A random vibration test method and system

Abstract

The present invention proposes a random vibration test method and system. Belonging to the field of vibration test technology, the method includes: using an image acquisition device to respectively acquire appearance images of a plurality of tested samples, and numbering the tested samples, and respectively stitching the acquired appearance images with different numbers into complete product pictures through an image stitching algorithm; performing preliminary processing on the product pictures through an image processing algorithm, extracting key feature information of the sample surface, and storing them in a feature library respectively; placing a plurality of tested samples on different experimental platforms respectively, performing random vibration tests, and collecting mechanical data in real time through sensors, while the image acquisition device continues to work to capture subtle changes on the sample surface in real time. The collected images and mechanical data are transmitted to the cloud platform through a multi-channel transmission protocol, and the best transmission path is selected using a performance evaluation and optimization algorithm to ensure the timeliness and reliability of the data.

Description

Technical Field

[0001] The invention provides a random vibration test method and system, belonging to the technical field of vibration testing. Background Art

[0002] In the process of product design and manufacturing, vibration testing is a key link in evaluating product durability and reliability. Traditional vibration testing methods often focus on the collection and analysis of mechanical data, ignoring the real-time monitoring and evaluation of changes in sample appearance. With the development of image processing technology, combined with high-speed image acquisition and processing technology, it is possible to accurately capture and analyze subtle changes on the sample surface, thereby more comprehensively evaluating the impact of vibration on the sample. However, how to ensure the clarity and continuity of image acquisition under high-frequency vibration conditions, and how to improve data transmission efficiency and ensure transmission quality, are still challenges facing current technology. Summary of the invention

[0003] The present invention provides a random vibration test method and system to solve the problems mentioned in the above background technology:

[0004] The present invention provides a random vibration test method, the method comprising:

[0005] S1. Use an image acquisition device to collect appearance images of multiple samples to be tested, and number the samples to be tested. Use an image stitching algorithm to stitch the collected appearance images with different numbers into a complete product picture;

[0006] S2. Perform preliminary processing on product images through image processing algorithms, extract key feature information of sample surfaces, and store them in feature libraries respectively;

[0007] S3, placing multiple samples on different experimental platforms for random vibration testing, collecting mechanical data in real time through sensors, while the image acquisition device works continuously to capture subtle changes on the sample surface in real time;

[0008] S4, transmitting the collected images and mechanical data to the cloud platform in parallel through multiple paths through a multi-channel transmission protocol;

[0009] S5. After receiving the images and mechanical data, the cloud platform conducts cross-analysis on the test results of different experimental platforms to obtain comprehensive analysis results;

[0010] S6. Based on the comprehensive analysis results, the intelligent evaluation model is used to automatically determine whether the sample is qualified or not, and generate targeted optimization suggestions or early warning signals based on the specific circumstances of the appearance changes.

[0011] Furthermore, the S1 includes:

[0012] S11, inspecting all the tested samples, and acquiring multi-angle appearance images of the multiple tested samples after inspection by an image acquisition device;

[0013] S12, after the appearance is collected, each sample is uniquely numbered and identified with an RFID tag;

[0014] S13, applying a feature-based image stitching algorithm to align and stitch images taken from multiple angles;

[0015] S14. Perform detail enhancement processing on the spliced ​​image, and separate key components and areas in the product image through an image segmentation algorithm.

[0016] Furthermore, the S2 includes:

[0017] S21, removing noise from the image by Gaussian filtering, adjusting the contrast of the image by an adaptive contrast enhancement algorithm; and unifying the image parameters to a preset standard;

[0018] S22, performing gradient calculation on the preprocessed image to obtain gradient maps of the image in the horizontal and vertical directions; based on the gradient map, applying an edge detection algorithm to extract edge information in the image;

[0019] S23, extracting texture features from the image after edge detection using a gray level co-occurrence matrix; extracting shape features based on the edge detection result;

[0020] S24, using principal component analysis to evaluate the extracted features, identifying the features that contribute to the classification or recognition task, and selecting a feature subset based on the evaluation results; and performing normalization and quantization on the selected features;

[0021] S25, combining the optimized features into a feature vector, and associating the feature vector of each sample with its unique identification code, and storing them in a feature library.

[0022] Furthermore, the S3 includes:

[0023] S31. Using an automated control system, multiple customized experimental platforms are automatically configured according to preset vibration test parameters;

[0024] S32, use the automated fixture system to place and fix the sample to be tested on the experimental platform;

[0025] S33, sending instructions through the central control system, the random vibration generators of all experimental platforms are started simultaneously, and vibration tests are performed according to preset parameters;

[0026] S34, the image acquisition device is automatically activated when the vibration test is started, and the key area of ​​the sample is locked using a preset tracking algorithm to capture the image;

[0027] S35. The sensor network automatically collects mechanical data and sends it to the data center in real time through wireless transmission technology, and pre-processes the data.

[0028] Furthermore, the S4 includes:

[0029] S41, selecting a multi-channel transmission protocol, and performing performance evaluation on each transmission channel based on a performance evaluation model;

[0030] S42, deploy monitoring points at key nodes and transmission paths of the network, collect real-time network data, process the collected network data through machine learning algorithms, and calculate real-time network status indicators;

[0031] S43, predicting the network status change in the future period of time through a network status prediction model based on historical data, and formulating a transmission strategy according to transmission requirements;

[0032] S44, through the channel selection algorithm based on multi-objective optimization, the best channel is selected from multiple candidate channels to transmit the image and mechanical data to the cloud platform;

[0033] S45. During the transmission process, when the network status changes, the dynamic adjustment mechanism is triggered to evaluate and reselect the transmission channel in real time.

[0034] Furthermore, the S5 includes:

[0035] S51, after receiving the image and mechanical data, the cloud platform performs image matching through a deep learning algorithm;

[0036] S52, processing the differential image, and combining it with the mechanical data, performing mechanical effect analysis on the changed area in the differential image to identify the specific damage type caused by the vibration;

[0037] S53. Cross-compare the test results of different experimental platforms to identify common and different problems; conduct a comprehensive evaluation of the samples through a comprehensive evaluation model based on machine learning;

[0038] S54, and visualize the data analysis results.

[0039] Furthermore, the S6 includes:

[0040] S61. Based on the comprehensive analysis results, the eligibility of the samples is automatically judged through the intelligent evaluation model, and the samples are divided into qualified and unqualified categories;

[0041] S62. Further classify the unqualified samples to identify the failure modes and potential problem areas;

[0042] S63, generating optimization suggestions according to the failure mode and the problem area, wherein the optimization suggestions include design improvement, material selection, and process optimization;

[0043] S64. Set up an early warning system to promptly issue early warning signals for samples with risks or potential failures, trigger emergency response mechanisms, and notify relevant personnel;

[0044] S65. Store the data, evaluation results and optimization suggestions of each test into the knowledge base, and continuously improve the test case library and evaluation model.

[0045] The present invention provides a random vibration test system, the system comprising:

[0046] Image acquisition module: The image acquisition device is used to collect appearance images of multiple samples to be tested, and the samples to be tested are numbered. The collected appearance images with different numbers are stitched into a complete product picture through the image stitching algorithm;

[0047] Data storage module: Use image processing algorithms to preliminarily process product images, extract key feature information on the sample surface, and store them in feature libraries;

[0048] Vibration test module: Multiple samples to be tested are placed on different experimental platforms for random vibration testing. Mechanical data is collected in real time through sensors. Meanwhile, the image acquisition device works continuously to capture subtle changes on the sample surface in real time.

[0049] Data transmission module: The collected images and mechanical data are transmitted to the cloud platform in parallel through multiple paths through a multi-channel transmission protocol;

[0050] Comparative analysis module: After receiving the images and mechanical data, the cloud platform conducts cross-analysis on the test results of different experimental platforms to obtain comprehensive analysis results;

[0051] Recommendation and warning module: Based on the comprehensive analysis results, the intelligent evaluation model is used to automatically determine whether the sample is qualified or not, and generate targeted optimization suggestions or warning signals based on the specific circumstances of the appearance changes.

[0052] The present invention provides a non-transitory computer-readable storage medium, on which a computer program is stored. The program is executed by a processor to implement any of the random vibration test methods described above.

[0053] A computer program product proposed by the present invention includes a computer program / instruction, and when the computer program / instruction is executed by a processor, the random vibration test method as described above is implemented.

[0054] The beneficial effects of the present invention are as follows: through image acquisition equipment and processing algorithms, the appearance characteristics and surface changes of the tested samples, as well as the real-time collection and processing of mechanical data can be fully captured to ensure the comprehensiveness and accuracy of the data; by using an automated control system and a fixture system, customized configuration and automatic vibration testing of multiple experimental platforms can be realized to improve test efficiency and consistency; a multi-channel transmission protocol is adopted, combined with performance evaluation and optimization algorithms, to ensure that images and mechanical data can be efficiently and timely transmitted to the cloud platform for comprehensive analysis; on the cloud platform, deep learning algorithms are used for image matching and differential image analysis, combined with machine learning models for comprehensive evaluation of data and problem identification, to improve the accuracy of fault mode identification and evaluation; based on the comprehensive analysis results, the eligibility of the samples is automatically judged through an intelligent evaluation model, and targeted optimization suggestions are generated, including design improvements, material selection and process optimization, to improve product quality and reliability; an early warning system is established to promptly issue early warning signals for samples with risks or potential failures, trigger an emergency response mechanism, and ensure timely resolution and management of problems. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 It is a step diagram of the method of the present invention;

[0056] Figure 2 This is a system module diagram of the present invention. DETAILED DESCRIPTION

[0057] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0058] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. The embodiments described are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0059] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used in the specification of the present invention herein are only for the purpose of describing specific embodiments and are not intended to limit the present invention.

[0060] One embodiment of the present invention, as Figure 1 As shown, a random vibration test method, the method comprising:

[0061] S1. Use an image acquisition device to collect appearance images of multiple samples to be tested, and number the samples to be tested. Use an image stitching algorithm to stitch the collected appearance images with different numbers into a complete product picture;

[0062] S2. Perform preliminary processing on product images through image processing algorithms, extract key feature information of sample surfaces, and store them in feature libraries respectively;

[0063] S3, placing multiple samples on different experimental platforms for random vibration testing, collecting mechanical data in real time through sensors, while the image acquisition device works continuously to capture subtle changes on the sample surface in real time;

[0064] S4, transmitting the collected images and mechanical data to the cloud platform in parallel through multiple paths through a multi-channel transmission protocol;

[0065] S5. After receiving the image and mechanical data, the cloud platform compares the image of the tested sample with the feature information of the tested sample stored in the feature library at the pixel level through image matching and difference algorithms to identify the appearance changes caused by vibration; and obtains comprehensive analysis results through cross-analysis of test results from different experimental platforms;

[0066] S6. Based on the comprehensive analysis results, the intelligent evaluation model is used to automatically determine whether the sample is qualified or not, and generate targeted optimization suggestions or early warning signals based on the specific circumstances of the appearance changes.

[0067] The working principle of the above technical solution is: use high-precision image acquisition equipment (such as high-definition cameras) to take appearance photos of multiple samples to be tested. Each sample is assigned a unique number for easy identification and management. Since the product may have a large size or complex shape, it is necessary to use an image stitching algorithm to stitch multiple local images into a complete, high-resolution product picture; the stitched product picture is preliminarily processed by an image processing algorithm (such as edge detection, feature extraction, etc.) to extract key feature information on the sample surface, such as texture, color, shape, etc. These feature information are crucial for subsequent vibration impact analysis, so they will be accurately extracted and stored in the feature library for later use; the tested samples are placed on different designed experimental platforms, which can simulate various random vibration environments. Vibration stimulation is applied to the sample through a vibration generator to simulate the vibration conditions that may be encountered during actual use or transportation; during the vibration test, sensors (such as accelerometers, displacement sensors, etc.) are used to collect mechanical data of the sample in real time, such as acceleration, displacement, stress, etc. At the same time, the image acquisition device also continues to work to capture subtle changes in the sample surface in real time, such as cracks, deformation, color changes, etc. These data provide an important basis for analyzing the impact of vibration on the appearance of samples. In order to ensure the real-time and accuracy of the data, a multi-channel transmission protocol is used to transmit the collected images and mechanical data to the cloud platform in parallel through multiple paths. This parallel transmission method can significantly improve the efficiency and reliability of data transmission. As the center of data processing and analysis, the cloud platform is responsible for receiving data from various experimental platforms. The cloud platform has powerful data storage and processing capabilities and can process a large amount of data at the same time, providing strong support for subsequent analysis and evaluation. After receiving the sample image after the test, the cloud platform uses image matching and differential algorithms to compare it with the original sample image stored in the feature library at the pixel level. This comparison can accurately identify the appearance changes caused by vibration, such as crack extension and deformation aggravation. In order to obtain more comprehensive analysis results, the cloud platform will also cross-analyze the test results of different experimental platforms. By comparing the response of samples under different vibration conditions, the intrinsic connection between vibration characteristics and appearance changes can be revealed, providing a strong basis for subsequent evaluation and optimization. Based on the comprehensive analysis results, the cloud platform uses an intelligent evaluation model to automatically judge the eligibility of the sample. The intelligent evaluation model combines multiple technologies such as machine learning and expert systems to accurately evaluate whether the performance of the sample in the vibration test meets the preset standards; based on the evaluation results and the specific circumstances of the appearance changes, the cloud platform will generate targeted optimization suggestions or early warning signals. These suggestions may involve sample design improvements, manufacturing process optimization, etc.; and early warning signals are used to promptly remind relevant personnel to pay attention to potential problems and risks.

[0068] The effect of the above technical solution is: multiple local images are stitched into a complete product picture through an image stitching algorithm, avoiding the problem of image loss caused by complex sample size or shape, thereby improving the accuracy of the test. Real-time capture of subtle changes in the sample surface, combined with real-time collection of mechanical data, can more comprehensively reflect the performance of the sample in a vibration environment, improving the comprehensiveness and accuracy of the test. Using image matching and differential algorithms for pixel-level comparison, it is possible to accurately identify appearance changes caused by vibration, such as cracks, deformation, etc., providing reliable data support for subsequent evaluation and optimization. Through cross-analysis of test results on different experimental platforms, the inherent connection between vibration characteristics and appearance changes can be revealed, which helps to discover potential problems and risks and improve the depth and breadth of data analysis. The intelligent evaluation model automatically determines whether the sample is qualified or not, reduces manual intervention, and improves the efficiency and objectivity of the evaluation. Generate targeted optimization suggestions or early warning signals based on the specific circumstances of the appearance changes, providing a clear guiding direction for subsequent improvements and optimizations. The tested samples are numbered and their key feature information is stored to ensure the traceability of the test process. When needed, the test data and results of each sample can be quickly accessed, which is convenient for problem tracking and cause analysis. Data is transmitted to the cloud platform in parallel through a multi-channel transmission protocol, realizing centralized storage and management of data. As the center of data processing and analysis, the cloud platform can easily share data and collaborate with other systems or teams, improving the overall efficiency of testing work. This technical solution is not only used for current test evaluation, but also provides a basis for subsequent continuous optimization and improvement. By analyzing test data and results, we can continuously optimize test methods, improve sample design or manufacturing processes, and improve product reliability and durability.

[0069] In one embodiment of the present invention, the S1 includes:

[0070] S11, inspecting all the tested samples, wherein the inspection includes thorough cleaning to remove dust and impurities on the surface; and acquiring multi-angle appearance images of the multiple tested samples after the inspection by an image acquisition device;

[0071] S12, after the appearance is collected, each sample is uniquely numbered and identified with an RFID tag;

[0072] S13. Apply feature-based image stitching algorithms (such as SURF+RANSAC) to accurately align and stitch images taken from multiple angles;

[0073] S14. Perform detail enhancement processing on the stitched image, including edge sharpening, noise suppression and contrast adjustment, and separate key components and areas in the product image through an image segmentation algorithm.

[0074] The working principle of the above technical solution is as follows: First, all the tested samples are thoroughly inspected to ensure the accuracy and effectiveness of image acquisition. During the inspection process, dust, impurities and other factors that may affect the image quality on the surface of the sample are removed. This step is an important preparation before image acquisition, which helps to improve image quality and reduce the difficulty of subsequent processing; then, the image acquisition equipment (such as a high-definition camera) is used to collect multi-angle appearance images of the inspected samples. Multi-angle shooting can ensure that comprehensive information of the sample is captured, including all sides, top and bottom, so as to avoid image loss or incomplete information due to viewing angle limitations; in order to achieve accurate management and tracking of each sample, each sample after appearance acquisition is uniquely numbered. This number will serve as the unique identifier of the sample throughout the test process, which is convenient for data collation and analysis; at the same time, RFID (radio frequency identification) tags are used to identify the samples. RFID tags have the advantages of non-contact reading, large storage capacity, and strong anti-pollution ability, which can greatly improve the efficiency and accuracy of sample management. Through RFID technology, the sample number and related information can be quickly read, reducing the time and error rate of manual operation; feature-based image stitching algorithms (such as SURF+RANSAC) are used to accurately align and stitch images taken from multiple angles. The SURF (Speeded-Up Robust Features) algorithm is a fast and robust feature point extraction algorithm that can detect stable feature points in an image and calculate its descriptor. The RANSAC (Random Sample Consensus) algorithm is used to estimate the parameters of the mathematical model from a data set containing "outliers", and iteratively estimate the optimal transformation matrix to achieve accurate alignment between images. Combining these two algorithms can achieve accurate stitching of multi-angle images and generate a complete product image; the stitched image is enhanced with details, including edge sharpening, noise suppression, and contrast adjustment. These processing steps are designed to improve the clarity and recognizability of the image, make the key features more prominent, and facilitate subsequent feature extraction and analysis; the key components and regions in the product image are separated by an image segmentation algorithm. Image segmentation is an important technology in image processing, which can divide an image into multiple regions or objects with similar properties. In this step, appropriate segmentation algorithms (such as region-based segmentation, edge-based segmentation, etc.) are selected according to the structure and characteristics of the product to separate key components (such as parts, connections, etc.) and regions (such as surface textures, color regions, etc.) in the product image. This helps to make subsequent feature extraction more accurate and efficient.

[0075] The effect of the above technical solution is: by thoroughly cleaning the surface of the sample to be tested and removing dust and impurities, the interference factors in the image can be significantly reduced, the clarity and contrast of the image can be improved, and a higher quality data basis can be provided for subsequent image processing and feature extraction. Multi-angle appearance image acquisition can ensure that comprehensive information of the sample is captured, including details such as the sides, top and bottom. This comprehensive image acquisition method helps to avoid information omission and improve the accuracy and reliability of the test. Uniquely numbering each sample and identifying it with an RFID tag can achieve accurate management and tracking of the sample. This not only facilitates the collation and analysis of data, but also helps to quickly locate the sample during the test and improve work efficiency. The application of feature-based image stitching algorithms (such as SURF+RANSAC) to accurately align and stitch images taken from multiple angles can generate complete and high-quality product pictures. This stitching method not only retains the detailed information of the image, but also avoids the problem of image loss caused by viewing angle limitations. Detail enhancement processing of the stitched image, including edge sharpening, noise suppression and contrast adjustment, can further improve the recognizability and clarity of the image. At the same time, the key components and areas in the product image are separated by the image segmentation algorithm, which helps to simplify the subsequent feature extraction process and improve processing efficiency. The key components and areas of the image processed by detail enhancement and image segmentation are more prominent and clear, which provides a more accurate and reliable data source for subsequent feature extraction. Accurate feature extraction is an important basis for vibration test analysis and helps to improve the accuracy and reliability of test results. By identifying samples with RFID tags, data can be quickly read and shared. During the test process, different departments or teams can easily obtain and use this data to promote collaborative work and data exchange.

[0076] In one embodiment of the present invention, the S2 includes:

[0077] S21, removing noise from the image by using Gaussian filtering or median filtering, adjusting the contrast of the image by using an adaptive contrast enhancement algorithm; and unifying image parameters to a preset standard, the parameters including size and resolution;

[0078] S22, performing gradient calculation on the preprocessed image to obtain a gradient map of the image in the horizontal and vertical directions; the gradient map reflects the rate of change of pixel values ​​in the image, and based on the gradient map, an edge detection algorithm (Canny, Laplacian of Gaussian) is applied to extract edge information in the image; wherein the image gradient TD is obtained by the following formula:

[0079]

[0080] Where H is the total number of pixels in the row, L is the total number of pixels in the column, Represents the horizontal gradient of the defect image, Represents the vertical gradient of the defect image.

[0081] S23, extracting texture features from the image after edge detection using a gray level co-occurrence matrix (GLCM); extracting shape features based on the edge detection result, wherein the shape features include contour perimeter, area, circularity, rectangularity, and Hu moment;

[0082] S24. Use principal component analysis (PCA) to evaluate the extracted features, identify the features that contribute to the classification or recognition task, and select the most representative feature subset based on the evaluation results; and perform normalization and quantization on the selected features.

[0083] S25, combining the optimized features into feature vectors in a certain order, wherein each feature vector represents feature information of a sample at a specific viewing angle; and associating the feature vector of each sample with its unique identification code, and storing them in a feature library.

[0084] The working principle of the above technical solution is: remove the noise in the image through algorithms such as Gaussian filtering or median filtering to improve the image quality. These filtering methods can smooth the random noise in the image while retaining the edge information of the image as much as possible; use an adaptive contrast enhancement algorithm to adjust the contrast of the image to make the features in the image more distinct. Contrast enhancement helps to improve the visual effect of the image and improve the accuracy of subsequent feature extraction; unify the parameters of the image (such as size and resolution) to the preset standards to ensure the comparability between different images. This step is an important part of image preprocessing and provides a standardized data basis for subsequent feature extraction; perform gradient calculation on the preprocessed image to obtain the gradient map of the image in the horizontal and vertical directions. The gradient map reflects the rate of change of pixel values ​​in the image and is an important basis for edge detection; based on the gradient map, edge detection algorithms (such as Canny algorithm and Laplacian of Gaussian algorithm) are applied to extract edge information in the image. Edge information is one of the most basic features in the image and is crucial for subsequent shape analysis, texture recognition, etc.; use algorithms such as gray level co-occurrence matrix (GLCM) to extract texture features from the image after edge detection. These features can describe the texture pattern and periodic structure of the image surface, providing important information for subsequent recognition and analysis; based on the edge detection results, shape features such as contour perimeter, area, circularity, rectangularity, Hu moment, etc. are extracted. These features can reflect the basic shape and geometric properties of the object, which are of great significance for tasks such as shape recognition and object classification; PCA is used to evaluate the extracted features and identify the features that contribute most to the classification or recognition task. PCA removes redundant information between features through dimensionality reduction technology, selects the most representative feature subset, reduces feature dimensions, and improves the speed and efficiency of subsequent processing; normalizes and quantizes the selected features to make different features comparable and reduce computational complexity. Feature processing is an important part of feature optimization, which helps to improve the accuracy of subsequent analysis and evaluation; the optimized features are combined into feature vectors in a certain order, and each feature vector represents the feature information of a sample at a specific perspective. The feature vector is a numerical representation of the sample feature, which provides data support for subsequent analysis and evaluation; the feature vector of each sample is associated with its unique identification code and stored in the feature library. The feature library supports fast retrieval and comparison, providing data support for subsequent vibration test analysis, sample identification, fault diagnosis and other tasks.

[0085] The effect of the above technical solution is: by removing image noise through Gaussian filtering or median filtering, the clarity and signal-to-noise ratio of the image are significantly improved, providing a higher quality data source for subsequent feature extraction. The adaptive contrast enhancement algorithm adjusts the image contrast, making the features in the image more distinct, further enhancing the recognizability of the image. The size and resolution of the image are unified to the preset standard, eliminating the processing deviation caused by different image sizes and resolutions, and improving the consistency and comparability of the processing results. The application of gradient calculation and edge detection algorithms (such as Canny and LoG) can accurately extract edge information in the image, which is the basis for subsequent shape analysis and texture recognition. The accuracy of edge detection directly affects the accuracy and reliability of subsequent feature extraction, and is a crucial step in image processing. The use of gray level co-occurrence matrix (GLCM) to extract texture features can describe the texture pattern and periodic structure of the image surface in detail, providing rich information for texture recognition. The extraction of shape features (such as contour perimeter, area, circularity, rectangularity, Hu moment) reflects the basic shape and geometric properties of the object, providing an important basis for shape analysis and object recognition. The application of principal component analysis (PCA) realizes the evaluation and selection of features, removes redundant information, retains the most representative feature subset, and effectively reduces the feature dimension. Feature dimension reduction not only improves the speed and efficiency of subsequent processing, but also reduces the computational complexity, making the entire processing process more efficient. The selected features are normalized and quantized to make different features comparable, further improving the consistency and reliability of the features. Normalization and quantization also help to reduce calculation errors and improve the accuracy of subsequent analysis and evaluation. The optimized features are combined into feature vectors and stored in the feature library in association with the unique identification code of the sample, realizing the rapid retrieval and comparison of feature data. The efficient construction of the feature library provides strong data support for subsequent analysis and evaluation, making the entire vibration test analysis process more systematic and automated. The above gradient formula fully reflects the rate of change of pixel values ​​in the image by calculating the gradient of each pixel in the horizontal and vertical directions and finding the square root of the sum of their squares. This rate of change is an important basis for edge detection, because edges often correspond to sharp changes in pixel values; by calculating the gradient map, the edge information in the image can be clearly displayed. Edges are the dividing lines between objects and backgrounds or different objects in an image, and are crucial for subsequent shape analysis, texture recognition, and other tasks. The use of gradient maps can significantly improve the accuracy and reliability of edge detection. Edge detection based on gradient maps can extract more accurate and rich edge information. As part of the features, these edge information can more accurately describe the objects and scenes in the image, providing strong support for subsequent feature extraction and recognition tasks. Before calculating the gradient, the image is usually preprocessed, such as Gaussian filtering or median filtering, to remove noise from the image.These preprocessing steps can reduce the impact of noise on gradient calculation and improve the clarity and accuracy of the gradient map. After applying the gradient formula, the edge information in the image becomes clearer and more prominent. The edge contour is clearly discernible, providing a good foundation for subsequent contour extraction, shape analysis and other tasks. Feature extraction based on gradient maps can significantly improve the efficiency of feature extraction. Because the gradient map already contains most of the important information in the image, feature extraction can be performed directly on the gradient map without complicated processing of the original image. The extracted features have stronger descriptive capabilities. Since the gradient map reflects the rate of change of pixel values ​​in the image, the features extracted based on the gradient map can more accurately describe the objects and scenes in the image, providing better support for subsequent classification, recognition and other tasks.

[0086] In one embodiment of the present invention, S3 includes:

[0087] S31. Using an automated control system, automatically configure multiple customized experimental platforms according to preset vibration test parameters (such as frequency, amplitude, and duration);

[0088] S32, using an automated fixture system to accurately place and fix the sample to be tested on the experimental platform;

[0089] S33, sending instructions through the central control system, the random vibration generators of all experimental platforms are started simultaneously, and vibration tests are performed according to preset parameters;

[0090] S34, the image acquisition device is automatically activated when the vibration test is started, and uses the preset tracking algorithm to lock the key area of ​​the sample to perform continuous and high-precision image capture;

[0091] S35. The sensor network automatically collects mechanical data (such as acceleration, displacement, stress, etc.), and sends it to the data center in real time through wireless transmission technology, and pre-processes the data.

[0092] The working principle of the above technical solution is as follows: the system automatically configures multiple customized experimental platforms according to preset vibration test parameters (such as frequency, amplitude, and duration). These parameters are set in advance according to the characteristics of the sample to be tested and the test requirements to ensure that the test can accurately simulate the actual working conditions; the experimental platform is configured according to the instructions of the control system, including adjusting the parameters of the vibration generator, setting the experimental environment, etc., to meet the test requirements of different samples; the system can accurately place and fix the sample to be tested on the experimental platform. The design of the fixture takes into account factors such as the shape, size, and weight of the sample to ensure that the sample will not shift or fall off due to vibration during the test; through automated control, the fixture can quickly and accurately position the sample to the predetermined position, providing a stable foundation for subsequent tests; the system is responsible for sending instructions to control the random vibration generators of all experimental platforms to start at the same time. This step ensures that all samples are tested under the same conditions, improving the reliability and comparability of the test results; the vibration generator performs vibration tests according to preset parameters (such as frequency, amplitude, and duration) to simulate the vibration environment in actual working conditions; when the vibration test is started, the image acquisition device is automatically activated. These devices use high-resolution cameras and advanced image processing technology to capture subtle changes in samples; preset tracking algorithms can lock key areas of samples and perform continuous, high-precision image capture. This step helps to observe dynamic changes such as deformation and damage of samples during vibration; the sensor network automatically collects mechanical data such as acceleration, displacement, stress, etc. These sensors are distributed at different locations on the experimental platform and can fully monitor the vibration response of the samples; the collected data is sent to the data center in real time via wireless transmission technology. This step reduces delays and interference during data transmission and improves the efficiency and reliability of data transmission; the data center pre-processes the received data, including data cleaning, denoising, calibration and other steps to ensure the quality of data used in subsequent analysis.

[0093] The effect of the above technical solution is: the automation of the vibration test process is realized by using the automatic control system and the central control system, which significantly improves the test efficiency. The configuration of multiple experimental platforms, sample placement and fixation, vibration test startup and other steps can be completed without manual intervention, which greatly saves manpower and time costs. At the same time, the random vibration generators of all experimental platforms can be started at the same time and tested according to the preset parameters, ensuring the consistency and efficiency of the test. The automated fixture system can accurately place and fix the tested samples on the experimental platform, effectively avoiding the test errors caused by sample position offset or looseness. This improves the accuracy and reliability of the test and makes the test results more accurate and reliable. The image acquisition device uses a preset tracking algorithm to lock the key area of ​​the sample and perform continuous and high-precision image capture. This step can observe the dynamic changes of the sample during the vibration process in real time, providing rich visual information for subsequent analysis and evaluation. The sensor network can automatically collect mechanical data (such as acceleration, displacement, stress, etc.) and send it to the data center in real time through wireless transmission technology. This step realizes the real-time collection and transmission of data, reduces the delay and interference of data transmission, and improves the real-time and accuracy of data. The data center preprocesses the received data, including data cleaning, denoising, calibration and other steps, to ensure the quality of the data used in subsequent analysis. This helps to reduce errors and uncertainties in the analysis process and improve the accuracy and reliability of the analysis results. The technical solution supports the configuration of multiple customized experimental platforms and can be flexibly adjusted according to different test requirements. This enables the technical solution to adapt to the test requirements of samples of different types and specifications and has a wide range of applicability. At the same time, the technical solution also has good scalability. With the continuous increase in test requirements and the continuous development of test technology, the technical solution can be expanded and upgraded by adding experimental platforms, upgrading control systems, etc. to meet higher test requirements. Through the comprehensive application of various technical means such as automated control systems, image acquisition equipment, sensor networks, etc., the technical solution can comprehensively and accurately obtain various information of the tested samples during the vibration process. This information not only helps to evaluate the vibration performance of the sample, but also provides strong support for the design, improvement and optimization of the product.

[0094] In one embodiment of the present invention, the S4 includes:

[0095] S41, selecting a multi-channel transmission protocol, wherein the multi-channel transmission protocol includes a reliable transmission channel based on TCP, an unreliable but high-speed transmission channel based on UDP, and a hybrid channel (such as a combination of TCP+UDP); performing a performance evaluation on each transmission channel based on an existing performance evaluation model, including key indicators such as bandwidth, delay, and packet loss rate,

[0096] S42. Deploy monitoring points at key nodes and transmission paths of the network to collect real-time network data, such as traffic, delay, packet loss rate, etc. Process the collected network data through machine learning algorithms to calculate real-time network status indicators; wherein the real-time network status indicator NSI is obtained by the following formula:

[0097]

[0098] Where T represents the time interval, F(T) represents the total traffic within the time interval T, L(T) represents the average delay within the time interval T, P(T) represents the packet loss rate within the time interval T, and B represents the total bandwidth of the network.

[0099] S43, predicting the network status change in the future period of time through a network status prediction model based on historical data, and formulating a transmission strategy according to transmission requirements;

[0100] S44, through the channel selection algorithm based on multi-objective optimization, the best channel is selected from multiple candidate channels to transmit the image and mechanical data to the cloud platform;

[0101] S45. During the transmission process, when the network status changes, the dynamic adjustment mechanism is triggered to evaluate and reselect the transmission channel in real time.

[0102] The working principle of the above technical solution is as follows: according to the specific needs of data transmission, select multi-channel transmission protocols including TCP, UDP and mixed channels (TCP+UDP). Each protocol has its own characteristics, such as TCP provides reliable but possibly slow transmission, and UDP provides unreliable but high-speed transmission; use the existing performance evaluation model to evaluate key indicators such as bandwidth, delay, and packet loss rate for each transmission channel. These evaluation results will serve as the basis for the formulation of subsequent transmission strategies; deploy monitoring points at key nodes and transmission paths of the network to collect real-time network data, such as traffic, delay, and packet loss rate; process the collected network data through machine learning algorithms to calculate real-time network status indicators, such as current bandwidth utilization, average delay, and packet loss rate trend. These indicators reflect the real-time status of the network and provide a basis for the adjustment of subsequent transmission strategies; use the network status prediction model based on historical data to predict changes in network status in the future. This helps to understand the network status in advance and reserve time for the adjustment of transmission strategies; formulate transmission strategies based on the prediction results and transmission requirements (including real-time network status, real-time requirements and accuracy requirements of data types). For example, in a network with high latency but low packet loss rate, TCP is used first to transmit mechanical data with high accuracy requirements; in a network with low latency but high packet loss rate, UDP is used to transmit image data with high real-time requirements but can tolerate certain packet loss, supplemented by a packet loss retransmission mechanism to improve data integrity; through a channel selection algorithm based on multi-objective optimization, multiple factors such as transmission efficiency, data integrity and real-time are comprehensively considered to select the best channel from multiple candidate channels for data transmission. This step is designed to ensure the efficiency and stability of data transmission; during the data transmission process, the changes in network status are continuously monitored; when the network status changes, the dynamic adjustment mechanism is triggered. This mechanism will evaluate the current network status in real time and reselect the transmission channel based on the new evaluation results. This step ensures that data transmission can quickly adapt to changes in network conditions and maintain efficiency and stability.

[0103] The effect of the above technical solution is: by selecting a multi-channel transmission protocol (TCP, UDP and mixed channels), and dynamically adjusting the transmission strategy according to the real-time network status and data type requirements, the data transmission process is made more flexible and adaptable. This flexibility can cope with various complex and changeable network environments and ensure the smooth progress of data transmission. By deploying monitoring points on key network nodes and transmission paths, and collecting and processing network data in real time, the current network status indicators such as bandwidth utilization, delay and packet loss rate trends can be accurately calculated. This provides a scientific basis for formulating efficient transmission strategies and helps to improve the efficiency and stability of data transmission. At the same time, the channel selection algorithm based on multi-objective optimization can comprehensively consider multiple factors such as transmission efficiency, data integrity and real-time, and select the best channel from multiple candidate channels for data transmission. This further ensures the efficiency and stability of data transmission. The above technical solution can flexibly select transmission protocols and formulate transmission strategies according to the real-time requirements and accuracy requirements of data types. For example, in a network with high latency but low packet loss rate, TCP is preferred to transmit mechanical data with high accuracy requirements; in a network with low latency but high packet loss rate, UDP is tried to transmit image data with high real-time requirements but a certain amount of packet loss can be tolerated, supplemented by a packet loss retransmission mechanism. This strategy helps to strike a balance between data integrity and real-time performance, and meets the data transmission requirements in different application scenarios. Through the network status prediction model based on historical data, the changes in network status in the future can be predicted, providing a basis for adjusting the transmission strategy. This intelligent prediction capability helps to plan data transmission plans in advance and reduce the risk of transmission interruption or data loss caused by sudden changes in network status. At the same time, during the transmission process, when the network status changes, the dynamic adjustment mechanism can be triggered to evaluate and reselect the transmission channel in real time. This dynamic adjustment capability ensures that data transmission can quickly adapt to changes in network conditions and maintain efficiency and stability. The above formula comprehensively considers the four key indicators of traffic, latency, packet loss rate and total network bandwidth, and can comprehensively and accurately reflect the performance of the network in a specific time interval. By adjusting the time interval T, it is possible to flexibly adapt to the network performance evaluation needs of different time scales, and both rapid changes in the short term and stable trends in the long term can be effectively reflected. Real-time network status indicators provide an important basis for the formulation and optimization of transmission strategies. By real-time monitoring and evaluation of network status, transmission strategies can be adjusted in a timely manner to ensure the real-time, reliability and efficiency of data transmission. When the network status is good, by selecting a high-bandwidth, low-latency transmission channel, the efficiency of data transmission can be significantly improved, the transmission time can be reduced, and the user experience can be improved. When the network status is poor, by selecting a transmission channel with a packet loss retransmission mechanism or adjusting the transmission strategy (such as adding a checksum, using a more reliable transmission protocol, etc.), the reliability of data transmission can be enhanced and the risk of data loss can be reduced.Real-time network status indicators help to allocate network resources reasonably. When the network is congested, the transmission of key data can be prioritized; when the network is idle, network resources can be fully utilized for operations such as batch data transmission or backup. When the network status changes, the real-time network status indicators can quickly reflect it and trigger a dynamic adjustment mechanism. This helps to adjust the transmission strategy in time to adapt to the new network environment and ensure the continuity and stability of data transmission. Combined with machine learning algorithms and network status prediction models, the application of real-time network status indicators can further enhance the intelligence of the system. By learning and predicting the changing trends of network status, the system can more actively adjust the transmission strategy to achieve more efficient and reliable data transmission.

[0104] In one embodiment of the present invention, S5 includes:

[0105] S51, after receiving the image and mechanical data, the cloud platform performs image matching through a deep learning algorithm (such as a convolutional neural network CNN);

[0106] S52, performing fine processing on the differential image, and combining mechanical data, performing mechanical effect analysis on the changed area in the differential image, and identifying specific damage types (such as fatigue cracks, looseness, etc.) caused by vibration;

[0107] S53. Cross-compare the test results of different experimental platforms to identify common and different problems; conduct a comprehensive evaluation of the samples through a comprehensive evaluation model based on machine learning;

[0108] S54, and visualize the data analysis results in the form of charts, heat maps, etc.

[0109] The working principle of the above technical solution is as follows: when the cloud platform receives image data from different experimental platforms, it first uses a deep learning algorithm (such as a convolutional neural network CNN) to perform image matching. The purpose of this step is to ensure that images collected at different time points or under different experimental conditions can be spatially aligned for subsequent differential analysis and change detection. By performing differential operations on the matched images, a differential image reflecting the changes in the sample during the vibration test is obtained. Next, the differential image is refined, including using morphological operations (such as corrosion, expansion, opening operations, closing operations, etc.) to remove noise, enhance the contrast of the change area, etc., to improve the accuracy of subsequent analysis. Combined with mechanical data (such as acceleration, displacement, stress, etc.), the mechanical effect analysis of the change area in the differential image is performed. By analyzing the correlation between mechanical data and image changes, the specific damage types caused by vibration, such as fatigue cracks, looseness, etc., are identified. The test results of different experimental platforms are cross-compared to identify common problems (i.e., damage or changes observed on multiple experimental platforms) and differential problems (abnormal conditions that only occur on individual experimental platforms). Through a comprehensive evaluation model based on machine learning, the samples are comprehensively evaluated by considering multiple evaluation indicators (such as damage degree, changes in mechanical properties, changes in appearance, etc.). The model can comprehensively consider various factors and give more comprehensive and accurate evaluation results. The data analysis results are visualized in the form of charts, heat maps, etc. These visualization tools can intuitively display the changes of samples during vibration testing, damage distribution, and comprehensive evaluation results, which helps researchers quickly understand the data, discover patterns, and formulate improvement measures.

[0110] The effect of the above technical solution is: by using deep learning algorithms (such as CNN) for image matching, the key features in the image can be automatically identified and aligned, which reduces human intervention and errors and improves the accuracy of image analysis. The refinement of differential images and mechanical effect analysis combined with mechanical data can more accurately identify specific damage types caused by vibration, such as fatigue cracks, looseness, etc. This comprehensive analysis method is more reliable and accurate than the analysis of a single data source. Cross-comparison of test results from different experimental platforms can identify common problems and differences, which helps to understand the response characteristics and damage mechanisms of samples under different experimental conditions. Through cross-comparison, the consistency and reliability of different experimental platforms can be verified, providing stronger support for subsequent testing and analysis. The comprehensive evaluation model based on machine learning can consider multiple evaluation indicators (such as damage degree, changes in mechanical properties, changes in appearance, etc.) to comprehensively evaluate the samples. This comprehensive evaluation method is more comprehensive and objective than single indicator evaluation, and can more accurately reflect the actual condition of the samples. Visualizing the data analysis results in the form of charts, heat maps, etc. can intuitively display the changes in the samples during the vibration test, the damage distribution, and the comprehensive evaluation results. This visual display method helps researchers quickly understand the data, discover patterns, and develop improvement measures. Through the above technical solutions, researchers can quickly obtain accurate damage detection results and comprehensive evaluation results to provide support for subsequent decision-making. This helps improve decision-making efficiency and accuracy and reduce decision-making errors caused by insufficient or incorrect information. The visual display of data analysis results can also promote knowledge sharing and collaboration. Researchers can communicate research results, discuss problems, and develop solutions by sharing visualization results, thereby promoting progress and development in the entire field.

[0111] In one embodiment of the present invention, S6 includes:

[0112] S61. Based on the comprehensive analysis results, the eligibility of the samples is automatically judged through the intelligent evaluation model, and the samples are divided into qualified and unqualified categories;

[0113] S62. Further classify the unqualified samples to identify the failure modes (such as fracture, wear and corrosion) and potential problem areas;

[0114] S63, generating optimization suggestions according to the failure mode and the problem area, wherein the optimization suggestions include design improvement, material selection, and process optimization;

[0115] S64. Set up an early warning system to promptly issue early warning signals for samples with major risks or potential failures, trigger emergency response mechanisms, and notify relevant personnel;

[0116] S65. Store the data, evaluation results and optimization suggestions of each test into the knowledge base, and continuously improve the test case library and evaluation model.

[0117] The working principle of the above technical solution is as follows: First, based on the comprehensive analysis results obtained in step S5, including data on multiple evaluation indicators such as damage degree, changes in mechanical properties, and changes in appearance; use a pre-trained intelligent evaluation model (possibly based on machine learning or deep learning algorithms) to automatically judge the eligibility of the sample. The model will take into account all relevant factors and divide the samples into two categories: qualified and unqualified according to the preset qualification standards; for samples judged to be unqualified, the intelligent evaluation model will also provide a specific description and location of the unqualified items to help users quickly locate the problem; further analyze the unqualified samples to identify specific failure modes, such as fracture, wear, corrosion, etc. This usually requires a combination of sample images, mechanical data, and professional knowledge; at the same time, potential problem areas that lead to failure are identified to provide a basis for subsequent optimization suggestions; based on failure modes and problem areas, design improvement suggestions are proposed, such as changing structural strength, optimizing material layout, etc.; if the problem is related to material performance, it is recommended to replace more suitable materials or adopt new material technology; if the problem is related to the production process, process optimization suggestions are proposed, such as improving the processing flow, adjusting process parameters, etc.; risk assessment is conducted on samples with major risks or potential failures to determine their urgency; early warning signals are issued in a timely manner to notify relevant personnel through emails, text messages, system notifications, etc.; emergency response mechanisms are triggered to ensure that relevant personnel can take prompt measures to prevent the problem from further expanding; the data, evaluation results, and optimization suggestions of each test are stored in the knowledge base to form a test case library; the evaluation model is continuously trained and optimized using new data and cases to improve its accuracy and generalization ability; the knowledge base can also serve as an internal knowledge sharing platform for the enterprise to promote communication and collaboration between different departments.

[0118] The effect of the above technical solution is: the qualification of samples is automatically judged through the intelligent evaluation model, which greatly reduces the time and cost of manual evaluation, and improves the accuracy and consistency of the evaluation. The automated evaluation process reduces human errors and makes the evaluation results more reliable. Further classification of unqualified samples, identification of specific failure modes (such as fracture, wear, corrosion) and potential problem areas, helps to deeply understand the failure mechanism of the product. This in-depth analysis provides precise goals and directions for subsequent improvements and optimizations. According to the failure mode and problem area, specific design improvement, material selection and process optimization suggestions are generated. These suggestions are targeted and operational, and can directly guide product improvement and upgrading. The generation of optimization suggestions helps enterprises to respond quickly to market changes and enhance product competitiveness. Set up an early warning system to monitor and warn samples with major risks or potential failures in real time. This helps enterprises to detect potential problems in a timely manner and avoid major accidents. The setting of the early warning system enhances the risk management capabilities of enterprises and ensures the safe operation of enterprises. Storing the data, evaluation results and optimization suggestions of each test in the knowledge base helps enterprises accumulate valuable testing experience and evaluation knowledge. The establishment of the knowledge base promotes the sharing and inheritance of knowledge, allowing new employees to quickly master relevant skills and methods. By continuously improving the test case library and evaluation model, the accuracy and adaptability of the evaluation model are improved. Continuous improvement and iteration enable the evaluation model to keep up with the development of products and technologies, providing more effective support for enterprises. By timely discovering and solving potential problems in products, the quality and reliability of products are improved, thereby improving customer satisfaction. Improved customer satisfaction helps to enhance the brand image and market share of the company.

[0119] One embodiment of the present invention, as Figure 2 As shown, a random vibration test system, the system comprising:

[0120] Image acquisition module: The image acquisition device is used to collect appearance images of multiple samples to be tested, and the samples to be tested are numbered. The collected appearance images with different numbers are stitched into a complete product picture through the image stitching algorithm;

[0121] Data storage module: Use image processing algorithms to preliminarily process product images, extract key feature information on the sample surface, and store them in feature libraries;

[0122] Vibration test module: Multiple samples to be tested are placed on different experimental platforms for random vibration testing. Mechanical data is collected in real time through sensors. Meanwhile, the image acquisition device works continuously to capture subtle changes on the sample surface in real time.

[0123] Data transmission module: The collected images and mechanical data are transmitted to the cloud platform in parallel through multiple paths through a multi-channel transmission protocol;

[0124] Comparative analysis module: After receiving the image and mechanical data, the cloud platform compares the sample image after the test with the feature information of the tested sample stored in the feature library at the pixel level through image matching and differential algorithms to identify the appearance changes caused by vibration; and obtains comprehensive analysis results through cross-analysis of test results from different experimental platforms;

[0125] Recommendation and warning module: Based on the comprehensive analysis results, the intelligent evaluation model is used to automatically determine whether the sample is qualified or not, and generate targeted optimization suggestions or warning signals based on the specific circumstances of the appearance changes.

[0126] The working principle of the above technical solution is: use high-precision image acquisition equipment (such as high-definition cameras) to take appearance photos of multiple samples to be tested. Each sample is assigned a unique number for easy identification and management. Since the product may have a large size or complex shape, it is necessary to use an image stitching algorithm to stitch multiple local images into a complete, high-resolution product picture; the stitched product picture is preliminarily processed by an image processing algorithm (such as edge detection, feature extraction, etc.) to extract key feature information on the sample surface, such as texture, color, shape, etc. These feature information are crucial for subsequent vibration impact analysis, so they will be accurately extracted and stored in the feature library for later use; the tested samples are placed on different designed experimental platforms, which can simulate various random vibration environments. Vibration stimulation is applied to the sample through a vibration generator to simulate the vibration conditions that may be encountered during actual use or transportation; during the vibration test, sensors (such as accelerometers, displacement sensors, etc.) are used to collect mechanical data of the sample in real time, such as acceleration, displacement, stress, etc. At the same time, the image acquisition device also continues to work to capture subtle changes in the sample surface in real time, such as cracks, deformation, color changes, etc. These data provide an important basis for analyzing the impact of vibration on the appearance of samples. In order to ensure the real-time and accuracy of the data, a multi-channel transmission protocol is used to transmit the collected images and mechanical data to the cloud platform in parallel through multiple paths. This parallel transmission method can significantly improve the efficiency and reliability of data transmission. As the center of data processing and analysis, the cloud platform is responsible for receiving data from various experimental platforms. The cloud platform has powerful data storage and processing capabilities and can process a large amount of data at the same time, providing strong support for subsequent analysis and evaluation. After receiving the sample image after the test, the cloud platform uses image matching and differential algorithms to compare it with the original sample image stored in the feature library at the pixel level. This comparison can accurately identify the appearance changes caused by vibration, such as crack extension and deformation aggravation. In order to obtain more comprehensive analysis results, the cloud platform will also cross-analyze the test results of different experimental platforms. By comparing the response of samples under different vibration conditions, the intrinsic connection between vibration characteristics and appearance changes can be revealed, providing a strong basis for subsequent evaluation and optimization. Based on the comprehensive analysis results, the cloud platform uses an intelligent evaluation model to automatically judge the eligibility of the sample. The intelligent evaluation model combines multiple technologies such as machine learning and expert systems to accurately evaluate whether the performance of the sample in the vibration test meets the preset standards; based on the evaluation results and the specific circumstances of the appearance changes, the cloud platform will generate targeted optimization suggestions or early warning signals. These suggestions may involve sample design improvements, manufacturing process optimization, etc.; and early warning signals are used to promptly remind relevant personnel to pay attention to potential problems and risks.

[0127] The effect of the above technical solution is: multiple local images are stitched into a complete product picture through an image stitching algorithm, avoiding the problem of image loss caused by complex sample size or shape, thereby improving the accuracy of the test. Real-time capture of subtle changes in the sample surface, combined with real-time collection of mechanical data, can more comprehensively reflect the performance of the sample in a vibration environment, improving the comprehensiveness and accuracy of the test. Using image matching and differential algorithms for pixel-level comparison, it is possible to accurately identify appearance changes caused by vibration, such as cracks, deformation, etc., providing reliable data support for subsequent evaluation and optimization. Through cross-analysis of test results on different experimental platforms, the inherent connection between vibration characteristics and appearance changes can be revealed, which helps to discover potential problems and risks and improve the depth and breadth of data analysis. The intelligent evaluation model automatically determines whether the sample is qualified or not, reduces manual intervention, and improves the efficiency and objectivity of the evaluation. Generate targeted optimization suggestions or early warning signals based on the specific circumstances of the appearance changes, providing a clear guiding direction for subsequent improvements and optimizations. The tested samples are numbered and their key feature information is stored to ensure the traceability of the test process. When needed, the test data and results of each sample can be quickly accessed, which is convenient for problem tracking and cause analysis. Data is transmitted to the cloud platform in parallel through a multi-channel transmission protocol, realizing centralized storage and management of data. As the center of data processing and analysis, the cloud platform can easily share data and collaborate with other systems or teams, improving the overall efficiency of testing work. This technical solution is not only used for current test evaluation, but also provides a basis for subsequent continuous optimization and improvement. By analyzing test data and results, we can continuously optimize test methods, improve sample design or manufacturing processes, and improve product reliability and durability.

[0128] An embodiment of the present invention is a non-transitory computer-readable storage medium, on which a computer program is stored. The program is executed by a processor to implement any of the random vibration testing methods described above.

[0129] An embodiment of the present invention is a computer program product, comprising a computer program / instruction, which implements any of the above-mentioned random vibration testing methods when executed by a processor.

[0130] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A random vibration test method, characterized in that: The method comprises: S1. Use an image acquisition device to collect appearance images of multiple samples to be tested, and number the samples to be tested. Use an image stitching algorithm to stitch the collected appearance images with different numbers into a complete product picture; S2. Perform preliminary processing on product images through image processing algorithms, extract key feature information of sample surfaces, and store them in feature libraries respectively; S3, placing multiple samples on different experimental platforms for random vibration testing, collecting mechanical data in real time through sensors, while the image acquisition device works continuously to capture subtle changes on the sample surface in real time; S4, transmitting the collected images and mechanical data to the cloud platform in parallel through multiple paths through a multi-channel transmission protocol; S5. After receiving the images and mechanical data, the cloud platform conducts cross-analysis on the test results of different experimental platforms to obtain comprehensive analysis results; S6. Based on the comprehensive analysis results, the intelligent evaluation model is used to automatically determine whether the sample is qualified or not, and generate targeted optimization suggestions or early warning signals according to the specific circumstances of the appearance changes; The S2 comprises: S21, removing noise from the image by Gaussian filtering, adjusting the contrast of the image by an adaptive contrast enhancement algorithm; and unifying the image parameters to a preset standard; S22, performing gradient calculation on the preprocessed image to obtain gradient maps of the image in the horizontal and vertical directions; based on the gradient map, applying an edge detection algorithm to extract edge information in the image; S23, extracting texture features from the image after edge detection using a gray level co-occurrence matrix; extracting shape features based on the edge detection result; S24, using principal component analysis to evaluate the extracted features, identifying the features that contribute to the classification or recognition task, and selecting a feature subset based on the evaluation results; and normalizing and quantizing the selected features; S25, combining the optimized features into a feature vector, and associating the feature vector of each sample with its unique identification code, and storing them in a feature library.

2. A random vibration testing method according to claim 1, characterized in that: Said S1 comprises: S11, inspecting all the tested samples, and acquiring multi-angle appearance images of the multiple tested samples after inspection by an image acquisition device; S12, after the appearance is collected, each sample is uniquely numbered and identified with an RFID tag; S13, applying a feature-based image stitching algorithm to align and stitch images taken from multiple angles; S14. Perform detail enhancement processing on the spliced ​​image, and separate key components and areas in the product image through an image segmentation algorithm.

3. A random vibration testing method according to claim 1, characterized in that: The S3 includes: S31. Using an automated control system, multiple customized experimental platforms are automatically configured according to preset vibration test parameters; S32, use the automated fixture system to place and fix the sample to be tested on the experimental platform; S33, sending instructions through the central control system, the random vibration generators of all experimental platforms are started simultaneously, and vibration tests are performed according to preset parameters; S34, the image acquisition device is automatically activated when the vibration test is started, and the key area of ​​the sample is locked using a preset tracking algorithm to capture the image; S35. The sensor network automatically collects mechanical data and sends it to the data center in real time through wireless transmission technology, and pre-processes the data.

4. A random vibration testing method according to claim 1, characterized in that: The S4 comprises: S41, selecting a multi-channel transmission protocol, and performing performance evaluation on each transmission channel based on a performance evaluation model; S42, deploy monitoring points at key nodes and transmission paths of the network, collect real-time network data, process the collected network data through machine learning algorithms, and calculate real-time network status indicators; S43, predicting the network status change in the future period of time through a network status prediction model based on historical data, and formulating a transmission strategy according to transmission requirements; S44, through the channel selection algorithm based on multi-objective optimization, the best channel is selected from multiple candidate channels to transmit the image and mechanical data to the cloud platform; S45. During the transmission process, when the network status changes, the dynamic adjustment mechanism is triggered to evaluate and reselect the transmission channel in real time.

5. A random vibration testing method according to claim 1, characterized in that: The S5 comprises: S51, after receiving the image and mechanical data, the cloud platform performs image matching through a deep learning algorithm; S52, processing the differential image, and combining it with the mechanical data, performing mechanical effect analysis on the changed area in the differential image to identify the specific damage type caused by the vibration; S53. Cross-compare the test results of different experimental platforms to identify common and different problems; conduct a comprehensive evaluation of the samples through a comprehensive evaluation model based on machine learning; S54, and visualize the data analysis results.

6. A random vibration testing method according to claim 1, characterized in that: The S6 comprises: S61. Based on the comprehensive analysis results, the eligibility of the samples is automatically judged through the intelligent evaluation model, and the samples are divided into qualified and unqualified categories; S62. Further classify the unqualified samples to identify the failure modes and potential problem areas; S63, generating optimization suggestions according to the failure mode and the problem area, wherein the optimization suggestions include design improvement, material selection, and process optimization; S64. Set up an early warning system to promptly issue early warning signals for samples with risks or potential failures, trigger emergency response mechanisms, and notify relevant personnel; S65. Store the data, evaluation results and optimization suggestions of each test into the knowledge base, and continuously improve the test case library and evaluation model.

7. A system for implementing the random vibration testing method according to claim 1, characterized in that: The system comprises: Image acquisition module: The image acquisition device is used to collect appearance images of multiple samples to be tested, and the samples to be tested are numbered. The collected appearance images with different numbers are stitched into a complete product picture through the image stitching algorithm; Data storage module: Use image processing algorithms to preliminarily process product images, extract key feature information on the sample surface, and store them in feature libraries; Vibration test module: Multiple samples to be tested are placed on different experimental platforms for random vibration testing. Mechanical data is collected in real time through sensors. Meanwhile, the image acquisition device works continuously to capture subtle changes on the sample surface in real time. Data transmission module: The collected images and mechanical data are transmitted to the cloud platform in parallel through multiple paths through a multi-channel transmission protocol; Comparative analysis module: After receiving the images and mechanical data, the cloud platform conducts cross-analysis on the test results of different experimental platforms to obtain comprehensive analysis results; Recommendation and warning module: Based on the comprehensive analysis results, the intelligent evaluation model is used to automatically determine whether the sample is qualified or not, and generate targeted optimization suggestions or warning signals based on the specific circumstances of the appearance changes.

8. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: The program is executed by a processor to implement the random vibration testing method as described in any one of claims 1-6.

9. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instructions are executed by a processor, the random vibration testing method as described in any one of claims 1 to 6 is implemented.

Images