A seamless steel pipe grinding process detection method and system based on machine vision

Abstract

The application relates to the technical field of quality detection, and discloses a seamless steel pipe grinding process detection method and system based on machine vision. The method comprises the following steps: acquiring detection data in a steel pipe grinding process under different abrasive particle sizes, inputting the detection data into a neural network model, extracting process variables and result variables of the grinding process, constructing a data relationship scatter diagram according to the extracted variables, performing trend analysis to calculate a dispersion degree, calculating a dispersion degree coefficient and an adjustment coefficient based on the dispersion degree, generating an overall deviation distribution scatter point value, comparing the overall deviation distribution scatter point value with a preset threshold value, and judging whether the grinding process is qualified. The method has the following effects: through multilevel analysis of data and capture of correlation, the accuracy and precision of detection are effectively improved, the reliability of grinding process evaluation is improved, and the demand of high-precision process control is met.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image processing, and in particular to a seamless steel pipe grinding process detection method and system based on machine vision. BACKGROUND

[0002] At present, in the production and processing process of seamless steel pipes, the grinding process is a key step to remove the oxide skin on the surface of the steel pipe and improve the surface quality. In this process, process adjustment needs to be made according to the grinding condition of the steel pipe surface, and the abrasive grain size in the grinding process is the main control parameter affecting the effect. Different abrasive grain sizes will cause different stress distribution and oxide skin distribution on the surface of the steel pipe, thereby affecting the final grinding effect. Therefore, by obtaining and analyzing the detection data corresponding to different abrasive grain sizes, generating the corresponding stress distribution map and oxide skin distribution map, and determining the optimal abrasive grain size, it has become an important link in process control.

[0003] The prior art usually judges the consistency of data by calculating the dispersion degree variance of detection data of different abrasive grain sizes, and takes it as an evaluation index. However, this method only focuses on the dispersion degree of data, ignores the correlation between detection data of different abrasive grain sizes, and is difficult to reflect the overall distribution characteristics of data, thereby resulting in low detection accuracy and precision, and being difficult to meet the high-precision process control demand. SUMMARY

[0004] The present application provides a seamless steel pipe grinding process detection method and system based on machine vision, which solves the problem of low detection accuracy and precision caused by only focusing on the dispersion degree of detection data and ignoring the correlation between detection data of different abrasive grain sizes in the prior art.

[0005] In a first aspect, to solve the above technical problems, the present application provides a seamless steel pipe grinding process detection method based on machine vision, comprising:

[0006] Obtaining detection data in the grinding process of the steel pipe under different abrasive grain sizes, the detection data including noise data generated in the grinding process and data after the grinding is completed;

[0007] Inputting the detection data into a neural network model, analyzing and extracting process variables and result variables of the grinding process;

[0008] According to the process variables and result variables, a data relationship scatter plot of the grinding process is constructed;

[0009] Trend analysis is performed on the data points in the data relationship scatter plot, the dispersion degree under each trend is calculated, and the dispersion degree value of the data relationship is obtained;

[0010] calculate a dispersion degree coefficient based on the dispersion degree value, obtain an adjustment coefficient of the dispersion degree of the data relationship by combining the dispersion degree coefficient and the distribution density of the points in the data relationship scatter plot;

[0011] generate the overall deviation distribution scatter value of the grinding process according to the dispersion degree value, the dispersion degree coefficient and the adjustment coefficient;

[0012] compare the overall deviation distribution scatter value with a preset threshold value, if the deviation distribution scatter value is lower than or equal to the preset threshold value, determine that the grinding process is qualified, if the deviation distribution scatter value is higher than the preset threshold value, determine that the grinding process is unqualified.

[0013] Preferably, the detection data is input into a neural network model, process variables and result variables of the grinding process are analyzed and extracted, including:

[0014] perform data preprocessing on the detection data to obtain preprocessed detection data, the data preprocessing includes removing noise and outliers in the detection data;

[0015] analyze the grinding process characteristics under different abrasive grain sizes based on the preprocessed detection data to obtain key parameters related to the grinding effect in the preprocessed detection data;

[0016] extract process noise and late noise features for combination according to the key parameters related to the grinding effect in the preprocessed detection data to obtain overall detection data;

[0017] input the overall detection data into a pre-trained neural network model to obtain process variables and result variables of the grinding process.

[0018] Preferably, a data relationship scatter plot of the grinding process is constructed according to the process variables and the result variables, including:

[0019] analyze the mutual relationship between the process variables and the result variables to obtain a correlation mode related to the grinding process effect;

[0020] perform data pairing on the process variables and the result variables based on the correlation mode to generate a plurality of data pairs, the data pairs can reflect the grinding process characteristics under different abrasive grain size conditions;

[0021] construct a data relationship scatter plot of the grinding process by taking the data pairs as coordinate points.

[0022] Preferably, trend analysis is performed on the data points in the data relationship scatter plot, the dispersion degree under each trend is calculated to obtain a dispersion degree value of the data relationship, including:

[0023] According to the change trend of the data points in the data relationship scatter diagram, the data points are divided into a gentle trend and a fluctuant trend, the gentle trend represents a region with smaller changes, and the fluctuant trend represents a region with larger changes;

[0024] According to different trends, the change amount of the overall change trend and the corresponding change times are calculated to quantify the fluctuation in different trends;

[0025] Based on the change amount of the overall change trend and the corresponding change times, the dispersion degree value in different trends is calculated by the following formula:

[0026]

[0027] Wherein, D r is the dispersion degree value, T represents different trends, the different trends include a gentle trend and a fluctuant trend, ΔV T represents the change amount of the overall change trend under the corresponding trend T, N T represents the corresponding change times of the change amount under the corresponding trend T, f(ΔV T ) and g(N T ) represent two linear normalization functions under the corresponding trend T, which are calculated by the following formula:

[0028]

[0029] Wherein, min(ΔV T ) is the minimum value of ΔV T , max(ΔV T ) is the maximum value of ΔV T , min(N T ) is the minimum value of N T , and max(N T ) is the maximum value of N T .

[0030] Preferably, a dispersion degree coefficient is calculated based on the dispersion degree value, and a regulation coefficient of the corresponding data relationship dispersion degree is obtained by combining the dispersion degree coefficient with the distribution density of the points in the data relationship scatter diagram, including:

[0031] Based on the dispersion degree value, the dispersion degree coefficient is calculated by the following formula:

[0032] D c = f(D r )

[0033] Wherein, D c represents the dispersion degree coefficient, D r is the dispersion degree value, and f is a linear normalization function;

[0034] Based on the dispersion degree coefficient, a sliding window parameter is set, and the dispersion degree in different sliding windows is calculated in segments, the sliding window parameter including a window width and a sliding step;

[0035] Based on the dispersion degree coefficient, a dispersion value in each sliding window is calculated by the following formula:

[0036]

[0037] wherein, D s represents the dispersion value, max(D c ) and min(D c ) are the maximum and minimum of the dispersion degree coefficient in the sliding window, and N is the number of the dispersion degree coefficient in the sliding window;

[0038] Based on the dispersion value, a normalized dispersion value is calculated by the following formula:

[0039]

[0040] wherein, D sn represents the normalized dispersion value, max(D s ) and min(D s ) are the maximum and minimum of the dispersion value in all sliding windows;

[0041] Based on the dispersion degree coefficient and the normalized dispersion value, an adjustment coefficient of the dispersion degree is calculated by the following formula:

[0042]

[0043] wherein, D adj represents the adjustment coefficient, and N max is the maximum number of scatter points in the data relationship scatter plot.

[0044] Preferably, according to the dispersion value, the dispersion degree coefficient and the adjustment coefficient, an overall deviation distribution scatter value of the grinding process is generated, including:

[0045] The dispersion value and the dispersion degree coefficient are used as input parameters of the overall deviation distribution scatter value, representing the overall dispersion degree and local features of the data distribution;

[0046] Based on the dispersion value, the dispersion degree coefficient and the adjustment coefficient, an overall deviation distribution scatter value is calculated by the following formula:

[0047] D total =D r ×D c ×D adj

[0048] wherein D total represents the overall deviation distribution scatter value, D r is the dispersion value, D c represents the dispersion coefficient, D adj is the adjustment coefficient.

[0049] Preferably, the overall deviation distribution scatter value is compared with a preset threshold value, if the deviation distribution scatter value is lower than or equal to the preset threshold value, it is judged that the grinding process is qualified, if the deviation distribution scatter value is higher than the preset threshold value, it is judged that the grinding process is unqualified, comprising:

[0050] Collecting historical data in multiple grinding processes, and statistically processing the overall deviation distribution scatter value of each grinding process to form a distribution data set;

[0051] According to the distribution data set, the mean and standard deviation of the deviation distribution scatter value are calculated to obtain a preliminary threshold range;

[0052] According to the preliminary threshold range, the mean is weighted and adjusted to obtain the final preset threshold value of the deviation distribution scatter value;

[0053] The overall deviation distribution scatter value is compared with the preset threshold value, if the deviation distribution scatter value is lower than or equal to the preset threshold value, it is judged that the grinding process is qualified, if the deviation distribution scatter value is higher than the preset threshold value, it is judged that the grinding process is unqualified.

[0054] In a second aspect, the present application provides a seamless steel pipe grinding process detection system based on machine vision, which is used to realize the machine vision-based seamless steel pipe grinding process detection method in any one of the above aspects, comprising:

[0055] A data acquisition module is configured to acquire detection data in a steel pipe grinding process under different abrasive grain sizes, wherein the detection data includes noise data generated in the grinding process and data after the grinding process ends;

[0056] A feature extraction module is configured to input the detection data into a neural network model, analyze and extract process variables and result variables of the grinding process;

[0057] A data relationship construction module is configured to construct a data relationship scatter plot of the grinding process according to the process variables and the result variables;

[0058] A trend analysis module is configured to perform trend analysis on data points in the data relationship scatter plot, calculate the dispersion degree under each trend, and obtain the dispersion degree value of the data relationship;

[0059] The discrete degree adjustment module is configured to calculate a discrete degree coefficient based on the discrete degree value, combine the discrete degree coefficient with the distribution density of the data relationship scatter plot point, and obtain an adjustment coefficient corresponding to the data relationship discrete degree.

[0060] The overall deviation calculation module is configured to generate an overall deviation distribution scatter value of the grinding process according to the discrete degree value, the discrete degree coefficient, and the adjustment coefficient.

[0061] The process evaluation module is configured to compare the overall deviation distribution scatter value with a preset threshold value, and if the deviation distribution scatter value is lower than or equal to the preset threshold value, the grinding process is determined to be qualified, and if the deviation distribution scatter value is higher than the preset threshold value, the grinding process is determined to be unqualified.

[0062] In a third aspect, the present application further provides an electronic device comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor implements the machine vision-based seamless steel pipe grinding process detection method according to any one of the above embodiments when executing the computer program.

[0063] In a fourth aspect, the present application further provides a computer-readable storage medium comprising a stored computer program, wherein the computer-readable storage medium controls the device where the computer-readable storage medium is located to execute the machine vision-based seamless steel pipe grinding process detection method according to any one of the above embodiments when the computer program is running.

[0064] Compared with the prior art, the present application has the following beneficial effects:

[0065] The present application provides a machine vision-based seamless steel pipe grinding process detection method, which comprises the steps of real-time acquisition of detection data, analysis and training of a neural network model, construction of a data relationship scatter plot, trend analysis, and discrete degree calculation, and finally generates a deviation distribution scatter value and performs qualification evaluation. Through this technical solution, the data characteristics under different abrasive particle sizes can be comprehensively analyzed, the correlation between the data and the overall distribution characteristics are captured, thereby improving the accuracy of detection.

[0066] Compared with the prior art, the present application not only focuses on the discrete degree of data, but also considers the correlation between different abrasive particle size data and the overall distribution characteristics, which significantly improves the accuracy and precision of detection. In addition, when evaluating the grinding process effect, the present application not only can detect the quality parameters after grinding, but also can predict potential quality problems in the grinding process through real-time data analysis. Therefore, the present application effectively solves the deficiencies of existing methods in detection precision, reliability, and adaptability, and meets the demand of high-precision grinding process control. BRIEF DESCRIPTION OF DRAWINGS

[0067] Figure 1 is a flow chart of a seamless steel pipe grinding process detection method based on machine vision provided by the first embodiment of the present application;

[0068] Figure 2 is a structural schematic diagram of a seamless steel pipe grinding process detection method based on machine vision provided by the second embodiment of the present application. DETAILED DESCRIPTION

[0069] The technical solutions in the embodiments of the present application will be clearly and completely described in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0070] With reference to Figure 1 The first embodiment of the present application provides a seamless steel pipe grinding process detection method based on machine vision, comprising the following steps:

[0071] S11, acquiring detection data in a steel pipe grinding process under different abrasive particle sizes, wherein the detection data comprises noise data generated in the grinding process and data after the grinding is completed;

[0072] S12, inputting the detection data into a neural network model to analyze and extract process variables and result variables of the grinding process;

[0073] S13, constructing a data relationship scatter plot of the grinding process according to the process variables and the result variables;

[0074] S14, performing trend analysis on data points in the data relationship scatter plot, calculating the dispersion degree under each trend, and obtaining a dispersion degree value of the data relationship;

[0075] S15, calculating a dispersion degree coefficient based on the dispersion degree value, combining the dispersion degree coefficient with the distribution density of points in the data relationship scatter plot, and obtaining an adjustment coefficient corresponding to the dispersion degree of the data relationship;

[0076] S16, generating an overall deviation distribution scatter value of the grinding process according to the dispersion degree value, the dispersion degree coefficient and the adjustment coefficient;

[0077] S17, comparing the overall deviation distribution scatter value with a preset threshold value, if the deviation distribution scatter value is lower than or equal to the preset threshold value, judging that the grinding process is qualified, and if the deviation distribution scatter value is higher than the preset threshold value, judging that the grinding process is unqualified.

[0078] In step S11, the detection data of the steel pipe grinding process under different abrasive particle sizes is obtained, including noise data generated during the grinding process and data after the grinding is completed, including:

[0079] Using multiple detection devices, real-time noise data is collected during the grinding operation to obtain dynamic information of the grinding process under different abrasive particle size conditions. The process of collecting real-time noise data during the grinding operation can capture transient changes in the grinding process and reflect the characteristic information generated during the grinding process.

[0080] After the grinding operation is completed, the detection device is used to continue collecting noise data after grinding to obtain stable state information after the grinding process is completed. The noise data after grinding can reflect the characteristics of the grinding result and provide a basis for subsequent analysis.

[0081] The collected noise data is labeled and classified according to the grinding process and the state after grinding, and is sorted according to different abrasive particle size conditions to ensure that the noise characteristic data corresponding to each abrasive particle size can be accurately distinguished during subsequent analysis.

[0082] In step S12, the detection data is input into a neural network model to analyze and extract process variables and result variables of the grinding process, including:

[0083] S121, data preprocessing is performed on the detection data to obtain preprocessed detection data, and the data preprocessing includes removing noise and outliers in the detection data;

[0084] S122, based on the preprocessed detection data, analyze the grinding process characteristics under different abrasive particle sizes to obtain key parameters related to the grinding effect in the preprocessed detection data;

[0085] S123, according to the key parameters related to the grinding effect in the preprocessed detection data, extract process noise and post-noise characteristics for combination to obtain overall detection data;

[0086] S124, input the overall detection data as input data into the pre-trained neural network model to obtain process variables and result variables of the grinding process.

[0087] In step S121, the detection data is preprocessed to obtain preprocessed detection data, and the data preprocessing includes removing noise and outliers in the detection data, including:

[0088] For noise signals in the detection data, a multi-stage filtering processing technology is used, including high-pass filtering, low-pass filtering, and band-pass filtering, which removes irrelevant background noise and high-frequency or low-frequency interference signals layer by layer.

[0089] After noise filtering, in order to further reduce the random fluctuations of the data, a moving average method is used for processing, that is, by taking the average value of the data within a certain range to smooth the data, so that the data fluctuation is more continuous, and the influence of the mutation point on the analysis is reduced;

[0090] In order to ensure the accuracy and consistency of the data, various algorithms are used to detect and eliminate abnormal values in the data;

[0091] After removing noise and abnormal values, in order to maintain the integrity of the data, an interpolation algorithm is used to compensate for missing data points;

[0092] After completing noise removal and abnormal value elimination, the data is standardized to ensure consistency of data from different batches or different detection equipment.

[0093] In steps S122 and S123, based on the preprocessed detection data, the grinding process characteristics under different abrasive particle sizes are analyzed to obtain key parameters related to the grinding effect in the data; According to the key parameters related to the grinding effect in the data, the process noise and post-process noise characteristics are combined to obtain the overall detection data. In the grinding process, the key parameters include grinding force, grinding temperature, grinding speed and abrasive particle size, which have a direct impact on the grinding effect. Illustratively, statistical analysis methods are used to identify these key parameters. For example, the average value, standard deviation, maximum value and minimum value of each parameter can be calculated to understand the trend and distribution of the parameter. Assuming there is a set of grinding force data, the stability of the grinding force during the grinding process can be evaluated by calculating the standard deviation of the data, and if the standard deviation is small, it means that the grinding force is stable and the grinding quality is high.

[0094] Further, the process noise refers to the noise caused by the change of grinding parameters during the grinding process, while the post-process noise is related to the cooling, measurement and other processes after grinding. By combining the process noise and post-process noise characteristics, the noise characteristics of the overall detection data can be obtained, which can be achieved by weighted sum of the intensity of different noise characteristics. The weight can be determined according to the degree of influence of the noise on the grinding effect. For example, if the influence of the process noise on the grinding effect is greater than that of the post-process noise, the process noise can be given a greater weight when combining the noise characteristics. In this way, the overall detection data will better reflect the main noise source in the grinding process.

[0095] In step S124, the neural network model is pre-trained, including:

[0096] The overall detection data is divided according to the ratio of 7:2:1, of which 70% is used for model training, 20% for model validation, and 10% for final testing to ensure the stability and generalization ability of the model training;

[0097] According to the complexity of the grinding process feature data, a neural network model architecture is designed, wherein the input layer is used to receive the preprocessed overall detection data, the feature dimension matches the feature quantity of the detection data, the output layer sets the number of nodes according to the number of target output variables, and the output layer adopts linear or nonlinear activation function to adapt to different types of prediction tasks;

[0098] A suitable loss function is selected to measure the error between the predicted value of the model and the actual output;

[0099] An Adam optimizer or a stochastic gradient descent optimizer is used for model training;

[0100] The predicted value of the model is calculated by forward propagation, and the model parameters are adjusted by the back propagation algorithm to minimize the loss function;

[0101] After training is completed, the model is evaluated using a test set, and the prediction error and various evaluation indicators are calculated.

[0102] In step S13, the data relationship scatter plot of the grinding process is constructed according to the process variables and the result variables, including:

[0103] S131, analyze the mutual relationship between the process variables and the result variables to obtain a correlation pattern related to the grinding process effect;

[0104] S132, based on the correlation pattern, data pairing is performed on the process variables and the result variables to generate a plurality of data pairs, which can reflect the grinding process characteristics under different abrasive grain size conditions;

[0105] S133, taking the data pairs as coordinate points, a data relationship scatter plot of the grinding process is constructed.

[0106] In step S131, the correlation pattern is a mathematical or visual structure describing the internal correlation and mutual influence between variables, which is used to reveal the influence and trend of different parameters in the grinding process on the final process effect, including:

[0107] The correlation strength between variables: through data analysis, the correlation strength between the process variables and the result variables under different abrasive grain size conditions is quantified, which can help identify the control parameters that have the greatest impact on process quality, so that the key variables can be adjusted during process optimization;

[0108] Trend type and distribution characteristics: The correlation mode divides data points into flat trends and fluctuating trends through trend analysis, the flat trends represent more stable process, and the fluctuating trends represent fluctuations in the process, different trends reflect how the size of abrasive particles affects the uniformity and consistency of surface roughness and stress distribution, for example, smaller abrasive particle size produces more stable and continuous surface effect, while larger abrasive particle size leads to fluctuations in surface roughness and stress distribution;

[0109] Discrete degree and distribution density: The correlation mode can further describe the distribution of data in the scatter plot by calculating the discrete degree value and the discrete coefficient, which represents the uniformity of grinding effect under different abrasive particle size conditions, where dense and regular distribution of data points indicates better process consistency, and discrete and sparse points indicate unevenness of process results, through the statistics of the discrete degree, the correlation mode can intuitively reveal the stability and controllability of the process conditions;

[0110] Quantitative basis for process optimization: The generated correlation mode not only intuitively displays the relationship between process parameters and quality results, but also provides a quantitative basis for further process optimization, the correlation mode can be used as a judgment standard to help identify and adjust the optimal abrasive particle size to ensure the stability and consistency of the process effect.

[0111] In step S133, a data relationship scatter plot of the grinding process is constructed with the data pairs as coordinate points, including:

[0112] Two key variables closely related to the grinding effect are selected as the horizontal axis and the vertical axis, respectively, where the horizontal axis selects variables representing process control, such as grinding speed, abrasive particle size, and feed rate, which have a direct impact on the grinding process and can better reflect the dynamic process of grinding, and the vertical axis selects variables reflecting the results, such as surface roughness, stress distribution, and oxide removal effect, which represent the final effect of the grinding process and are important indicators for evaluating process quality.

[0113] It should be noted that the data point density in the scatter plot reflects the data distribution under different abrasive particle size conditions, and the display of density is crucial to understanding the uniformity of grinding process characteristics and effect:

[0114] Data density changes under different abrasive particle size conditions: In actual operation, different abrasive particle size conditions will produce different numbers of data points. For example, smaller abrasive particle size generates more concentrated and dense data points, reflecting the uniformity of the grinding process; while larger abrasive particle size leads to more sparse or fluctuating distribution of data points, reflecting the unevenness of the grinding process;

[0115] Density display method: Reflect the data density on the scatter plot by the color or transparency of the data points. The areas with higher density can be represented by dark or opaque points, while the areas with lower density are represented by light or semi-transparent points. In this way, the process personnel can intuitively understand the uniformity and stability of the grinding process under different abrasive grain size conditions by observing the color depth or transparency change in the scatter plot;

[0116] Statistics of density information: Further, in order to ensure the readability of the graph and the accuracy of the information, the point density under each abrasive grain size condition needs to be counted, such as calculating the number of data points in each unit area, and presenting the density distribution in the scatter plot with color gradient or point size. The above design can more accurately reflect the concentration and dispersion of data under different grinding conditions, thereby more effectively assisting process optimization.

[0117] In step S14, the data points in the data relationship scatter plot are analyzed for trends, the dispersion degree under each trend is calculated, and the dispersion degree value of the data relationship is obtained, including:

[0118] According to the change trend of the data points in the data relationship scatter plot, the data points are divided into a gentle trend and a fluctuating trend, the gentle trend represents an area with smaller changes, and the fluctuating trend represents an area with larger changes;

[0119] According to different trends, the change amount and the corresponding change number of the overall change trend are calculated, and the fluctuation under different trends is quantified;

[0120] Based on the change amount and the corresponding change number of the overall change trend, the dispersion degree value under different trends is calculated by the following formula:

[0121]

[0122] Wherein, D r is the dispersion degree value, T represents different trends, the different trends include a gentle trend and a fluctuating trend, ΔV T represents the change amount of the overall change trend under the corresponding trend T, N T represents the corresponding change number of the change amount under the corresponding trend T, f(ΔV T ) and g(N T ) represent two linear normalization functions under the corresponding trend T, which are calculated by the following formula:

[0123]

[0124] Wherein, min(ΔV T ) is the minimum value of ΔV T , max(ΔV T ) is the maximum value of ΔV Tmax(N T ) is the maximum value of N T , min(N T ) is the minimum value of N T , and max(N c ) is the maximum value of N r .

[0125] In this embodiment, the division of the flat trend and the fluctuant trend can be achieved by calculating the slope between data points. For each pair of consecutive data points in the scatter plot, the slope between them (i.e., the change divided by the time interval or position interval) is calculated. A threshold is set, and when the absolute value of the slope is lower than the threshold, the data point is considered to belong to the flat trend; when the absolute value of the slope is higher than the threshold, it is considered to belong to the fluctuant trend. For example, if the slope between data points (1, 2) and (2, 3) is 0.5, and the slope between data points (3, 3) and (4, 6) is 2.5, the change of the latter is greater than the preset threshold 2, and thus it belongs to the fluctuant trend.

[0126] In step S15, the dispersion degree coefficient is calculated based on the dispersion degree value, and the adjustment coefficient of the dispersion degree of the data relationship is obtained by combining the dispersion degree coefficient with the distribution density of the points in the data relationship scatter plot, including:

[0127] The dispersion degree coefficient is calculated based on the dispersion degree value by the following formula:

[0128] D c =f(D r )

[0129] wherein D c represents the dispersion degree coefficient, D r is the dispersion degree value, and f is a linear normalization function;

[0130] Based on the dispersion degree coefficient, the sliding window parameters are set, and the dispersion degree in different sliding windows is calculated segmentally, wherein the sliding window parameters include the window width and the sliding step;

[0131] The dispersion condition value in each sliding window is calculated based on the dispersion degree coefficient by the following formula:

[0132]

[0133] wherein D s represents the dispersion condition value, max(D c ) and min(D c ) are the maximum value and the minimum value of the dispersion degree coefficient in the sliding window, respectively, and N is the number of the dispersion degree coefficients in the sliding window;

[0134] The normalized dispersion condition value is calculated based on the dispersion condition value by the following formula:

[0135]

[0136] wherein D sn represents the normalized discrete condition value, max(D s ) and min(D s ) are the maximum and minimum values of the discrete condition value in all sliding windows, respectively;

[0137] Based on the discrete degree coefficient and the normalized discrete condition value, the adjustment coefficient of the discrete degree is calculated by the following formula:

[0138]

[0139] wherein D adj represents the adjustment coefficient, N max is the maximum number of scatter points in the data relationship scatter plot.

[0140] In step S16, the overall deviation distribution scatter point value of the grinding process is generated according to the discrete degree value, the discrete degree coefficient and the adjustment coefficient, including:

[0141] The discrete degree value and the discrete degree coefficient are used as input parameters of the overall deviation distribution scatter point value, representing the overall discrete degree and local characteristics of the data distribution;

[0142] Based on the discrete degree value, the discrete degree coefficient and the adjustment coefficient, the overall deviation distribution scatter point value is calculated by the following formula:

[0143] D total = D r × D c × D adj

[0144] wherein D total represents the overall deviation distribution scatter point value, D r is the discrete degree value, D c represents the discrete degree coefficient, and D adj is the adjustment coefficient.

[0145] It should be noted that the discrete degree value D r represents the discrete degree of the data points in the grinding process, i.e. the dispersion degree of the data points relative to the average value or the expected value. The higher the discrete degree value, the more dispersed the distribution of the data points, and the poorer the consistency of the grinding process. The discrete degree coefficient D c is used to adjust the discrete degree value to reflect the discrete degree under different grinding processes or different conditions. For example, different grinding speeds or grinding tools may affect the discrete degree, and the discrete degree coefficient can be adjusted according to different conditions to more accurately reflect the actual discrete degree. The adjustment coefficient Dadj The deviation distribution discrete point value is used to further fine-tune the body deviation. Thus, by multiplying the discrete degree value, the discrete degree coefficient and the adjustment coefficient, a comprehensive index can be obtained for evaluating the overall deviation degree of the grinding process. The higher this value is, the greater the deviation degree of the grinding process is, and process adjustment is needed to improve product quality.

[0146] In step S17, the overall deviation distribution discrete point value is compared with a preset threshold value. If the deviation distribution discrete point value is lower than or equal to the preset threshold value, it is judged that the grinding process is qualified. If the deviation distribution discrete point value is higher than the preset threshold value, it is judged that the grinding process is unqualified, including:

[0147] S171, historical data in multiple grinding processes are collected, the overall deviation distribution discrete point value of each grinding process is counted, and a distribution data set is formed;

[0148] S172, the mean value and the standard deviation of the deviation distribution discrete point value are calculated according to the distribution data set, and a preliminary threshold range is obtained;

[0149] S173, the mean value is weighted and adjusted according to the preliminary threshold range, and a final preset threshold value of the deviation distribution discrete point value is obtained;

[0150] S174, the overall deviation distribution discrete point value is compared with the preset threshold value. If the deviation distribution discrete point value is lower than or equal to the preset threshold value, it is judged that the grinding process is qualified. If the deviation distribution discrete point value is higher than the preset threshold value, it is judged that the grinding process is unqualified.

[0151] In the above steps, the mean value of the deviation distribution discrete point value is used to represent the average level of the overall deviation, and the standard deviation of the deviation distribution discrete point value is used to measure the fluctuation degree of the deviation distribution discrete point value. Then, the threshold value can be tightened to the sum of the mean value and the standard deviation multiplied by a preset adjustment coefficient, to obtain a preliminary threshold range. The preset adjustment coefficient represents the tolerance range of the deviation discrete point value, for example, it is set to 1, 2, 3, etc. according to the precision requirement, which is not limited in the present application.

[0152] For example, data in the past 100 grinding processes are collected. These data include the deviation between the actual size and the target size of the product after each grinding. For the 100 grindings, the deviation value of each grinding is calculated and recorded to form a distribution data set. For example, the deviation value set [0.02, -0.01, 0.03, -0.02, 0.00, …, 0.01]. The mean value (average deviation) and the standard deviation of this data set are calculated. Assuming that the calculation results are as follows: mean value (average deviation): 0.005 mm; standard deviation: 0.01 mm. Then, the mean value ± 3 times the standard deviation can be used to determine a preliminary threshold range, which is:

[0153] Lower limit: 0.005mm - 3*0.01mm = -0.025mm

[0154] Upper limit: 0.005mm + 3*0.01mm = 0.035mm

[0155] It should be noted that the present embodiment determines a weighting factor to adjust the mean value, so as to obtain the final preset threshold value of the scattered points of the deviation distribution. For example, considering the uncertainty of the process and the potential risks, a weighting factor less than 1 is selected to reduce the threshold value, so as to ensure higher product quality. Exemplarily, using a safety factor of 0.8 (i.e. 80% of 3 times the standard deviation), then the final threshold value is:

[0156] Lower limit: 0.005mm - 0.8*3*0.01mm = -0.019mm

[0157] Upper limit: 0.005mm + 0.8*3*0.01mm = 0.029mm

[0158] Therefore, the final preset threshold value range is [-0.019, 0.029].

[0159] In an implementation mode, the machine vision-based seamless steel pipe grinding process detection method of the present application is suitable for real-time quality detection in a seamless steel pipe production line. The surface grinding process of the seamless steel pipe directly affects the quality of the final product, such as the surface finish, the removal effect of the oxide skin, and the uniformity of the stress distribution. Therefore, it is crucial to conduct high-precision real-time monitoring during the grinding process. The present application uses a non-contact method to collect grinding process noise data under different abrasive grain sizes in real time by arranging multiple noise detection devices on the production line, including dynamic noise data generated during the grinding process and static noise data after the grinding is completed. These data truly reflect the dynamic change state and the final result state of the grinding process, providing a reliable data basis for comprehensive evaluation of the process quality.

[0160] After the data collection is completed, the obtained grinding process noise data will enter the data preprocessing stage. Through filtering, denoising and outlier rejection data preprocessing techniques, the environmental interference and accidental abnormal noise in the original grinding process noise data are removed, so as to ensure that the data more truly reflect the characteristics of the grinding process. This preprocessing step can improve the quality and consistency of the data, laying a solid foundation for subsequent data analysis. The preprocessed data to a certain extent reduces the noise interference, and more concentratedly embodies the influence of abrasive grain size on the grinding effect.

[0161] Next, the processed detection data is input into a well-trained neural network model. The neural network model has been trained based on data of different abrasive particle sizes and various grinding conditions, and can accurately identify and extract key process variables and result variables in the grinding process. The key process variables and result variables in the grinding process not only represent the current state of the grinding process, but also provide important data support for subsequent trend analysis and scatter plot construction, ensuring the accuracy and reliability of the analysis results.

[0162] After extracting the key process variables and result variables in the grinding process, the present application visualizes these variables through a data relationship scatter plot. The data relationship scatter plot is used to show the correlation between different variables. In the present application, process variables and result variables are used as coordinate axes, and the grinding effect under different abrasive particle sizes is intuitively displayed in the form of coordinate points. The data relationship scatter plot can clearly present the data distribution characteristics under different abrasive particle size conditions. For example, smaller abrasive particle sizes will correspond to smoother surface effects and uniform stress distribution, while larger abrasive particle sizes will lead to increased surface roughness and uneven stress distribution. Through the data relationship scatter plot, production operators can intuitively observe the relationship between different variables, providing a basis for process parameter optimization.

[0163] Subsequently, the system performs trend analysis on the data points in the data relationship scatter plot to calculate the dispersion degree under each trend. The process of trend analysis includes identifying flat trends and fluctuating trends, and quantitatively calculating these trends. The flat trend represents a region with small changes, and the fluctuating trend represents a region with large changes. The division of flat trends and fluctuating trends helps to reveal the fluctuation characteristics and consistency of data under different abrasive particle size conditions, reflecting the stability and controllability of the process under different conditions. By calculating the change amount and corresponding change times of the overall change trend, the system can quantify the data fluctuation under different trends.

[0164] Based on the results of trend analysis, the system will calculate the dispersion degree value. The dispersion degree value not only represents the volatility of the data, but also reflects the stability and consistency of the process. Next, by calculating the dispersion degree coefficient, the dispersion degree values under different trends are normalized, so that the data under different abrasive particle size conditions can be compared within a standard range. The calculation of the dispersion degree coefficient also takes into account the distribution density of the data points in the data relationship scatter plot, ensuring that the data under different process conditions can be compared fairly and accurately.

[0165] After obtaining the dispersion degree coefficient, the system will calculate the adjustment coefficient. The adjustment coefficient is based on the data distribution density and is calculated in combination with the dispersion degree value and the dispersion degree coefficient to ensure that the overall deviation distribution scatter point value can fully reflect the overall deviation of the current grinding process. By calculating the adjustment coefficient, the system can effectively adjust the deviation between the data, making the final deviation distribution scatter point value more representative.

[0166] Finally, the generated overall deviation distribution scatter point value is compared with a preset threshold to evaluate the eligibility of the grinding process. The preset threshold is set based on process requirements, historical data, and quality control standards. Comparing the overall deviation distribution scatter point value with the threshold, if the overall deviation distribution scatter point value is lower than or equal to the threshold, it indicates that the grinding process is within the allowable quality range and is judged to be eligible; otherwise, if the deviation value exceeds the threshold, the process quality is judged to be unqualified. This quality evaluation method combines multi-level data analysis and variable relationship calculation, significantly improving the accuracy and reliability of process quality judgment.

[0167] In another implementation, the detection method of the present application is not only suitable for the grinding process of seamless steel pipes, but also can be applied to the quality control of other metal surface processing processes, such as polishing, polishing process of castings or surface treatment of plates, etc. By adjusting the parameters of the data acquisition module and the feature extraction module, the system of the present application can adapt to different metal surface processing conditions, thereby providing a widely applicable quality control scheme. In other application scenarios, the present application can achieve high-precision process evaluation through similar noise detection equipment and neural network models, effectively identify potential quality problems in the production process, and reduce the defective rate of the later products.

[0168] To better illustrate the working principle of the present application, the implementation process of the present application is described below in combination with a specific application scenario. Assuming that in a typical seamless steel pipe grinding workshop, real-time data is collected at different stages of the grinding process by installing the detection system of the present application. During the grinding process, the detection equipment collects noise data in real time to analyze the current grinding state; after the grinding is completed, the system further collects noise data after grinding to compare the changes before and after the process. All data are transmitted to the data processing center through the network and input into the neural network model for analysis to extract the key process variables and result variables of the current process. Subsequently, the system generates a data relationship scatter plot for trend analysis and dispersion degree calculation, and finally obtains the overall deviation distribution scatter point value. The system feeds back the process quality judgment result to the operator in time according to the comparison result of the deviation distribution scatter point value and the threshold. If the process is judged to be unqualified, the system will issue an alarm so that the operator can immediately adjust the process.

[0169] In summary, with reference to Figure 2The second embodiment of the present application provides a seamless steel pipe grinding process detection system based on machine vision, comprising:

[0170] A data acquisition module is configured to acquire detection data in the steel pipe grinding process under different abrasive particle sizes, wherein the detection data includes noise data generated in the grinding process and data after the grinding is completed.

[0171] A feature extraction module is configured to input the detection data into a neural network model, analyze and extract process variables and result variables of the grinding process.

[0172] A data relationship construction module is configured to construct a data relationship scatter plot of the grinding process according to the process variables and the result variables.

[0173] A trend analysis module is configured to perform trend analysis on data points in the data relationship scatter plot, calculate the dispersion degree under each trend, and obtain a dispersion degree value of the data relationship.

[0174] A dispersion degree adjustment module is configured to calculate a dispersion degree coefficient based on the dispersion degree value, combine the dispersion degree coefficient with the distribution density of points in the data relationship scatter plot, and obtain an adjustment coefficient of the dispersion degree of the data relationship.

[0175] An overall deviation calculation module is configured to generate an overall deviation distribution scatter point value of the grinding process according to the dispersion degree value, the dispersion degree coefficient and the adjustment coefficient.

[0176] A process evaluation module is configured to compare the overall deviation distribution scatter point value with a preset threshold value, and if the deviation distribution scatter point value is lower than or equal to the preset threshold value, it is determined that the grinding process is qualified, and if the deviation distribution scatter point value is higher than the preset threshold value, it is determined that the grinding process is unqualified.

[0177] Preferably, the data acquisition module includes multiple detection devices arranged at different stations of the grinding process, which are configured to collect noise data in real time during the grinding process and collect noise data after the grinding is completed. This module can comprehensively monitor the noise of the grinding process and the result under different abrasive particle sizes, providing sufficient input information for subsequent data analysis.

[0178] Preferably, the feature extraction module includes a data preprocessing unit and a neural network model processing unit. The data preprocessing unit is configured to remove noise interference and outliers in the detection data to improve the reliability of the data; and the neural network model processing unit is configured to extract key process variables and result variables of the grinding process from the preprocessed data to provide basic data for the construction of the data relationship.

[0179] Preferably, the trend analysis module further comprises a trend identification unit and a dispersion calculation unit. The trend identification unit is used for identifying the trend area of the smooth and tortuous trend in the data relationship scatter diagram, and the dispersion calculation unit is used for calculating the dispersion degree under different trends, so as to quantify the volatility and consistency of the data.

[0180] Preferably, the dispersion degree adjustment module comprises a sliding window setting unit and a normalization calculation unit. The sliding window setting unit is used for setting the window width and sliding step length for calculating the dispersion degree in different windows; and the normalization calculation unit is used for performing normalization processing on the dispersion condition value in each window, so as to obtain the dispersion degree coefficient with unified scale, thereby laying a foundation for the final adjustment coefficient calculation.

[0181] Preferably, the process evaluation module comprises a threshold setting unit and a comparison judgment unit. The threshold setting unit presets the threshold of the deviation distribution scatter diagram according to the process requirement, historical data and quality control standard, and the comparison judgment unit is used for comparing the overall deviation distribution scatter diagram value with the preset threshold, so as to judge the passability of the grinding process and output the evaluation result.

[0182] It should be noted that the seamless steel pipe grinding process detection device based on machine vision provided by the embodiment of the present application is used to execute all process steps of the seamless steel pipe grinding process detection method based on machine vision provided by the above-mentioned embodiment, and the working principles and beneficial effects of the two are one-to-one correspondence, so the working principles and beneficial effects of the two are not described again.

[0183] The embodiment of the present application further provides an electronic device. The electronic device comprises a processor, a memory and a computer program, such as a grinding process detection program, stored in the memory and executable on the processor. The processor implements the steps in each of the above-mentioned various seamless steel pipe grinding process detection methods based on machine vision when executing the computer program, such as the step S11 shown in the figure. Figure 1 Alternatively, the processor implements the functions of each module / unit in each of the above-mentioned various devices when executing the computer program, such as the data acquisition module.

[0184] 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.

[0185] The electronic device can be a computing device such as a desktop computer, a notebook computer, a palm computer, 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 the electronic device 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.

[0186] 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 device, discrete hardware component, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc. The processor is the control center of the electronic device, and connects various parts of the electronic device through various interfaces and lines.

[0187] 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.; 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 non-volatile 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.

[0188] 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.

[0189] 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.

[0190] 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 are not used to 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 machine vision-based inspection method for seamless steel pipe grinding process, characterized in that, Includes the following steps: The detection data during the grinding process of steel pipes with different abrasive grain sizes are obtained. The detection data includes noise data generated during the grinding process and data after the grinding is completed. The detection data is input into a neural network model to analyze and extract process variables and result variables of the grinding process. Based on the process variables and result variables, construct a scatter plot of data relationships for the grinding process; Perform trend analysis on the data points in the scatter plot of the data relationship, calculate the degree of dispersion under each trend, and obtain the degree of dispersion value of the data relationship; Based on the dispersion value, a dispersion coefficient is calculated. By combining the dispersion coefficient with the distribution density of points in the data relationship scatter plot, an adjustment coefficient for the dispersion of the corresponding data relationship is obtained. Based on the discreteness value, the discreteness coefficient, and the adjustment coefficient, the overall deviation distribution scatter point value of the grinding process is generated; The overall deviation distribution scatter point value is compared with a preset threshold. If the deviation distribution scatter point value is lower than or equal to the preset threshold, the grinding process is deemed qualified. If the deviation distribution scatter point value is higher than the preset threshold, the grinding process is deemed unqualified. The formula for calculating the degree of dispersion is as follows: in, The dispersion value is T, which represents different trends, including flat trends and volatile trends. This represents the amount of change in the overall trend under the corresponding trend T. This indicates the number of changes in the quantity of change under the corresponding trend T. and These represent two linear normalized functions corresponding to trend T; The formula for calculating the dispersion coefficient is as follows: in, Indicates the coefficient of dispersion. This represents the degree of dispersion. It is a linear normalization function.

2. The machine vision-based inspection method for seamless steel pipe grinding process according to claim 1, characterized in that, The detection data is input into a neural network model to analyze and extract process variables and outcome variables of the grinding process, including: The detection data is preprocessed to obtain preprocessed detection data. The data preprocessing includes removing noise and outliers from the detection data. Based on the pre-processed test data, the grinding process characteristics under different abrasive sizes are analyzed, and the key parameters related to the grinding effect in the pre-processed test data are obtained. Based on the key parameters related to grinding effect in the preprocessed test data, process noise and post-process noise features are extracted and combined to obtain the overall test data. Using the overall detection data as input data, the data is fed into a pre-trained neural network model to obtain the process variables and result variables of the grinding process.

3. The seamless steel pipe grinding process inspection method based on machine vision according to claim 1, characterized in that, Based on the process variables and result variables, a scatter plot of data relationships for the grinding process is constructed, including: Analyze the relationships between the process variables and the outcome variables to obtain the correlation patterns related to the grinding process effect; Based on the association pattern, the process variables and result variables are paired to generate multiple data pairs, which can reflect the grinding process characteristics under different abrasive grain size conditions. Using the data pairs as coordinate points, a scatter plot of the data relationships in the grinding process is constructed.

4. The machine vision-based inspection method for seamless steel pipe grinding process according to claim 1, characterized in that, Perform trend analysis on the data points in the scatter plot of the data relationship, calculate the degree of dispersion under each trend, and obtain the dispersion value of the data relationship, including: Based on the changing trends of the data points in the scatter plot of the data relationship, the data points are divided into flat trends and volatile trends. The flat trend represents the area with small changes, and the volatile trend represents the area with large changes. Based on different trends, calculate the amount of change in the overall trend and the corresponding number of changes to quantify the fluctuations under different trends; Calculate the dispersion value under different trends based on the amount of change and the corresponding number of changes in the overall trend; Two linear normalization functions corresponding to the trend and Calculate using the following formulas respectively: in, for The minimum value, for The maximum value; for The minimum value, for The maximum value.

5. The machine vision-based inspection method for seamless steel pipe grinding process according to claim 1, characterized in that, Based on the dispersion value, a dispersion coefficient is calculated. Combining this dispersion coefficient with the distribution density of points in the data relationship scatter plot, an adjustment coefficient for the corresponding data relationship dispersion is obtained, including: Calculate the dispersion coefficient based on the dispersion value; Based on the dispersion coefficient, sliding window parameters are set, and the dispersion within different sliding windows is calculated in segments. The sliding window parameters include window width and sliding step size. Based on the aforementioned dispersion coefficient, the discrete value within each sliding window is calculated using the following formula: in, Represents discrete values. and These represent the maximum and minimum values ​​of the dispersion coefficient within the sliding window, respectively. This represents the number of dispersion coefficients within the sliding window; Based on the discrete values, the normalized discrete values ​​are calculated using the following formula: in, This represents the normalized discrete values. and These represent the maximum and minimum values ​​for the discrete cases across all sliding windows, respectively. Based on the aforementioned dispersion coefficient and the normalized dispersion value, the dispersion adjustment coefficient is calculated using the following formula: in, This represents the adjustment coefficient. This represents the maximum number of scatter points in the scatter plot of the data relationships.

6. The machine vision-based inspection method for seamless steel pipe grinding process according to claim 1, characterized in that, Based on the aforementioned dispersion value, dispersion coefficient, and adjustment coefficient, the overall deviation distribution scatter point value of the grinding process is generated, including: The dispersion value and the dispersion coefficient are used as input parameters for the overall deviation distribution scatter point value to represent the overall dispersion and local characteristics of the data distribution; Based on the dispersion value, the dispersion coefficient, and the adjustment coefficient, the overall deviation distribution scatter point value is calculated using the following formula: in, This represents the overall deviation from the distribution scatter points. This represents the degree of dispersion. Indicates the coefficient of dispersion. This is the adjustment coefficient.

7. The machine vision-based inspection method for seamless steel pipe grinding process according to claim 1, characterized in that, The overall deviation distribution scatter point value is compared with a preset threshold. If the deviation distribution scatter point value is lower than or equal to the preset threshold, the grinding process is deemed qualified; if the deviation distribution scatter point value is higher than the preset threshold, the grinding process is deemed unqualified, including: Historical data from multiple grinding processes were collected, and the overall deviation distribution scatter points of each grinding process were statistically analyzed to form a distribution dataset. Based on the distributed dataset, the mean and standard deviation of the deviation from the scatter points are calculated to obtain a preliminary threshold range; Based on the initial threshold range, the mean is weighted and adjusted to obtain the final preset threshold for the deviation from the distribution scatter point value; The overall deviation distribution scatter point value is compared with the preset threshold. If the deviation distribution scatter point value is lower than or equal to the preset threshold, the grinding process is deemed qualified. If the deviation distribution scatter point value is higher than the preset threshold, the grinding process is deemed unqualified.

8. A machine vision-based inspection system for seamless steel pipe grinding processes, characterized in that, The method for implementing the seamless steel pipe grinding process inspection method based on machine vision as described in any one of claims 1 to 7 includes: The data acquisition module is used to acquire detection data during the steel pipe grinding process under different abrasive sizes. The detection data includes noise data generated during the grinding process and data after the grinding is completed. The feature extraction module is used to input the detection data into the neural network model to analyze and extract the process variables and result variables of the grinding process; The data relationship construction module is used to construct a scatter plot of data relationships for the grinding process based on the process variables and result variables. The trend analysis module is used to perform trend analysis on the data points in the data relationship scatter plot, calculate the degree of dispersion under each trend, and obtain the degree of dispersion value of the data relationship. The dispersion adjustment module is used to calculate the dispersion coefficient based on the dispersion value, and combine the dispersion coefficient with the distribution density of points in the data relationship scatter plot to obtain the adjustment coefficient of the dispersion of the corresponding data relationship; The overall deviation calculation module is used to generate the overall deviation distribution scatter point value of the grinding process based on the dispersion value, dispersion coefficient and adjustment coefficient; The process evaluation module is used to compare the overall deviation distribution scatter point value with a preset threshold. If the deviation distribution scatter point value is lower than or equal to the preset threshold, the grinding process is judged to be qualified; if the deviation distribution scatter point value is higher than the preset threshold, the grinding process is judged to be unqualified.

9. An electronic device, characterized in that, The system includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor, when executing the computer program, implements the machine vision-based seamless steel pipe grinding process inspection method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein, when the computer program is executed, it controls the device containing the computer-readable storage medium to perform the machine vision-based seamless steel pipe grinding process inspection method as described in any one of claims 1 to 7.

Images