Static arc contact workpiece rusty spot defect detection and recognition processing method

Abstract

The invention discloses a rusty spot defect detection and recognition processing method for a static arc contact machined part, and relates to the technical field of rusty part processing. The hyper-spectral camera collects images at multiple angles, adaptively adjusts the angle, and screens clear images; quantum denoising, GAN and MSRN are combined to enhance the image; using QCNN to extract rusty spot features, and using QSVM to identify rusty spot types and severity; introducing a quantum annealing algorithm to make a processing decision; rusty spots are repaired by combining a nano material and laser; after processing, acquiring an image verification effect again; sample library construction and real-time monitoring are covered, and the quality and adaptability are improved through quantum coding and quantum reinforcement learning. According to the method, comprehensive images are collected through hyper-spectral imaging, quantum algorithm preprocessing, feature extraction and recognition are carried out, the detection precision and efficiency are improved, the quantum annealing decision is scientific, and nanometer and laser repair damage is small; the sample and monitoring module improves quality and adaptability, ensures stable operation of electrical equipment, and is environment-friendly and efficient.

Description

Technical Field

[0001] The present invention relates to the technical field of rust treatment, and in particular to a method for detecting, identifying and treating rust defects of a static arc contact workpiece. Background Art

[0002] In modern industrial production, static arc contacts are key components of electrical equipment, and their quality and performance directly impact the stability and reliability of the entire electrical system. However, during production, storage, and use, static arc contact parts are susceptible to rust defects due to the effects of ambient humidity, oxygen in the air, and other chemicals. Rust not only affects the appearance of static arc contacts but, more importantly, reduces their electrical conductivity, mechanical properties, and corrosion resistance. In severe cases, it can even cause electrical equipment failures, leading to safety incidents and significant economic losses.

[0003] Traditional methods for detecting rust on static arc contacts rely primarily on manual visual inspection and simple physical testing. Manual visual inspection is subject to high subjectivity, low efficiency, and a high risk of missed detections. It also struggles to accurately identify tiny rust spots or those hidden within workpieces. While simple physical testing methods, such as magnetic particle inspection and ultrasonic testing, can detect some internal defects, they have limited accuracy for surface defects like rust and are unable to accurately determine the type and severity of rust spots.

[0004] With the development of computer vision and artificial intelligence technologies, several image recognition-based rust detection methods have emerged. These methods capture images of static arc contact workpieces and utilize image processing and machine learning algorithms to identify and classify rust spots. However, existing image recognition methods still have some shortcomings when faced with complex industrial environments and diverse rust characteristics. For example, the image acquisition process is easily affected by factors such as lighting and angle, resulting in unstable image quality; image processing algorithms are less robust to noise and interference, prone to misidentification and omission; and machine learning models have limited training samples, making them incapable of recognizing newly emerging rust types and characteristics.

[0005] Furthermore, traditional methods for treating rust stains primarily include chemical derusting and mechanical polishing. While chemical derusting can quickly remove rust, it produces significant amounts of wastewater and exhaust gas, causing significant environmental pollution. Mechanical polishing can easily damage the surface of workpieces, impacting performance and precision. While our company utilizes more advanced equipment (such as laser derusting), we still cannot achieve an optimal treatment solution based on rust analysis and feedback control of our derusting equipment. Therefore, those skilled in the art urgently need to address these technical issues. Summary of the Invention

[0006] Based on the fact that existing processing decisions are mainly based on experience and simple rules, lack scientific basis and optimization algorithm, it is impossible to formulate the best processing plan according to the specific situation of the rust spots; the present invention proposes a method for detecting and identifying rust spots on static arc contact workpieces, which combines advanced hyperspectral imaging, quantum computing and other technologies to solve the problems of existing methods, improve the accuracy and efficiency of rust spot detection and identification, and at the same time cooperate with rust removal equipment to realize scientific and reasonable feedback control decisions, thereby achieving high-quality production and use of static arc contact workpieces.

[0007] To solve the problems mentioned in the above-mentioned prior art.

[0008] In order to achieve the above-mentioned object, the present invention adopts the following technical solution: a method for detecting and identifying rust defects of a static arc contact workpiece, comprising the following steps:

[0009] A camera with hyperspectral imaging capability is used to capture images of the static arc contact workpiece from different angles. In addition to visible light and infrared light, it also captures ultraviolet spectrum information to construct a hyperspectral image dataset; an improved image clarity evaluation formula is used. Among them S ij is the texture complexity of the area where the pixel (i, j) is located, and α is the complexity influence coefficient;

[0010] The quantum denoising algorithm is used to denoise the collected images, and the generative adversarial network (GAN) and the multi-scale residual network (MSRN) are combined for image enhancement. GAN generates high-quality images, and MSRN enhances image features at different scales. An improved image enhancement coefficient formula is used. Among them C ij is the color contrast at pixel (i, j), β is the contrast influence coefficient;

[0011] The color, shape, and texture features of rust spots are extracted by combining quantum convolutional neural network (QCNN) with multi-scale feature fusion technology; the feature significance evaluation formula based on information entropy is used. where w k is the weight of the kth feature, f k is the value of the kth feature, and s is the number of features;

[0012] The extracted rust spot features are classified and identified using a classifier based on quantum support vector machine (QSVM), and the improved recognition accuracy evaluation formula is used. Where T rec is the identification time, γ is the time influence coefficient;

[0013] The static arc contact parts are processed according to the treatment decision. For the repaired rust spots, a combination of nanomaterial repair technology and laser rust removal technology is used. Nanomaterials fill the tiny pores in the rust spots, and laser rust removal accurately removes the rust layer without damaging the substrate. For scrapped parts, a green recycling process is adopted.

[0014] The treated static arc contact workpiece is subjected to hyperspectral image acquisition and analysis again, and the treatment effect evaluation formula based on quantum state comparison is used Among them, M diff is the difference measurement value of the quantum state before and after processing, and ∈ is the difference influence coefficient.

[0015] Furthermore, the method further includes the following steps:

[0016] Sample library construction steps: Collect rust spot images of different types and severity of static arc contact parts as samples, classify and label the samples using quantum clustering algorithm, and use the improved sample quality evaluation formula Among them H entropy is the information entropy of the sample, and ζ is the entropy influence coefficient.

[0017] Real-time monitoring steps: During the static arc contact processing, an industrial camera with quantum imaging technology is used to collect images of the workpiece in real time. The deep learning model and classifier are continuously updated using the quantum reinforcement learning algorithm. A real-time monitoring accuracy evaluation formula based on quantum entanglement is used. Evaluate the real-time monitoring effect, where E entanglement is the quantum entanglement measurement value, and η is the entanglement influence coefficient.

[0018] Furthermore, in the image acquisition step, polarization-hyperspectral fusion imaging technology is used to obtain polarization and hyperspectral information of the surface of the static arc contact workpiece, and the polarization information fusion formula F=λ×I based on quantum state fusion is used. p ×(1+θ×Q p )+(1-λ)×I m ×(1+θ×Q m ), where I p is the polarization image information, I m is the hyperspectral image information, Q p and Q m are the quantum state eigenvalues of polarization and hyperspectral information respectively, λ is the fusion coefficient, θ is the quantum state influence coefficient, and the two types of information are fused.

[0019] Furthermore, in the image preprocessing step, the quantum generative adversarial network QGAN is used to denoise and enhance the image; the QGAN processing effect evaluation formula based on quantum fidelity is used The quantum fidelity of the image before and after processing, where F ijis the pixel point (i, j).

[0020] Furthermore, in the rust spot feature extraction step, the convolutional neural network QACNN with quantum attention mechanism is used to focus on the key features of rust spots through the superposition and entanglement of quantum states; the attention weight distribution formula based on the quantum state probability distribution is used. where f k is the value of the kth feature, Q k is the quantum state probability value of the kth feature, s is the number of features, and attention weights are assigned to different features.

[0021] Furthermore, based on the type and severity of the identified rust spots, the ant colony algorithm is used to control the laser head to achieve precise rust removal operations, including:

[0022] Based on the type and severity of the identified rust spots, the ant colony algorithm is used to control the laser head to achieve precise rust removal operations, including:

[0023] Initialize the current position of the first laser head and calculate the spatial distance from the current laser head to each rust spot;

[0024] For all current rust spots, use the K-means clustering algorithm or the DBSCAN algorithm to perform cluster analysis based on severity, grouping rust spots with similar colors into a rust spot cluster sub-region, and then divide them into N rust spot cluster sub-regions in sequence;

[0025] In the current rust cluster sub-area, the path planning algorithm is combined with the cooling time optimization of power switching to implement calculation processing to obtain the planned path within the cluster.

[0026] A system using the method for detecting, identifying, and processing rust defects on a static arc contact workpiece comprises:

[0027] Image acquisition module: This module uses hyperspectral imaging and polarization-hyperspectral fusion imaging technologies, is equipped with an adaptive angle adjustment algorithm, and uses an improved image clarity evaluation formula and a polarization information fusion formula based on quantum state fusion to collect and screen images.

[0028] Image preprocessing module: Use quantum denoising algorithm, QGAN combined with multi-scale residual network for denoising and enhancement processing, and use improved image enhancement coefficient formula and quantum fidelity-based QGAN processing effect evaluation formula to evaluate the effect;

[0029] Rust spot feature extraction module: Utilizes QCNN and QACNN combined with multi-scale feature fusion technology to extract rust spot features. It uses a feature significance evaluation formula based on information entropy and an attention weight allocation formula based on quantum state probability distribution to screen and assign weights.

[0030] Rust spot recognition module: uses the QSVM classifier to classify and identify rust spot features, and uses the improved recognition accuracy evaluation formula to evaluate the recognition effect;

[0031] Processing decision module: Based on the type and severity of the identified rust spots, the ant colony algorithm is used to control the laser head to achieve precise rust removal operations;

[0032] Processing execution module: Use nanomaterial repair technology and laser rust removal technology to repair rust spots, and use green recycling technology to process scrapped parts;

[0033] Result verification module: Hyperspectral images of the processed workpiece are collected and analyzed again, and the treatment effect is verified using the treatment effect evaluation formula based on quantum state comparison.

[0034] Furthermore, it also includes:

[0035] Sample library construction module: collect and annotate rust spot image samples, classify and annotate them using quantum clustering algorithm, and use improved sample quality assessment formula to screen samples for model training;

[0036] Real-time monitoring module: Use industrial cameras with quantum imaging technology to collect images in real time, apply quantum reinforcement learning algorithms to update models and classifiers, and use real-time monitoring accuracy evaluation formulas based on quantum entanglement to evaluate the results.

[0037] Compared with the existing technology, the beneficial effects of the present invention are:

[0038] In terms of image acquisition, the use of hyperspectral imaging and polarization-hyperspectral fusion imaging technologies, combined with an adaptive angle adjustment algorithm, enables more comprehensive and accurate rust information to be obtained, effectively mitigating the effects of lighting and angle on image quality. Improved image clarity assessment formulas and a polarization information fusion formula based on quantum state fusion further enhance image quality and information richness, providing a solid foundation for subsequent analysis.

[0039] During the image preprocessing stage, the application of quantum denoising algorithms and quantum generative adversarial networks (QGANs) can effectively remove noise interference while generating high-quality enhanced images. The QGAN processing effect evaluation formula based on quantum fidelity ensures the accuracy and reliability of preprocessing and improves the precision of subsequent feature extraction and recognition. During the rust spot feature extraction and recognition phase, the use of quantum convolutional neural networks (QCNNs), quantum attention-based convolutional neural networks (QACNNs), and quantum support vector machines (QSVMs) leverages the powerful parallel capabilities of quantum computing to accelerate the feature extraction and recognition process, improving recognition accuracy and efficiency. Improved feature saliency evaluation formulas and attention weight allocation formulas based on quantum state probability distributions enable more precise capture and utilization of key features.

[0040] In terms of treatment decisions, an ant colony algorithm controls the laser head to achieve precise rust removal based on the type and severity of the identified rust spots. The combination of nanomaterial repair technology and laser rust removal technology achieves efficient rust repair while minimizing damage to the workpiece. A green recycling process enables environmentally friendly treatment and resource reuse of scrapped workpieces.

[0041] The application of technologies such as quantum coding, quantum clustering, and quantum reinforcement learning in the sample library construction and real-time monitoring modules improves the security of sample data and the adaptability of the model. Evaluation formulas based on quantum entanglement and quantum state correlation ensure the accuracy and reliability of real-time monitoring and data management.

[0042] In summary, the method and system of the present invention comprehensively improve the level of detection, identification and processing of rust defects in static arc contact workpieces, have the advantages of high efficiency, accuracy and environmental protection, and can provide strong guarantees for the stable operation of electrical equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 This is a schematic block diagram of a system for detecting, identifying, and processing rust defects on static arc contact workpieces proposed by the present invention;

[0044] Figure 2 This is a flow chart of a method for detecting, identifying, and processing rust defects on a static arc contact workpiece proposed by the present invention;

[0045] Figure 3 This is a flowchart of a specific implementation process of a method for detecting and identifying rust defects on a static arc contact workpiece proposed by the present invention;

[0046] Figure 4 This is a schematic diagram of the effect of clustering multiple rust spot cluster sub-regions in a method for detecting and identifying rust spot defects on a static arc contact workpiece proposed by the present invention;

[0047] Figure 5 This is a schematic diagram of the path planning effect of the current rust spot cluster sub-region in the rust spot defect detection, identification and processing method of the static arc contact workpiece proposed by the present invention. DETAILED DESCRIPTION

[0048] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0049] Example 1

[0050] Reference Figure 1-2 The embodiment of the present invention provides a method for detecting and identifying rust defects in a static arc contact workpiece, comprising the following steps:

[0051] S101, image acquisition step: Use a camera with hyperspectral imaging capabilities to capture images of the static arc contact workpiece from multiple different angles. In addition to visible light and infrared light, it can also capture special spectral information such as ultraviolet light to construct a hyperspectral image dataset.

[0052] Using the adaptive angle adjustment algorithm, the optimal shooting angle is automatically determined according to the contour of the workpiece, improving the integrity of image information acquisition.

[0053] Adopting an improved image clarity evaluation formula Among them S ij is the texture complexity of the area where the pixel point (i, j) is located, and α is the complexity influence coefficient, which is used to more accurately screen clear images for subsequent analysis.

[0054] S102, image preprocessing step: Use quantum denoising algorithm to denoise the collected image, taking advantage of the superposition and entanglement characteristics of quantum bits to efficiently remove noise interference and retain image details to the greatest extent. Combine generative adversarial network (GAN) and multi-scale residual network (MSRN) for image enhancement. GAN generates high-quality images, and MSRN enhances image features at different scales. Use improved image enhancement coefficient formula Among them C ij is the color contrast at pixel (i, j), and β is the contrast influence coefficient, which can be used to evaluate the enhancement effect more scientifically.

[0055] S103, Rust Spot Feature Extraction Step: Utilizes a Quantum Convolutional Neural Network (QCNN) combined with multi-scale feature fusion technology to extract features such as color, shape, and texture of the rust spots. Leveraging the parallel computing power of quantum states, QCNN accelerates feature extraction and improves feature differentiation.

[0056] Adopt the feature significance evaluation formula based on information entropy where w k is the weight of the kth feature, f k is the value of the kth feature, s is the number of features, and it is used to more accurately screen features with high significance.

[0057] S104, rust spot identification step: Use a classifier based on quantum support vector machine (QSVM) to classify and identify the extracted rust spot features. QSVM uses quantum algorithms to optimize the classification decision boundary to improve the accuracy and efficiency of identification. Use the improved recognition accuracy evaluation formula Where T recis the recognition time, and γ is the time influence coefficient, which comprehensively considers the recognition time and accuracy.

[0058] S105, processing decision step: according to the type and severity of the identified rust spots, the ant colony algorithm is used to control the laser head to achieve precise rust removal processing operations.

[0059] S106: Processing steps: The static arc contact parts are processed according to the processing decision. For repairable rust spots, a combination of nanomaterial repair technology and laser rust removal technology is used. Nanomaterials can fill the tiny pores in the rust spots, and laser rust removal can accurately remove the rust layer without damaging the substrate. For scrapped parts, green recycling processes are used to achieve efficient material recovery and reuse.

[0060] Result verification steps: The treated static arc contact workpiece is subjected to hyperspectral image acquisition and analysis again, and the treatment effect evaluation formula based on quantum state comparison is used Among them, M diff is the measurement value of the difference between the quantum states before and after processing, ∈ is the difference influence coefficient, which can more accurately verify the processing effect.

[0061] The present invention further comprises the following steps:

[0062] Sample library construction steps: Collect a large number of rust spot images of different types and severity of static arc contact parts as samples; classify and label the samples using a quantum clustering algorithm that uses the characteristics of quantum states to achieve more accurate clustering. Use an improved sample quality assessment formula Among them H entropy is the information entropy of the sample, ζ is the entropy influence coefficient, and high-quality samples are screened out for training the deep learning model to improve the recognition accuracy of the model.

[0063] Real-time monitoring steps: During the static arc contact processing, industrial cameras equipped with quantum imaging technology are used to capture images of the workpiece in real time. Quantum imaging technology can capture weaker optical signals and improve the ability to detect early rust spots. The deep learning model and classifier are continuously updated using the quantum reinforcement learning algorithm, which achieves rapid learning and optimization through the evolution of quantum states. A real-time monitoring accuracy evaluation formula based on quantum entanglement is used. Among them E entanglement is the quantum entanglement measurement value, η is the entanglement influence coefficient, which can more accurately evaluate the real-time monitoring effect.

[0064] In the present invention, in the image acquisition step, polarization-hyperspectral fusion imaging technology is used to obtain polarization and hyperspectral information of the surface of the static arc contact workpiece, which more comprehensively reflects the characteristics of the rust spots. The polarization information fusion formula F=λ×I based on quantum state fusion is used. p ×(1+θ×Q p)+(1-λ)×I m ×(1+θ×Q m ), where I p is the polarization image information, I m is the hyperspectral image information, Q p and Q m are the quantum state eigenvalues of polarization and hyperspectral information, λ is the fusion coefficient, and θ is the quantum state influence coefficient. The two types of information are fused to improve the information richness of the image.

[0065] In this invention, in the image preprocessing step, a quantum generative adversarial network (QGAN) is used to denoise and enhance the image. QGAN uses the characteristics of quantum states to generate high-quality images that are closer to real images, and the discriminator uses quantum measurements to determine the authenticity of the image. The QGAN processing effect evaluation formula based on quantum fidelity is used. Among them F ij It is the quantum fidelity of the image before and after processing at pixel point (i, j), which can more accurately evaluate the processing effect.

[0066] In the present invention, in the step of extracting rust spot features, a convolutional neural network with quantum attention mechanism (QACNN) is used to achieve more efficient attention to the key features of rust spots through the superposition and entanglement of quantum states. where f k is the value of the kth feature, Q k is the quantum state probability value of the kth feature, s is the number of features, and attention weights are assigned to different features to improve the accuracy of feature extraction.

[0067] Example 2

[0068] See also Figure 2 as well as Figure 3 Further research revealed that the core of laser rust removal lies in precision, specifically removing only the rust layer without damaging the substrate. The aforementioned image detection and feature analysis steps provide crucial information for laser rust removal. For example, based on the image detection results, the location of rust spots on the workpiece can be determined and their distribution can be mapped. Furthermore, through feature extraction (color, shape, texture, etc.) and classification, the depth and thickness of rust spots can be determined, providing a basis for adjusting the laser rust removal power and operating parameters.

[0069] The present invention provides a solution to this problem. By extracting and classifying features (such as color, shape, and texture), the depth and thickness of rust spots are determined, providing a basis for adjusting the power and operating parameters of laser rust removal. Multi-scale feature fusion technology accurately defines the boundaries of rust spots, preventing excessive laser irradiation on the substrate.

[0070] During specific processing, detection and identification → data is input to the laser equipment, and the results of rust spot detection and classification are used as input data to guide the design of the working path and parameters of the laser rust removal equipment. Then the feedback control is output to form a closed-loop control between laser rust removal and rust spot detection, and the power of the laser is adjusted dynamically. At the same time, the following principles are followed: for rust spots with thicker thickness, the laser power needs to be higher; for slight surface rust spots, the power can be appropriately reduced to avoid damaging the substrate. The laser power is set according to the severity of the rust spots. The laser scanning path is planned according to the shape and position of the rust spots. The laser equipment removes the rust spots according to the set parameters and path. The rust removal effect is monitored by an optical sensor, and the laser power is adjusted dynamically;

[0071] However, further research also found that the laser equipment used by the researchers is a large-power equipment, which has certain usage conditions; namely, the path planning problem when processing multiple similar but unconnected rust spots, and taking into account the cooling time required for laser output power adjustment (because the laser equipment of this company requires a long time to cool down or adjust and preheat when the output power adjustment range is too large, but when the output power adjustment range is not large, there is often no need for cooling time or preheating time). This embodiment requires a path planning algorithm to optimize the scanning path and power adjustment frequency of the laser equipment, thereby improving the rust removal efficiency. There are multiple similar but unconnected rust spots, and the severity of each rust spot is different, requiring different laser powers. The laser equipment requires a certain amount of cooling or stabilization time when adjusting the output power (cooling time or preheating time is collectively referred to as adjustment time or cooling time in the industry). Frequent power adjustment will reduce work efficiency. A planned path is used to reduce the ineffective movement of the laser equipment and avoid frequent power switching.

[0072] Considering that the laser equipment needs cooling time when adjusting the laser output power significantly; therefore, a planning path algorithm is used to deal with the situation where multiple rust spots are not connected. Path planning is selected for rust spots with different degrees of severity to improve the rust removal efficiency.

[0073] The following model output objectives were set: First, efficient path planning was achieved, enabling the laser to cover all rust spots using the shortest possible path through reasonable path arrangement. Second, the frequency of power adjustments was reduced, prioritizing rust spots of similar severity to reduce the number of power switches, thereby improving efficiency.

[0074] S1051, initializing the current position of the first laser head and calculating the spatial distance from the current laser head to each rust spot;

[0075] During the research process, the design idea is as follows: before path planning, the rust spots are first grouped and sorted: that is, grouping based on severity; the detected rust spots are grouped according to severity, and the grouping results are as follows: Mild rust spot group: requires low-power laser processing. Medium rust spot group: requires medium-power laser processing. Severe rust spot group: requires high-power laser processing. By grouping, it is ensured that the laser equipment can continuously process rust spots with the same power requirements and reduce the power switching frequency. Then, in the specific execution, the rust spots should be processed in order from high severity to low severity. However, the above-mentioned embodiment of the present application believes that the above-mentioned simple division of rust spot groups into three levels still cannot guarantee the purpose of shortening the cooling time corresponding to the power adjustment. For this reason, the embodiment of the present application adopts a solution to use the K-means clustering algorithm to cluster and divide multiple cluster groups;

[0076] That is, all current rust spots are clustered according to severity (i.e., rust spot color). Specifically, the K-means clustering algorithm or the DBSCAN algorithm is used to group rust spots of similar rust color into a rust spot cluster sub-region, and then N rust spot cluster sub-regions (i.e., N cluster groups) are obtained by sub-dividing. After the current rust spot cluster sub-region is determined, path planning processing is performed within the cluster (the second embodiment of the present invention clusters rust spots of similar severity and then processes them in batches to avoid frequent power switching). Then, the current rust spot cluster sub-region is further divided into a sub-mild rust spot group, a sub-medium rust spot group, and a sub-severe rust spot group. Generally, priority sorting is considered: within each rust spot cluster sub-region, rust spots are processed in order from high severity to low severity. However, the embodiment of the present application adopts the following scheme, namely, secondary subdivision and combination, as detailed in S1052.

[0077] S1052: In the current rust cluster sub-area, a path planning algorithm is combined with power switching cooling time optimization to perform calculations to obtain a planned path within the cluster:

[0078] See also Figure 4 and Figure 5 , Figure 4 Schematic diagram of the effect after clustering multiple rust cluster sub-regions; Figure 5 Schematic diagram of the path planning effect for the current rust cluster sub-area.

[0079] The goal of the ant colony algorithm is to minimize the total path length and the cooling time of the power switching. That is, the path optimization goal is to minimize the moving path length of the laser head and the cooling time of the power switching. The input of the above ant colony algorithm: the spatial coordinates of the rust spot, the spatial distance from rust spot i to rust spot j, the rust severity information (or rust severity Si ). Output: optimal path sequence.

[0080] S10521 sets the objective function for path optimization: The goal of the ant colony algorithm is to minimize the total path length and power switching cost. In rust spot path planning, the goal of the third embodiment of the present invention is to plan the optimal path for the laser device based on the distribution and severity of the rust spots, minimizing the following two objectives: 1. Total path length: the total distance the laser device travels between rust spots. 2. Power switching cost: the time it takes to adjust the laser head power when treating rust spots of varying severity.

[0081] Input parameters: Rust spot set N: rust spot location and severity information, defined as N = {n1, n2, ..., n m}, where each rust spot n i Contains the following properties:

[0082] Position coordinates (x i ,y i ); severity S i (Rust severity information): corresponds to the laser power requirement; distance matrix D: the Euclidean distance between rust spots Power switching cost matrix P: Cooling time or overhead p of power adjustment ij , when S i ≠S j The above corrosion severity information S i It is obtained by quantitative detection of the rust color. The color features are extracted and quantified in the rust area. For example, the average color value is calculated, that is, the color mean of the rust area is calculated in the RGB space. This will not be described in detail.

[0083] Objective function; define the total cost Z, including path length and power switching cost:

[0084]

[0085] d ij : The distance between rust spots i and j.

[0086] p ij : The power switching cost from rust spot i to rust spot j (the power switching cost mentioned here can be the cooling time cost of power switching or other reference costs).

[0087] λ: importance weight of power switching.

[0088] Optimization goal: minimize the objective function Z to obtain the optimal path of the laser equipment.

[0089] Generally speaking, the ant colony algorithm (ACO) seeks the global optimal solution by simulating the behavior of ants releasing pheromones and selecting paths while foraging. Its core steps include: Path selection probability: Ants choose their next target based on pheromone concentration and distance. Pheromone update: After completing a path, ants strengthen (increase) or dilute (evaporate) pheromones based on path quality. Iterative optimization: Through multiple iterations, the algorithm gradually converges to the optimal path.

[0090] S10522, perform parameter initialization processing of the ant colony algorithm: that is, perform parameter settings: α: pheromone influence factor (controls the importance of pheromone); β: heuristic function influence factor (controls the importance of distance); ρ: pheromone evaporation coefficient (controls the decay rate of old pheromone); Q: pheromone release intensity; ant number M and maximum number of iterations T max ;

[0091] Execute initialization of pheromone matrix: define the initial pheromone τ ij =τ0(mean). Initialize the heuristic function η ij =1 / d ij (reciprocal of distance);

[0092] Get the initial position of the ants: randomly assign M ants to any starting point in the rust spot set.

[0093] S10523, execute the path construction of the ant colony algorithm: the probability P of the ant moving from the current position i to the next node j ij : Among them, τ ij : pheromone concentration; η ij : Heuristic function (inverse of distance); allowed: the set of rust spots that have not been visited; the ant will choose the next rust spot based on this probability until all rust spots are visited.

[0094] S10524, perform path completion and cost calculation; each ant completes a complete path (visits all rust spots) and records the path sequence Path k ; Then calculate the total cost Z of the path k , including path length and power switching cost:

[0095]

[0096] S10525, execute pheromone update of ant colony algorithm including pheromone evaporation and pheromone increase: pheromone evaporation (to prevent excessive accumulation): τ ij =(1-ρ)·τ ij ; Where, ρ: evaporation coefficient; pheromone increase (strengthening excellent path): Only add pheromones to the edges on path k.

[0097] S10526, repeat path construction and pheromone update until the ant colony algorithm converges to the optimal path to achieve iterative optimization output; repeat the process of path selection, path completion, and pheromone update until one of the following conditions is met: the maximum number of iterations T is reached max。 And the global optimal path no longer changes (converges).

[0098] In summary, the general scheme for executing the ant colony algorithm is to first initialize: the location and severity of the rust spot are used as the input of the ant colony; execute the path construction of the ant colony algorithm: each ant selects the next rust spot based on the spatial distance and the power switching cooling time cost; execute the pheromone update of the ant colony algorithm: update the pheromone based on the path length and the number of power switching times; repeat the path construction and pheromone update until the ant colony algorithm converges to the optimal path. This will not be repeated in detail.

[0099] Example 3

[0100] The third embodiment of the present invention is based on the second embodiment. In its application, it was found that the path planning still had the problem of insufficient time utilization. In response to this, an improved technical solution of the third embodiment is proposed, namely, preparing a second backup laser head. This solves the problem of cross-use between the first laser head and the second backup laser head and rationally planning time.

[0101] S1051a, initialize the current position of the laser head and calculate the spatial distance from the current laser head to each rust spot; pre-group the rust spots within the current rust spot cluster sub-region, and obtain the sub-minor rust spot group, sub-moderate rust spot group, and sub-severe rust spot group divided within the current rust spot cluster sub-region; obtain the tasks within the sub-minor rust spot group and simultaneously create a first minor rust spot area list (i.e., sort out the tasks within the minor rust spot group within the rust spot cluster sub-region); simultaneously introduce and set up a second backup laser head and start the second backup laser head;

[0102] S1052a: Within the current rust spot cluster sub-region, a path planning algorithm is combined with power switching cooling time optimization to perform calculations and processing to obtain a planned path within the cluster. Simultaneously, tasks within the remaining minor rust spot groups within the current rust spot cluster sub-region are continuously monitored.

[0103] S10521a, that is, setting the objective function of path optimization:

[0104] The goal of the ant colony algorithm is to minimize the total path length and power switching cost. In rust spot path planning, the design goal of the third embodiment of the present invention is to plan the optimal path for the laser device based on the distribution and severity of the rust spots, minimizing the following two objectives: 1. Total path length: the total distance the laser device travels between rust spots; 2. Power switching cost: the time it takes to adjust the laser head power when treating rust spots of varying severity.

[0105] Input parameters: Rust spot set N: rust spot location and severity information, defined as N = {n1, n2, ..., n m}, where each rust spot n i Contains the following properties:

[0106] Position coordinates (x i ,y i ); severity S i (i.e., rust severity information): corresponds to the laser power requirement; distance matrix D: the Euclidean distance between rust spots Power switching cost matrix P: Cooling time or overhead p of power adjustment ij , when S i ≠S j Time is non-zero.

[0107] Objective function; define the total cost Z, including path length and power switching cost:

[0108]

[0109] d ij : The distance between rust spots i and j.

[0110] p ij : The power switching cost from rust spot i to rust spot j (the power switching cost mentioned here can be the cooling time cost of power switching or other reference costs).

[0111] λ: importance weight of power switching.

[0112] Optimization goal: minimize the objective function Z to obtain the optimal path of the laser equipment.

[0113] S10522a, perform parameter initialization processing of the ant colony algorithm: that is, perform parameter settings: α: pheromone influence factor (controls the importance of pheromone); β: heuristic function influence factor (controls the importance of distance); ρ: pheromone evaporation coefficient (controls the decay rate of old pheromone); Q: pheromone release intensity; ant number M and maximum number of iterations T max ;

[0114] Execute initialization of pheromone matrix: define the initial pheromone τij =τ0(mean). Initialize the heuristic function η ij =1 / d ij (reciprocal of distance);

[0115] Get the initial position of the ants: randomly assign M ants to any starting point in the rust spot set.

[0116] S10523a, execute the path construction of the ant colony algorithm: the probability P of the ant moving from the current position i to the next node j ij : Among them, τ ij : pheromone concentration; η ij : Heuristic function (inverse of distance); allowed: the set of rust spots that have not been visited; the ant will choose the next rust spot based on this probability until all rust spots are visited.

[0117] S10524a, perform path completion and cost calculation; each ant completes a complete path (visits all rust spots) and records the path sequence Path k ; Then calculate the total cost Z of the path k , including path length and power switching cost:

[0118]

[0119] S10525a, executing the pheromone update of the ant colony algorithm including pheromone evaporation and pheromone increase: pheromone evaporation (to prevent excessive accumulation): τ ij =(1-ρ)·τ ij ; Where, ρ: evaporation coefficient; pheromone increase (strengthening excellent path): Only add pheromones to the edges on path k.

[0120] S10526a, repeat path construction and pheromone update until the ant colony algorithm converges to the optimal path to achieve iterative optimization output; repeat the process of path selection, path completion, and pheromone update until one of the following conditions is met: the maximum number of iterations T is reached max。 And the global optimal path no longer changes (converges).

[0121] S10527a, when cross-cluster execution is performed between the current rust spot cluster sub-region and the adjacent rust spot cluster sub-region, tasks within the remaining minor rust spot groups within the current rust spot cluster sub-region are obtained in advance in real time (a first minor rust spot area list is established), and then the sum of the remaining areas to be removed of the minor rust spots within the tasks within the minor rust spot groups is obtained in real time within the rust spot cluster sub-region (i.e., the rust spots are circular by default); the cooling time of the sum of the remaining areas to be removed of the current rust spot cluster sub-region is calculated, and the number of remaining rust spots within the current rust spot cluster sub-region is calculated;

[0122] S10528a, then executing a cooling time cross compensation process operation between the first laser head and the second backup laser head, specifically including:

[0123] When preparing for cross-cluster power switching, the cooling time of the total remaining area to be derusted and the number of remaining rust spots in the sub-area of the current rust spot cluster are obtained in real time;

[0124] When it is detected that the cooling time of the total remaining area to be rusted is less than the standard cross-cluster cooling time of the first laser head for the first time, and the number of remaining rust spots in the current rust cluster sub-area is less than the preset total threshold, the second backup laser head is inserted to execute the rust treatment task of the remaining slight rust spots corresponding to the remaining slight rust spots in the first slight rust area list (i.e., the cooling time of the total remaining area to be rusted), so as to compensate for the cooling time of the first backup laser head, and then let the first laser head enter the time adjustment program in advance, and then directly execute the scanning and rust removal operation of the next rust cluster sub-area after cooling.

[0125] Thus, analyzing the above technical solution, we can see that in Example 3 of the present invention, the backup laser head (second laser head) intervenes and takes over the remaining tasks when it detects that the main laser head (first laser head) needs to cross clusters in advance, thus avoiding the idleness of the main laser head due to cooling. This cross-compensation significantly shortens task interruptions and improves overall operational efficiency.

[0126] Based on the real-time data on the remaining rust area, number, and cooling time, the system can dynamically adjust the task allocation of the laser head to ensure that high-priority tasks (such as cross-cluster operations) are quickly connected and task switching delays are reduced. At the same time, entering the adjustment program in advance allows the main laser head sufficient time to cool down, extending its service life. This system realizes seamless cluster switching. After completing the core task of the current sub-area, the main laser head can enter the adjustment program and cool down in advance to prepare for the execution of the task of the next rust cluster sub-area. This "preheating" switching reduces the time overhead of cross-cluster operations and improves system continuity. Then, after the first laser head has finished cooling, it directly enters the next rust cluster sub-area to perform the rust removal task.

[0127] Example 4

[0128] A fourth embodiment of the present invention provides a system for detecting, identifying, and processing rust defects on a static arc contact workpiece, comprising:

[0129] Image acquisition module: It adopts hyperspectral imaging and polarization-hyperspectral fusion imaging technology, is equipped with an adaptive angle adjustment algorithm, and uses an improved image clarity evaluation formula and a polarization information fusion formula based on quantum state fusion to collect and screen images.

[0130] Image acquisition module: The module innovatively adopts hyperspectral imaging and polarization-hyperspectral fusion imaging technology, which can capture rich spectral information and polarization characteristics of the surface of the static arc contact workpiece. The adaptive angle adjustment algorithm can intelligently adjust the angle of the image acquisition device according to the shape and placement of the workpiece to ensure comprehensive and no-dead-angle shooting. The improved image clarity evaluation formula is used to quantitatively judge the clarity of the collected images and eliminate blurred images. At the same time, the polarization information fusion formula based on quantum state fusion efficiently fuses information of different polarization dimensions to screen out the most representative images. Through advanced imaging technology and intelligent algorithms, high-resolution images rich in information are obtained, which greatly improves the quality and usability of the images. Comprehensive image acquisition provides accurate and complete data support for subsequent rust spot detection and identification, reducing detection errors and omissions caused by image quality problems.

[0131] Image preprocessing module: Using the quantum denoising algorithm, it can specifically remove noise interference in the image and retain the key details of the image. At the same time, the image is enhanced by combining QGAN (quantum generative adversarial network) with a multi-scale residual network. The improved image enhancement coefficient formula can accurately adjust the degree and direction of enhancement according to the actual situation of the image. The QGAN processing effect evaluation formula based on quantum fidelity strictly evaluates the processed image to ensure that the image enhancement effect meets the requirements. After preprocessing, the noise in the image is effectively removed, and key information such as rust spots is significantly enhanced, which greatly improves the visual effect and data quality of the image. This provides a clearer and easier-to-analyze image for the subsequent rust spot feature extraction module, improving the accuracy and efficiency of feature extraction.

[0132] Rust Feature Extraction Module: Leveraging the strengths of QCNN (Quantum Convolutional Neural Network) and QACNN (Quantum Adaptive Convolutional Neural Network), combined with multi-scale feature fusion technology, rust features are extracted from different scales and angles. A feature significance evaluation formula based on information entropy quantitatively evaluates the extracted features and selects the most significant ones. An attention weight allocation formula based on the quantum state probability distribution rationally assigns weights based on feature importance, highlighting key features. This advanced feature extraction method can deeply explore the essential characteristics of rust spots and improve the accuracy and completeness of feature extraction. The screening and weight allocation mechanism highlights key features, reduces the interference of redundant information, and lays a solid foundation for accurate rust spot identification.

[0133] Rust Spot Recognition Module: This module uses a QSVM (Quantum Support Vector Machine) classifier to classify and identify rust spot features processed by the feature extraction module. An improved recognition accuracy evaluation formula rigorously evaluates recognition results from multiple dimensions to ensure accuracy and reliability. During the recognition process, the QSVM classifier quickly and accurately determines the type of rust spot based on the distribution and patterns of rust spot features. This module achieves high-precision rust spot recognition, quickly and accurately distinguishing different types of rust spots. Accurate recognition results provide a key basis for subsequent treatment decisions, facilitate targeted treatment measures, and improve the effectiveness and efficiency of rust spot treatment.

[0134] Processing decision module: Based on the type and severity of the identified rust spots, the ant colony algorithm is used to control the laser head to achieve precise rust removal operations.

[0135] Processing Execution Module: This module utilizes a combination of nanomaterial repair technology and laser rust removal technology to repair repairable rust spots. Nanomaterials can penetrate deep into rust spots, chemically reacting with the rusty material to achieve repair. Laser rust removal utilizes high-energy laser beams to precisely remove rust. Scrapped workpieces are processed using green recycling processes to ensure the rational use of resources and environmental protection. This module achieves efficient rust repair and environmentally friendly treatment of scrapped workpieces. The combination of these two advanced technologies improves the effectiveness and efficiency of rust repair. At the same time, the green recycling process embodies environmental protection, ensuring the quality of the workpiece while reducing the impact on the environment.

[0136] The Result Verification Module collects and analyzes hyperspectral images of the treated workpiece again. Using a treatment effect evaluation formula based on quantum state comparison, the treated image is compared with the pre-treatment image and the standard image. This precise quantum state comparison determines whether the rust has been effectively removed and whether the repair has met expectations. This module provides scientific and accurate verification of rust treatment results. Through rigorous image acquisition and analysis, problems in the treatment process can be promptly identified, providing feedback for further process optimization and quality improvement, ensuring that the quality of the static arc contact workpiece meets requirements.

[0137] The present invention also includes:

[0138] Sample library construction module: collect and label a large number of rust spot image samples, classify and label them using quantum clustering algorithm, and use improved sample quality assessment formula to screen high-quality samples for model training.

[0139] Real-time monitoring module: Use industrial cameras equipped with quantum imaging technology to collect images in real time, apply quantum reinforcement learning algorithms to update models and classifiers, and use real-time monitoring accuracy evaluation formulas based on quantum entanglement to evaluate the results.

[0140] The above are only preferred specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with this technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solutions and inventive concepts of the present invention, should be covered by the scope of protection of the present invention.

Claims

1. A method for detecting and identifying rust defects in static arc contact workpieces, characterized in that: The following steps are involved: A camera with hyperspectral imaging capability is used to capture images of the static arc contact workpiece from different angles. In addition to visible light and infrared light, it also captures ultraviolet spectrum information to construct a hyperspectral image dataset. The quantum denoising algorithm is used to denoise the collected images, and the generative adversarial network (GAN) and the multi-scale residual network (MSRN) are combined for image enhancement. GAN generates high-quality images, and MSRN enhances image features at different scales. The color, shape, and texture features of rust spots are extracted by combining quantum convolutional neural network (QCNN) with multi-scale feature fusion technology. The extracted rust spot features are classified and identified using a classifier based on quantum support vector machine (QSVM); According to the type and severity of the identified rust spots, the ant colony algorithm is used to control the laser head to achieve precise rust removal operations; The static arc contact parts are processed according to the treatment decision. For the repaired rust spots, a combination of nanomaterial repair technology and laser rust removal technology is used. Nanomaterials fill the tiny pores in the rust spots, and laser rust removal accurately removes the rust layer without damaging the substrate. For scrapped parts, a green recycling process is adopted. The treated static arc contact workpiece is subjected to hyperspectral image acquisition and analysis again, and the treatment effect evaluation formula based on quantum state comparison is used Among them, M diff is the difference measurement value of the quantum state before and after processing, and ∈ is the difference influence coefficient.

2. The method for detecting and identifying rust defects on a static arc contact workpiece according to claim 1, characterized in that: Also includes: Rust spot images of different types and severity of static arc contact parts are collected as samples, and the samples are classified and labeled using a quantum clustering algorithm. The improved sample quality evaluation formula is used. Among them H entropy is the information entropy of the sample, and ζ is the entropy influence coefficient.

3. The method for detecting and identifying rust defects on a static arc contact workpiece according to claim 1, characterized in that: Also includes: Real-time monitoring steps: During the static arc contact processing, an industrial camera with quantum imaging technology is used to collect images of the workpiece in real time. The deep learning model and classifier are continuously updated using the quantum reinforcement learning algorithm. A real-time monitoring accuracy evaluation formula based on quantum entanglement is used. Evaluate the real-time monitoring effect, where E entanglement is the quantum entanglement measurement value, and η is the entanglement influence coefficient.

4. The method for detecting and identifying rust defects in a static arc contact workpiece according to claim 1, characterized in that: In the image acquisition step, polarization-hyperspectral fusion imaging technology is used to obtain the polarization and hyperspectral information of the surface of the static arc contact workpiece, and the polarization information fusion formula F=λ×I based on quantum state fusion is used. p ×(1+θ×Q p )+(1-λ)×I m ×(1+θ×Q m ); Among them I p is the polarization image information, I m is the hyperspectral image information, Q p and Q m are the quantum state eigenvalues of polarization and hyperspectral information respectively, λ is the fusion coefficient, θ is the quantum state influence coefficient, and the two types of information are fused.

5. The method for detecting and identifying rust defects on a static arc contact workpiece according to claim 1, characterized in that: In the image preprocessing step, the quantum generative adversarial network QGAN is used to denoise and enhance the image; the QGAN processing effect evaluation formula based on quantum fidelity is used The quantum fidelity of the image before and after processing, where F ij is the pixel point (i, j), n is the number of rows in the image, and m is the number of columns in the image.

6. The method for detecting and identifying rust defects in a static arc contact workpiece according to claim 1, characterized in that: In the rust spot feature extraction step, the convolutional neural network QACNN with quantum attention mechanism is used to focus on the rust spot features through the superposition and entanglement of quantum states; the attention weight distribution formula based on the quantum state probability distribution is used. where f k is the value of the kth feature, Q k is the quantum state probability value of the kth feature, s is the number of features, and attention weights are assigned to different features.

7. The method for detecting and identifying rust defects on a static arc contact workpiece according to claim 1, characterized in that: Based on the type and severity of the identified rust spots, the ant colony algorithm is used to control the laser head to achieve precise rust removal operations, including: Initialize the current position of the first laser head and calculate the spatial distance from the current laser head to each rust spot; For all current rust spots, use the K-means clustering algorithm or the DBSCAN algorithm to perform cluster analysis based on severity, grouping rust spots with similar colors into a rust spot cluster sub-region, and then divide them into N rust spot cluster sub-regions in sequence; In the current rust cluster sub-area, the path planning algorithm is combined with the cooling time optimization of power switching to implement calculation processing to obtain the planned path within the cluster.

8. A system for implementing the method for detecting, identifying, and processing rust defects on a static arc contact workpiece according to any one of claims 1 to 7, characterized in that: include: Image acquisition module: This module uses hyperspectral imaging and polarization-hyperspectral fusion imaging technologies, equipped with an adaptive angle adjustment algorithm, and uses an improved image clarity evaluation formula and a polarization information fusion formula based on quantum state fusion to collect and filter images. Image preprocessing module: This module uses a quantum denoising algorithm, QGAN combined with a multi-scale residual network for denoising and enhancement, and uses an improved image enhancement coefficient formula and a QGAN processing effect evaluation formula based on quantum fidelity to evaluate the effect. Rust spot feature extraction module: Utilizes QCNN and QACNN combined with multi-scale feature fusion technology to extract rust spot features. It uses a feature significance evaluation formula based on information entropy and an attention weight allocation formula based on quantum state probability distribution to screen and assign weights. Rust spot recognition module: uses the QSVM classifier to classify and identify rust spot features, and uses an improved recognition accuracy evaluation formula to evaluate the recognition effect; processing decision module: uses the ant colony algorithm to control the laser head to achieve precise rust removal operations based on the type and severity of the identified rust spots; Processing execution module: Use a combination of nanomaterial repair technology and laser rust removal technology to repair rust spots, and use green recycling technology to treat scrapped workpieces; result verification module: Perform hyperspectral image acquisition and analysis on the treated workpieces again, and use the treatment effect evaluation formula based on quantum state comparison to verify the treatment effect.

Images