Power station boiler heat-resistant alloy steel hardness detection method and system based on large model

Abstract

The present application relates to the technical field of hardness detection, and discloses a power station boiler heat-resistant alloy steel hardness detection method and system based on a large model, the method comprising: obtaining original data and correcting the Rockwell hardness original value for roughness to obtain an initial data set; grouping alloy element proportions, correcting environmental noise interference to obtain Rockwell hardness reference values of different material types; correcting the Rockwell hardness reference values for temperature and vibration interference to obtain Rockwell hardness correction values; obtaining conversion coefficients matching the Rockwell hardness correction values and performing dynamic optimization and feature enhancement to obtain final conversion coefficients; using support vector regression to perform fitting processing on the Rockwell hardness correction values and the final conversion coefficients to obtain Brinell hardness prediction values; and calculating a remaining life index in combination with the Brinell hardness prediction values. The present method can realize high-precision conversion from Rockwell hardness to Brinell hardness and improve the reliability of heat-resistant alloy steel remaining life prediction.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of hardness detection, and in particular to a power station boiler heat-resistant alloy steel hardness detection method and system based on a large model. BACKGROUND

[0002] At present, as the core equipment of the thermal power generation system, the safe and stable operation of the power station boiler is directly related to the reliability and economic benefits of the entire power system. The heat-resistant alloy steel components inside the boiler are subjected to high temperature and high pressure environment for a long time, and accurate detection of the material hardness is of decisive significance for evaluating the equipment state and preventing safety accidents.

[0003] In one prior art, Leeb hardness detection is used for on-site nondestructive testing, hardness readings on the surface of the component are collected, and Leeb hardness is converted to Brinell hardness using an empirical formula or a linear conversion table. The detection results are stored and preliminarily analyzed by means of a data logger. However, in the complex boiler operating environment of high temperature, high pressure and vibration, the system based on the traditional conversion method and single data processing cannot comprehensively consider and dynamically compensate the multi-dimensional factors such as material composition, environmental interference and measurement conditions, resulting in the accuracy of hardness conversion and state evaluation being affected.

[0004] In summary, the prior art cannot establish an accurate conversion relationship considering the influence of material type and environmental factors, and cannot realize high-precision conversion of Leeb hardness to Brinell hardness, and the reliability of the remaining life prediction of the heat-resistant alloy steel of the power station boiler is low. SUMMARY

[0005] The present application provides a power station boiler heat-resistant alloy steel hardness detection method and system based on a large model to realize high-precision conversion of Leeb hardness to Brinell hardness and improve the reliability of the remaining life prediction of the heat-resistant alloy steel.

[0006] In a first aspect, to solve the above technical problems, the present application provides a power station boiler heat-resistant alloy steel hardness detection method based on a large model, comprising:

[0007] Obtain the original data, and correct the Leeb hardness original value in the original data for surface roughness influence to obtain an initial data set; the original data includes: Leeb hardness original value, alloy element proportion, vibration interference information and temperature fluctuation record;

[0008] According to the alloy element proportion, the initial data set is grouped and processed, the Leeb hardness value under different surface roughness of each group is calculated and corrected for environmental noise interference to obtain the Leeb hardness reference value under different material types;

[0009] According to the vibration interference information and the temperature fluctuation record, the leeb hardness reference value is corrected for temperature fluctuation and vibration interference to obtain a leeb hardness correction value;

[0010] A conversion coefficient matching the leeb hardness correction value is obtained and is subjected to dynamic optimization and feature enhancement processing to obtain a final conversion coefficient;

[0011] The leeb hardness correction value and the final conversion coefficient are subjected to fitting processing by support vector regression to obtain a brinell hardness prediction value;

[0012] A brinell hardness theoretical value corresponding to the alloy element proportion is obtained, and the brinell hardness prediction value is combined to calculate a residual life index.

[0013] In an optional implementation, the original data is obtained, and the leeb hardness original value in the original data is subjected to surface roughness influence correction to obtain an initial data set, including:

[0014] The leeb hardness original value, the alloy element proportion, the vibration interference information and the temperature fluctuation record are obtained to obtain the original data;

[0015] The roughness parameter is measured, and the roughness correction coefficient is calculated in combination with the probe contact pressure;

[0016] The leeb hardness original value is corrected according to the roughness correction coefficient to obtain a leeb hardness initial value, and the initial data set is obtained in combination with the alloy element proportion, the vibration interference information and the temperature fluctuation record.

[0017] In an optional implementation, the initial data set is grouped according to the alloy element proportion, the leeb hardness value of each group under different surface roughness is calculated, and correction is made for environmental noise interference to obtain the leeb hardness reference value under different material types, including:

[0018] The initial data set is subjected to frequency domain processing, the interference features of environmental noise are extracted, and noise reduction processing is performed to obtain a noise reduction data set;

[0019] The interference correction coefficient is determined according to the interference features;

[0020] The feature vector of the alloy element proportion is extracted from the noise reduction data set and is subjected to feature screening to obtain a feature data set;

[0021] The feature data set is subjected to grouping processing to obtain a grouped data set;

[0022] According to the group data set, the Leeb hardness value of each group of materials under the influence of different surface roughness is calculated, and the Leeb hardness reference value under different material types is obtained by correcting the interference correction coefficient.

[0023] In an optional implementation, the temperature fluctuation and vibration interference correction of the Leeb hardness reference value according to the vibration interference information and the temperature fluctuation record obtains a Leeb hardness correction value, and the method comprises the following steps:

[0024] According to the vibration interference information, vibration interference features are extracted;

[0025] According to the temperature fluctuation record, temperature fluctuation features are extracted;

[0026] According to the vibration interference features and the temperature fluctuation features, a comprehensive correction coefficient is calculated;

[0027] According to the comprehensive correction coefficient, the Leeb hardness reference value is corrected to obtain a Leeb hardness correction value.

[0028] In an optional implementation, the method further comprises the following steps:

[0029] A material type record matched with the Leeb hardness correction value is obtained from a preset Brinell hardness database, and a corresponding material conversion coefficient is extracted to obtain a first conversion coefficient;

[0030] The first conversion coefficient is adjusted to obtain a second conversion coefficient in view of environmental noise interference;

[0031] The second conversion coefficient is smoothed to obtain a third conversion coefficient;

[0032] The matching degree of the Leeb hardness correction value and the material type is calculated, if the matching degree is lower than a preset matching degree threshold, dynamic optimization and feature enhancement processing are performed to obtain a final conversion coefficient, and if the matching degree is not lower than the matching degree threshold, the third conversion coefficient is determined as the final conversion coefficient.

[0033] In an optional implementation, the dynamic optimization and feature enhancement processing to obtain the final conversion coefficient comprises the following steps:

[0034] The uniformity features of the distribution of the measurement points are obtained;

[0035] A material type record matched with the alloy element proportion is obtained from a preset material property library, a corresponding element conversion coefficient is extracted, the element conversion coefficient and the uniformity features are integrated to obtain a first coefficient set;

[0036] analyze spatial correlation characteristics of the alloy element proportion and the measurement point distribution, obtain a first kernel function weight corresponding to the measurement point distribution, adjust the first coefficient set according to the first kernel function weight, and obtain a second coefficient set;

[0037] smooth the second coefficient set to obtain a third coefficient set;

[0038] adjust an offset of the third conversion coefficient with respect to the measurement point distribution according to the third coefficient set, and fuse the first kernel function weight to obtain an optimized conversion coefficient;

[0039] perform feature enhancement processing on the optimized conversion coefficient to obtain a final conversion coefficient.

[0040] In an optional implementation, the feature enhancement processing on the optimized conversion coefficient to obtain the final conversion coefficient includes:

[0041] obtain a material type record matched with the alloy element proportion from the material property library, extract and integrate distribution characteristics of a sampling frequency and a probe contact pressure value to obtain a first feature set;

[0042] obtain a second kernel function weight corresponding to the sampling frequency according to the first feature set to obtain a second feature set;

[0043] smooth the second feature set to obtain a third feature set;

[0044] adjust the optimized conversion coefficient according to the third feature set, and fuse the second kernel function weight to obtain the final conversion coefficient.

[0045] In an optional implementation, the fitting processing on the Rockwell hardness correction value and the final conversion coefficient by the support vector regression to obtain the Brinell hardness prediction value includes:

[0046] integrate features of the Rockwell hardness correction value and the measurement point distribution to obtain a first hardness value set;

[0047] fit the Rockwell hardness correction value and the final conversion coefficient by support vector regression according to the first hardness value set, obtain a third kernel function weight, and obtain a second hardness value set;

[0048] smooth the second hardness value set to obtain a third hardness value set;

[0049] perform interpolation processing according to the third hardness value set, and combine the third kernel function weight to obtain the Brinell hardness prediction value.

[0050] In an alternative embodiment, the Brinell hardness theoretical value corresponding to the alloy element proportion is obtained, and the residual life index is calculated in combination with the Brinell hardness prediction value, comprising:

[0051] The Brinell hardness theoretical value corresponding to the alloy element proportion is obtained from a preset heat-resistant alloy database;

[0052] The residual life index is calculated according to the Brinell hardness theoretical value and the Brinell hardness prediction value.

[0053] In a second aspect, the present application provides a large model-based power plant boiler heat-resistant alloy steel hardness detection system for realizing the large model-based power plant boiler heat-resistant alloy steel hardness detection method described above, comprising:

[0054] A data acquisition and correction module is configured to obtain original data, correct the Leeb hardness original value in the original data according to the surface roughness influence, and obtain an initial data set;

[0055] A reference value determination module is configured to group process the initial data set according to the alloy element proportion, calculate the Leeb hardness value under different surface roughness for each group, and correct the environmental noise interference to obtain the Leeb hardness reference value under different material types;

[0056] A reference value correction module is configured to correct the Leeb hardness reference value according to the vibration interference information and the temperature fluctuation record to obtain a Leeb hardness correction value;

[0057] A conversion coefficient determination module is configured to obtain a conversion coefficient matched with the Leeb hardness correction value and perform dynamic optimization and feature enhancement processing to obtain a final conversion coefficient;

[0058] A hardness prediction module is configured to perform fitting processing on the Leeb hardness correction value and the final conversion coefficient by using support vector regression to obtain a Brinell hardness prediction value;

[0059] A life evaluation module is configured to obtain the Brinell hardness theoretical value corresponding to the alloy element proportion, and calculate the residual life index in combination with the Brinell hardness prediction value.

[0060] Compared with the prior art, the present application has the following beneficial effects:

[0061] (1) The present application corrects the surface roughness influence of the original data hardness value, obtains the initial data set, and then groups and processes according to the proportion of alloy elements, and combines environmental noise interference for noise reduction processing, to obtain the hardness reference value of different material types. Finally, the hardness reference value is subjected to frequency domain analysis and fusion processing of temperature fluctuation and vibration interference, to obtain the hardness correction value. This series of operations effectively reduces the interference of surface state and environmental interference on hardness measurement, significantly improves the representativeness and stability of hardness data, and lays a clean and regular data foundation for subsequent accurate conversion, thereby solving the hardness value misalignment problem caused by large environmental interference and insufficient consideration of material differences in the prior art.

[0062] (2) The present application dynamically optimizes the conversion coefficient, and enhances the processing by fusing the measurement point distribution, sampling frequency and probe contact pressure and other multi-dimensional features, to obtain the final conversion coefficient. Then, the support vector regression is used to fit the hardness correction value and the final conversion coefficient to obtain the Brinell hardness prediction value. This process fully utilizes the spatial correlation and physical correlation between data, realizes high-precision and self-adaptive mapping from hardness to Brinell hardness, overcomes the limitations of poor applicability and large prediction deviation of traditional precision conversion models under complex working conditions, thereby significantly improving the accuracy and robustness of the hardness conversion model.

[0063] (3) The present application dynamically determines the remaining life index by correlating the Brinell hardness prediction value with the material state evaluation, and comprehensively predicting the error range and fitting calculation time. This process realizes end-to-end reasoning from hardness data to life evaluation, converts the detection data into a life index that can be used for maintenance decision-making, and at the same time, by introducing error and time consumption as evaluation factors, enhances the reliability and practicality of the life prediction result, thereby solving the problem of low reliability of life prediction in the prior art due to the lack of real-time data support. BRIEF DESCRIPTION OF DRAWINGS

[0064] Figure 1 is a flowchart of one embodiment of the hardness detection method of the power station boiler heat-resistant alloy steel provided by the present application based on a large model.

[0065] Figure 2 is a flowchart of one embodiment of the hardness detection method of the power station boiler heat-resistant alloy steel provided by the present application based on a large model.

[0066] Figure 3 is a structural schematic diagram of one embodiment of the hardness detection system of the power station boiler heat-resistant alloy steel provided by the present application based on a large model. DETAILED DESCRIPTION

[0067] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0068] Reference Figure 1 The first embodiment of the present invention provides a method for hardness testing of heat-resistant alloy steel for power plant boilers based on a large model, including the following steps:

[0069] S11, Obtain the original data, and perform surface roughness influence correction on the original Leeb hardness values ​​in the original data to obtain the initial dataset;

[0070] S12, the initial dataset is grouped according to the alloy element ratio, the Leeb hardness value of each group under different surface roughness is calculated and corrected for environmental noise interference, and the Leeb hardness reference value under different material types is obtained.

[0071] S13, Based on the vibration interference information and the temperature fluctuation record, the Leeb hardness reference value is corrected for temperature fluctuation and vibration interference to obtain the Leeb hardness correction value.

[0072] S14, obtain the conversion coefficient that matches the Leeb hardness correction value and perform dynamic optimization and feature enhancement processing to obtain the final conversion coefficient;

[0073] S15, support vector regression is used to fit the Leeb hardness correction value and the final conversion coefficient to obtain the Brinell hardness prediction value.

[0074] S16, obtain the theoretical Brinell hardness value corresponding to the alloy element ratio, and calculate the remaining life index by combining the predicted Brinell hardness value.

[0075] In step S11, raw data is acquired, and surface roughness influence correction is applied to the raw Leeb hardness values ​​in the raw data to obtain an initial dataset, including:

[0076] The raw data is obtained by acquiring the original Leeb hardness value, alloy element ratio, vibration interference information, and temperature fluctuation records.

[0077] The roughness parameters are measured, and the roughness correction coefficient is calculated by combining the probe contact pressure.

[0078] The original Leeb hardness value is corrected according to the roughness correction coefficient to obtain the initial Leeb hardness value. Combined with the alloy element ratio, the vibration interference information and the temperature fluctuation record, an initial dataset is obtained.

[0079] Firstly, the original data containing the original value of the Leeb hardness, the proportion of alloying elements, the vibration interference information and the temperature fluctuation record are acquired, and it should be noted that the original value of the Leeb hardness is collected by a portable Leeb hardness tester; the proportion of alloying elements can be detected in real time by an online X-ray fluorescence spectrometer; the vibration interference information can be collected by an accelerometer; and the temperature fluctuation record can be collected by a K-type thermocouple sensor deployed on the surface of the component and in the detection environment. The collected data are integrated to obtain the original data.

[0080] Subsequently, the roughness correction coefficient is calculated. Exemplarily, the core roughness parameters of the detection area of the component are measured by a contact surface roughness meter, including the arithmetic average deviation. Three different points are selected for measurement in each detection area, and the average value is taken as the representative value of the roughness parameter of the area. The probe contact pressure is collected in real time by a micro pressure sensor integrated in the probe of the Leeb hardness tester. The surface roughness influence factor represents the hardness deviation rate, and is obtained by the formula: wherein is the calculated surface roughness influence factor, is the representative value of the roughness parameter, is the actual contact pressure, is the reference contact pressure, and the initial default value is set to 0.5, which can be adjusted according to the standard contact pressure of the actual working condition. It should be noted that the coefficients 0.03 and 0.012 are the optimal coefficient combinations obtained by least square fitting based on a large amount of historical data covering different representative values of roughness parameters, contact pressure deviations and measured hardness deviation rates, which are suitable for ∈[0.1,10]μm, ∈[0,1]Mpa of the conventional processing environment. Then, the roughness correction coefficient is calculated based on , and the calculation formula is: wherein is the surface roughness influence factor, is the calculated roughness correction coefficient.

[0081] Exemplarily, when Ra=0.9μm, the actual contact pressure =0.47MPa, the influence factor is: =0.03×0.9+0.012×|0.47-0.5=0.027+0.00036=0.027 (retaining 3 decimal places), and the corresponding roughness correction coefficient = 1.028 (retaining 3 decimal places).

[0082] Then, the roughness correction coefficient is multiplied by the original Leeb hardness value to obtain the initial Leeb hardness value. The historical initial hardness values, alloy element ratios, vibration interference information and temperature fluctuation records are integrated into a dataset to obtain the initial dataset.

[0083] In step S12, the initial dataset is grouped according to the alloy element ratio, the Leeb hardness value of each group under different surface roughness is calculated, and correction is made for environmental noise interference to obtain the Leeb hardness reference value for different material types, including:

[0084] The initial dataset is subjected to frequency domain processing to extract the interference features of environmental noise, and then noise reduction processing is performed to obtain a noise-reduced dataset.

[0085] Based on the interference characteristics, determine the interference correction coefficient;

[0086] Based on the noise reduction dataset, feature vectors of the alloy element ratios are extracted and feature filtering is performed to obtain a feature dataset.

[0087] The feature dataset is grouped to obtain a grouped dataset;

[0088] Based on the grouped dataset, the Leeb hardness value of each group of materials under the influence of different surface roughness is calculated, and the value is corrected by the interference correction coefficient to obtain the Leeb hardness benchmark value for different material types.

[0089] First, the initial dataset is processed in the frequency domain to extract environmental noise interference features, and then denoised to obtain a denoised dataset. It should be noted that the time-domain signals in the initial dataset are digitized to form a discrete time series, which is then calculated using a Fourier transform algorithm to obtain a spectrum. The spectrum identifies mechanical vibration peaks in the 20-80Hz range and electromagnetic interference peaks around 150Hz. Subsequently, a band-stop filter with a stopband frequency matching the detected noise frequency and a stopband width set to ±10Hz of the main frequency is constructed for denoising, preserving the core signal components related to alloy element proportions, thus obtaining the denoised dataset.

[0090] Subsequently, the interference correction coefficient is determined. Illustratively, according to the ambient noise frequency characteristics, 1024 consecutive data points in each detection period are selected as an analysis window, the root mean square value of the noise band in the analysis window is calculated, and the maximum noise intensity of all the analysis windows is taken as the representative value of the noise intensity of the detection; the environmental noise influence factor represents the hardness reading deviation rate caused by noise, which is obtained by the formula: g = 0.025 x AT + 0.008 x N, wherein: g is the calculated environmental noise influence factor, N is the representative value of the noise intensity, and AT is the temperature fluctuation amplitude in the detection process, i.e. |T_max-T_min|. It should be noted that the coefficients 0.025 and 0.008 in the formula are the optimal coefficient combination obtained by multiple linear regression and least squares fitting based on a large number of historical data sets covering different noise intensities, temperature fluctuation amplitudes and measured hardness deviation rates, which are suitable for the conventional working scene of ATe[0, 5] °C, Ne[0, 50] mV. It is worth noting that when calculating N, the unit is unified to mV, and when calculating AT, the unit is unified to °C. After unifying the units, only the numerical value is used for calculation when the formula is calculated, and the obtained g is a dimensionless value.

[0091] Then, based on g, the interference correction coefficient is calculated, and the calculation formula is , wherein is the environmental noise influence factor, is the calculated interference correction coefficient. Illustratively, when the representative value of the noise intensity N = 12.5 mV and the temperature fluctuation amplitude AT = 3.2 °C during the detection process are measured, the environmental noise influence factor is: g = 0.025 x 3.2 + 0.008 x 12.5 = 0.080 + 0.100 = 0.180 (retaining 3 decimal places), and the corresponding interference correction coefficient is: = 1 / (1-0.180) = 1 / 0.82≈1.220 (retaining 3 decimal places).

[0092] After that, the feature vector of the alloy element ratio is extracted from the noise reduction data set, and the features that significantly affect the Rockwell hardness are screened to obtain a feature data set. It should be noted that the feature screening is performed by principal component analysis method, the contents of each alloy element in the initial data set are first standardized by Z-score, then the covariance matrix of the standardized data is calculated, and then the characteristic value and the corresponding feature vector are obtained by characteristic decomposition of the covariance matrix. Finally, the characteristic values are sorted, and the first three principal components are extracted as the main feature vector to obtain the feature data set.

[0093] Subsequently, according to the proportion of alloy elements, the feature data set is grouped by material type using the K-means clustering algorithm. It should be noted that the clustering dimension is three-dimensional, the preset grouping number K=3, and it is divided into low Cr-Mo system, high Cr-Ni system and composite strengthening system, to obtain the grouped data set.

[0094] Finally, according to the formula , the correction is carried out for each group in the grouped data set, and the Leeb hardness reference value under different material types is obtained. It should be noted that is the calculated Leeb hardness reference value, is the Leeb hardness value in the grouped data set, is the interference correction coefficient, is the roughness correction coefficient. For example, for a certain high Cr-Ni alloy group, the Leeb hardness value is 615HL, the interference coefficient =0.98, the roughness coefficient =0.96, and the Leeb hardness reference value of this group is: =615×0.98×0.96=579.8HL.

[0095] In step S13, the Leeb hardness reference value is corrected for temperature fluctuation and vibration interference according to the vibration interference information and the temperature fluctuation record, and the Leeb hardness correction value is obtained, including:

[0096] According to the vibration interference information, the vibration interference feature is extracted;

[0097] According to the temperature fluctuation record, the temperature fluctuation feature is extracted;

[0098] According to the vibration interference feature and the temperature fluctuation feature, the comprehensive correction coefficient is calculated;

[0099] According to the comprehensive correction coefficient, the Leeb hardness reference value is corrected to obtain the Leeb hardness correction value.

[0100] First, for example, the mean square value of acceleration is extracted from the vibration interference information composed of acceleration sequence as the vibration interference feature, and the low frequency band of temperature fluctuation is extracted from the temperature fluctuation record by a low-pass filter with a cutoff frequency of 1Hz, and then the temperature change amplitude is extracted from the low frequency band as the temperature fluctuation feature.

[0101] Subsequently, the temperature fluctuation feature and the vibration interference feature are fused by weighted average method, and the comprehensive correction coefficient is calculated. The specific calculation formula is: , wherein E is the calculated comprehensive correction coefficient, is the temperature fluctuation feature, is the vibration interference feature, and are weight coefficients, and the size of which is set according to the relative influence degree of temperature fluctuation and vibration interference on the hardness value in the Leeb hardness data, and exemplarily, if it is shown by analysis of historical data that temperature fluctuation is the main influencing factor, then .

[0102] Finally, according to the comprehensive correction coefficient, the Leeb hardness reference value is corrected by the formula: to obtain the Leeb hardness correction value. Wherein, is the calculated Leeb hardness correction value, is the Leeb hardness reference value obtained in the foregoing, and the coefficients a and b are the optimal coefficient combination obtained by least square fitting based on a large amount of historical data covering the influence factors of different temperature fluctuations and vibration interference fusion and the measured hardness deviation rate, and exemplarily, the coefficients and are respectively 2HL and 1.5HL.

[0103] Exemplarily, the reference value of the Leeb hardness value of a certain system for a certain heat-resistant steel pipe is 655HL, and when detected, the sensor captures a temperature fluctuation of 0.5Hz and a vibration interference of 50Hz. For the temperature fluctuation and the vibration interference, the temperature fluctuation amplitude of the low frequency band is extracted as the temperature fluctuation feature, and the vibration interference feature is 0.65 . The comprehensive correction coefficient calculated by substituting into the formula is 0.74. Finally, the reference value is compensated, and the corrected Leeb hardness value is output as 658HL (integer is retained).

[0104] As shown in Figure 2 , in step S14, the conversion coefficient matching the Leeb hardness correction value is obtained and dynamic optimization and feature enhancement processing are performed to obtain the final conversion coefficient, including:

[0105] Obtain the material type record matching the Leeb hardness correction value from the pre-set Brinell hardness database, extract the corresponding material conversion coefficient to obtain the first conversion coefficient;

[0106] Adjust the first conversion coefficient for environmental noise interference to obtain the second conversion coefficient;

[0107] Smooth the second conversion coefficient to obtain the third conversion coefficient;

[0108] Calculate the matching degree of the Leeb hardness correction value and the material type, if the matching degree is lower than the pre-set matching degree threshold, perform dynamic optimization and feature enhancement processing to obtain the final conversion coefficient, if the matching degree is not lower than the matching degree threshold, determine the third conversion coefficient as the final conversion coefficient.

[0109] Firstly, a Brinell hardness database of multi-materials is established based on experimental data, which contains a plurality of typical heat-resistant alloy steel samples, each group of samples records the element proportion, the Rockwell hardness, the corresponding Brinell hardness, the environmental parameters and the material conversion coefficient, and these sample vectors are stored in the database. In the database, the corresponding material conversion coefficient is extracted by the Rockwell hardness correction value and its corresponding alloy element characteristics as the retrieval condition, to obtain the first conversion coefficient. It should be noted that the material conversion coefficient is a proportional factor for converting the Rockwell hardness value to the Brinell hardness value.

[0110] Subsequently, the first conversion coefficient is integrated, and the second conversion coefficient is obtained by the formula: , wherein, is the calculated second conversion coefficient, is the first conversion coefficient, is the interference correction coefficient obtained in the foregoing.

[0111] Then, the second conversion coefficient is calculated by the formula: using the sliding weighted average method, to obtain the third conversion coefficient. Wherein, is the size of the sliding window, taking 5 measurement points, is the Gaussian weight, and the calculation formula is: , is the calculated third conversion coefficient. Exemplarily, if the values of the five consecutive measurement points are [0.0781, 0.0790, 0.0775, 0.0801, 0.0783], the smoothed value is =0.0786.

[0112] Finally, the matching degree of the Rockwell hardness correction value and the material type is calculated, and it should be noted that the definition of the matching degree is: , wherein, represents the nearest neighbor distance of the Rockwell hardness value and the material type matching, is the maximum distance in the database. Based on a large amount of experimental historical data, the matching degree threshold is set to 0.85, if the calculated M is less than 0.85, the subsequent dynamic optimization and enhancement processing steps are performed, and if M is greater than 0.85, the final conversion coefficient is .

[0113] Wherein, the dynamic optimization and feature enhancement processing are performed to obtain the final conversion coefficient, including:

[0114] obtaining the uniformity feature of the measurement point distribution;

[0115] ​obtaining a material type record matched with the alloy element proportion from a preset material characteristic library, extracting a corresponding element conversion coefficient, and integrating the element conversion coefficient and the uniformity feature to obtain a first coefficient set;

[0116] analyzing spatial correlation features of the alloy element proportion and the measurement point distribution, obtaining a first kernel function weight corresponding to the measurement point distribution, and adjusting the first coefficient set according to the first kernel function weight to obtain a second coefficient set;

[0117] performing smoothing processing on the second coefficient set to obtain a third coefficient set;

[0118] adjusting an offset of the third conversion coefficient changing with the measurement point distribution according to the third coefficient set, and fusing the first kernel function weight to obtain an optimized conversion coefficient;

[0119] performing feature enhancement processing on the optimized conversion coefficient to obtain a final conversion coefficient.

[0120] First, based on the coordinate information of the measurement points, the uniformity feature of the spatial distribution is calculated by the formula It should be noted that and in the formula are the average distance and the standard deviation between the three-dimensional coordinates of the measurement points collected by the laser range finder.

[0121] Subsequently, a multi-material characteristic library is established based on experimental data and material detection results, which contains multiple typical heat-resistant alloy steel samples. Each group of samples records the alloy element proportion, the Rockwell hardness, the corresponding Brinell hardness, and the corresponding element conversion coefficient. These sample vectors are stored in the database. In this library, the corresponding element conversion coefficient is extracted according to the material type record matched with the alloy element proportion, and then the element conversion coefficient representing the intrinsic properties of the material is linearly integrated with the uniformity feature of the measurement point distribution to obtain a first coefficient set. It should be noted that the element conversion coefficient is a proportional factor representing the influence degree of the content of each alloy element on the hardness conversion relationship. For example, if three materials are matched, the Cr content is close, the Mo and V content is slightly different, the corresponding conversion coefficients K are [0.081, 0.084, 0.083], and U=0.18, then the first coefficient set after integration is [0.081, 0.084, 0.083, 0.18].

[0122] Then, based on the first coefficient set, the spatial correlation features of the alloy element proportion and the measurement point distribution are further analyzed to obtain a first kernel function weight. It should be noted that this step uses a Gaussian kernel function: wherein is the distance between adjacent points a first kernel function weight of a target point s, is the Euclidean distance between two points, and σ is the bandwidth parameter of the kernel function, which is set according to the average distance between the measurement points, specifically, the average distance divided by 2. The first kernel function weight is spliced with the first coefficient set to obtain a second coefficient set.

[0123] Subsequently, the second coefficient set is smoothed by a regularization method through the formula: to obtain a third coefficient set. Wherein is the third coefficient set after smoothing, is the second coefficient set, is a regularization parameter, which is determined by cross-validation minimization of prediction error, and the preferred value range is [0.01, 0.5], is a penalty term, .

[0124] Finally, according to the third coefficient set, a spatial offset adjustment function is established: to calculate the coefficient offset caused by uneven distribution of measurement points, wherein is the offset sensitivity coefficient, which is determined by regression analysis and error minimization fitting on experimental data under different measurement point distribution conditions, and is preferably 0.05, and d is the average distance between measurement points. The offset is weighted and fused with the first kernel function weight to obtain an optimized conversion coefficient. It should be noted that the formula: is used to fuse the offset and the first kernel function weight, wherein is the optimized conversion coefficient, T(s) is the offset, is the fusion factor, which is determined by cross-validation by comparing the prediction errors of different weight fusion results in multiple historical samples, and the preferred value range is [0.5, 1.0].

[0125] For example, the basic conversion coefficient of a certain boiler pipe wall is 1.120, the system integrates the uniformity characteristics of the measurement point distribution to obtain 0.85, and the preliminary coefficient is 1.093. Subsequently, the spatial weight of the target point and the surrounding 8 adjacent points is calculated by the Gaussian kernel function, and the weighted average coefficient is 1.097. Subsequently, the regularization smoothing processing is adopted, the regularization parameter is taken as 0.1, and the coefficient is corrected to 1.095. Finally, according to the offset caused by uneven distribution of measurement points, the fusion factor is taken as 0.7, and the optimized conversion coefficient is 1.094.

[0126] Wherein, the optimized conversion coefficient is subjected to feature enhancement processing to obtain a final conversion coefficient, including:

[0127] obtaining a material type record matched with the alloy element proportion from the material characteristic library, extracting and integrating distribution characteristics of corresponding sampling frequency and probe contact pressure values to obtain a first feature set;

[0128] According to the first feature set, a second kernel function weight corresponding to the sampling frequency is obtained to obtain a second feature set;

[0129] The second feature set is smoothed to obtain a third feature set;

[0130] According to the third feature set, the optimized conversion coefficient is adjusted, and the second kernel function weight is fused to obtain a final conversion coefficient.

[0131] First, a material type record matched with the alloy element proportion is obtained from the material characteristic library, and corresponding sampling frequency and probe contact pressure values are extracted in the record. Then, the distribution statistical characteristics of probe contact pressure on all measurement points on the surface of the component are calculated, including mean, standard deviation and skewness. Finally, the sampling frequency and the pressure distribution characteristics are vector spliced to form a first feature set.

[0132] Then, according to the first feature set, a second kernel function weight corresponding to the sampling frequency is obtained to obtain a second feature set. It should be noted that, first, the sampling frequency in the first sampling set is normalized by formula to calculate the frequency ratio, wherein is the calculated frequency ratio, is the sampling frequency in the first feature set, is the standard frequency, which is determined by selecting the mean of the frequency that minimizes the hardness measurement error in the statistical analysis of multiple sets of material sampling frequency under stable measurement conditions, and then calculating the second kernel function weight by Gaussian kernel function , wherein, is the calculated second kernel function weight, is the kernel bandwidth, which is determined by cross-validation minimization of prediction error on historical data, and the preferred value range is [0.08, 0.30], is the calculated frequency ratio. Finally, the calculated second kernel function weight is multiplied by the first feature set to obtain a second feature set.

[0133] Then, the second feature set is smoothed to obtain a third feature set. It should be noted that this step still uses the regularization smoothing method for smoothing, and the processing formula is the same as the above step, which is determined by cross-validation minimization of prediction error, and the preferred value range is [0.2, 0.4] to obtain the smoothed third feature set.

[0134] Finally, according to the third feature set, the optimized conversion coefficient is adjusted, and the second kernel function weight is fused to obtain the final conversion coefficient. It should be noted that the final conversion coefficient is obtained by calculating the offset of the optimized conversion coefficient and weighting and fusing the offset and the second kernel function weight. The calculation formula of the offset and the weighting and fusing formula are the same as the above steps. The offset sensitivity coefficient is determined by regression analysis and error minimization fitting of experimental data under different measurement point distribution conditions. The fusion factor is determined by comparing the prediction errors of different weight fusion results in a plurality of experimental data by cross-validation.

[0135] In step S15, the support vector regression is used to fit the Leeb hardness correction value and the final conversion coefficient to obtain the Brinell hardness prediction value, including:

[0136] Integrate the Leeb hardness correction value and the measurement point distribution features to obtain a first hardness value set;

[0137] According to the first hardness value set, the Leeb hardness correction value and the final conversion coefficient are fitted by support vector regression to obtain a third kernel function weight, and a second hardness value set is obtained;

[0138] The second hardness value set is smoothed to obtain a third hardness value set;

[0139] According to the third hardness value set, an interpolation processing is performed to obtain a Brinell hardness prediction value in combination with the third kernel function weight.

[0140] First, the Leeb hardness correction value and the measurement point distribution features are integrated to obtain a first hardness value set. It should be noted that the spatial coordinates, the Leeb hardness correction value and the final conversion coefficient of each measurement point are collected, the spatial distance between each measurement point and the center of the overall region is calculated, and the features are integrated into a feature vector: wherein, is the Leeb hardness correction value, is the final conversion coefficient, is the distance between the measurement point and the center of the overall region, and the feature vectors of all measurement points form the first hardness value set.

[0141] Then, according to the first hardness value set, a pre-trained support vector regression model is used for nonlinear fitting. It should be noted that the support vector regression model uses a radial basis function as a kernel function, and the function expression is wherein is the kernel function parameter, which is determined by regression analysis and error minimization fitting of experimental data under different The cross-validation results of the predicted error are compared, and the value that minimizes the average error is selected, and the preferred value range is [0.05, 0.1]. Then, according to the training sample , the optimization objective function is trained, wherein , , is the Leeb hardness correction value, is the final conversion coefficient, is the measured Brinell hardness value, and the optimization objective function satisfies: , , , wherein is a penalty factor, preferably in the range of [100, 200], is an insensitive interval, preferably in the range of [0.1, 0.5], is mapped to a high-dimensional space by a kernel function. After training, the kernel weight corresponding to each support vector and the predicted hardness value are output, and the obtained kernel weight set is the third kernel function weight, and the predicted hardness value set is the second hardness value set.

[0142] It should be noted that the pre-training process of the support vector regression model includes: on a plurality of typical power plant boiler heat-resistant alloy steel parts, the Leeb hardness correction value and the final conversion coefficient obtained by the foregoing steps (S11 to S14) are collected as input features at the same time, and the corresponding Brinell hardness measured value is obtained by offline destructive test or measurement on the calibrated desktop Brinell hardness tester at the same point, as the target label for model training, thereby completing the construction of the training data set. The data set needs to cover different material types, service states and environmental conditions to ensure the generalization ability of the model; then the data set is randomly divided into a training set and a test set in a ratio of 7:3, the given super parameter optimization range is used, the average mean square error is used as the evaluation index, the model performance is evaluated by 5-fold cross-validation, the optimal super parameter combination is determined by grid search combined with cross-validation, and the model is trained.

[0143] Then, the second hardness value set is smoothed by a regularization method, and the calculation formula is , wherein is a regularization parameter, and the error of the smoothed prediction result is compared by taking different values, and the value that minimizes the error is selected, and the preferred value range is [0.2, 0.5], , and the calculated result is the third hardness value set.

[0144] Finally, based on the third hardness value set, interpolation processing is performed combined with the third kernel function weight to obtain the predicted Brinell hardness value. It should be noted that the interpolation function is wherein is a third kernel function weight, is a Brinell hardness value in the third hardness value set. Illustratively, if the Brinell hardness values of the three adjacent measurement points in the third hardness value set are [202, 198, 205], the corresponding kernel weights are [0.32, 0.27, 0.41], then the interpolation prediction is: = 0.32 x 202 + 0.27 x 198 + 0.41 x 205 = 202.9 HB.

[0145] In step S16, a Brinell hardness theoretical value corresponding to the alloy element proportion is obtained, and a residual life index is calculated based on the Brinell hardness prediction value, including:

[0146] obtaining the Brinell hardness theoretical value corresponding to the alloy element proportion from a preset heat-resistant alloy database;

[0147] calculating the residual life index based on the Brinell hardness theoretical value and the Brinell hardness prediction value.

[0148] First, a heat-resistant alloy database is established based on experimental data and material detection results. The database contains multiple typical heat-resistant alloy steel samples. Each group of samples records the alloy element proportion and the corresponding Brinell hardness theoretical value. The alloy element proportion is used as a retrieval condition to obtain the corresponding Brinell hardness theoretical value.

[0149] Subsequently, the residual life index is calculated based on the Brinell hardness theoretical value and the Brinell hardness prediction value. It should be noted that the determination of the residual life index is based on an empirical formula: is calculated, wherein is the calculated residual life index, is the Brinell hardness prediction value, is the Brinell hardness theoretical value, is a life correction coefficient determined based on the service time of the component and the rated life time. The specific calculation method is to subtract the quotient of the service time divided by the rated life time from 1, and the calculated difference is used as the life correction coefficient. It should be noted that the rated life time refers to the usable time of the material under the designed rated working condition, which can be obtained from a material manual or industry standard. Illustratively, the Brinell hardness theoretical value is 200 HB, the current Brinell hardness prediction value is 184 HB, the service time is 208 days, and the rated service life is 1000 days. Substituting the formula calculation can obtain the residual life index as =184 / 200×(1-208 / 1000)=0.73 (rounded to two decimal places). Based on the calculated remaining lifespan index, a decision is made in conjunction with a pre-determined decision rule. This decision rule can be determined based on historical experience. By analyzing the operation and maintenance records of each component in historical data, rules are generated to guide operation and maintenance decisions. For example, the decision rule is: when... When the value is ≥0.7, the component is in a safe state and requires no action; when 0.4≤ When the value is <0.7, the component performance may degrade, requiring close monitoring; when 0.2 ≤ When <0.4, the component may have a problem and needs immediate inspection; when When the value is less than 0.2, the availability of the component is low, and a replacement assessment should be initiated. Therefore, the calculated remaining life index can guide engineering decisions and ensure the normal operation of the project.

[0150] It should be noted that the above formula for calculating remaining lifetime indicators is determined within a reasonable margin of error. For scenarios requiring more accurate lifetime prediction, more precise formulas can be introduced for calculation, such as: ,in , , and The meaning of is the same as the meaning of the parameters in the life index calculation formula mentioned above; It is a non-linear index of hardness decay, reflecting the degree of accelerated hardness decay with aging. Its value is 1.2-2.0, which can be obtained by referring to steel grade-specific reports or mainstream engineering experience values. It is the material activation energy, which can be obtained by consulting the material handbook; It is the universal gas constant; It is the cumulative average temperature during operation; It is a time aging index, generally measured by... Obtained through experience matching, for example when At ≥600℃, Take 1.4, 500℃≤ At <600℃, Take 1.25, At <500℃, Take 1.15.

[0151] In summary, the present application discloses a kind of based on big model's power station boiler heat-resistant alloy steel hardness detection method, comprising: obtaining original data, the original value of Rockwell hardness in the original data is corrected to surface roughness influence, obtain initial dataset;The original data includes: Rockwell hardness original value, alloy element proportion, vibration interference information and temperature fluctuation record;According to the alloy element proportion, the initial dataset is grouped and processed, the Rockwell hardness value under different surface roughness of each group is calculated and corrected for environmental noise interference, obtain Rockwell hardness reference value under different material types;According to the vibration interference information and the temperature fluctuation record, the Rockwell hardness reference value is corrected to temperature fluctuation and vibration interference, obtain Rockwell hardness correction value;Obtain the conversion coefficient matched with the Rockwell hardness correction value and carry out dynamic optimization and feature enhancement processing, obtain final conversion coefficient;Using support vector regression, the Rockwell hardness correction value and the final conversion coefficient are fitted and processed, obtain Brinell hardness prediction value;Obtain the Brinell hardness theoretical value corresponding to the alloy element proportion, combined with the Brinell hardness prediction value, calculate to obtain residual life index.The present application realizes high-precision conversion from Rockwell hardness to Brinell hardness through hardness detection, and improves the reliability of heat-resistant alloy steel residual life prediction.

[0152] Reference Figure 3 The second embodiment of the present application provides a kind of based on big model's power station boiler heat-resistant alloy steel hardness detection system, comprising:

[0153] Data acquisition and correction module, for obtaining original data, the original value of Rockwell hardness in the original data is corrected to surface roughness influence, obtain initial dataset;

[0154] Reference value determination module, for according to the alloy element proportion, the initial dataset is grouped and processed, the Rockwell hardness value under different surface roughness of each group is calculated and corrected for environmental noise interference, obtain Rockwell hardness reference value under different material types;

[0155] Reference value correction module, for according to the vibration interference information and the temperature fluctuation record, the Rockwell hardness reference value is corrected to temperature fluctuation and vibration interference, obtain Rockwell hardness correction value;

[0156] Conversion coefficient determination module, for obtaining the conversion coefficient matched with the Rockwell hardness correction value and carrying out dynamic optimization and feature enhancement processing, obtain final conversion coefficient;

[0157] Hardness prediction module, for using support vector regression, the Rockwell hardness correction value and the final conversion coefficient are fitted and processed, obtain Brinell hardness prediction value;

[0158] A life assessment module is configured to obtain a Brinell hardness theoretical value corresponding to the alloy element proportion, and combine the Brinell hardness predicted value to calculate a residual life index.

[0159] It should be noted that the large model-based power station boiler heat-resistant alloy steel hardness detection system provided by the embodiments of the present application is used to execute all process steps of the large model-based power station boiler heat-resistant alloy steel hardness detection method provided by the above embodiments, and the working principles and beneficial effects of the two are one-to-one correspondence, so they will not be repeated.

[0160] The embodiments of the present application also provide an electronic device. The electronic device includes a processor, a memory, and a computer program, such as a data acquisition program, stored in the memory and executable on the processor. The processor implements the steps in each of the above various large model-based power station boiler heat-resistant alloy steel hardness detection method embodiments when executing the computer program, such as Figure 1 The step S11 shown. Alternatively, the processor implements the functions of each module / unit in each of the above device embodiments when executing the computer program, such as the data acquisition and correction module.

[0161] For example, the computer program can be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present application. The one or more modules / units can be a series of computer program instruction segments capable of completing a specific function, which are used to describe the execution process of the computer program in the electronic device.

[0162] The electronic device can be a desktop computer, a notebook, a palm computer, and a smart tablet, etc. The electronic device can include, but is not limited to, a processor, a memory. Those skilled in the art can understand that the above components are only examples of the electronic device and do not constitute a limitation on the electronic device, and can include more or fewer components than the above, or combine certain components, or different components, for example, the electronic device can also include an input / output device, a network access device, a bus, etc.

[0163] The processor can be a central processing unit (CPU), and can also be other general-purpose processors, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor. The processor is a control center of the electronic device, and connects various parts of the electronic device through various interfaces and lines.

[0164] The memory can be used to store the computer program and / or modules, and the processor realizes various functions of the electronic device by running or executing the computer program and / or modules stored in the memory, and calling data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application program required by a function (such as a sound playing function, an image playing function, etc.), etc.; and the data storage area can store data created according to the use of the mobile phone (such as audio data, a phone book, etc.), etc. In addition, the memory can include a high-speed random access memory, and can also include a nonvolatile memory, for example, a hard disk, a memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, at least one disk storage device, a flash memory device, or other volatile solid-state memory device.

[0165] The modules / units integrated in the electronic device, if realized in the form of software function units and sold or used as independent products, can be stored in a computer readable storage medium. Based on such understanding, all or part of the processes in the above-mentioned embodiment methods can also be completed by a computer program instructing related hardware, and the computer program can be stored in a computer readable storage medium. The computer program can implement the steps of each method embodiment when executed by a processor. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or some intermediate forms, etc. The computer readable medium can include any entity or device, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium, etc. that can carry the computer program code. It should be noted that the contents included in the computer readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction, for example, in some jurisdictions, according to legislation and patent practice, the computer readable medium does not include electrical carrier signals and telecommunication signals.

[0166] It should be noted that the above-described device embodiments are only schematic, and the units described as separate components can or can not be physically separated, and the components shown as units can or can not be physical units, that is, they can be located in one place, or distributed on multiple network units. Part or all of the modules can be selected according to actual needs to achieve the purpose of the embodiment. In addition, the connection relationship between the modules in the device embodiment provided by the present application indicates that there is a communication connection between them, which can be realized as one or more communication buses or signal lines. Those skilled in the art can understand and implement it without creative labor.

[0167] The above-described specific embodiments further illustrate the purpose, technical solutions and beneficial effects of the present application. It should be understood that the above-described specific embodiments are only for the specific embodiments of the present application and do not limit the protection scope of the present application. It is particularly pointed out that any modification, equivalent replacement, improvement, etc. made by those skilled in the art within the spirit and principles of the present application should be included in the protection scope of the present application.

Claims

1. A method for detecting the hardness of heat-resistant alloy steel of a power plant boiler based on a large model, characterized by, The method comprises the following steps: obtaining original data, correcting the original hardness value in the original data according to the influence of surface roughness to obtain an initial data set; the original data comprises: original hardness value, alloy element proportion, vibration interference information and temperature fluctuation record; grouping the initial data set according to the alloy element proportion, calculating the hardness value of each group under different surface roughness and correcting the environmental noise interference to obtain the hardness reference value of different material types; correcting the hardness reference value according to the vibration interference information and the temperature fluctuation record to obtain the corrected hardness value; obtaining the conversion coefficient matched with the corrected hardness value and performing dynamic optimization and feature enhancement processing to obtain the final conversion coefficient; using support vector regression to fit the corrected hardness value and the final conversion coefficient to obtain the predicted Brinell hardness value; obtaining the theoretical Brinell hardness value corresponding to the alloy element proportion, combining the predicted Brinell hardness value to calculate the residual life index.

2. The large model-based power plant boiler heat-resistant alloy steel hardness detection method according to claim 1, characterized in that, The method comprises the following steps: obtaining original data, correcting the original hardness value in the original data according to the influence of surface roughness to obtain an initial data set; obtaining the original hardness value, the alloy element proportion, the vibration interference information and the temperature fluctuation record to obtain the original data; measuring the roughness parameter and combining the probe contact pressure to calculate the roughness correction coefficient; 3. The large model-based power plant boiler heat-resistant alloy steel hardness detection method according to claim 2, characterized in that, correcting the original hardness value according to the roughness correction coefficient to obtain the initial hardness value, combining the alloy element proportion, the vibration interference information and the temperature fluctuation record to obtain the initial data set. The method comprises the following steps: performing frequency domain processing on the initial data set, extracting the interference characteristics of environmental noise and performing noise reduction processing to obtain a noise reduction data set; determining the interference correction coefficient according to the interference characteristics; extracting the feature vector of the alloy element proportion from the noise reduction data set and performing feature screening to obtain a feature data set; grouping the feature data set to obtain a grouped data set; 4. The large model-based power plant boiler heat-resistant alloy steel hardness detection method according to claim 1, characterized in that, calculating the hardness value of each group of materials under the influence of different surface roughness, combining the interference correction coefficient to obtain the hardness reference value of different material types. The method comprises the following steps: extracting the vibration interference characteristics according to the vibration interference information; extracting the temperature fluctuation characteristics according to the temperature fluctuation record; calculating the comprehensive correction coefficient according to the vibration interference characteristics and the temperature fluctuation characteristics; correcting the hardness reference value according to the comprehensive correction coefficient to obtain the corrected hardness value.

5. The large model-based power plant boiler heat-resistant alloy steel hardness detection method according to claim 1, characterized in that, The acquisition and the matching of the conversion coefficient of the Rockwell hardness correction value and the dynamic optimization and feature enhancement processing are obtained, including: From the pre-set Brinell hardness database, the material type record matched with the Rockwell hardness correction value is obtained, the corresponding material conversion coefficient is extracted, and the first conversion coefficient is obtained. Adjust the first conversion coefficient according to the environmental noise interference, and obtain the second conversion coefficient. Smooth the second conversion coefficient, and obtain the third conversion coefficient. Calculate the matching degree of the Rockwell hardness correction value and the material type, if the matching degree is lower than the pre-set matching degree threshold, the dynamic optimization and feature enhancement processing is carried out, and the final conversion coefficient is obtained, if the matching degree is not lower than the matching degree threshold, the third conversion coefficient is determined as the final conversion coefficient.

6. The large model-based power plant boiler heat-resistant alloy steel hardness detection method according to claim 5, characterized in that, The dynamic optimization and feature enhancement processing is carried out, and the final conversion coefficient is obtained, including: Obtain the uniformity feature of the measurement point distribution; From the pre-set material characteristic library, the material type record matched with the alloy element proportion is obtained, the corresponding element conversion coefficient is extracted, the element conversion coefficient and the uniformity feature are integrated, and the first coefficient set is obtained. Analyze the spatial correlation feature of the alloy element proportion and the measurement point distribution, obtain the first kernel function weight corresponding to the measurement point distribution, adjust the first coefficient set according to the first kernel function weight, and obtain the second coefficient set. Smooth the second coefficient set, and obtain the third coefficient set. According to the third coefficient set, adjust the offset of the third conversion coefficient with the measurement point distribution, fuse the first kernel function weight, and obtain the optimized conversion coefficient. The feature enhancement processing is carried out on the optimized conversion coefficient, and the final conversion coefficient is obtained.

7. The large model-based power plant boiler heat-resistant alloy steel hardness detection method according to claim 6, characterized in that, The feature enhancement processing is carried out on the optimized conversion coefficient, and the final conversion coefficient is obtained, including: From the material characteristic library, the material type record matched with the alloy element proportion is obtained, the distribution features of the sampling frequency and the probe contact pressure value are extracted and integrated, and the first feature set is obtained. According to the first feature set, the second kernel function weight corresponding to the sampling frequency is obtained, and the second feature set is obtained. Smooth the second feature set, and obtain the third feature set. According to the third feature set, adjust the optimized conversion coefficient, fuse the second kernel function weight, and obtain the final conversion coefficient.

8. The large model-based power plant boiler heat-resistant alloy steel hardness detection method according to claim 1, characterized in that, The Rockwell hardness correction value and the final conversion coefficient are fitted by support vector regression, and the Brinell hardness prediction value is obtained, including: Integrate the Rockwell hardness correction value and the feature of the measurement point distribution, and obtain the first hardness value set. According to the first hardness value set, the Rockwell hardness correction value and the final conversion coefficient are fitted by support vector regression, the third kernel function weight is obtained, and the second hardness value set is obtained. Smooth the second hardness value set, and obtain the third hardness value set. According to the third hardness value set, interpolation processing is carried out, and the third kernel function weight is combined to obtain the Brinell hardness prediction value.

9. The large model-based power plant boiler heat-resistant alloy steel hardness detection method according to claim 1, characterized by, The Brinell hardness theoretical value corresponding to the alloy element proportion is obtained, and a residual life index is calculated by combining the Brinell hardness prediction value, including: obtaining the Brinell hardness theoretical value corresponding to the alloy element proportion from a preset heat-resistant alloy database; calculating the residual life index according to the Brinell hardness theoretical value and the Brinell hardness prediction value.

10. A power plant boiler heat-resistant alloy steel hardness detection system based on a large model, characterized in that, The method for detecting the hardness of heat-resistant alloy steel of a power station boiler based on a large model comprises the following steps: a data acquisition and correction module is configured to obtain original data, and to correct the original data in the original data set to obtain an initial data set; a reference value determination module is configured to group the initial data set according to the alloy element proportion, to calculate the Leeb hardness value of each group under different surface roughness, and to correct the Leeb hardness value for environmental noise interference to obtain the Leeb hardness reference value under different material types; a reference value correction module is configured to correct the Leeb hardness reference value for temperature fluctuation and vibration interference according to the vibration interference information and the temperature fluctuation record to obtain a Leeb hardness correction value; a conversion coefficient determination module is configured to obtain a conversion coefficient matched with the Leeb hardness correction value and to perform dynamic optimization and feature enhancement processing to obtain a final conversion coefficient; a hardness prediction module is configured to perform fitting processing on the Leeb hardness correction value and the final conversion coefficient by using support vector regression to obtain a Brinell hardness prediction value; a life evaluation module is configured to obtain the Brinell hardness theoretical value corresponding to the alloy element proportion, and to calculate a residual life index by combining the Brinell hardness prediction value.

Images