Method for optimizing laser-assisted grinding of ceramic matrix composites

Abstract

The application discloses a kind of ceramic matrix composite laser-assisted grinding process optimization methods, comprising the following steps: laser ablation modification is carried out on ceramic matrix composite by laser-assisted grinding;Surface and subsurface damage after grinding are identified based on machine vision, and the quantitative evaluation index including damage number, defect area ratio, defect dispersion, defect type entropy and dominant damage is established;With the lowest grinding damage as the goal, the mapping relationship between process parameters and damage index is established, and machine learning algorithm is used to optimize processing parameters.The ceramic matrix composite laser-assisted grinding process optimization method of the application realizes the scientific optimization of ceramic matrix composite laser-assisted grinding process parameters by constructing multi-dimensional damage quantitative evaluation system and combining intelligent optimization algorithm, which can significantly reduce grinding damage, improve processing efficiency and surface forming quality, and provides a systematic solution for high-precision, low-damage processing of ceramic matrix composite.

Description

Technical Field

[0001] This invention belongs to the field of ceramic matrix composite material processing technology, and more specifically, it relates to an optimization method for laser-assisted grinding of ceramic matrix composite materials based on intelligent damage quantification and process parameter optimization. Background Technology

[0002] Ceramic matrix composites possess excellent thermodynamic properties such as high temperature resistance, low density, high strength, and oxidation resistance, making them irreplaceable in key equipment fields such as hot-end components of aero-engines, thermal protection systems for hypersonic vehicles, structural components of space launch vehicles, and high-reliability friction braking systems. As a core material in high-end equipment manufacturing, the processing quality and forming precision of ceramic matrix composite components directly determine the operational stability, reliability, and service life of the equipment.

[0003] However, the inherent anisotropy, heterogeneity, and high hardness and brittleness of ceramic matrix composites make them highly susceptible to severe brittle fracture damage during conventional grinding processes. This damage includes surface / subsurface cracks, fiber breakage, fiber pull-out, and matrix fragmentation. Such processing damage not only significantly reduces the surface finish and dimensional accuracy of components but also substantially weakens their overall mechanical properties and structural integrity, thus severely limiting the application range and reliability of ceramic matrix composites in extremely harsh service environments.

[0004] Laser-assisted grinding is an effective means to improve the processing performance of ceramic matrix composites. This technology uses a laser to pre-ablate and modify the area to be processed, forming an oxide layer or softened layer of a specific depth on the material surface, which is then removed by grinding with an abrasive tool. Compared to traditional single grinding methods, laser-assisted grinding can significantly reduce grinding force, inhibit the generation and propagation of grinding damage, while improving processing efficiency and reducing abrasive tool wear.

[0005] However, existing laser-assisted grinding technology still faces the following key challenges in engineering applications.

[0006] First, the damage forms of ceramic matrix composites during grinding are complex and diverse, encompassing multiple scales and types of damage, from microcracks to macro-fiber fractures. Currently, the industry lacks a unified, standardized, and quantifiable processing quality evaluation standard, making it difficult to comprehensively, accurately, and quantitatively determine and assess the processing effects after laser-assisted grinding. This severely restricts the clarity of process optimization directions.

[0007] Secondly, the laser-assisted grinding process is influenced by a multitude of strongly coupled factors, including material properties (such as fiber orientation and matrix properties), laser irradiation conditions (such as power, frequency, and scanning strategy), grinding wheel properties (such as abrasive grain size and bond type), and grinding process parameters (such as feed rate, depth of cut, and spindle speed). In practical engineering applications, the lack of systematic and scientific process optimization methods makes it difficult to efficiently and accurately search for the optimal process combination within this multivariate and strongly coupled parameter space. Consequently, the low-damage and high-efficiency advantages of laser-assisted grinding cannot be maximized.

[0008] Therefore, developing a method that can scientifically evaluate processing damage and systematically optimize process parameters is of great theoretical significance and engineering application value for breaking through the technical bottleneck of high-precision, low-damage, and high-efficiency processing of ceramic matrix composites and ensuring their reliable service performance in high-end equipment. Summary of the Invention

[0009] In view of this, the purpose of this invention is to provide a method for optimizing the laser-assisted grinding process of ceramic matrix composites. By constructing a multi-dimensional quantitative evaluation index that includes damage quantity, area ratio, defect dispersion, defect type entropy, and dominant damage, and combining it with machine learning to establish a mapping relationship between process parameters and damage, the grinding parameters can be intelligently optimized, thereby significantly improving the processing quality and efficiency of ceramic matrix composites.

[0010] To achieve the above objectives, the present invention provides the following technical solution: An optimization method for laser-assisted grinding of ceramic matrix composites includes the following steps: Step 1: Laser-assisted grinding of ceramic matrix composites: The ceramic matrix composites are ablated and modified by laser to form an oxide layer on the material surface, and then the oxide layer is ground using a grinding tool; Step 2: Grinding damage identification and quantitative evaluation: Based on machine vision, the surface and subsurface damage characteristics of ceramic matrix composites after laser-assisted grinding are identified, and quantitative evaluation indicators for different types of damage are established. Step 3: Grinding process parameter optimization: Establish the mapping relationship between laser-assisted grinding process parameters and grinding damage. With the goal of minimizing grinding damage, optimize the processing parameters using machine learning algorithms.

[0011] Furthermore, in step one, the laser-assisted grinding method consists of the following steps: 11) Material pretreatment: Pretreatment of ceramic matrix composite samples is performed to ensure uniform surface condition during subsequent processing; 12) Laser treatment: Laser is used to ablate and modify the surface of the material. By setting the laser output parameters, spot parameters and scanning strategy, an oxide layer of a specific depth is formed on the surface of the material. 13) Grinding: The laser oxide layer is removed by changing the grinding wheel type, processing parameters and laser parameters.

[0012] Furthermore, in step two, the method for identifying and quantifying surface and subsurface damage based on machine vision includes the following steps: 21) Dataset Construction: Use detection equipment to collect images of surface and subsurface damage of materials after laser-assisted grinding, and label each damage feature to form standardized label samples to construct a dataset; 22) Model training: Using machine vision algorithm models, a smart recognition model for surface and subsurface damage of ceramic matrix composites is constructed to achieve automatic damage localization, classification and contour extraction; 23) Quantitative indicators: Establish quantitative evaluation indicators for surface damage, including the number of damages and the proportion of defect area; and establish quantitative evaluation indicators for subsurface damage, including the number of damages, the proportion of damage area, defect dispersion, defect type entropy, and dominant damage.

[0013] Furthermore, the defect dispersion is expressed as: in: Indicates the defect dispersion; The sample standard deviation, representing the total proportion of damages, reflects the degree of dispersion of damage distributions across multiple sample groups relative to the mean. The sample mean represents the total proportion of damages, characterizing the average level of damage distribution across multiple sample groups; The defect type entropy is represented as: in: Determines the entropy of the defect type; This represents the normalized area fraction of surface fiber fracture. The normalized area fraction of the fractured normal fiber; Let be the normalized area fraction of the crack; and: in: The pixel area of ​​the surface fiber fracture region; The pixel area of ​​the normal fiber fracture region; The pixel area of ​​the crack region; This represents the total pixel area.

[0014] Furthermore, the dominant damage includes surface fiber fracture dominance, normal fiber fracture dominance, and crack dominance, expressed as: in: , and These represent the surface fiber fracture dominance, normal fiber fracture dominance, and crack dominance, respectively.

[0015] Furthermore, in step three, the method for optimizing grinding process parameters is as follows: 31) Establishment of optimization algorithm model: Taking the minimum grinding damage as the optimization objective, a grinding parameter process optimization model is constructed by combining machine learning algorithms; 32) Optimal solution output: Set the algorithm parameters and output the optimal combination of process parameters; 33) Experimental verification: Process the sample according to the optimal parameters to verify its reliability.

[0016] Furthermore, in step 31), the method for establishing the optimization algorithm model is as follows: a BP neural network regression model is used to establish a nonlinear mapping relationship between laser-assisted grinding process parameters and machining damage, with feed rate as the basis. Grinding speed Laser scanning spacing As input layer features, with damage quantity Percentage of damaged area Defect dispersion Defect type entropy Surface fiber fracture dominance Normal fiber fracture dominance Crack dominance The output layer features are used; the hidden layer of the BP neural network regression model adopts a three-layer fully connected structure, with the number of neurons in the three fully connected layers being 64, 32, and 16 respectively. The ReLU function is selected as the activation function, and the mean squared error is used as the loss function and AdamW is used as the optimizer to complete the training.

[0017] Furthermore, based on the aforementioned BP neural network regression model, a hybrid intelligent optimization model is constructed by integrating the gray wolf optimization algorithm and the particle swarm optimization algorithm, and a weighted objective function is established with the goal of minimizing damage: in: For decision variables; This represents the number of damages under normal operating conditions. This represents the percentage of the damaged area under normal operating conditions. This represents the defect dispersion under normal operating conditions. This refers to the defect type entropy under normal operating conditions. , , , , , and These are the weighting coefficients, and .

[0018] Furthermore, in step 32), the method for outputting the optimal solution is as follows: using the BP neural network mapping model as a proxy model, the gray wolf optimization algorithm and the particle swarm optimization algorithm are fused to iteratively search for optimization in the constraint space, and the optimal combination of process parameters is calculated.

[0019] Furthermore, in step three, the established laser-assisted grinding process parameters include at least the feed rate, grinding speed, and laser scanning distance; the grinding damage includes at least the number of damages, the proportion of damage area, the defect dispersion, the defect type entropy, and the dominance damage.

[0020] The beneficial effects of this invention are as follows: The laser-assisted grinding process optimization method for ceramic matrix composites of the present invention achieves the following technical effects through a synergistic approach combining "laser modification," "visual quantification," and "intelligent optimization": (1) Significantly suppress processing damage and improve surface quality: By laser ablation modification, an easily removable oxide layer is formed on the material surface, which weakens the surface strength of the material and reduces grinding resistance from the processing mechanism level, effectively suppressing the generation and expansion of brittle damage such as surface and subsurface cracks and fiber breakage during grinding. (2) Establish a comprehensive and objective damage quantification evaluation system: In response to the problem of complex damage forms and diverse characteristics of ceramic matrix composites, a multi-dimensional quantitative index system covering damage quantity, defect area ratio, defect dispersion, defect type entropy and dominant damage was constructed. This system realizes the transformation of damage evaluation from single size measurement to comprehensive characterization of statistical characteristics such as distribution dispersion and type entropy. It can objectively and comprehensively reflect the degree and distribution characteristics of processing damage, and provide accurate and reliable data support and evaluation basis for process optimization. (3) Achieving scientific optimization and adaptive optimization of process parameters: By establishing a nonlinear mapping relationship between laser-assisted grinding process parameters and multi-dimensional damage indicators, with low damage as the optimization goal, the machine learning algorithm is integrated to efficiently iterate and optimize within the constrained space, outputting the optimal combination of process parameters, forming a closed-loop control of "detection", "quantification" and "optimization", effectively solving the problem that traditional trial and error methods cannot balance efficiency and quality, and significantly improving the controllability and consistency of the processing process. Under the premise of ensuring processing quality, the processing efficiency is maximized, and the wear of grinding tools and processing costs are reduced, providing reliable technical support for the precision manufacturing of key components in aerospace and other fields. Attached Figure Description

[0021] To make the objectives, technical solutions, and beneficial effects of this invention clearer, the following figures are provided for illustration: Figure 1 This is a flowchart of the laser-assisted grinding process optimization method for ceramic matrix composites according to the present invention; Figure 2 The experiment and results of laser-assisted grinding are shown in Figure 1; (a) is a diagram of the experimental process; (b) is the surface morphology after grinding. Figure 3 The images show the damage identification results; (a) shows the surface damage identification results; (b) shows the subsurface damage identification results. Figure 4 This is a comparison image of the surface after grinding process optimization and the surface before optimization. Detailed Implementation

[0022] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, so that those skilled in the art can better understand and implement the present invention. However, the embodiments described are not intended to limit the present invention.

[0023] This embodiment of the laser-assisted grinding process optimization method for ceramic matrix composites aims to address the problems of low processing quality and imperfect damage quantification evaluation system in existing technologies. By combining laser ablation modification with grinding and machine vision, intelligent identification and quantification of surface and subsurface damage are achieved. Simultaneously, a mapping relationship between process parameters and processing damage is constructed to optimize grinding parameters and processes, thereby improving the processing quality and efficiency of ceramic matrix composites and providing a systematic solution for their high-precision, low-damage, and high-efficiency processing.

[0024] Specifically, such as Figure 1 As shown in this embodiment, the method for optimizing the laser-assisted grinding process of ceramic matrix composites includes the following steps.

[0025] Step 1: Laser-assisted grinding. A laser is used to ablate and modify the ceramic matrix composite material, forming an oxide layer of a specific depth on the material surface. Then, an abrasive tool is used to grind this oxide layer, completing the surface treatment. This process removes material while minimizing grinding damage. Figure 2 As shown in (a).

[0026] Specifically, in this embodiment, the steps of the laser-assisted grinding method are as follows.

[0027] 11) Material pretreatment: The ceramic matrix composite sample is pretreated, including but not limited to material size, quantity, and original surface, to ensure uniform surface condition during subsequent processing. In this embodiment, 10 ceramic matrix composite samples with dimensions of 10 mm × 20 mm × 5 m (length × width × height) are processed.

[0028] 12) Laser Processing: Laser ablation modification of the material surface is performed. By setting laser output parameters, spot parameters, and scanning strategies, an oxide layer of a specific depth is formed on the material surface, achieving material modification in the area to be processed. In this embodiment, a picosecond laser was used for the experiment. The laser emitter's emission frequency was 50 kHz, the laser spot overlap rate was 67%, and the pulse laser density was 13.81–34.54 J / cm². 2 The laser power was set to 12.2 W, the laser spot diameter at the focal point was 30 μm, the minimum ablation linewidth was 15–20 μm, the scanning speed was 500 mm / s, the number of scans was 800 (refocusing every 400 scans), and the scanning mode was parallel scanning.

[0029] 13) Grinding: By changing the grinding wheel type, processing parameters, and laser parameters, the laser-induced oxide layer is removed, ensuring material removal efficiency while minimizing the generation and propagation of grinding damage. In this embodiment, an electroplated diamond belt with a grit size of 120# and a contact width of 10 mm is used to grind the laser-induced ablation layer on a laser-abrasive belt co-processing equipment. Feed rate Four parameters were selected: 70 mm / min, 100 mm / min, 130 mm / min, and 160 mm / min. Grinding speed Four parameters were set: 1500 r / min (11.8 m / s), 1800 r / min (14.2 m / s), 2100 r / min (16.5 m / s), and 2400 r / min (18.9 m / s). Additionally, the laser scanning interval was... Four parameters were also set: 80 μm, 130 μm, 180 μm, and 230 μm. The grinding depth was determined based on the maximum ablation depth of the laser.

[0030] Step Two: Grinding Damage Identification and Quantitative Evaluation. Based on machine vision, the surface and subsurface damage characteristics of the ceramic matrix composite material after laser-assisted grinding are identified, and quantitative evaluation indicators for different types of damage are established. Specifically, in this embodiment, an intelligent recognition algorithm is applied to classify and identify surface fiber fractures, normal fiber fractures, cracks, and other damage on the surface and subsurface of the ceramic matrix composite material, and a quantitative evaluation index system is established to quantify the degree of damage. The specific steps of the machine vision-based surface and subsurface damage identification and quantitative evaluation method are as follows.

[0031] 21) Dataset Construction: Using detection equipment, images of material surface and subsurface damage covering multiple scales and working conditions after laser-assisted grinding are collected, and relevant annotation software is used to accurately annotate each damage feature to form standardized label samples in order to construct an accurate and rich dataset.

[0032] In this embodiment, scanning electron microscopy (SEM) is used to acquire images of the surface / subsurface, with no fewer than 400 images in different background scenes. The dataset will then be expanded to 2000 images through data augmentation.

[0033] In this embodiment, Labelme software is used to manually annotate damage such as surface fiber fractures, normal fiber fractures, and cracks. For approximately circular normal fiber fractures, a circular tool is used to draw along the actual boundary to ensure that the diameter strictly matches the actual size of the fracture. For surface fiber fractures with complex contours, a polygon mode is used to approximate the fracture point by point, and vertices are appropriately added in areas with drastic curvature changes to ensure fitting accuracy. For linear or irregularly expanding cracks, a polygon tool is used to annotate along the actual crack direction and edge to completely cover the crack area and accurately reflect the crack length and morphological characteristics. After drawing, COCO format JSON tags are generated.

[0034] 22) Model training: Using machine vision algorithm models, a smart recognition model for surface and subsurface damage of ceramic matrix composites is constructed to realize automatic damage localization, classification and contour extraction.

[0035] This embodiment employs machine vision algorithms such as YOLOv8, using the AdamW optimizer during training and a composite loss function combining classification and regression losses. The training epochs are set to 450, the batch size to 32, and the input image resolution to a uniform 640×640. The initial learning rate is set to 0.0003, with a learning decay factor of 0.05 to ensure the stability and accuracy of training convergence. The weight decay coefficient is set to 0.0001 to suppress overfitting. During the model inference phase, the confidence threshold is set to 0.35, and the intersection-over-union (IoU) threshold is set to 0.62 to ensure the accuracy and robustness of damage identification and localization, enabling correct identification and classification of surface / subsurface damage. Meanwhile, during training, mean precision (mAP@0.5, mAP@0.5-0.95), precision, recall, training loss, and validation loss are used as core evaluation metrics to monitor and validate model performance in real time, ensuring stable model convergence and reliable recognition performance.

[0036] 23) Quantitative indicators: Establish quantitative evaluation indicators for surface damage, including the number of damages and the proportion of defect area; and establish quantitative evaluation indicators for subsurface damage, including the number of damages, the proportion of damage area, defect dispersion, defect type entropy, and dominant damage.

[0037] Specifically, surface damage evaluation indicators mainly include the number of damages (N) and the percentage of defect area (A%). Subsurface damage indicators mainly include the number of damages (N), the percentage of damage area (A%), the defect dispersion (CV), and the defect type entropy. Damage is categorized by its dominance (including surface fiber fracture-dominated FFD, normal fiber fracture-dominated NFFD, and crack-dominated CFD). Damage dispersion (CV) is a core indicator characterizing the spatial uniformity of damage distribution on the subsurface. It quantifies the degree of aggregation of damage regions on the material. A larger CV value indicates that the damage is more concentrated in a local area; a smaller CV value indicates a more uniform spatial distribution of damage. The defect dispersion CV index is calculated by the ratio of the sample standard deviation to the sample mean, eliminating the influence of damage quantity and size, and enabling horizontal comparison of damage distribution characteristics between different samples. The formula is as follows: in: Indicates the defect dispersion; The sample standard deviation, representing the total proportion of damages, reflects the degree of dispersion of damage distributions across multiple sample groups relative to the mean. The sample mean represents the total proportion of damages, characterizing the average level of damage distribution across multiple sample groups.

[0038] Damage type entropy It is a key indicator for quantifying the uniformity of subsurface damage type distribution. Its core significance lies in reflecting the degree of balance in the composition of various damage types. The larger the value, the more balanced the distribution of different damage types, and the absence of a clear dominant damage type; The smaller the value, the more concentrated the damage types are, indicating that one or a few types of damage dominate. This index requires first normalizing to obtain the area fraction of each type of damage, using the total pixel area of ​​all damage types as the basis for the calculation. Based on this, the normalized area fraction of various types of damage is calculated using the following formula: in: This represents the normalized area fraction of surface fiber fracture. The normalized area fraction of the fractured normal fiber; This represents the normalized area fraction of the crack. The pixel area of ​​the surface fiber fracture region; The pixel area of ​​the normal fiber fracture region; The pixel area of ​​the crack region; This represents the total pixel area.

[0039] Based on the normalized area fractions of the above-mentioned types of damage, the damage type entropy The calculation formula is: in: This represents the defect type entropy.

[0040] Defect Type Entropy When these indicators are combined with the proportion of damaged area and the CV of defect dispersion, the three-dimensional properties of damage type, coverage, and distribution characteristics can be revealed more comprehensively, providing more systematic quantitative support for material quality assessment and performance optimization.

[0041] when A large value indicates that the material is simultaneously subjected to the combined effects of three types of damage: surface fiber fracture, normal fiber fracture, and cracking. This only reflects the distribution balance of damage types and cannot clearly identify which type of damage has a dominant impact on material properties. To overcome this deficiency and accurately analyze the core failure factors under conditions of multiple coexisting damage, three types of damage dominance indices are proposed: surface fiber fracture dominance (FFD), normal fiber fracture dominance (NFFD), and crack dominance (CFD). Since the dominance indices are essentially a direct mapping of the area proportion of each type of damage, they are expressed by quantifying the area proportion weight of a single damage type, i.e., the normalized area fraction of each of the above damage types, as shown in the following formula: in: , and These represent the surface fiber fracture dominance, normal fiber fracture dominance, and crack dominance, respectively.

[0042] The core advantage of the dominance index lies in its intuitiveness and targeting. The higher the area ratio of the damage type, the greater its dominance value, which can be directly identified as the core factor affecting material performance, providing a clear direction for engineering optimization without the need for complex calculations.

[0043] Step 3: Grinding Process Parameter Optimization. Establish a mapping relationship between laser-assisted grinding process parameters and grinding damage. With the goal of minimizing grinding damage, optimize the processing parameters using machine learning algorithms. In this embodiment, the established laser-assisted grinding process parameters include at least feed rate, grinding speed, and laser scanning interval; grinding damage includes at least the number of damages, damage area ratio, defect dispersion, defect type entropy, and dominant damage.

[0044] 31) Establishment of optimization algorithm model: Taking the minimum grinding damage as the optimization objective, a grinding parameter process optimization model is constructed by combining machine learning algorithms; In this embodiment, the optimization algorithm model is established by using a BP neural network regression model to establish a nonlinear mapping relationship between laser-assisted grinding process parameters and machining damage. The model uses the feed rate... Grinding speed Laser scanning spacing As input layer features, with damage quantity Percentage of damaged area Defect dispersion Defect type entropy Surface fiber fracture dominance Normal fiber fracture dominance Crack dominance The output layer features are used; the hidden layers of the network adopt a three-layer fully connected structure with 64, 32 and 16 neurons respectively. The ReLU function is used as the activation function, and the mean squared error is used as the loss function and AdamW is used as the optimizer to complete the training.

[0045] Based on the nonlinear mapping relationship, the gray wolf optimization algorithm and the particle swarm optimization algorithm are integrated to construct a hybrid intelligent optimization model, and a weighted objective function is established with the goal of minimizing damage: in: The parameters are decision variables and satisfy the processing condition: 70 mm / min ≤ ≤200mm / min; 12 m / s≤ ≤22 m / s; 70 μm≤ D ≤250 μm; This represents the number of damages under normal operating conditions. This represents the percentage of the damaged area under normal operating conditions. This represents the defect dispersion under normal operating conditions. This refers to the defect type entropy under normal operating conditions. , , , , , and These are the weighting coefficients, and .

[0046] 32) Optimal solution output: Set the algorithm parameters and output the optimal combination of process parameters.

[0047] In this embodiment, the BP neural network mapping model is used as a surrogate model, and the Gray Wolf optimization algorithm and the Particle Swarm Optimization algorithm are fused to iteratively optimize within the constrained space to calculate the optimal combination of process parameters. Specifically, the Gray Wolf optimization algorithm is configured with a population size of 10, a maximum number of iterations of 10, and an exploration factor a of 0.2; the Particle Swarm Optimization algorithm is configured with a population size of 15, a maximum number of iterations of 10, an inertia weight w of 0.5, individual cognitive coefficient c1 and social cognitive coefficient c2 both of 1.5, a mutation rate of 0.2, and a crossover rate of 0.7, to balance the local development capability and global exploration capability of the algorithm. Based on the above parameter configuration, the BP neural network mapping model is used as a surrogate model, and the Gray Wolf optimization and Particle Swarm Optimization algorithms are fused to iteratively optimize within the constrained space to calculate the optimal combination of process parameters.

[0048] 33) Experimental verification: Process the sample according to the optimal parameters and set conventional grinding as the control group. Test the surface roughness, surface and subsurface micromorphology and fatigue performance, etc., to verify its reliability.

[0049] The technical effects of the laser-assisted grinding process optimization method for ceramic matrix composites of the present invention are explained below through sample characterization and other results.

[0050] (1) Surface morphology analysis after laser-assisted grinding.

[0051] During the laser ablation stage, SiC fibers and the matrix absorb high energy, sublimate, and react with oxygen to generate a large amount of SiO2 oxide. Furthermore, the ceramic matrix composite material continuously absorbs energy under laser irradiation, forming an inverted conical groove with a narrow bottom and a wide top. In traditional grinding, the removal mechanism of composite materials mostly consists of brittle fracture, with fibers exhibiting large-area, layered, brittle spalling accompanied by fiber pull-out. Figure 2 (b) shows the microscopic surface morphology after laser-assisted grinding. It was found that material removal was mainly fiber crushing, fiber breakage and interface damage. A coating layer was generated at high feed and grinding speeds, which made the processed surface smooth.

[0052] (3) Analysis of damage identification results. like Figure 3 The image shows the damage recognition results from machine vision. The YOLOv8 algorithm is used for intelligent recognition and feature extraction of surface and subsurface damage in ceramic matrix composites after grinding, enabling accurate localization and efficient classification of various types of damage. Figure 3 (a) shows the results of material surface damage identification. Figure 3(b) shows the results of subsurface damage identification. As can be seen from the figure, the YOLOv8 algorithm has excellent identification performance for both surface and subsurface damage in ceramic matrix composites. The accuracy of surface damage identification can reach about 90%, and the accuracy of subsurface damage identification can reach about 95%. The damage boundaries are clear, the location is accurate, and the classification is reliable. It can completely and realistically reflect the characteristics of surface and subsurface damage of the material, and can provide accurate and stable data support for subsequent damage quantification evaluation and process parameter optimization.

[0053] (3) Optimize the results analysis.

[0054] Laser-induced ablation-assisted grinding micro-surfaces after optimization of processing parameters, such as Figure 4 As shown in the figure, by comparing the surfaces of the original control group and those treated with optimized processing parameters, a significant improvement in surface morphology can be observed. In the original control group, the microscopic surface after grinding exhibited obvious fiber breakage, interface layer peeling, cracks, and a large amount of fiber breakage, resulting in poor surface quality. In contrast, although some degree of fiber breakage and interface layer peeling still occurred with the optimized processing parameters, the frequency and severity of these damages were significantly reduced. Further analysis revealed that under the optimized processing parameters, fiber crushing occurred during grinding, forming numerous flat areas; this feature was almost absent in the original control group. The formation of these flat areas can be attributed to the fact that the optimized processing parameters effectively reduced the cutting force and heat during grinding, thereby mitigating the degree of fiber damage.

[0055] (4) Technical effects.

[0056] The laser-assisted grinding process optimization method for ceramic matrix composites in this embodiment has significant technical advantages and outstanding application value compared with existing technologies. Firstly, this embodiment employs laser ablation modification and abrasive grinding. Laser treatment controllably modifies the surface of the ceramic matrix composite, forming an oxide layer structure that is easily removed by grinding. This effectively weakens the surface strength of the material and reduces grinding resistance, thereby significantly inhibiting the generation and propagation of surface and subsurface damage during processing. This improves the grinding characteristics of ceramic matrix composites from a processing mechanism perspective, enhancing the stability of processing quality. Secondly, this embodiment uses machine vision to achieve intelligent identification of grinding damage in ceramic matrix composites. It can accurately locate and classify surface and subsurface damage characteristics, constructing a quantitative characterization method covering multiple types of damage. This forms a complete and standardized damage quantitative evaluation system, objectively and comprehensively reflecting the degree of processing damage, and providing reliable data support and evaluation basis for process optimization. Meanwhile, this embodiment establishes a mapping relationship between laser-assisted grinding process parameters and processing damage. With low damage as the optimization goal, machine learning algorithms are used to scientifically optimize the process parameters, outputting optimal parameter combinations suitable for engineering applications. This improves the controllability and consistency of the processing, reduces processing costs, extends the service life of the grinding wheel, and enhances both processing quality and efficiency. The overall solution of this embodiment can meet the manufacturing requirements of key fields such as aerospace for high-precision, high-reliability, and long-service-life ceramic matrix composite components.

[0057] In summary, the laser-assisted grinding process optimization method for ceramic matrix composites in this embodiment systematically solves the problems of low processing quality and imperfect damage quantification evaluation system in the prior art by organically combining laser-assisted grinding, intelligent damage identification and quantification, and process parameter optimization. It significantly improves the processing level and application potential of ceramic matrix composites and has outstanding innovation, practicality, and engineering promotion value.

[0058] The above-described embodiments are merely preferred embodiments provided to fully illustrate the present invention, and the scope of protection of the present invention is not limited thereto. Equivalent substitutions or modifications made by those skilled in the art based on the present invention are all within the scope of protection of the present invention. The scope of protection of the present invention is defined by the claims.

Claims

1. A method for optimizing laser-assisted grinding processes of ceramic matrix composites, characterized in that: Includes the following steps: Step 1: Laser-assisted grinding of ceramic matrix composites: The ceramic matrix composites are ablated and modified by laser to form an oxide layer on the material surface, and then the oxide layer is ground using a grinding tool; Step 2: Grinding damage identification and quantitative evaluation: Based on machine vision, the surface and subsurface damage characteristics of ceramic matrix composites after laser-assisted grinding are identified, and quantitative evaluation indicators for different types of damage are established. Step 3: Grinding process parameter optimization: Establish the mapping relationship between laser-assisted grinding process parameters and grinding damage. With the goal of minimizing grinding damage, optimize the processing parameters using machine learning algorithms.

2. The method for optimizing laser-assisted grinding process of ceramic matrix composites according to claim 1, characterized in that: In step one, the laser-assisted grinding method consists of the following steps: 11) Material pretreatment: Pretreatment of ceramic matrix composite samples is performed to ensure uniform surface condition during subsequent processing; 12) Laser treatment: Laser is used to ablate and modify the surface of the material. By setting the laser output parameters, spot parameters and scanning strategy, an oxide layer of a specific depth is formed on the surface of the material. 13) Grinding: The laser oxide layer is removed by changing the grinding wheel type, processing parameters and laser parameters.

3. The method for optimizing laser-assisted grinding process of ceramic matrix composites according to claim 1, characterized in that: In step two, the method for identifying and quantifying surface and subsurface damage based on machine vision consists of the following steps: 21) Dataset Construction: Use detection equipment to collect images of surface and subsurface damage of materials after laser-assisted grinding, and label each damage feature to form standardized label samples to construct a dataset; 22) Model training: Using machine vision algorithm models, a smart recognition model for surface and subsurface damage of ceramic matrix composites is constructed to achieve automatic damage localization, classification and contour extraction; 23) Quantitative indicators: Establish quantitative evaluation indicators for surface damage, including the number of damages and the proportion of defect area; and establish quantitative evaluation indicators for subsurface damage, including the number of damages, the proportion of damage area, defect dispersion, defect type entropy, and dominant damage.

4. The method for optimizing laser-assisted grinding process of ceramic matrix composites according to claim 3, characterized in that: The defect dispersion is expressed as: in: Indicates the defect dispersion; The sample standard deviation, representing the total proportion of damages, reflects the degree of dispersion of damage distributions across multiple sample groups relative to the mean. The sample mean represents the total proportion of damages, characterizing the average level of damage distribution across multiple sample groups; The defect type entropy is represented as: in: Determines the entropy of the defect type; This represents the normalized area fraction of surface fiber fracture. The normalized area fraction of the fractured normal fiber; Let be the normalized area fraction of the crack; and: in: The pixel area represents the region of fiber breakage on the surface. The pixel area of ​​the normal fiber fracture region; The pixel area of ​​the crack region; This represents the total pixel area.

5. The method for optimizing laser-assisted grinding process of ceramic matrix composites according to claim 4, characterized in that: The dominant damage includes surface fiber fracture dominance, normal fiber fracture dominance, and crack dominance, expressed as: in: , and These represent the surface fiber fracture dominance, normal fiber fracture dominance, and crack dominance, respectively.

6. The method for optimizing laser-assisted grinding process of ceramic matrix composites according to claim 1, characterized in that: In step three, the method for optimizing grinding process parameters is as follows: 31) Establishment of optimization algorithm model: Taking the minimum grinding damage as the optimization objective, a grinding parameter process optimization model is constructed by combining machine learning algorithms; 32) Optimal solution output: Set the algorithm parameters and output the optimal combination of process parameters; 33) Experimental verification: Process the sample according to the optimal parameters to verify its reliability.

7. The method for optimizing laser-assisted grinding process of ceramic matrix composites according to claim 6, characterized in that: In step 31), the method for establishing the optimization algorithm model is as follows: a BP neural network regression model is used to establish a nonlinear mapping relationship between laser-assisted grinding process parameters and machining damage, with feed rate as the key factor. Grinding speed Laser scanning spacing As input layer features, with damage quantity Percentage of damaged area Defect dispersion Defect type entropy Surface fiber fracture dominance Normal fiber fracture dominance Crack dominance The output layer features are used; the hidden layer of the BP neural network regression model adopts a three-layer fully connected structure, with the number of neurons in the three fully connected layers being 64, 32, and 16 respectively. The activation function is the ReLU function, and the training is completed using the mean squared error as the loss function and AdamW as the optimizer.

8. The method for optimizing laser-assisted grinding process of ceramic matrix composites according to claim 7, characterized in that: Based on the BP neural network regression model, a hybrid intelligent optimization model is constructed by integrating the gray wolf optimization algorithm and the particle swarm optimization algorithm, and a weighted objective function is established with the goal of minimizing damage: in: For decision variables; This represents the number of damages under normal operating conditions. This represents the percentage of the damaged area under normal operating conditions. This represents the defect dispersion under normal operating conditions. This refers to the defect type entropy under normal operating conditions. , , , , , and These are the weighting coefficients, and .

9. The method for optimizing laser-assisted grinding process of ceramic matrix composites according to claim 7, characterized in that: In step 32), the optimal solution is output by using the BP neural network mapping model as a proxy model, integrating the gray wolf optimization algorithm and the particle swarm optimization algorithm to iteratively search for optimization in the constraint space, and calculating the optimal combination of process parameters.

10. The method for optimizing laser-assisted grinding process of ceramic matrix composites according to claim 1, characterized in that: In step three, the established laser-assisted grinding process parameters include at least the feed rate, grinding speed, and laser scanning spacing; the grinding damage includes at least the number of damages, the proportion of damage area, the defect dispersion, the defect type entropy, and the dominant damage.

Images