Aluminum decorative strip color coating optimization system based on automatic control

Abstract

The invention relates to the technical field of aluminum decorative strip coating, and discloses an aluminum decorative strip color coating optimization system based on automatic control, which comprises a material pretreatment layer, a coating control layer and a quality detection layer. The material pretreatment layer is used for treating the surface of the aluminum decorative strip through surface cleaning, base material activation, roughness detection and adhesive force test; the coating control layer realizes coating control by utilizing a color blending unit, a spraying robot cluster, a dynamic temperature control module and a distributed controller in combination with a real-time color matching algorithm and a self-adaptive path planning mechanism; and the quality detection layer adopts an image analysis technology and a deep learning algorithm to construct a coating quality feature model, and color difference detection, defect positioning and process parameter compensation decision are realized. According to the system, automation, precision and intellectualization of the aluminum decoration strip coating process are achieved, the coating quality and the production efficiency are improved, and the system is suitable for the field of aluminum decoration strip color coating.

Description

Technical Field

[0001] The present invention relates to the technical field of aluminum trim strip coating, and in particular to an aluminum trim strip color coating optimization system based on automatic control. Background Art

[0002] In modern industrial production, aluminum trim is widely used in the automotive, construction, electronics, and other fields due to its lightweight, corrosion-resistant, and easy-to-process advantages. As consumers' demands for product appearance and quality continue to rise, the color coating process for aluminum trim faces greater challenges. Traditional aluminum trim coating processes have many shortcomings and cannot meet the requirements of refined, automated, and high-quality production.

[0003] From a material pretreatment perspective, traditional processes lack the precision to refine the surface treatment of aluminum trim strips. Manual or simple mechanical cleaning methods struggle to completely remove surface contaminants such as grease, oxides, and dust, leading to poor adhesion between the subsequent coating and the substrate, and prone to issues such as flaking and blistering. Furthermore, traditional methods lack precise detection and control methods for substrate surface activation and roughness control, making them unable to adapt to the varying material and process requirements of aluminum trim strips, impacting the coating's bonding strength and appearance quality.

[0004] In terms of coating control, traditional color matching relies on manual experience, making it difficult to accurately replicate coating formulas, resulting in poor color consistency across batches of aluminum trim. During the spraying process, manual operation or fixed-path mechanical spraying cannot dynamically adjust the spray trajectory and parameters based on the coating type and process stage, which can easily lead to problems such as uneven coating thickness and inconsistent color distribution. Furthermore, traditional processes lack real-time temperature control and collaborative operation mechanisms, making the curing environment difficult to precisely control, affecting the coating's curing quality and production efficiency.

[0005] Quality inspection is another weak link in traditional processes. Traditional manual inspection methods are inefficient and subject to subjective factors, making it difficult to accurately detect quality issues such as color variations and defects. The lack of scientific models and algorithms to support coating quality analysis and process parameter adjustments hinders dynamic optimization and precise control of the coating process, resulting in high scrap rates and increased production costs.

[0006] With the rapid development of automated control technology, computer vision technology, and deep learning algorithms, applying these advanced technologies to the color coating process for aluminum trim strips has become an inevitable trend to address the shortcomings of traditional processes and improve coating quality and production efficiency. Therefore, it is urgent to develop an automated control-based aluminum trim strip color coating optimization system to achieve automation, precision, and intelligence in the aluminum trim strip coating process, meeting the demand for high-quality aluminum trim strips in modern industrial production. Summary of the Invention

[0007] The purpose of the present invention is to provide an aluminum trim color coating optimization system based on automated control to solve the problems raised in the above background technology.

[0008] To achieve the above objectives, the present invention provides the following technical solution: an aluminum trim color coating optimization system based on automated control, the system comprising:

[0009] Material pretreatment layer, coating control layer and quality inspection layer;

[0010] The material pretreatment layer includes a surface cleaning device, a substrate activator, a roughness detector, and a coating adhesion tester, which are used to perform physical and chemical treatment on the surface of the aluminum trim strip, analyze the micromorphology, and perform pretreatment operations according to process parameters;

[0011] The coating control layer includes a color mixing unit, a spray robot cluster, a dynamic temperature control module, and a distributed controller. It is equipped with a real-time color matching algorithm and an adaptive path planning mechanism to establish a multi-station collaborative operation channel, realize the synchronization of coating parameters and the classified transmission of process instructions, and adopt an adaptive path planning mechanism to dynamically adjust the coating trajectory and perform layered operations according to the coating type and process stage. The integrated visual sensing interface realizes the coordinated feedback of the physical coating status and digital process.

[0012] The quality detection layer is used to use image analysis technology combined with a standard color difference database and a real-time coating data set to iteratively optimize the color distribution model, thereby constructing a multi-dimensionally mapped coating quality feature model. Based on this feature model, a deep learning algorithm is used to perform color difference detection, defect location, and process parameter compensation decisions.

[0013] Preferably, the surface cleaning device includes at least a plasma processor, an ultrasonic cleaning tank, an electrostatic precipitator and a chemical degreasing tank, which are used to remove pollutants from the surface of the aluminum substrate; the color mixing unit includes at least a color paste metering pump, a spectral analysis module, a viscosity regulator and a mixing reactor, which are used to accurately reproduce the paint formula; the spray robot cluster includes at least a six-axis robotic arm, a high-voltage electrostatic spray gun, an atomization pressure controller and a flow monitoring sensor, which are used to execute the spraying operation instructions issued by the coating control layer.

[0014] Preferably, in the coating control layer, the pre-treated aluminum trim strips are coated by a spray robot cluster using a preset trajectory program, and then the curing environment is regulated by a dynamic temperature control module. The surface data and the surface data recorded by the roughness detector are transmitted to a distributed controller for process optimization. The process optimization adopts a multi-parameter coupling analysis technology, and the collected coating characteristics are input into the parameter space established by different optimization algorithms for joint processing.

[0015] The method of realizing the coordinated feedback of the physical coating state and the digital process through the integrated visual sensing interface includes: realizing the coordinated feedback of the physical coating state and the digital process through the integrated visual sensing interface, exchanging the real-time imaging data of the coating, and jointly analyzing the acquired colorimetric features, realizing dynamic mapping of process parameters, completing instruction issuance, state monitoring and parameter correction, and issuing activation intensity parameters for the material pretreatment layer.

[0016] Preferably, the construction of a multi-dimensional mapping coating quality feature model includes:

[0017] Establish parameter correlation channels and bidirectional calibration mechanisms between coating physical properties and digital feature space;

[0018] The actual coating state is measured through optical inspection and texture analysis. Based on the color distribution, glossiness value, and film thickness measurement values ​​obtained from the quality inspection layer, a coating quality benchmark feature library and a defect feature library are constructed. Model parameter iteration and process simulation are performed based on online inspection data to convert the actual coating state into a high-precision digital feature space.

[0019] Perform parameter identification on the coating quality characteristic model, input the real-time detected coating data into the established model, and use the feature comparison algorithm to dynamically compensate the analytical results of the model to obtain the optimized coating quality characteristic model;

[0020] The coating quality feature model includes physical coating space, digital feature space, process database and interaction mechanism between modules;

[0021] The physical coating space is the data source of the feature model and contains the actual state characteristics of the coating; the digital feature space forms a mapping relationship with the physical coating space, and the coating quality characteristics are mathematically characterized through multi-dimensional parametric modeling; the process database integrates historical process data and real-time detection information, and provides a benchmark data set including a color difference threshold library, a defect pattern library, and an equipment parameter library; the interaction mechanism realizes data communication between modules, and the physical coating space and the process database realize real-time collection and model update of quality parameters through standardized protocols, the physical coating space and the digital feature space transfer feature parameters through a data interface, and the digital feature space and the process database realize information interaction through a data bus.

[0022] Preferably, the color difference detection using a deep learning algorithm based on the feature model includes:

[0023] Based on the coating quality feature model, standard color plate data, ambient lighting parameters, and coating reflectivity values ​​are obtained to construct a color difference sample set;

[0024] After feature enhancement processing, the color difference sample set is divided into a training set and a validation set;

[0025] Establish a convolutional neural network-support vector machine-residual network hybrid model architecture, set the model's initial parameters, input the training set into the hybrid model for joint training, perform feature extraction through the convolutional neural network, perform boundary delineation through the support vector machine, and perform detail compensation through the residual network. Use the dynamic gradient mechanism to balance the recognition errors of different algorithms in specific detection scenarios until the model accuracy reaches the set threshold or the scheduled training rounds are completed;

[0026] Input the validation set into the trained hybrid model, calculate the comprehensive performance index of the model, and select the optimal color difference detection model;

[0027] Based on the optimal color difference detection model, the spatial distribution map of abnormal color areas is output, and the coordinates of the color difference exceeding the standard area are determined by combining the color diffusion model.

[0028] Preferably, the defect location is performed using a deep learning algorithm based on the feature model, including:

[0029] Based on the coating quality feature model, the texture features, shape parameters, and grayscale distribution data of the defect area are extracted to construct a defect feature vector set;

[0030] Independent component analysis is used to reduce the dimension of the defect feature vector to obtain the core defect feature components;

[0031] Establish a defect classification model based on density clustering and determine the optimal number of classifications using the silhouette coefficient algorithm;

[0032] The classified defect features are input into the preset process adjustment library to match the optimal compensation solution and generate a targeted parameter correction instruction set.

[0033] Preferably, the process parameter compensation decision-making using a deep learning algorithm based on the feature model includes:

[0034] Based on the coating quality characteristic model, the motion parameters of the spraying equipment, the rheological properties of the coating, and the ambient temperature and humidity data are collected to build a process compensation feature library;

[0035] Reorganize the data in the process compensation feature library into time series to generate process sample segments;

[0036] Establish a long short-term memory network model, set the number of memory units and forget gate parameters, and obtain the correlation characteristics of process parameters through back propagation calculation;

[0037] The associated features are input into the classifier for compensation decision classification, and the discrimination results of process parameter maintenance and dynamic adjustment are output.

[0038] Preferably, the construction of a multi-dimensional mapping coating quality feature model further includes:

[0039] A stratified sampling mechanism is used to segment the continuous inspection data, and the data segments of each process stage are independently subjected to feature extraction;

[0040] Establish a correlation matrix between data segment features and equipment parameters to record the corresponding coating quality patterns under different process conditions;

[0041] The feature model is updated online through a transfer learning algorithm, and the model parameter adaptive correction mechanism is triggered when an unrecorded quality pattern is detected.

[0042] Preferably, the method for generating the parameter correction instruction includes:

[0043] Establish a mapping relationship table between defect types and compensation methods, including compensation methods for orange peel defects corresponding to atomization pressure adjustment and sagging defects corresponding to spray speed correction;

[0044] Genetic algorithm is used to search for the best compensation parameter combination, including spray gun movement speed, paint flow rate and curing temperature;

[0045] The compensation effect is evaluated in real time through a closed-loop control mechanism, and the parameter re-optimization process is triggered when the quality indicators do not meet the standards.

[0046] Preferably, the training process of the hybrid model architecture includes:

[0047] The Boosting method is used to generate training sample sets for multiple base learners, and the classification performance of each base learner is evaluated through stratified sampling.

[0048] The adaptive coefficient allocation algorithm is used to calculate the combined contribution coefficient of each base learner, and the gradient boosting mechanism is used to integrate and optimize the output results of the base learners.

[0049] Compared with the prior art, the present invention has the following beneficial effects:

[0050] In terms of material pretreatment, the surface cleaning system integrates multiple devices, including a plasma processor, ultrasonic cleaning tank, electrostatic precipitator, and chemical degreasing tank. This system thoroughly removes contaminants from the aluminum substrate surface both physically and chemically, providing a clean, tidy surface for subsequent painting. A substrate activator activates the surface of aluminum trim according to process parameters, increasing surface energy and enhancing adhesion between the coating and the substrate. A roughness tester and coating adhesion tester monitor the substrate's surface micromorphology and adhesion data in real time, performing pretreatment operations based on this data. This enables precise control of the pretreatment process, ensuring that the substrate surface condition meets painting requirements and improving coating quality stability from the source.

[0051] The coating control layer achieves precise control of coating parameters and multi-station collaborative operation through the coordinated operation of a color mixing unit, a spray robot cluster, a dynamic temperature control module, and a distributed controller. The color mixing unit, consisting of components such as a colorant metering pump, a spectral analysis module, a viscosity regulator, and a mixing reactor, accurately reproduces coating formulas, ensuring color consistency across batches. The spray robot cluster, equipped with a six-axis robotic arm, a high-voltage electrostatic spray gun, an atomizer pressure controller, and a flow monitoring sensor, dynamically adjusts the coating trajectory based on coating type and process stage through an adaptive path planning mechanism, achieving layered coating and ensuring uniform coating thickness and consistent color distribution. The dynamic temperature control module precisely regulates the curing environment. Combined with the distributed controller's multi-parameter coupling analysis technology, it optimizes process parameters based on real-time collected coating characteristic data, improving curing quality and production efficiency. Furthermore, an integrated visual sensing interface enables collaborative feedback between the physical coating status and the digital process. By exchanging real-time coating imaging data and jointly analyzing colorimetric characteristics, it monitors the coating status in real time and dynamically adjusts process parameters, further enhancing the precision and intelligence of the coating process.

[0052] The quality inspection layer utilizes image analysis technology combined with a standard color difference database and a real-time coating dataset to construct a multi-dimensional mapping coating quality feature model. Based on this model, deep learning algorithms are employed to implement color difference detection, defect location, and process parameter compensation decisions. By establishing a parameter correlation channel and a bidirectional calibration mechanism between the coating's physical properties and the digital feature space, the actual coating state is converted into a high-precision digital feature space, enabling comprehensive and accurate analysis of coating quality. The application of deep learning algorithms, such as a convolutional neural network-support vector machine-residual network hybrid model architecture for color difference detection, a density-based defect classification model for defect location, and a long short-term memory network model for process parameter compensation decisions, enables rapid and accurate detection of color differences and defects and generates targeted parameter correction instructions, enabling dynamic optimization and precise control of the coating process. Furthermore, the application of a stratified sampling mechanism and transfer learning algorithms enables real-time updating of the feature model and adaptive parameter correction based on online inspection data, improving the system's adaptability to diverse process conditions and quality models, further reducing scrap rates and improving production efficiency and product quality stability. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 This is a working principle diagram of the aluminum trim color coating optimization system based on automatic control according to the present invention;

[0054] Figure 2 Design drawings constructed for coating quality characterization models;

[0055] Figure 3Design diagram for training color difference detection model;

[0056] Figure 4 Design diagram for defect location process;

[0057] Figure 5 Design diagram for process parameter compensation decision. DETAILED DESCRIPTION

[0058] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0059] See also Figure 1-Figure 5 The present invention relates to an aluminum trim color coating optimization system based on automatic control, which includes:

[0060] The material pretreatment layer, coating control layer and quality inspection layer work together to achieve automated optimization control of the color coating of aluminum trim strips.

[0061] The material pretreatment layer undergoes physical and chemical treatment, micromorphology analysis, and pretreatment based on process parameters through a surface cleaning device, substrate activator, roughness tester, and coating adhesion tester. The surface cleaning device removes contaminants from the aluminum substrate, providing a clean surface for subsequent painting. The substrate activator activates the aluminum trim surface to enhance surface activity. The roughness tester measures the micromorphology of the treated surface and obtains surface roughness data. The coating adhesion tester performs coating adhesion testing on the pretreated aluminum trim based on process parameters to ensure that the pretreatment meets painting requirements.

[0062] The coating control layer includes a color matching unit, a spray robot cluster, a dynamic temperature control module, and a distributed controller. A real-time color matching algorithm and an adaptive path planning mechanism are set up to establish a multi-station collaborative operation channel, synchronize paint parameters, and transmit process instructions in a classified manner. An adaptive path planning mechanism is used to dynamically adjust the coating trajectory and perform layered operations based on the coating type and process stage. The integrated visual sensor interface enables collaborative feedback between the physical coating state and the digital process. The color matching unit accurately reproduces the paint formula, providing paint that meets the requirements for spraying. The spray robot cluster executes the spray operation instructions issued by the coating control layer and covers the pre-treated aluminum trim with paint according to the preset trajectory program. The dynamic temperature control module regulates the curing environment to ensure the coating curing effect. The distributed controller coordinates and controls each module to achieve multi-station collaborative operation.

[0063] The quality inspection layer uses image analysis technology combined with a standard color difference database and a real-time coating dataset to iteratively optimize the color distribution model. This model then constructs a multi-dimensional coating quality feature model. Based on this feature model, a deep learning algorithm is employed for color difference detection, defect location, and process parameter compensation decisions. By inspecting the coating using image analysis technology and combining it with database data, an accurate coating quality feature model is constructed, providing a basis for subsequent inspections and decision-making.

[0064] The present invention is further described in detail below with reference to specific embodiments.

[0065] Example 1:

[0066] Example 1 of an automated control-based color coating optimization system for aluminum trim strips involves the partial equipment components and implementation methods for the material pretreatment layer and the coating control layer. This implementation utilizes multiple devices working in concert to achieve precise control of both the pretreatment and coating process for the aluminum trim strips, laying the foundation for subsequent high-quality coating.

[0067] The surface cleaning device for the material pretreatment layer plays a crucial role in the pre-coating process for aluminum trim strips. A plasma processor generates plasma to treat the aluminum substrate surface. The high-energy particles in the plasma collide with the aluminum substrate surface, effectively removing organic matter and oxides, resulting in a clean and activated surface. During the treatment process, plasma parameters such as power and gas flow can be adjusted based on the aluminum substrate's material and surface condition to ensure consistent and stable treatment results.

[0068] Ultrasonic cleaning tanks utilize the cavitation effect created by ultrasound waves in liquids to remove dirt from aluminum surfaces. When ultrasound waves propagate through liquids, they produce alternating cycles of compression and expansion. During the expansion cycles, tiny bubbles form. These bubbles quickly collapse during the compression cycles, generating powerful shock waves that dislodge dirt adhering to the aluminum surface. The cleaning fluid in ultrasonic cleaning tanks can be formulated appropriately based on the type of dirt to enhance cleaning effectiveness. Parameters such as cleaning time and temperature can also be precisely controlled to ensure an efficient and stable cleaning process.

[0069] Electrostatic precipitators (ESPs) remove dust and other tiny particles from the surface of aluminum substrates through the principle of electrostatic attraction. As the aluminum substrate passes through the ESP, the dust particles become electrically charged and then attracted to an oppositely charged collection plate. Parameters such as the ESP's electric field strength and airflow velocity can be adjusted based on the size and concentration of the dust particles to ensure effective removal of dust of varying sizes.

[0070] Chemical degreasing tanks use chemicals to remove grease from the aluminum substrate surface. Chemical degreasing agents typically contain alkaline substances and surfactants, which react with grease, breaking it down into water-soluble substances and achieving the desired degreasing effect. During the chemical degreasing process, parameters such as degreasing agent concentration, temperature, and immersion time must be strictly controlled to ensure effective degreasing without damaging the aluminum substrate surface.

[0071] The color mixing unit in the coating control layer plays a key role in accurately reproducing paint formulas. Colorant metering pumps utilize high-precision metering technology to precisely control colorant dosage. Through advanced sensors and control systems, these pumps accurately deliver the appropriate amount of colorant according to preset recipe requirements, ensuring accurate and consistent paint color. During the delivery process, the pumps also monitor and adjust the flow rate and pressure of the colorant in real time to ensure metering accuracy.

[0072] The spectral analysis module performs spectral analysis on the prepared paint, detecting the paint's absorption and reflection of light of different wavelengths to determine its color parameters. Using a high-resolution spectrometer and advanced analysis algorithms, the module accurately measures the paint's color coordinates, color difference, and other parameters. Comparing these with a standard color palette, the module determines whether the paint's color meets the required specifications. If color deviation is detected, the module promptly provides feedback to the control system for adjustment.

[0073] A viscosity regulator is used to adjust the viscosity of a coating to the optimal range for spraying. Paint viscosity significantly impacts spraying results. If the viscosity is too high, the paint may not atomize evenly, resulting in coating defects such as sagging and orange peel. If the viscosity is too low, the paint may over-atomize, causing waste and environmental pollution. Viscosity regulators precisely adjust the viscosity of the coating through heating, stirring, or the addition of diluents, ensuring proper fluidity and atomization during the spraying process.

[0074] Mixing reactors thoroughly blend the various coating components to ensure a uniform and consistent coating. These reactors are typically equipped with powerful stirring mechanisms and heating systems, enabling the coating components to fully mix and react under high temperature and pressure. During the mixing process, parameters such as the stirring speed and duration can be adjusted based on the coating's formulation and properties to ensure uniformity and stability. Furthermore, the mixing reactor can monitor and control parameters such as temperature and pressure in real time to ensure a safe and efficient mixing process.

[0075] The spray robot cluster achieves high-precision spraying during the painting process. The six-axis robotic arm, with six degrees of freedom, enables flexible movement in three-dimensional space, enabling complex spraying trajectories. Leveraging advanced motion control systems and programming technology, the six-axis robotic arm can precisely spray according to pre-programmed trajectories, ensuring consistent coating thickness and uniformity. The high repeatability of the six-axis robotic arm ensures consistent spraying results, improving coating quality.

[0076] The high-voltage electrostatic spray gun uses the principle of electrostatic adsorption to ensure a more even adhesion of paint to the surface of the aluminum trim. When the paint is sprayed through the high-voltage electrostatic spray gun, it becomes charged. The aluminum trim is grounded, creating an electric field. Under the influence of this electric field, the charged paint particles are attracted to the surface of the aluminum trim, forming a uniform coating. Parameters such as the voltage and current of the high-voltage electrostatic spray gun can be adjusted based on the properties of the paint and the shape of the aluminum trim to optimize the electrostatic adsorption effect and reduce paint waste and environmental pollution.

[0077] Atomizing pressure controllers control the atomization of paint, ensuring uniform coating. Paint atomization directly impacts coating quality. Poor atomization can result in defects such as particles and pinholes. Atomizing pressure controllers precisely regulate the pressure and flow rate of the spray gun, ensuring a smooth and uniform coating by forming fine, uniform droplets. Furthermore, atomizing pressure controllers can adaptively adjust parameters such as spray distance and angle to achieve optimal atomization.

[0078] Flow monitoring sensors monitor paint flow in real time, ensuring a stable paint supply during the spraying process. Utilizing high-precision flow measurement technology, these sensors accurately measure paint flow in real time and transmit this data to the control system. If an abnormal paint flow is detected, the control system promptly adjusts the paint pump speed or pressure to ensure a stable paint supply. The use of flow monitoring sensors not only improves coating quality but also reduces paint waste and production costs.

[0079] During implementation, the various devices collaborated through automated control systems. The surface cleaning device in the material pretreatment layer treated the aluminum substrate in the order of a plasma processor, ultrasonic cleaning tank, electrostatic precipitator, and chemical degreasing tank. The processing parameters of each device were precisely adjusted based on the material and surface condition of the aluminum substrate. The coating control layer also seamlessly integrated and coordinated the color mixing unit, spray robot cluster, and dynamic temperature control module. After the color mixing unit prepared the paint according to the preset formula, it transferred the paint to the spray robot cluster for spraying. The dynamic temperature control module precisely controlled the curing environment according to the requirements of the spraying process to ensure that the coating was fully cured.

[0080] The entire implementation process utilizes advanced automated control and sensor technologies, enabling precise control and real-time monitoring of each device. Through the automated control system, operators can remotely monitor and adjust each device's operating parameters, improving production efficiency and management. Furthermore, the application of sensor technology enables each device to sense operating status and environmental changes in real time, automatically adjusting based on this feedback, ensuring the stability and reliability of the coating process.

[0081] Example 2:

[0082] Example 2 of the automated control-based aluminum trim color coating optimization system involves process optimization and a collaborative feedback mechanism for the coating control layer. This implementation utilizes multi-parameter coupling analysis technology and visual sensor feedback to achieve dynamic optimization and precise control of the coating process, ensuring stable and consistent coating quality.

[0083] In the painting control layer, the pre-treated aluminum trim is coated by a cluster of spray robots. These robots operate according to a pre-set trajectory program, precisely planned based on the shape, size, and painting requirements of the aluminum trim. During the spraying process, each joint of the six-axis robotic arm moves precisely, ensuring the spray gun maintains the appropriate distance and angle from the surface of the aluminum trim, thereby ensuring uniform coating thickness. The high-voltage electrostatic spray gun uses the principle of electrostatic adsorption to ensure that the paint adheres more evenly to the surface of the aluminum trim, reducing paint waste and environmental pollution. An atomization pressure controller and flow monitoring sensor monitor and control the paint atomization effect and flow rate in real time, ensuring the stability and consistency of the spraying process.

[0084] After spraying, the aluminum trim enters the dynamic temperature control module for curing environment regulation. The dynamic temperature control module can accurately control environmental parameters such as temperature and humidity during the curing process. Temperature is one of the key factors affecting the curing effect of the coating. Different types of coatings require different curing temperature ranges. The dynamic temperature control module uses a heating system and temperature sensors to monitor and adjust the temperature of the curing environment in real time to ensure that the temperature remains stable within the range required by the coating. Humidity will also affect the curing of the coating. Excessive humidity may cause defects such as bubbles and whitening on the coating surface. The dynamic temperature control module controls the humidity of the curing environment through a dehumidification system and humidity sensors to provide good environmental conditions for coating curing.

[0085] Surface data recorded by the roughness detector and environmental data collected by the dynamic temperature control module are transmitted to the distributed controller for process optimization. The distributed controller uses multi-parameter coupling analysis technology to comprehensively analyze and process this data. Multi-parameter coupling analysis considers the interplay between multiple parameters, enabling a more comprehensive and accurate analysis of the process. Coating quality analysis considers not only surface roughness and parameters such as the curing environment temperature and humidity, but also factors such as the coating's rheological properties and spraying parameters. This coupled analysis of these parameters identifies key factors influencing coating quality and enables targeted optimization.

[0086] The collected coating features are fed into the parameter spaces established by different optimization algorithms for joint processing. Different optimization algorithms have different characteristics and scopes of application. Combining multiple algorithms can fully leverage their strengths and improve optimization results. Genetic algorithms have strong global search capabilities, enabling them to find optimal solutions within large parameter spaces; particle swarm optimization algorithms have a fast convergence rate, enabling them to quickly find optimal solutions within local regions. By simultaneously feeding the collected coating features into the parameter spaces established by both the genetic and particle swarm algorithms, a more optimal combination of process parameters can be obtained through joint processing.

[0087] An integrated visual sensing interface enables coordinated feedback between the physical coating status and the digital process. This interface utilizes a high-resolution camera and advanced image processing technology to capture real-time imaging data of the coating. The acquired colorimetric features are then jointly analyzed to determine whether the coating color meets the required specifications by analyzing the color information contained in the imaging data. Furthermore, dynamic mapping of process parameters is implemented, associating the coating's colorimetric features with the process parameters and establishing a mapping relationship between them. Based on this mapping relationship, when deviations in coating color are detected, corresponding process parameters, such as the paint formula and spraying parameters, can be adjusted promptly.

[0088] During the collaborative feedback process, the integrated visual sensing interface collects real-time imaging data of the coating and transmits it to the control system for processing. The control system analyzes and processes the imaging data, extracting information such as color characteristics and comparing it with a standard color palette. If color deviation is detected, the control system calculates the necessary process parameters based on a preset algorithm and issues adjustment instructions to the corresponding modules. Upon receiving the instructions, each module promptly adjusts its operating parameters, such as the spray robot cluster adjusting the spray trajectory and parameters, and the color mixing unit adjusting the paint formula. The control system also monitors the effects of these adjustments in real time to ensure improved coating quality.

[0089] The integrated visual sensing interface enables coordinated feedback between the physical coating status and the digital process, including distributing adjusted process parameters to each module, enabling command issuance, status monitoring, and parameter correction. The control system distributes the adjusted process parameters to each module via the communication network. Upon receiving the command, each module executes the corresponding operation and provides feedback to the control system. The control system monitors the status of each module in real time to ensure that it operates according to the command requirements. If a module's operating status is detected as abnormal, the control system promptly intervenes and makes corrections, ensuring the stability and reliability of the entire coating process.

[0090] After the adjusted process parameters are distributed to each module, activation strength parameters are also distributed to the material pretreatment layer. The activation strength of the material pretreatment layer significantly impacts the adhesion of the coating. By analyzing the relationship between the coating's quality characteristics and the process parameters, appropriate activation strength parameters are determined and distributed to the material pretreatment layer. Based on these activation strength parameters, the material pretreatment layer adjusts the operating parameters of equipment such as the plasma processor and substrate activator to improve the activation level of the aluminum trim surface, thereby enhancing the coating's adhesion.

[0091] Example 3:

[0092] Example 3 of the automated aluminum trim color coating optimization system involves constructing a multi-dimensionally mapped coating quality feature model at the quality inspection layer. This implementation achieves a precise mapping of the physical coating state and digital feature space by establishing parameter association channels, building a feature library, identifying model parameters, and designing an interactive mechanism, providing a foundation for intelligent coating quality detection and process optimization.

[0093] 1. Establishment of parameter association channel and bidirectional calibration mechanism

[0094] The first step in constructing a multi-dimensional mapping model of coating quality characteristics is to establish a parameter correlation channel and a bidirectional calibration mechanism between the coating's physical properties and the digital feature space. Physical properties include measurable physical quantities such as the coating's color distribution, gloss, film thickness, and surface roughness. The digital feature space, through mathematical modeling, converts these physical quantities into computable digital parameters, such as RGB color values, spectral reflectance curves, and texture feature vectors. The bidirectional calibration mechanism ensures consistency between physical measurements and digital model parameters. For example, by regularly calibrating the spectral response curve of optical inspection equipment, the collected color data matches the digital characteristics of a standard color palette.

[0095] 2. Feature Library Construction and Real-time Data-Driven Model Iteration

[0096] The actual coating state is obtained through optical detection (such as spectrometer, colorimeter) and texture analysis (such as laser confocal microscopy) to obtain basic data, and the color distribution of the quality detection layer is collected , glossiness value , film thickness measurement value With the feature library as the core, we build the benchmark feature library and defect feature library.

[0097] Benchmark feature library stores multi-dimensional parameters of qualified coatings, such as standard color range , gloss threshold , film thickness tolerance , forming a feature template with qualified quality.

[0098] The defect feature library records the characteristic combinations of common defects, such as the chromaticity offset corresponding to the color difference defect , Texture roughness peak value of sagging defect , wavelength-amplitude distribution of orange peel defects, etc.

[0099] During the online detection process, the real-time collected coating data will trigger the model parameter iteration mechanism. For example, when the film thickness measurement values ​​of 10 samples are detected continuously Deviate from the baseline value Exceed When the system automatically incorporates the new data into the benchmark feature library, and refits the film thickness and process parameters (such as spray flow) through the least squares method. , atomization pressure ) to dynamically update the benchmark features. Simultaneously, the process simulation module, based on the updated feature library, uses finite element analysis to predict the coating state under different process parameter combinations, assisting in process optimization.

[0100] 3. Parameter identification and structural design of coating quality characteristic model

[0101] When performing parameter identification on the coating quality characteristic model, the coating data detected in real time is input into the established model and a feature matching algorithm (such as Euclidean distance matching) is used. ,in is the measured value, is the model prediction value) to dynamically compensate the analytical results of the model. If the calculated distance Exceeding the preset threshold , the system automatically triggers the parameter correction process, adjusts the weight coefficient or boundary conditions in the model until .

[0102] The structure of the coating quality characteristic model includes four core modules:

[0103] Physical coating space: As a data source, it is directly related to the actual coating status of the aluminum trim and contains raw data of physical quantities detected in real time, such as the pixel values ​​of the coating surface image collected by a charge-coupled device (CCD) camera and the thickness distribution measured by a contact film thickness meter.

[0104] Digital feature space: forms a mapping relationship with the physical coating space, and mathematically characterizes the coating quality characteristics through multi-dimensional parameterized modeling. For example, the two-dimensional chromaticity distribution Convert to a one-dimensional feature vector ,in is the average chromaticity value of different areas; and film thickness After normalization, it is used as a vector component to form a comprehensive feature vector .

[0105] Process database: Integrates historical process data and real-time inspection information, including three types of benchmark data sets:

[0106] Color difference threshold library: stores the allowable chromaticity deviation range corresponding to different color standards;

[0107] Defect pattern library: records the mapping relationship between defect types and feature vectors. For example, the high-frequency texture component in the feature vector corresponding to the orange peel defect accounts for more than 30%;

[0108] Equipment parameter library: saves historical effective process parameter combinations such as spray robot trajectory parameters, paint formula parameters, curing temperature curves, etc.

[0109] Interaction mechanism: realizes data communication between modules, including:

[0110] The physical coating space and process database collects quality parameters and updates the model in real time through standardized protocols (such as OPCUA). For example, the film thickness detection data is synchronized to the process database every 10 seconds.

[0111] The physical coating space and the digital feature space transfer feature parameters through a data interface (such as REST API) to convert image pixel values ​​into feature vectors ;

[0112] The digital feature space and the process database exchange information through a data bus (such as a CAN bus), and the feature comparison results are fed back to the process parameter adjustment module.

[0113] 4. Model Optimization for Stratified Sampling and Transfer Learning

[0114] In order to improve the adaptability of the model to complex process, Example 3 uses a stratified sampling mechanism to segment the continuous detection data. According to the coating process stage (such as primer spraying, color paint spraying, and varnish curing), the detection data is divided into different data segments, and each data segment is independently extracted. For example, the primer spraying stage focuses on extracting the film thickness uniformity feature, the color paint spraying stage focuses on the color consistency feature, and the varnish curing stage focuses on the gloss and surface roughness features. By establishing a correlation matrix between data segment features and equipment parameters (in is the number of features, is the number of equipment parameters), record different process conditions (such as spray gun moving speed , paint flow , curing temperature ) corresponding coating quality mode.

[0115] When an undocumented quality pattern is detected (e.g., the texture features of a new orange peel defect don't match the existing defect pattern library), the system triggers a transfer learning algorithm to update the feature model online. Transfer learning leverages the existing knowledge of the defect feature library to quickly adapt to the new quality pattern by adjusting some model parameters (such as the shallow weights of the convolutional neural network). The specific process involves inputting the new feature vector into the feature extraction layer of the pre-trained model, freezing the weights of the first few layers, and fine-tuning only the last few layers. The model parameters are then optimized using the backpropagation algorithm until the new feature can be correctly classified.

[0116] 5. Formula Description and Application Scenarios

[0117] In the feature comparison algorithm, the Euclidean distance formula is used to calculate the difference between the measured value and the model predicted value:

[0118]

[0119] in, : Euclidean distance, which represents the degree of difference between the measured features and the model predicted features. The smaller the value, the higher the matching degree. : Feature dimension, such as chromaticity, glossiness, and film thickness ; : No. The measured value of a feature, such as the measured chromaticity value , measured glossiness ; : No. The model prediction value of each feature is generated by a mathematical model of the digital feature space.

[0120] This formula is used to quantify the deviation between the measured value and the expected value of the coating quality characteristics, providing data support for process parameter adjustment. For example, when the Euclidean distance between the measured color value of a batch of aluminum trim and the model predicted value is detected Exceeding the threshold When the system automatically prompts that the output parameters of the color paste metering pump or the trajectory speed of the spraying robot need to be adjusted.

[0121] Example 4:

[0122] Example 4 of the automated control-based aluminum trim color coating optimization system involves a deep learning color difference detection and defect location process based on a coating quality feature model. This implementation achieves accurate identification and location of aluminum trim coating color deviations and surface defects by constructing a sample set, designing a hybrid model architecture, and optimizing a classification algorithm, providing a clear decision-making basis for process adjustments. The following is an explanation based on specific scenarios:

[0123] 1. Implementation process and application of color difference detection

[0124] In the color difference detection process, the system first obtains standard color plate data, ambient lighting parameters, and coating reflectance values ​​based on the coating quality feature model to construct a color difference sample set. For example, when producing a certain model of automotive aluminum trim, the standard color plate is defined as Pantone 186C red, with the corresponding RGB values ​​of (227, 26, 33), and the spectral reflectance presents a specific curve in the range of 450-700nm. The ambient lighting parameters need to simulate the D65 standard light source (color temperature 6500K, illuminance 3000Lux) in the actual production workshop, and the light fluctuations are monitored in real time by a spectroradiometer. The coating reflectance values ​​are collected by an integrating sphere spectrometer. The surface of the aluminum trim is evenly divided into 100 detection areas, and 10 reflectance samples are obtained in each area to form a single sample containing 1000 data points.

[0125] After the sample set is constructed, feature enhancement is performed to improve model generalization. This includes adding ±5% Gaussian noise to the colorimetric data to simulate detection error, interpolating and expanding the spectral reflectance curve to uniform data length, and normalizing the ambient lighting parameters (e.g., mapping illuminance values ​​to the [0, 1] range). The processed samples are then divided into a training set and a validation set in an 8:2 ratio. The training set contains 800 samples for model training, and the validation set contains 200 samples for performance evaluation.

[0126] A hybrid convolutional neural network (CNN), support vector machine (SVM), and residual network (ResNet) model architecture was then established. The CNN module uses the classic VGG16 network structure, consisting of 13 convolutional layers and 3 fully connected layers, to extract spatial features of chromaticity distribution, such as the edges, texture, and uniformity of color blocks. The SVM module, based on the RBF kernel function, demarcates the feature vectors output by the CNN to distinguish between acceptable and abnormal color areas. The ResNet module contains 50 layers of residual blocks to compensate for detailed features, such as subtle differences in spectral reflectance in areas of minor color difference. The initial model parameters were set as follows: CNN learning rate 0.001, SVM regularization parameter C=10, and ResNet weight decay coefficient 0.0001.

[0127] During training, the training set is fed into the hybrid model for joint training. The CNN first extracts features from the input colorimetric image, generating a 512-dimensional feature vector. This vector is then fed into the SVM for binary classification (pass / fail), outputting classification probabilities. The ResNet then performs deep feature extraction on the raw spectral data, outputting a 256-dimensional detail feature vector. These two features are concatenated into a 768-dimensional vector, which is then fed through a fully connected layer to produce the final color difference prediction. Training utilizes a dynamic gradient mechanism, automatically adjusting weights based on the recognition error of different algorithms. For example, when the CNN's recognition error in complex curved areas exceeds 15%, the gradient update step size for that module is automatically increased. Training continues until the model's accuracy on the training set consistently exceeds 95% or 50 iterations are completed.

[0128] During the validation phase, the validation set is fed into the trained hybrid model, and comprehensive performance metrics such as precision, recall, and F1 value are calculated. If the model achieves an F1 value of 0.92 or higher on the validation set, it is considered the optimal color difference detection model. This model can output a spatial distribution map of abnormal color areas, such as marking red patches (indicating excessive color difference) at the corners of aluminum trim. Combined with a color diffusion model (based on Fick's diffusion law to simulate color diffusion paths), the coordinate range of the excessive color difference area (e.g., X = 10-25mm, Y = 5-15mm) is determined, providing precise location information for manual re-evaluation or robotic re-spraying.

[0129] 2. Implementation Process and Application of Defect Localization

[0130] The defect localization process first extracts texture features, shape parameters, and grayscale distribution data from the defect area based on the coating quality characteristic model, constructing a set of defect feature vectors. Taking the common sag defect as an example, texture features are extracted using a two-dimensional Fourier transform to determine the proportion of low-frequency components (typically, the low-frequency texture energy of a sag defect exceeds 60%). Shape parameters include the length-to-width ratio of the defect area (sag defects are often long strips with an aspect ratio >3:1). Grayscale distribution data is processed using histogram equalization to obtain the mean and variance (the grayscale mean of a sag area is typically 10-15 grayscale levels lower than that of a normal coating, with a variance increased by 5-8 units). A vector containing 20 features is constructed for each defect sample, such as [low-frequency energy, aspect ratio, grayscale mean, grayscale variance, etc.].

[0131] Because the original feature vectors are high-dimensional (20 dimensions), independent component analysis (ICA) is used for dimensionality reduction. ICA decomposes the original features into independent core components by maximizing the non-Gaussianity between features. For example, after ICA processing 1,000 sagging defect samples, the first five principal components were extracted, with a cumulative variance contribution of 90%, forming a five-dimensional core defect feature component [s1, s2, s3, s4, s5]. S1 corresponds to texture complexity, S2 to shape regularity, and S3 to grayscale uniformity.

[0132] The reduced feature vector is then fed into a density-based defect classification model (such as the DBSCAN algorithm). The optimal number of clusters is determined using the silhouette coefficient algorithm. For a sample set containing three defects (sagging, orange peel, and particles), the number of clusters is initially set from k=2 to k=5. The silhouette coefficient corresponding to each k value is calculated (ranging from -1 to 1, with larger values ​​indicating better clustering). When k=3, the silhouette coefficient reaches a maximum of 0.78, thus determining the optimal number of clusters to be 3, corresponding to each of the three defect types.

[0133] The classified defect features are input into the preset process adjustment library to match the optimal compensation solution. The process adjustment library stores the mapping relationship between defect types and compensation measures, for example:

[0134] Sagging defect: match the compensation instruction of "reducing the spraying speed by 10%-15% and increasing the atomization pressure by 5kPa";

[0135] Orange peel defect: Match the compensation instruction of "increasing the coating viscosity by 10 mPa·s and adjusting the curing temperature curve (reducing the heating rate by 2°C / min)"

[0136] Particle defects: Match the compensation instruction of "start secondary cleaning of electrostatic precipitator and check paint filter device".

[0137] Based on the matching results, the system generates a set of targeted parameter correction instructions, which are then sent to the coating control layer via a distributed controller for execution. For example, if a sagging defect is detected on the surface of an aluminum trim strip, the system automatically sends a command to the spray robot cluster to adjust the spray gun movement speed from 200mm / s to 170mm / s. It also sends a command to the atomizer pressure controller to increase the pressure from 200kPa to 205kPa. The adjusted process parameters are then recorded in the process database for future reference.

[0138] 3. Collaborative Application of Hybrid Models and Clustering Algorithms

[0139] In actual production, color difference detection and defect location are often performed simultaneously. For example, after a batch of aluminum trim was painted, the hybrid model first detected a localized color deviation (ΔEab = 4.5, exceeding the threshold ΔEab = 3.0). Simultaneously, the defect classification model identified the presence of a particle defect in this area. Combining these two results, the system determined that the color difference was caused by a particle defect (abnormal reflectivity on the particle surface caused the color deviation). Therefore, the particle defect compensation solution was prioritized: after a partial cleaning of the area, a spare spray gun was used for re-spraying. During re-spraying, the process parameters of low flow (50ml / min) and high atomization pressure (250kPa) were used to avoid secondary sagging.

[0140] Throughout the entire process, the system implements a closed-loop inspection and control system through real-time data streaming: the quality inspection layer outputs inspection results every 500ms, and the coating control layer completes parameter adjustments and provides feedback to the actuator within 1s. Inspection accuracy can reach: color difference positioning error ≤ 0.5mm, and defect classification accuracy ≥ 90%, meeting the high-precision requirements of automotive parts coating.

[0141] Example 5:

[0142] Example 5 of the automated control-based aluminum trim color coating optimization system involves deep learning process parameter compensation decision-making and parameter correction instruction generation based on a coating quality feature model. This implementation achieves intelligent adjustment and dynamic optimization of coating process parameters by constructing a process compensation feature library, designing a long-short-term memory network model, and optimizing parameter search algorithms. The following describes the process parameters in detail using specific scenarios:

[0143] 1. Construction and data processing of process compensation feature library

[0144] In a production line for automotive aluminum trim, the system collects data on spray equipment motion parameters, paint rheological properties, and ambient temperature and humidity based on a coating quality characteristic model to construct a process compensation feature library. For example, spray equipment motion parameters include spray gun movement speed (range: 150-300 mm / s), spray distance (range: 100-200 mm), and spray gun angle (range: 0-90°); paint rheological properties include viscosity (range: 20-50 mPa·s), solids content (range: 40%-60%), and density (range: 1.1-1.3 g / cm³); and ambient temperature and humidity data include temperature (range: 20-30°C) and relative humidity (range: 40%-60%). Each parameter is collected in a time series format at a sampling frequency of 10 Hz, forming a multidimensional time series dataset.

[0145] The data in the process compensation feature library are reorganized into time series to generate process sample segments. For example, 10 seconds of continuous data (including 100 sampling points) is used as a sample segment, and each sample segment contains the spray gun movement speed sequence [v1, v2, ..., v 100 ]、Paint viscosity sequence [μ1,μ2,…,μ 100 ] and other multi-dimensional parameters. For different process stages (such as primer spraying, paint spraying, and varnish curing), independent sample segment sets are established to ensure that each sample segment corresponds to a complete process operation cycle.

[0146] 2. Training and Application of Long Short-Term Memory Network Model

[0147] A long short-term memory (LSTM) model was established to analyze the correlation characteristics of process parameters. The LSTM model was configured with 128 memory cells, and the forget gate parameter was initialized to 0.8. The input layer received multidimensional time series data from the process compensation feature library, and the output layer was a fully connected layer used to predict coating quality characteristics (such as film thickness, gloss, and color difference). During training, a stochastic gradient descent algorithm was used, with a learning rate of 0.001, a batch size of 32, and 50 training rounds.

[0148] Backpropagation calculations are used to extract the associated features of process parameters. For example, after model training, it was found that the spray gun movement speed has a negative correlation with film thickness (the faster the speed, the thinner the film), while the paint viscosity has a positive correlation with surface roughness (the higher the viscosity, the rougher the surface). When the film thickness of a batch of aluminum trim is detected to be below the standard value, the system automatically identifies the possibility of excessive spray gun movement speed based on the associated feature analysis of the LSTM model, and makes this a key parameter to adjust.

[0149] The associated features are input into a classifier for compensation decision classification. The classifier, using a support vector machine (SVM), categorizes process parameters into two categories: "maintain" and "dynamically adjust." For example, if the LSTM model predicts a film thickness deviation exceeding ±5%, the SVM classifier outputs a "dynamically adjust" decision; if the predicted deviation is within ±2%, the SVM classifier outputs a "maintain" decision. For parameters requiring adjustment, the parameter correction instruction generation module further determines the specific adjustment value.

[0150] 3. Generation of parameter correction instructions and closed-loop control

[0151] The generation of parameter correction instructions is based on a mapping table of defect types and compensation methods. For example, when an orange peel defect is detected on the surface of an aluminum trim, the system matches the corresponding compensation method in the mapping table:

[0152] If the texture wavelength of the orange peel defect is in the range of 0.5-1mm, the corresponding method is to "increase the atomization pressure by 5-10kPa";

[0153] If the texture wavelength is in the range of 1-2mm, the corresponding setting is "reducing the spray gun movement speed by 10-15%";

[0154] If a decrease in glossiness is detected at the same time, the corresponding step is to "adjust the curing temperature curve (reduce the heating rate by 1-2°C / min)".

[0155] A genetic algorithm was used to search for the optimal compensation parameter combination. For example, the three parameters of spray gun speed, paint flow rate, and curing temperature were searched within the following ranges: spray gun speed 150-250 mm / s, paint flow rate 80-120 ml / min, and curing temperature 120-160°C. The genetic algorithm was set with a population size of 50, a crossover probability of 0.8, a mutation probability of 0.05, and 50 iterations. Each iteration evaluated the fitness function corresponding to each parameter combination (based on the coating quality characteristics predicted by the LSTM model), gradually evolving the optimal parameter combination.

[0156] Compensation effectiveness is evaluated in real time through a closed-loop control mechanism. Once the system generates and executes parameter correction instructions, coating quality characteristics are continuously monitored. For example, if, after adjusting the spray gun speed, the film thickness still does not reach the target value and the deviation direction remains unchanged (e.g., still too thin), a parameter reoptimization process is triggered: the search range is expanded (e.g., the spray gun speed is adjusted to 140-260 mm / s), the number of genetic algorithm iterations is increased (to 100), and the optimal parameter combination is recalculated.

[0157] 4. Training and Optimization of Hybrid Model Architecture

[0158] In practical applications, training of hybrid model architectures uses boosting methods to generate training sample sets for multiple base learners. For example, 80% of the data from the process compensation feature library is randomly sampled as the training set, and the remaining 20% ​​is used as the validation set. Stratified sampling is performed on the training set to generate five subsets, and a base learner (such as a decision tree or neural network) is trained on each subset.

[0159] Evaluate the classification performance of each base learner and select the best performing base learners for ensemble. For example, calculate the precision, recall, and F1 value of each base learner on the validation set and select the three base learners with the highest F1 values. Use an adaptive coefficient allocation algorithm to calculate the combined contribution coefficient of each base learner, assigning higher weights to base learners with better performance.

[0160] The gradient boosting mechanism is used to integrate and optimize the outputs of the base learners. For example, the first base learner predicts a 10% reduction in the spray gun's movement speed, the second base learner predicts a 15% reduction, and the third base learner predicts a 12% reduction. By comprehensively considering the predictions and weights of each base learner, the gradient boosting mechanism ultimately determines a 13% reduction in the spray gun's movement speed as the corrective instruction.

[0161] V. Collaborative Application in Actual Production

[0162] During the production of a batch of automotive aluminum trim strips, the system detected a slight orange peel defect on the coating surface, and the film thickness was slightly lower than the standard value. Based on the analysis of the coating quality characteristic model, the system initiated the parameter compensation decision process:

[0163] LSTM model analysis shows that the spray gun movement speed is negatively correlated with film thickness, while the paint viscosity is positively correlated with orange peel defects;

[0164] The SVM classifier outputs a “dynamic adjustment” decision, which requires adjusting both the spray gun movement speed and the paint viscosity;

[0165] The genetic algorithm searched for the optimal parameter combination, which resulted in a 12% reduction in the spray gun movement speed (from 200 mm / s to 176 mm / s) and a 5 mPa·s reduction in the paint viscosity (from 35 mPa·s to 30 mPa·s).

[0166] After executing the parameter correction instructions, the system continuously monitors the coating quality and confirms after 20 samples that the defects have been eliminated and the film thickness meets the standards.

[0167] Throughout the entire process, the system continuously optimizes parameter compensation strategies through real-time data updates from the process compensation feature library and dynamic learning of the LSTM model. For example, when ambient temperature and humidity change, the system automatically adjusts the curing temperature parameters to ensure that coating quality is not affected by environmental factors. Furthermore, through the integrated optimization of the hybrid model architecture, the accuracy and robustness of parameter compensation decisions are improved, reducing misjudgments caused by the limitations of a single model.

[0168] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "includes," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0169] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. An aluminum trim color coating optimization system based on automated control, characterized in that: It includes: Material pretreatment layer, coating control layer and quality inspection layer; The material pretreatment layer includes a surface cleaning device, a substrate activator, a roughness detector, and a coating adhesion tester, which are used to perform physical and chemical treatment on the surface of the aluminum trim strip, analyze the micromorphology, and perform pretreatment operations according to process parameters; The coating control layer includes a color mixing unit, a spray robot cluster, a dynamic temperature control module, and a distributed controller. It is equipped with a real-time color matching algorithm and an adaptive path planning mechanism to establish a multi-station collaborative operation channel, realize the synchronization of coating parameters and the classified transmission of process instructions, and adopt an adaptive path planning mechanism to dynamically adjust the coating trajectory and perform layered operations according to the coating type and process stage. The integrated visual sensing interface realizes the coordinated feedback of the physical coating status and digital process. The quality detection layer is used to use image analysis technology combined with a standard color difference database and a real-time coating data set to iteratively optimize the color distribution model, thereby constructing a multi-dimensionally mapped coating quality feature model. Based on this feature model, a deep learning algorithm is used to perform color difference detection, defect location, and process parameter compensation decisions.

2. The aluminum trim color coating optimization system based on automatic control according to claim 1 is characterized in that: The surface cleaning device includes at least a plasma processor, an ultrasonic cleaning tank, an electrostatic precipitator and a chemical degreasing tank, which are used to remove pollutants from the surface of the aluminum substrate; the color mixing unit includes at least a color paste metering pump, a spectral analysis module, a viscosity regulator and a mixing reactor, which are used to accurately reproduce the paint formula; the spray robot cluster includes at least a six-axis robotic arm, a high-voltage electrostatic spray gun, an atomization pressure controller and a flow monitoring sensor, which are used to execute the spray operation instructions issued by the coating control layer.

3. The aluminum trim color coating optimization system based on automatic control according to claim 1 is characterized in that: In the coating control layer, the pre-treated aluminum trim strips are coated by a spray robot cluster using a preset trajectory program. The curing environment is then regulated by a dynamic temperature control module and transmitted together with the surface data recorded by the roughness detector to a distributed controller for process optimization. The process optimization adopts multi-parameter coupling analysis technology to input the collected coating characteristics into the parameter space established by different optimization algorithms for joint processing; The method of realizing the coordinated feedback of the physical coating state and the digital process through the integrated visual sensing interface includes: realizing the coordinated feedback of the physical coating state and the digital process through the integrated visual sensing interface, exchanging the real-time imaging data of the coating, and jointly analyzing the acquired colorimetric features, realizing dynamic mapping of process parameters, completing instruction issuance, state monitoring and parameter correction, and issuing activation intensity parameters for the material pretreatment layer.

4. The aluminum trim color coating optimization system based on automated control according to claim 1 is characterized in that: The construction of the multi-dimensional mapping coating quality feature model includes: Establish parameter correlation channels and bidirectional calibration mechanisms between coating physical properties and digital feature space; The actual coating state is measured through optical inspection and texture analysis. Based on the color distribution, glossiness value, and film thickness measurement values ​​obtained from the quality inspection layer, a coating quality benchmark feature library and a defect feature library are constructed. Model parameter iteration and process simulation are performed based on online inspection data to convert the actual coating state into a high-precision digital feature space. Perform parameter identification on the coating quality characteristic model, input the real-time detected coating data into the established model, and use the feature comparison algorithm to dynamically compensate the analytical results of the model to obtain the optimized coating quality characteristic model; The coating quality feature model includes physical coating space, digital feature space, process database and interaction mechanism between modules; The physical coating space is the data source of the feature model and contains the actual state characteristics of the coating; the digital feature space forms a mapping relationship with the physical coating space, and the coating quality characteristics are mathematically characterized through multi-dimensional parametric modeling; the process database integrates historical process data and real-time detection information, and provides a benchmark data set including a color difference threshold library, a defect pattern library, and an equipment parameter library; the interaction mechanism realizes data communication between modules, and the physical coating space and the process database realize real-time collection and model update of quality parameters through standardized protocols, the physical coating space and the digital feature space transfer feature parameters through a data interface, and the digital feature space and the process database realize information interaction through a data bus.

5. The aluminum trim color coating optimization system based on automatic control according to claim 1 is characterized in that: The color difference detection using a deep learning algorithm based on the feature model includes: Based on the coating quality feature model, standard color plate data, ambient lighting parameters, and coating reflectivity values ​​are obtained to construct a color difference sample set; After feature enhancement processing, the color difference sample set is divided into a training set and a validation set; Establish a convolutional neural network-support vector machine-residual network hybrid model architecture, set the model's initial parameters, input the training set into the hybrid model for joint training, perform feature extraction through the convolutional neural network, perform boundary delineation through the support vector machine, and perform detail compensation through the residual network. Use the dynamic gradient mechanism to balance the recognition errors of different algorithms in specific detection scenarios until the model accuracy reaches the set threshold or the scheduled training rounds are completed; Input the validation set into the trained hybrid model, calculate the comprehensive performance index of the model, and select the optimal color difference detection model; Based on the optimal color difference detection model, the spatial distribution map of abnormal color areas is output, and the coordinates of the color difference exceeding the standard area are determined by combining the color diffusion model.

6. The aluminum trim color coating optimization system based on automatic control according to claim 1 is characterized in that: Defect location using a deep learning algorithm based on the feature model includes: Based on the coating quality feature model, the texture features, shape parameters, and grayscale distribution data of the defect area are extracted to construct a defect feature vector set; Independent component analysis is used to reduce the dimension of the defect feature vector to obtain the core defect feature components; Establish a defect classification model based on density clustering and determine the optimal number of classifications using the silhouette coefficient algorithm; The classified defect features are input into the preset process adjustment library to match the optimal compensation solution and generate a targeted parameter correction instruction set.

7. The aluminum trim color coating optimization system based on automatic control according to claim 1 is characterized in that: The process parameter compensation decision-making using a deep learning algorithm based on the feature model includes: Based on the coating quality characteristic model, the motion parameters of the spraying equipment, the rheological properties of the coating, and the ambient temperature and humidity data are collected to build a process compensation feature library; Reorganize the data in the process compensation feature library into time series to generate process sample segments; Establish a long short-term memory network model, set the number of memory units and forget gate parameters, and obtain the correlation characteristics of process parameters through back propagation calculation; The associated features are input into the classifier for compensation decision classification, and the discrimination results of process parameter maintenance and dynamic adjustment are output.

8. The aluminum trim color coating optimization system based on automatic control according to claim 1 is characterized in that: The construction of the multi-dimensional mapping coating quality feature model further includes: A stratified sampling mechanism is used to segment the continuous inspection data, and the data segments of each process stage are independently subjected to feature extraction; Establish a correlation matrix between data segment features and equipment parameters to record the corresponding coating quality patterns under different process conditions; The feature model is updated online through a transfer learning algorithm, and the model parameter adaptive correction mechanism is triggered when an unrecorded quality pattern is detected.

9. The aluminum trim color coating optimization system based on automatic control according to claim 1 is characterized in that: The method for generating the parameter correction instruction includes: Establish a mapping relationship table between defect types and compensation methods, including compensation methods for orange peel defects corresponding to atomization pressure adjustment and sagging defects corresponding to spray speed correction; Genetic algorithm is used to search for the best compensation parameter combination, including spray gun movement speed, paint flow rate and curing temperature; The compensation effect is evaluated in real time through a closed-loop control mechanism, and the parameter re-optimization process is triggered when the quality indicators do not meet the standards.

10. The aluminum trim color coating optimization system based on automatic control according to claim 1 is characterized in that: The training process of the hybrid model architecture includes: The Boosting method is used to generate training sample sets for multiple base learners, and the classification performance of each base learner is evaluated through stratified sampling. The adaptive coefficient allocation algorithm is used to calculate the combined contribution coefficient of each base learner, and the gradient boosting mechanism is used to integrate and optimize the output results of the base learners.

Images