A multi-system coordinated low-magnification test system and method

Abstract

This invention relates to the field of quality inspection technology, specifically to a multi-system collaborative low-magnification testing system and method. The system includes a central intelligent scheduling subsystem configured to coordinate the operation of the entire system; a sample pretreatment and identification domain configured to identify and archive information from input samples; an intelligent acid etching process control domain configured to perform acid etching treatment and process monitoring on the samples; an adaptive post-processing and quality closed-loop domain configured to perform post-processing and quality testing on the etched samples, and to control rework when tests fail; and a result evaluation and knowledge management domain configured to perform image evaluation on the processed samples and optimize the system's process knowledge based on the evaluation results. The central intelligent scheduling subsystem is connected to the sample pretreatment and identification domain, the intelligent acid etching process control domain, the adaptive post-processing and quality closed-loop domain, and the result evaluation and knowledge management domain, respectively. This technical solution can improve the efficiency and accuracy of metallurgical quality inspection of metallic materials.

Description

Technical Field

[0001] This invention relates to the field of quality inspection technology, specifically to a multi-system collaborative low-magnification testing system and method. Background Technology

[0002] Low-magnification acid etching is a core inspection method for evaluating the metallurgical quality, internal defects, and surface condition of steel, alloys, and other metallic materials. It is widely used in the fields of steel, aerospace, and special metal processing. This test exposes macroscopic defects (such as segregation, inclusions, and cracks) in materials through selective corrosion of the sample surface by acid. The accuracy of the test results directly affects the material grade determination and the formulation of subsequent processing technology.

[0003] In the current technological environment, to improve inspection efficiency and consistency, there are mainly two technological evolution trends in the industry: The first trend is the widespread adoption of automated equipment for single processes. For example, automatic pickling machines, automatic cleaning and drying lines, and robotic arm sampling devices have gradually replaced traditional manual operations. These devices achieve automated operation at specific workstations (such as sample introduction, acid etching, cleaning, and imaging), significantly reducing the intensity of manual labor and improving the standardization of single-point operations.

[0004] The second trend is localized improvements targeting specific pain points. To improve the working environment, environmental patents have emerged specifically for acid mist collection and treatment; to monitor corrosion conditions, static image acquisition devices based on machine vision have appeared; and to improve detection accuracy, defect recognition algorithms based on deep learning have emerged.

[0005] However, despite the progress made in certain aspects of the above technologies, existing low-magnification acid etching test systems still have some problems, mainly in the following aspects: Existing automated equipment is mostly composed of independent functional units, lacking a unified central control system to connect sample introduction systems, etching tanks, cleaning lines, and testing instruments. Each unit operates based on its own pre-set logic, and the connection between processes relies on physical transmission or simple signal triggers, rather than dynamic coordination based on real-time data. When faced with mixed production lines of multiple batches and specifications of samples, the system cannot perform global task scheduling and optimal resource allocation based on the real-time load of each station, equipment health status, and sample characteristics, resulting in low overall workflow efficiency and difficulty in responding to sudden incoming material anomalies or equipment failures.

[0006] The control logic of most current mainstream equipment is based on a hard-coded sequential execution mode, meaning that process parameters (such as etching time, temperature, and concentration) are fixed before the test begins. Although some equipment incorporates visual monitoring, it is mostly used for post-processing recording rather than real-time feedback control. In actual corrosion processes, due to slight fluctuations in the composition of the metal material, differences in the initial surface state, and the dynamic decay of acid effectiveness, the actual corrosion rate generally deviates from the theoretical model. Because the system lacks real-time perception and closed-loop feedback capabilities for the corrosion process, it cannot dynamically correct the process path based on real-time image feature values ​​or environmental parameters, making it prone to under-corrosion (defects not exposed) or over-corrosion (blurred surface), leading to unstable and unreliable test results.

[0007] Each low-magnification test generates massive amounts of process data (including sample 3D dimensions, material composition, environmental parameters during corrosion, real-time image sequences, and final rating results). However, in existing systems, this data is typically scattered across different subsystems or devices, forming "data silos." The system lacks the ability to deeply mine and learn the relationships between process parameters, intermediate states, and final results, and cannot automatically extract the optimal combination of process parameters for samples of different materials and specifications from historical data. Updating the process knowledge base relies primarily on summaries of human experience, which is inefficient and difficult to quantify, preventing the system from becoming "smarter" with increasing experiment counts, resulting in a situation where there is data but no wisdom. Summary of the Invention

[0008] The purpose of this invention is to propose a multi-system collaborative low-magnification testing system and method, which can improve the efficiency and accuracy of metallurgical quality inspection of metallic materials.

[0009] To achieve the above objectives, in a first aspect, the present invention proposes a multi-system collaborative low-magnification testing system and method, comprising: The central intelligent scheduling subsystem includes a digital twin model for virtual mapping of the physical system, a dynamic process path planner for generating a baseline path based on sample information and generating the actual execution path based on real-time feedback, and a multi-objective optimizer for batch task scheduling and resource allocation based on optimization objectives. The sample pretreatment and identification domain includes an information acquisition module for acquiring sample ID, three-dimensional dimensions, surface morphology, and material composition, as well as a report generation module for generating a sample digital passport containing sample material, dimensions, and initial surface condition. The intelligent acid etching process control domain includes an acid mist monitoring module for monitoring the environment inside the etching tank, an acid liquid management module for managing acid liquid efficiency, and an etching process control module for dynamically adjusting process parameters based on the comparison results of real-time image feature values ​​and expected models. The adaptive post-processing and quality closed-loop domain includes a cleaning and drying line for cleaning and drying samples, a quality detection point for detecting post-processing quality, and a rework traceability decision-maker for tracing the source of non-conformity based on the non-conformity signal and instructing targeted rework. The results evaluation and knowledge management domain includes a standardized imaging module for acquiring standardized images, an analysis and preliminary evaluation module for defect identification and grade assessment of images, and a process knowledge self-evolution engine for linking and mining data from the entire process to optimize the process knowledge base.

[0010] Beneficial effects of the basic solution: This technical solution achieves full-chain collaborative control of sample pretreatment, acid etching process, post-treatment and result evaluation, combined with real-time environmental monitoring, image feature comparison and digital passport traceability, effectively eliminating inspection deviations caused by manual operation, process fluctuations and sample differences, realizing the standardization and stabilization of low-magnification test results, and ensuring inspection quality from the source.

[0011] Based on digital twin virtual mapping and dynamic process path planning, the system can automatically generate an appropriate execution path according to information such as sample ID, size, and material. Combined with a multi-objective optimizer, it can complete batch task scheduling and resource allocation without the need for frequent manual program switching. It can quickly adapt to the differentiated testing needs of multiple varieties and specifications of samples, and significantly improve the flexibility of the test production line and the overall utilization rate of equipment.

[0012] By monitoring acid mist, managing acid performance, and comparing real-time images of the corrosion process, the corrosion parameters can be dynamically adjusted to avoid problems such as over-corrosion and under-corrosion. At the same time, the acid usage cycle can be optimized and the risk of acid mist leakage can be reduced. This improves the test results while reducing consumable consumption and safety management pressure.

[0013] By relying on the post-processing quality inspection and rework traceability decision-making system, combined with sample digital passports, the system can quickly locate the problem links of non-conforming samples, automatically issue targeted rework instructions, avoid blindly repeating experiments, shorten the problem handling cycle, reduce ineffective waste, and form a complete quality closed loop of inspection, judgment, traceability and correction.

[0014] The process knowledge self-evolution engine can mine full-process test data, continuously iterate and optimize the process knowledge base, reduce the threshold for personnel operation and training costs, and realize the standardized and intelligent execution of high-quality test processes.

[0015] As a feasible preferred embodiment, the digital twin model includes a geometric layer and a physical layer. The geometric layer includes the kinematic parameters of the robotic arm and the spatial coordinates of the workstation. The physical layer includes an acid flow field model, a temperature field heat transfer model, and a robotic arm dynamics model, which are used for process simulation and collision detection.

[0016] As a feasible preferred embodiment, the information acquisition module includes a 2D / 3D vision unit, a barcode scanner, and a laser-induced breakdown spectroscopy or miniature X-ray fluorescence probe; the dynamic process path planner is configured to dynamically insert or skip specific process nodes in the baseline path based on visual judgment and sensor values.

[0017] As a feasible preferred solution, the acid mist monitoring module includes a high-definition camera installed above the corrosion tank, which, together with a surrounding temperature-controlled inert gas curtain and an automatically opening and closing physical sealing cover, forms a visual monitoring environment.

[0018] As a feasible preferred solution, the acid management module includes a circulation loop and an acid performance decay model. The acid performance decay model calculates the remaining acid performance based on the monitored acid density, pH value, and cumulative area of ​​the corroded sample, and controls the replenishment or replacement of acid accordingly.

[0019] As a feasible preferred embodiment, the corrosion process control module is configured to extract the corrosion front expansion rate and surface texture change feature values ​​through image algorithms, and compare these feature values ​​with the expected model of the current process stage. When the deviation exceeds the limit, the liquid flow pattern in the tank, the sample swing amplitude, or the corrosion temperature is dynamically adjusted.

[0020] As a feasible preferred embodiment, the cleaning and drying line includes pre-cleaning, main cleaning, multi-stage rinsing and hot air / vacuum drying stations; the quality inspection points include a microdroplet conductivity meter set after rinsing and a multispectral surface detector set after drying.

[0021] As a feasible preferred solution, the rework traceability decision-maker is configured to receive non-conforming signals from quality inspection points, combine the sample's digital passport and the executed process history, use a rule engine to calculate the responsibility probability of each process to determine the source of contamination, and instruct the robotic arm to send the sample back to that process for targeted rework.

[0022] As a feasible preferred solution, the analysis and preliminary evaluation module is configured to automatically identify defects and make preliminary assessments of their levels on the final low-magnification images; the process knowledge self-evolution engine is configured to perform correlation mining between the final image quality quantitative score and the complete data chain of the entire process, and discover strong correlation rules between the results and specific process parameter combinations through machine learning.

[0023] Secondly, the present invention also provides a multi-system collaborative low-magnification testing method, applied to the aforementioned multi-system collaborative low-magnification testing system, comprising: The system receives inspection task orders through a central intelligent scheduling subsystem, performs simulations using a digital twin model, formulates scheduling plans using a multi-objective optimizer, and generates baseline paths for samples using a dynamic process path planner. In the sample pretreatment and identification domain, the sample's ID, size, surface morphology, and material composition information are collected to generate a digital passport containing the sample's initial state. In the intelligent acid etching process control domain, the acid mist monitoring module monitors the corrosion environment, the acid liquid management module manages the acid liquid efficiency, and the corrosion process control module dynamically adjusts the acid etching process parameters based on the comparison results of real-time image features and expected models. The samples are cleaned and dried in the adaptive post-processing and quality closed-loop domain, and quality testing is performed. If the test fails, the source is traced and analyzed based on the digital passport and process history, and the rework is instructed to be targeted. In the result evaluation and knowledge management domain, samples are standardized for imaging and defect assessment. The assessment results are then correlated with the entire process data for analysis and updating of the process knowledge base to optimize subsequent processes. Attached Figure Description

[0024] Figure 1 This is a schematic diagram of the architecture of a low-magnification experimental system that involves multi-system collaboration. Detailed Implementation

[0025] To make the technical solution and advantages of this application clearer, the technical solution of the present invention will be further described in detail below with reference to the accompanying drawings. It is understood that the specific embodiments described herein are only some embodiments of the present invention, and are only used to explain this application, not to limit it. It should be noted that the technical features or combinations of technical features described in the following embodiments should not be considered isolated; they can be combined with each other to achieve better technical effects. The same reference numerals appearing in the accompanying drawings of the following embodiments represent the same features or components, and can be applied to different embodiments.

[0026] Furthermore, unless otherwise defined, the technical or scientific terms used in this invention description shall have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains.

[0027] The present invention will now be described in further detail with reference to the accompanying drawings.

[0028] This disclosure provides a multi-system collaborative low-magnification testing system, referring to... Figure 1 It includes a central intelligent scheduling subsystem, a sample pretreatment and identification domain, an intelligent acid etching process control domain, an adaptive post-processing and quality closed-loop domain, and a result evaluation and knowledge management domain.

[0029] The central intelligent scheduling subsystem includes a digital twin model, a dynamic process path planner, and a multi-objective optimizer.

[0030] The digital twin model provides a 1:1 virtual mapping of the physical system (each workstation, robotic arm, and fluid path) for process simulation and collision detection. The digital twin module comprises a geometric layer and a physical layer. The geometric layer includes the robotic arm's DH parameters and the workstation's spatial coordinates; the physical layer includes a CFD model of the acid flow field, a heat transfer model of the temperature field, and a robotic arm dynamics model. In this embodiment, during the process simulation phase, the digital twin model simulates the processing paths of different samples, predicts potential collision points, and adjusts the path planning in advance to avoid collisions during actual operation.

[0031] The dynamic process path planner uses initial sample information to retrieve a baseline path from the process knowledge base. At each subsequent process node, it dynamically determines the next process node and parameters based on real-time feedback data (such as visual judgment and sensor values), forming a unique execution path for that sample. For example, for a stainless steel sample with thick oxide scale, the initial baseline path is: blasting → hot acid etching process A → hot water + brushing → rinsing. After blasting, visual inspection reveals significant oxide scale residue. The dynamic process path planner immediately intervenes, generating a corrective path that inserts a second blasting step to ensure complete oxide scale removal.

[0032] After each process node is completed, a path adjustment trigger decision is made, and the specific logic is as follows: Triggering conditions include: A corrective adjustment is triggered when real-time feedback data is greater than the correction threshold and less than or equal to the scrap threshold. For example, after sandblasting, visual inspection feedback indicates that the residual oxide scale area accounts for 15% (>5% correction threshold) and the residual thickness is 0.05mm (>0.02mm correction threshold), but is less than the 30% scrap threshold, thus triggering a corrective adjustment.

[0033] When real-time feedback data is better than the acceptable threshold, and historical data supports simplifying subsequent processes, optimization adjustments are triggered. For example, if after sandblasting, the system detects "0% oxide scale residue and roughness Ra=1.2μm (better than the Ra≤2.0μm standard)," and the system determines that the subsequent acid etching process can be shortened, optimization adjustments are triggered.

[0034] Emergency termination of adjustment is triggered when real-time feedback data exceeds the scrapping threshold or a safety hazard is detected.

[0035] For example, if a penetrating scratch with a length >10mm or a sample deformation >2mm is found after sandblasting, an emergency termination is triggered.

[0036] The threshold is the core criterion for path adjustment, used to quantify whether the feedback data meets the standards. Its setting needs to be combined with process objectives, sample characteristics, equipment accuracy, and historical data. A dynamic calibration method for the base threshold is used to ensure its rationality and adaptability. The specific steps are as follows: Based on the standard process requirements in the process knowledge base, and combined with the initial information of the sample (material, specifications, initial defects, etc.), the basic thresholds of each feedback indicator are set, and the thresholds are divided into three levels: qualified threshold, correction threshold, and scrap threshold, so as to clarify the judgment boundary.

[0037] Example of a stainless steel sample with thick oxide scale: Oxide scale residue: Acceptable threshold (residual area percentage ≤ 5%, residual thickness ≤ 0.02 mm); Correction threshold (residual area percentage > 5% or residual thickness > 0.02 mm); Scrap threshold (residual area percentage > 30% or residual thickness > 0.1 mm, cannot be removed by correction process); Sandblasting pressure: acceptable threshold (0.6-0.8MPa); correction threshold (<0.5MPa or >0.9MPa); scrap threshold (<0.4MPa or >1.0MPa, which may cause sample deformation or equipment damage).

[0038] Based on real-time feedback data, historical process data, and individual sample differences, the threshold is dynamically calibrated to improve the accuracy of judgment and avoid false triggering or missed triggering.

[0039] Statistically analyze the process feedback data of similar samples in the past, calculate the mean and standard deviation of the threshold, and if a batch of samples generally meets the standard but triggers adjustment or does not meet the standard but does not trigger adjustment, then fine-tune the threshold range. Based on the initial defects of the sample (such as uneven oxide thickness), the threshold of each individual sample is adjusted individually (for example, in areas with thicker oxide, the correction threshold can be appropriately relaxed to avoid over-adjustment). If environmental factors (temperature, humidity) affect the process effect, adjust the threshold accordingly (e.g., in low temperature environments, the acid etching effect decreases, so the qualified threshold for acid etching effect can be appropriately reduced to avoid frequent triggering of correction).

[0040] After each process is completed, record the feedback data, threshold judgment results, and path adjustment effects. Verify the thresholds periodically (e.g., monthly). If the process pass rate and efficiency improve after adjustment, retain the calibrated thresholds. If misjudgment occurs, backtrack the data and recalibrate, forming a closed loop of setting-calibration-verification-update.

[0041] The multi-objective optimizer is used to assign weights to objectives such as "highest quality," "shortest time," and "lowest energy / material consumption," globally optimizing batch task queuing and resource allocation. For example, when production tasks are urgent, setting a higher weight for "shortest time" will prioritize processing samples with shorter processing times, while also considering the "highest quality" objective to ensure the accuracy of test results. The formula is as follows:

[0042] in, Let be the weight of the i-th objective. Let x be a function of the i-th objective, and let x be the decision variable.

[0043] After receiving the inspection task order, the central intelligent scheduling subsystem uses a multi-objective optimizer to formulate a preliminary batch scheduling plan based on equipment status and optimization objectives. Once the sample enters the system, the dynamic process path planner generates a baseline path based on the initial sample information. After each process node is completed, the path is dynamically adjusted based on real-time feedback data. A digital twin model monitors the physical system's operational status in real time, ensuring the feasibility and safety of the path planning.

[0044] The sample pretreatment and identification domain includes an information acquisition module and a report generation module.

[0045] The information acquisition module includes a 2D / 3D vision unit integrated into the sample inlet, a barcode scanner, and a laser-induced breakdown spectroscopy or miniature X-ray fluorescence probe. This enables rapid, non-destructive identification of the sample ID, precise three-dimensional dimensions, surface morphology, and material composition. For example, for a stainless steel sample, the 2D / 3D vision unit acquires its surface morphology and three-dimensional dimensions, the barcode scanner reads the sample ID, and the laser-induced breakdown spectroscopy probe analyzes its material composition to determine that the sample is 304 stainless steel.

[0046] The report generation module is used to synthesize the collected information to generate a digital passport for the sample, including material, size, and initial surface condition (flatness, oxide scale coverage score). For example, for the 304 stainless steel sample mentioned above, the generated "digital passport" records the material as 304 stainless steel, the size as 100mm long, 50mm wide, and 10mm thick, the initial surface condition as good flatness, and the oxide scale coverage score as 3 out of 5.

[0047] After the sample enters the information acquisition module, all acquisition devices work simultaneously to obtain comprehensive information about the sample. Once information acquisition is complete, the system automatically generates an initial state digitized report and uploads it to the central intelligent scheduling system. The central intelligent scheduling system uses the digital passport as an index to query the process knowledge base and generate a baseline process blueprint for the sample.

[0048] The intelligent acid etching process control domain includes an acid mist monitoring module, an acid liquid management module, and a corrosion process control module.

[0049] The acid mist monitoring module includes a high-definition camera positioned above the corrosion tank, which, together with a surrounding temperature-controlled inert gas curtain (such as nitrogen) and an automatically opening and closing physical sealing cover, creates a clear viewing window with controllable mist layer during monitoring, enabling truly effective process observation. For example, during hot acid corrosion, the high-definition camera captures real-time images of the corrosion tank, while the temperature-controlled inert gas curtain and physical sealing cover effectively suppress acid mist, ensuring clear images. By observing the corrosion process through the monitoring screen, anomalies can be detected promptly.

[0050] The acid management module includes a circulation loop and an acid performance decay model, enabling automatic replenishment or replacement of acid as needed.

[0051] The circulation loop is used for automatic solution preparation and online monitoring of density and pH. An acid performance degradation model is used to monitor data and the cumulative area of ​​corroded samples. For example, the acid mist monitoring module monitors the density and pH of the acid solution in real time (using an online density meter and pH meter). When the density or pH exceeds the set range, the acid performance degradation model calculates the remaining acid performance based on the monitoring data and the cumulative area of ​​corroded samples, determining whether the acid needs to be replenished or replaced. The formula is as follows:

[0052] in, For the remaining efficiency of acid solution, The density of the acid solution, The pH value of the acid solution. The cumulative area of ​​the corroded sample. For initial performance, For density increment, For the initial density, This represents the amount of pH decay in the acid solution. , and The attenuation coefficient is... This is the temperature-time correction factor.

[0053] The corrosion process control module receives high-resolution images of the process and extracts features such as the corrosion front expansion rate and surface texture changes using image algorithms. These real-time features are compared with the expected model for the current process stage; if the deviation exceeds the limit, the process parameters are dynamically adjusted. Image processing includes preprocessing, segmentation, and tracking. Preprocessing mainly addresses acid fog interference, using the Dark Channel Prior algorithm for defogging; segmentation mainly extracts the corrosion region using U-Net or traditional thresholding; tracking mainly involves matching the corrosion front point set (optical flow method or feature point tracking). (Expansion rate...) The calculation formula is as follows:

[0054] in, This is the leading edge displacement (pixel displacement converted to actual distance). This represents the time corresponding to the leading-edge displacement.

[0055] For example, adjusting the liquid flow pattern in the tank, changing the sample oscillation amplitude, or fine-tuning the corrosion temperature can form a real-time small closed-loop control within the corrosion process. For instance, during hot acid corrosion, if the image algorithm extracts a slow corrosion front expansion rate that deviates significantly from the expected model, the corrosion process adaptive controller can dynamically adjust the process parameters, increasing the corrosion temperature by 5°C to accelerate the corrosion process.

[0056] After the sample enters the etching tank, the acid mist monitoring module activates, providing a clear monitoring image. The acid solution management module monitors the acid solution status in real time to ensure acid quality. The etching process control module dynamically adjusts process parameters based on high-definition images and expected models to achieve precise control of the etching process. The central intelligent scheduling subsystem receives information from each module in real time for global coordination and management.

[0057] The adaptive post-processing and quality closed-loop domain includes the washing and drying line, quality inspection points, and rework traceability decision-makers.

[0058] The cleaning and drying line includes multiple independent and controllable stations such as pre-cleaning, main cleaning (using different modes such as cold / hot water, ultrasonic cleaning, and brushing), multi-stage rinsing, and hot air / vacuum drying. For example, for a stainless steel sample that has undergone hot acid corrosion, pre-cleaning is performed first to remove residual acid from the surface; then, main cleaning is performed using hot water and ultrasonic modes to thoroughly clean the sample surface; next, multi-stage rinsing is performed to ensure that there are no residual impurities on the sample surface; finally, hot air drying is performed to dry the sample surface.

[0059] After key processes, quality inspection points are set up with rapid detection sensors. For example, a microdroplet conductivity meter is set up after rinsing to detect the concentration of residual ions on the sample surface; after drying, a multispectral surface detector is set up to intelligently identify and classify residual stains (such as water-based residues, oil films, solid particles, and salt crystals) through imaging with different wavelength light sources. For example, if the microdroplet conductivity meter detects a residual ion concentration of 10 μS / cm on the sample surface, which is lower than the set threshold of 20 μS / cm, it indicates that the rinsing effect is good; if the multispectral surface detector detects a water stain on the sample surface, it is classified and recorded.

[0060] The rework traceability decision-maker receives non-conformance signals and stain classification information from quality inspection points. Combining the sample's digital passport and the executed process history, the decision-maker uses a rule engine to determine the most likely source of contamination and instructs a robotic arm to return the sample to that process for targeted rework, rather than simply starting from scratch. Specifically, it uses Bayesian networks or DS evidence theory to calculate the responsibility probability of each process to achieve traceability. For example, for the sample detected with water stains, the intelligent rework traceability decision-maker, combining the digital passport and process history, determines that the water stains are caused by incomplete drying and instructs the robotic arm to return the sample to the drying station to execute a parameter-enhanced drying mode.

[0061] After entering the washing and drying line, samples undergo post-processing at various stations. Quality checks are conducted at key stations after each process. If a non-conforming signal is detected, the rework traceability decision-maker immediately intervenes to make traceability decisions and implement targeted rework. The central intelligent scheduling system monitors the post-processing process in real time to ensure sample quality.

[0062] The results evaluation and knowledge management domain includes a standardized imaging module, an analysis and preliminary evaluation module, and a process knowledge self-evolution engine.

[0063] The standardized imaging module includes an automated imaging chamber under constant illumination. For example, after a sample enters the automated imaging chamber, it is photographed under constant lighting conditions to ensure stable image quality.

[0064] The image intelligent analysis and preliminary assessment unit is used to automatically identify defects (porosity, cracks, segregation, etc.) and preliminarily assess their severity on the final low-magnification photographs. For example, the image intelligent analysis algorithm analyzes the photograph, identifies a crack defect on the sample surface, and preliminarily assesses it as of moderate severity based on the length and width of the crack.

[0065] The process knowledge self-evolution engine serves as the system's learning hub. Continuously running background analysis, it correlates the final image quality quantification score of each sample with the complete data chain of the entire process, including initial state, dynamic process path, and data from all sensors and visual feedback. Through machine learning, it automatically discovers strong correlation rules between high-quality results and specific combinations of process parameters, continuously optimizing and enriching the process knowledge base to form a system-level closed loop that drives the entire system to become increasingly intelligent with use. For example, the process knowledge self-evolution engine analysis reveals that when the corrosion temperature is 80℃ and the corrosion time is 30 minutes, the inspection results for 304 stainless steel samples are of high quality. This rule is recorded in the process knowledge base for optimizing the processing technology of subsequent similar samples.

[0066] After sample processing, the samples are taken in a standardized imaging unit. The image intelligent analysis and preliminary evaluation unit analyzes and evaluates the images. The process knowledge self-evolution engine is activated, correlating the test results with the entire process data to optimize the process knowledge base. The central intelligent scheduling system adjusts the processing procedures for subsequent samples based on the optimized process knowledge base.

[0067] The above content is merely an embodiment of the present invention. Commonly known structures and characteristics of the solutions are not described in detail here. Those skilled in the art are aware of all common technical knowledge in the field prior to the application date or priority date, are aware of all existing technologies in that field, and have the ability to apply conventional experimental methods prior to that date. Those skilled in the art can improve and implement this solution based on the guidance provided in this application and their own capabilities. Some typical known structures or methods should not be obstacles for those skilled in the art to implement this application. It should be noted that those skilled in the art can make several modifications and improvements without departing from the structure of the present invention. These should also be considered within the scope of protection of the present invention, and will not affect the effectiveness of the implementation of the present invention or the practicality of the patent. The scope of protection claimed in this application should be determined by the content of its claims, and the specific embodiments described in the specification can be used to interpret the content of the claims.

Claims

1. A low-magnification experimental system with multi-system collaboration, characterized in that, include: The central intelligent scheduling subsystem includes a digital twin model for virtual mapping of the physical system, a dynamic process path planner for generating a baseline path based on sample information and generating the actual execution path based on real-time feedback, and a multi-objective optimizer for batch task scheduling and resource allocation based on optimization objectives. The sample pretreatment and identification domain includes an information acquisition module for acquiring sample ID, three-dimensional dimensions, surface morphology, and material composition, as well as a report generation module for generating a sample digital passport containing sample material, dimensions, and initial surface condition. The intelligent acid etching process control domain includes an acid mist monitoring module for monitoring the environment inside the etching tank, an acid liquid management module for managing acid liquid efficiency, and an etching process control module for dynamically adjusting process parameters based on the comparison results of real-time image feature values ​​and expected models. The adaptive post-processing and quality closed-loop domain includes a cleaning and drying line for cleaning and drying samples, a quality detection point for detecting post-processing quality, and a rework traceability decision-maker for tracing the source of non-conformity based on the non-conformity signal and instructing targeted rework. The results evaluation and knowledge management domain includes a standardized imaging module for acquiring standardized images, an analysis and preliminary evaluation module for defect identification and grade assessment of images, and a process knowledge self-evolution engine for linking and mining data from the entire process to optimize the process knowledge base.

2. The low-magnification experimental system with multi-system collaboration according to claim 1, characterized in that, The digital twin model includes a geometric layer and a physical layer. The geometric layer includes the kinematic parameters of the robotic arm and the spatial coordinates of the workstation. The physical layer includes an acid flow field model, a temperature field heat transfer model, and a robotic arm dynamics model, which are used for process simulation and collision detection.

3. The low-magnification experimental system with multi-system collaboration according to claim 1, characterized in that, The information acquisition module includes a 2D / 3D vision unit, a barcode scanner, and a laser-induced breakdown spectroscopy or miniature X-ray fluorescence probe; the dynamic process path planner is configured to dynamically insert or skip specific process nodes in the baseline path based on visual judgment and sensor values.

4. The low-magnification experimental system with multi-system collaboration according to claim 1, characterized in that, The acid mist monitoring module includes a high-definition camera installed above the corrosion tank, which, together with a surrounding temperature-controlled inert gas curtain and an automatically opening and closing physical sealing cover, forms a visual monitoring environment.

5. The low-magnification experimental system with multi-system collaboration according to claim 1, characterized in that, The acid management module includes a circulation loop and an acid performance decay model. The acid performance decay model calculates the remaining acid performance based on the monitored acid density, pH value, and cumulative area of ​​the corroded sample, and controls the replenishment or replacement of acid accordingly.

6. The low-magnification experimental system with multi-system collaboration according to claim 1, characterized in that, The corrosion process control module is configured to extract the corrosion front expansion rate and surface texture change feature values ​​through image algorithms, and compare these feature values ​​with the expected model of the current process stage. When the deviation exceeds the limit, the liquid flow pattern in the tank, the sample swing amplitude, or the corrosion temperature is dynamically adjusted.

7. The low-magnification experimental system with multi-system collaboration according to claim 1, characterized in that, The cleaning and drying line includes pre-cleaning, main cleaning, multi-stage rinsing, and hot air / vacuum drying stations; the quality inspection points include a microdroplet conductivity meter installed after rinsing and a multispectral surface detector installed after drying.

8. The low-magnification experimental system with multi-system collaboration according to claim 1, characterized in that, The rework traceability decision-maker is configured to receive non-conforming signals from quality inspection points, combine the sample's digital passport and the history of executed processes, use a rule engine to calculate the responsibility probability of each process to determine the process that is the root cause of the contamination, and instruct the robotic arm to send the sample back to that process for targeted rework.

9. The low-magnification experimental system with multi-system collaboration according to claim 1, characterized in that, The analysis and preliminary evaluation module is configured to automatically identify defects and make preliminary assessments of their levels in the final low-magnification images; the process knowledge self-evolution engine is configured to correlate the final image quality quantitative score with the complete data chain of the entire process and discover strong correlation rules between the results and specific process parameter combinations through machine learning.

10. A low-magnification experimental method involving multiple systems, characterized in that, The system is applied to a low-magnification testing system with multi-system collaboration as described in any one of claims 1-9, comprising: The system receives inspection task orders through a central intelligent scheduling subsystem, performs simulations using a digital twin model, formulates scheduling plans using a multi-objective optimizer, and generates baseline paths for samples using a dynamic process path planner. In the sample pretreatment and identification domain, the sample's ID, size, surface morphology, and material composition information are collected to generate a digital passport containing the sample's initial state. In the intelligent acid etching process control domain, the acid mist monitoring module monitors the corrosion environment, the acid liquid management module manages the acid liquid efficiency, and the corrosion process control module dynamically adjusts the acid etching process parameters based on the comparison results of real-time image features and expected models. The samples are cleaned and dried in the adaptive post-processing and quality closed-loop domain, and quality testing is performed. If the test fails, the source is traced and analyzed based on the digital passport and process history, and the rework is instructed to be targeted. In the result evaluation and knowledge management domain, samples are standardized for imaging and defect assessment. The assessment results are then correlated with the entire process data for analysis and updating of the process knowledge base to optimize subsequent processes.

Images