Intelligent non-invasive workpiece traceability system based on ultrasonic phased array technology

Abstract

The invention relates to the technical field of industrial production traceability, in particular to an intelligent non-invasive workpiece traceability system based on an ultrasonic phased array technology, which comprises a data collection module, a feature extraction module and a traceability module, the interior of the workpiece is scanned at multiple angles, and full-matrix data are generated; the feature extraction module adopts a convolutional neural network to automatically extract internal structure and component distribution features, and generates a unique original traceability code; the traceability module associatively stores the data and the codes, re-extracts features in subsequent processing, matches the features with the database, and determines the identity of the workpiece; according to the system, the problems of surface treatment interference and internal information loss are solved, full-life-cycle traceability and quality monitoring are realized through noninvasive accurate acquisition, feature extraction and dynamic matching mechanisms, and the process reliability and the data security are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of industrial production traceability, in particular to an intelligent non-invasive workpiece traceability system based on ultrasonic phased array technology. BACKGROUND

[0002] Traditional workpiece traceability technology relies on barcodes, radio frequency identification or surface detection methods, which are difficult to obtain internal information of workpieces, especially after the workpiece undergoes surface treatment such as polishing and painting, the identification is easily damaged or cannot penetrate the surface layer, resulting in traceability failure. The existing detection method based on ultrasonic wave usually adopts fixed scanning of single array probe, the detection angle and range are limited, which cannot fully reflect the internal structure characteristics of complex workpieces, and lacks adaptability to dynamic processing process.

[0003] In addition, the existing system relies on manual setting of parameters in the feature extraction link, which is difficult to accurately identify the subtle differences in the internal workpiece, resulting in repeated or misjudged traceability coding; the database architecture is single, which cannot meet the mass data storage and secure access requirements, and lacks real-time feedback mechanism of processing quality deviation, making it difficult to realize closed-loop optimization of process. SUMMARY

[0004] The purpose of the present application is to provide an intelligent non-invasive workpiece traceability system based on ultrasonic phased array technology to solve the problems raised in the background art. The specific technology includes how to non-invasively obtain internal full-dimensional information of workpieces and generate unique identification to avoid surface treatment interference, and how to establish a dynamic quality monitoring mechanism to realize closed-loop optimization of processing technology.

[0005] To achieve the above purpose, the present application provides the following technical solutions: The workpiece traceability system includes a data collection module, a feature extraction module and a traceability module, wherein: The data collection module uses ultrasonic phased array equipment to control the excitation order and time delay of the probe array elements, so that ultrasonic waves are incident to the internal workpiece at different angles and directions, and full-matrix data containing internal structure and composition distribution are generated by receiving reflected echoes in all directions; according to the workpiece material and shape, the preset parameters are automatically matched, for example, the high-frequency transmission frequency, power and probe angle are automatically set for high-hardness alloy steel, to ensure that the ultrasonic wave penetrates the complex structure and obtains high-precision information; The intelligent data acquisition interface automatically calibrates the equipment parameters (such as frequency, power, probe angle) before acquisition, and monitors the data quality in real time during acquisition. If an abnormality is found, parameter adjustment or reacquisition is triggered immediately to ensure data accuracy and stability.

[0006] The feature information extraction unit in the feature extraction module adopts a multi-layer convolutional neural network model, extracts internal structure features and component distribution features through convolution layers sliding on full matrix data, reduces dimensions of the features and retains key information through a pooling layer, and integrates the features through a full connection layer to generate a unique feature vector representing the workpiece; The encoding unit in the feature extraction module generates a unique original traceability code based on the feature vector using a trained convolutional neural network encoder; the encoder quantifies the difference between the predicted code and the real code through a multi-class cross-entropy loss function, optimizes the network parameters combined with a back propagation algorithm, and ensures the uniqueness and anti-interference ability of the code; the formula of the multi-class cross-entropy loss function is as follows: , wherein is the number of training samples; represents the number of different workpiece feature combination categories; represents the real probability that the training sample belongs to the category , is the probability that the model predicts that the sample belongs to the category .

[0007] The traceability module stores full matrix data, feature information and original traceability codes in a distributed database, supports high concurrency access and data expansion, sets multiple access permissions (administrator, engineer, ordinary operator), the administrator can manage data, the engineer can analyze data for process improvement, and the ordinary operator is limited to query function, and data security is ensured through regular data backup and fast recovery mechanism.

[0008] In the subsequent processing link of the workpiece, the ultrasonic parameters are optimized according to the surface treatment layer characteristics, for example, the emission power is increased to penetrate the surface for a paint spraying layer, the frequency range is adjusted to reduce attenuation, and the real internal information is ensured to be obtained; The feature weight is dynamically adjusted through a multi-feature fusion similarity calculation algorithm, the component distribution feature weight susceptible to influence is reduced, the internal structure feature weight is enhanced, the matching result is determined combined with a comprehensive similarity threshold, and non-invasive traceability is realized.

[0009] Quality warning is triggered based on the similarity difference value, when the deviation of the current workpiece feature and the original feature in the database exceeds a preset threshold, the abnormal workpiece is automatically marked and warning is pushed; the root cause of the processing defect is located reversely through feature correlation analysis, parameter correction instructions are generated, real-time feedback is fed back to the processing equipment through an industrial internet of things interface, process parameters such as numerical control machine tools and welding equipment are dynamically optimized, a "detection-analysis-regulation" closed loop is formed, and historical data is aggregated to train a self-learning model to improve the prediction accuracy.

[0010] Compared with the prior art, the present application has the following advantages: Through multi-angle scanning of ultrasonic phased array and intelligent parameter adaptation, non-invasive and accurate collection of internal information of the workpiece is realized; based on the feature extraction and coding mechanism of the convolutional neural network, the uniqueness and anti-interference ability of the traceability are improved; the distributed database and multi-level permission management ensure data security and scalability; dynamic quality deviation early warning and process parameter closed-loop optimization effectively solve the quality drift problem in the processing process, and the application has high-precision traceability and real-time process control ability, and is suitable for full life cycle quality management in complex industrial scenes. BRIEF DESCRIPTION OF DRAWINGS

[0011] Figure 1 A full matrix collection flowchart in the data collection process of the application is shown in the figure. Figure 2 An example diagram of the device probe detecting a workpiece in a batch is shown in the figure. Figure 3 An example diagram of the device detecting a workpiece is shown in the figure. Figure 4 An image background diagram of the same workpiece at the initial stage detected by the device is shown in the figure. Figure 5 An image background diagram of the same workpiece at the processing stage detected by the device is shown in the figure. Figure 6 A system workflow diagram of the application is shown in the figure. Figure 7 A schematic diagram of the overall module unit of the application is shown in the figure.

[0012] In the figure: 100, data collection module; 200, feature extraction module; 201, feature information extraction unit; 202, coding unit; 300, traceability module. DETAILED DESCRIPTION

[0013] The technical solutions in the embodiments of the application will be described in detail below with reference to the drawings in the application. Obviously, the described embodiments are only a part of the embodiments of the application, not all. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the application.

[0014] Next, the application provides a technical solution: an intelligent non-invasive workpiece traceability system based on ultrasonic phased array technology, including a data collection module 100, a feature extraction module 200 and a traceability module 300.

[0015] At the initial stage of processing a batch of workpieces, the same device workpoint is specified for the same batch of workpieces, and the data collection module 100 collects the internal original information of each workpiece using an ultrasonic phased array device to form a full matrix data, which specifically includes: Referring to Figures 1-2 , by using ultrasonic phased array technology, by controlling the excitation sequence and time delay of the probe array elements, the ultrasonic waves are incident to the workpiece interior at different angles and directions, and the reflected echoes are received in all directions, so as to obtain comprehensive and detailed internal original information of the workpiece; the probe of the ultrasonic phased array equipment is composed of multiple array elements, by accurately controlling the excitation time and intensity of each array element, the focusing and scanning of ultrasonic waves are realized, for example, when detecting large mechanical parts, by adjusting the excitation sequence and time delay, the ultrasonic waves can cover the entire internal part of the part, and the information of each part is obtained.

[0016] In the initial stage of workpiece processing, the data collection module 100 is composed of an ultrasonic phased array equipment and an intelligent data acquisition interface, and according to the material and shape of the workpiece, the preset parameters are automatically matched, the internal original information is accurately collected, and the full matrix data is generated. For example, when processing a batch of special alloy steel workpieces, according to the characteristics of alloy steel, the transmission frequency of the ultrasonic phased array equipment is set to 7MHz, the power is set to 35W, and the probe angle is set to 45°, in order to collect internal original information, because the hardness of alloy steel is high and the internal structure is complex, the frequency of 7MHz can better penetrate and obtain detailed internal information; the power of 35W can ensure the signal strength, so that the ultrasonic waves still have enough energy to reflect the echoes when penetrating the alloy steel; the probe angle of 45° is conducive to comprehensive detection, which can cover all parts of the alloy steel workpiece and avoid detection blind area; the intelligent data acquisition interface has automatic calibration function, before collecting data, the parameters of the equipment will be automatically calibrated to ensure the accuracy and stability of the collected data; at the same time, the intelligent data acquisition interface can also monitor the data quality in real time during the collection process, and once the abnormality is found, the adjustment and re-collection are carried out in time.

[0017] The feature information extraction unit 201 in the feature extraction module 200 extracts features from the full matrix data using a convolutional neural network-based feature extraction method to generate feature information, which specifically includes: Referring to Figure 3 For the collected full matrix data, a convolutional neural network model with multiple convolutional layers, pooling layers and fully connected layers is constructed by using a convolutional neural network-based feature extraction method; the convolutional layer automatically extracts key information such as internal structure features and composition distribution features by sliding the convolutional kernel on the full matrix data; the pooling layer reduces the dimension of the data output by the convolutional layer while retaining the main features; the fully connected layer integrates the features output by the pooling layer to form a feature vector used to generate the original traceability code as the feature information.

[0018] The encoding unit 202 in the feature extraction module 200 generates a unique original traceability code based on the feature information, i.e., the feature information, using a convolutional neural network-based encoder. During the training of the convolutional neural network-based encoder, a multi-class cross-entropy loss function is used to measure the difference between the model's predicted encoding and the true encoding. The function formula is as follows: wherein is the number of training samples (e.g., the number of alloy steel workpiece training samples); represents the number of different workpiece feature combination categories (e.g., the number of categories corresponding to different feature combinations of alloy steel workpieces); represents the true probability (using one-hot encoding) of the training sample belonging to the category , is the probability of the model predicting that the sample belongs to the category , and the network parameters are continuously adjusted and optimized through the backpropagation algorithm to improve the performance of the encoder, ensuring the accuracy and uniqueness of the generated original traceability code.

[0019] The traceability module 300 stores the internal original information, feature information, and original traceability code corresponding to them in the database, including: The internal original information, feature information, and original traceability code are stored in the database using a distributed database storage method, which has high reliability and scalability. The internal original information, feature information, and original traceability code of the workpiece are stored, and multiple levels of access permissions are set to ensure data security. The multiple levels of access permissions include administrators, engineers, and ordinary operators. Administrators have the highest level of access and can manage and modify data, including adding, deleting, and updating data. Engineers can view and analyze data for product quality research and production process improvement. Ordinary operators can only perform simple data queries, such as querying the original traceability code based on the workpiece number. This setup effectively protects data security and privacy, preventing data leakage and misuse. In addition, the distributed database has data backup and recovery functions, which periodically backup data. In the event of data loss or damage, data can be quickly recovered to ensure normal system operation.

[0020] The traceability module 300 extracts feature information from the current workpiece at any subsequent processing stage and matches it with the feature information stored in the database to determine the original traceability code corresponding to the current workpiece, including: In any subsequent processing link of the workpiece, when it is necessary to trace the workpiece, the ultrasonic phased array equipment is used to collect the internal original information of the current workpiece, and corresponding full matrix data is generated; in view of the surface treatment conditions such as polishing, polishing and paint spraying of the workpiece, the parameters of the ultrasonic phased array equipment are further optimized according to the characteristics of the surface treatment layer, such as appropriately increasing the transmission power and adjusting the frequency range, to ensure that the ultrasonic wave can penetrate the surface treatment layer and obtain accurate internal information of the workpiece; different surface treatment layers have different attenuation and reflection characteristics of ultrasonic waves, for example, for a paint layer with large thickness, the transmission power needs to be increased and the frequency needs to be adjusted to ensure that the ultrasonic wave can penetrate and obtain internal real information; Referring to Figures 4-5 , the feature information extraction unit 201 uses the feature extraction method based on the convolutional neural network to extract the feature information of the full matrix data of the current workpiece (processing link), and matches it with the feature information stored in the database (initial stage). The image detected by the analysis instrument is matched visually, the image background is observed and compared, a multi-feature fusion similarity calculation algorithm is used, the similarity of the workpiece in terms of internal structure, component distribution and other characteristics is considered comprehensively, and a reasonable similarity threshold is set. When the workpiece undergoes surface treatment, the multi-feature fusion similarity calculation algorithm can effectively exclude the interference of surface treatment on feature matching through dynamic adjustment of feature weights, for example, appropriately increasing the weight of the internal structure feature which is less affected by surface treatment; for the component distribution feature which may be affected by surface treatment, the surface treatment process and material characteristics are corrected, and then the similarity is calculated; when the comprehensive similarity of the current workpiece and the original state exceeds the threshold, it is determined that the matching is successful, and the non-invasive tracing of the workpiece is realized.

[0021] Referring to Figure 6 , the process starts with the adjustment of the ultrasonic phased array equipment parameters based on the workpiece material and shape, and the internal original information of the workpiece is collected by scanning to form full matrix data (initial information collection); then it enters the data processing and coding stage, and the convolutional neural network is used to extract feature information from the full matrix data to generate a unique original traceability code and store it; when it is necessary to trace the workpiece, the ultrasonic phased array equipment is used again to collect the full matrix data of the current workpiece, the new feature information is extracted by the convolutional neural network, and the multi-feature fusion similarity calculation is performed with the original features stored in the database, and finally the non-invasive traceability verification is completed through algorithm matching.

[0022] The traceability module 300 further judges the quality deviation of the workpiece in the machining process according to the similarity difference between the characteristic information of the current workpiece and the characteristic information in the database, and if the similarity difference exceeds a preset threshold, triggers a warning and marks an abnormal workpiece; through the multi-feature fusion algorithm, the similarity difference between the current workpiece characteristic information and the original characteristic information in the database is dynamically compared, a weight model is constructed based on core features such as internal structure stability and composition distribution uniformity, and the characteristic deviation caused by machining error is automatically identified, for example, a higher weight is assigned to the grain morphology feature which is easily affected by heat treatment, and a lower weight is assigned to the surface oxide layer interference feature; when the comprehensive similarity difference exceeds the preset threshold, a hierarchical warning mechanism is automatically triggered, the abnormal workpiece is marked as a test state and a deviation report is pushed to the quality management terminal, and the deviation type (such as internal crack exceeding or composition segregation anomaly) is recorded for process tracing, and the threshold is set according to the characteristic fluctuation range of the historical qualified workpiece and the industry standard dynamic optimization, to ensure the accuracy of the warning.

[0023] The traceability module 300 compares the characteristic information of the current workpiece with the characteristic information of the historical machining qualified workpiece, reversely adjusts the process parameters of the current machining equipment according to the comparison result, and performs multi-dimensional matching between the current workpiece characteristic information and the characteristic database of the historical machining qualified workpiece, and locates the machining defect source through feature correlation analysis. For example, when the internal porosity of the workpiece is detected to be abnormal, the temperature and pressure parameter characteristics of the qualified workpieces in the same batch in the smelting stage are automatically extracted, the heating time or cooling rate deviation of the current casting equipment is reversely deduced by using a decision tree model, and a parameter correction instruction is generated; the instruction is fed back to the machining equipment controller in real time through the industrial Internet of Things interface, and the core parameters such as the feed amount and laser welding power of the numerical control machine tool are dynamically adjusted, forming a “detection-analysis-control” closed loop, and at the same time, the multi-batch process optimization data is aggregated periodically, the self-learning model is trained to improve the prediction accuracy of parameter adjustment, and the adaptive optimization of the machining process is realized.

[0024] In addition, the present embodiment introduces the application scene of the present application in actual production, wherein the detection equipment in the data collection module 100 integrates a high-resolution ultrasonic phased array detection device, has an automatic parameter adjustment function, and can quickly adapt to the detection needs of different workpieces; the equipment is connected with the data processing system of the background through a high-speed encrypted network communication connection, to ensure the safety and stability of data transmission; the high-speed encrypted network adopts an advanced encryption protocol such as SSL / TLS protocol, to encrypt and decrypt the transmitted data in real time, to prevent data from being stolen or tampered with during transmission; at the same time, the network has the characteristics of high bandwidth and low delay, and can quickly transmit a large amount of detection data to ensure the efficiency of the detection process.

[0025] In the automobile engine cylinder block production workshop, the detection equipment plays a key role, including: The cylinder on the production line will pass through the detection equipment at different stages of processing. After rough machining, the detection equipment collects the internal information of the cylinder according to the preset parameters, and judges whether the cylinder has raw material defects or problems introduced in the processing process through the traceability system. If it is found that there is a small crack inside a certain cylinder, the traceability system quickly locates the cylinder at the casting link, and the local temperature of the mold causes internal defects. The original traceability code is generated and related information is associated. According to the traceability result, the technician adjusts the casting process parameters in time, such as optimizing the cooling system of the mold to ensure uniform mold temperature and avoid the generation of more defective products in subsequent processing, effectively improving product quality and production efficiency. At the same time, in product quality traceability and after-sales service, the traceability system can quickly and accurately locate the problem, saving the enterprise a lot of time and cost. For example, when the customer feedbacks that the engine cylinder has a fault, the detection equipment detects the cylinder, and the traceability system can quickly determine the production batch, processing link and possible problems of the cylinder. The enterprise can carry out targeted maintenance or replacement. For problems caused by raw material defects, the enterprise can negotiate with the supplier to solve the problem; for problems in the processing process, the enterprise can optimize the production process to avoid similar problems from occurring again. Through accurate traceability, the enterprise can improve customer satisfaction, maintain the market reputation of the enterprise, and enhance the market competitiveness of the enterprise; In the field of aerospace component manufacturing, the quality requirements for components are extremely high. This detection equipment and traceability system also plays an important role. For example, when manufacturing aircraft engine blades, through detection and traceability of the blades at different stages of processing, it is ensured that each blade meets strict quality standards, ensuring aviation safety.

[0026] Please refer to Figure 7 The intelligent non-invasive workpiece traceability system based on ultrasonic phased array technology realizes non-invasive detection and tracking through multi-module cooperation. The data collection module 100 adopts an ultrasonic phased array device with adjustable array element excitation parameters, intelligently collects full matrix data according to the material characteristics of the workpiece, and combines the automatic calibration function of the intelligent data acquisition interface to ensure data quality; the feature extraction module 200 includes a feature information extraction unit 201 and a coding unit 202. The feature information extraction unit 201 extracts structural features and composition distribution features from the full matrix data through a convolutional neural network (including convolutional layers, pooling layers and fully connected layers); the coding unit 202 uses an encoder trained by a multi-class cross-entropy loss function to convert the feature vector into a unique original traceability code; the traceability module 300 stores multi-dimensional data in a distributed database and sets hierarchical permissions. In subsequent processing links, it overcomes surface treatment interference by optimizing ultrasonic penetration parameters, dynamically matches feature similarity using a multi-feature fusion algorithm, and realizes workpiece tracking by combining threshold judgment. At the same time, through feature deviation analysis, quality warning is triggered and processing parameters are adjusted in reverse, forming a closed-loop traceability mechanism of detection-analysis-optimization.

[0027] The foregoing generally describes the basic principles, main features and advantages of the present application. It should be understood that the present application is not limited to the above-mentioned embodiments, and the above-mentioned embodiments and descriptions in the specification are only preferred examples of the present application and are not intended to limit the present application. Various changes and improvements can be made to the present application without departing from the spirit and scope of the present application, and such changes and improvements are all within the scope of the present application. The scope of protection of the present application is defined by the appended claims and their equivalents.

Claims

1. An intelligent non-invasive workpiece traceability system based on ultrasonic phased array technology, characterized in that, The system comprises a data collection module (100), a feature extraction module (200) and a traceability module (300), wherein: The data collection module (100) collects internal original information of each workpiece by using an ultrasonic phased array device to form full matrix data; The feature extraction module (200) extracts features from the full matrix data by using a feature extraction method based on a convolutional neural network to generate feature information; and generates a unique original traceability code for each workpiece based on the feature information; The traceability module (300) stores the internal original information, the feature information and the original traceability code in the database in correspondence; at any subsequent processing stage of the workpiece, the feature information of the current workpiece is extracted, and the feature information stored in the database is matched to determine the original traceability code corresponding to the current workpiece.

2. The intelligent non-intrusive workpiece traceability system based on ultrasonic phased array technology of claim 1, wherein, The data collection module (100) controls the excitation sequence and time delay of the probe elements of the ultrasonic phased array device to make the ultrasonic waves incident to the inside of the workpiece at different angles and directions, and receives the reflected echoes to obtain the internal original information of the workpiece.

3. The intelligent non-intrusive workpiece traceability system based on ultrasonic phased array technology of claim 1, wherein, The data collection module (100) automatically matches the preset parameters including the transmission frequency, power and probe angle according to the material and shape of the workpiece to collect the internal original information and generate the full matrix data.

4. The intelligent non-intrusive workpiece traceability system based on ultrasonic phased array technology of claim 1, wherein, The data collection module (100) comprises an intelligent data acquisition interface, which automatically calibrates the equipment parameters before data acquisition, and monitors the data quality in real time during the acquisition process, and adjusts and reacquires when an abnormality is found.

5. The intelligent non-intrusive workpiece traceability system based on ultrasonic phased array technology of claim 1, wherein, The feature extraction module (200) comprises a feature information extraction unit (201), which extracts internal structure features and composition distribution features from the full matrix data by using a convolutional neural network model with multiple convolutional layers, pooling layers and fully connected layers, and generates a feature vector as the feature information.

6. The intelligent non-intrusive workpiece traceability system based on ultrasonic phased array technology of claim 5, wherein, The feature extraction module (200) comprises an encoding unit (202), which generates a unique original traceability code based on the feature information by using a trained convolutional neural network encoder; the training of the encoder optimizes the network parameters by using a multi-class cross-entropy loss function and a back propagation algorithm, wherein the formula of the multi-class cross-entropy loss function is as follows: wherein is the number of training samples; represents the number of different workpiece feature combination classes; denotes the training samples belonging to class with true probability is the probability that the model predicts that a sample belongs to class .

7. The intelligent non-intrusive workpiece traceability system based on ultrasonic phased array technology of claim 1, wherein, The traceability module (300) stores the internal original information, the feature information and the original traceability code in a distributed database, and sets multiple levels of access permissions, including administrator permissions, engineering permissions and ordinary operator permissions.

8. The intelligent non-intrusive workpiece traceability system based on ultrasonic phased array technology of claim 1, wherein, The traceability module (300) optimizes the parameters of the ultrasonic phased array device according to the characteristics of the surface treatment layer at the subsequent processing stage of the workpiece, extracts the feature information of the current workpiece, and matches the feature information in the database by using a multi-feature fusion similarity calculation algorithm to determine the original traceability code.

9. The intelligent non-intrusive workpiece traceability system based on ultrasonic phased array technology of claim 8, wherein, The multi-feature fusion similarity calculation algorithm dynamically adjusts the weights of different features, excludes the interference of surface treatment on matching, and determines the matching result based on a comprehensive similarity threshold.

10. The intelligent non-intrusive workpiece traceability system based on ultrasonic phased array technology of claim 1, wherein, The traceability module (300) judges quality deviation according to a similarity difference value of current workpiece characteristic information and characteristic information in the database, triggers a pre-warning and marks an abnormal workpiece if the similarity difference value exceeds a preset threshold value, and simultaneously reversely adjusts process parameters of a machining device through characteristic correlation analysis to form a closed-loop optimization.

Images