Steel cutting quality inspection device for ship steel structure machining

By performing data projection and feature extraction in the steel cutting quality inspection device, and collaborative analysis combined with internal and external inspection data sets, the problem of inaccurate inspection results in the prior art is solved, and more efficient and reliable detection results are achieved.

CN119065318APending Publication Date: 2024-12-03NANTONG YILONG STEEL STRUCTURE CO LTD
View PDF 12 Cites 0 Cited by

Patent Information

Application Number
CN202411564268.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-11-05
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

In the prior art, steel cutting quality inspection relies on manual visual inspection, which is inefficient and prone to errors, and cannot guarantee the accuracy and consistency of inspection results.

Method used

It provides a steel cutting quality inspection device for marine steel structure processing, extracts projection point features through data projection, builds internal and external feature inspection channels, conducts coordinated quality analysis, and uses internal and external inspection data sets to determine, so as to realize abnormal traceability.

Benefits of technology

It improves the comprehensiveness and accuracy of the test results, reduces human error, and ensures the reliability and consistency of the test results.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119065318A_ABST
    Figure CN119065318A_ABST
Patent Text Reader

Abstract

The invention discloses a steel cutting quality inspection device for ship steel structure machining, and relates to the related field of steel cutting, and the device comprises a steel cutting module used for cutting target steel; the projection point feature extraction module is used for extracting projection point features; the quality inspection channel construction module is used for constructing a quality inspection channel; the internal inspection data set generation module is used for synchronizing the sample internal characteristics and material matching information to the internal quality inspection branch; the appearance inspection data set generation module is used for synchronizing sample appearance characteristics and spatial structure information to the appearance quality inspection branch; the quality collaborative analysis module is used for performing quality collaborative analysis; and the abnormality tracing module is used for performing abnormality tracing. The technical problem that the detection result is poor in comprehensiveness and accuracy in existing steel cutting quality inspection is solved, and the technical effect of improving the comprehensiveness and accuracy of the detection result is achieved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of steel cutting, and in particular to a steel cutting quality inspection device used for ship steel structure processing. Background Art

[0002] In the field of ship steel structure processing, steel cutting is a key process, and its quality directly affects the strength and stability of the entire ship structure. Therefore, strict inspection of steel cutting quality is particularly important. Under the existing technical background, the inspection of steel cutting quality mainly relies on manual visual inspection, which is inefficient, error-prone, and greatly affected by the experience and subjective judgment of the inspectors, and cannot guarantee the accuracy and consistency of the inspection results.

[0003] In the current related technologies, there are technical problems in steel cutting quality inspection, such as poor comprehensiveness and accuracy of test results. Summary of the invention

[0004] The present application provides a steel cutting quality inspection device for ship steel structure processing, which obtains steel cutting sample information, performs data projection on the steel cutting sample information, extracts projection point features, obtains multiple sample internal features and multiple sample appearance features, constructs a quality inspection channel to inspect the multiple sample internal features and multiple sample appearance features, combines internal inspection data sets and appearance inspection data sets to perform quality collaborative analysis, and performs abnormality tracing based on quality collaborative judgment results. By comprehensively analyzing the internal and appearance characteristics of steel, the technical effect of improving the comprehensiveness and accuracy of the detection results is achieved.

[0005] The present application provides a steel cutting quality inspection device for ship steel structure processing, comprising: A steel cutting module, the steel cutting module is used to determine a steel processing plan based on the ship steel structure, execute the steel processing plan to cut the target steel, and obtain steel cutting sample information; a projection point feature extraction module, the projection point feature extraction module is used to perform data projection based on the steel cutting sample information, extract projection point features according to the projection results, and obtain multiple sample features, the multiple sample features include multiple sample internal features and multiple sample appearance features; a quality inspection channel construction module, the quality inspection channel construction module is used to construct a quality inspection channel, the quality inspection channel includes an internal quality inspection branch and an appearance quality inspection branch; an internal inspection data set generation module, the internal inspection data set generation module is used to extract the multiple sample internal features The material ratio information of the target cut piece is synchronized to the internal quality inspection branch to inspect the target cut piece and generate an internal inspection data set; an appearance inspection data set generation module, the appearance inspection data set generation module is used to synchronize the appearance features of the multiple samples in combination with the spatial structure information of the target cut piece to the appearance quality inspection branch to inspect the target cut piece and generate an appearance inspection data set; a quality collaborative analysis module, the quality collaborative analysis module is used to combine the internal inspection data set with the appearance inspection data set to perform quality collaborative analysis on the target cut piece and generate a quality collaborative judgment result; an abnormality tracing module, the abnormality tracing module is used to trace the abnormality according to the quality collaborative judgment result, and process and identify the target cut piece according to the traceability result.

[0006] In a possible implementation, data projection is performed based on the steel cutting sample information, projection point features are extracted according to the projection results, multiple sample features are obtained, and the following processing is performed: The steel cutting sample information is converted into multiple data points to construct a sample data lattice; the sample data lattice is dimensionally reduced to generate a sample reduced-dimensional data lattice; based on wavelet transform, the sample reduced-dimensional data lattice is projected to a low-dimensional space to obtain a projection result; it is determined whether the projection point in the projection result is greater than or equal to a preset threshold; multiple projection points greater than or equal to the preset threshold are extracted for analysis to generate multiple sample appearance features; multiple projection points less than the preset threshold are extracted for analysis to generate multiple sample internal features; the multiple sample appearance features and the multiple sample internal features are integrated to obtain the multiple sample features.

[0007] In a possible implementation, the sample data lattice is dimensionally reduced to generate a sample dimensionally reduced data lattice, and the following processing is performed: The sample data lattice is preprocessed, and data is standardized according to the preprocessing result to determine the sample standard data lattice; the sample standard data lattice is traversed, and the ith data point and the jth data point are extracted and calculated through the covariance calculation expression to obtain the covariance between the ith data point and the jth data point, thereby iterating to obtain a covariance matrix; characteristic decomposition is performed based on the covariance matrix to obtain multiple data eigenvalues ​​and multiple data eigenvectors, and the multiple data eigenvalues ​​correspond to the multiple data eigenvectors; the multiple data eigenvalues ​​are sorted in descending order to obtain an eigenvalue sequence, and the first m eigenvalues ​​and m eigenvectors are extracted according to the eigenvalue sequence; based on the sample data lattice, the sample dimension reduction data lattice is generated by combining the first m eigenvalues ​​and m eigenvectors.

[0008] In a possible implementation, the sample data point array is preprocessed to perform the following processing: The sample data dot matrix is ​​denoised by using a median filtering method to generate a sample denoised data dot matrix; the sample denoised data dot matrix is ​​divided into a training data point set and a test training point set; based on the training data point set and the test training point set, the sample denoised data points are evaluated through cross-validation to generate a denoising evaluation result; and the preprocessing result is generated according to the denoising evaluation result.

[0009] In a possible implementation, the plurality of sample internal features are combined with the material ratio information of the target cut piece and synchronized to the internal quality inspection branch to inspect the target cut piece, generate an internal inspection data set, and perform the following processing: Determine the material type information of the target steel based on the ship steel structure, analyze the material property information according to the material type information, and determine the material ratio information of the target cutting part according to the analysis result; establish data association between the internal characteristics of the multiple samples and the material ratio information to generate an association factor; based on the association factor, synchronize the internal characteristics of the multiple samples in combination with the material ratio information of the target cutting part to the internal quality inspection branch; formulate a sample internal test plan according to the physical and chemical properties of the target steel through the internal quality inspection branch; execute the sample internal test plan to generate sample test results, traverse the material ratio information and match the sample test results, evaluate the quality of the target cutting part according to the matching results, and generate an internal quality score; add the internal quality score to the internal inspection data set.

[0010] In a possible implementation, the plurality of sample appearance features are combined with the spatial structure information of the target cut piece and synchronized to the appearance quality inspection branch to inspect the target cut piece, generate an appearance inspection data set, and perform the following processing: The spatial structure information is traversed to determine multiple structural dimension information and multiple spatial position relationships; historical appearance quality inspection reports are retrieved to extract historical structural dimension information and historical spatial position relationships to formulate inspection standards; the multiple sample appearance features are combined with the spatial structure information of the target cutting piece and synchronized to the appearance quality inspection branch, the multiple sample appearance features are traversed, the multiple structural dimension information and the multiple spatial position relationships are compared with the inspection standards, and the multiple sample appearance features are classified and labeled according to the comparison results to obtain multiple category labels; multiple abnormal labels are extracted based on the multiple category labels, and multiple abnormal appearance features are added to the appearance inspection data set according to the multiple abnormal labels.

[0011] In a possible implementation, the internal inspection data set is combined with the appearance inspection data set to perform a quality collaborative analysis on the target cut part to generate a quality collaborative determination result, and the following processing is performed: A quality influence coefficient is generated by performing calculation based on the internal inspection evaluation index and the appearance inspection evaluation index; a weight is assigned to the internal inspection data set and the appearance inspection data set according to the quality influence coefficient to generate a weight assignment result; the internal inspection data set and the appearance inspection data set are combined with the weight assignment result to generate a joint inspection data set; based on the joint inspection data set, it is determined whether the inspection result of the target cut part is within a preset quality standard range; if the inspection result of the target cut part is within the preset quality standard range, the quality of the target cut part is determined to be qualified, and the quality collaborative determination result is generated; if the inspection result of the target cut part is not within the preset quality standard range, the quality of the target cut part is determined to be unqualified, and a plurality of abnormal levels are divided according to the inspection result that is not within the preset quality standard range, and the plurality of abnormal levels are traversed to match the target cut part with unqualified quality to generate the quality collaborative determination result.

[0012] In a possible implementation, the abnormality is traced according to the quality collaborative determination result, and the following processing is performed: Based on the multiple abnormality levels combined with the quality collaborative determination result, the internal inspection data set of the target cut part is traced for abnormality to generate internal traceability data; based on the multiple abnormality levels combined with the quality collaborative determination result, the appearance inspection data set of the target cut part is traced for abnormality to generate external traceability data; based on the internal traceability data, internal abnormality feedback is performed to determine the first quality abnormality source; based on the external traceability data, external abnormality feedback is performed to determine the second quality abnormality source; the first quality abnormality source and the second quality abnormality source are added to the traceability result.

[0013] It is intended to adopt the steel cutting quality inspection device for ship steel structure processing proposed in this application, determine the steel processing plan based on the ship steel structure through the steel cutting module, execute the steel processing plan to cut the target steel, obtain steel cutting sample information, perform data projection based on the steel cutting sample information through the projection point feature extraction module, extract the projection point features according to the projection results, obtain multiple sample features, the multiple sample features include multiple sample internal features and multiple sample appearance features, construct a quality inspection channel through the quality inspection channel construction module, the quality inspection channel includes an internal quality inspection branch and an appearance quality inspection branch, and combine the multiple sample internal features through the internal inspection data set generation module The material ratio information of the target cut part is synchronized to the internal quality inspection branch for inspection of the target cut part to generate an internal inspection data set; the appearance features of multiple samples are combined with the spatial structure information of the target cut part through the appearance inspection data set generation module to be synchronized to the appearance quality inspection branch for inspection of the target cut part to generate an appearance inspection data set; the quality collaborative analysis module combines the internal inspection data set with the appearance inspection data set to perform quality collaborative analysis on the target cut part to generate a quality collaborative judgment result; the abnormality tracing module traces the abnormality according to the quality collaborative judgment result, and the target cut part is processed and marked according to the traceability result, thereby achieving the technical effect of improving the comprehensiveness and accuracy of the detection results. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solution of the embodiment of the present invention, the accompanying drawings of the embodiment of the present invention will be briefly introduced below. The structural diagram is used in the present application to illustrate the operations performed by the device according to the embodiment of the present application. It should be understood that the previous or following operations are not necessarily performed accurately in order. On the contrary, various steps can be processed in reverse order or simultaneously as needed. At the same time, other operations can also be added to these processes, or one or more operations can be removed from these processes.

[0015] Figure 1 A schematic structural diagram of a steel cutting quality inspection device for ship steel structure processing provided in an embodiment of the present application.

[0016] Figure 2 A schematic diagram of a process for collaborative quality analysis of a target cut part in a steel cutting quality inspection device for ship steel structure processing provided in an embodiment of the present application.

[0017] Explanation of the accompanying drawings: steel cutting module 10, projection point feature extraction module 20, quality inspection channel construction module 30, internal inspection data set generation module 40, appearance inspection data set generation module 50, quality collaborative analysis module 60, abnormality tracing module 70. DETAILED DESCRIPTION

[0018] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below.

[0019] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings. The described embodiments should not be regarded as limiting the present application. All other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of this application.

[0020] In the following description, reference is made to "some embodiments", which describe a subset of all possible embodiments, but it is understood that "some embodiments" may be the same subset or different subsets of all possible embodiments, and may be combined with each other without conflict, and the terms "first\second" involved are merely to distinguish similar objects and do not represent a specific ordering of objects. The terms "including" and "having" and any variations are intended to cover non-exclusive inclusions, for example, a process, method, device, product, or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, devices, products, or equipment. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those generally understood by technicians in the technical field of this application. The terms used herein are for the purpose of describing the embodiments of the present application only.

[0021] The present application embodiment provides a steel cutting quality inspection device for ship steel structure processing, such as Figure 1 As shown, the device comprises: The steel cutting module 10 is used to determine a steel processing scheme based on the ship steel structure, execute the steel processing scheme to cut the target steel, and obtain steel cutting sample information. Specifically, according to the design drawings, material lists, technical specifications, etc. of the ship steel structure (the skeleton structure of the ship, composed of various steels, with supporting and protective functions), the specific requirements of the ship steel structure for steel, such as size, shape, material, etc., are analyzed to generate a steel processing scheme, including cutting method, cutting sequence, cutting parameters, etc. According to the steel processing scheme, the parameters of the cutting machine are set, such as saw blade selection, rotation speed, feed speed, cutting angle, etc., and the target steel is placed on the workbench of the cutting machine, and is positioned and fixed as needed to ensure that the target steel does not move or shake during the cutting process. The cutting machine is started, and the steel is cut according to the preset parameters. After the cutting is completed, the steel sample information after cutting is obtained, including the size, shape, surface quality and other data of the cut steel, wherein the steel cutting sample is a steel sample cut according to the steel processing scheme, which is used to inspect and evaluate the cutting quality.

[0022] The projection point feature extraction module 20 is used to perform data projection based on the steel cutting sample information, extract projection point features according to the projection results, and obtain multiple sample features, wherein the multiple sample features include multiple sample internal features and multiple sample appearance features. Specifically, the steel cutting sample information includes detailed data such as size, shape, and material. The steel cutting sample information is preprocessed as necessary, such as data cleaning and format conversion. The steel cutting sample information is projected into a specific two-dimensional or three-dimensional space using data projection methods such as projection based on geometric shapes and projection based on textures. The projected data can be represented in an image, point cloud, or other data format. Key points or feature points are identified in the projected data. These points can represent important features of the steel cutting sample. The projection point features are extracted from the projection data using techniques such as image processing and point cloud processing, including the position, color, shape, and texture of the points. The extracted projection point features are classified according to internal features and appearance features. The internal features include the internal structure, texture, and material of the steel; and the appearance features include shape, size, and surface morphology. The internal features and appearance features of multiple samples are integrated to form a complete sample feature set.

[0023] In a possible implementation, data projection is performed based on the steel cutting sample information, and projection point features are extracted according to the projection results to obtain multiple sample features, including: converting the steel cutting sample information into multiple data points to construct a sample data lattice; reducing the dimension of the sample data lattice to generate a sample reduced-dimensional data lattice; projecting the sample reduced-dimensional data lattice to a low-dimensional space based on a wavelet transform to obtain a projection result; determining whether the projection points in the projection result are greater than or equal to a preset threshold; extracting multiple projection points greater than or equal to the preset threshold for analysis to generate multiple sample appearance features; extracting multiple projection points less than the preset threshold for analysis to generate multiple sample internal features; integrating the multiple sample appearance features and the multiple sample internal features to obtain the multiple sample features.

[0024] Specifically, the information of steel cutting samples includes dimensions, weight, hardness, surface roughness, etc. For each aspect, key data or indicators are extracted and quantified into specific values. These values ​​are data points. All extracted data points are combined to form a multidimensional data set, namely the sample data lattice. Each data point represents a characteristic or attribute of the sample, and the entire data lattice fully describes the comprehensive information of the sample. Select a dimensionality reduction algorithm (such as PCA, LDA, etc.) to reduce the dimensionality of the sample data lattice. The data point set after dimensionality reduction is the sample dimensionality reduction data lattice. The sample dimensionality reduction data lattice retains the main features of the original data (sample data lattice), but has a lower dimension. The wavelet transform is used to further process the dimensionality reduced data and project it into a low-dimensional space. The wavelet transform is used to reveal the local and global characteristics of the data.

[0025] A threshold is preset based on domain knowledge, expert experience or historical data, and each projection point in the projection result is traversed to determine whether its value is greater than or equal to the preset threshold. For projection points greater than or equal to the preset threshold, it is considered that they represent the appearance characteristics of the steel cutting sample, and these projection points are analyzed using statistical methods or machine learning techniques to extract features related to the appearance. For example, if the projection points reflect the surface roughness or color of the sample, then these features are classified as appearance features. For projection points less than the threshold, it is considered that they represent the internal characteristics of the steel cutting sample, and these projection points are analyzed using statistical methods or machine learning techniques to extract features related to the internal quality, structure or material properties. For example, if the projection points reflect the hardness, chemical composition or internal structure of the sample, then these features are classified as internal features. The appearance features and internal features extracted from the projection points are merged, and the integrated feature set represents multiple sample features of the steel cutting sample. This implementation method distinguishes the characteristics of the projection points by setting a threshold, conveniently extracts features related to the appearance and internal quality, and achieves the technical effect of improving the accuracy of feature extraction.

[0026] In a possible implementation, the sample data lattice is dimensionally reduced to generate a sample reduced-dimensional data lattice, including: preprocessing the sample data lattice, performing data standardization based on the preprocessing result, and determining a sample standard data lattice; traversing the sample standard data lattice, extracting the ith data point and the jth data point, calculating through a covariance calculation expression, obtaining the covariance between the ith data point and the jth data point, and iterating to obtain a covariance matrix; performing feature decomposition based on the covariance matrix to obtain multiple data eigenvalues ​​and multiple data eigenvectors, wherein the multiple data eigenvalues ​​correspond to the multiple data eigenvectors; sorting the multiple data eigenvalues ​​in descending order to obtain an eigenvalue sequence, and extracting the first m eigenvalues ​​and m eigenvectors according to the eigenvalue sequence; and generating the sample reduced-dimensional data lattice based on the sample data lattice combined with the first m eigenvalues ​​and m eigenvectors.

[0027] Specifically, for the sample data matrix, check for missing values, outliers or noise in the data, and clean, fill or smooth as needed. Convert the data to the same scale by subtracting the mean and dividing by the standard deviation, and eliminate the scale differences between different features by standardization so that the weights are equal in the calculation. Traverse the sample standard data matrix, analyze each pair of data points in the sample standard data matrix one by one, and for the i-th data point and the j-th data point, use the covariance calculation expression (the average of the deviation product of the two data points) to calculate the covariance between the two, and repeat the above process until the covariance between all pairs of data points is calculated. Fill the calculated covariance values ​​into the matrix, the rows and columns of the matrix correspond to the data points, and the elements of the matrix are the covariance between the i-th and j-th data points. The covariance matrix is ​​subjected to eigendecomposition, that is, its eigenvalues ​​(scalar values ​​obtained after eigendecomposition, representing the degree of variation of the data in a certain direction) and eigenvectors (vectors corresponding to the eigenvalues, representing the change pattern of the data in this direction) are solved. The result of eigendecomposition is a set of eigenvalues ​​and eigenvectors, and there is a one-to-one correspondence between the eigenvalues ​​and the eigenvectors. The eigenvalues ​​are sorted in descending order to obtain an eigenvalue sequence. According to the required degree of dimensionality reduction or interpretability, the first m eigenvalues ​​and the corresponding m eigenvectors in the eigenvalue sequence are selected, where m is the number of features to be retained. The linear combination result obtained by linearly combining the data points in the original data lattice (sample data lattice) and the selected m eigenvectors is the data lattice after dimensionality reduction, in which each data point is represented by the coordinates of the original data point in the new feature space. This implementation method can retain the main information in the data while reducing the dimension by selecting the most important eigenvalues ​​and eigenvectors, thereby achieving the technical effect of reducing information loss.

[0028] In one possible implementation, the sample data dot matrix is ​​preprocessed, including: denoising the sample data dot matrix using a median filtering method to generate a sample denoised data dot matrix; dividing the sample denoised data dot matrix into a training data point set and a test training point set; based on the training data point set and the test training point set, evaluating the sample denoised data points through cross-validation to generate a denoising evaluation result; and generating the preprocessing result based on the denoising evaluation result.

[0029] Specifically, determine a neighborhood size (e.g., 3x3, 5x5, etc.), which is used to calculate the median of each data point. For each data point in the sample data lattice, perform the following steps: take the current data point as the center, select all points within the determined neighborhood size as the neighborhood of the data point, sort the values ​​of all points in the selected neighborhood, take the median, replace the value of the current data point with the median of its neighborhood, repeat the above steps until all points in the sample data lattice have been processed, and after all points are replaced, a new data lattice is obtained, i.e., the sample denoised data lattice. According to the needs and data volume, determine the division ratio of the training data point set and the test training point set (e.g., 80% training set and 20% test set), and use a random algorithm to randomly assign the data points in the sample denoised data lattice to the training data point set and the test training point set. Use a cross-validation method, such as K-fold cross-validation, and determine the number of folds K. According to the cross-validation method, further divide the training data point set into K subsets of similar size. For each iteration, select K-1 subsets as the current training set and the remaining subset as the validation set. Use the training set data of the current iteration to train an evaluation model, evaluate the performance of the model on the validation set, measure the denoising effect by calculating performance indicators (such as mean square error, correlation coefficient, etc.), and record the evaluation results of each iteration. Average or otherwise synthesize the evaluation results of K iterations to obtain an overall denoising effect evaluation. Use the best model (or evaluation method) trained on the entire training data point set to evaluate on the test training point set, calculate performance indicators to verify the generalization ability of the denoising effect on unseen data, use the evaluation results on the test training point set as part of the denoising evaluation results, and combine the cross-validation results on the training data point set and the evaluation results on the test training point set to generate the final denoising evaluation results. Check the denoising evaluation results to evaluate whether the denoising effect meets the requirements, that is, determine whether the main features are retained. If the denoising effect meets the requirements, the sample denoising data dot matrix is ​​output as the preprocessing result; otherwise, adjust the denoising parameters or methods and re-execute the denoising process. This implementation method evaluates the denoised data through cross-validation to verify whether the denoising method is effective and whether the main features of the denoised data are retained, achieving the technical effect of ensuring that the denoising process does not lose important information.

[0030] The quality inspection channel construction module 30 is used to construct a quality inspection channel, and the quality inspection channel includes an internal quality inspection branch and an appearance quality inspection branch. Specifically, the quality inspection channel is a system or process including multiple quality inspection links and branches, which is used to comprehensively evaluate the quality of steel. According to the characteristics of steel cutting samples, inspection requirements, quality inspection standards, and customer needs, the specific content and standards of internal quality inspection and appearance quality inspection are determined. According to the internal characteristics of steel (such as material, structure, chemical composition, etc.), design corresponding inspection methods, inspection equipment, and inspection processes, determine the key points and control points of internal quality inspection, make the internal quality of steel meet the standards and customer needs, and construct internal quality inspection branches. According to the appearance characteristics of steel (such as size, shape, surface quality, etc.), design corresponding inspection methods, inspection equipment, and inspection processes, determine the inspection points and evaluation standards of appearance quality inspection, make the appearance quality of steel meet the standards and customer needs, and construct appearance quality inspection branches. Among them, the internal quality inspection branch is used to inspect the internal characteristics of steel, such as material, structure, chemical composition, etc.; the appearance quality inspection branch is used to inspect the appearance characteristics of steel, such as size, shape, surface quality, etc. The internal quality inspection branch and the appearance quality inspection branch are integrated into a unified quality inspection channel so that the two branches can work together.

[0031] An internal inspection data set generation module 40 is used to synchronize the internal features of the multiple samples with the material ratio information of the target cut piece to the internal quality inspection branch to inspect the target cut piece and generate an internal inspection data set. Specifically, query the material ratio information of the target cut piece, including the composition ratio and addition amount of various raw materials. Match the internal features of the multiple samples with the material ratio information of the target cut piece, find the sample that is closest or similar to the material ratio information of the target cut piece, synchronize the matched internal features of the multiple samples to the internal quality inspection branch, and perform quality inspection in combination with the material ratio information of the target cut piece. According to the detection standard, threshold, etc., use the internal quality inspection branch to inspect the target cut piece and record the inspection data. Sort the inspection data generated by the internal quality inspection branch, including inspection results, data parameters, etc., and construct an internal inspection data set based on the sorted data.

[0032] In a possible implementation, the multiple internal features of the samples are combined with the material ratio information of the target cutting piece and synchronized to the internal quality inspection branch to inspect the target cutting piece to generate an internal inspection data set, including: determining the material type information of the target steel based on the ship steel structure, analyzing the material property information according to the material type information, and determining the material ratio information of the target cutting piece according to the analysis result; establishing data association between the multiple internal features of the samples and the material ratio information to generate an association factor; based on the association factor, synchronizing the multiple internal features of the samples in combination with the material ratio information of the target cutting piece to the internal quality inspection branch; formulating a sample internal test plan according to the physical and chemical properties of the target steel through the internal quality inspection branch; executing the sample internal test plan to generate sample test results, traversing the material ratio information and matching the sample test results, performing quality evaluation on the target cutting piece according to the matching results to generate an internal quality score; and adding the internal quality score to the internal inspection data set.

[0033] Specifically, identify the type of steel used in the ship steel structure, such as carbon steel, stainless steel, alloy steel, etc., and query or calculate its basic information such as chemical composition and physical properties according to the type of steel. According to the type of steel, further analyze the material's strength, toughness, corrosion resistance and other properties, and determine the material ratio (such as the content ratio of alloy elements) required for the target cutting part according to the ship design requirements and purpose. Use data analysis methods (such as regression analysis, cluster analysis, etc.) to associate the internal characteristics of the sample with the material ratio information and generate a correlation factor, which refers to a mathematical or statistical model that can reflect the relationship between the internal characteristics of the sample and the material ratio. The established correlation factor and the material ratio information of the target cutting part are synchronized to the internal quality inspection branch. The internal quality inspection branch formulates the sample internal test plan according to the physical and chemical properties of the target steel, executes the test plan, and generates the sample test results. Traverse the material ratio information and match the sample test results. According to the matching results, the quality of the target cutting part is evaluated. Through matching and evaluation, it is determined whether the target cutting part meets the design requirements and material ratio requirements, and an internal quality score is generated. The generated internal quality score is added to the internal inspection data set. This implementation method establishes data association between the internal characteristics of multiple samples and material ratio information, generates correlation factors, and helps the internal quality inspection branch better understand the relationship between sample characteristics and material ratios, achieving the technical effect of improving the targetedness and accuracy of inspections.

[0034] The appearance inspection data set generation module 50 is used to synchronize the appearance features of the multiple samples with the spatial structure information of the target cut piece to the appearance quality inspection branch to inspect the target cut piece and generate an appearance inspection data set. Specifically, a measuring tool or device is used to extract the spatial structure of the target cut piece, such as length, width, height, angle, etc. The correlation between the appearance features of the multiple samples and the spatial structure information of the target cut piece is analyzed. According to the analysis results, samples with similar appearance or structure to the target cut piece are found, and the matched appearance features of the multiple samples are synchronized to the appearance quality inspection branch, and the inspection is combined with the spatial structure information of the target cut piece. According to the inspection standards, such as dimensional tolerance, shape requirements, surface quality standards, etc., the appearance quality inspection branch is used to inspect the target cut piece, and the inspection data, including qualified or unqualified items, etc., are recorded. The inspection data generated by the appearance quality inspection branch is sorted, including inspection results, description of unqualified items, inspection time, etc., and an appearance inspection data set is constructed based on the sorted data.

[0035] In a possible implementation, the multiple sample appearance features are combined with the spatial structure information of the target cut part and synchronized to the appearance quality inspection branch to inspect the target cut part to generate an appearance inspection data set, including: traversing the spatial structure information to determine multiple structural dimension information and multiple spatial position relationships; retrieving historical appearance quality inspection reports to extract historical structural dimension information and historical spatial position relationships to formulate inspection standards; synchronizing the multiple sample appearance features with the spatial structure information of the target cut part to the appearance quality inspection branch, traversing the multiple sample appearance features, comparing the multiple structural dimension information and the multiple spatial position relationships with the inspection standards, classifying and labeling the multiple sample appearance features according to the comparison results to obtain multiple category labels; extracting multiple abnormal labels based on the multiple category labels, and adding multiple abnormal appearance features to the appearance inspection data set according to the multiple abnormal labels.

[0036] Specifically, read the spatial structural information of the target cutting part, including CAD drawings, 3D models or actual measurement data, traverse this information, identify and extract key structural dimension information, such as length, width, height, angle, etc., and analyze the spatial position relationship between these structural elements, such as relative position, parallelism, verticality, etc. Among them, the spatial structural information describes the layout, size and shape of the target cutting part in space, the structural dimension information refers to the size of each component on the target cutting part, and the spatial position relationship describes the relative position and arrangement of each structural element in the target cutting part. Access the historical database or file storage system, retrieve historical appearance quality inspection reports similar to the target cutting part, extract key structural dimension information and spatial position relationship data from these reports, and formulate the appearance quality inspection standard of the current target cutting part based on these data, combined with industry standards and customer needs, that is, the criteria or specifications for judging whether the appearance quality of the target cutting part is qualified. Combine the appearance feature data of multiple samples with the spatial structure information of the target cutting parts, synchronize the combined data to the appearance quality inspection branch, check the appearance features of each sample one by one, compare the structural size information and spatial position relationship of the sample with the established inspection standards, evaluate whether the sample meets the inspection standards, and record the comparison results. According to the comparison results, the appearance features of the samples are divided into different categories, such as qualified, slightly unqualified, seriously unqualified, etc., and the corresponding category labels are assigned to the appearance features of each category of samples. Identify the labels representing abnormalities (such as unqualified, seriously unqualified, etc.) from the category labels, extract the sample appearance feature data related to these abnormal labels, and add these abnormal appearance feature data to the appearance inspection data set. This implementation method ensures the consistency and repeatability of the inspection process by formulating inspection standards, and achieves the technical effect of improving the reliability of the inspection results.

[0037] The quality collaborative analysis module 60 is used to combine the internal inspection data set with the appearance inspection data set to perform a quality collaborative analysis on the target cut part and generate a quality collaborative judgment result. Specifically, the internal inspection data set and the appearance inspection data set are associated according to the unique identifier of the target cut part (such as batch number, serial number, etc.) so that the internal and appearance data correspond to the same target cut part. The two inspection data sets are cleaned to remove duplicate, erroneous or invalid data, and the data is standardized for easy analysis and comparison. A quality collaborative analysis model is selected or developed to integrate the internal and appearance inspection results and give a comprehensive quality judgment result. The internal and appearance inspection data of the target cut part are input into the trained quality collaborative analysis model. The model performs calculations and inferences based on the input data to give a quality collaborative judgment result of the target cut part, that is, the overall quality assessment result of the target cut part.

[0038] like Figure 2As shown, in a possible implementation, the internal inspection data set is combined with the appearance inspection data set to perform a quality collaborative analysis on the target cut part to generate a quality collaborative judgment result, including: calculating based on the internal inspection evaluation index and the appearance inspection evaluation index to generate a quality influence coefficient; weighting the internal inspection data set and the appearance inspection data set according to the quality influence coefficient to generate a weight allocation result; combining the internal inspection data set and the appearance inspection data set in combination with the weight allocation result to generate a joint inspection data set; based on the joint inspection data set, determine whether the inspection result of the target cut part is in a preset quality standard range; if the inspection result of the target cut part is in the preset quality standard range, determine that the quality of the target cut part is qualified and generate the quality collaborative judgment result; if the inspection result of the target cut part is not in the preset quality standard range, determine that the quality of the target cut part is unqualified, and divide it into multiple abnormal levels according to the inspection result that is not in the preset quality standard range, traverse the multiple abnormal levels to match the target cut part with unqualified quality, and generate the quality collaborative judgment result.

[0039] Specifically, the internal inspection evaluation indicators include material properties, etc.; the appearance inspection evaluation indicators include color, texture, scratches, etc. According to historical data, expert opinions or industry standards, a weight is assigned to each evaluation indicator. By calculating the weighted average of all evaluation indicators in each data set (internal inspection and appearance inspection), the respective quality impact coefficients are obtained. The quality impact coefficient is a quantitative indicator that indicates the degree of influence of a certain data set (internal inspection or appearance inspection) on the overall quality. According to the calculated quality impact coefficient, corresponding weights are assigned to the internal inspection data set and the appearance inspection data set. The size of the weight reflects the importance of each in the overall quality assessment. The internal inspection data set and the appearance inspection data set are matched according to the same sample or batch. According to the weight allocation result, the two data sets are weighted and merged to generate a joint inspection data set. Each data point in the joint inspection data set contains the weighted information of the internal inspection and appearance inspection. According to industry standards, customer needs or enterprise regulations, one or more quality standard intervals are set, and the data in the joint inspection data set are compared with the preset quality standard intervals. If the comparison result shows that the inspection result of the target cutting part is within the preset quality standard interval, its quality is determined to be qualified, and the qualified quality result is recorded and output as the quality collaborative determination result. If the comparison results show that the inspection results of the target cut parts are not within the preset quality standard range, then the quality is judged to be unqualified. According to the degree and type of unqualified, the unqualified inspection results are divided into multiple abnormal levels (such as slight unqualified, serious unqualified, etc.), and these abnormal levels are traversed to find the abnormal level that best matches the unqualified inspection results of the target cut parts, and the matched abnormal level is output as the quality collaborative judgment result. This implementation method quantifies the importance of internal inspection and appearance inspection by calculating the quality impact coefficient and assigning weights, making the evaluation results more objective and accurate, and achieving the technical effect of making the quality evaluation of the target cut parts more accurate and comprehensive.

[0040] The abnormality tracing module 70 is used to trace the abnormality according to the quality collaborative determination result, and to process and identify the target cut parts according to the traceability result. Specifically, according to the quality collaborative determination result, determine which target cut parts have quality abnormalities, and collect detailed information of the abnormal target cut parts, including but not limited to: internal inspection data: such as dimensional deviation, material properties, etc.; appearance inspection data: such as surface defects, color differences, etc.; other relevant information in the production process: such as machine operation records, operator information, process parameters, etc. According to the collected information, the causes of the quality abnormality are gradually analyzed and determined by methods such as flow chart analysis and fault tree analysis. According to the traceability results, the abnormal target cut parts are classified, and can be divided into different levels or categories according to the nature and severity of the abnormality. For different types of abnormal target cut parts, corresponding processing plans are formulated, including rework, scrapping, re-inspection, etc. According to the processing plan, the abnormal target cut parts are processed accordingly, and the processed target cut parts are identified, including processing results, processing date, processing personnel, etc. The embodiment of the present application adopts technical means such as obtaining steel cutting sample information, performing data projection on the steel cutting sample information, extracting projection point features, obtaining multiple sample internal features and multiple sample appearance features, constructing a quality inspection channel to inspect multiple sample internal features and multiple sample appearance features, jointly performing quality collaborative analysis on the internal inspection data set and the appearance inspection data set, and tracing the source of abnormalities based on the quality collaborative judgment results. By comprehensively analyzing the internal and appearance characteristics of steel, the technical effect of improving the comprehensiveness and accuracy of the detection results is achieved.

[0041] In a possible implementation, abnormality tracing is performed according to the quality collaborative determination result, including: tracing the abnormality of the internal inspection data set of the target cut part based on the multiple abnormality levels in combination with the quality collaborative determination result to generate internal traceability data; tracing the abnormality of the appearance inspection data set of the target cut part based on the multiple abnormality levels in combination with the quality collaborative determination result to generate external traceability data; performing internal abnormality feedback based on the internal traceability data to determine the first source of quality abnormality; performing external abnormality feedback based on the external traceability data to determine the second source of quality abnormality; adding the first source of quality abnormality and the second source of quality abnormality to the traceability result.

[0042] Specifically, according to the quality collaborative judgment result, determine the abnormality level corresponding to the target cut piece, screen out the data related to the target cut piece from the internal inspection data set, combine the abnormality level, conduct in-depth analysis on the screened internal inspection data set, find the internal factors that cause quality abnormalities, such as material properties, production processes, etc., record the analyzed internal abnormal factors, and form internal traceability data. Screen out the data related to the target cut piece from the appearance inspection data set, combine the abnormality level, conduct in-depth analysis on the screened appearance inspection data set, find the external factors that cause quality abnormalities, such as surface scratches, color deviation, etc., record the analyzed external abnormal factors, and form external traceability data. Feedback the internal traceability data to relevant departments or personnel, such as production departments, material suppliers, etc., and determine the main factors that cause quality abnormalities, that is, the first source of quality abnormalities, based on the feedback results. Feedback the external traceability data to relevant departments or personnel, and determine the main factors that cause quality abnormalities, that is, the second source of quality abnormalities, based on the feedback results. Integrate the determined first source of quality abnormalities and the second source of quality abnormalities into the same traceability result, and output the integrated traceability result to relevant personnel or systems for subsequent quality improvement and monitoring. This implementation method combines traceability analysis with abnormality levels, accurately locates the key factors that lead to quality abnormalities, and provides a clear direction for subsequent improvements. By using the traceability results as an important basis for quality improvement, it achieves the technical effect of achieving continuous quality improvement.

[0043] Although the present application makes various references to certain modules in the device according to the embodiments of the present application, any number of different modules may be used and run on the user terminal and / or server, and the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the present invention.

[0044] The above specific implementation manner does not constitute a limitation to the protection scope of the present application. It should be understood by those skilled in the art that various modifications, combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present application should be included in the protection scope of the present application. In some cases, the actions or steps recorded in the present application can be performed in an order different from that in the embodiment and can still achieve the desired results. In addition, the process depicted in the accompanying drawings does not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. A steel cutting quality inspection device for ship steel structure processing, the device comprising: A steel cutting module, which is used to determine a steel processing plan based on the ship steel structure, execute the steel processing plan to cut the target steel, and obtain steel cutting sample information; A projection point feature extraction module, the projection point feature extraction module is used to perform data projection based on the steel cutting sample information, extract projection point features according to the projection results, and obtain multiple sample features, wherein the multiple sample features include multiple sample internal features and multiple sample appearance features; A quality inspection channel construction module, wherein the quality inspection channel construction module is used to construct a quality inspection channel, wherein the quality inspection channel includes an internal quality inspection branch and an appearance quality inspection branch; An internal inspection data set generation module, the internal inspection data set generation module is used to synchronize the internal features of the multiple samples with the material ratio information of the target cutting piece to the internal quality inspection branch to inspect the target cutting piece and generate an internal inspection data set; An appearance inspection data set generation module, the appearance inspection data set generation module is used to synchronize the appearance features of the multiple samples with the spatial structure information of the target cut piece to the appearance quality inspection branch to inspect the target cut piece and generate an appearance inspection data set; A quality collaborative analysis module, the quality collaborative analysis module is used to combine the internal inspection data set with the appearance inspection data set to perform a quality collaborative analysis on the target cut part to generate a quality collaborative determination result; The abnormality tracing module is used to trace the abnormality according to the quality collaborative determination result, and to mark the target cutting part according to the traceability result.

2. The steel cutting quality inspection device for ship steel structure processing according to claim 1, characterized in that: The projection point feature extraction module is used for: Convert steel cutting sample information into multiple data points and construct a sample data dot matrix; Performing dimensionality reduction on the sample data lattice to generate a sample dimensionality reduction data lattice; Based on wavelet transform, the sample dimension reduction data lattice is projected into a low-dimensional space to obtain a projection result; Determine whether the projection point in the projection result is greater than or equal to a preset threshold; Extracting a plurality of projection points greater than or equal to the preset threshold for analysis to generate a plurality of sample appearance features; Extracting a plurality of projection points smaller than the preset threshold for analysis to generate a plurality of sample internal features; The multiple sample appearance features and the multiple sample internal features are integrated to obtain the multiple sample features.

3. The steel cutting quality inspection device for ship steel structure processing according to claim 2, characterized in that: The projection point feature extraction module is used for: Preprocessing the sample data dot matrix, performing data standardization according to the preprocessing result, and determining the sample standard data dot matrix; Traversing the sample standard data point matrix, extracting the ith data point and the jth data point, calculating through the covariance calculation expression, obtaining the covariance between the ith data point and the jth data point, thereby iterating to obtain a covariance matrix; Performing eigendecomposition based on the covariance matrix to obtain a plurality of data eigenvalues ​​and a plurality of data eigenvectors, wherein the plurality of data eigenvalues ​​and the plurality of data eigenvectors have a corresponding relationship; Sorting the multiple data eigenvalues ​​in descending order to obtain an eigenvalue sequence, and extracting first m eigenvalues ​​and m eigenvectors according to the eigenvalue sequence; The sample dimension reduction data lattice is generated based on the sample data lattice combined with the first m eigenvalues ​​and m eigenvectors.

4. The steel cutting quality inspection device for ship steel structure processing according to claim 3, characterized in that: The projection point feature extraction module is used for: De-noising the sample data dot matrix using a median filter method to generate a sample de-noised data dot matrix; Dividing the sample denoised data point matrix into a training data point set and a test training point set; Based on the training data point set and the test training point set, the sample denoising data points are evaluated by cross-validation to generate a denoising evaluation result; The preprocessing result is generated according to the denoising evaluation result.

5. The steel cutting quality inspection device for ship steel structure processing according to claim 1, characterized in that: The internal verification data set generation module is used for: Determine the material type information of the target steel material based on the ship steel structure, analyze the material property information according to the material type information, and determine the material ratio information of the target cutting part according to the analysis result; Establishing data association between the internal features of the plurality of samples and the material ratio information to generate an association factor; Based on the correlation factor, synchronizing the plurality of sample internal features with the material ratio information of the target cutting piece to the internal quality inspection branch; Formulate internal testing plans for samples according to the physical and chemical properties of the target steel through the internal quality inspection branch; Execute the sample internal test scheme to generate sample test results, traverse the material ratio information and match it with the sample test results, evaluate the quality of the target cutting piece according to the matching results, and generate an internal quality score; The internal quality score is added to the internal validation data set.

6. The steel cutting quality inspection device for ship steel structure processing according to claim 1, characterized in that: The appearance inspection data set generation module is used for: Traversing the spatial structure information to determine a plurality of structural dimension information and a plurality of spatial position relationships; Retrieve historical appearance quality inspection reports to extract historical structural dimension information and historical spatial position relationships to formulate inspection standards; Synchronize the multiple sample appearance features in combination with the spatial structure information of the target cut piece to the appearance quality inspection branch, traverse the multiple sample appearance features, compare the multiple structural dimension information and the multiple spatial position relationships with the inspection standard, and classify and label the multiple sample appearance features according to the comparison results to obtain multiple category labels; A plurality of abnormal labels are extracted based on the plurality of category labels, and a plurality of abnormal appearance features are added to the appearance inspection dataset according to the plurality of abnormal labels.

7. The steel cutting quality inspection device for ship steel structure processing according to claim 1, characterized in that: The quality collaborative analysis module is used to: Calculate based on internal inspection evaluation index and appearance inspection evaluation index to generate quality impact coefficient; Performing weight allocation on the internal inspection data set and the appearance inspection data set according to the quality influence coefficient to generate a weight allocation result; Combining the internal inspection data set with the appearance inspection data set in combination with the weight distribution result to generate a joint inspection data set; Based on the joint inspection data set, determining whether the inspection result of the target cut part is within a preset quality standard range; If the inspection result of the target cut piece is within the preset quality standard range, the quality of the target cut piece is determined to be qualified, and the quality collaborative determination result is generated; If the inspection result of the target cut part is not within the preset quality standard interval, the quality of the target cut part is determined to be unqualified, and a plurality of abnormal levels are divided according to the inspection result that is not within the preset quality standard interval. The plurality of abnormal levels are traversed to match the target cut parts with unqualified quality, and the quality collaborative judgment result is generated.

8. The steel cutting quality inspection device for ship steel structure processing according to claim 7, characterized in that: The exception tracing module is used to: Based on the multiple abnormality levels and the quality collaborative determination result, the internal inspection data set of the target cut part is traced for abnormality to generate internal traceability data; Based on the multiple abnormality levels and the quality collaborative determination result, the appearance inspection data set of the target cut part is traced for abnormality, and external traceability data is generated; Performing internal abnormality feedback based on the internal traceability data to determine the first quality abnormality source; Performing external abnormality feedback based on the external traceability data to determine the second quality abnormality source; The first quality abnormality source and the second quality abnormality source are added to the traceability result.

Citation Information

Patent Citations

  • Automatic control system for nickel-based target material production

    CN113485267A

  • Seamless steel tube one-by-one tracking production system

    CN114425561A

  • Quality detection method and system for nuclear grade alloy steel elbow

    CN115077618A

  • Method and device for measuring dimension of steel rail piece

    CN115824070A

  • Stainless steel chain product quality inspection method and system

    CN116165213A