A pre-warning correction method, system and device for an aluminum alloy anodic oxidation process

Abstract

The application relates to the field of metal surface treatment, and provides a pre-warning correction method, system and equipment for an aluminum alloy anodic oxidation process.The pre-warning correction method comprises collecting multidimensional data in an anodic oxidation process, wherein the multidimensional data at least comprises workpiece surface images, film layer cross-section images and process parameter time sequence data; the multidimensional data is pretreated, the pretreated image data is input into a preset image recognition model to obtain an image recognition result; the pretreated time sequence data is input into a preset time sequence prediction model to obtain a time sequence prediction result; the image recognition result and the time sequence prediction result are subjected to weighted fusion processing to generate fault pre-warning information; and corresponding oxidation process correction instructions are generated according to the fault pre-warning information, so that intelligent pre-warning and self-correction of the aluminum alloy anodic oxidation process are realized, product quality and production efficiency are significantly improved, and the waste product rate and the artificial intervention cost are reduced.

Description

Technical Field

[0001] This application relates to the field of metal surface treatment, and in particular to a method, system and equipment for early warning correction of aluminum alloy anodizing process. Background Technology

[0002] Aluminum alloy anodizing is a surface treatment process that generates an oxide film on the surface of aluminum alloys through electrochemical methods. During the anodizing process, fluctuations in process parameters can directly lead to defects in the film layer, such as microcracks, pinholes, and scorching. These defects not only affect the appearance quality of the product but can also reduce the corrosion resistance and mechanical strength of the film layer, resulting in product scrap. Traditional anodizing process monitoring mainly relies on the experience and judgment of operators and periodic spot checks. By the time defects are discovered, batch scrapping has often already occurred, making real-time early warning and online repair impossible. This makes it difficult to meet the stringent film quality requirements of high-end fields such as aerospace. Summary of the Invention

[0003] This application aims to improve at least one technical problem in the background art.

[0004] This application provides a method for early warning correction in an aluminum alloy anodizing process, comprising: Collect multi-dimensional data during the anodizing process; Multi-dimensional data are preprocessed to obtain preprocessed image data and time-series data; The preprocessed image data is input into a preset image recognition model to obtain the image recognition result; The preprocessed time series data is input into a preset time series prediction model to obtain the time series prediction results; The image recognition results and time-series prediction results are weighted and fused to generate fault warning information; Based on the fault warning information, generate corresponding oxidation process correction instructions.

[0005] According to some technical solutions of this application, the preprocessing of multi-dimensional data to obtain preprocessed image data and time-series data specifically includes: The multi-dimensional data includes at least workpiece surface images, film cross-sectional images, and time-series data of process parameters; Multi-dimensional data were extracted to obtain workpiece surface images, film cross-sectional images, and time-series data of process parameters; The workpiece surface image and the film cross-sectional image are subjected to grayscale conversion, filtering and edge processing to obtain preprocessed image data; Outlier removal and normalization are performed on the time series data of process parameters to obtain preprocessed time series data.

[0006] According to some technical solutions of this application, the weighted fusion processing of image recognition results and time-series prediction results to generate fault warning information specifically includes: Obtain the first confidence level from the image recognition results; Obtain the second confidence level from the time series prediction results; The first confidence level and the second confidence level are weighted and summed according to preset weights to obtain the warning confidence level; The warning confidence level is compared with the preset warning threshold to generate fault warning information.

[0007] According to some technical solutions of this application, the step of comparing the warning confidence level with a preset warning threshold to generate fault warning information specifically includes: The preset warning threshold includes a first preset threshold and a second preset threshold, and the first preset threshold is less than the second preset threshold; The warning confidence level is compared with the first preset threshold and the second preset threshold; When the warning confidence level is less than the first preset threshold, a fault-free warning message is generated. When the warning confidence level is greater than or equal to the first preset threshold and less than the second preset threshold, a minor fault warning message is generated. When the warning confidence level is greater than or equal to the second preset threshold, a severe fault warning message is generated.

[0008] According to some technical solutions of this application, the step of generating corresponding oxidation process correction instructions based on fault warning information specifically includes: The fault warning information is analyzed to obtain the fault level information; Match the corresponding correction strategy from the preset correction strategy library based on the fault level information; The corresponding oxidation process correction instructions are generated based on the correction strategy.

[0009] According to some technical solutions of this application, before collecting multi-dimensional data during the anodizing process, the method further includes: Acquire historical operating condition data of the anodizing process, including historical workpiece surface images, historical film cross-sectional images, and historical process parameter time series data; Defect areas are annotated on historical workpiece surface images and historical film cross-sectional images to obtain annotated image samples; Construct the initial identification model and the initial prediction model; The initial recognition model is trained using labeled image samples to obtain a preset image recognition model; The preset time series prediction model is obtained by training with historical process parameter time series data.

[0010] According to some technical solutions of this application, the step of inputting the preprocessed image data into a preset image recognition model to obtain the image recognition result specifically includes: The preset image recognition model is a convolutional neural network model; The preprocessed image data is input into the convolutional neural network model; Image features are obtained by performing convolution and pooling on image data based on a convolutional neural network model. The image features are classified and identified to obtain the fault type and the corresponding first confidence level; The fault type and the first confidence level are integrated to obtain the image recognition result.

[0011] According to some technical solutions of this application, the step of inputting the preprocessed time series data into a preset time series prediction model to obtain the time series prediction result specifically includes: The preset time-series prediction model is a long short-term memory network model; The preprocessed time series data is input into the long short-term memory network model; Feature extraction of time-series data is performed based on a long short-term memory network model to obtain time-series features; Predict the probability of failure occurrence and the corresponding second confidence level by analyzing time-series features; The probability of failure occurrence and the second confidence level are integrated to obtain the time series prediction result. This application also provides an early warning and correction system for an aluminum alloy anodizing process, comprising: The data acquisition module is used to collect multi-dimensional data during the anodizing process; The preprocessing module is used to preprocess multi-dimensional data to obtain preprocessed image data and time-series data; The recognition module is used to input the preprocessed image data into a preset image recognition model to obtain the image recognition result; The prediction module is used to input the preprocessed time series data into a preset time series prediction model to obtain the time series prediction results; The fusion module is used to perform weighted fusion processing on image recognition results and time-series prediction results to generate fault warning information; The correction module is used to generate corresponding oxidation process correction instructions based on fault warning information.

[0012] This application also provides a warning correction device for an aluminum alloy anodizing process, the warning correction device for an aluminum alloy anodizing process comprising: a memory and at least one processor, the memory storing instructions; at least one processor calling the instructions in the memory to cause the warning correction device for an aluminum alloy anodizing process to perform the various steps of the warning correction method for an aluminum alloy anodizing process as described above.

[0013] The early warning and correction method for aluminum alloy anodizing process provided in this application has at least the following beneficial effects: by generating a comprehensive early warning confidence level and classifying early warning information of different degrees according to the confidence level, after generating fault early warning information, the corresponding anodizing process correction instruction is directly generated, realizing intelligent early warning and self-correction of aluminum alloy anodizing process, significantly improving product quality and production efficiency, and reducing scrap rate and manual intervention cost. Attached Figure Description

[0014] Figure 1 A flowchart of an early warning correction method for an aluminum alloy anodizing process provided in this application embodiment; Figure 2 This is a structural diagram of the early warning and correction system for the aluminum alloy anodizing process provided in the embodiments of this application; Figure 3 This is a schematic diagram of the structure of the early warning and correction device for the aluminum alloy anodizing process provided in the embodiments of this application. Detailed Implementation

[0015] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0016] In the description of this application, it should be understood that the orientation descriptions, such as up, down, front, back, left, right, etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation or be constructed or operated in a specific orientation. Therefore, they should not be construed as limiting the present invention.

[0017] In the description of this application, unless otherwise expressly defined, terms such as "setup," "installation," and "connection" should be interpreted broadly, and those skilled in the art can reasonably determine the specific meaning of the above terms in this application in conjunction with the specific content of the technical solution.

[0018] The following is combined with Figures 1 to 3 Embodiments of the present invention will be described.

[0019] This application provides a method for early warning correction in an aluminum alloy anodizing process, comprising: S100, collect multi-dimensional data during the anodizing process, the multi-dimensional data including at least workpiece surface image, film cross-sectional image and process parameter timing data; For example, in the oxidation tank, an in-situ industrial camera for capturing surface images, a spectroscopic ellipsometry for acquiring cross-sectional images of the film layer, and various other sensors are provided. The workpiece is immersed in the electrolyte for anodizing. The in-situ industrial camera captures surface images of the workpiece in real time; the spectroscopic ellipsometry simultaneously acquires cross-sectional images of the film layer; the sensors in the electrolyte acquire pH value, Al³⁺ concentration, and temperature every 200ms; the pulse power acquisition module records current and voltage values ​​every 100ms; and the film thickness sensor acquires real-time film thickness every 1 second.

[0020] S200 preprocesses multi-dimensional data to obtain preprocessed image data and time-series data. During preprocessing, the image data is converted to grayscale, Gaussian filtered, and Canny edge detected to extract defect feature regions. Outliers in the time-series data are removed. For example, data in the time-series data that exceed the mean ± 3 times the standard deviation are replaced with the moving average of the previous 5 data and normalized to obtain the preprocessed image data and time-series data.

[0021] S300: The preprocessed image data is input into a preset image recognition model to obtain the image recognition result; for example, a convolutional neural network (CNN) is used, with an architecture of 3 convolutional layers, 2 max pooling layers, 2 fully connected layers, and 1 output layer. The convolutional kernels of the convolutional layers are 3×3, with 32, 64, and 128 respectively; the pooling kernels of the pooling layers are 2×2; and the neurons of the fully connected layers are 512 and 128. The preprocessed image data is input into the CNN, and the model outputs the image recognition result. The image recognition result includes the fault type of the current image and the corresponding first confidence level. For example, fault type 0 represents no fault, fault type 1 represents microcracks, fault type 2 represents pinholes, and fault type 3 represents charring.

[0022] S400: The preprocessed time-series data is input into a pre-defined time-series prediction model to obtain the time-series prediction results. For example, the time-series prediction model uses a Long Short-Term Memory (LSTM) network, whose architecture consists of one input layer, two LSTM hidden layers, and one fully connected output layer. The input layer includes five neurons, corresponding to five features: pH, Al³⁺ concentration, current, voltage, and membrane thickness. Each LSTM hidden layer has 64 neurons. The preprocessed time-series data is input into the LSTM, and the model outputs the probability of a fault occurring within the next 5 minutes and the corresponding second confidence level.

[0023] S500 performs weighted fusion processing on image recognition results and temporal prediction results to generate fault warning information. For example, it adjusts the weight ratio of CNN and LSTM in real time according to the fault type, and combines the confidence calculation formula Conf = w1×Acc_CNN + w2×Acc_LSTM to ensure the reliability of fusion, where Conf is the warning confidence, Acc_CNN is the recognition accuracy of the CNN model, Acc_LSTM is the prediction accuracy of the LSTM model, w1 is the weight coefficient of the CNN model, and w2 is the weight coefficient of the LSTM model.

[0024] S600 generates corresponding oxidation process correction instructions based on fault warning information. By parsing the fault warning information and determining the fault level to be severe, it matches a severe correction strategy from the correction strategy library and generates corresponding correction instructions. These instructions include pausing oxidation, discharging 50% of the old electrolyte, replenishing with fresh self-healing electrolyte, immersing in a 5% nitric acid solution for 30 seconds to remove the charred layer, and then restarting oxidation with initial parameters. Optionally, the receiving and executing devices include an electrolyte replenishment metering pump, a pulse power supply, a discharge valve, and a surface pretreatment unit.

[0025] Therefore, by generating a comprehensive early warning confidence level and classifying early warning information into different levels according to the confidence level, the corresponding anodizing process correction command is directly generated after generating fault early warning information. This achieves intelligent early warning and self-correction of the aluminum alloy anodizing process, significantly improving product quality and production efficiency, and reducing scrap rate and manual intervention costs. It helps to solve the technical problems in existing anodizing process monitoring technologies, such as single fault identification dimension, low early warning accuracy, lack of online repair capability, and poor model generalization. In addition, it does not depend on a specific aluminum alloy grade or process formula, and can be compatible with the anodizing process of all aluminum alloys from 1 series to 7 series through model training and transfer learning of multi-grade data. At the same time, each module in the solution can be independently optimized and upgraded, facilitating the subsequent introduction of new sensor types or improved algorithm models, and has good technical continuity and expansion space.

[0026] In some embodiments, S200 involves preprocessing the multi-dimensional data to obtain preprocessed image data and time-series data, specifically including: the multi-dimensional data includes at least workpiece surface images, film cross-sectional images, and process parameter time-series data; S210: Extract multi-dimensional data to obtain workpiece surface images, film cross-sectional images, and time-series data of process parameters; separate image data and time-series data from the received raw data packet. The image data includes workpiece surface images and film cross-sectional images, while the time-series data includes: pH value, Al³⁺ concentration, temperature, current, voltage, and film thickness.

[0027] S220 performs grayscale conversion, filtering, and edge processing on the workpiece surface image and the film cross-section image to obtain preprocessed image data. The RGB three-channel image is converted to a single-channel grayscale image using the formula Gray = 0.299R + 0.587G + 0.114B. Gaussian filtering is used to remove noise with a kernel size of 5×5 and a standard deviation σ=1.0, resulting in a smooth image while preserving edge information. Edge processing uses the Canny edge detection algorithm with a low threshold of 50 and a high threshold of 150 to extract the edge contours of film defects and mark potential cracks and pinholes. The processed image data is saved as preprocessed image data.

[0028] S230. Outlier removal and normalization are performed on the time-series data of process parameters to obtain preprocessed time-series data. Outlier removal involves calculating the mean μ and standard deviation σ for each parameter sequence. Data points exceeding the range [μ-3σ, μ+3σ] are considered outliers and replaced with the moving average of the previous 5 normal values. For example, if the pH value suddenly changes to 2.5 at a certain moment, while the normal range is 3.5-4.5, then the average of the previous 5 pH values ​​(3.8) is used instead. Normalization normalizes each parameter to the [0,1] interval using the formula X_norm = (X - X_min) / (X_max - X_min), where X_min and X_max are the minimum and maximum values ​​of the parameter's historical data. Normalized time-series data eliminates the influence of dimensions, facilitating model processing. Thus, through the above steps, preprocessed image data and time-series data are obtained for subsequent model use.

[0029] In some embodiments, step S500 involves weighted fusion processing of the image recognition results and the time-series prediction results to generate fault warning information, specifically including: S510, Obtain the first confidence level from the image recognition result; for example, the image recognition model outputs the fault type as "microcrack" and the softmax probability value of this category as the first confidence level. For example, for the current image, the model outputs the probabilities of each category as no fault 0.02, microcrack 0.95, pinhole 0.02, and char 0.01, then the first confidence level is 0.95.

[0030] S520, Obtain the second confidence level from the time series prediction results; for example, the time series prediction model outputs the probability of a fault occurring within the next 5 minutes, and the historical accuracy obtained by evaluating the model on the validation set is used as the second confidence level. For example, if the LSTM outputs a fault probability of 98%, and the model's accuracy on the validation set is 0.98, then the second confidence level is 0.98.

[0031] S530, Weighted sum the first confidence level and the second confidence level according to a preset weight to obtain a warning confidence level; the preset weight coefficient is optimized and determined from historical data. In this embodiment, the preset w1 = 0.4 and the weight w2 = 0.6. Exemplarily, calculate the warning confidence level Conf = 0.4×0.95 + 0.6×0.98 = 0.968.

[0032] S540, Compare the warning confidence level with a preset warning threshold to generate a fault warning message. Exemplarily, the preset warning threshold includes a first threshold 0.8 and a second threshold 0.95. Since Conf = 0.968 ≥ 0.95, it is determined as a severe warning, and a fault warning message including "severe warning" and the fault type "micro crack" is generated. If Conf is between 0.8 and 0.95, a mild warning message is generated; if Conf < 0.8, a no-fault message is generated.

[0033] In some embodiments, in S540, comparing the warning confidence level with a preset warning threshold to generate a fault warning message specifically includes: The preset warning threshold includes a first preset threshold and a second preset threshold, and the first preset threshold is less than the second preset threshold; based on the statistical analysis of historical fault data, set the first preset threshold T1 = 0.8, the second preset threshold T2 = 0.95, and T1 < T2. T1 corresponds to a relatively low probability of fault occurrence, and T2 corresponds to a high probability.

[0034] S541, Compare the warning confidence level with the first preset threshold and the second preset threshold; compare the warning confidence level Conf with T1 and T2. For example, if the calculation result of Conf is 0.88, the following judgment is made: S542, When the warning confidence level is less than the first preset threshold, generate a no-fault warning message; when Conf < T1, generate a no-fault warning message. If Conf = 0.75, generate a "no fault" message, and the system continues to run normally.

[0035] S543, When the warning confidence level is greater than or equal to the first preset threshold and less than the second preset threshold, then generate a mild fault warning message; when T1 ≤ Conf < T2, generate a mild fault warning message. If Conf = 0.88, generate a "mild warning" message and attach the fault type.

[0036] S544, When the warning confidence level is greater than or equal to the second preset threshold, then generate a severe fault warning message. When Conf ≥ T2, generate a severe fault warning message. If Conf = 0.968, generate a "severe warning" message and attach the fault type.

[0037] In some embodiments, step S600 generates a corresponding oxidation process correction instruction based on the fault warning information, specifically including: S610, parse the fault warning information to obtain fault level information; for example, if the current fault warning information is a severe warning and the fault type is microcrack, parse the warning information and extract the fault level as "severe" and the fault type as "microcrack".

[0038] S620 matches the corresponding correction strategy from the preset correction strategy library based on the fault level information; the correction strategy library pre-stores correction schemes corresponding to different fault levels: For mild warnings, replenish the self-healing electrolyte, for example, by adjusting the pulse power on / off ratio from 1:1.5 to 1:1.8, reducing the peak voltage by 2V, and continuing for 3 minutes. For severe warnings, suspend oxidation, for example, by draining 50% of the old electrolyte, replenishing with freshly prepared self-healing electrolyte, injecting a 5% nitric acid solution to soak for 30 seconds to remove the charred layer, and restarting oxidation at 90% of the initial parameters.

[0039] S630 generates corresponding oxidation process correction instructions based on the correction strategy. For example, in the event of a severe warning, it sends an instruction to "start discharge, rate 5L / min, discharge to 50% liquid level".

[0040] In some embodiments, before collecting multi-dimensional data during the anodizing process, step S100 further includes: S101, acquire historical operating condition data of the anodizing process, including historical workpiece surface images, historical film cross-sectional images, and historical process parameter time series data; extract multiple sets of anodizing production data within a certain period of time from the database, including historical workpiece surface images, historical film cross-sectional images, and historical process parameter time series data of normal and defective samples, such as pH, Al³⁺ concentration, temperature, current, voltage, film thickness, etc., sampling frequency of 1Hz, etc., and record the final film thickness deviation, salt spray resistance time, and other product quality test results for each production.

[0041] S102, mark the defect areas of historical workpiece surface images and historical film cross-sectional images to obtain marked image samples; use the annotation tool to select the defect areas in the images and classify them as microcracks, pinholes, scorching, etc., and generate a corresponding annotation file in XML format for each image, recording the defect type and location coordinates, forming marked image samples.

[0042] S103, Construct the initial recognition model and the initial prediction model. The initial recognition model uses an untrained convolutional neural network with an architecture of 3 convolutional layers, 2 max-pooling layers, 2 fully connected layers, and 1 output layer. The convolutional kernels are 3×3, with 32, 64, and 128 neurons respectively; the pooling kernels are 2×2; and the fully connected layers have 512 and 128 neurons respectively. The preprocessed image data is input into the CNN, and the model outputs the image recognition result, which includes the fault type of the current image and the corresponding first confidence score.

[0043] The initial prediction model architecture is an untrained Long Short-Term Memory (LSTM) network, consisting of one input layer, two LSTM hidden layers, and one fully connected output layer. The input layer comprises five neurons corresponding to five features: pH, Al³⁺ concentration, current, voltage, and membrane thickness. Each LSTM hidden layer has 64 neurons. Preprocessed time-series data is input into the LSTM, and the model outputs the probability of a fault occurring within the next five minutes and the corresponding second confidence level.

[0044] S104: Train the initial recognition model using labeled image samples to obtain the preset image recognition model. Divide the labeled image samples into training, validation, and test sets in a 7:2:1 ratio. Supervised training of the CNN is performed using the training set, with cross-entropy loss as the loss function, Adam as the optimizer, and a learning rate of 0.001. The accuracy is evaluated on the validation set after each training round, and training stops when the validation set accuracy no longer improves. Finally, the model is evaluated using the test set, yielding a recognition accuracy of 99.2%, which meets the preset requirements. This model is then saved and confirmed as the preset image recognition model.

[0045] S105 utilizes historical process parameter time-series data for training to obtain a pre-defined time-series prediction model. The historical process parameter time-series data is divided into time windows, each containing data from the first 30 minutes, predicting the probability of failure occurring in the next 5 minutes. Fault events are labeled on the data; for example, based on actual production records, if a fault occurs within the next 5 minutes, it is labeled as 1, otherwise 0. The training set, validation set, and test set are divided in a 7:2:1 ratio. The LSTM is trained using the training set, with a binary cross-entropy loss function, an Adam optimizer, and a learning rate of 0.001. The accuracy on the validation set is monitored during training to prevent overfitting. The test set evaluation shows a prediction accuracy of 98.8%, meeting the pre-defined requirements, and this is saved as the pre-defined time-series prediction model.

[0046] In some embodiments, step S300 involves inputting the preprocessed image data into a preset image recognition model to obtain an image recognition result, specifically including: The preset image recognition model is a convolutional neural network model; S310, input the preprocessed image data into the convolutional neural network model; S320 performs convolution and pooling processing on image data based on a convolutional neural network model to obtain image features; S330 classifies and identifies image features to obtain the fault type and its corresponding first confidence level. Four neurons correspond to four fault types. The activation function Softmax outputs the probability of each category, and the category with the highest probability is taken as the fault type. This probability value is the first confidence level. For example, if the output probability is [0.02, 0.95, 0.02, 0.01], the fault type is "microcrack," and the first confidence level is 0.95.

[0047] S340 integrates the fault type and the first confidence level to obtain the image recognition result. The fault type code and the first confidence level are encapsulated into JSON format data and output as the image recognition result.

[0048] In some embodiments, step S400 involves inputting the preprocessed time series data into a preset time series prediction model to obtain time series prediction results, specifically including: The preset time-series prediction model is a long short-term memory network model. S410, input the preprocessed time series data into the long short-term memory network model; S420 uses a long short-term memory network model to extract features from time-series data to obtain time-series features. The preprocessed time-series data consists of five features: pH, Al³⁺ concentration, current, voltage, film thickness, and a sequence of the past 60 time steps, i.e., a sequence of one point every 10 seconds over the past 10 minutes, with a shape of 60×5, which is input into the input layer of the LSTM.

[0049] S430 predicts the time-series features to obtain the probability of failure and the corresponding second confidence level; the output value is the probability of failure occurring within the next 5 minutes, which is between 0 and 1. Simultaneously, the model internally stores the accuracy evaluated on the validation set, which serves as the second confidence level. For example, if the current input yields a failure probability of 0.98, and the model validation accuracy is 0.988, then the second confidence level is 0.988.

[0050] S440 integrates the failure probability and the second confidence level to obtain the time series prediction result. The failure probability and the second confidence level are encapsulated as JSON data and output as the time series prediction result.

[0051] In some optional embodiments, before processing, the material information and target film parameters of the workpiece to be processed are first obtained; the reference process parameters are matched from the process matching library according to the material information and target film parameters; the corresponding acquisition device start command is generated according to the reference process parameters; for example, on the production line of aerospace aluminum alloy parts, the material of the workpiece to be processed is 7075 aluminum alloy, the target film thickness is 30μm, and the salt spray resistance time is 1200h. After obtaining the material information and target film parameters, the process database is called according to the input workpiece information, the reference process parameters are matched, and the hardware device is started to collect data.

[0052] This application also provides an early warning and correction system for aluminum alloy anodizing process, which includes a data acquisition module 100, a preprocessing module 200, an identification module 300, a prediction module 400, a fusion module 500, and a correction module 600. The system comprises the following modules: an acquisition module for acquiring multi-dimensional data during the anodizing process; a preprocessing module for preprocessing the multi-dimensional data to obtain preprocessed image data and time-series data; an identification module for inputting the preprocessed image data into a preset image recognition model to obtain image recognition results; a prediction module for inputting the preprocessed time-series data into a preset time-series prediction model to obtain time-series prediction results; a fusion module for performing weighted fusion processing on the image recognition results and time-series prediction results to generate fault warning information; and a correction module for generating corresponding anodizing process correction instructions based on the fault warning information.

[0053] Figure 3 This is a schematic diagram of a pre-warning and correction device for an aluminum alloy anodizing process provided in an embodiment of the present invention. The pre-warning and correction device 700 for the aluminum alloy anodizing process can vary significantly due to different configurations or performance. It may include one or more central processing units (CPUs) 710 (e.g., one or more processors) and a memory 720, and one or more storage media 730 (e.g., one or more mass storage devices) storing application programs 733 or data 732. The memory 720 and storage media 730 can be temporary or persistent storage. The program stored in the storage media 730 may include one or more modules (not shown in the diagram), each module may include a series of instruction operations on the pre-warning and correction device 700 for the aluminum alloy anodizing process. Furthermore, the processor 710 may be configured to communicate with the storage media 730 and execute a series of instruction operations in the storage media 730 on the pre-warning and correction device 700 for the aluminum alloy anodizing process to implement the steps of the pre-warning and correction method for the aluminum alloy anodizing process provided in the above-described method embodiments.

[0054] The early warning and correction device 700 for aluminum alloy anodizing processes may also include one or more power supplies 740, one or more wired or wireless network interfaces 750, one or more input / output interfaces 760, and / or one or more operating systems 731, such as Windows Server, Mac OS X, Unix, Linux, FreeBSD, etc. Those skilled in the art will understand that... Figure 3 The illustrated structure of the early warning correction device for the aluminum alloy anodizing process does not constitute a limitation on the early warning correction device for the aluminum alloy anodizing process. It may include more or fewer components than illustrated, or combine certain components, or have different component arrangements.

[0055] The present invention also provides a computer-readable storage medium, which can be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium, wherein the computer-readable storage medium stores instructions that, when executed on a computer, cause the computer to perform the steps of a warning correction method for an aluminum alloy anodizing process.

[0056] The preferred embodiments of the present invention have been described in detail above, but the present disclosure is not limited to the embodiments described. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of the present invention, and these equivalent modifications or substitutions are all included within the scope defined by the claims of the present disclosure.

Claims

1. A method for early warning and correction in an aluminum alloy anodizing process, characterized in that: include: Collect multi-dimensional data during the anodizing process; Multi-dimensional data are preprocessed to obtain preprocessed image data and time-series data; The preprocessed image data is input into a preset image recognition model to obtain the image recognition result; The preprocessed time series data is input into a preset time series prediction model to obtain the time series prediction results; The image recognition results and time-series prediction results are weighted and fused to generate fault warning information; Based on the fault warning information, generate corresponding oxidation process correction instructions.

2. The early warning and correction method for aluminum alloy anodizing process according to claim 1, characterized in that: The preprocessing of multi-dimensional data to obtain preprocessed image data and time-series data specifically includes: The multi-dimensional data includes at least workpiece surface images, film cross-sectional images, and time-series data of process parameters; Multi-dimensional data were extracted to obtain workpiece surface images, film cross-sectional images, and time-series data of process parameters; The workpiece surface image and the film cross-sectional image are subjected to grayscale conversion, filtering, and edge processing to obtain preprocessed image data; Outlier removal and normalization are performed on the time series data of process parameters to obtain preprocessed time series data.

3. The early warning and correction method for aluminum alloy anodizing process according to claim 1, characterized in that: The weighted fusion processing of image recognition results and time-series prediction results to generate fault warning information specifically includes: Obtain the first confidence level from the image recognition results; Obtain the second confidence level from the time series prediction results; The first confidence level and the second confidence level are weighted and summed according to preset weights to obtain the warning confidence level; The warning confidence level is compared with the preset warning threshold to generate fault warning information.

4. The early warning and correction method for aluminum alloy anodizing process according to claim 1, characterized in that: The step of comparing the warning confidence level with a preset warning threshold to generate fault warning information specifically includes: The preset warning threshold includes a first preset threshold and a second preset threshold, and the first preset threshold is less than the second preset threshold; The warning confidence level is compared with the first preset threshold and the second preset threshold; When the warning confidence level is less than the first preset threshold, a fault-free warning message is generated; When the warning confidence level is greater than or equal to the first preset threshold and less than the second preset threshold, a minor fault warning message is generated. When the warning confidence level is greater than or equal to the second preset threshold, a severe fault warning message is generated.

5. The early warning and correction method for aluminum alloy anodizing process according to claim 1, characterized in that: The step of generating corresponding oxidation process correction instructions based on fault warning information specifically includes: The fault warning information is analyzed to obtain the fault level information; Match the corresponding correction strategy from the preset correction strategy library based on the fault level information; The corresponding oxidation process correction instructions are generated based on the correction strategy.

6. The early warning and correction method for aluminum alloy anodizing process according to claim 1, characterized in that: Before collecting multi-dimensional data during the anodizing process, the following steps are also included: Acquire historical operating condition data of the anodizing process, including historical workpiece surface images, historical film cross-sectional images, and historical process parameter time series data; Defect areas are annotated on historical workpiece surface images and historical film cross-sectional images to obtain annotated image samples; Construct the initial identification model and the initial prediction model; The initial recognition model is trained using labeled image samples to obtain a preset image recognition model; The preset time series prediction model is obtained by training with historical process parameter time series data.

7. The early warning and correction method for aluminum alloy anodizing process according to claim 1, characterized in that: The step of inputting the preprocessed image data into a preset image recognition model to obtain the image recognition result specifically includes: The preset image recognition model is a convolutional neural network model; The preprocessed image data is input into the convolutional neural network model; Image features are obtained by performing convolution and pooling on image data based on a convolutional neural network model. The image features are classified and identified to obtain the fault type and the corresponding first confidence level; The fault type and the first confidence level are integrated to obtain the image recognition result.

8. The early warning and correction method for aluminum alloy anodizing process according to claim 1, characterized in that: The step of inputting the preprocessed time series data into a preset time series prediction model to obtain the time series prediction result specifically includes: The preset time-series prediction model is a long short-term memory network model; The preprocessed time series data is input into the long short-term memory network model; Feature extraction of time-series data is performed based on a long short-term memory network model to obtain time-series features; Predict the probability of failure occurrence and the corresponding second confidence level by analyzing time-series features; The probability of failure occurrence and the second confidence level are integrated to obtain the time series prediction results.

9. A warning and correction system for anodizing aluminum alloy processes, characterized in that: include: The data acquisition module is used to collect multi-dimensional data during the anodizing process; The preprocessing module is used to preprocess multi-dimensional data to obtain preprocessed image data and time-series data; The recognition module is used to input the preprocessed image data into a preset image recognition model to obtain the image recognition result; The prediction module is used to input the preprocessed time series data into a preset time series prediction model to obtain the time series prediction results; The fusion module is used to perform weighted fusion processing on image recognition results and time-series prediction results to generate fault warning information; The correction module is used to generate corresponding oxidation process correction instructions based on fault warning information.

10. A warning and correction device for an aluminum alloy anodizing process, characterized in that, The early warning and correction device for the aluminum alloy anodizing process includes: a memory and at least one processor, wherein the memory stores instructions; At least one of the processors invokes the instructions in the memory to cause the early warning correction device for the aluminum alloy anodizing process to perform the steps of the early warning correction method for the aluminum alloy anodizing process as described in any one of claims 1-8.

Images