Industrial production defect assessment method and system fused with deep learning

Abstract

The invention relates to the technical field of industrial production defect detection, in particular to an industrial production defect evaluation method and system fused with deep learning. The method comprises the steps of receiving a defect evaluation instruction, determining a target workpiece, fixing the workpiece, starting a deep learning analysis device, analyzing the instruction to obtain material characteristics and defect types, calculating illumination intensity and acquisition frequency, and performing real-time monitoring and completing evaluation. Through intelligent parameter calculation and real-time monitoring, the precision, efficiency and adaptability of defect evaluation are remarkably improved, the efficient and accurate requirements of modern industry are met, meanwhile, personal errors and detection cost are reduced, and the reliability and automation level of the evaluation process are improved.

Description

Technical Field

[0001] This invention belongs to the field of industrial production defect detection technology, specifically a method and system for industrial production defect assessment that integrates deep learning. Background Technology

[0002] With the rapid development of automation and intelligence in industrial production, deep learning-based defect detection and evaluation methods have played a significant role in improving product quality and reducing labor costs. Modern industry demands higher efficiency and accuracy in defect evaluation systems; however, existing technologies still have significant shortcomings in terms of detection accuracy, adaptability, real-time performance, and system integration. For example, patent CN111080622B proposes a method and apparatus for classifying and detecting workpiece surface defects based on few-sample deep learning. This method achieves defect classification and quantification by quickly locating the pixel coordinate regions of defects and extracting irregular defect regions. However, this technical solution relies heavily on few-sample learning, and its detection accuracy and generalization ability may be limited when facing complex, diverse, or low-contrast defects. Furthermore, this method does not fully consider the dynamic correlation between defect evaluation results and production processes, making it difficult to directly guide process optimization and quality control. In addition, patent CN115937147B proposes a method and apparatus for determining defect detection parameters. This method generates multiple fused images by fusing defect images and their mask images, and compares these images with reference images to determine the target defect image and its optical detection parameters. While this method can effectively improve image quality and meet practical industrial needs, its focus is primarily on image processing and parameter optimization, lacking multi-dimensional analysis and comprehensive evaluation of defect assessment results. Furthermore, this method has high requirements for hardware and imaging conditions, potentially leading to a significant increase in system deployment costs and complexity.

[0003] The aforementioned issues indicate that existing industrial production defect assessment methods still have certain shortcomings in terms of detection accuracy, multi-dimensional assessment capabilities, system adaptability, and integration with production processes. Especially when facing diverse defect types and dynamically changing production environments in complex industrial scenarios, current technologies struggle to fully meet the demands of modern industry for efficient and intelligent defect assessment. Therefore, an innovative technical solution is urgently needed that integrates deep learning models with multi-source data processing techniques to achieve high-precision defect detection, multi-dimensional assessment, and dynamic feedback. This would optimize production processes, improve product quality, and reduce production costs, ultimately meeting the pressing need of modern industry for efficient and intelligent defect assessment systems. Summary of the Invention

[0004] This invention provides a method and system for industrial production defect assessment that integrates deep learning, with the main objective of improving the accuracy, efficiency, and adaptability of defect detection and assessment in industrial production. To achieve the above objective, this invention provides a method for industrial production defect assessment that integrates deep learning, comprising: receiving an industrial production defect assessment instruction; determining the target workpiece to be assessed based on the defect assessment instruction, wherein the target workpiece is a metal component with a complex surface structure; fixing the target workpiece in a pre-constructed multi-dimensional scanning device, wherein the multi-dimensional scanning device is a component of a pre-constructed defect detection system, the defect detection system further comprising a light source adjustment module, an image acquisition module, and a data processing unit, wherein the multi-dimensional scanning device is connected to the light source adjustment module, the light source adjustment module is used to dynamically adjust the light intensity, and the image acquisition module is connected to the multi-dimensional scanning device; when the target workpiece is successfully fixed in the multi-dimensional scanning device, activating a pre-constructed deep learning analysis device, wherein the deep learning analysis device includes a feature extraction unit, a defect classification unit, and a result feedback unit, the feature extraction unit is connected to the image acquisition module, and the result processing unit is activated. The defect classification unit is connected to the feature extraction unit; the defect assessment command is transmitted to the result feedback unit, which parses the defect assessment command to obtain the surface material characteristics, defect type, and assessment time of the target workpiece, wherein the surface material characteristics and defect type are response features; the illumination intensity value of the light source adjustment module and the acquisition frequency of the image acquisition module are calculated based on the surface material characteristics and defect type; the light source adjustment module and the image acquisition module are started based on the pre-acquired actual assessment strategy, the illumination intensity value, and the acquisition frequency; the actual illumination intensity of the light source adjustment module, the actual acquisition frequency of the image acquisition module, and the actual detection time of the light source adjustment module and the image acquisition module are monitored; when the actual illumination intensity is equal to the illumination intensity value, the actual acquisition frequency is equal to the acquisition frequency, and the actual time is the same as the assessment time, the operation of the light source adjustment module and the image acquisition module is stopped, and the defect assessment of the target workpiece is completed.

[0005] Optionally, the step of transmitting the defect assessment command to the result feedback unit and using the result feedback unit to parse the defect assessment command to obtain the surface material characteristics, defect type, and assessment time of the target workpiece includes: constructing multiple signal transmission channels, wherein the multiple signal transmission channels are independent of each other, and each signal transmission channel can independently transmit the same defect assessment command; performing aggregation on the multiple signal transmission channels to obtain a signal transmission channel set, and performing target identification on each signal transmission channel in the signal transmission channel set to obtain an identification transmission channel set, wherein each identification transmission channel in the identification transmission channel set has a unique identifier; calculating the transmission quality index of each identification transmission channel in the identification transmission channel set, and aggregating the transmission quality indices of each identification transmission channel to obtain a transmission quality index set. The transmission quality index set is judged based on a preset quality judgment function to obtain a set of high-quality transmission channels. A transmission channel is randomly selected from the set of high-quality transmission channels, and signal transmission is performed using the selected transmission channel while recording the time to obtain the signal transmission time. It is then determined whether the signal transmission time is greater than a preset time threshold. If the signal transmission time is greater than the time threshold, the selected transmission channel is removed from the set of high-quality transmission channels to obtain an updated transmission channel set. The high-quality transmission channel set is updated using the updated transmission channel set, and the process returns to the step of randomly selecting a transmission channel from the set of high-quality transmission channels. If the signal transmission time is not greater than the time threshold, the defect evaluation instruction is parsed using the result feedback unit to obtain the surface material characteristics, defect type, and evaluation time of the target workpiece.

[0006] Optionally, the method for calculating the transmission quality index is as follows: randomly select an identification transmission channel from the identification transmission channel set, perform signal transmission based on the identification transmission channel, capture all signal waveforms during the signal transmission process using a preset signal monitoring tool, wherein the signal waveforms include complete waveforms and distorted waveforms; calculate the signal distortion rate using the complete waveforms and distorted waveforms; set a signal transmission interval and a signal transmission period, perform periodic signal transmission based on the signal transmission interval and signal transmission period, and obtain a periodic signal dataset; perform delay recording on the periodic signal dataset to obtain a signal delay set, perform mean calculation on the signal delay set to obtain the signal delay mean; obtain the maximum signal delay, bandwidth utilization, distortion weight coefficient, delay weight coefficient, and bandwidth weight coefficient, and calculate the transmission quality index based on the maximum signal delay, bandwidth utilization, distortion weight coefficient, delay weight coefficient, bandwidth weight coefficient, signal distortion rate, and signal delay mean.

[0007] Optionally, the calculation of the transmission quality index based on the maximum signal delay, bandwidth utilization, distortion weighting coefficient, delay weighting coefficient, bandwidth weighting coefficient, signal distortion rate, and mean signal delay includes:

[0008]

[0009] Where Q represents the transmission quality index, w1, w2, and w3 represent the distortion weighting coefficient, bandwidth weighting coefficient, and delay weighting coefficient, respectively, U represents the bandwidth utilization rate, and L... m L represents the maximum signal delay. a This represents the average signal delay.

[0010] Optionally, the step of determining the transmission quality index set based on a preset quality determination function to obtain a set of high-quality transmission channels includes: calculating the index mean and index standard deviation of the transmission quality index set; obtaining an adjustment factor; and constructing a quality determination function based on the adjustment factor, the index mean, and the index standard deviation.

[0011]

[0012] Among them, F(Q) i ) represents the output of the quality judgment function, Q i Let μ represent a single transmission quality index, σ represent the index mean, σ represent the index standard deviation, and k represent an adjustment factor. Transmission quality indices are extracted sequentially from the set of transmission quality indices, and the extracted indices are substituted into the quality determination function to obtain a determination output. When the determination output is 0, the transmission quality index is identified as a poor transmission index, and it is removed from the set of transmission quality indices to obtain an updated index set. The updated index set is used to update the transmission quality index set, and the process of extracting transmission quality indices sequentially from the set of transmission quality indices is returned. When the determination output is 1, the transmission quality index is identified as a high-quality transmission index, and the process of extracting transmission quality indices sequentially from the set of transmission quality indices is returned. When all transmission quality indices in the set have been extracted and substituted into the quality determination function, the transmission quality index set is identified as a high-quality transmission channel set.

[0013] Optionally, the method for calculating the light intensity value is as follows: obtain the surface reflectivity of the target workpiece, and calculate the surface light absorption rate based on the surface reflectivity and surface material properties.

[0014] α=1-ρ·β

[0015] Where α represents the surface light absorptivity, ρ represents the surface material properties, and γ represents the surface reflectivity; the light source power, light source angle, ambient light interference, and light source attenuation coefficient of the light source adjustment module are obtained, and the light intensity value is calculated based on the light source power, light source angle, ambient light interference, and light source attenuation coefficient.

[0016] Optionally, calculating the illuminance value based on the light source power, light source angle, ambient light interference, and light source attenuation coefficient includes: calculating the illuminance value using the following formula:

[0017] I = P·cos(θ) - E·λ

[0018] Where I represents the illuminance value, P represents the power of the light source, θ represents the angle of the light source, E represents the ambient light interference, and λ represents the light source attenuation coefficient.

[0019] Optionally, the method for calculating the acquisition frequency is as follows: obtain the acquisition resolution of the image acquisition module, obtain the surface roughness value and optical transmittance of the target workpiece, obtain the overall response time of the image acquisition module and the multi-dimensional scanning device, and calculate the acquisition frequency based on the overall response time, acquisition resolution, surface roughness value, optical transmittance, defect type, and surface material characteristics.

[0020]

[0021] Where f represents the acquisition frequency, R represents the acquisition resolution, T represents the overall response time, τ represents the optical transmittance, σ represents the surface roughness value, and δ represents the defect type.

[0022] Optionally, before activating the light source adjustment module and image acquisition module based on the pre-acquired actual evaluation strategy, the light intensity value, and the acquisition frequency, the method further includes: after completing the calculation of the light intensity value of the light source adjustment module and the acquisition frequency of the image acquisition module, generating an evaluation strategy and transmitting the evaluation strategy back to the initiator of the defect evaluation command, wherein the evaluation strategy includes: Strategy A: first evaluate the surface material characteristics of the target workpiece, then evaluate the defect type of the target workpiece; Strategy B: first evaluate the defect type of the target workpiece, then evaluate the surface material characteristics of the target workpiece; Strategy C: simultaneously evaluate the surface material characteristics and defect type of the target workpiece; receiving the evaluation strategy selected by the initiator of the defect evaluation command to obtain the actual evaluation strategy.

[0023] To achieve the above objectives, the present invention also provides an industrial production defect assessment system integrating deep learning, comprising: a workpiece fixing module for receiving industrial production defect assessment instructions and determining the target workpiece to be assessed according to the defect assessment instructions, wherein the target workpiece is a metal part with a complex surface structure; fixing the target workpiece in a pre-constructed multi-dimensional scanning device, wherein the multi-dimensional scanning device is a component of a pre-constructed defect detection system, the defect detection system further comprising a light source adjustment module, an image acquisition module, and a data processing unit, wherein the multi-dimensional scanning device is connected to the light source adjustment module, the light source adjustment module is used to dynamically adjust the light intensity, and the image acquisition module is connected to the multi-dimensional scanning device; and a signal transmission module for activating a pre-constructed deep learning analysis device when the target workpiece is successfully fixed in the multi-dimensional scanning device, wherein the deep learning analysis device includes a feature extraction unit, a defect classification unit, and a result feedback unit, the feature extraction unit is connected to the image acquisition module, and the result processing unit is used to activate .... The defect classification unit is connected to the feature extraction unit; the defect evaluation command is transmitted to the result feedback unit, which parses the defect evaluation command to obtain the surface material characteristics, defect type, and evaluation time of the target workpiece, wherein the surface material characteristics and defect type are response features; the parameter calculation module is used to calculate the illumination intensity value of the light source adjustment module and the acquisition frequency of the image acquisition module based on the surface material characteristics and defect type; the workpiece evaluation module is used to start the light source adjustment module and the image acquisition module based on the pre-acquired actual evaluation strategy, the illumination intensity value, and the acquisition frequency, and monitor the actual illumination intensity of the light source adjustment module, the actual acquisition frequency of the image acquisition module, and the actual detection time of the light source adjustment module and the image acquisition module. When the actual illumination intensity is equal to the illumination intensity value, the actual acquisition frequency is equal to the acquisition frequency, and the actual time is the same as the evaluation time, the operation of the light source adjustment module and the image acquisition module is stopped, and the defect evaluation of the target workpiece is completed.

[0024] To address the aforementioned problems, the present invention also provides an electronic device comprising: a memory storing at least one instruction; and a processor executing the instructions stored in the memory to implement the aforementioned industrial production defect assessment method incorporating deep learning.

[0025] To address the aforementioned problems, the present invention also provides a computer-readable storage medium storing at least one computer-executable instruction, which is executed by a processor in an electronic device to implement the aforementioned industrial production defect assessment method incorporating deep learning.

[0026] To address the problems described in the background art, this invention combines receiving industrial production defect assessment instructions with constructing an intelligent assessment process encompassing multiple steps, including fixing the target workpiece, deep learning analysis, and parameter calculation. This process comprehensively assesses the defects of the target workpiece, improving the accuracy and efficiency of defect assessment. Through intelligent parameter calculation and real-time monitoring, this method significantly enhances the reliability of the assessment process, meeting the demands of modern industry for efficient and accurate defect assessment. First, the system receives industrial production defect assessment instructions and determines the target workpiece based on these instructions. By receiving specific defect assessment instructions, the system can identify the target workpiece, improving the accuracy of the assessment process and enabling rapid adjustment of the assessment object according to actual needs and application environment. This optimizes resource allocation and utilization efficiency, reducing testing time and costs. Second, the target workpiece is fixed in a multi-dimensional scanning device. Using a multi-dimensional scanning device to fix the target workpiece ensures stability during the detection process. The design of the multi-dimensional scanning device enhances the safety of the detection, preventing accidental displacement or damage during the detection process, thus ensuring the reliability of the detection results. Finally, once the target workpiece is successfully fixed in the multi-dimensional scanning device, deep learning analysis is initiated. The deep learning analysis device automates the defect assessment process, reducing the risk of errors from human operation. By precisely controlling the light source adjustment module and image acquisition module, it enables rapid and efficient detection, improving overall detection efficiency. Then, it parses the defect assessment command to obtain the surface material characteristics and defect types of the target workpiece. This provides key performance indicators for defect assessment, which help evaluate the defect performance of the target workpiece. Furthermore, based on the surface material characteristics and defect types, it calculates the light intensity value of the light source adjustment module and the acquisition frequency of the image acquisition module. By calculating the light intensity value and acquisition frequency, it can automatically adjust the detection conditions according to actual needs, ensuring the accuracy and effectiveness of the assessment. This intelligent adjustment mechanism improves the adaptability of the detection to handle target workpieces with different standards. Finally, it monitors the actual light intensity and acquisition frequency, stopping the detection. Real-time monitoring of the actual light intensity and acquisition frequency ensures that the defect assessment of the target workpiece is performed under set conditions, avoiding detection errors caused by improper operation. When the actual conditions match the preset values, the detection automatically stops, improving the efficiency of defect assessment. Therefore, the present invention can improve the efficiency of industrial production defect assessment. Attached Figure Description

[0027] Figure 1 A flowchart illustrating an industrial production defect assessment method incorporating deep learning, provided as an embodiment of the present invention;

[0028] Figure 2A functional block diagram of an industrial production defect assessment system integrating deep learning provided in an embodiment of the present invention;

[0029] Figure 3 This is a schematic diagram of the structure of an electronic device that implements the industrial production defect assessment method based on fused deep learning, according to an embodiment of the present invention. Detailed Implementation

[0030] This invention provides a method and system for industrial production defect assessment that integrates deep learning. Its core lies in using an intelligent process to perform comprehensive and accurate defect detection and assessment of target workpieces in industrial production. The following is in conjunction with the appendix... Figure 1 To be continued Figure 3 The technical solution of the present invention will be described in detail with reference to specific embodiments.

[0031] As attached Figure 1 As shown, the industrial production defect assessment method integrating deep learning provided by this invention includes several core steps, which together constitute a complete defect assessment process. First, an industrial production defect assessment instruction is received, and the target workpiece to be assessed is determined based on this instruction. The target workpiece is a metal component with a complex surface structure, such as a key component of an automotive engine or a precision part in the aerospace field. In practical applications, the defect assessment instruction is usually issued by the quality control system on the production line and transmitted to the defect assessment system described in this invention via a network. This instruction contains basic information about the target workpiece, such as the workpiece number, material type, and possible types of defects. After receiving the instruction, the system automatically identifies and locks onto the target workpiece, ensuring that subsequent operations are accurately targeted at the object to be assessed.

[0032] Next, the target workpiece is placed and fixed within a pre-constructed multidimensional scanning device. The multidimensional scanning device is one of the core components of the entire defect detection system, and its design fully considers the complex surface structure of the target workpiece and the stability requirements during the detection process. (See attached image) Figure 2 As shown, the multi-dimensional scanning device, together with the light source adjustment module, image acquisition module, and data processing unit, constitutes the hardware architecture of the defect detection system. The multi-dimensional scanning device firmly fixes the target workpiece using mechanical clamps or vacuum adsorption, preventing workpiece displacement due to vibration or other external interference during the inspection process. Furthermore, the multi-dimensional scanning device supports multi-angle rotation and displacement adjustment to comprehensively capture the surface details of the target workpiece from different perspectives. This design not only improves the reliability of the inspection results but also significantly enhances the system's adaptability, enabling it to handle workpieces with various complex shapes.

[0033] Once the target workpiece is successfully fixed, the system activates a pre-built deep learning analysis device. This device includes a feature extraction unit, a defect classification unit, and a result feedback unit. The feature extraction unit, connected to the image acquisition module, is responsible for extracting key features from the acquired image data, such as surface texture, edge contours, and reflectivity. The defect classification unit, based on the extracted features, uses a pre-trained deep learning model to classify the defect types of the target workpiece. The result feedback unit parses the defect evaluation instructions, extracting response features such as the surface material properties, defect type, and evaluation time of the target workpiece. These features provide crucial information for subsequent parameter calculations and optimization of detection conditions.

[0034] To ensure the efficiency and accuracy of the inspection process, the system needs to calculate the illumination intensity value of the light source adjustment module and the acquisition frequency of the image acquisition module based on the surface material characteristics and defect type of the target workpiece. Regarding the calculation of the illumination intensity value, the system first obtains the surface reflectivity of the target workpiece and then calculates the surface light absorption rate based on the surface reflectivity and surface material characteristics. The formula for calculating the surface light absorption rate is:

[0035] α=1-ρ·β

[0036] Where α represents the surface light absorptivity, ρ represents the surface material properties, and β represents the surface reflectivity. Subsequently, the system, considering the light source power, light source angle, ambient light interference, and light source attenuation coefficient from the light source adjustment module, calculates the illuminance value using the following formula:

[0037] I = P·cos(θ) - E·λ

[0038] Where I represents the light intensity, P represents the light source power, θ represents the light source angle, E represents ambient light interference, and λ represents the light source attenuation coefficient. This calculation process ensures that the lighting conditions can be dynamically adjusted according to the specific characteristics of the target workpiece, thereby improving the adaptability and accuracy of the detection.

[0039] Simultaneously, the system also needs to calculate the acquisition frequency of the image acquisition module. The calculation of the acquisition frequency depends on several factors, including the acquisition resolution of the image acquisition module, the surface roughness value of the target workpiece, the optical transmittance, and the overall response time of the image acquisition module and the multi-dimensional scanning device. The formula for calculating the acquisition frequency is:

[0040]

[0041] Where f represents the acquisition frequency, R represents the acquisition resolution, T represents the overall response time, τ represents the optical transmittance, σ represents the surface roughness value, and δ represents the defect type. Using this calculation formula, the system can optimize the acquisition frequency according to the specific characteristics of the target workpiece, thereby improving detection efficiency while ensuring image quality.

[0042] After calculating the illumination intensity and acquisition frequency, the system generates an evaluation strategy and sends it back to the initiator of the defect evaluation command. The evaluation strategy includes three selectable modes: Strategy A evaluates the surface material properties of the target workpiece first, then the defect type; Strategy B evaluates the defect type first, then the surface material properties; and Strategy C evaluates both surface material properties and defect type simultaneously. The initiator selects the appropriate evaluation strategy based on actual needs. Upon receiving the selected evaluation strategy, the system activates the light source adjustment module and the image acquisition module based on the pre-acquired evaluation strategy, illumination intensity, and acquisition frequency. At this point, the system begins real-time monitoring of the actual illumination intensity of the light source adjustment module, the actual acquisition frequency of the image acquisition module, and their running time. When the actual illumination intensity equals the calculated illumination intensity, the actual acquisition frequency equals the calculated acquisition frequency, and the actual running time is the same as the evaluation time, the system automatically stops the operation of the light source adjustment module and the image acquisition module, completing the defect evaluation of the target workpiece.

[0043] In the above process, signal transmission quality control is a crucial step in ensuring the accurate delivery of evaluation instructions. The system achieves efficient signal transmission by constructing multiple independent signal transmission channels. Each signal transmission channel has a unique identifier, and its performance is evaluated by calculating a transmission quality index. The calculation of the transmission quality index involves several parameters, including signal distortion rate, bandwidth utilization, average signal delay, and maximum signal delay. Specifically, the system randomly selects one channel from the set of identified transmission channels for signal transmission and uses a pre-set signal monitoring tool to capture the complete waveform and distorted waveform during transmission. The formula for calculating the signal distortion rate is:

[0044]

[0045] Where D represents the signal distortion rate, N d N represents the number of distorted waveforms. t This represents the total number of waveforms. Subsequently, the system sets the signal transmission interval and signal transmission period, records the signal delay based on the periodically transmitted dataset, and calculates the average signal delay. Finally, the formula for calculating the transmission quality index is:

[0046]

[0047] Where Q represents the transmission quality index, w1, w2, and w3 represent the distortion weighting coefficient, bandwidth weighting coefficient, and delay weighting coefficient, respectively, U represents the bandwidth utilization rate, and L... m L represents the maximum signal delay. a This represents the average signal delay. Using this formula, the system can comprehensively evaluate the performance of each transmission channel and select the best transmission channels for actual signal transmission.

[0048] To further ensure the reliability of signal transmission, the system determines the transmission quality index set based on a preset quality judgment function. The quality judgment function is constructed based on the mean and standard deviation of the exponents in the transmission quality index set, and its formula is:

[0049]

[0050] Among them, F(Q) i ) represents the output of the quality judgment function, Q i Let μ represent the mean of a single transmission quality index, σ represent the standard deviation of the index, and k represent the adjustment factor. The system sequentially extracts the transmission quality indices and substitutes them into the quality judgment function for evaluation. When the judgment output is 0, the index is identified as a poor transmission index and removed from the set; when the judgment output is 1, it is identified as a good transmission index. After multiple iterations, the system finally obtains a set of high-quality transmission channels for subsequent signal transmission.

[0051] As attached Figure 3 As shown, the present invention also provides an electronic device for implementing the above-described industrial production defect assessment method incorporating deep learning. The electronic device includes a memory and a processor. The memory stores at least one computer-executable instruction, which the processor executes to implement the various steps of the method. Furthermore, the present invention also provides a computer-readable storage medium in which the stored instructions can also be executed by the processor in the electronic device to implement the above-described method.

[0052] In summary, this invention significantly improves the accuracy, efficiency, and adaptability of defect assessment by combining multiple steps, including receiving industrial production defect assessment instructions, constructing an intelligent assessment process encompassing the fixation of the target workpiece to be assessed, deep learning analysis, and parameter calculation. In practical applications, this invention has been successfully applied to the engine parts production line of an automobile manufacturing company. Through comprehensive inspection of the target workpiece, it has effectively reduced the defect rate and improved production efficiency.

Claims

1. A method for assessing defects in industrial production that integrates deep learning, characterized in that, The method includes: Receive an industrial production defect assessment instruction, and determine the target workpiece to be assessed based on the defect assessment instruction, wherein the target workpiece is a metal part with a complex surface structure; The target workpiece is placed and fixed in a pre-constructed multi-dimensional scanning device, and the multi-dimensional scanning device is one of the components of the pre-constructed defect detection system. The defect detection system also includes a light source adjustment module, an image acquisition module, and a data processing unit. The multi-dimensional scanning device is connected to the light source adjustment module, which is used to dynamically adjust the light intensity. The image acquisition module is connected to the multi-dimensional scanning device. When the target workpiece is successfully fixed in the multi-dimensional scanning device, the pre-built deep learning analysis device is activated. The deep learning analysis device includes a feature extraction unit, a defect classification unit, and a result feedback unit. The feature extraction unit is connected to the image acquisition module, and the defect classification unit is connected to the feature extraction unit. The defect assessment command is transmitted to the result feedback unit, which then parses the defect assessment command to obtain the surface material characteristics, defect type, and assessment time of the target workpiece. The surface material characteristics and defect type are the response features. The illumination intensity value of the light source adjustment module and the acquisition frequency of the image acquisition module are calculated based on the surface material characteristics and defect types. Based on the pre-acquired actual evaluation strategy, the light intensity value, and the acquisition frequency, the light source adjustment module and the image acquisition module are started. The actual light intensity of the light source adjustment module, the actual acquisition frequency of the image acquisition module, and the actual detection time of the light source adjustment module and the image acquisition module are monitored. When the actual light intensity is equal to the light intensity value, the actual acquisition frequency is equal to the acquisition frequency, and the actual time is the same as the evaluation time, the operation of the light source adjustment module and the image acquisition module is stopped, and the defect evaluation of the target workpiece is completed.

2. The industrial production defect assessment method integrating deep learning as described in claim 1, characterized in that, The process of transmitting the defect assessment command to the result feedback unit, and using the result feedback unit to parse the defect assessment command to obtain the surface material characteristics, defect type, and assessment time of the target workpiece includes: Multiple signal transmission channels are constructed, wherein the multiple signal transmission channels are independent of each other, and each signal transmission channel can independently transmit the same defect assessment command; The multiple signal transmission channels are aggregated to obtain a signal transmission channel set, and each signal transmission channel in the signal transmission channel set is target identified to obtain an identified transmission channel set, wherein each identified transmission channel in the identified transmission channel set has a unique identifier; Calculate the transmission quality index of each identification transmission channel in the identification transmission channel set, aggregate the transmission quality indices of each identification transmission channel to obtain a transmission quality index set, and determine the transmission quality index set based on a preset quality judgment function to obtain a high-quality transmission channel set. Randomly select a transmission channel from the set of high-quality transmission channels, use the first transmission channel to transmit the signal and record the time to obtain the signal transmission time; Determine whether the signal transmission time is greater than a preset time threshold; If the signal transmission time is greater than the time threshold, the first transmission channel is removed from the set of high-quality transmission channels to obtain an updated transmission channel set. The high-quality transmission channel set is updated using the updated transmission channel set, and the above steps of randomly selecting the first transmission channel from the set of high-quality transmission channels are returned. If the signal transmission time is not greater than the time threshold, the result feedback unit is used to parse the defect evaluation command to obtain the surface material characteristics, defect type and evaluation time of the target workpiece.

3. The industrial production defect assessment method integrating deep learning as described in claim 2, characterized in that, The method for calculating the transmission quality index is as follows: One identification transmission channel is randomly selected from the set of identification transmission channels, and signal transmission is performed based on the identification transmission channel. All signal waveforms during the signal transmission process are captured using a preset signal monitoring tool, wherein the signal waveforms include complete waveforms and distorted waveforms. The signal distortion rate is calculated using the complete waveform and the distorted waveform. Set the signal transmission interval and signal transmission period, and perform periodic signal transmission based on the signal transmission interval and signal transmission period to obtain a periodic signal dataset; Delay recording is performed on the periodic signal dataset to obtain a signal delay set, and the mean is calculated on the signal delay set to obtain the signal delay mean; Obtain the maximum signal delay, bandwidth utilization, distortion weight coefficient, delay weight coefficient, and bandwidth weight coefficient, and calculate the transmission quality index based on the maximum signal delay, bandwidth utilization, distortion weight coefficient, delay weight coefficient, bandwidth weight coefficient, signal distortion rate, and average signal delay.

4. The industrial production defect assessment method integrating deep learning as described in claim 3, characterized in that, The calculation of the transmission quality index based on the maximum signal delay, bandwidth utilization, distortion weighting coefficient, delay weighting coefficient, bandwidth weighting coefficient, signal distortion rate, and mean signal delay includes: Where Q represents the transmission quality index, w1, w2, and w3 represent the distortion weighting coefficient, bandwidth weighting coefficient, and delay weighting coefficient, respectively, U represents the bandwidth utilization rate, and L... m L represents the maximum signal delay. a This represents the average signal delay.

5. The industrial production defect assessment method integrating deep learning as described in claim 2, characterized in that, The process of determining the transmission quality index set based on a preset quality determination function to obtain a set of high-quality transmission channels includes: Calculate the mean and standard deviation of the indices of the transmission quality index set; Obtain the adjustment factor, and construct a quality judgment function based on the adjustment factor, the index mean, and the index standard deviation: Among them, F(Q) i ) represents the output of the quality judgment function, Q i σ represents a single transmission quality index, μ represents the index mean, σ represents the index standard deviation, and k represents the adjustment factor. The transmission quality indices are extracted sequentially from the set of transmission quality indices, and the extracted transmission quality indices are substituted into the quality determination function to obtain the determination output; When the output is 0, the transmission quality index is confirmed as a poor transmission index. The poor transmission index is removed from the transmission quality index set to obtain an updated index set. The transmission quality index set is updated using the updated index set and the above steps of extracting transmission quality indices sequentially from the transmission quality index set are returned. When the output is 1, the transmission quality index is confirmed as a high-quality transmission index, and the process of extracting transmission quality indices from the set of transmission quality indices is returned. Once all transmission quality indices in the transmission quality index set are extracted and substituted into the quality determination function, the transmission quality index set is confirmed as a high-quality transmission channel set.

6. The industrial production defect assessment method integrating deep learning as described in claim 1, characterized in that, The method for calculating the light intensity value is as follows: Obtain the surface reflectance of the target workpiece, and calculate the surface light absorption rate based on the surface reflectance and surface material properties: α=1-ρ·β Where α represents the surface light absorptivity, ρ represents the surface material properties, and β represents the surface reflectivity; The light source power, light source angle, ambient light interference, and light source attenuation coefficient of the light source adjustment module are obtained, and the illuminance value is calculated based on the light source power, light source angle, ambient light interference, and light source attenuation coefficient.

7. The industrial production defect assessment method integrating deep learning as described in claim 6, characterized in that, The calculation of the illuminance value based on the light source power, light source angle, ambient light interference, and light source attenuation coefficient includes: The light intensity value can be calculated using the following formula: I = P·cos(θ) - E·λ Where I represents the illuminance value, P represents the power of the light source, θ represents the angle of the light source, E represents the ambient light interference, and λ represents the light source attenuation coefficient.

8. The industrial production defect assessment method integrating deep learning as described in claim 1, characterized in that, The method for calculating the sampling frequency is as follows: The acquisition resolution of the image acquisition module is obtained, as are the surface roughness value and optical transmittance of the target workpiece. The overall response time of the image acquisition module and the multi-dimensional scanning device is also obtained. Based on the overall response time, acquisition resolution, surface roughness value, optical transmittance, defect type, and surface material characteristics, the acquisition frequency is calculated. Where f represents the acquisition frequency, R represents the acquisition resolution, T represents the overall response time, τ represents the optical transmittance, σ represents the surface roughness value, and δ represents the defect type.

9. The industrial production defect assessment method integrating deep learning as described in claim 1, characterized in that, Before the pre-acquired actual evaluation strategy, the light intensity value, and the acquisition frequency are used to activate the light source adjustment module and the image acquisition module, the method further includes: After the light intensity value of the light source adjustment module and the acquisition frequency of the image acquisition module are calculated, an evaluation strategy is generated and the evaluation strategy is sent back to the initiator of the defect evaluation command. The evaluation strategy includes: Strategy A: first evaluate the surface material characteristics of the target workpiece, and then evaluate the defect type of the target workpiece; Strategy B: first evaluate the defect type of the target workpiece, and then evaluate the surface material characteristics of the target workpiece; Strategy C: evaluate the surface material characteristics and defect type of the target workpiece at the same time. The initiator of the defect assessment instruction selects the assessment strategy and obtains the actual assessment strategy.

10. An industrial production defect assessment system integrating deep learning, characterized in that, The system includes: A workpiece fixing module is used to receive industrial production defect assessment instructions and determine the target workpiece to be assessed according to the defect assessment instructions. The target workpiece is a metal part with a complex surface structure. The target workpiece is fixed in a pre-constructed multi-dimensional scanning device, which is a component of a pre-constructed defect detection system. The defect detection system also includes a light source adjustment module, an image acquisition module, and a data processing unit. The multi-dimensional scanning device is connected to the light source adjustment module, which is used to dynamically adjust the light intensity. The image acquisition module is connected to the multi-dimensional scanning device. The signal transmission module is used to activate a pre-built deep learning analysis device when the target workpiece is successfully fixed in the multi-dimensional scanning device. The deep learning analysis device includes a feature extraction unit, a defect classification unit, and a result feedback unit. The feature extraction unit is connected to the image acquisition module, and the defect classification unit is connected to the feature extraction unit. The defect evaluation command is transmitted to the result feedback unit, which parses the defect evaluation command to obtain the surface material characteristics, defect type, and evaluation time of the target workpiece. The surface material characteristics and defect type are response features. The parameter calculation module is used to calculate the light intensity value of the light source adjustment module and the acquisition frequency of the image acquisition module based on the surface material characteristics and defect type. The workpiece evaluation module is used to activate the light source adjustment module and the image acquisition module based on the pre-acquired actual evaluation strategy, the light intensity value, and the acquisition frequency. It monitors the actual light intensity of the light source adjustment module, the actual acquisition frequency of the image acquisition module, and the actual detection time of the light source adjustment module and the image acquisition module. When the actual light intensity is equal to the light intensity value, the actual acquisition frequency is equal to the acquisition frequency, and the actual time is the same as the evaluation time, the operation of the light source adjustment module and the image acquisition module is stopped, and the defect evaluation of the target workpiece is completed.

Images