Cable bridge weld joint intelligent detection system

Abstract

The invention belongs to the technical field of cable bridge detection, and particularly relates to a cable bridge weld joint intelligent detection system which comprises a multi-modal data fusion acquisition module, a weld joint feature intelligent extraction module, a defect type intelligent identification module, a weld joint threat decision module, an automatic sorting module and a visual feedback supervision end. According to the invention, the multi-modal data of the cable bridge welding seam is acquired through the multi-modal data fusion acquisition module, the key features of the cable bridge welding seam are extracted based on the multi-modal data, and the welding seam defect is intelligently identified based on the extracted welding seam feature vector. The welding line threat is judged based on the defect recognition result, the quality condition of the cable bridge is evaluated, the detected cable bridge is automatically sorted by controlling the manipulator, automatic and intelligent detection of the welding line of the cable bridge is achieved, the welding line of the cable bridge can be rapidly and accurately detected, the detection efficiency is high, the defect omission ratio is reduced, and the detection cost is reduced. And the production quality and the production efficiency of the cable bridge can be improved.

Description

Technical Field

[0001] This invention relates to the field of cable tray inspection technology, specifically an intelligent inspection system for cable tray welds. Background Technology

[0002] Cable trays are rigid structural systems composed of supports, brackets, and installation accessories. They are made of metal or non-metal materials and are mainly used to support, fix, and protect cable lines. They are widely used in industrial, commercial, and civil buildings to provide an orderly and safe laying channel for power, communication, and control cables. Quality inspection is an essential step in the production of cable trays. For example, Chinese invention patent CN114387272A discloses a method for detecting defective cable trays based on image processing. This invention uses processing and analysis to distinguish between the spot areas and defect areas in the gray-scale abrupt change areas of the image to be inspected, thereby detecting cable trays that are truly defective and achieving quality inspection of cable trays. However, in practical applications, the above-mentioned invention mainly focuses on the overall appearance inspection of cable tray quality. It cannot identify all the defects of each weld on the cable tray in detail and judge the risk of cable tray use. It also cannot automatically sort cable trays with quality problems and accurately evaluate the sorting performance, which is not conducive to realizing the automated and intelligent inspection of cable tray welds. To address the aforementioned technical shortcomings, a solution is proposed. Summary of the Invention

[0003] The purpose of this invention is to provide an intelligent inspection system for cable tray welds, which solves the problems of existing technologies that cannot accurately identify defects in all welds on cable trays and determine the risk of cable tray use, and cannot automatically sort cable trays with quality problems and accurately evaluate sorting performance, thus hindering the realization of automated and intelligent inspection of cable tray welds.

[0004] To achieve the above objectives, the present invention provides the following technical solution: A smart inspection system for cable tray welds includes a multimodal data fusion acquisition module, a weld feature intelligent extraction module, a defect intelligent identification module, a weld threat decision-making module, an automated sorting module, and a visual feedback monitoring terminal. The multimodal data fusion acquisition module acquires multimodal data of the cable tray welds and sends the acquired multimodal data to the weld feature intelligent extraction module. The weld feature intelligent extraction module extracts key features of the cable tray welds based on the received multimodal data and sends the weld feature vectors to the defect intelligent identification module. The intelligent defect identification module uses machine learning algorithms to intelligently identify weld defects based on extracted weld feature vectors, and sends the defect identification results to the weld threat decision module and the visual feedback monitoring terminal. The weld threat decision module judges the threat of welds and evaluates the quality status of cable trays based on the defect identification results, and sends the analysis results to the automated sorting module and the visual feedback monitoring terminal. Based on the analysis results of the weld threat decision module, the automated sorting module controls a robotic arm to sort the cable trays, so that the cable trays corresponding to the quality non-conforming signal and the quality excellent signal enter the corresponding output channels respectively.

[0005] Furthermore, the multimodal data fusion acquisition module uses a high-resolution industrial camera to perform a panoramic scan of the surface of the cable tray weld, acquiring two-dimensional image data of the weld. It also uses a laser 3D scanner to measure the three-dimensional shape of the weld, acquiring geometric parameters including weld height, width, and depth to form three-dimensional point cloud data of the weld. Additionally, it uses an ultrasonic flaw detector to emit ultrasonic waves into the weld and receive the reflected echo signals to obtain the internal structural information of the weld. Finally, the acquired two-dimensional images, three-dimensional point clouds, and ultrasonic echo signals are synchronized in time and aligned in space to achieve multimodal data fusion.

[0006] Furthermore, the intelligent weld feature extraction module uses a deep learning-based convolutional neural network to extract features from surface image data. Through multi-layer convolution and pooling operations, it automatically learns the texture features of the weld surface. For 3D point cloud data, point cloud processing algorithms are used to extract the geometric shape features of the weld, and principal component analysis is used to reduce the dimensionality of the point cloud data to extract representative shape features. For ultrasonic echo signals, time-frequency analysis is used to extract the time-domain and frequency-domain features of the signal, and short-time Fourier transform is used to convert the signal into a time-frequency spectrum to further extract the time-frequency features of the signal. The extracted surface texture features, geometric shape features and internal structure features are spliced ​​together to form a comprehensive weld feature vector.

[0007] Furthermore, the specific analysis process of the weld threat decision module includes: All welds on the cable tray are obtained, and the corresponding welds are marked as i, where i is a natural number greater than or equal to 1; the defect identification results of weld i are retrieved, all defects distributed on weld i are obtained, the detection data of various parameters corresponding to the corresponding defects are collected, and the detection data of various parameters are compared with the corresponding preset data requirements. If there are parameters whose detection data do not meet the corresponding preset data requirements, the threat judgment symbol ZY-1 is assigned to the corresponding defect. If there is a defect on weld i that is assigned the threat assessment symbol ZY-1, then weld i is marked as a high-threat weld; if there is no defect on weld i that is assigned the threat assessment symbol ZY-1, then weld i is divided into several inspection areas. If the corresponding inspection area involves a defect, then the corresponding inspection area is marked as an uninspected area. The number of abnormal inspection areas on weld i is collected and the ratio is calculated with the total number of areas to be inspected to obtain the weld abnormal inspection decision value. The weld abnormal inspection decision value is compared with the corresponding preset weld abnormal inspection decision threshold. If the weld abnormal inspection decision value exceeds the corresponding preset weld abnormal inspection decision threshold, weld i is marked as a high-threat weld; if the weld abnormal inspection decision value does not exceed the corresponding preset weld abnormal inspection decision threshold, weld i is marked as a low-threat weld.

[0008] Furthermore, the specific analysis process of the weld threat decision module also includes: The marking information of all welds on the cable tray is obtained. If there are high-threat welds on the cable tray, a corresponding cable tray quality failure signal is generated. If there are no high-threat welds on the cable tray, the weld risk analysis value is obtained by comparing the weld defect decision value of weld i with the corresponding preset weld defect decision threshold. Each group of welds is pre-set to correspond to a set of preset importance weight values. The weld risk analysis value of weld i is multiplied by the corresponding preset importance weight value to obtain the weld risk value. The weld risk values ​​of all welds on the cable tray are summed to obtain the cable tray quality assessment value. The cable tray quality assessment value is compared with the preset cable tray quality assessment threshold. If the cable tray quality assessment value exceeds the preset cable tray quality assessment threshold, a quality failure signal is generated for the corresponding cable tray; if the cable tray quality assessment value does not exceed the preset cable tray quality assessment threshold, a quality excellent signal is generated for the corresponding cable tray.

[0009] Furthermore, the automated sorting module is connected to the sorting monitoring and alarm module. The sorting monitoring and alarm module monitors the operation of the automated sorting module, analyzes the performance of the corresponding automated sorting process, and determines whether a sorting alarm signal is generated. When a sorting alarm signal is generated, it is sent to the visual feedback monitoring terminal. When the visual feedback monitoring terminal receives the sorting alarm signal, it issues a corresponding warning.

[0010] Furthermore, the specific analysis process of the sorting monitoring and alarm module includes: The monitoring information of the corresponding automatic sorting process is obtained, and the actual movement path of the robot is compared with the corresponding planned movement path. Based on this, the length ratio of the non-overlapping path is obtained and marked as the sorting path deviation measurement value. The number of times the robot deviates from the path during the corresponding automatic sorting process is obtained and marked as the sorting deviation frequency detection value. The sorting path deviation measurement value and sorting deviation frequency detection value are compared with the preset sorting path deviation measurement threshold and preset sorting deviation detection threshold, respectively. If the sorting path deviation measurement value or sorting deviation frequency detection value exceeds the corresponding preset threshold, a sorting alarm signal for the corresponding automatic sorting process is generated. If neither the sorting path deviation measurement value nor the sorting deviation frequency detection value exceeds the corresponding preset threshold, a sorting auxiliary judgment value is obtained through sorting auxiliary judgment analysis. The sorting auxiliary judgment value is compared with the preset sorting auxiliary judgment threshold. If the sorting auxiliary judgment value exceeds the preset sorting auxiliary judgment threshold, a sorting alarm signal for the corresponding automatic sorting process is generated.

[0011] Furthermore, the specific analysis process for sorting-assisted judgment analysis is as follows: The vibration amplitude and noise decibel value of the robotic arm in the automated sorting module are collected and marked as sorting vibration value and sorting noise value, respectively. The sorting vibration value and sorting noise value are compared with the preset sorting vibration threshold and preset sorting noise threshold, respectively. If the sorting vibration value or sorting noise value exceeds the corresponding preset threshold, it is determined that the robotic arm is in an abnormal sorting state. The percentage of time the robotic arm is in an abnormal sorting state during the corresponding automated sorting process is obtained and marked as the sorting time anomaly detection value. The average value of sorting vibration and the average value of sorting noise during the corresponding automated sorting process are marked as sorting vibration detection value and sorting noise detection value, respectively. The sorting auxiliary judgment value is obtained by weighted summation of the sorting time anomaly detection value, sorting vibration detection value and sorting noise detection value.

[0012] Furthermore, the sorting monitoring and alarm module is communicatively connected to the sorting performance detection and output module. The sorting monitoring and alarm module sends sorting alarm signals to the sorting performance detection and output module. The sorting performance detection and output module analyzes the sorting performance of the cable tray during the detection period, generates a sorting performance qualified signal or a sorting performance abnormal signal through analysis, and sends the sorting performance qualified signal or sorting performance abnormal signal to the visual feedback monitoring terminal. When the visual feedback monitoring terminal receives the sorting performance abnormal signal, it issues a corresponding warning.

[0013] Furthermore, the specific analysis process of the sorting performance detection output module is as follows: The number of sorting alarm signals generated during the detection period is obtained and marked as the total number of automated sorting processes. The ratio is calculated to obtain the sorting alarm statistics value. The sorting alarm statistics value is compared with the preset sorting alarm statistics threshold. If the sorting alarm statistics value exceeds the preset sorting alarm statistics threshold, a sorting performance abnormality signal is generated. If the sorting alarm statistics value does not exceed the preset sorting alarm statistics threshold, the moment when the automated sorting module receives the quality non-conforming signal or the quality excellent signal is marked as the first moment, and the moment when the automated sorting module starts sorting is marked as the second moment. The sorting response value is obtained by calculating the time difference between the first moment and the second moment. The sorting reaction value is compared with the preset sorting reaction threshold. If the sorting reaction value exceeds the preset sorting reaction threshold, the corresponding sorting reaction value is marked as a sorting defect value. The ratio of the number of sorting defect values ​​to the number of sorting reaction values ​​in the detection period is calculated to obtain the sorting anomaly value. The average of all sorting reaction values ​​in the detection period is calculated to obtain the sorting effectiveness value. The sorting performance output value is calculated by weighting and summing the sorting alarm statistics, sorting anomaly test values, and sorting effectiveness test values. The sorting performance output value is then compared with a preset sorting performance output threshold. If the sorting performance output value exceeds the preset sorting performance output threshold, a sorting performance abnormality signal is generated; if the sorting performance output value does not exceed the preset sorting performance output threshold, a sorting performance qualified signal is generated.

[0014] Compared with the prior art, the beneficial effects of the present invention are: 1. In this invention, multimodal data of cable tray welds are collected, key features of cable tray welds are extracted based on the multimodal data, weld defects are intelligently identified based on the extracted weld feature vectors, the threat of welds is judged based on the defect identification results, and the quality status of cable trays is evaluated. The inspected cable trays are automatically sorted, realizing automated and intelligent inspection of cable tray welds. The inspection efficiency is high and the defect omission rate is reduced, which is conducive to improving the production quality and efficiency of cable trays. 2. In this invention, the sorting monitoring and alarm module monitors the operation of the automated sorting module, analyzes the performance of the corresponding automated sorting process, continuously monitors the subsequent sorting status when a sorting alarm signal is generated, and performs timely checks and adjustments as needed. In addition, the sorting performance detection and output module analyzes the sorting performance of the cable trays during the detection period, and inspects and repairs the automated sorting module when an abnormal sorting performance signal is generated, so as to ensure the high efficiency and stability of automated sorting of cable trays. Attached Figure Description

[0015] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings; Figure 1 This is a system block diagram of Embodiment 1 of the present invention; Figure 2 This is a system block diagram of Embodiments 2 and 3 of the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] Example 1: As Figure 1 As shown, the present invention proposes an intelligent inspection system for cable tray welds, which includes a multimodal data fusion acquisition module, an intelligent weld feature extraction module, an intelligent defect identification module, a weld threat decision-making module, an automated sorting module, and a visual feedback monitoring terminal. The multimodal data fusion acquisition module collects multimodal data of cable tray welds, including surface images, three-dimensional morphology and ultrasonic echo signals of the welds, and sends the collected multimodal data to the weld feature intelligent extraction module. Through multimodal data fusion acquisition, the surface and internal information of the welds can be comprehensively obtained, avoiding the limitations of a single detection method and improving the accuracy and reliability of defect detection. Specifically, firstly, the multimodal data fusion acquisition module uses a high-resolution industrial camera to perform a panoramic scan of the surface of the cable tray weld, acquiring two-dimensional image data of the weld. The image resolution reaches the micrometer level, which can clearly show the tiny defects on the weld surface. At the same time, a laser 3D scanner is used to measure the three-dimensional shape of the weld, acquiring geometric parameters such as the height, width, and depth of the weld, forming three-dimensional point cloud data of the weld. In addition, an ultrasonic flaw detector emits ultrasonic waves into the weld and receives the reflected echo signals, thereby acquiring the structural information inside the weld, such as the reflection characteristics of defects such as pores, slag inclusions, and cracks. Finally, the acquired two-dimensional images, three-dimensional point clouds, and ultrasonic echo signals are synchronized in time and aligned in space to achieve multimodal data fusion.

[0018] The intelligent weld feature extraction module extracts key features of cable tray welds based on received multimodal data, including surface texture features, geometric features, and internal structural features. It then sends the weld feature vector to the intelligent defect identification module to provide effective feature vectors for defect identification. Through intelligent feature extraction, the module can automatically learn the key features of the weld, reducing the complexity and subjectivity of manual feature design and improving the effectiveness and discriminability of the features. It should be noted that the intelligent weld feature extraction module uses a deep learning-based convolutional neural network (CNN) to extract features from surface image data. Through multi-layer convolution and pooling operations, it automatically learns the texture features of the weld surface, such as roughness, uniformity, and directionality. For 3D point cloud data, point cloud processing algorithms are used to extract the geometric features of the weld, such as straightness, roundness, and parallelism. Principal component analysis (PCA) is then used to reduce the dimensionality of the point cloud data and extract the most representative shape features. For ultrasonic echo signals, time-frequency analysis methods are used to extract the time-domain and frequency-domain features of the signal, such as echo amplitude, frequency, and phase. The signal is then converted into a time-frequency spectrum using short-time Fourier transform (STFT) to further extract the time-frequency features of the signal. The extracted surface texture features, geometric features, and internal structure features are then stitched together to form a comprehensive weld feature vector.

[0019] The intelligent defect identification module uses machine learning algorithms to intelligently identify weld defects based on the extracted weld feature vectors, including common defects such as porosity, slag inclusions, cracks, and lack of fusion. The defect identification results are sent to the weld threat decision module and the visual feedback monitoring terminal. By using machine learning algorithms to achieve intelligent identification of defect types, the subjectivity and uncertainty of human judgment can be avoided, thereby improving the accuracy and consistency of defect identification.

[0020] The weld threat decision-making module assesses the threat level of welds and evaluates the quality of cable trays based on defect identification results. It then sends the analysis results to the automated sorting module and the visual feedback monitoring terminal. This not only allows for a reasonable assessment of the risk status of each weld but also accurately evaluates the quality of the cable trays, preventing the outflow of substandard products. Furthermore, it facilitates targeted improvement and remedial measures for the cable trays, demonstrating a high level of intelligence. The specific analysis process of the weld threat decision-making module includes: All welds on the cable tray are obtained, and the corresponding welds are marked as i, where i is a natural number greater than or equal to 1; the defect identification results of weld i are retrieved, all defects distributed on weld i are obtained, and the detection data of various parameters corresponding to the corresponding defects (such as the length, depth, width, etc. of crack defects) are collected. The detection data of each parameter are compared with the corresponding preset data requirements. If there are parameters whose detection data do not meet the corresponding preset data requirements, it indicates that the safety risk posed by the corresponding defect is relatively large, and the threat judgment symbol ZY-1 is assigned to the corresponding defect. If there is a defect on weld i that is assigned the threat assessment symbol ZY-1, it indicates that weld i poses a high safety threat, and weld i is marked as a high-threat weld; if there is no defect on weld i that is assigned the threat assessment symbol ZY-1, weld i is divided into several inspection areas. If the corresponding inspection area involves a defect, the corresponding inspection area is marked as an uninspected area. The number of non-inspection areas on weld i is collected and the ratio is calculated to the total number of areas to be inspected to obtain the weld non-inspection decision value. The weld non-inspection decision value is compared with the corresponding preset weld non-inspection decision threshold. If the weld non-inspection decision value exceeds the corresponding preset weld non-inspection decision threshold, it indicates that the safety threat posed by weld i is high, and weld i is marked as a high-threat weld. If the weld non-inspection decision value does not exceed the corresponding preset weld non-inspection decision threshold, it indicates that the safety threat posed by weld i is low, and weld i is marked as a low-threat weld.

[0021] Furthermore, the specific analysis process of the weld threat decision module also includes: obtaining the marking information of all welds on the cable tray; if there are high-threat welds on the cable tray, it indicates that the quality of the corresponding cable tray is poor, and a quality non-compliance signal for the corresponding cable tray is generated; if there are no high-threat welds on the cable tray, the weld defect decision value of weld i is calculated by the ratio of the weld defect decision value to the corresponding preset weld defect decision threshold to obtain the weld risk analysis value. Each group of welds is pre-set to have a set of preset importance weight values ​​that are greater than zero. It should be noted that the higher the importance of the position corresponding to the weld, the larger the value of the preset importance weight value. The weld risk analysis value of weld i is multiplied by the corresponding preset importance weight value to obtain the weld risk value. The weld risk values ​​of all welds on the cable tray are summed to obtain the cable tray quality assessment value. The cable tray quality assessment value is compared with the preset cable tray quality assessment threshold. If the cable tray quality assessment value exceeds the preset threshold, it indicates that the overall quality risk of the corresponding cable tray is relatively high, and a quality failure signal is generated for the corresponding cable tray. If the cable tray quality assessment value does not exceed the preset threshold, it indicates that the overall quality risk of the corresponding cable tray is relatively low, and a quality excellent signal is generated for the corresponding cable tray.

[0022] Based on the analysis results of the weld threat decision module, the automated sorting module controls a robotic arm to sort cable trays. Cable trays corresponding to quality failure signals and quality excellence signals are sent to their respective output channels. This means that cable trays with quality failure signals and quality excellence signals are automatically sent to their respective output channels and transported out through different channels, eliminating the need for manual sorting and improving automation and sorting efficiency. Furthermore, the automated sorting module sends sorting execution information to a visual feedback monitoring terminal to facilitate management personnel's understanding of the sorting execution status.

[0023] Example 2: Figure 2As shown, the difference between this embodiment and Embodiment 1 is that the automated sorting module is connected to the sorting monitoring and alarm module. The sorting monitoring and alarm module monitors the operation of the automated sorting module, analyzes the operation performance of the corresponding automated sorting process, and determines whether to generate a sorting alarm signal through analysis. When a sorting alarm signal is generated, it is sent to the visual feedback monitoring terminal. When the visual feedback monitoring terminal receives a sorting alarm signal, it issues a corresponding warning to remind managers to continuously monitor the subsequent sorting status and make timely checks and adjustments as needed to ensure the efficient and stable operation of the automated sorting module. The specific analysis process of the sorting monitoring alarm module is as follows: The monitoring information of the corresponding automatic sorting process is obtained, and the actual movement path of the robot is compared with the corresponding planned movement path. Based on this, the length ratio of the non-overlapping path is obtained and marked as the sorting path deviation measurement value. The number of times the robot deviates from the path during the corresponding automatic sorting process is obtained and marked as the sorting deviation frequency detection value. The sorting path deviation measurement value and sorting deviation frequency detection value are compared with the preset sorting path deviation measurement threshold and preset sorting deviation detection threshold respectively. If the sorting path deviation measurement value or sorting deviation frequency detection value exceeds the corresponding preset threshold, it indicates that the path execution status of the corresponding automatic sorting process is poor, and a sorting alarm signal for the corresponding automatic sorting process is generated. If the sorting deviation measurement value and the sorting deviation frequency detection value do not exceed the corresponding preset threshold, the vibration amplitude and noise decibel value of the robot in the automated sorting module are collected and marked as sorting vibration value and sorting noise value, respectively. The sorting vibration value and sorting noise value are compared with the preset sorting vibration threshold and preset sorting noise threshold, respectively. If the sorting vibration value or sorting noise value exceeds the corresponding preset threshold, it indicates that the real-time operating status of the robot is not good, and the robot is judged to be in an abnormal sorting state. The percentage of time the robot arm is in an abnormal sorting state during the corresponding automated sorting process is obtained and marked as the sorting time difference detection value. The average value of sorting vibration and the average value of sorting noise during the corresponding automated sorting process are marked as sorting vibration detection value and sorting noise detection value, respectively. The sorting auxiliary judgment value is obtained by weighted summation of the sorting time difference detection value, sorting vibration condition detection value, and sorting noise condition detection value. Specifically, the sorting time difference detection value, sorting vibration condition detection value, and sorting noise condition detection value are each assigned a corresponding preset weight coefficient, and the sorting time difference detection value, sorting vibration condition detection value, and sorting noise condition detection value are multiplied by the corresponding preset weight coefficient. The sum of the three sets of product results is marked as the sorting auxiliary judgment value. It should be noted that the larger the value of the sorting assistance judgment value, the worse the overall operation of the corresponding automatic sorting process. The sorting assistance judgment value is compared with the preset sorting assistance judgment threshold. If the sorting assistance judgment value exceeds the preset sorting assistance judgment threshold, it indicates that the overall operation of the corresponding automatic sorting process is poor, and a sorting alarm signal for the corresponding automatic sorting process is generated.

[0024] Example 3: Figure 2 As shown, the difference between this embodiment and Embodiment 1 and Embodiment 2 is that the sorting monitoring alarm module is communicatively connected to the sorting performance detection output module. The sorting monitoring alarm module sends the sorting alarm signal to the sorting performance detection output module. The sorting performance detection output module analyzes the sorting performance of the cable tray during the detection period and generates a sorting performance qualified signal or a sorting performance abnormal signal through analysis. Furthermore, the system sends either a pass / fail signal or an abnormal sorting signal to the visual feedback monitoring terminal. Upon receiving an abnormal sorting signal, the visual feedback monitoring terminal issues a corresponding warning to remind management personnel to inspect and maintain the automated sorting module, further ensuring the high efficiency and stability of the automated cable tray sorting process. The specific analysis process of the sorting performance detection output module is as follows: The number of sorting alarm signals generated during the detection period is obtained and marked as the total number of automated sorting processes. The ratio is calculated to obtain the sorting alarm statistics value. The sorting alarm statistics value is compared with the preset sorting alarm statistics threshold. If the sorting alarm statistics value exceeds the preset sorting alarm statistics threshold, it indicates that the sorting performance of the automated sorting module during the detection period is poor, and an abnormal sorting performance signal is generated. If the sorting alarm statistics value does not exceed the preset sorting alarm statistics threshold, the moment when the automated sorting module receives the quality failure signal or the quality excellence signal is marked as the first moment, and the moment when the automated sorting module starts sorting is marked as the second moment. The time difference between the first moment and the second moment is calculated to obtain the sorting response value. The larger the value of the sorting response value, the slower the response to the sorting action of the corresponding cable tray. The sorting response value is compared with the preset sorting response threshold. If the sorting response value exceeds the preset sorting response threshold, it indicates that the sorting action for the corresponding cable tray is slow. The corresponding sorting response value is then marked as a sorting defect value. The ratio of the number of sorting defect values ​​to the number of sorting response values ​​during the detection period is calculated to obtain the sorting anomaly value. The average of all sorting response values ​​during the detection period is calculated to obtain the sorting effectiveness value. The sorting performance output value is obtained by weighted summation of sorting alarm statistics, sorting anomaly test values, and sorting effectiveness test values. Specifically, each sorting alarm statistics, sorting anomaly test value, and sorting effectiveness test value is assigned a corresponding preset weight coefficient, and each of these values ​​is multiplied by its corresponding preset weight coefficient. The sum of the three products is then marked as the sorting performance output value. It should be noted that the larger the sorting performance output value, the worse the overall sorting performance of the automated sorting module during the detection period. The sorting performance output value is compared with the preset sorting performance output threshold. If the sorting performance output value exceeds the preset sorting performance output threshold, it indicates that the overall sorting performance of the automated sorting module during the detection period is poor, and an abnormal sorting performance signal is generated. If the sorting performance output value does not exceed the preset sorting performance output threshold, it indicates that the overall sorting performance of the automated sorting module during the detection period is good, and a qualified sorting performance signal is generated.

[0025] The working principle of this invention is as follows: In use, a multimodal data fusion acquisition module collects multimodal data of cable tray welds. A weld feature intelligent extraction module extracts key features of the cable tray welds based on the received multimodal data. A defect intelligent identification module intelligently identifies weld defects based on the extracted weld feature vectors and uses machine learning algorithms. A weld threat decision module judges the threat of the welds based on the defect identification results and assesses the quality status of the cable trays. Based on the cable tray quality analysis results, a robotic arm is controlled to sort the cable trays, allowing them to enter the corresponding output channels. By integrating multimodal data acquisition, defect intelligent identification, threat assessment, and automated sorting functions, automated and intelligent detection of cable tray welds is achieved. This enables rapid and accurate detection of cable tray welds, resulting in high detection efficiency and a reduced defect miss rate, which is beneficial for improving the production quality and efficiency of cable trays.

[0026] In this invention, the threshold, preset value, or preset range settings are for result comparison and analysis to determine whether the result is good or bad. The magnitude of these values ​​is determined by a combination of large-scale model analysis of sample data and human experience, and can also be appropriately adjusted based on seasonal or common-sense influence conditions. Similarly, the preset weight coefficients and influence factors are assigned specific values ​​based on the magnitude of each parameter's influence on the result, ultimately reflecting the impact on the result. These settings are also determined by a combination of large-scale model analysis of sample data and human experience, and can also be appropriately adjusted based on seasonal or common-sense influence conditions.

[0027] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, enabling those skilled in the art to better understand and utilize it. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. An intelligent inspection system for cable tray welds, characterized in that, It includes a multimodal data fusion acquisition module, a weld feature intelligent extraction module, a defect intelligent identification module, a weld threat decision module, an automated sorting module, and a visual feedback monitoring terminal; the multimodal data fusion acquisition module collects multimodal data of cable tray welds, and the weld feature intelligent extraction module extracts key features of cable tray welds based on the received multimodal data; The defect intelligent identification module uses machine learning algorithms to intelligently identify weld defects based on the extracted weld feature vectors. The weld threat decision module judges the weld threat based on the defect identification results and evaluates the quality status of the cable tray. The automated sorting module, based on the analysis results of the weld threat decision module, controls a robotic arm to sort cable trays.

2. The intelligent inspection system for cable tray welds according to claim 1, characterized in that, The multimodal data fusion acquisition module uses a high-resolution industrial camera to perform panoramic scanning of the surface of the cable tray weld, and uses a laser 3D scanner to measure the 3D morphology of the weld, and uses an ultrasonic flaw detector to emit ultrasonic waves into the weld and receive the reflected echo signals.

3. The intelligent inspection system for cable tray welds according to claim 1, characterized in that, The intelligent weld feature extraction module uses a deep learning-based convolutional neural network to extract features from surface image data; for 3D point cloud data, it uses point cloud processing algorithms to extract the geometric shape features of the weld and performs dimensionality reduction processing on the point cloud data through principal component analysis. For ultrasonic echo signals, time-frequency analysis is used to extract the time-domain and frequency-domain features of the signal, and the signal is converted into a time-spectrum map through short-time Fourier transform. The extracted surface texture features, geometric features and internal structure features are spliced ​​together to form a comprehensive weld feature vector.

4. The intelligent inspection system for cable tray welds according to claim 1, characterized in that, The specific analysis process of the weld threat decision module includes: All welds on the cable tray are acquired, and each weld is marked as i, where i is a natural number greater than or equal to 1. All defects distributed on weld i are acquired. If a defect with the threat judgment symbol ZY-1 exists on weld i, weld i is marked as a high-threat weld. If no defect with the threat judgment symbol ZY-1 exists on weld i, the ratio of the number of non-inspection areas on weld i to the total number of areas to be inspected is calculated to obtain the weld non-inspection decision value. If the weld non-inspection decision value exceeds the corresponding preset weld non-inspection decision threshold, weld i is marked as a high-threat weld; otherwise, weld i is marked as a low-threat weld.

5. The intelligent inspection system for cable tray welds according to claim 4, characterized in that, The specific analysis process of the weld threat decision module also includes: if there are high-threat welds on the cable tray, a corresponding cable tray quality failure signal is generated; if there are no high-threat welds on the cable tray, the weld hazard values ​​of all welds on the cable tray are summed to obtain the cable tray quality assessment value; if the cable tray quality assessment value exceeds the preset cable tray quality assessment threshold, a corresponding cable tray quality failure signal is generated; otherwise, a corresponding cable tray quality excellent signal is generated.

6. The intelligent inspection system for cable tray welds according to claim 1, characterized in that, The automated sorting module is connected to the sorting monitoring and alarm module. The sorting monitoring and alarm module monitors the operation of the automated sorting module, analyzes the performance of the corresponding automated sorting process, and sends the sorting alarm signal to the visual feedback monitoring terminal when it is generated.

7. The intelligent inspection system for cable tray welds according to claim 6, characterized in that, The specific analysis process of the sorting monitoring and alarm module includes: If the monitoring information of the corresponding automatic sorting process is obtained, and the sorting path deviation measurement value or the sorting frequency deviation detection value exceeds the corresponding preset threshold, a sorting alarm signal for the corresponding automatic sorting process is generated; if neither the sorting path deviation measurement value nor the sorting frequency deviation detection value exceeds the corresponding preset threshold, a sorting auxiliary judgment value is obtained through sorting auxiliary judgment analysis; if the sorting auxiliary judgment value exceeds the preset sorting auxiliary judgment threshold, a sorting alarm signal for the corresponding automatic sorting process is generated.

8. The intelligent inspection system for cable tray welds according to claim 7, characterized in that, The specific analysis process of sorting auxiliary judgment analysis is as follows: obtain the percentage of time the robot is in an abnormal sorting state during the corresponding automated sorting process and mark it as the sorting time difference detection value. The sorting auxiliary judgment value is obtained by weighted summation of the sorting time difference detection value, sorting vibration detection value and sorting noise detection value.

9. The intelligent inspection system for cable tray welds according to claim 6, characterized in that, The sorting monitoring and alarm module communicates with the sorting performance detection and output module. The sorting monitoring and alarm module sends sorting alarm signals to the sorting performance detection and output module. The sorting performance detection and output module analyzes the sorting performance of the cable tray during the detection period and sends either a qualified sorting performance signal or an abnormal sorting performance signal to the visual feedback monitoring terminal.

10. The intelligent inspection system for cable tray welds according to claim 9, characterized in that, The specific analysis process of the sorting performance detection output module is as follows: If the sorting alarm statistics value exceeds the preset sorting alarm statistics threshold, a sorting performance abnormality signal is generated; if the sorting alarm statistics value does not exceed the preset sorting alarm statistics threshold, the sorting performance output value is obtained by weighted summation of the sorting alarm statistics value, the sorting anomaly measurement value, and the sorting effectiveness measurement value. If the sorting performance output value exceeds the preset sorting performance output threshold, a sorting performance abnormality signal is generated; otherwise, a sorting performance qualified signal is generated.

Images