A method of preparing a composite coating for a back mold insert

Abstract

The present application relates to a kind of preparation methods of post mould insert composite coating in the field of mould accessory manufacturing, comprising: extracting grain orientation and interface bonding state information from initial transition layer surface, using ion beam scanning technology to analyze grain distribution uniformity and interface stress, determine coating bonding force basic parameter;According to microdefect distribution data, the influence degree of defect on thermal fatigue resistance is analyzed by combining preset threshold range, and the real-time evaluation result of coating performance is obtained;Improved gradient composite coating is generated by optimized deposition process, and the change trend of microdefect is scanned again using online detection system, to determine the improvement range of coating durability and production efficiency;Key performance indicators are extracted from simulation results, combined with the data log of online detection system, process parameters are iteratively optimized by gradient descent algorithm, and the final stable coating preparation scheme is obtained.

Description

Technical Field

[0001] The invention relates to the field of mold parts manufacturing, and in particular to a method for preparing a composite coating of a rear mold insert. Background Art

[0002] In mold production, back-mold inserts, as core components, directly determine the mold's service life and product quality. The quality of their surface coating technology has a crucial impact on production efficiency and economic benefits. As die-casting processes evolve toward higher temperatures, higher pressures, and higher speeds, coatings must withstand extreme operating conditions, including thermal shock, mechanical wear, and molten metal erosion. Traditional surface treatment methods are no longer able to meet these increasingly demanding performance requirements. Therefore, the development of high-performance coatings, as well as their preparation and testing technologies, has become a key focus for improving the reliability and intelligence of die-casting molds.

[0003] Currently, while commonly used surface treatment techniques such as nitriding and chrome plating can improve the surface hardness of inserts to a certain extent, coating peeling and insufficient thermal fatigue resistance are common problems. Furthermore, the detection of defects such as microcracks and porosity within the coating relies on offline metallographic analysis, which is not only time-consuming and inefficient, but also lacks timely feedback on process defects, making quality control during production more difficult. These limitations restrict the stability and durability of back-mold inserts under complex operating conditions. The core challenges in this research area focus on achieving high coating adhesion, high density, and real-time defect monitoring. Traditional processes struggle to form a strong interface between the substrate and the coating, making the coating susceptible to peeling during thermal cycling. Furthermore, the disordered grain orientation and accumulation of interface defects during coating preparation further reduce its thermal fatigue resistance. Furthermore, offline microdefect detection methods cannot form a closed-loop optimization loop with the preparation process, resulting in delayed process parameter adjustments and impacting yield. These technical challenges collectively constitute a bottleneck in improving the performance of back-mold insert coatings.

[0004] Therefore, how to construct a high-bonding, high-density gradient composite coating through a multi-energy field synergistic deposition process, and integrate it with an online detection system to identify micro-defects in real time and provide feedback for process optimization, has become a key issue in improving the durability and production efficiency of rear mold inserts. Solving this problem requires not only breakthroughs in interface strengthening and structural control technologies in coating preparation, but also the dynamic coupling of detection and process technology to provide a reliable solution for the die-casting industry. Summary of the Invention

[0005] The present invention provides a method for preparing a composite coating of a rear mold insert, comprising the following steps:

[0006] Obtain surface data of the rear mold insert substrate, apply plasma field and thermal field synergistically through multi-energy field deposition equipment to deposit an initial transition layer on the substrate surface, and obtain a composite coating prototype with a gradient structure;

[0007] Extract the grain orientation and interface bonding state information from the initial transition layer surface, use ion beam scanning technology to analyze the grain distribution uniformity and interface stress, and determine the basic parameters of coating bonding strength;

[0008] Based on the basic parameters of coating adhesion, the deposition rate and energy input are adjusted using a multi-energy field deposition device to continue depositing a high-density functional layer on the transition layer to obtain an enhanced gradient composite coating.

[0009] An online detection system is deployed on the surface of the enhanced gradient composite coating to obtain micro-defect distribution data inside the coating through laser interferometry and ultrasonic scanning to determine the specific location and scale of microcracks and pores;

[0010] Based on the micro-defect distribution data and the preset threshold range, the impact of defects on thermal fatigue resistance is analyzed to obtain real-time evaluation results of coating performance;

[0011] If the evaluation result is lower than the preset performance standard, the defect distribution characteristics are extracted from the online detection system and fed back to the multi-energy field deposition equipment to adjust the deposition temperature and ion energy to optimize the density and adhesion of the next batch of coatings;

[0012] An improved gradient composite coating was generated through an optimized deposition process. The online inspection system was used to scan the micro-defect trends again to determine the extent of improvements in coating durability and production efficiency.

[0013] Obtain thermal cycling test data for the improved gradient composite coating, use finite element analysis to simulate the stress distribution of the coating under high temperature and high pressure conditions, and determine the actual improvement level in thermal fatigue resistance and interface strengthening;

[0014] Key performance indicators are extracted from the simulation results and combined with the data logs of the online detection system. The process parameters are iteratively optimized through the gradient descent algorithm to obtain the final stable coating preparation solution.

[0015] The technical solution provided by the embodiment of the present invention may have the following beneficial effects:

[0016] The present invention discloses a method for preparing a composite coating for a rear mold insert. The method forms a composite coating with a gradient structure on the surface of a substrate through a multi-energy field deposition device, and uses an online detection system to monitor the coating quality in real time. The present invention first deposits an initial transition layer on the surface of the substrate, and then analyzes the grain distribution and interface stress to determine the coating adhesion parameters. Based on these parameters, the deposition process is adjusted to form a high-density functional layer. The present invention innovatively uses laser interference and ultrasonic scanning technology to detect micro-defects in the coating, and dynamically optimizes the deposition parameters based on the detection results. Through an iterative optimization process, the present invention can continuously improve the coating performance and enhance its thermal fatigue resistance and durability. Finally, the present invention verifies the coating performance through thermal cycle testing and finite element analysis, and uses a machine learning algorithm to optimize the process parameters to achieve high-quality and high-efficiency gradient composite coating preparation. DETAILED DESCRIPTION

[0017] The following will be combined with the embodiments of the present invention to clearly and in detail describe the technical solutions in the embodiments of the present invention. The described embodiments are only part of the embodiments of the present invention.

[0018] The preparation method of a composite coating for a rear mold insert in this embodiment may specifically include:

[0019] Step S1, obtaining the surface data of the rear mold insert substrate, applying the plasma field and thermal field synergistically through the multi-energy field deposition equipment, depositing an initial transition layer on the substrate surface, and obtaining a composite coating prototype with a gradient structure.

[0020] Acquire characteristic data of the substrate surface to obtain a surface state distribution; apply plasma field treatment to the surface state distribution using a multi-energy field deposition device to obtain a plasma-activated surface; perform thermal field modulation on the plasma-activated surface to obtain a thermal field modulated surface; if the uniformity of the thermal field modulated surface reaches a preset threshold, deposit a transition layer using a deposition device to obtain an initial coating structure; obtain gradient distribution data of the initial coating structure; adjust field synergy parameters based on the gradient distribution data to obtain an optimized composite coating prototype; perform surface detection on the optimized composite coating prototype to obtain final coating data.

[0021] Exemplarily, characteristic data is extracted from the substrate surface by a sensor to obtain a surface state distribution.

[0022] It is understandable that the core of this step lies in accurately sensing the physical properties of the substrate surface.

[0023] For example, assuming that the substrate is a metal part, the sensor can use a laser scanner to scan the surface roughness. The data obtained shows that the roughness of a certain area is Ra2.5 microns, while that of another area is Ra1.8 microns. This distribution reflects the unevenness of the surface state and provides a basis for subsequent processing. The beneficial effect of doing so is that it can lay a data foundation for customized surface treatment and avoid blind operation. A multi-energy field deposition device is used to apply plasma field treatment to the surface state distribution to obtain a plasma-activated surface.

[0024] Specifically, multi-energy field devices may integrate radio frequency discharge devices to generate plasma to bombard the surface.

[0025] In one possible implementation, nitrogen is used as the working gas, the pressure is controlled at 0.5 Pa, and the power is set to 200 W. After 10 minutes of plasma action, the surface active sites increase and the chemical bond energy is stimulated. The benefit of this step is that it increases the reactivity of the surface and creates conditions for subsequent energy injection. Energy is injected into the plasma-activated surface through the action of a thermal field, resulting in a thermal field modulated surface.

[0026] It should be noted that the thermal field can be achieved through resistance heating.

[0027] For example, by placing the substrate in a hot environment of 300 degrees Celsius and keeping it warm for 15 minutes, the surface molecules will rearrange themselves due to the thermal energy, and the uniformity will be improved.

[0028] Preferably, if microcracks exist on the initial surface, the thermal field can partially repair them, improving structural stability. This treatment provides a more uniform substrate for subsequent coating deposition. If the thermal field modulated surface uniformity reaches a predetermined threshold, a transition layer is deposited using a deposition device, yielding the initial coating structure.

[0029] In one embodiment, the uniformity threshold is set at a roughness variation of less than 0.3 microns. At this point, a 50-nanometer titanium transition layer can be deposited using magnetron sputtering equipment. The titanium layer, due to its excellent adhesion and corrosion resistance, effectively connects the substrate to the outer coating. This design enhances the adhesion between the coating and the substrate, extending its service life. Gradient analysis is performed on the initial coating structure to generate gradient distribution data.

[0030] For example, scanning electron microscopy (SEM) cross-section observations reveal a smooth transition in elemental concentration at the interface between the titanium layer and the substrate, while concentrations in the outer regions change abruptly. This data reflects the layered nature of the coating structure and provides a basis for optimization. This beneficial effect clarifies the direction of adjustment and avoids imbalances in coating performance. Based on this gradient distribution data, field synergy parameters are adjusted to produce a prototype of an optimized composite coating.

[0031] Specifically, if the outer layer is found to be insufficiently hard, the bias voltage during deposition can be increased to -100V, while the operating pressure can be raised to 1Pa, resulting in a harder titanium nitride layer. This adjustment achieves a balance between hardness and toughness, improving the coating's wear resistance. Surface inspection of the optimized composite coating prototype is performed using image processing technology to obtain final coating data.

[0032] In one possible implementation, an industrial camera captures surface images and, combined with an edge detection algorithm, analyzes coating defects. Micropores as small as 10 microns in diameter were discovered in a specific area. Further process optimization reduced the defects to negligible levels. This type of inspection ensures coating quality and meets the demands of high-precision applications.

[0033] It is understandable that from feature extraction to final inspection, each step is closely linked to each other and jointly supports the realization of high-quality coatings.

[0034] For example, the initial surface state distribution directly affects the depth of plasma treatment, while the uniformity of thermal field modulation determines the adhesion of the transition layer. This multi-faceted collaborative logic ensures the rigor of the process and the reliability of the results.

[0035] Step S2: extracting grain orientation and interface bonding state information from the surface of the initial transition layer, analyzing grain distribution uniformity and interface stress using ion beam scanning technology, and determining basic coating bonding parameters.

[0036] The grain orientation information and interface bonding information on the transition layer surface are obtained, and the distribution data of the grains and the stress distribution data of the interface are obtained by using ion beam scanning technology; the uniformity of the grain distribution is judged according to a preset threshold value to obtain the grain uniformity distribution characteristics; based on the uniformity distribution characteristics, the change trend of the interface stress is determined to obtain the interface stress distribution law; based on the stress distribution law, the coating bonding force parameters are extracted to obtain the bonding strength quantitative value; if the bonding strength quantitative value is lower than the preset threshold, the grain orientation characteristics are adjusted by using a machine learning algorithm to obtain the optimized grain distribution data; based on the optimized distribution data, the interface stress distribution is re-analyzed to obtain the adjusted bonding force parameters; based on the adjusted bonding force parameters, the coating bonding state is judged to determine the final bonding force characteristics of the coating.

[0037] For example, when extracting grain orientation and interface bonding information from the transition layer surface, specific data can be obtained through electron backscatter diffraction technology.

[0038] For example.

[0039] In one possible implementation, a high-resolution microscope is used to scan the transition layer surface, producing a two-dimensional distribution map of grain orientation. Combined with the atomic arrangement characteristics at the interface, this method provides a preliminary assessment of the bond strength between the grains and the matrix. This method can intuitively reflect the anisotropy of grain orientation. For example, a grain angle deviation between 5 and 15 degrees indicates relatively consistent orientation. Ion beam scanning technology is used to obtain grain distribution data and interface stress distribution.

[0040] Preferably, a focused ion beam device can be used to perform layer-by-layer peeling analysis on the surface.

[0041] Specifically, with the ion beam energy set at 30 keV, a 100 μm x 100 μm area was scanned to obtain a grain size distribution ranging from 0.5 μm to 2 μm, while simultaneously recording changes in interface stress values ​​from 50 MPa to 200 MPa. This method helps reveal areas of stress concentration and provides data support for subsequent analysis. A preset threshold is used to determine grain distribution uniformity.

[0042] For example, the uniformity threshold may be set to 80%.

[0043] In one embodiment, the distribution is considered uniform if the standard deviation of the grain size within the scanned area is less than 0.3 microns.

[0044] It should be noted that this judgment method can quickly screen out areas with abnormal grain distribution, facilitating targeted optimization. When analyzing the trend of interface stress changes through uniformity distribution characteristics.

[0045] It is understandable that the correspondence between high stress points and grain boundaries can be observed in conjunction with the stress distribution diagram.

[0046] For example, in areas with high uniformity, stress fluctuations are small, typically around 10 MPa, while at boundaries, stress can rise to 150 MPa due to grain mismatch. This regularity analysis helps clarify the driving factors of stress distribution. When extracting coating adhesion parameters based on stress distribution patterns,

[0047] In one possible implementation, the bonding strength may be measured by nanoindentation technology.

[0048] Specifically, under a loading force of 10 mN, the indentation depth and rebound characteristics are recorded, resulting in a quantitative bond strength value of approximately 2 GPa. This quantitative value reflects the level of adhesion between the coating and the substrate. If the bond strength value falls below a preset threshold, for example, below 1.5 GPa, the grain orientation characteristics are adjusted using a machine learning algorithm.

[0049] For example, a convolutional neural network is used to analyze the grain distribution image, optimize the orientation angle deviation to within 10 degrees, and generate new distribution data. This adjustment can improve the stability of the interface bonding. When the interface stress is re-analyzed based on the optimized distribution data.

[0050] Preferably, the ion beam scan may be repeated to observe whether the stress value drops below 100 MPa.

[0051] In one embodiment, the stress concentration points were reduced by 30% after adjustment, indicating a significant optimization effect. This method provides a reliable basis for subsequent parameter adjustments. The adjusted bonding strength parameters are used to determine the bonding state of the coating.

[0052] For example, if the bonding strength is increased to 1.8 GPa, the coating bonding state is considered to be good.

[0053] Specifically, this judgment method can effectively verify the feasibility of optimization measures and ensure the durability of the coating during use.

[0054] In step S3, based on the basic parameters of the coating adhesion, the deposition rate and energy input are adjusted using a multi-energy field deposition device, and a high-density functional layer is continuously deposited on the transition layer to obtain an enhanced gradient composite coating.

[0055] Obtain the initial data of the deposition rate and energy input generated by the multi-energy field deposition equipment to determine the basic parameters of the coating adhesion. Adjust the deposition rate and energy input according to the initial data to obtain the deposition conditions of the transition layer. Use the adjusted deposition conditions to deposit a high-density functional layer on the transition layer to determine the density distribution of the functional layer. Determine the deposition parameters of the gradient composite structure based on the density distribution of the functional layer to obtain a preliminary composite coating. For the preliminary composite coating, use the random forest algorithm to analyze the enhanced effect of the coating adhesion and determine the enhanced characteristics. Obtain the final parameters of the composite coating from the enhanced characteristics to determine the structure of the enhanced gradient composite coating. If the adhesion distribution of the composite coating is uniform, use image processing technology to analyze the surface morphology of the composite coating to obtain the final coating data.

[0056] For example, obtaining initial data on deposition rate and energy input using a multi-energy field deposition device is an important basis for determining coating adhesion parameters.

[0057] In one possible implementation, plasma-enhanced chemical vapor deposition equipment can be used to measure initial data at a deposition rate of 5 nm / min and an energy input of 200 W by adjusting the RF power and gas flow rate. This data reflects the initial conditions for transition layer formation and provides a basis for subsequent parameter adjustments.

[0058] It should be noted that the balance between deposition rate and energy input directly affects the microstructure of the transition layer. Too high an energy may lead to coarse grains, while too low a rate may result in insufficient interlayer bonding. When adjusting the deposition rate and energy input based on the initial data.

[0059] It is understood that this process is intended to optimize the deposition conditions of the transition layer.

[0060] Specifically, the deposition rate can be increased to 8nm / min, while the energy input can be adjusted to 250W to enhance the migration ability of atoms and form a denser transition layer structure.

[0061] In one embodiment, scanning electron microscopy observation of the adjusted transition layer revealed a decrease in porosity from 5% to 2%, demonstrating that the optimized deposition conditions significantly improved interlayer quality. After depositing a high-density functional layer using the adjusted deposition conditions, determining the density distribution of the functional layer is particularly critical.

[0062] For example, X-ray diffraction analysis of the functional layer's crystal structure reveals a density distribution uniformity exceeding 90%. This high density improves the coating's wear and corrosion resistance, laying the foundation for subsequent composite structures. The density distribution of the functional layer is used to determine the deposition parameters for the gradient composite structure.

[0063] Preferably, a multi-layer deposition scheme can be designed according to a density gradient.

[0064] For example, three functional layers, each 50nm, 70nm, and 100nm thick, are sequentially deposited on a transition layer to form a gradient composite coating. This structure effectively disperses interfacial stress and improves overall adhesion. Using a random forest algorithm to analyze the adhesion enhancement effect of preliminary composite coatings is a highly effective method.

[0065] In one example, characteristic parameters such as coating thickness, density, and deposition rate were input, and the algorithm output a bonding enhancement factor of 1.5. This demonstrates that a data-driven approach can accurately identify key variables influencing bonding strength. The final parameters of the composite coating are derived from the enhanced features.

[0066] For example, a combination of a functional layer thickness of 100 nm and a deposition rate of 10 nm / min was determined as the final solution. This parameter combination ensured a uniformity of adhesion distribution of over 95% for the enhanced gradient composite coating. If the adhesion distribution was uniform, image processing techniques were used to analyze the surface topography and obtain the final coating data.

[0067] Specifically, an atomic force microscope scan of the coating surface revealed a surface roughness Ra of 0.8 nm. This smooth morphology not only reflects the coating's high quality but also enhances its service life in harsh environments.

[0068] It can be understood that the optimization of the surface morphology further verifies the rationality of the deposition parameters and provides a guarantee for the long-term stability of the coating performance.

[0069] In step S4, an online detection system is deployed on the surface of the enhanced gradient composite coating to obtain micro-defect distribution data inside the coating through laser interferometry and ultrasonic scanning to determine the specific location and scale of microcracks and pores.

[0070] Laser interferometry is used to obtain interference fringe data on the coating surface to determine the initial profile of the microdefect distribution. Ultrasonic scanning is used to perform deep probing within the coating to obtain a complete dataset of microdefect distribution. A convolutional neural network is applied to this microdefect distribution data to determine the coordinates of the microcrack locations. Based on this microdefect distribution data, pore boundary characteristics are calculated to determine the distribution range of pore size. If the overlap between the microcrack location and the pore size exceeds a preset threshold, an image segmentation algorithm is used to separate the overlapping defects and obtain independent defect features. These independent defect features are then compared with the layered characteristics of the gradient composite material to determine the depth distribution of the defects within the coating. The difference between this depth distribution and the initial profile is obtained, and defect types are classified using a support vector machine to determine the final microdefect distribution features.

[0071] For example, obtaining interference fringe data on the coating surface by laser interferometry can be understood as utilizing the reflection and interference phenomenon of the laser beam on the coating surface to generate a fringe pattern.

[0072] In one possible implementation, a 632.8nm helium-neon laser is used to illuminate the coating surface. The fringe spacing is recorded using an interferometer. This initial profile reveals the distribution of micro-defects such as scratches or pits. This method is highly sensitive and can quickly detect surface topography changes, enhancing the accuracy of subsequent analysis. Ultrasonic scanning is also used to provide a non-destructive method for deep probing within the coating.

[0073] Specifically, an ultrasonic probe with a frequency of 5 MHz can be used to move on the coating surface at a scanning speed of 2 mm / s to obtain the echo signal of internal defects.

[0074] For example, a sudden change in signal intensity may correspond to a crack or pore. The complete dataset provides three-dimensional information for defect location. This approach offers the advantage of wide coverage depth, revealing issues hidden beneath the surface. A convolutional neural network is applied to the micro-defect distribution data.

[0075] Preferably, defect features can be identified by training the network.

[0076] For example, ultrasonic and interferometric data are fed into the network, and the convolution kernel size is set to 3×3. The edge features of microcracks are extracted, and their coordinates are output, such as (x: 50μm, y: 75μm, z: 20μm). This method can efficiently process complex data, improve positioning accuracy, and provide a reliable basis for subsequent judgment. The boundary features of pores are calculated using microdefect distribution data.

[0077] It should be noted that the pore edge can be determined based on the grayscale change of the image.

[0078] For example, the grayscale value at a pore boundary is typically 20% lower than the surrounding area. Combined with the algorithm, this allows us to calculate the pore diameter to be between 10 and 50 μm. This analysis helps assess the impact of defects on coating performance. If the overlap between the microcrack location and the pore size exceeds a preset threshold, for example, if the overlap exceeds 30%, an image segmentation algorithm is used to separate the overlapping defects.

[0079] In one embodiment, a watershed algorithm can be used to divide the overlapping area into independent cracks and pores, and extract features such as crack length 15μm, pore area 25μm, etc. 2 This separation allows for clearer analysis of defect characteristics and avoids confusion. Defect depth can be determined by comparing the individual defect signatures with the layered characteristics of the gradient composite.

[0080] For example, if the thickness of the transition layer is 100 μm and the crack depth reaches 80 μm, it may affect the bonding strength.

[0081] For example, by comparing the layer density distribution, it can be inferred that defects are concentrated in high-density functional layers, improving the targeted nature of the analysis. After obtaining the difference data between the depth distribution and the initial profile, a support vector machine can be used to classify the defect type.

[0082] Specifically, depth differences, such as 10μm and 20μm, are used as input features to classify defects into types such as cracks, pores, or delamination. This method has clear logic and can quickly classify defects, providing guidance for optimizing the coating process.

[0083] In one possible implementation, the final micro-defect distribution characteristics can be intuitively presented through three-dimensional modeling.

[0084] For example, the defect distribution map shows that cracks are concentrated within 50μm of the surface, while pores are mostly distributed in the 80-120μm depth range. This detailed feature helps to improve deposition parameters and enhance coating durability.

[0085] Step S5: Analyze the influence of defects on thermal fatigue resistance based on the micro-defect distribution data and the preset threshold range to obtain a real-time evaluation result of the coating performance.

[0086] Obtain micro-defect distribution data, extract the micro-defect distribution data to obtain a defect feature set; based on the defect feature set combined with a preset threshold, determine the area where the defects exceed the preset threshold range to obtain the defect impact range; with respect to the defect impact range, analyze the change trend of thermal fatigue resistance to obtain a resistance degree distribution; obtain key indicators of coating performance from the resistance degree distribution to determine performance change characteristics; if the performance change characteristics exceed the preset threshold, use a support vector machine algorithm to classify the coating performance to obtain a performance grade division; combine the performance grade division with real-time evaluation requirements to determine the real-time status of the coating performance to obtain a final evaluation result; update the defect impact analysis model based on the final evaluation result to obtain an optimized thermal fatigue resistance prediction result.

[0087] For example, the distribution data is extracted through micro-defect distribution, which usually converts the micro-crack and pore information inside the coating into a quantifiable feature set.

[0088] For example, a three-dimensional distribution map can be generated by scanning data, in which each defect point is marked with its spatial coordinates and size. Assuming that the length of the microcrack in a certain area is 0.5mm and the pore diameter is 0.2mm, these data form the basis of the defect feature set.

[0089] In a possible implementation, a preset threshold, such as a crack length greater than 0.3 mm or a pore diameter greater than 0.15 mm, is used to screen the critical defect areas.

[0090] For example, if 60% of the defects in a certain area exceed a threshold, the area can be determined to be the core area of ​​the defect impact range.

[0091] Specifically, when analyzing the impact range of defects, we can start from the perspective of thermal fatigue resistance.

[0092] Preferably, the temperature distribution change trend of the defect area is observed by simulating thermal cycle loading. For example, if the temperature of a defect area increases by more than 20%, it indicates that the thermal fatigue resistance is reduced.

[0093] In one embodiment, the resistance distribution can be divided into three levels: high, medium, and low. If the resistance of a certain area drops to a low level, its key indicators such as thermal conductivity reduction of 15% are recorded.

[0094] It should be noted that the extraction of performance change characteristics can be based on these indicators to determine whether the coating still meets the usage requirements.

[0095] In one possible implementation, if the performance change characteristic exceeds a preset threshold, such as a decrease in thermal conductivity exceeding 10%, a support vector machine algorithm may be used to classify the performance level.

[0096] For example, coating performance is divided into three levels: A, B, and C. Level A indicates normal, Level C indicates that replacement is required, and a 12% decrease in thermal conductivity in a certain area is classified as Level B.

[0097] Understandably, this classification helps to quickly identify the coating status and improve the efficiency of assessment.

[0098] For example, to meet real-time assessment requirements, sensors can collect temperature and stress data to determine the real-time status of coating performance. If an area experiences an abnormally high temperature and concentrated stress, this indicates a defect that has impacted performance, and the real-time status is marked as "Needs Attention."

[0099] In one embodiment, the final evaluation results may be used to update the analysis model.

[0100] Preferably, the model is trained with historical data to improve its accuracy in predicting thermal fatigue resistance by 10%.

[0101] Specifically, the updated model can analyze the impact of defects from multiple aspects when making predictions.

[0102] For example, considering the defect density, distribution uniformity and material fatigue properties, assuming that the defect density in a certain area is 5 / cm 2 , its resistance is predicted to decrease by 25%, while the density of the other area is 2 / cm 2 , resistance is only reduced by 8%.

[0103] For example, this multi-dimensional analysis ensures more reliable prediction results, while the optimized model can provide early warning of potential failure areas, providing a basis for coating maintenance.

[0104] It should be noted that the advantage of this method is that it forms a closed-loop analysis chain from defect distribution to performance evaluation and then to model optimization.

[0105] Preferably, by combining real-time status monitoring and prediction, the coating service life can be effectively extended and maintenance costs can be reduced.

[0106] In one embodiment, if a coating is evaluated and the process is adjusted in a timely manner, 30% of potential failure risks can be avoided.

[0107] In step S6, if the evaluation result is lower than the preset performance standard, the defect distribution characteristics are extracted from the online detection system and fed back to the multi-energy field deposition equipment to adjust the deposition temperature and ion energy to optimize the density and adhesion of the next batch of coatings.

[0108] Obtain the defect distribution characteristics of the online detection system, use image processing technology to extract the boundaries of the defect distribution characteristics, and obtain defect distribution data; analyze the abnormal areas in the deposition process according to the defect distribution data, use the support vector machine algorithm to determine the relationship between the abnormal areas and the deposition temperature, and determine the temperature adjustment direction; update the control parameters of the deposition equipment according to the temperature adjustment direction, obtain the adjusted deposition temperature value, and obtain a new temperature configuration; analyze the influence of ion energy on the binding force through the defect distribution data, use the linear regression algorithm to determine the energy adjustment amplitude, and determine the ion energy optimization value; obtain the current ion energy parameters of the deposition equipment, adjust the equipment settings in combination with the ion energy optimization value, and obtain a new energy configuration; use the adjusted temperature configuration and the energy configuration to operate the multi-energy field deposition equipment to obtain the next batch of coating data; determine the coating density and binding force change trends according to the next batch of coating data; compare the next batch of coating data with the preset standard, determine the optimization effect, and obtain batch optimization results.

[0109] In one possible implementation, when the inspection result is lower than a preset standard, obtaining defect distribution characteristics from the online inspection system is a key step.

[0110] For example, a high-resolution camera can be used to capture microscopic images of the coating surface, and image processing techniques can be used to extract feature boundaries. For example, edge detection algorithms can be used to identify the outline of defects, providing data such as their shape, size, and distribution. For example, if multiple micropores with diameters between 0.1 and 0.5 mm are detected on the surface of a batch of coatings, this data can intuitively reflect the density and regional characteristics of the defect distribution, providing a basis for subsequent analysis.

[0111] Specifically, when analyzing abnormal areas in the deposition process through defect distribution data, it is possible to focus on the areas where defects are concentrated.

[0112] For example, if the number of micropores detected in a certain area accounts for 70% of the total, it may indicate that the deposition conditions in this area are abnormal. When using the support vector machine algorithm to determine the relationship between the abnormal area and the deposition temperature.

[0113] It's easy to understand that by training the model using historical data, inputting defect distribution characteristics and corresponding temperature values, the output is a judgment of whether the temperature is too high or too low. For example, if the analysis shows that the deposition temperature in the defect-concentrated area reaches 650°C, exceeding the normal range by 50°C, the temperature adjustment direction is to reduce the temperature. This method helps quickly locate the root cause of the problem.

[0114] Preferably, when the control parameters of the deposition equipment are updated according to the temperature adjustment direction, the temperature may be adjusted down from 650 degrees Celsius to 600 degrees Celsius.

[0115] In one embodiment, the device achieves this change by adjusting the heating power, and the new temperature configuration takes effect immediately. This can effectively reduce the generation of defects caused by high temperatures and improve coating quality.

[0116] It should be noted that the linear regression algorithm can be used when analyzing the effect of ion energy on binding force through defect distribution data.

[0117] For example, suppose the defect map shows an ion energy of 200 electron volts in areas with weak adhesion, while the normal area is 250 electron volts. Analysis indicates that the energy needs to be increased by 20% to 240 electron volts as the optimized value. After obtaining the current parameters from the deposition equipment, adjusting to the new values ​​can significantly improve the adhesion of the coating to the substrate.

[0118] In one example, after running the multi-energy field deposition equipment with the adjusted temperature and energy profiles, the next batch of coating data might show an increase in density from 92% to 95% and an increase in bonding force from 50 Newtons to 60 Newtons. This trend indicates that the optimization measures were effective. The batch optimization results can be compared to a preset standard, such as a density requirement of 94% or higher, to determine whether they meet the standards.

[0119] For example, when considering the adjustment effects of temperature and ion energy from multiple aspects, it can be found that lowering the temperature reduces the microcracks caused by thermal stress, while increasing the energy improves the uniformity of atomic deposition. The synergistic effect of the two makes the coating performance more stable.

[0120] It is understandable that this optimization not only improves the reliability of the coating batch, but also provides parameter reference for subsequent production and reduces trial and error costs.

[0121] In one embodiment, the coating density and adhesion trends can be determined by observing microstructural changes using a scanning electron microscope and measuring specific values ​​using an adhesion tester. This multi-dimensional verification approach ensures comprehensive analysis and provides a reliable basis for process improvement.

[0122] Preferably, the optimized batch results can also feed back into further fine-tuning of equipment parameters, forming a closed loop of continuous improvement.

[0123] In step S7, an improved gradient composite coating is generated through the optimized deposition process, and the trend of micro-defect changes is scanned again using the online detection system to determine the extent of improvement in coating durability and production efficiency.

[0124] Obtain a gradient composite coating sample, wherein the gradient composite coating sample is prepared using preset deposition process parameters; for the gradient composite coating sample, use an online detection system to scan its surface to obtain distribution characteristic information of surface microdefects; determine preliminary data of microdefect changes based on the distribution characteristic information; use a time series algorithm to process the continuous scanning results of the online detection system to obtain trend characteristic values ​​of the microdefect changes; judge the durability index of the gradient composite coating sample based on the trend characteristic values; if the durability index is lower than a preset threshold, adjust the deposition process parameters to generate an improved gradient composite coating sample; use the online detection system to scan the improved gradient composite coating sample to obtain updated microdefect change data and determine the improvement in durability; use a production efficiency evaluation tool to analyze the generation process of the improved gradient composite coating sample to obtain production efficiency improvement data; extract key features from the production efficiency improvement data, and combine them with the durability improvement to judge whether the optimization of the deposition process parameters meets the expected standards.

[0125] Exemplarily, when gradient composite coating data is generated through an optimized deposition process.

[0126] It can be understood that this process aims to form a coating with a gradient structure by adjusting the process parameters.

[0127] For example, in the initial stage, the process may set the deposition temperature to 800 degrees and the ion energy to 50eV, and obtain the initial coating sample after running it through the multi-energy field deposition equipment.

[0128] For example, such a sample may exhibit a characteristic of gradually increasing hardness from the bottom layer to the surface layer, but may also have microcracks due to incomplete optimization of parameters. When obtaining micro-defect distribution characteristics from the initial coating sample.

[0129] Preferably, the sample surface can be scanned using a high-resolution optical scanner of an online detection system.

[0130] In a possible implementation, the scanning resolution is set to 10 microns, the number of micro defects detected on the sample surface is 15 per square centimeter, and the average defect size is 20 microns.

[0131] Specifically, these data reflect the uneven stress distribution within the coating, which may be related to the high deposition temperature. The features extracted in this way provide a basis for subsequent analysis. When performing trend analysis on the preliminary data of micro-defect changes.

[0132] For example, a time series algorithm may be used to process the results of five consecutive scans.

[0133] In one embodiment, algorithm analysis shows that the number of defects increased from 15 to 18 per square centimeter after three scans, indicating that the coating is unstable under sustained stress. A trend characteristic value might indicate a defect growth rate of 20%, suggesting that further adjustment of process parameters is required. This analysis helps to dynamically assess coating performance. When determining coating durability indicators based on trend characteristic values.

[0134] It should be noted that durability may be assessed through simulated wear tests.

[0135] In one embodiment, if the preset durability threshold is 500 friction cycles, and the current coating only withstands 400 cycles before significant wear occurs, the parameters need to be adjusted.

[0136] For example, lowering the deposition temperature to 750°C and increasing the ion energy to 60eV produced an improved coating sample. This adjustment effectively reduced internal stress and improved durability. The online inspection system was run again using the improved coating sample.

[0137] Understandably, updated micro-defect data might show a drop to 10 defects per square centimeter, resulting in a 25% improvement in durability.

[0138] Specifically, this improvement indicates progress in process optimization to reduce micro-defects, while coating adhesion may also be enhanced by adjusting ion energy, contributing to longer service life. When analyzing the improved coating sample generation process using a production efficiency assessment tool.

[0139] For example, it can be evaluated by counting the sample output per unit time.

[0140] In one possible implementation, 50 samples were produced per hour before optimization, and this number increased to 60 after optimization, representing a 20% efficiency improvement. Key features were extracted from the efficiency improvement data.

[0141] Preferably, attention can be paid to equipment operation stability and energy consumption changes.

[0142] For example, energy consumption dropped from 10 kWh per wafer to 9 kWh, indicating that the optimization not only improved efficiency but also reduced costs. The improvement in durability and efficiency can be combined to determine whether the process optimization has met the expected standards.

[0143] For example, if the expected goals are to improve durability by 20% and efficiency by 15%, and the actual results exceed expectations, the optimization effect is significant.

[0144] In one embodiment, this comprehensive assessment also provides data support for the next round of process iteration, ensuring continuous improvement in coating performance and production efficiency. This multi-dimensional analysis approach forms a complete logical chain of parameter adjustment and performance improvement.

[0145] Step S8, obtaining thermal cycle test data of the improved gradient composite coating, using a finite element analysis algorithm to simulate the stress distribution of the coating under high temperature and high pressure conditions, and determining the actual improvement level of thermal fatigue resistance and interface strengthening.

[0146] Obtain data on the relationship between temperature changes and the number of cycles in multiple thermal cycle experiments to determine the performance of the coating in multiple cycles; based on the thermal cycle test data, use a finite element analysis algorithm to construct a three-dimensional model of the coating under high temperature and high pressure conditions to obtain stress distribution characteristics in the model; simulate the changes in the stress distribution through the finite element analysis algorithm to determine the thermal fatigue resistance level of the coating under high temperature and high pressure conditions; based on the simulation results of the stress distribution and the thermal fatigue resistance, obtain a quantitative index of the coating interface strengthening to determine the degree of improvement of thermal fatigue caused by interface strengthening; use a machine learning regression algorithm to process the thermal cycle test data and the simulation results to obtain the correlation between thermal fatigue resistance and interface strengthening; analyze the changing trend of coating performance under different working conditions based on the correlation rule to determine the overall improvement level of the improved gradient composite coating; adjust the parameters of the simulation analysis according to the changing trend to obtain optimized coating performance data.

[0147] Exemplary, thermal cycling test data acquisition for an improved gradient composite coating.

[0148] For example, the experiment can be conducted in a thermal cycle furnace with a set temperature range of 200°C to 1000°C and a cycle number of 500. The temperature change of the coating surface is recorded in each cycle and monitored in real time using thermocouples to ensure data continuity.

[0149] For example, in the initial 100 cycles, the temperature was increased from 200°C to 800°C, kept at this temperature for 30 minutes, and then dropped to 200°C. It was found that no obvious cracks were observed on the coating surface, indicating its initial resistance to thermal shock.

[0150] In one possible implementation, when extracting key parameters from thermal cycle test data, attention may be paid to the rate of change of coating thickness and the initiation time of surface microcracks.

[0151] Specifically, assuming an initial thickness of 50 microns, the thickness decreases by 2 microns after 300 cycles, and microcracks first appear at cycle 250. These parameters provide basic data for subsequent analysis.

[0152] It should be noted that when using the finite element analysis algorithm to construct a three-dimensional model, you can choose commercial software such as ANSYS and divide the coating into two parts: the substrate and the gradient layer for input.

[0153] For example, under high-temperature and high-pressure conditions of 900°C and 10 MPa, simulation results show that stress concentration occurs at the coating-substrate interface, with a maximum stress of 150 MPa. This stress distribution indicates that interfacial bonding strength is crucial for performance.

[0154] Specifically, when simulating changes in stress distribution, thermal fatigue resistance can be observed by adjusting the number of cycles and temperature gradient.

[0155] For example, when the number of cycles increased to 600, the stress peak rose to 180 MPa, but the coating did not peel off, indicating that it has strong thermal fatigue resistance. This helps to determine the stability of the coating under extreme conditions.

[0156] In one embodiment, when obtaining a quantitative index of coating interface strengthening, an interface shear strength test may be performed to measure that the shear strength of the improved coating is increased from 50 MPa to 70 MPa.

[0157] Preferably, this enhancement indicates that the interface strengthening significantly improves thermal fatigue resistance and extends the service life of the coating.

[0158] It is understandable that when the machine learning regression algorithm processes data, a random forest model can be selected to input the number of thermal cycles, temperature and stress data, and output the predicted value of thermal fatigue resistance.

[0159] For example, the prediction results show that when the number of cycles exceeds 400, the resistance decreases by 10%, which reveals the law of performance changes with working conditions.

[0160] For example, when analyzing coating performance trends through correlation patterns, it was found that resistance was stable below 700°C, but dropped significantly above this temperature. This provides guidance for process optimization, such as adjusting the composition ratio of the gradient layer.

[0161] In one example, simulations after optimization showed a 15% improvement in resistance, validating the effectiveness of the adjustments. This approach not only improves coating performance but also provides data support for production.

[0162] Step S9: extract key performance indicators from the simulation results, combine them with the data log of the online detection system, and iteratively optimize the process parameters through the gradient descent algorithm to obtain the final stable coating preparation plan.

[0163] Acquire simulation results, use key extraction methods to determine main performance parameters, and obtain performance indicator data; obtain data logs in the online detection system, and determine initial values ​​of process parameters in combination with the performance indicator data; for the records in the data logs, use a gradient descent algorithm to iteratively optimize the process parameters to obtain an adjusted parameter set; based on the adjusted parameter set, obtain the change trend in the coating preparation process and determine the direction of parameter adjustment; if the change trend exceeds a preset threshold, re-extract the online detection data through a system combination method to obtain an updated log set; use the updated log set, combined with the performance indicator data, to optimize the process parameters again through the gradient descent algorithm to obtain a stable parameter solution; based on the stable parameter solution, obtain the final result of coating preparation and determine whether it meets the requirements of the stable solution.

[0164] For example, when obtaining performance indicator data through simulation results, a key extraction method can be used to determine the main performance parameters.

[0165] For example, after simulating the stress distribution of the coating under high temperature and high pressure, indicators such as the maximum stress value and the number of heat-resistant cycles can be extracted.

[0166] In one possible implementation, suppose a simulation shows that a coating experiences a peak stress of 300 MPa after 500 cycles at 1000°C. This data can serve as the basis for performance evaluation. The key to key extraction is to identify the parameters that most significantly impact coating performance, avoiding redundant data that could interfere with subsequent analysis. When acquiring data logs from an online inspection system, performance indicator data can be combined to determine initial values ​​for process parameters.

[0167] Specifically, the detection system may record information such as the temperature during coating deposition is 800°C and the deposition rate is 5 μm / min.

[0168] For example, if the performance indicator shows that the number of heat-resistant cycles is correlated with the deposition rate, the initial process parameters may be set to a historical optimal record close to this value, such as 4.8 μm / min.

[0169] It should be noted that this initial value setting relies on the completeness and accuracy of the log data, which helps reduce the number of subsequent optimization iterations. When iteratively optimizing process parameters using the gradient descent algorithm based on the records in the data log, its implementation can be considered from multiple perspectives.

[0170] For example, the initial deposition rate is set to 5 μm / min, but the log shows that microcracks appear in the coating at high temperature. The algorithm can be used to gradually adjust the rate to 4.5 μm / min and observe the trend of crack reduction.

[0171] In one embodiment, if the parameter set is adjusted by 0.1 μm / min per iteration, and the number of cracks decreases from 10 to 2 after five adjustments, the parameter set becomes stable. This data-driven approach gradually approaches the optimal solution, improving process controllability. By analyzing the changing trends during the coating preparation process based on the adjusted parameter set, the direction of parameter adjustment can be determined.

[0172] Preferably, if the coating thickness uniformity is improved by 15% after the deposition rate is reduced from 5 μm / min to 4.5 μm / min, the adjustment direction can continue to explore the reduction of the rate.

[0173] It is understood that trend analysis not only reflects the impact of parameter changes on the results, but also provides a basis for subsequent optimization, avoiding blind adjustments. If the change trend exceeds the preset threshold, for example, the thickness uniformity requirement is 90%, but the actual measurement is only 85%, then the online inspection data is re-extracted through a combination of system methods.

[0174] In one possible implementation, the update log might show excessive temperature fluctuations, such as from 800°C to 820°C, affecting uniformity. Data is then recollected to ensure that the temperature is within ±5°C. This update ensures that the data reflects the latest operating conditions, improving analysis accuracy. Using this updated log set in conjunction with performance indicator data, a gradient descent algorithm can be used to optimize process parameters again, resulting in a stable parameter solution.

[0175] For example, after adjusting the temperature to 810°C and the rate to 4.6 μm / min, the uniformity reaches 92%, which meets the requirements.

[0176] Specifically, this iterative optimization ensures parameter stability through repeated verification and avoids local optimal traps. When the final result of coating preparation is obtained through a stable parameter solution, it can be judged whether it meets the requirements.

[0177] For example, the final coating showed no obvious failure after 600 cycles at 1000°C, indicating that the solution is feasible.

[0178] It should be noted that this judgment relies on the matching of performance indicators with actual needs, which can effectively guide production practice and improve the reliability of the coating under harsh working conditions.

[0179] The preferred embodiments of the present invention disclosed above are intended only to help illustrate the present invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the present invention to specific embodiments. Obviously, many modifications and variations are possible based on the contents of this specification. These embodiments are selected and described in detail in this specification to better explain the principles and practical applications of the present invention, thereby enabling those skilled in the art to better understand and utilize the present invention. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A method for preparing a composite coating of a rear mold insert, characterized in that: The method comprises the following steps: Step S1: obtaining surface data of the rear mold insert substrate, applying a plasma field and a thermal field synergistically through a multi-energy field deposition device to deposit an initial transition layer on the substrate surface, thereby obtaining a composite coating prototype with a gradient structure; Step S2: extracting grain orientation and interface bonding state information from the surface of the initial transition layer, analyzing grain distribution uniformity and interface stress using ion beam scanning technology, and determining basic coating bonding parameters; Step S3: Based on the basic parameters of the coating adhesion, the deposition rate and energy input are adjusted using a multi-energy field deposition device, and a high-density functional layer is continuously deposited on the transition layer to obtain an enhanced gradient composite coating; Step S4: deploying an online detection system on the surface of the enhanced gradient composite coating to obtain micro-defect distribution data inside the coating through laser interferometry and ultrasonic scanning to determine the specific location and scale of microcracks and pores; Step S5: analyzing the influence of the defects on thermal fatigue resistance based on the micro-defect distribution data and a preset threshold range to obtain a real-time evaluation result of the coating performance; Step S6: If the evaluation result is lower than the preset performance standard, the defect distribution characteristics are extracted from the online detection system and fed back to the multi-energy field deposition equipment to adjust the deposition temperature and ion energy to optimize the density and adhesion of the next batch of coatings; Step S7: Generate an improved gradient composite coating through the optimized deposition process, and use the online detection system to scan the trend of micro-defect changes again to determine the improvement in coating durability and production efficiency; Step S8: Obtain thermal cycle test data of the improved gradient composite coating, simulate the stress distribution of the coating under high temperature and high pressure conditions using a finite element analysis algorithm, and determine the actual improvement level of thermal fatigue resistance and interface strengthening; Step S9: extract key performance indicators from the simulation results, combine them with the data log of the online detection system, iteratively optimize the process parameters through the gradient descent algorithm, and obtain the final stable coating preparation plan.

2. The method for preparing a composite coating for a rear mold insert according to claim 1, characterized in that: The step S1 comprises: Acquire characteristic data of the substrate surface and obtain surface state distribution; According to the surface state distribution, a plasma field treatment is applied using a multi-energy field deposition device to obtain a plasma activated surface; performing thermal field modulation on the plasma activated surface to obtain a thermal field modulated surface; If the uniformity of the thermal field modulation surface reaches a preset threshold, a transition layer is deposited by a deposition device to obtain an initial coating structure; Acquiring gradient distribution data of the initial coating structure; According to the gradient distribution data, the field synergy parameters are adjusted to obtain an optimized composite coating prototype; The optimized composite coating prototype is subjected to surface inspection to obtain final coating data.

3. The method for preparing a composite coating for a rear mold insert according to claim 1, characterized in that: The step S2 comprises: Acquiring grain orientation information and interface bonding information on the surface of the transition layer, and using ion beam scanning technology to acquire distribution data of the grains and stress distribution data of the interface; Determining the uniformity of the grain distribution according to a preset threshold value to obtain a grain uniformity distribution feature; Determining the change trend of the interface stress according to the uniform distribution characteristics to obtain the interface stress distribution law; According to the stress distribution law, the coating bonding force parameters are extracted to obtain the quantitative value of the bonding strength; If the bond strength quantified value is lower than a preset threshold, a machine learning algorithm is used to adjust the grain orientation characteristics to obtain optimized grain distribution data; Reanalyzing the interfacial stress distribution according to the optimized distribution data to obtain adjusted bonding force parameters; The coating bonding state is judged based on the adjusted bonding parameters to determine the final bonding characteristics of the coating.

4. The method for preparing a composite coating for a rear mold insert according to claim 1, characterized in that: The step S3 comprises: Obtain initial data on deposition rate and energy input from multi-energy field deposition equipment to determine the basic parameters of coating adhesion; Adjusting the deposition rate and energy input according to the initial data to obtain deposition conditions for the transition layer; depositing a high-density functional layer on the transition layer using the adjusted deposition conditions, and determining the density distribution of the functional layer; determining deposition parameters of the gradient composite structure according to the density distribution of the functional layer to obtain a preliminary composite coating; For the preliminary composite coating, a random forest algorithm is used to analyze the enhancement effect of the coating adhesion and determine the enhancement characteristics; obtaining final parameters of the composite coating from the enhanced features and determining the structure of the enhanced gradient composite coating; If the bonding force of the composite coating is evenly distributed, image processing technology is used to analyze the surface morphology of the composite coating to obtain final coating data.

5. The method for preparing a composite coating of a rear mold insert according to any one of claims 1 to 4, characterized in that: The step S4 comprises: Obtain interference fringe data on the coating surface through laser interferometry to determine the initial profile of micro-defect distribution; Ultrasonic scanning is used to deeply detect the interior of the coating to obtain a complete data set of micro-defect distribution; Applying a convolutional neural network to the micro-defect distribution data to determine the coordinate information of the micro-crack position; Calculating the boundary characteristics of the pores based on the micro-defect distribution data to determine the distribution range of the pore size; If the overlapping area between the microcrack position and the pore size exceeds a preset threshold, the overlapping defects are separated by an image segmentation algorithm to obtain independent defect features; Determining the depth distribution of the defect within the coating based on a comparison of the independent defect characteristics with the layered characteristics of the gradient composite material; The difference data between the depth distribution and the initial profile is obtained, and the defect type is classified by a support vector machine to determine the final micro-defect distribution characteristics.

6. The method for preparing a composite coating for a rear mold insert according to any one of claims 1 to 4, characterized in that: The step S5 comprises: Acquiring micro-defect distribution data, and extracting the micro-defect distribution data to obtain a defect feature set; According to the defect feature set and a preset threshold, determining the area where the defect exceeds the preset threshold range, and obtaining the defect impact range; Analyze the changing trend of thermal fatigue resistance according to the defect impact range and obtain the resistance degree distribution; Obtaining key indicators of coating performance from the resistance distribution and determining performance change characteristics; If the performance change characteristic exceeds a preset threshold, a support vector machine algorithm is used to classify the coating performance to obtain a performance grade classification; Combining the performance level classification with real-time evaluation requirements, determining the real-time status of the coating performance, and obtaining a final evaluation result; The analysis model of the defect impact is updated according to the final evaluation result to obtain an optimized thermal fatigue resistance prediction result.

7. The method for preparing a composite coating for a rear mold insert according to any one of claims 1 to 4, characterized in that: The step S6 comprises: Obtaining defect distribution characteristics of the online detection system, extracting boundaries of the defect distribution characteristics using image processing technology, and obtaining defect distribution data; Analyzing abnormal areas during the deposition process based on the defect distribution data, using a support vector machine algorithm to determine the relationship between the abnormal areas and the deposition temperature, and determining a temperature adjustment direction; updating the control parameters of the deposition equipment according to the temperature adjustment direction, obtaining the adjusted deposition temperature value, and obtaining a new temperature configuration; Analyzing the effect of ion energy on binding force through the defect distribution data, using a linear regression algorithm to determine the energy adjustment range, and determining the optimized ion energy value; Obtaining current ion energy parameters of the deposition equipment, adjusting equipment settings based on the ion energy optimization value, and obtaining a new energy configuration; operating a multi-energy field deposition device using the adjusted temperature configuration and energy configuration to obtain coating data for a next batch; Determine the coating density and bonding strength change trend based on the next batch of coating data; The coating data of the next batch is compared with the preset standard to judge the optimization effect and obtain the batch optimization result.

8. The method for preparing a composite coating for a rear mold insert according to any one of claims 1 to 4, characterized in that: The step S7 comprises: Obtaining a gradient composite coating sample, wherein the gradient composite coating sample is prepared using preset deposition process parameters; For the gradient composite coating sample, an online detection system is used to scan its surface to obtain distribution characteristic information of surface micro defects; Determining preliminary data of micro-defect changes based on the distribution characteristic information; Using a time series algorithm to process the continuous scanning results of the online detection system to obtain the trend characteristic value of the micro-defect change; determining a durability index of the gradient composite coating sample according to the trend characteristic value; If the durability index is below a preset threshold, adjusting the deposition process parameters to generate an improved gradient composite coating sample; Scanning the improved gradient composite coating sample using the online detection system to obtain updated micro-defect change data and determine the improvement in durability; Using a production efficiency evaluation tool to analyze the generation process of the improved gradient composite coating sample to obtain data on the improvement of production efficiency; Key features are extracted from the production efficiency improvement data, and combined with the durability improvement data to determine whether the optimization of the deposition process parameters meets the expected standards.

9. The method for preparing a composite coating on a rear mold insert according to any one of claims 1 to 4, characterized in that: The step S8 comprises: Obtain the relationship between temperature change and cycle number in multiple thermal cycling experiments to determine the performance of the coating in multiple cycles; Based on the thermal cycle experimental data, a finite element analysis algorithm is used to construct a three-dimensional model of the coating under high temperature and high pressure conditions to obtain the stress distribution characteristics in the model; Simulating the change of the stress distribution by the finite element analysis algorithm to determine the thermal fatigue resistance level of the coating under high temperature and high pressure working conditions; According to the simulation results of the stress distribution and the thermal fatigue resistance, a quantitative index of coating interface strengthening is obtained to determine the degree of improvement of thermal fatigue caused by interface strengthening; Using a machine learning regression algorithm to process the thermal cycle test data and the simulation results to obtain a correlation between thermal fatigue resistance and interface strengthening; Analyze the variation trend of coating performance under different working conditions according to the correlation law, and judge the overall improvement level of the improved gradient composite coating; The parameters of the simulation analysis are adjusted according to the change trend to obtain optimized coating performance data.