Titanium alloy bar quality detection system and method

Abstract

The invention relates to the technical field of titanium alloy bar detection, and discloses a titanium alloy bar quality detection system and method. The system comprises a multi-modal data acquisition module, a detection reference dynamic optimization module, a data standardization processing module, a multi-modal feature fusion module, an intelligent defect identification and evaluation module, a quality grade evaluation module and a quality detection report generation module. When the quality of the titanium alloy bar is detected, through a multi-mode nondestructive detection mechanism integrating ultrasonic detection data, optical surface detection data and eddy current electromagnetic detection data, all-dimensional defect information of the interior, the surface and the near surface of the bar is synchronously obtained, and complementation and cross validation of the multi-mode detection data are achieved; the problem of limitation of a single sensor detection visual angle is solved, the missed detection and misjudgment risks are reduced, and the comprehensiveness and reliability of a defect detection result are ensured.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of titanium alloy rod detection, in particular to a titanium alloy rod quality detection system and method. BACKGROUND

[0002] Titanium alloy rods are important metal materials, and are widely used in many fields due to excellent performance and low cost. At present, the global titanium alloy production is steadily increasing, and titanium alloy materials are indispensable. Titanium alloy is used in important fields such as automobiles, aerospace and ships. Therefore, the production quality detection of titanium alloy is particularly important.

[0003] At present, due to the complex defect morphology and random distribution of titanium alloy rods, when multi-modal non-destructive detection is carried out, only a single sensor can obtain the feature information of a specific dimension of the rod, and cannot realize the synchronous and collaborative perception and data correlation of internal, surface and near-surface defects. When single detection mode data distortion and environmental interference occur, it will lead to missed detection and misjudgment of defect characteristics, and cannot guarantee the comprehensiveness and reliability of the multi-modal detection result.

[0004] Therefore, the present application provides a titanium alloy rod quality detection system and method to solve the above problems. SUMMARY

[0005] In view of the deficiencies of the prior art, the present application provides a titanium alloy rod quality detection system and method. The present application solves the problem of missed detection and misjudgment of defect characteristics in the background art, and cannot guarantee the comprehensiveness and reliability of the multi-modal detection result.

[0006] To achieve the above purpose, the present application provides the following technical scheme: a titanium alloy rod quality detection system and method, the method comprising the following steps:

[0007] S1, collecting multi-modal non-destructive detection data of the titanium alloy rod, including ultrasonic detection data, optical surface detection data and eddy current electromagnetic detection data;

[0008] S2, based on the ultrasonic detection data, optical surface detection data and eddy current electromagnetic detection data, respectively, detecting the reference calibration processing, generating dynamic optimized ultrasonic detection reference parameters, optical detection reference parameters and electromagnetic detection reference parameters;

[0009] S3, according to the dynamic optimized ultrasonic detection reference parameters, optical detection reference parameters and electromagnetic detection reference parameters, the corresponding detection data is subjected to real-time standardization processing, generating standardized ultrasonic detection data, standardized optical detection data and standardized electromagnetic detection data;

[0010] S4, performing multimodal feature fusion processing on the standardized ultrasonic detection data, standardized optical detection data and standardized electromagnetic detection data, extracting stereoscopic defect feature information of the titanium alloy bar, and generating a multimodal fusion defect feature matrix;

[0011] S5, based on the multimodal fusion defect feature matrix and a pre-stored typical defect feature database, performing intelligent matching analysis to generate a defect type recognition result and defect quantification evaluation data;

[0012] S6, according to the defect type recognition result and the defect quantification evaluation data, combining a titanium alloy bar quality rating standard to perform quality grade determination processing, and generating a titanium alloy bar quality grade evaluation result;

[0013] S7, constructing a full-factor quality detection report containing multimodal detection data, defect feature information, recognition result and quality evaluation, and outputting titanium alloy bar quality detection certification data.

[0014] Preferably, the S1 of collecting multimodal non-destructive detection data comprises the following steps:

[0015] S11, synchronously collecting ultrasonic detection data, optical surface detection data and eddy current electromagnetic detection data of the titanium alloy bar through a multi-sensor integrated detection platform:

[0016] The ultrasonic detection data adopts a phased array ultrasonic probe to obtain echo signal characteristics of internal defects of the bar;

[0017] The optical surface detection data adopts a high-resolution industrial camera array to obtain a bar surface micro-topography image;

[0018] The eddy current electromagnetic detection data adopts a multi-frequency eddy current probe to obtain near-surface electromagnetic characteristic parameters of the bar.

[0019] Preferably, the S2 of performing detection reference calibration processing comprises the following steps:

[0020] S21, establishing a dynamic reference adjustment model based on material characteristics, the model input including titanium alloy grade, heat treatment state, bar specification parameters and environmental temperature parameters;

[0021] S22, using an adaptive filtering algorithm to perform noise suppression processing on the original detection signal to generate detection data with optimized signal-to-noise ratio;

[0022] S23, acquiring system error parameters in real time through a reference calibration device, and using a compensation algorithm to generate the dynamically optimized ultrasonic detection reference parameters, optical detection reference parameters and electromagnetic detection reference parameters.

[0023] Preferably, the real-time standardization of the detection data in S3 comprises the following steps:

[0024] S31, establishing a data standardization conversion function according to the dynamically optimized detection reference parameters:

[0025] ;

[0026] wherein is the standardized detection data, is the original detection data, is the reference offset, is the range coefficient, is the sensitivity factor, is the environmental compensation amount;

[0027] S32, performing time domain normalization on the ultrasonic detection data, pixel gray scale standardization on the optical detection data, and impedance plane normalization on the electromagnetic detection data.

[0028] Preferably, the pixel gray scale standardization on the optical detection data in S32 comprises the following steps:

[0029] S321, obtaining a pre-stored surface image of a defect-free standard titanium alloy bar, and extracting a gray scale histogram thereof as a standard distribution template;

[0030] S322, calculating a probability density function of a gray scale histogram of a to-be-processed optical surface detection image;

[0031] S323, mapping the gray scale distribution of the to-be-processed image to the standard distribution template through a cumulative distribution function transformation;

[0032] S324, performing local contrast enhancement on the mapped image by using a limited contrast adaptive histogram equalization algorithm, to generate enhanced standardized optical detection data.

[0033] Preferably, the multi-modal feature fusion processing in S4 comprises the following steps:

[0034] S41, extracting time-frequency features of the ultrasonic detection data by using a wavelet transform algorithm, to generate an ultrasonic defect feature vector;

[0035] S42, extracting texture features of the optical surface detection data by using a deep learning convolutional neural network, to generate a surface defect feature vector;

[0036] S43, extracting impedance features of the electromagnetic detection data by using a principal component analysis method, to generate an electromagnetic defect feature vector;

[0037] S44, fusing the ultrasonic defect feature vector, the surface defect feature vector and the electromagnetic defect feature vector into a unified multimodal fusion defect feature matrix through a feature level fusion algorithm:

[0038] ;

[0039] wherein is a fusion feature vector, is an ultrasonic feature vector, is an optical feature vector, is an electromagnetic feature vector, are weights assigned to the ultrasonic, optical and electromagnetic features, respectively.

[0040] Preferably, the intelligent matching analysis in S5 includes the following steps:

[0041] S51, establishing a typical defect feature database, the database containing multimodal feature templates of four typical defects of pores, inclusions, cracks and segregation;

[0042] S52, calculating the similarity measure of the multimodal fusion defect feature matrix and the typical defect feature template using an improved dynamic time warping algorithm;

[0043] S53, generating a defect type recognition result and a defect confidence index through a fuzzy inference system;

[0044] S54, generating quantitative evaluation data of defect size, depth and orientation based on feature parameter regression analysis.

[0045] Preferably, the quality grade determination process in S6 includes the following steps:

[0046] S61, establishing a quality grading rule library according to titanium alloy bar application standards, the rule library containing quality requirements of different application grades of aerospace grade, medical grade and industrial grade;

[0047] S62, comprehensively evaluating defect type, defect size, defect location and defect density parameters using a multi-criteria decision analysis method;

[0048] S63, generating a titanium alloy bar quality grade evaluation result through a weighted scoring model, and outputting a quality compliance certification indication.

[0049] Preferably, the construction of the full-factor quality detection report in S7 includes the following steps:

[0050] S71, constructing a full-factor quality data chain containing original detection data, processing process data and analysis result data;

[0051] S72, generating an unforgeable quality detection digital fingerprint by using a blockchain technology;

[0052] S73, outputting a titanium alloy bar quality detection report and digital certification conforming to international standards.

[0053] Preferably, the system comprises:

[0054] A multi-modal data acquisition module acquires ultrasonic, optical and electromagnetic detection data of the titanium alloy bar and outputs original multi-modal detection data;

[0055] A detection reference dynamic optimization module receives the original multi-modal detection data and material process parameters, processes through a dynamic reference adjustment model and an adaptive filtering algorithm, and outputs dynamically optimized ultrasonic, optical and electromagnetic detection reference parameters;

[0056] A data standardization processing module receives the original multi-modal detection data and the dynamically optimized detection reference parameters, processes through a data standardization conversion function, and outputs standardized ultrasonic detection data, standardized optical detection data and standardized electromagnetic detection data;

[0057] A multi-modal feature fusion module receives the standardized multi-modal detection data, processes through wavelet transform, convolutional neural network and principal component analysis algorithm for feature extraction and fusion, and outputs a multi-modal fusion defect feature matrix;

[0058] An intelligent defect identification and evaluation module receives the multi-modal fusion defect feature matrix, calls a typical defect feature database and performs intelligent matching and regression analysis, and outputs a defect type identification result and defect quantitative evaluation data;

[0059] A quality grade evaluation module receives the defect type identification result and the defect quantitative evaluation data, processes based on a quality grading rule library and a multi-criteria decision analysis method, and outputs a titanium alloy bar quality grade evaluation result;

[0060] A quality detection report generation module integrates the original multi-modal detection data, the multi-modal fusion defect feature matrix, the defect type identification result, the defect quantitative evaluation data and the quality grade evaluation result, generates and outputs a full-element quality detection report and certification data.

[0061] Advantages

[0062] Compared with the prior art, the titanium alloy bar quality detection system and method provided by the present application have the following advantages:

[0063] 1. In the present application, when the titanium alloy bar quality is detected, through the multi-modal non-destructive testing mechanism of integrating ultrasonic detection data, optical surface detection data and eddy current electromagnetic detection data, the comprehensive defect information of the bar inside, surface and near surface is synchronously obtained, the complementary and cross verification of multi-modal detection data are realized, the problem of single sensor detection perspective limitation is overcome, the risk of missed detection and misjudgment is reduced, and the comprehensiveness and reliability of the defect detection result are ensured.

[0064] 2. In the present application, when the titanium alloy bar quality is detected, through the establishment of a dynamic reference adjustment model based on titanium alloy grade, heat treatment state, bar specification parameters and environmental temperature parameters, dynamic optimized ultrasonic detection reference parameters, optical detection reference parameters and electromagnetic detection reference parameters can be generated in real time according to specific process parameters, and the detection data is standardized in real time to generate standardized ultrasonic detection data, standardized optical detection data and standardized electromagnetic detection data, so that the system can automatically adapt to the detection requirements of different batches and different specifications of bars, eliminate systematic measurement deviation, and ensure the consistency of the detection process and the comparability of the results.

[0065] 3. In the present application, when the titanium alloy bar quality is detected, a multi-modal fusion defect feature matrix is constructed by multi-modal feature fusion processing, and intelligent matching analysis is performed with a pre-stored typical defect feature database, a multi-criteria decision analysis method is used to comprehensively evaluate the defect type, defect size, defect position and defect density parameters, generate defect type recognition results and defect quantitative evaluation data, and combine the titanium alloy bar quality rating standard to perform quality grade determination processing, realize the objective recognition of complex defects and the generation of titanium alloy bar quality grade evaluation results, avoid the dependence on artificial experience of traditional methods, and improve the objectivity, accuracy and overall detection efficiency of quality detection. BRIEF DESCRIPTION OF DRAWINGS

[0066] Fig. 1 The flowchart of the titanium alloy bar quality detection method of the present application;

[0067] Fig. 2 The framework diagram of the titanium alloy bar quality detection system of the present application. DETAILED DESCRIPTION

[0068] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0069] Specific embodiments: please refer to Figs. 1-2A titanium alloy bar quality detection system and method, the method comprising the following steps:

[0070] S1, collecting multi-modal non-destructive testing data of the titanium alloy bar, including ultrasonic testing data, optical surface testing data and eddy current electromagnetic testing data;

[0071] S2, based on the ultrasonic testing data, the optical surface testing data and the eddy current electromagnetic testing data, respectively, the detection reference calibration processing is carried out, and the dynamic optimized ultrasonic testing reference parameter, the optical testing reference parameter and the electromagnetic testing reference parameter are generated;

[0072] S3, according to the dynamic optimized ultrasonic testing reference parameter, the optical testing reference parameter and the electromagnetic testing reference parameter, the corresponding detection data is carried out real-time standardization processing, and the standardized ultrasonic testing data, the standardized optical testing data and the standardized electromagnetic testing data are generated;

[0073] S4, the multi-modal feature fusion processing is carried out on the standardized ultrasonic testing data, the standardized optical testing data and the standardized electromagnetic testing data, the three-dimensional defect feature information of the titanium alloy bar is extracted, and a multi-modal fusion defect feature matrix is generated;

[0074] S5, based on the multi-modal fusion defect feature matrix and the pre-stored typical defect feature database, intelligent matching analysis is carried out, and defect type recognition result and defect quantitative evaluation data are generated;

[0075] S6, according to the defect type recognition result and the defect quantitative evaluation data, combined with the titanium alloy bar quality rating standard, quality grade judgment processing is carried out, and titanium alloy bar quality grade evaluation result is generated;

[0076] S7, a full-factor quality detection report containing multi-modal detection data, defect feature information, recognition result and quality evaluation is constructed, and titanium alloy bar quality detection certification data is output.

[0077] The multi-modal non-destructive testing data collected in S1 includes the following steps:

[0078] S11, the ultrasonic testing data, the optical surface testing data and the eddy current electromagnetic testing data of the titanium alloy bar are synchronously collected through a multi-sensor integrated detection platform:

[0079] The ultrasonic testing data adopts phased array ultrasonic probe to obtain the echo signal characteristics of the internal defects of the bar;

[0080] The optical surface testing data adopts high-resolution industrial camera array to obtain the micro-topography image of the bar surface;

[0081] The eddy current electromagnetic testing data adopts multi-frequency eddy current probe to obtain the near-surface electromagnetic characteristic parameters of the bar.

[0082] The calibration process for the detection benchmark in S2 includes the following steps:

[0083] S21. Establish a dynamic benchmark adjustment model based on material properties. The model inputs include titanium alloy grade, heat treatment state, bar specifications and ambient temperature parameters.

[0084] S22. An adaptive filtering algorithm is used to suppress noise in the original detection signal to generate detection data with optimized signal-to-noise ratio. This includes the following steps:

[0085] S221. Using the original detection signal as the desired response signal of the adaptive filter. ;

[0086] S222, Using the signal collected by the environmental noise sensor as the reference input signal of the filter. ;

[0087] S223. Update the filter weight vector using the least mean square algorithm. Its iterative formula is:

[0088] ;

[0089] Where is the time. The expected response signal is at time . The reference input noise signal is at time . The filter weight vector, For a moment The input signal vector, Step size factor For a moment The output error signal;

[0090] S224, Output the filter The signal is used as the denoised signal, and the signal-to-noise ratio improvement is calculated. If the improvement does not reach the preset threshold, the step size factor is automatically adjusted. Then iterate again until detection data with optimized signal-to-noise ratio is generated;

[0091] S23. The system error parameters are obtained in real time through the reference calibration device, and the compensation algorithm is used to generate dynamically optimized ultrasonic detection reference parameters, optical detection reference parameters and electromagnetic detection reference parameters.

[0092] The real-time standardization of detection data in S3 includes the following steps:

[0093] S31. Based on the dynamically optimized detection benchmark parameters, establish a data standardization transformation function:

[0094] ;

[0095] wherein is the normalized detection data, is the original detection data, is the reference offset, is the range coefficient, is the sensitivity factor, is the environmental compensation amount;

[0096] S32, the ultrasonic detection data is processed by time domain normalization, the optical detection data is processed by pixel gray standardization, and the electromagnetic detection data is processed by impedance plane normalization.

[0097] In S32, the optical detection data is processed by pixel gray standardization, and the surface image gray distribution is mapped to a standard distribution template through histogram specification technology, specifically including the following steps:

[0098] S321, a pre-stored surface image of a defect-free standard titanium alloy bar is obtained, and a gray histogram thereof is extracted as a standard distribution template;

[0099] S322, a gray histogram probability density function of the optical surface detection image to be processed is calculated;

[0100] S323, the gray distribution of the image to be processed is mapped to the standard distribution template through cumulative distribution function transformation;

[0101] S324, a limited contrast adaptive histogram equalization algorithm is used to enhance the local contrast of the mapped image, and the enhanced standardized optical detection data is generated.

[0102] The multi-modal feature fusion processing in S4 includes the following steps:

[0103] S41, a wavelet transform algorithm is used to extract the time-frequency features of the ultrasonic detection data, and an ultrasonic defect feature vector is generated, specifically including the following steps:

[0104] S411, Db4 wavelet is selected as the base wavelet, the standardized ultrasonic detection data is decomposed by 3-layer wavelet packet, and the wavelet packet coefficients of each node are obtained

[0105] S412, the energy values of the wavelet packet coefficients of each frequency band are calculated , and the calculation formula is:

[0106] ;

[0107] wherein is the wavelet packet coefficient of the th node in the th layer, is The first The first The energy value of the wavelet packet coefficient of the first node, The number of layers of wavelet packet decomposition, The node serial number index in the specified layer, and N is the coefficient length;

[0108] S413, arrange the energy values in frequency band order to form an ultrasonic defect feature vector , which is represented as:

[0109] ;

[0110] S42, use a deep learning convolutional neural network to extract the texture features of the optical surface detection data and generate a surface defect feature vector, which includes the following steps:

[0111] S421, construct a convolutional neural network model based on the ResNet-18 architecture, with the input being the surface image in the standardized optical detection data;

[0112] S422, use the pre-trained model weight for transfer learning, and use the titanium alloy surface image dataset containing defects for fine-tuning training of the last two layers of the network;

[0113] S423, output the global average pooling layer before the last fully connected layer of the network as a high-dimensional texture feature;

[0114] S424, perform dimensionality reduction processing on the high-dimensional texture feature to generate the final surface defect feature vector ;

[0115] S43, use principal component analysis to extract the impedance features of the electromagnetic detection data and generate an electromagnetic defect feature vector, which includes the following steps:

[0116] S431, organize the standardized electromagnetic detection data into a sample-feature matrix , where each row represents a sampling point and each column represents an impedance value at a frequency;

[0117] S432, center the matrix and calculate its covariance matrix

[0118] S433, solve the eigenvalues and eigenvectors of the covariance matrix , and arrange them in descending order of eigenvalue size;

[0119] S434, select the first principal components, so that the cumulative variance contribution rate , the calculation formula is:

[0120] ;

[0121] wherein is the total number of original features, is the th feature, is the number of selected principal components;

[0122] S435, projecting the original data to the selected principal component direction to generate the dimensionality-reduced electromagnetic defect feature vector ;

[0123] S44, fusing the ultrasonic defect feature vector, surface defect feature vector and electromagnetic defect feature vector into a unified multi-modal fusion defect feature matrix through a feature-level fusion algorithm:

[0124] ;

[0125] wherein is the fusion feature vector, is the ultrasonic feature vector, is the optical feature vector, is the electromagnetic feature vector, respectively, are the weights assigned to the ultrasonic, optical and electromagnetic features.

[0126] The intelligent matching analysis in S5 includes the following steps:

[0127] S51, establishing a typical defect feature database, the database containing multi-modal feature templates of four typical defects of pores, inclusions, cracks and segregation;

[0128] S52, using an improved dynamic time warping algorithm to calculate the similarity measure of the multi-modal fusion defect feature matrix and the typical defect feature template, wherein the improved dynamic time warping algorithm introduces a weight coefficient to strengthen the matching of key feature areas, which includes the following steps:

[0129] S521, calculating the Euclidean distance matrix between the multi-modal fusion defect feature matrix and the typical defect feature template in each dimension feature sequence ;

[0130] S522, according to the physical saliency of the defect feature, assigning dynamic weight coefficients to different areas in the distance matrix, wherein the weight coefficient of the defect center area is greater than the weight coefficient of the edge area ;

[0131] S523, constructing a cumulative cost matrix , the recursive calculation formula of which is:

[0132] ;

[0133] wherein is the value of the point in the cumulative cost matrix, is the Euclidean distance between the point in the feature sequence and the point in the template sequence, is the corresponding dynamic weight coefficient; S524, find the optimal normalization path by backtracking the minimum path of the cumulative cost matrix, and output the final similarity measure value;

[0134] S524, find the optimal normalization path by backtracking the minimum path of the cumulative cost matrix, and output the final similarity measure value;

[0135] S53, generate defect type identification results and defect confidence indicators through the fuzzy inference system, specifically including the following steps:

[0136] S531, set the input variable of the fuzzy inference system as the similarity measure value, and its fuzzy subset includes {low, medium, high};

[0137] S532, set the output variable as the defect type confidence, and its fuzzy subset includes {pore, inclusion, crack, segregation};

[0138] S533, establish a fuzzy inference rule base to define the mapping relationship between the fuzzy subset of the similarity measure value and the defect type confidence. When the similarity measure value belongs to the "high" fuzzy subset, the output mapped is the high confidence of the "crack" defect type.

[0139] S534, use the barycentric method to de-fuzzify the confidence output by the fuzzy inference system, and output the clear numerical value as the defect confidence indicator;

[0140] S54, generate quantitative evaluation data of defect size, depth, and orientation based on regression analysis of feature parameters.

[0141] The quality grade determination process in S6 includes the following steps:

[0142] S61, establish a quality grading rule base according to titanium alloy bar application standards, which includes quality requirements for different application levels such as aerospace level, medical level, and industrial level;

[0143] S62, use multi-criteria decision analysis method to comprehensively evaluate defect type, defect size, defect location, and defect density parameters, wherein the multi-criteria decision analysis method is specifically the outranking distance method for comprehensive evaluation, which specifically includes the following steps:

[0144] ​​S621, construct a decision matrix, taking defect type, defect size, defect position and defect density parameters as evaluation indexes;

[0145] S622, normalize the decision matrix to eliminate the influence of different dimensions;

[0146] S623, determine the positive ideal solution and negative ideal solution of each evaluation index according to the quality grading rule base;

[0147] S624, calculate the Euclidean distance of each evaluation scheme and the positive ideal solution and negative ideal solution, and calculate the relative closeness, taking the relative closeness as the basis for comprehensive evaluation;

[0148] S63, generate the titanium alloy bar quality grade evaluation result through the weighted scoring model, and output the quality compliance certification indication, specifically using the linear weighted sum algorithm, and the calculation formula is:

[0149]

[0150] wherein is the final quality comprehensive score, is the weight coefficient of the i-th evaluation index, is the standardized score of the i-th evaluation index, is the total number of evaluation indexes; The quality grading rule base pre-stores the comprehensive score threshold interval corresponding to different quality grades, matches the calculated value with the threshold interval, and generates the titanium alloy bar quality grade evaluation result and the quality compliance certification indication.

[0151] The quality grading rule base pre-stores the comprehensive score threshold interval corresponding to different quality grades, matches the calculated value with the threshold interval, and generates the titanium alloy bar quality grade evaluation result and the quality compliance certification indication.

[0152] S7 in the construction of the full element quality detection report includes the following steps:

[0153] S71, establish a full life cycle quality data chain containing original detection data flow, data processing intermediate parameters, defect feature analysis results and quality judgment conclusion;

[0154] S72, based on time sequence alignment technology, associate and map the multi-modal fusion defect feature matrix with the corresponding spatial position coordinates of the bar to generate a defect distribution atlas with positioning information;

[0155] S73, use data encryption hash algorithm to operate the integrated detection data and generated report to generate an unforgeable digital digest and block chain storage fingerprint;

[0156] S74, automatically generate a structured detection report and visual interactive interface containing defect atlas, quantitative data, grade evaluation and digital authentication information according to industry standard format.​​​

[0157] Wherein the data encryption hash algorithm in S73 adopts SHA-256 algorithm to generate a digital digest, which specifically includes the following steps:

[0158] S731, combine the integrated detection data and the generated report in a pre-defined format, and convert it into a binary data stream;

[0159] S732, padding and message grouping processing is performed on the binary data stream;

[0160] S733, initialize the eight hash initial values of the SHA-256 algorithm;

[0161] S734, perform cyclic compression processing on each message packet, including logical function operation and modulo addition operation;

[0162] S735, finally output a tamper-proof digital digest with a length of 256 bits;

[0163] The blockchain storage evidence fingerprint specifically includes the following steps:

[0164] S736, combine the digital digest, timestamp and device identity to generate storage transaction data;

[0165] S737, broadcast the storage transaction data to the blockchain network;

[0166] S738, after the consensus verification of the blockchain network is passed, the new block hash value containing the storage transaction is fed back as the blockchain storage evidence fingerprint and stored.

[0167] The system comprises:

[0168] A multi-modal data acquisition module acquires ultrasonic, optical and electromagnetic detection data of titanium alloy bars and outputs raw multi-modal detection data;

[0169] A detection reference dynamic optimization module receives raw multi-modal detection data and material process parameters, processes them through a dynamic reference adjustment model and an adaptive filtering algorithm, and outputs dynamically optimized ultrasonic, optical and electromagnetic detection reference parameters;

[0170] A data standardization processing module receives raw multi-modal detection data and dynamically optimized detection reference parameters, processes them through a data standardization conversion function, and outputs standardized ultrasonic detection data, standardized optical detection data and standardized electromagnetic detection data;

[0171] A multi-modal feature fusion module receives standardized multi-modal detection data, extracts and fuses features through wavelet transform, convolutional neural network and principal component analysis algorithm, and outputs a multi-modal fusion defect feature matrix;

[0172] Intelligent defect recognition and evaluation module, receiving multi-modal fusion defect feature matrix, calling typical defect feature database and performing intelligent matching and regression analysis, outputting defect type recognition result and defect quantitative evaluation data;

[0173] Quality grade evaluation module, receiving defect type recognition result and defect quantitative evaluation data, processing based on quality grading rule base and multi-criteria decision analysis method, outputting titanium alloy bar quality grade evaluation result;

[0174] Quality detection report generation module integrates original multi-modal detection data, multi-modal fusion defect feature matrix, defect type recognition result, defect quantitative evaluation data and quality grade evaluation result, generates and outputs full-factor quality detection report and certification data.

[0175] The operation steps of the titanium alloy bar quality detection system and method are as follows:

[0176] Step one, multi-modal data collaborative acquisition

[0177] The system synchronously acquires multi-modal non-destructive detection data of titanium alloy bars through a multi-sensor integrated detection platform. In this process, the phased array ultrasonic probe is used to obtain the echo signal characteristics of internal defects of the bar, the high-resolution industrial camera array is used to capture the surface micro-topography image, and the multi-frequency eddy current probe is used to extract the near-surface electromagnetic characteristic parameters, forming a comprehensive data acquisition system covering the interior, surface and near-surface.

[0178] Step two, detection reference dynamic optimization and data standardization

[0179] A dynamic reference adjustment model based on material characteristics is established, which inputs include titanium alloy grade, heat treatment state, bar specification parameters and environmental temperature parameters. Adaptive filtering algorithm is used to suppress noise of original detection signal, generating detection data with optimized signal-to-noise ratio. The system error parameters are obtained in real time by reference calibration device, and compensation algorithm is used to generate dynamic optimized detection reference parameters. According to the optimized reference parameters, the multi-modal detection data is standardized in real time to eliminate the dimensional difference and generate standardized detection data.

[0180] Step three, multi-modal feature fusion and intelligent recognition

[0181] The system extracts time-frequency features from standardized ultrasonic data using wavelet transform algorithm, extracts texture features from optical images using deep learning convolutional neural network, and extracts impedance features from electromagnetic data using principal component analysis method. The feature vectors of different modalities are fused into a unified multi-modal fusion defect feature matrix through a feature-level fusion algorithm. The feature matrix is intelligently matched with a pre-stored typical defect feature database, an improved dynamic time warping algorithm is used to calculate the similarity measure, and a fuzzy reasoning system is used to generate the defect type recognition result and confidence index.

[0182] Step four, quality decision and report generation

[0183] The system establishes a quality grading rule library according to the titanium alloy bar application standard, uses a multi-criteria decision analysis method to comprehensively evaluate the defect type, size, position and density parameters. The quality grade evaluation result is generated through a weighted scoring model, and the quality compliance certification indication is output. The whole process detection data and analysis results are integrated, the data encryption hash algorithm is used to generate an unforgeable digital digest and a block chain storage fingerprint, and finally a structured detection report containing defect map, quantitative data and authentication information is generated.

[0184] It should be noted that, in this document, relational terms such as first and second and the like can be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises", "comprising", or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without limitation, an element preceded by "comprises... a" does not, without more constraints, foreclose the existence of additional identical elements in the process, method, article, or apparatus that comprises the element.

[0185] Although embodiments of the present application have been shown and described, it is to be understood that various modifications, substitutions, replacements and changes can be made to these embodiments without departing from the principles and spirit of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A method of titanium alloy bar quality detection, characterized in that: The method comprises the following steps: S1, collecting multi-modal non-destructive testing data of titanium alloy bars, including ultrasonic testing data, optical surface testing data and eddy current electromagnetic testing data; S2, based on the ultrasonic testing data, optical surface testing data and eddy current electromagnetic testing data, respectively, performing detection reference calibration processing to generate dynamically optimized ultrasonic testing reference parameters, optical testing reference parameters and electromagnetic testing reference parameters; S3, according to the dynamically optimized ultrasonic testing reference parameters, optical testing reference parameters and electromagnetic testing reference parameters, performing real-time standardization processing on the corresponding detection data to generate standardized ultrasonic testing data, standardized optical testing data and standardized electromagnetic testing data; S4, performing multi-modal feature fusion processing on the standardized ultrasonic testing data, standardized optical testing data and standardized electromagnetic testing data, extracting stereoscopic defect feature information of the titanium alloy bars, and generating a multi-modal fusion defect feature matrix; S5, based on the multi-modal fusion defect feature matrix and a pre-stored typical defect feature database, performing intelligent matching analysis to generate a defect type recognition result and defect quantification evaluation data; S6, according to the defect type recognition result and the defect quantification evaluation data, combining a titanium alloy bar quality rating standard to perform quality grade determination processing to generate a titanium alloy bar quality grade evaluation result; S7, constructing a full-factor quality detection report containing multi-modal detection data, defect feature information, recognition result and quality evaluation, and outputting titanium alloy bar quality detection certification data.

2. The method for detecting the quality of titanium alloy bar according to claim 1, characterized in that: The S1 of collecting multi-modal non-destructive testing data comprises the following steps: S11, synchronously collecting ultrasonic testing data, optical surface testing data and eddy current electromagnetic testing data of titanium alloy bars through a multi-sensor integrated detection platform: The ultrasonic testing data uses a phased array ultrasonic probe to obtain echo signal characteristics of internal defects of the bar; The optical surface testing data uses a high-resolution industrial camera array to obtain a bar surface micro-topography image; The eddy current electromagnetic testing data uses a multi-frequency eddy current probe to obtain near-surface electromagnetic characteristic parameters of the bar.

3. The method of claim 1, wherein: The S2 of performing detection reference calibration processing comprises the following steps: S21, establishing a dynamic reference adjustment model based on material characteristics, the model input comprising titanium alloy grade, heat treatment state, bar specification parameters and environmental temperature parameters; S22, using an adaptive filtering algorithm to perform noise suppression processing on original detection signals to generate detection data with optimized signal-to-noise ratio; S23, real-time acquiring system error parameters through a reference calibration device, and using a compensation algorithm to generate the dynamically optimized ultrasonic testing reference parameters, optical testing reference parameters and electromagnetic testing reference parameters.

4. The method of claim 1, wherein: The S3 of performing real-time standardization processing on the detection data comprises the following steps: S31, according to the dynamically optimized detection reference parameters, establishing a data standardization conversion function: ； wherein is the standardized detection data, is the original detection data, is the reference offset, is the range coefficient, is the sensitivity factor, is the environmental compensation. S32, using time domain normalization processing on the ultrasonic testing data, pixel gray scale standardization processing on the optical testing data, and impedance plane normalization processing on the electromagnetic testing data.

5. The method of claim 4, wherein: The pixel gray scale normalization processing is adopted in the optical detection data in S32, and the gray scale distribution of the surface image is mapped to a standard distribution template through histogram specification technology, specifically including the following steps: S321, a pre-stored surface image of a defect-free standard titanium alloy bar is acquired, and a gray scale histogram thereof is extracted as a standard distribution template; S322, a gray scale histogram probability density function of a to-be-processed optical surface detection image is calculated; S323, the gray scale distribution of the to-be-processed image is mapped to the standard distribution template through a cumulative distribution function transformation; S324, a limited contrast adaptive histogram equalization algorithm is adopted to perform local contrast enhancement on the mapped image, thereby generating enhanced normalized optical detection data.

6. The method of claim 1, wherein: The multi-modal feature fusion processing in S4 includes the following steps: S41, a wavelet transform algorithm is adopted to extract time-frequency features of the ultrasonic detection data, thereby generating an ultrasonic defect feature vector; S42, a deep learning convolutional neural network is adopted to extract texture features of the optical surface detection data, thereby generating a surface defect feature vector; S43, a principal component analysis method is adopted to extract impedance features of the electromagnetic detection data, thereby generating an electromagnetic defect feature vector; S44, a feature-level fusion algorithm is adopted to fuse the ultrasonic defect feature vector, the surface defect feature vector and the electromagnetic defect feature vector into a unified multi-modal fusion defect feature matrix: ； wherein is the fusion feature vector, is the ultrasonic feature vector, is the optical feature vector, is the electromagnetic feature vector, are the weights assigned to the ultrasonic, optical and electromagnetic features, respectively.

7. The method of claim 1, wherein: The intelligent matching analysis in S5 includes the following steps: S51, a typical defect feature database is established, and the database contains multi-modal feature templates of four typical defects, i.e., pores, inclusions, cracks and segregation; S52, an improved dynamic time warping algorithm is adopted to calculate the similarity measurement between the multi-modal fusion defect feature matrix and the typical defect feature template; S53, a fuzzy inference system is adopted to generate a defect type recognition result and a defect confidence index; S54, a feature parameter regression analysis is adopted to generate quantitative evaluation data of defect size, depth and orientation.

8. The method of claim 1, wherein: The quality grade determination processing in S6 includes the following steps: S61, a quality grading rule library is established according to titanium alloy bar application standards, and the rule library contains quality requirements of different application grades, such as aerospace grade, medical grade and industrial grade; S62, a multi-criteria decision analysis method is adopted to comprehensively evaluate defect type, defect size, defect location and defect density parameters; S63, a weighted scoring model is adopted to generate a titanium alloy bar quality grade evaluation result, and a quality conformity certification indication is output.

9. The method of claim 1, wherein: The construction of the full-element quality detection report in S7 includes the following steps: S71, a full-element quality data chain is constructed, which contains original detection data, processing process data and analysis result data; S72, a blockchain technology is adopted to generate an unalterable quality detection digital fingerprint; S73, a titanium alloy bar quality detection report and digital certification conforming to international standards are output.

10. A titanium alloy bar quality detection system for implementing the titanium alloy bar quality detection method of any one of claims 1-9, characterized in that: The system includes: A multi-modal data acquisition module acquires ultrasonic, optical and electromagnetic detection data of a titanium alloy bar, and outputs original multi-modal detection data; A detection reference dynamic optimization module receives the original multi-modal detection data and material process parameters, processes them through a dynamic reference adjustment model and an adaptive filtering algorithm, and outputs dynamically optimized ultrasonic, optical, and electromagnetic detection reference parameters; A data standardization processing module receives the original multi-modal detection data and the dynamically optimized detection reference parameters, processes them through a data standardization conversion function, and outputs standardized ultrasonic detection data, standardized optical detection data, and standardized electromagnetic detection data; A multi-modal feature fusion module receives the standardized multi-modal detection data, extracts and fuses features through wavelet transform, convolutional neural network, and principal component analysis algorithm, and outputs a multi-modal fusion defect feature matrix; An intelligent defect identification and evaluation module receives the multi-modal fusion defect feature matrix, calls a typical defect feature database, and performs intelligent matching and regression analysis, and outputs defect type identification results and defect quantitative evaluation data; A quality grade evaluation module receives the defect type identification results and defect quantitative evaluation data, processes them based on a quality grading rule library and a multi-criteria decision analysis method, and outputs titanium alloy bar quality grade evaluation results; A quality detection report generation module integrates the original multi-modal detection data, multi-modal fusion defect feature matrix, defect type identification results, defect quantitative evaluation data, and quality grade evaluation results, generates and outputs a full-element quality detection report and certification data.

Images