failure probability threshold for ceramic parts includes the defect removal threshold and the qualified protection threshold for each small ceramic part area within each ceramic part.

[0033] Set a failure probability threshold for ceramic parts, substitute the failure probability threshold for ceramic parts into the Weibull statistical distribution model, and deduce the internal water pressure load of each small ceramic part area in each ceramic part corresponding to different failure probability thresholds for ceramic parts.

[0034] The internal water pressure load of each small ceramic part area corresponding to the defect rejection threshold is used as the lower limit of the internal water pressure load screening range of the ceramic part, and the internal water pressure load of each small ceramic part area corresponding to the qualified protection threshold is used as the upper limit of the internal water pressure load screening range of the ceramic part, thus obtaining the internal water pressure load screening range of the ceramic part.

[0035] Preferably, in S5, the process of constructing a regression model for the physical properties of ceramic parts and outputting the regression coefficients for the physical properties of ceramic parts includes:

[0036] The regression coefficients of the physical properties of ceramic parts include the regression coefficients of hardness, temperature, humidity, compressive strength and elastic modulus of each small ceramic part area in each ceramic part;

[0037] Extract the physical property data of ceramic pieces from the Weibull statistical data set and transform them into a linear training set and a linear test set;

[0038] The multiple linear regression algorithm is adopted. The hardness, temperature, humidity, compressive strength and elastic modulus of each small ceramic region in each ceramic piece in the linear training set are used as inputs, and the internal water pressure load of each small ceramic region in each ceramic piece in the linear training set is used as output. The linear relationship between the hardness, temperature, humidity, compressive strength and elastic modulus of each small ceramic region in each ceramic piece and the corresponding internal water pressure load is learned respectively, and the regression model of the physical properties of the ceramic piece is obtained by training.

[0039] The linear test set data is input into the regression model of the physical properties of ceramic parts. An adaptive moment estimator optimizer is used to adjust the regression coefficients and intercept terms of the regression model of the physical properties of ceramic parts, optimize the performance of the regression model of the physical properties of ceramic parts, obtain the final regression model of the physical properties of ceramic parts, and then obtain the regression coefficients of the physical properties of ceramic parts. The regression coefficients of the physical properties of ceramic parts are then integrated into the Weibull statistical data set.

[0040] Preferably, in step S6, the process of constructing a calibration model for the internal water pressure load of the ceramic part, outputting the calibration coefficient for the internal water pressure load, and calibrating the screening range for the internal water pressure load of the ceramic part includes:

[0041] The regression coefficients of the physical properties of ceramic pieces and the test environment data were extracted from the Weibull statistical data set and divided into a nonlinear training set and a nonlinear test set.

[0042] Using the random forest algorithm, the nonlinear training set data is set as input and the internal water pressure load calibration coefficient is set as output. The nonlinear relationship between the regression coefficient of the physical properties of the ceramic part, the test environment data and the internal water pressure load calibration coefficient is learned, and the internal water pressure load calibration model of the ceramic part is trained and obtained.

[0043] The nonlinear training set data is input into the calibration model of internal water pressure load of ceramic parts obtained through training. The stochastic gradient descent optimizer is used to adjust the parameters of the calibration model of internal water pressure load of ceramic parts, optimize the calibration model of internal water pressure load of ceramic parts, and obtain the final calibration model of internal water pressure load of ceramic parts.

[0044] Combining the regression coefficients of the physical properties of the ceramic parts and the test environment data, the corresponding internal water pressure load calibration coefficients are output. Based on the output internal water pressure load calibration coefficients, the internal water pressure load screening range of the ceramic parts is calibrated.

[0045] Preferably, in S7, the screening process for the optimal internal water pressure load on the ceramic component includes:

[0046] Based on the internal water pressure load of the ceramic part within the calibrated internal water pressure load screening range, an objective function is defined and a maximum number of iterations is set. The particle swarm optimization algorithm is adopted to initialize the particle swarm and evaluate each particle in the particle swarm through the objective function.

[0047] The position and velocity of each particle in the particle swarm are updated to guide the selection process of the optimal internal water pressure load for the ceramic part. After each update and optimization, the fitness of each particle in the particle swarm is continuously compared iteratively to update the global optimal solution. When the number of iterations reaches the maximum number of iterations, the update process of the particle swarm is stopped, and the optimal internal water pressure load for the ceramic part is output through the particle swarm optimization algorithm.

[0048] Preferably, in S8, the process of optimizing the water pressure load inside the ceramic component and generating the final output report of the water pressure load inside the ceramic component includes:

[0049] The distribution of internal water pressure load in each small ceramic part area of ​​each actual ceramic part is analyzed within the calibrated internal water pressure load screening range. The qualification level of the internal water pressure load of each actual ceramic part is evaluated and corresponding operations are performed. The qualification level of the internal water pressure load of each ceramic part and the ceramic part operation record are obtained to achieve optimization of the internal water pressure load of the ceramic part.

[0050] By using report generation technology, the water pressure load qualification level of each ceramic component and the operation record of the ceramic component are integrated to generate the final output report of water pressure load in the ceramic component.

[0051] An internal water pressure load optimization system based on Weibull statistics, used to implement the above method, includes:

[0052] The data acquisition module is used to collect monitoring data on internal water pressure load in various small ceramic parts, including internal water pressure strength test data, ceramic part image data, ceramic part physical property data, and test environment data.

[0053] The feature extraction module extracts features from the porcelain image data to obtain the porcelain feature data;

[0054] The distribution application module is used to fit the distribution parameters of the Weibull statistics to obtain the Weibull statistical distribution model, and then obtain the screening range of water pressure load inside the ceramic part. It is divided into the following units:

[0055] The distribution fitting unit, referencing the test data of internal water pressure resistance and the characteristic data of ceramic parts, uses the maximum likelihood estimation method to fit the distribution parameters of Weibull statistics, and obtains the Weibull statistical distribution model;

[0056] The interval is defined as a unit, the failure probability threshold of the ceramic component is set, and the water pressure load screening interval inside the ceramic component is derived based on the Weibull statistical distribution model.

[0057] The load calibration and screening module is used to calibrate the internal water pressure load screening range of ceramic parts and screen out the optimal internal water pressure load for ceramic parts. It is divided into the following units:

[0058] The interval calibration unit constructs a regression model of the physical properties of ceramic parts using physical property data and a multiple linear regression algorithm, outputs regression coefficients of the physical properties of ceramic parts, and constructs a calibration model of internal water pressure load of ceramic parts using experimental environment data and a random forest algorithm, outputs calibration coefficients of internal water pressure load, and calibrates the selected interval of internal water pressure load of ceramic parts.

[0059] The load screening unit, based on the calibrated internal water pressure load screening range of the ceramic part, uses a particle swarm optimization algorithm to screen out the optimal internal water pressure load of the ceramic part.

[0060] The output module is optimized by using the calibrated internal water pressure load screening range and the optimal internal water pressure load of the ceramic part as references to optimize the internal water pressure load of the ceramic part and generate the final output report of the internal water pressure load of the ceramic part.

[0061] The beneficial effects of this invention are:

[0062] This invention, through the deep integration of refined data acquisition and the Weibull statistical model, effectively solves the drawbacks of traditional internal water pressure tests that rely on empirically set fixed loads. By fitting the Weibull statistical distribution parameters using the maximum likelihood estimation method and combining the goodness-of-fit test to optimize the model, it accurately characterizes the strength dispersion of brittle ceramic parts, and inversely derives a load screening range based on failure probability thresholds (defect rejection, qualification protection). This avoids both defect omissions caused by excessively low loads and damage to qualified ceramic parts caused by excessively high loads. It is particularly suitable for risk management of stress concentration in micro-defects under extreme environments such as extreme cold and high temperatures, improving the scientific rigor and safety of internal water pressure load assessment for ceramic parts.

[0063] This invention relies on multi-algorithm collaborative calibration and optimization to further improve the accuracy and intelligence level of internal water pressure load optimization, forming a practical application closed loop. A regression model of the physical properties of ceramic components is constructed using a multiple linear regression algorithm to quantify the linear relationship between physical parameters such as hardness and compressive strength and internal water pressure load, outputting accurate regression coefficients. Then, combined with experimental environment data, a calibration model is constructed using a random forest algorithm, outputting calibration coefficients to eliminate the interference of environmental factors (such as temperature changes and vibrations) on the load range, ensuring that the range is suitable for different application scenarios. Finally, based on the calibrated range, the optimal internal water pressure load is efficiently selected using a particle swarm optimization algorithm. Simultaneously, the actual load distribution is analyzed to assess the qualification level of the ceramic components and generate output reports. This achieves full-process intelligentization from data acquisition, model construction, range calibration to optimal load selection and result output, providing practical technical support for the quality control and reliability assurance of brittle ceramic components such as ceramic insulators in ultra-high voltage power transmission projects. Attached Figure Description

[0064] Figure 1 This is a schematic flowchart of the method of the present invention;

[0065] Figure 2 This is a schematic diagram of the modules of the system of the present invention. Detailed Implementation

[0066] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings.

[0067] Example 1: As Figure 1 As shown, a method for optimizing internal water pressure load based on Weibull statistics includes the following steps:

[0068] S1. Divide the surface of the target ceramic part into several small ceramic parts regions of the same area, and collect the internal water pressure load monitoring data of each small ceramic part region, including internal water pressure strength test data, ceramic part image data, ceramic part physical property data and test environment data, so as to achieve uniform coverage of the internal water pressure load monitoring range and complete data dimension collection, laying a solid data foundation for subsequent analysis steps.

[0069] S2. Extract features from the porcelain image data to obtain porcelain feature data, which builds a key bridge for the implementation of subsequent steps.

[0070] S3. Based on the test data of internal water pressure resistance and the characteristic data of ceramic parts, the maximum likelihood estimation method is used to fit the distribution parameters of Weibull statistics to obtain the Weibull statistical distribution model, which provides technical support for the scientific and accurate analysis of internal water pressure loads in ceramic parts.

[0071] S4. Set the failure probability threshold for ceramic parts, and based on the Weibull statistical distribution model, deduce the screening range of water pressure load inside the ceramic parts, clarify the boundary range of water pressure load inside the ceramic parts that meets the requirements, and lay the foundation for subsequent screening of water pressure load inside the ceramic parts.

[0072] S5. By using the physical property data of ceramic parts and the multiple linear regression algorithm, a regression model of the physical properties of ceramic parts is constructed, and the regression coefficients of the physical properties of ceramic parts are output. The linear relationship between the actual water pressure load inside the ceramic parts and various physical property data of the ceramic parts is clearly quantified, providing a data basis for constructing a calibration model of water pressure load inside the ceramic parts.

[0073] S6. Using the random forest algorithm, a calibration model for the internal water pressure load of ceramic parts is constructed by using the regression coefficients of the physical properties of ceramic parts and the test environment data. The calibration coefficients of the internal water pressure load are output to calibrate the screening range of the internal water pressure load of ceramic parts, which greatly improves the accuracy of the screening range of the internal water pressure load of ceramic parts. Environmental factors are included in the consideration of the internal water pressure load of ceramic parts, which improves the robustness of the screening range of the internal water pressure load of ceramic parts.

[0074] S7. Based on the calibrated internal water pressure load screening range of the ceramic parts, the optimal internal water pressure load of the ceramic parts is screened out through the particle swarm optimization algorithm, and the optimal internal water pressure load of the ceramic parts that takes into account both safety and stability is quickly located, so as to achieve efficient optimization of load configuration.

[0075] S8. Using the calibrated internal water pressure load screening range and the optimal internal water pressure load of the ceramic part as references, optimize the internal water pressure load of the ceramic part and generate the final output report of the internal water pressure load of the ceramic part, forming an optimization scheme for the internal water pressure load of the ceramic part that takes into account both scientificity and practicality, providing technical support for the production control and practical application of ceramic parts.

[0076] In S1, the data acquisition process for monitoring the internal water pressure load in each small ceramic component area includes:

[0077] The surface of the target ceramic part is divided into several small ceramic parts regions of the same area, and each small ceramic part region is assigned a number.

[0078] Different types of data acquisition equipment are deployed to collect monitoring data on the internal water pressure load of various small ceramic parts. The data acquisition equipment includes water pressure sensors, pressure sensors, high-speed data acquisition instruments, laser rangefinders, internal diameter measuring instruments, high-resolution digital cameras, microhardness testers, miniature patch temperature sensors, miniature patch humidity sensors, miniature pressure testing machines, force sensors, dynamic elastic modulus testers, low-temperature sensors, high-temperature sensors, temperature recorders, gas sensors, liquid concentration sensors, corrosion resistance test chambers, time counters, vibration sensors, acceleration sensors, light intensity sensors, and ultraviolet radiation meters.

[0079] The internal water pressure resistance test data includes the burst water pressure value of each small ceramic part in each test ceramic piece, the pressure drop value when the ceramic piece bursts, the main body inner diameter, main body outer diameter, main body effective height, main body effective length, wall thickness, flange inner diameter, flange thickness, flange sealing surface diameter, bolt hole center circle diameter, arc transition radius between the main body and the flange, arc transition radius between the main body and the end cap, curvature radius of the end cap, and end cap thickness; ceramic piece image data includes real-time surface images of each test ceramic piece; ceramic piece physical property data includes the internal water pressure load, hardness, temperature, humidity, compressive strength, and elastic modulus of each small ceramic part in each ceramic piece; test environment data includes the real-time steady-state low temperature value and its holding time, the real-time steady-state high temperature value and its holding time, the number of steady-state high and low temperature difference cycles, real-time relative humidity, real-time concentration of corrosive media, cumulative contact time with corrosive media, real-time external vibration frequency, real-time external vibration acceleration, real-time solar radiation intensity, and cumulative ultraviolet radiation dose of each small ceramic part in each ceramic piece;

[0080] Several test ceramic pieces were selected and subjected to internal water pressure strength tests. A water pressure sensor was used to collect the rupture water pressure values ​​of each small ceramic part area within each test ceramic piece. A pressure sensor and a high-speed data acquisition instrument were used to collect the pressure drop values ​​at the moment of rupture of each small ceramic part area within each test ceramic piece. A laser rangefinder was used to collect the main body inner diameter, main body outer diameter, main body effective height, main body effective length, wall thickness, flange thickness, and end cap thickness of each test ceramic piece. An inner diameter measuring instrument was used to collect the flange inner diameter and flange sealing surface diameter of each test ceramic piece. An optical surface profilometer was used to collect the bolt hole center circle diameter, the arc transition radius between the main body and the flange, the arc transition radius between the main body and the end cap, and the curvature radius of the end cap of each test ceramic piece.

[0081] Real-time surface images of each test ceramic piece were acquired using a high-resolution digital camera; the internal water pressure load of each small ceramic area within each piece was acquired using a water pressure sensor; the hardness, temperature, humidity, and elastic modulus of each small ceramic area within each piece were acquired using a microhardness tester, a micro-patch temperature sensor, a micro-patch humidity sensor, and a dynamic elastic modulus tester; and the compressive strength of each small ceramic area within each piece was acquired using a micro-compression testing machine and a force sensor. The internal water pressure load of each small ceramic area within each piece encompasses the internal water pressure load of each small ceramic area within each test ceramic piece.

[0082] Using a low-temperature sensor and a high-speed data acquisition instrument, the real-time steady-state low-temperature value and its holding time of the environment in each small ceramic part of each ceramic component are collected; using a high-temperature sensor and a high-speed data acquisition instrument, the real-time steady-state high-temperature value and its holding time of the environment in each small ceramic part of each ceramic component are collected; using a temperature recorder and a miniature patch-type humidity sensor, the steady-state high and low temperature difference cycle number and real-time relative humidity of the environment in each small ceramic part of each ceramic component are collected; using a gas sensor and a liquid concentration sensor, the real-time concentration of corrosive media in the environment of each small ceramic part of each ceramic component is collected; using a corrosion resistance test chamber and a time counter, the cumulative contact time between each small ceramic part of each ceramic component and the corrosive media is collected; using a vibration sensor, an acceleration sensor, a light intensity sensor, and an ultraviolet radiation meter, the real-time external vibration frequency, real-time external vibration acceleration, real-time solar radiation intensity, and cumulative ultraviolet radiation dose of the environment in each small ceramic part of each ceramic component are collected;

[0083] Image enhancement and denoising were performed on the ceramic image data. Data cleaning and standardization were performed on the water pressure strength test data, ceramic physical property data, and test environment data. The pre-processed internal water pressure load monitoring data were integrated to generate the Weibull statistical data set.

[0084] In S2, the process of obtaining the ceramic feature data includes:

[0085] The ceramic feature data is the crack length of each small ceramic region in each experimental ceramic piece. Feature extraction is performed on the real-time surface image of each experimental ceramic piece. Based on the semantic segmentation technology of deep learning, the small ceramic regions in each experimental ceramic piece are identified in the real-time surface image of each experimental ceramic piece.

[0086] A U-shaped convolutional neural network was used to segment the crack regions in each small ceramic region of each identified test ceramic piece, and the cross-entropy loss function was used to optimize the crack region segmentation results.

[0087] The skeleton of the crack region segmentation result is extracted. Combined with the Zhang-Suen thinning algorithm, the crack region segmentation result is reduced to a skeleton line with a width of one pixel, while preserving the topological structure of the crack. The skeleton line with a width of one pixel is traversed at the pixel level, and the distance between adjacent skeleton pixels is calculated. The distances of all adjacent pixels are accumulated to obtain the total pixel length of the crack in each small ceramic part region of each test ceramic piece. According to the image scale, the total pixel length of the crack in each small ceramic part region of each test ceramic piece is converted into the actual physical length to obtain the crack length of each small ceramic part region of each test ceramic piece.

[0088] In S3, the process of fitting the distribution parameters of the Weibull statistics using the maximum likelihood estimation method to obtain the Weibull statistical distribution model includes:

[0089] The distribution parameters of Weibull statistics are the shape parameters, scale parameters, and location parameters of the Weibull statistical distribution model.

[0090] The Weibull statistical distribution model is defined, and the maximum likelihood estimation method is adopted. Based on the probability density function of the Weibull statistical distribution, the expression of the probability density function of the Weibull statistical distribution for each experimental porcelain piece is obtained. The likelihood function is fitted, and the natural logarithm of the likelihood function is taken to obtain the log-likelihood function, thereby simplifying the calculation. The likelihood function is the product of the probability density function expressions of the Weibull statistical distribution for each experimental porcelain piece. The expressions for the Weibull statistical distribution model, the probability density function of the Weibull statistical distribution, the likelihood function, and the log-likelihood function are as follows:

[0091] ;

[0092] ;

[0093] ;

[0094] ;

[0095] in, , , and These are the Weibull statistical distribution model, the probability density function, the likelihood function, and the log-likelihood function of the Weibull statistical distribution, respectively. Describe the failure probability of each small ceramic part area in each ceramic part; S is the internal water pressure load of each small ceramic part area in each ceramic part. The shape parameters of the Weibull statistical distribution model describe the trend of failure probability with load. At that time, the failure probability increases rapidly with increasing load. At that time, the failure probability increases linearly with the load. At that time, the failure probability increases more slowly with increasing load; The scale parameter of the Weibull statistical distribution model corresponds to the load value when the failure probability is 63.2%. It is an inherent property of the Weibull distribution and is determined by the material and structure of the ceramic piece itself. The location parameters of the Weibull statistical distribution model indicate that the internal water pressure load is lower than When the ceramic component fails, the probability is 0. In engineering, the model can be fitted based on actual data. When simplifying the model, it can be set to... =0, which means a two-parameter Weibull distribution, but this embodiment still selects a three-parameter Weibull distribution; The probability density function of the Weibull statistical distribution corresponding to each experimental ceramic piece; The internal water pressure load of each small ceramic part area in each test ceramic piece; n represents the total number of test ceramic pieces, and i can be a positive integer between 1 and n;

[0096] By maximizing the log-likelihood function, the estimated values ​​of the shape, scale, and location parameters of the Weibull statistical distribution model are obtained. The estimated values ​​of the shape, scale, and location parameters of the Weibull statistical distribution model are then substituted into the Weibull statistical distribution model to obtain the Weibull statistical distribution model.

[0097] The goodness-of-fit test method was used to verify the fit between the internal water pressure strength test data, ceramic component characteristic data and the Weibull statistical distribution model. The shape parameters, scale parameters and position parameters of the Weibull statistical distribution model were further optimized to obtain the final Weibull statistical distribution model.

[0098] In S4, the process of deducing the selection range of water pressure load inside the ceramic part based on the Weibull statistical distribution model includes:

[0099] The failure probability threshold of ceramic parts includes the defect rejection threshold and the qualified protection threshold for each small ceramic part area in each ceramic part. The defect rejection threshold is the lower limit threshold of the water pressure load screening range in the ceramic part, and the qualified protection threshold is the upper limit threshold of the water pressure load screening range in the ceramic part.

[0100] A failure probability threshold for ceramic components is set, and then substituted into the Weibull statistical distribution model to deduce the internal water pressure load corresponding to each small ceramic component area within each ceramic component for different failure probability thresholds. Specifically, the defect rejection threshold for each small ceramic component area within each ceramic component is set to 99.9%, and the qualified protection threshold is set to 0.1%. The defect rejection threshold ensures that defective components break, allowing for timely screening of defective components, while the qualified protection threshold ensures that qualified components suffer zero damage, reducing subsequent usage risks. Therefore, the defect rejection threshold for each small ceramic component area within each ceramic component is set to 99.9%, and the qualified protection threshold is set to 0.1%. This value can be adjusted according to actual application conditions.

[0101] The internal water pressure load of each small ceramic part area corresponding to the defect rejection threshold is used as the lower limit of the internal water pressure load screening range of the ceramic part, and the internal water pressure load of each small ceramic part area corresponding to the qualified protection threshold is used as the upper limit of the internal water pressure load screening range of the ceramic part, thus obtaining the internal water pressure load screening range of the ceramic part.

[0102] In S5, the process of constructing a regression model for the physical properties of ceramic parts and outputting the regression coefficients for the physical properties of ceramic parts includes:

[0103] The regression coefficients of the physical properties of ceramic parts include the regression coefficients of hardness, temperature, humidity, compressive strength and elastic modulus of each small ceramic part area in each ceramic part, which respectively describe the linear relationship between the hardness, temperature, humidity, compressive strength and elastic modulus of each small ceramic part area in each ceramic part and the corresponding internal water pressure load.

[0104] Physical property data of ceramic pieces were extracted from the Weibull statistical dataset and transformed into linear training set and linear test set, with the ratio of linear training set data to linear test set data being 7:3.

[0105] The multiple linear regression algorithm is used. The hardness, temperature, humidity, compressive strength and elastic modulus of each small ceramic region in each ceramic piece in the linear training set are used as input data, and the internal water pressure load of each small ceramic region in each ceramic piece in the linear training set is used as output data. The linear relationship between the hardness, temperature, humidity, compressive strength and elastic modulus of each small ceramic region in each ceramic piece and the corresponding internal water pressure load is learned respectively, and the regression model of the physical properties of the ceramic piece is obtained.

[0106] The linear test set data is input into the regression model of the physical properties of ceramic parts. An adaptive moment estimator optimizer is used to adjust the regression coefficients and intercept terms of the regression model of the physical properties of ceramic parts, optimize the performance of the regression model of the physical properties of ceramic parts, obtain the final regression model of the physical properties of ceramic parts, and then obtain the regression coefficients of the physical properties of ceramic parts. The regression coefficients of the physical properties of ceramic parts are then integrated into the Weibull statistical data set.

[0107] The expression for the regression model of the physical properties of this ceramic piece is:

[0108] ;

[0109] in, , , , and These are the regression coefficients for the hardness, temperature, humidity, compressive strength, and elastic modulus of each small ceramic part within each ceramic piece. , , , and The parameters are the hardness, temperature, humidity, compressive strength, and elastic modulus of each small ceramic component area within each ceramic piece. and These are the intercept and error terms of the regression model for the physical properties of ceramic parts, respectively.

[0110] In S6, the process of constructing a calibration model for the internal water pressure load of ceramic parts, outputting the internal water pressure load calibration coefficients, and calibrating the internal water pressure load screening range of ceramic parts includes:

[0111] The regression coefficients of the physical properties of ceramic pieces and the test environment data were extracted from the Weibull statistical data set and divided into a nonlinear training set and a nonlinear test set, with the ratio of the nonlinear training set data to the nonlinear test set data being 8:2.

[0112] Using the random forest algorithm, the regression coefficients of the physical properties of ceramic parts and the test environment data are set as input data, and the calibration coefficients of the internal water pressure load are set as output data. The nonlinear relationship between the regression coefficients of the physical properties of ceramic parts, the test environment data and the calibration coefficients of the internal water pressure load is learned, and the internal water pressure load calibration model of ceramic parts is trained and obtained.

[0113] The nonlinear training set data is input into the calibration model of internal water pressure load of ceramic parts obtained through training. The stochastic gradient descent optimizer is used to adjust the parameters of the calibration model of internal water pressure load of ceramic parts, optimize the calibration model of internal water pressure load of ceramic parts, and obtain the final calibration model of internal water pressure load of ceramic parts.

[0114] Combining the regression coefficients of the physical properties of the ceramic parts and the test environment data, the corresponding internal water pressure load calibration coefficients are output. Based on the output internal water pressure load calibration coefficients, the internal water pressure load screening range of the ceramic parts is calibrated.

[0115] The specific calibration process includes: calibrating the internal water pressure load screening range of the ceramic parts based on the output internal water pressure load calibration coefficient. This actually involves calibrating the internal water pressure load of each small ceramic part region corresponding to the defect rejection threshold and the pass protection threshold within the internal water pressure load screening range. A first internal water pressure load calibration threshold and a second internal water pressure load calibration threshold are set. When the internal water pressure load calibration coefficient is lower than the first internal water pressure load calibration threshold, the internal water pressure load of each small ceramic part region corresponding to the defect rejection threshold is calibrated according to... The formula calibrates the internal water pressure load of each small ceramic part area within each ceramic part corresponding to the defect removal threshold. When the internal water pressure load calibration coefficient is between the first and second internal water pressure load calibration thresholds, no calibration is performed on the internal water pressure load screening interval of the ceramic part. When the internal water pressure load calibration coefficient is higher than the second internal water pressure load calibration threshold, the internal water pressure load of each small ceramic part area within each ceramic part corresponding to the qualified protection threshold is calibrated, based on... The formula defines the internal water pressure load of each small ceramic component area within each ceramic component corresponding to the calibration qualified protection threshold, where... and These represent the internal water pressure loads in each small ceramic part region of each ceramic part, corresponding to the defect removal thresholds before and after calibration. and These represent the internal water pressure loads in each small ceramic component area of ​​each ceramic component, corresponding to the qualified protection thresholds before and after calibration. This is the calibration coefficient for internal water pressure load.

[0116] In S7, the process of selecting the optimal internal water pressure load for ceramic components includes:

[0117] Based on the internal water pressure load of the ceramic part within the calibrated internal water pressure load screening range, an objective function is defined and a maximum number of iterations is set. The particle swarm optimization algorithm is adopted to initialize the particle swarm, and each particle in the particle swarm is evaluated through the objective function. That is, the performance of the ceramic part under the internal water pressure load is calculated. The evaluation result represents the fitness of the particle. The higher the fitness, the better the performance of the ceramic part.

[0118] The position and velocity of each particle in the particle swarm are updated to guide the screening process of the optimal internal water pressure load for the ceramic part. After each update and optimization, the fitness of each particle in the particle swarm is continuously compared iteratively to update the global optimal solution. When the number of iterations reaches the maximum number of iterations, the update process of the particle swarm is stopped, and the optimal internal water pressure load for the ceramic part is output through the particle swarm optimization algorithm.

[0119] The particle swarm update formula, which guides the screening process by updating the position and velocity of each particle in the particle swarm, is as follows:

[0120] ;

[0121] ;

[0122] in, It is the velocity of the particle. It refers to the position of the particles (i.e., the water pressure load inside the ceramic piece). It is the optimal position in the history of particles. It is the global optimal position of the particle swarm. It is inertial weight. and It is the acceleration constant. and It is a random number. It is the velocity of the particles after the update. This refers to the updated particle positions. The particle velocity and position are based on the calibrated internal water pressure load within the ceramic component's screening range, obtained after initializing the particle swarm using a particle swarm optimization algorithm. and It is generated and constantly changes during the update of the position and velocity of each particle in the particle swarm.

[0123] In S8, the process of optimizing the water pressure load inside the ceramic component and generating the final output report of the water pressure load inside the ceramic component includes:

[0124] The distribution of internal water pressure load in each small ceramic part area of ​​each actual ceramic part is analyzed within the calibrated internal water pressure load screening range. The qualification level of the internal water pressure load of each actual ceramic part is evaluated and corresponding operations are performed accordingly. The qualification level of the internal water pressure load of each ceramic part and the ceramic part operation record are obtained to achieve optimization of the internal water pressure load of the ceramic part.

[0125] If no small ceramic part has an internal water pressure load outside the calibrated internal water pressure load screening range, the ceramic part is considered excellent and will be kept. If one small ceramic part has an internal water pressure load outside the calibrated internal water pressure load screening range, the ceramic part is considered good. If two small ceramic parts have internal water pressure loads outside the calibrated internal water pressure load screening range, the ceramic part is considered qualified. Both good and qualified ceramic parts will be continuously monitored. If any small ceramic part is found to be outside the calibrated internal water pressure load screening range, the corresponding ceramic part will be rejected. If two or more small ceramic parts have water pressure loads outside the calibrated internal water pressure load screening range, the ceramic part is considered unqualified and will be directly rejected.

[0126] By using report generation technology, the water pressure load qualification level of each ceramic component and the operation record of the ceramic component are integrated to generate the final output report of water pressure load in the ceramic component.

[0127] Example 2: Figure 2 As shown, an internal water pressure load optimization system based on Weibull statistics is used to implement the method in Example 1. It includes a data acquisition module, a feature extraction module, a distributed application module, a load calibration and screening module, and an optimization output module. The modules are interconnected.

[0128] The data acquisition module is used to collect monitoring data on internal water pressure load in various small ceramic parts, including internal water pressure strength test data, ceramic part image data, ceramic part physical property data, and test environment data, providing comprehensive and multi-dimensional solid data support for the entire internal water pressure load analysis process of ceramic parts;

[0129] The feature extraction module extracts features from the porcelain image data to obtain porcelain feature data, thus achieving accurate mining of porcelain feature information.

[0130] The distribution application module is used to fit the distribution parameters of Weibull statistics to obtain the Weibull statistical distribution model, and then obtain the water pressure load screening interval inside the ceramic part. The distribution application module is divided into a distribution fitting unit and an interval delineation unit. In a unit-based collaborative mode, it completes the entire process from the construction of the Weibull statistical distribution model to the definition of the water pressure load interval inside the ceramic part.

[0131] The distribution fitting unit, referencing the internal water pressure strength test data and ceramic component characteristic data, uses the maximum likelihood estimation method to fit the distribution parameters of the Weibull statistics, obtaining the Weibull statistical distribution model. It deeply integrates the internal water pressure strength test data and ceramic component characteristic data to construct a quantitative analysis tool that conforms to the variation law of internal water pressure load in ceramic components.

[0132] The interval is defined as a unit, the failure probability threshold of the ceramic component is set, and the water pressure load screening interval inside the ceramic component is derived based on the Weibull statistical distribution model.

[0133] The load calibration and screening module is used to calibrate the internal water pressure load screening range of the ceramic part and screen out the optimal internal water pressure load of the ceramic part. The load calibration and screening module is divided into a range calibration unit and a load screening unit, achieving the dual goals of accurate correction of the load range and efficient positioning of the optimal load.

[0134] The interval calibration unit constructs a regression model of the physical properties of ceramic parts using physical property data and a multiple linear regression algorithm, and outputs the regression coefficients of the physical properties of ceramic parts. It also constructs a calibration model of the internal water pressure load of ceramic parts using experimental environment data and a random forest algorithm, and outputs the internal water pressure load calibration coefficients. This calibrates the internal water pressure load screening interval of ceramic parts, comprehensively considering the influence of the physical properties of ceramic parts and environmental factors, and significantly improves the accuracy of the load screening interval in matching with actual application scenarios.

[0135] The load screening unit, based on the calibrated internal water pressure load screening range of the ceramic part, uses the particle swarm optimization algorithm to screen out the optimal internal water pressure load of the ceramic part, thereby achieving efficient optimization within the calibrated internal water pressure load screening range and determining the best internal water pressure load configuration of the ceramic part.

[0136] The output module is optimized by using the calibrated internal water pressure load screening range and the optimal internal water pressure load of the ceramic part as references to optimize the internal water pressure load of the ceramic part and generate the final output report of the internal water pressure load of the ceramic part.

Claims

1. A Weibull statistics-based internal water ballast load optimization method, characterized by, The method comprises the following steps: S1, divide the surface of the target detection porcelain piece into a plurality of small porcelain piece regions with the same area, collect internal water pressure load monitoring data of each small porcelain piece region, including internal water pressure strength test data, porcelain piece image data, porcelain piece physical property data, and test environment data; S2, perform feature extraction on the porcelain piece image data to obtain porcelain piece feature data; S3, refer to the internal water pressure strength test data and the porcelain piece feature data, fit the distribution parameters of Weibull statistics by using a maximum likelihood estimation method to obtain a Weibull statistical distribution model, the process comprising: The distribution parameters of Weibull statistics are shape parameters, scale parameters, and position parameters of the Weibull statistical distribution model; According to the probability density function of Weibull statistical distribution, the expression of the probability density function of Weibull statistical distribution of each test porcelain piece is obtained by using the maximum likelihood estimation method, the likelihood function is fitted, and the log-likelihood function is obtained by taking the natural logarithm of the likelihood function; The shape parameters, scale parameters, and position parameters of the Weibull statistical distribution model are estimated by maximizing the log-likelihood function, and the estimated values of the shape parameters, scale parameters, and position parameters of the Weibull statistical distribution model are substituted into the Weibull statistical distribution model to obtain the Weibull statistical distribution model; The goodness-of-fit test method is used to verify the fit of the internal water pressure strength test data, the porcelain piece feature data, and the Weibull statistical distribution model, and the shape parameters, scale parameters, and position parameters of the Weibull statistical distribution model are further optimized to obtain the final Weibull statistical distribution model; S4, set a porcelain piece failure probability threshold, and inversely deduce a porcelain piece internal water pressure load screening interval based on the Weibull statistical distribution model; S5, construct a porcelain piece physical property regression model by using the porcelain piece physical property data and a multiple linear regression algorithm, and output porcelain piece physical property regression coefficients; S6, construct a porcelain piece internal water pressure load calibration model by using the random forest algorithm, the porcelain piece physical property regression coefficients, and the test environment data, output internal water pressure load calibration coefficients, and calibrate the porcelain piece internal water pressure load screening interval; S7, based on the calibrated porcelain piece internal water pressure load screening interval, screen out the optimal internal water pressure load of the porcelain piece by using the particle swarm optimization algorithm; S8, use the calibrated porcelain piece internal water pressure load screening interval and the optimal internal water pressure load of the porcelain piece as a reference to optimize the porcelain piece internal water pressure load, and generate a final output report of the porcelain piece internal water pressure load.

2. The Weibull statistics-based internal water ballast load optimization method according to claim 1, characterized in that, In S1, the collection process of the internal water pressure load monitoring data of each small porcelain piece region comprises: Assign numbers to each small porcelain piece region, deploy different types of data collection equipment, and collect internal water pressure load monitoring data of each small porcelain piece region; The anti-internal water pressure strength test data includes the fracture water pressure value of each small porcelain piece area in each test porcelain piece, the pressure drop value when the porcelain piece is fractured, the inner diameter of the main body of each test porcelain piece, the outer diameter of the main body, the effective height of the main body, the effective length of the main body, the wall thickness, the inner diameter of the flange, the thickness of the flange, the diameter of the sealing surface of the flange, the diameter of the bolt hole center circle, the arc transition radius between the main body and the flange, the arc transition radius between the main body and the head, the curvature radius of the head, and the thickness of the head; The porcelain piece image data is a real-time surface image of each test porcelain piece; The porcelain piece physical property data includes the internal water pressure load, hardness, temperature, humidity, compressive strength and elastic modulus of each small porcelain piece area in each porcelain piece; The test environment data includes the real-time stable low temperature value and its holding time, the real-time stable high temperature value and its holding time, the number of stable high-low temperature difference cycles, the real-time relative humidity, the real-time concentration of corrosive medium, the cumulative contact time with corrosive medium, the real-time external vibration frequency, the real-time external vibration acceleration, the real-time solar radiation intensity and the cumulative radiation dose of ultraviolet rays of each small porcelain piece area environment in each porcelain piece; The porcelain piece image data is subjected to image enhancement and image denoising, the anti-internal water pressure strength test data, the porcelain piece physical property data and the test environment data are subjected to data cleaning and data standardization processing, and the pre-processed internal water pressure load monitoring data are integrated to generate a Weibull statistical data set.

3. The Weibull statistics-based internal water ballast load optimization method according to claim 2, characterized in that, In S2, the porcelain piece feature data acquisition process includes: The porcelain piece feature data is the crack length of each small porcelain piece area in each test porcelain piece, and the real-time surface image of each test porcelain piece is subjected to feature extraction, and each small porcelain piece area in each test porcelain piece in the real-time surface image of each test porcelain piece is identified based on a deep learning semantic segmentation technology; A U-shaped convolutional neural network is used to segment the crack area in each small porcelain piece area in each test porcelain piece, and a cross-entropy loss function is used to optimize the crack area segmentation result; The crack area segmentation result is subjected to skeleton extraction, and the crack area segmentation result is reduced to a single-pixel-width skeleton line by combining a Zhang-Suen thinning algorithm, the topological structure of the crack is retained, the single-pixel-width skeleton line is subjected to pixel-level traversal, the distance between adjacent skeleton pixels is calculated, and the total pixel length of the crack in each small porcelain piece area in each test porcelain piece is obtained by accumulating the distances between all adjacent pixels, and the total pixel length of the crack in each small porcelain piece area in each test porcelain piece is converted into an actual physical length according to an image scale to obtain the crack length of each small porcelain piece area in each test porcelain piece.

4. The Weibull statistics-based internal water ballast load optimization method according to claim 1, characterized by, In S4, the process of inversely deducing the internal water pressure load screening interval of the porcelain piece based on the Weibull statistical distribution model includes: The porcelain piece failure probability threshold includes a defect rejection threshold and a qualified protection threshold of each small porcelain piece area in each porcelain piece; The porcelain piece failure probability threshold is set, and the porcelain piece failure probability threshold is substituted into the Weibull statistical distribution model to inversely deduce the internal water pressure load of each small porcelain piece area in each porcelain piece corresponding to different porcelain piece failure probability thresholds, so as to obtain the internal water pressure load of each small porcelain piece area in each porcelain piece corresponding to the porcelain piece failure probability threshold. The defect rejection threshold corresponds to the internal water pressure load of each small porcelain piece region in each porcelain piece as the lower limit value of the porcelain piece internal water pressure load screening interval, and the qualified protection threshold corresponds to the internal water pressure load of each small porcelain piece region in each porcelain piece as the upper limit value of the porcelain piece internal water pressure load screening interval, to obtain the porcelain piece internal water pressure load screening interval.

5. The Weibull statistics-based internal water ballast load optimization method according to claim 2, characterized by, In S5, the process of constructing a porcelain piece physical property regression model and outputting porcelain piece physical property regression coefficients includes: The porcelain piece physical property regression coefficients include the regression coefficients of hardness, temperature, humidity, compressive strength and elastic modulus of each small porcelain piece region in each porcelain piece; The porcelain piece physical property data in the Weibull statistical data set is extracted and converted into a linear training set and a linear test set; Using a multiple linear regression algorithm, the hardness, temperature, humidity, compressive strength and elastic modulus of each small porcelain piece region in each porcelain piece in the linear training set are taken as input, and the internal water pressure load of each small porcelain piece region in each porcelain piece in the linear training set is taken as output, and the linear relationship between the hardness, temperature, humidity, compressive strength and elastic modulus of each small porcelain piece region in each porcelain piece is learned, respectively, and the internal water pressure load corresponding to the internal water pressure load is learned, respectively, and the porcelain piece physical property regression model is trained; The linear test set data is input into the porcelain piece physical property regression model, an adaptive moment estimation optimizer is used to adjust the regression coefficients and intercept terms of the porcelain piece physical property regression model, the performance of the porcelain piece physical property regression model is optimized, the final porcelain piece physical property regression model is obtained, and the porcelain piece physical property regression coefficients are obtained, and the porcelain piece physical property regression coefficients are integrated into the Weibull statistical data set.

6. The Weibull statistics-based internal water ballast load optimization method of claim 1, wherein, In S6, the process of constructing a porcelain piece internal water pressure load calibration model and outputting internal water pressure load calibration coefficients for calibrating the porcelain piece internal water pressure load screening interval includes: The porcelain piece physical property regression coefficients and test environment data in the Weibull statistical data set are extracted and divided into a nonlinear training set and a nonlinear test set; Using a random forest algorithm, the nonlinear training set data is set as input, and the internal water pressure load calibration coefficient is set as output, the nonlinear relationship between the porcelain piece physical property regression coefficient, the test environment data and the internal water pressure load calibration coefficient is learned, and the porcelain piece internal water pressure load calibration model is trained and obtained; The nonlinear training set data is input into the porcelain piece internal water pressure load calibration model trained and obtained, a stochastic gradient descent optimizer is used to adjust the porcelain piece internal water pressure load calibration model parameters, the porcelain piece internal water pressure load calibration model is optimized, and the final porcelain piece internal water pressure load calibration model is obtained; Combined with the current porcelain piece physical property regression coefficients and test environment data, the corresponding internal water pressure load calibration coefficients are output, and based on the output internal water pressure load calibration coefficients, the porcelain piece internal water pressure load screening interval is calibrated.

7. The Weibull statistics-based internal water ballast load optimization method of claim 1, wherein, In S7, the process of screening the optimal internal water pressure load of the porcelain piece includes: Based on the porcelain piece internal water pressure load located in the calibrated porcelain piece internal water pressure load screening interval, a target function is defined, a maximum number of iterations is set, a particle swarm optimization algorithm is used, a particle swarm is initialized, and each particle in the particle swarm is evaluated through the target function. The position and speed of each particle in the particle swarm are updated to guide the implementation of the optimal inner water pressure load screening process of the porcelain piece. After each optimization update, the fitness of each particle in the particle swarm is iteratively compared, and the global optimal solution is updated. When the number of iterations reaches the maximum number of iterations, the updating process of the particle swarm is stopped, and the optimal inner water pressure load of the porcelain piece is output through the particle swarm optimization algorithm.

8. The Weibull statistics-based internal water ballast load optimization method of claim 1, wherein, In S8, the process of optimizing the inner water pressure load of the porcelain piece and generating the final output report of the inner water water pressure load of the porcelain piece includes: Analyzing the distribution of the inner water pressure load of each small porcelain piece area in each actual porcelain piece in the calibrated porcelain piece inner water pressure load screening interval, evaluating the qualified level of the actual inner water pressure load of each porcelain piece and performing corresponding operations, obtaining the inner water pressure load qualified level of each porcelain piece and the porcelain piece operation record, and optimizing the inner water pressure load of the porcelain piece; Using report generation technology, integrating the inner water pressure load qualified level of each porcelain piece and the porcelain piece operation record, and generating the final output report of the inner water pressure load of the porcelain piece.

9. A Weibull statistics based internal water ballast optimization system for implementing the method of any one of claims 1-8, characterized by, It includes: A data acquisition module for acquiring inner water pressure load monitoring data of each small porcelain piece area, including inner water pressure strength test data, porcelain piece image data, porcelain piece physical property data, and test environment data; A feature extraction module for extracting features from porcelain piece image data to obtain porcelain piece feature data; A distribution application module for fitting the distribution parameters of Weibull statistics to obtain a Weibull statistical distribution model, and then obtaining the porcelain piece inner water pressure load screening interval, which is divided into the following units: A distribution fitting unit that uses the maximum likelihood estimation method to fit the distribution parameters of Weibull statistics based on the inner water pressure strength test data and porcelain piece feature data to obtain a Weibull statistical distribution model; An interval setting unit that sets a porcelain piece failure probability threshold and inversely deduces the porcelain piece inner water pressure load screening interval based on the Weibull statistical distribution model; A load calibration screening module for calibrating the porcelain piece inner water pressure load screening interval and screening the optimal inner water pressure load of the porcelain piece, which is divided into the following units: An interval calibration unit that uses porcelain piece physical property data and multiple linear regression algorithm to construct a porcelain piece physical property regression model, outputs porcelain piece physical property regression coefficients, combines test environment data and random forest algorithm to construct a porcelain piece inner water pressure load calibration model, and outputs inner water pressure load calibration coefficients to calibrate the porcelain piece inner water pressure load screening interval; A load screening unit that uses the calibrated porcelain piece inner water pressure load screening interval and the particle swarm optimization algorithm to screen the optimal inner water pressure load of the porcelain piece; An optimization output module that optimizes the inner water pressure load of the porcelain piece and generates the final output report of the inner water pressure load of the porcelain piece based on the calibrated porcelain piece inner water pressure load screening interval and the optimal inner water pressure load of the porcelain piece.

Images