Remote control method and system for ceramic capacitor laminating machine

Abstract

The invention belongs to the field of remote control, and discloses a remote control method and system for a ceramic capacitor laminating machine. Micron-sized alignment precision data and crack gap detection data in the multi-layer ceramic stacking process are obtained from an automatic feeding system and a defect detection module of the ceramic capacitor stacking machine through a distributed data acquisition interface; a data cleaning algorithm is adopted to remove redundant records and abnormal values, if the data quality comprehensive score is lower than a preset threshold value, a data restoration mechanism is triggered, and a standardized ceramic lamination data set is generated; calculating a ceramic lamination complexity weight coefficient based on the feature vector of the multilayer ceramic stack, and if the weight coefficient exceeds a preset threshold, starting a high-performance computing resource pool containing a graphic processing unit cluster and distributed storage to generate a ceramic lamination complexity analysis result; generating a load prediction value according to a ceramic lamination complexity analysis result; and an optimized ceramic lamination production resource configuration state is generated, and remote control is realized.

Description

Technical Field

[0001] The present invention belongs to the field of remote control and discloses a remote control method and system for a ceramic capacitor lamination machine. Background Art

[0002] The production of ceramic capacitor stacks faces a complex technical challenge: how to achieve high production efficiency while maintaining high precision and quality. This challenge involves multiple aspects: First, the micron-level alignment accuracy and crack and void detection require extremely high precision, while large-scale production requires rapid processing of massive amounts of data, creating a conflict between the two.

[0003] Secondly, the complexity of ceramic laminate structures varies, and how to flexibly allocate computing resources based on the characteristics of different products while ensuring processing efficiency and avoiding resource waste is a thorny issue. Furthermore, achieving real-time monitoring, rapid response, and intelligent decision-making to address various anomalies in the collaborative production process of multiple devices is also a major challenge.

[0004] Furthermore, optimizing production parameters involves multiple, intertwined factors. Finding the optimal balance within a dynamically changing production environment to improve product yield is equally challenging. These interrelated issues collectively constitute the core technical challenges in the intelligent and digital transformation of ceramic capacitor stack production. Solving this series of digital economy transformation challenges not only impacts the production efficiency and product quality of individual companies but also impacts the technological advancement and competitiveness of the entire industry.

[0005] Therefore, there is an urgent need to design a remote control method and system for a ceramic capacitor stacking machine. Summary of the Invention

[0006] In order to overcome the above-mentioned defects of the prior art and to achieve the above-mentioned purpose, the present invention provides the following technical solution: a remote control method for a ceramic capacitor stacking machine, comprising: Acquire micron-level alignment accuracy data and crack and void detection data during the multilayer ceramic stacking process from the automated feeding system and defect detection module of the ceramic capacitor stacking machine through a distributed data acquisition interface; A data cleaning algorithm was used to remove redundant records and outliers. The dataset was verified based on the evaluation criteria of completeness (missing rate less than 5%), consistency (format uniformity greater than 95%), and accuracy (error rate less than 2%). If the overall data quality score fell below the preset threshold T1 = 0.85, the data repair mechanism was triggered to generate a standardized ceramic laminate dataset. The standardized ceramic stack dataset is sharded and processed in parallel using an in-memory computing engine. The ceramic stack complexity weight coefficient W = Σ(ai × xi) is calculated based on the feature vectors of the multilayer ceramic stack. Here, ai is the feature importance coefficient calculated based on information gain, ranging from [0, 1], and xi is the normalized micron-level alignment accuracy and crack and void detection feature value, ranging from [0, 1]. If W exceeds the preset threshold T2 = 0.7, a high-performance computing resource pool consisting of a graphics processing unit cluster and distributed storage is activated to generate the ceramic stack complexity analysis results. Based on the results of the ceramic stacking complexity analysis, an adaptive load balancing algorithm is used to dynamically adjust the task allocation ratio of computing nodes. A support vector machine algorithm is used to predict the load trend of the ceramic stacking production system based on environmental temperature and humidity control data and the operating status of the high-speed cutting device, generating a load prediction value R in tasks per second. Based on R and the current system utilization U (range [0,1]), the expansion factor S = R × (1 + U) / 1000 is calculated, where 1000 is the preset benchmark load value, in tasks per second. If S exceeds the preset threshold T3 = 1.2, the elastic expansion mechanism is triggered to generate an optimized ceramic laminate production resource configuration state.

[0007] Preferably, the method of acquiring micron-level alignment accuracy data and crack and gap detection data during the multilayer ceramic stacking process from the automated feeding system and defect detection module of the ceramic capacitor stacking machine through the distributed data acquisition interface specifically includes: Acquire alignment accuracy data from the feeding system of the ceramic capacitor stacking machine through the OPCUA protocol, obtain crack and gap data from the defect detection module, and merge them to generate the original data set; The drop_duplicates function of Pandas is used to remove duplicate records from the original dataset, and the quantile function is used to remove outliers outside the upper and lower 1% quantiles to generate a cleaned dataset; If the missing value rate of the cleaned dataset is higher than the preset threshold of 15%, linear interpolation is used to fill in the missing values ​​to generate a completed dataset; According to pre-established format rules, the field types and unit uniformity of the completed data set are verified to determine a format uniformity rate.

[0008] Preferably, the sharding and parallel processing of the standardized ceramic laminate dataset by the memory computing engine specifically includes: The Apache Spark memory computing framework is used to process the standardized ceramic stack dataset; The data output by Spark is partitioned using consistent hashing using the libketama library and stored in 64 data shards of the HDFS distributed file system. Based on the sharding distribution structure, 16 computing nodes were started using the MPI parallel framework. Each node used OpenCV's SIFT algorithm to extract micron-level alignment coordinate features and the ResNet50 pre-trained model to extract crack and void probability features, generating a 256-dimensional feature vector set. Input the feature vector set into the mutual_info_classif function of scikit-learn, calculate the information gain of each feature, and generate a set of feature importance coefficients; MinMaxScaler is used to perform linear normalization on the eigenvector set, compressing all eigenvalues ​​to the range of 0 to 1 to generate a normalized eigenvalue set.

[0009] Preferably, the dynamically adjusting the task allocation ratio of computing nodes by the adaptive load balancing algorithm specifically includes: The allocation weight of each computing node is calculated by a weighted round-robin algorithm. The weight formula is 1 / (1+the current load rate of the node) to obtain a node task allocation table. Acquire real-time temperature and humidity data from the ambient temperature and humidity sensor, analyze equipment fault codes from the cutting device log, generate compensation coefficients if the temperature and humidity exceed the preset range, and mark the abnormal node if the fault code appears three times consecutively; The compensation coefficient, the fault mark and the node CPU usage are combined into an input vector, and a load prediction model is trained using a support vector machine algorithm to output a load prediction value for the next 5 minutes.

[0010] Preferably, the ceramic laminate production resource configuration status and ceramic laminate complexity analysis results adopt a remote control data encryption protocol and real-time instruction transmission function, decouple the functional modules of the hot pressing forming process and the high-speed cutting device through a microservice architecture, and use an application programming interface gateway to uniformly manage the data interaction and user authority verification of multi-device collaborative management, and process asynchronous production requests through a message queue. If the service response time exceeds the delay threshold T4 determined by the ceramic laminate complexity analysis results, in the range of [100,300] milliseconds, the service degradation mechanism is activated to generate a highly available ceramic laminate production platform.

[0011] Preferably, the data encryption protocol and real-time instruction transmission function through remote control specifically include: Obtaining the ceramic laminate production resource configuration status through AES-256-CBC encryption, and parsing the configuration status to obtain a production task allocation table; Using RabbitMQ message queue to distribute the task instructions of the production task allocation table to the hot pressing forming module and the cutting device module, and obtain the execution status of the modules; The real-time production request sequence of the module execution state is processed through the Kafka message queue, the request processing time of each module is counted, if the processing time exceeds the preset delay threshold T4, the module is marked as a high delay module, and a high delay module list is obtained; The task proportion of the high delay module is adjusted by using a weighted round robin algorithm, and an optimized task allocation table is generated.

[0012] Preferably, the high-availability ceramic stack production platform collects running state data and multi-layer ceramic stack production indicators through real-time monitoring modules and state feedback mechanisms, uses a time series analysis algorithm to perform trend prediction and anomaly detection on central processor usage rate, memory occupancy rate, and environmental temperature and humidity control data, generates an automated operation and maintenance instruction set, including parameter adjustment and resource reallocation, and if any performance indicator deviates from the normal range by more than a preset tolerance T5=10%, triggers an automatic optimization program to adjust the ceramic stack production parameters, and generates stable digital service capabilities for ceramic capacitor stacks through remote fault diagnosis and log auditing functions.

[0013] Preferably, the running state data of the high-availability ceramic stack production platform collected through the real-time monitoring modules and the state feedback mechanisms specifically includes: The real-time monitoring module collects central processor usage rate, memory occupancy rate, and environmental temperature and humidity data from the ceramic stack production platform, writes them into the InfluxDB database at every minute interval, and generates a structured state data set; An ARIMA algorithm is used to perform 7-day rolling training on the structured state data set, to obtain a trend prediction model and output a prediction data set; If any indicator in the prediction data set exceeds a dynamic threshold value calculated based on the 3σ principle, a PID controller is triggered to calculate ceramic stack production parameter adjustment values, and a parameter adjustment instruction set containing temperature compensation values and pressure correction values is generated; The parameter adjustment instruction set is distributed to the temperature control execution module and the pressure regulation module through the MQTT protocol, and an updated running state data set is obtained.

[0014] Preferably, the trend prediction of the central processor usage rate using the time series analysis algorithm specifically includes: The real-time monitoring module collects central processor usage rate data and writes it into the InfluxDB time series database, extracts the last 24 hours of data from the database, and uses the Prophet algorithm to predict the usage rate trend for the next hour; If the predicted value exceeds the preset threshold of 85%, a rule engine based on a decision tree is triggered to calculate temperature adjustment values and pressure correction values for the ceramic stack production parameters, and writes them into the RabbitMQ message queue; The historical operating status data of the past 30 days is extracted from the InfluxDB and merged with the latest collected CPU usage data. The K-means clustering algorithm is used to match similar working conditions, and a digital service instruction set is generated. The set is transmitted to the control terminal via the ModbusTCP protocol to update the production parameters.

[0015] A remote control system for a ceramic capacitor stacking machine, comprising: A distributed data acquisition module is used to acquire micron-level alignment accuracy data and crack and void detection data from the ceramic capacitor stacking machine's automated feeding system and defect detection module during the multilayer ceramic stacking process. A data cleaning algorithm is used to remove redundant records and outliers. The dataset is verified based on the evaluation criteria of completeness (missing rate less than 5%), consistency (format uniformity greater than 95%), and accuracy (error rate less than 2%). If the overall data quality score falls below the preset threshold T1 = 0.85, the data repair mechanism is triggered to generate a standardized ceramic stacking dataset. An in-memory computing module is used to perform sharded storage and parallel processing of a standardized ceramic stack dataset. The module calculates the ceramic stack complexity weight coefficient W = Σ(ai × xi) based on the feature vectors of the multilayer ceramic stack, where ai is the feature importance coefficient calculated based on information gain, ranging from [0, 1], and xi is the normalized micron-level alignment accuracy and crack and void detection feature value, ranging from [0, 1]. If W exceeds a preset threshold T2 = 0.7, a high-performance computing resource pool including a graphics processing unit cluster and distributed storage is activated to generate ceramic stack complexity analysis results. The load balancing module is used to dynamically adjust the task allocation ratio of computing nodes through an adaptive load balancing algorithm based on the results of the ceramic stacking complexity analysis. It also uses a support vector machine algorithm to predict the load trend of the ceramic stacking production system based on environmental temperature and humidity control data and the operating status of the high-speed cutting device, and generates a load prediction value R in tasks per second; The expansion control module is used to calculate the expansion factor S = R × (1 + U) / 1000 based on R and the current system utilization U (range [0, 1]), where 1000 is the preset benchmark load value, in tasks per second. If S exceeds the preset threshold T3 = 1.2, the elastic expansion mechanism is triggered to generate an optimized ceramic laminate production resource configuration state. The technical effects and advantages of the remote control method for a ceramic capacitor lamination machine of the present invention are as follows: Through distributed data collection interfaces and data cleaning algorithms, we ensure data integrity, consistency and accuracy, set clear evaluation standards (missing rate <5%, format uniformity rate >95%, error rate <2%), ensure data quality, and have a data repair mechanism that can handle abnormal data in a timely manner.

[0016] It uses an in-memory computing engine for sharded storage and parallel processing, and utilizes a high-performance computing resource pool, including GPU clusters and distributed storage, to achieve efficient management of 64 data shards through the HDFS distributed file system.

[0017] Dynamically adjust task allocation based on an adaptive algorithm, predict load trends using a support vector machine algorithm, and design a reasonable expansion mechanism to ensure optimal allocation of system resources. AES-256-CBC encryption is used to protect data transmission, and a complete permission verification and user management mechanism is implemented, with remote fault diagnosis and log auditing capabilities.

[0018] It supports real-time monitoring of multi-dimensional data, uses multiple algorithms (ARIMA, Prophet, K-means) for trend forecasting, has anomaly detection and automatic optimization capabilities, supports multi-device collaborative management, has an elastic expansion mechanism, realizes automatic optimization of ceramic laminate production parameters, provides complete digital service capabilities, and supports real-time monitoring and adjustment of the production process.

[0019] Reduce resource waste through intelligent resource allocation, reduce the risk of equipment failure through predictive maintenance, have a high degree of automation, reduce manual intervention, accurately control micron-level positioning accuracy, effectively monitor and control cracks and gaps, achieve precise adjustment of production parameters, provide automated operation and maintenance instruction sets, support remote control and fault diagnosis, and have a complete status feedback mechanism. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 : Schematic diagram of the process of the present invention. DETAILED DESCRIPTION

[0021] 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.

[0022] Example 1 See also Figure 1 As shown, the remote control method of a ceramic capacitor lamination machine described in this embodiment includes: S101,1 Obtain micron-level alignment accuracy data and crack gap detection data in the multi-layer ceramic stacking process from the automated feeding system and defect detection module of the ceramic capacitor laminating machine through the distributed data acquisition interface, remove redundant records and outliers using a data cleaning algorithm, verify the data set according to the evaluation standards of less than 5% missing rate for integrity, more than 95% format uniformity for consistency, and less than 2% error rate for accuracy, if the comprehensive score of data quality is less than the preset threshold T1=0.85, trigger the data repair mechanism, and generate a standardized ceramic lamination data set; S102,2 Store and process the standardized ceramic lamination data set in parallel through the memory computing engine, calculate the ceramic lamination complexity weight coefficient W=∑(ai×xi) based on the feature vector of multi-layer ceramic stacking, where ai is the feature importance coefficient calculated based on information gain, range [0,1], xi is the normalized micron-level alignment accuracy and crack gap detection feature value, range [0,1], if W exceeds the preset threshold T2=0.7, start the high-performance computing resource pool containing a graphics processing unit cluster and distributed storage, and generate ceramic lamination complexity analysis results; S103,3 According to the ceramic lamination complexity analysis results, dynamically adjust the task allocation ratio of the computing nodes through the adaptive load balancing algorithm, predict the ceramic lamination production system load trend based on the environmental temperature and humidity control data and high-speed cutting device running state using the support vector machine algorithm, generate the load prediction value R, unit: task per second, according to R and the current system utilization rate U, range [0,1], calculate the expansion factor S=R×(1+U) / 1000, where 1000 is the preset baseline load value, unit: task per second, if S exceeds the preset threshold T3=1.2, trigger the elastic expansion mechanism, and generate the optimized ceramic lamination production resource configuration state; S104,4 According to the optimized ceramic lamination production resource configuration state and ceramic lamination complexity analysis results, decouple the function modules of hot pressing forming process and high-speed cutting device through micro-service architecture, use application programming interface gateway to uniformly manage data interaction and user permission verification of multi-device collaborative management, process asynchronous production requests through message queue, if the service response time exceeds the delay threshold T4 determined by the ceramic lamination complexity analysis results, range [100,300] milliseconds, start the service degradation mechanism, and generate a highly available ceramic lamination production platform; S105,5 collects the operating status data of the high-availability ceramic stack production platform and the production indicators of multilayer ceramic stacks through the real-time monitoring module and status feedback mechanism. A time series analysis algorithm is used to perform trend prediction and anomaly detection on the CPU usage rate, memory occupancy rate, and ambient temperature and humidity control data. This generates an automated operation and maintenance instruction set, including parameter adjustment and resource reallocation. If any performance indicator deviates from the normal range by more than the preset tolerance T5 = 10%, an automatic optimization program is triggered to adjust the ceramic stack production parameters. Remote fault diagnosis and log auditing functions are used to generate stable digital service capabilities for ceramic capacitor stacks. Step S101, 1. Obtain micron-level alignment accuracy data and crack and gap detection data during the multilayer ceramic stacking process from the automated feeding system and defect detection module of the ceramic capacitor stacking machine through a distributed data acquisition interface. Use a data cleaning algorithm to remove redundant records and outliers. Verify the data set based on the evaluation criteria of completeness with a missing rate below 5%, consistency with a format uniformity rate above 95%, and accuracy with an error rate below 2%. If the comprehensive data quality score is lower than the preset threshold T1=0.85, trigger the data repair mechanism to generate a standardized ceramic stacking data set.

[0023] The alignment accuracy data is obtained from the feeding system of the ceramic capacitor stacking machine through the OPCUA protocol, and the crack and gap data is obtained from the defect detection module, and the data are merged to generate the original data set. The drop_duplicates function of Pandas is used to remove duplicate records of the original data set, and the outliers outside the upper and lower 1% quantiles are removed through the quantile function to generate a cleaned data set. If the missing rate of the cleaned data set is higher than the preset threshold of 15%, linear interpolation is used to fill the missing values ​​to generate a completed data set. According to the pre-established format rules, the field type and unit consistency of the completed data set are verified to determine the format uniformity rate.

[0024] Specifically, the OPCUA protocol is used to obtain the alignment accuracy data from the feeding system of the ceramic capacitor stacking machine, and the crack and gap data is obtained from the defect detection module, and the data are merged to generate the original data set. The drop_duplicates function of Pandas is used to remove duplicate records of the original data set, and the outliers outside the upper and lower 1% quantiles are removed by the quantile function to generate a cleaned data set. If the missing rate of the cleaned data set is higher than the preset threshold of 15%, linear interpolation is used to fill the missing values ​​to generate a completed data set. According to the pre-established format rules, the field type and unit uniformity of the completed data set are verified to determine the format uniformity rate.

[0025] We used the OPCUA protocol to obtain alignment accuracy data from the feeder system of a ceramic capacitor lamination machine, and crack and void data from the defect detection module. These data were then merged into the original dataset. We used the drop_duplicates function in Pandas to remove duplicate records, and the quantile function to remove outliers outside the upper and lower 1% quantiles to generate a cleaned dataset. We calculated the missing rate of the cleaned dataset, which is the number of missing records divided by the total number of records multiplied by 100.

[0026] If the missing value rate exceeds the preset threshold of 15%, linear interpolation is used to fill in the missing values ​​and generate a completed dataset. The field type and unit consistency of the completed dataset are verified according to pre-established formatting rules. The format consistency rate is calculated as the number of records that conform to the format divided by the total number of records multiplied by 100. If the format consistency rate is below the preset threshold of 98%, Z-score standardization is used to adjust the numeric fields and generate a standardized dataset.

[0027] The error rate of the normalized dataset is calculated as the number of outliers divided by the total number of samples multiplied by 100. If the error rate exceeds the preset threshold of 2%, the outlier values ​​are recalculated based on the Euclidean distance using the K-nearest neighbor algorithm with k=5 to generate a repaired dataset. The overall quality score of the repaired dataset is calculated as 0.3 times the completeness rate plus 0.7 times the accuracy rate.

[0028] If the comprehensive score is lower than the preset threshold of 95%, the K-nearest neighbor repair process is repeated until the final ceramic stacking data set is output. When acquiring data from the automated feeding system and defect detection module of the ceramic capacitor stacking machine, the MQTT protocol is used through the distributed data acquisition interface to obtain micron-level alignment accuracy data from the servo motor encoder of the feeding system at a sampling frequency of 100Hz, with an accuracy of ±1μm. At the same time, crack and gap detection data is obtained from the CCD camera of the defect detection module with a resolution of 0.5μm / pixel. The data is transmitted in JSON format through the Kafka message queue, processing 1,000 records per second. The data cleaning algorithm uses an outlier detection method based on the Z score, setting the threshold Z=3 to eliminate records with alignment accuracy deviations exceeding 3μm or crack width anomalies exceeding 5μm. Redundant records are eliminated through timestamp synchronization checks, and data with a time difference of less than 10ms is retained. Data integrity assessment is performed by calculating the missing rate. 1 million records are scanned and a missing rate below 5% (i.e., fewer than 50,000 missing records) is considered qualified. The consistency assessment uses regular expressions to verify the JSON format uniformity, which must be higher than 95% (i.e., fewer than 5,000 records with format errors). The accuracy assessment is performed by comparing against a standard template, with an error rate of less than 2% (i.e., fewer than 20,000 records with errors). The comprehensive score is calculated as S = 0.4 × completeness + 0.3 × consistency + 0.3 × accuracy. If S is less than 0.85, the data repair mechanism is triggered. The KNN algorithm (K = 5) is used to fill missing values ​​based on nearest neighbor interpolation and perform mean replacement on outliers. A standardized ceramic stack dataset containing alignment accuracy, crack width, and timestamps is generated and stored in Parquet format. Each batch of data is approximately 1GB to support subsequent analysis. The alignment accuracy data is used to optimize the servo control of the feeding system, and the crack data is used to adjust the pressing parameters to ensure the stability of the ceramic stack quality.

[0029] The micron-level alignment accuracy and crack and gap detection data of the ceramic capacitor stacking machine are obtained through a distributed data acquisition interface. The integrity, consistency and accuracy of the dataset are verified according to the missing rate, format unification rate and error rate evaluation standards. If the comprehensive score is lower than the preset threshold, the data repair mechanism is used to obtain a standardized ceramic stacking dataset.

[0030] Micron-level alignment accuracy data is obtained from the automated feeding system of the ceramic capacitor stacking machine, and crack and gap data is obtained from the defect detection module. These data are combined to generate an initial dataset. For the initial dataset, the mean filling method is used to complete missing records to generate a completed dataset. For the completed dataset, a regular expression is used to verify that the alignment accuracy field is a floating point type with a unit of micron, and the crack and gap field is an integer with a unit of square micron, to generate a formatted dataset. For the formatted dataset, a random forest classifier is used to repair outliers to generate a repaired dataset. Based on the repaired dataset, the completeness rate, format consistency rate, and accuracy rate are calculated. If the combined score is lower than the preset threshold, the outliers are iteratively repaired to generate an optimized dataset.

[0031] Specifically, a distributed data acquisition interface is used to acquire micron-level alignment accuracy data from the automated feeding system of the ceramic capacitor lamination machine, and crack and void data is obtained from the defect detection module. These data are then combined to generate an initial dataset. The missing rate of the initial dataset is calculated, which is equal to the number of missing records divided by the total number of records. If the missing rate exceeds a preset threshold, a mean-filling method is used to fill in missing records based on the mean of the field's non-missing values ​​to generate a completed dataset.

[0032] For the completed dataset, regular expression is used to check the format of each field. The format consistency is equal to the number of records that meet the predefined format rules divided by the total number of records. The predefined format rules include that the bit accuracy field is a floating-point type with units of microns, and the crack gap field is an integer type with units of square microns. If the format consistency is lower than a preset threshold, the field type is converted, and a formatted dataset is generated. For the formatted dataset, the error rate is calculated, which is equal to the number of abnormal value records divided by the total number of records. The abnormal value determination criterion is that the bit accuracy exceeds ±3 times the standard deviation range or the crack gap is negative.

[0033] If the error rate is higher than a preset threshold, a random forest classifier is used to repair abnormal values based on feature importance ranking, and a repaired dataset is generated. Based on the repaired dataset, the integrity rate, the format consistency rate, and the accuracy rate are calculated. The integrity rate is equal to the proportion of records without missing values. The format consistency rate is defined as before. The accuracy rate is equal to the proportion of records without abnormalities. The comprehensive score is equal to the integrity rate multiplied by weight a, plus the format consistency rate multiplied by weight b, plus the accuracy rate multiplied by weight c. If the comprehensive score is lower than a preset threshold, the abnormal value repair step is iteratively performed, and an optimized dataset is generated. For the optimized dataset, it is verified whether each field meets the business specifications, such as the bit accuracy error not exceeding a specified range and the crack gap area not exceeding a specified value, and a standardized ceramic laminate dataset is generated.

[0034] When acquiring data from the automated feeding system and defect detection module of the ceramic capacitor laminate machine, the distributed data acquisition interface can use the RESTful API method to acquire micron-level bit accuracy data from the servo motor of the feeding system, with an accuracy control within ±1.5 microns. At the same time, crack gap data is acquired from the high-resolution CCD camera of the defect detection module, with an accuracy of 0.8 square microns. The initial dataset is stored in JSON format, containing bit accuracy, crack gap area, and timestamp fields. The missing rate calculation is based on scanning 500,000 records. If the missing records exceed 25,000, i.e., the missing rate is higher than 5%, the filling mechanism is triggered. The Pandas fillna function is used to fill in the missing records based on the non-missing value mean of the bit accuracy field, such as 2.3 microns, to ensure data integrity. This method can effectively reduce the impact of data missing on subsequent analysis.

[0035] Specifically, the format consistency check verifies the field format through regular expressions.

[0036] For example, the alignment accuracy field must be a floating-point type with units in microns, such as 1.234; the crack gap field must be an integer type with units in square microns, such as 45. If format consistency falls below 95%, meaning there are more than 5,000 records with incorrect formats, use the Pandas astype function to convert non-compliant fields to a standard type, such as converting the string "1.234um" to the floating-point type 1.234. This format unification helps ensure a standardized dataset and facilitates subsequent processing.

[0037] In one embodiment, the error rate calculation is performed by counting the number of outlier records.

[0038] For example, if alignment accuracy exceeds ±3 times the standard deviation (e.g., if the mean is 2 microns and the standard deviation is 0.5 microns), the outlier range is between 0.5 and 3.5 microns. Negative cracks and voids, such as -10 square microns, are also considered outliers. If the error rate exceeds 3%, meaning there are more than 15,000 outlier records, a RandomForestClassifier is used to prioritize alignment accuracy correction based on feature importance. This approach effectively improves data accuracy by identifying key features through machine learning models.

[0039] It will be appreciated that the composite score calculation combines the completeness rate, format consistency rate, and accuracy rate.

[0040] For example, if the completeness rate is 98%, the format consistency rate is 96%, and the accuracy rate is 95%, the combined score is 0.3 × 0.98 + 0.2 × 0.96 + 0.5 × 0.95 = 0.961. If the score falls below 90%, outlier repair is repeated for up to five iterations to ensure dataset quality. Dataset optimization requires verification of business specifications, such as alignment accuracy error of no more than ±2 microns and crack void area of ​​no more than 50 square microns. This verification ensures that the data meets the requirements of the ceramic lamination process.

[0041] For example, when generating standardized ceramic stackup datasets, the data is stored in CSV format, approximately 500 MB per batch. The alignment accuracy data can be used to optimize the servo motor control logic of the feeding system, and the crack gap data can guide the adjustment of press process parameters. The high quality of this dataset supports the stability of subsequent process optimization.

[0042] Step S102, 2 The standardized ceramic stacking data set is sharded and processed in parallel through the memory computing engine, and the ceramic stacking complexity weight coefficient W=Σ(ai×xi) is calculated based on the feature vector of the multi-layer ceramic stack, where ai is the feature importance coefficient calculated based on information gain, ranging from [0,1], and xi is the normalized micron-level alignment accuracy and crack and gap detection feature value, ranging from [0,1]. If W exceeds the preset threshold T2=0.7, a high-performance computing resource pool including a graphics processing unit cluster and distributed storage is started to generate the ceramic stacking complexity analysis results.

[0043] The Apache Spark in-memory computing framework was used to process a standardized ceramic laminate dataset. The Spark output data was partitioned using consistent hashing using the libketama library and stored in 64 data shards of the HDFS distributed file system.

[0044] Based on the sharded distribution structure, 16 compute nodes were launched using the MPI parallel framework. Each node invoked OpenCV's SIFT algorithm to extract micron-level alignment coordinate features and a pre-trained ResNet50 model to extract crack and void probability features, generating a 256-dimensional feature vector set. This feature vector set was then input into the scikit-learn mutual_info_classif function to calculate the information gain of each feature and generate a set of feature importance coefficients. Linear normalization was performed on the feature vector set using the MinMaxScaler function, compressing all eigenvalues ​​to the range of 0 to 1 to generate a set of normalized eigenvalues. The ceramic stack complexity weight W was calculated using the formula W = the sum of the feature importance coefficients ai and the normalized eigenvalues ​​xi. If the complexity weight W exceeded the preset threshold T2, a GPU cluster consisting of eight NVIDIA Tesla V100s and HDFS storage resources were allocated. The GPU cluster then invoked a PyTorch convolutional neural network to process the feature data, outputting a three-dimensional complexity distribution map of the ceramic stack, which was stored in a designated HDFS shard.

[0045] Specifically, processing a ceramic laminate dataset involves multiple technical topics, including distributed computing, feature extraction, feature selection, normalization, and complexity analysis. The following analysis and examples address each technical topic, focusing on the business area of ​​ceramic laminate quality analysis. This analysis closely integrates the data collection and cleaning background of historical information to ensure logical coherence and mutual support.

[0046] For example, the Apache Spark in-memory computing framework can efficiently process large amounts of data using its distributed computing capabilities when processing standardized ceramic stack datasets. Spark loads the data into memory and uses RDDs (Resilient Distributed Datasets) to parallelize the alignment accuracy and crack gap data.

[0047] In one possible implementation, assume a dataset contains 10 million records, each of which includes micron-level alignment coordinates and crack widths. Spark uses the DataFrame API to group and aggregate the data, calculating the mean alignment deviation and crack width distribution for each batch. This approach can quickly process datasets up to 1GB, significantly improving computational efficiency. For data sharding and storage, the libketama library implements consistent hashing, evenly distributing data across 64 shards in HDFS.

[0048] For example, 10 million records are hashed and distributed based on bitwise precision, with each shard storing approximately 156,000 records. HDFS ensures data reliability through multiple replication mechanisms, making it suitable for the high availability requirements of ceramic laminate data. Sharded data can be quickly retrieved by timestamp or batch number, facilitating subsequent feature extraction.

[0049] Specifically, the MPI parallel framework activates 16 computing nodes to extract features. OpenCV's SIFT algorithm is used to analyze the alignment coordinate image and generate a 128-dimensional keypoint descriptor to capture micron-level alignment deviation features.

[0050] For example, the servo motor in the feeding system provides coordinate images with an accuracy of ±1μm. The SIFT algorithm identifies key points and generates feature descriptors for analyzing alignment stability. The pre-trained ResNet50 model processes crack and void images and outputs a crack probability distribution.

[0051] For example, a CCD camera provides images with a resolution of 0.5μm / pixel. ResNet50 extracts crack edge features, generating a 256-dimensional feature vector. This feature extraction method accurately characterizes the quality of ceramic laminates. For feature selection, scikit-learn's mutual_info_classif function calculates information gain to assess feature importance.

[0052] For example, in a 256-dimensional feature vector set, the information gain for the position deviation feature is 0.8, and the information gain for the crack width feature is 0.6. High-gain features are prioritized. This approach ensures that subsequent analysis focuses on key features and reduces redundant computations.

[0053] For example, MinMaxScaler normalization compresses feature values ​​to the range of 0 to 1. For example, if the alignment deviation ranges from -3μm to 3μm, normalization maps it to 0 to 1, unifying feature data of different dimensions. This process improves the consistency of model inputs and optimizes subsequent complexity analysis.

[0054] In one possible implementation, the complexity weight coefficient W is calculated by weighted summation of the feature importance coefficient and the normalized feature value.

[0055] For example, if the alignment deviation feature weight is 0.5 and the crack width is 0.3, the normalized values ​​are 0.8 and 0.6, respectively, resulting in a W value of 0.58. If W exceeds the threshold T2, for example, 0.7, GPU cluster resources are allocated. A GPU cluster consisting of eight NVIDIA Tesla V100s runs a convolutional neural network based on PyTorch, processing the feature data to generate a three-dimensional complexity distribution map.

[0056] For example, the distribution map shows areas with concentrated alignment deviations and high crack incidence, which are stored in a designated HDFS shard to facilitate subsequent optimization of ceramic stacking process parameters.

[0057] It is understandable that this multi-level processing method, from data sharding to feature extraction, selection, normalization, and then to complexity analysis, forms a complete technical chain, ensuring efficient processing and high-quality analysis of ceramic laminate data sets, and supporting process optimization needs.

[0058] A standardized ceramic stacking dataset is obtained through an in-memory computing engine, and the feature vectors of the multilayer ceramic stack are calculated using sharded storage and parallel processing technology. The complexity weight coefficient is compared with the preset threshold to determine whether to start the high-performance computing resource pool.

[0059] Spark batch reads distributed storage shards, and principal component analysis is used to generate a feature vector set of a multilayer ceramic stack to obtain feature vector data. Based on the feature vector data, scikit-learn is used to calculate the information gain of each feature, generate a feature importance coefficient set, and obtain feature weight data. Through the feature weight data and the feature vector data normalized by MinMax, the weighted summation formula W=Σ(ai×xi) is used, where ai is the feature importance coefficient and xi is the normalized eigenvalue, to calculate the complexity weight coefficient and obtain the weight coefficient value. If the weight coefficient value exceeds the preset threshold, the YARN resource manager is scheduled to allocate a GPU computing node to generate complexity distribution data.

[0060] Specifically, a standardized ceramic stack dataset was acquired from a multilayer ceramic stack using the Spark in-memory computing engine. HDFS was used to shard the dataset and generate three replicas, resulting in a sharded dataset. HDFS sharded storage was used to manage the sharded dataset, and consistent hashing was used to distribute the data across multiple DataNodes, resulting in distributed storage shards. Spark was used to batch read the distributed storage shards, and principal component analysis was used to generate a set of feature vectors for the multilayer ceramic stack, resulting in feature vector data. Based on the feature vector data, scikit-learn was used to calculate the information gain of each feature, generating a set of feature importance coefficients and obtaining feature weights.

[0061] The complexity weight coefficient is calculated using the weighted summation formula W = Σ(ai × xi), where ai represents the feature importance coefficient and xi represents the normalized feature value, using the feature weight data and MinMax normalized feature vector data. If the weight coefficient exceeds the preset threshold of 0.75, the YARN resource manager is dispatched to allocate a GPU compute node. The feature vector data is processed using ResNet18 to generate a ceramic stack complexity distribution and obtain the complexity distribution data. Based on the complexity distribution data, the results are saved to the specified column family using HBase by hashing the row key, resulting in the final storage result.

[0062] Processing a ceramic laminate dataset involves multiple technical topics, including distributed storage, feature extraction, feature selection, normalization, and complexity analysis. The following analysis and examples for each technical topic are presented, focusing on the business area of ​​ceramic laminate quality analysis and considering the data collection and cleaning background of historical information. This ensures a logically rigorous and mutually supportive discussion.

[0063] For example, the Apache Spark in-memory computing engine can leverage its distributed computing capabilities to efficiently load data when processing standardized ceramic stackup datasets. Spark accelerates data processing through in-memory computing, making it suitable for large-scale alignment accuracy and crack detection data in ceramic stackups.

[0064] In one possible implementation, the dataset contains 5 million records, each of which includes micron-level alignment coordinates and crack depth. Spark uses the DataFrame API to batch read and preprocess the data, for example, grouping by batch number and extracting alignment deviation statistics. This approach is suitable for rapidly processing large datasets. For data sharding and storage, HDFS distributes data across multiple DataNodes using a consistent hashing algorithm.

[0065] For example, 5 million records are hashed and distributed across 32 shards based on bitwise precision, with each shard storing approximately 156,000 records. HDFS generates three replicas to ensure high data availability, which is suitable for the high reliability requirements of ceramic laminate data. Sharded data can be quickly retrieved by timestamp, facilitating subsequent analysis.

[0066] It can be understood that feature extraction generates a feature vector set through principal component analysis.

[0067] For example, the PCA algorithm is used to process the alignment coordinates and crack depth data to generate a 64-dimensional feature vector to capture the main quality features.

[0068] In an embodiment, the feeding system provides coordinate data with a precision of ±2 μm, and the PCA analysis extracts the main components of the alignment deviation to form a low-dimensional feature representation. This way effectively reduces the data dimension and retains key information. In feature selection, the mutual_info_classif function of scikit-learn calculates the information gain to generate a set of feature importance coefficients.

[0069] For example, in a 64-dimensional feature vector, the information gain of the alignment deviation feature is 0.75, and the information gain of the crack depth feature is 0.55. High-gain features are preferentially retained. This method focuses on key features and improves analysis efficiency.

[0070] For example, MinMax normalization maps feature values to the interval [0, 1]. Assuming that the alignment deviation range is -5 μm to 5 μm, after normalization, it is mapped to [0, 1], and the feature data of different dimensions is unified. This processing ensures the consistency of model input. In complexity analysis, the weight coefficient W is calculated by weighted sum of feature importance coefficients and normalized feature values.

[0071] For example, the alignment deviation feature weight is 0.6, and the crack depth is 0.4. The normalized values are 0.7 and 0.5, respectively, and the W value is 0.62. If W exceeds the threshold value 0.75, the YARN resource manager allocates a GPU node. ResNet18 processes the feature vector data to generate a complexity distribution map, showing the alignment deviation and crack distribution area.

[0072] It can be understood that the complexity distribution data is stored in the specified column family by row key hash value in HBase.

[0073] For example, the distribution map is stored by batch number, the row key is the timestamp, and the column family saves the complexity distribution features, which facilitates fast query. This storage method supports subsequent process optimization analysis.

[0074] In a possible implementation, from the core scheme to the extended scheme, the data processing link from sharding storage to feature extraction, selection, normalization, and then to complexity analysis and storage forms a complete technical process, supporting the ceramic lamination quality analysis requirements.

[0075] Step S103, 3: Based on the results of the ceramic stacking complexity analysis, the task allocation ratio of the computing nodes is dynamically adjusted through an adaptive load balancing algorithm. The support vector machine algorithm is used to predict the load trend of the ceramic stacking production system based on the environmental temperature and humidity control data and the operating status of the high-speed cutting device, and a load prediction value R is generated in tasks per second. Based on R and the current system utilization U in the range of [0,1], the expansion factor S=R×(1+U) / 1000 is calculated, where 1000 is the preset benchmark load value in tasks per second. If S exceeds the preset threshold T3=1.2, the elastic expansion mechanism is triggered to generate an optimized ceramic stacking production resource configuration state.

[0076] A weighted round-robin algorithm calculates the allocation weight of each computing node using the formula 1 / (1 + the node's current load rate), resulting in a node task allocation table. Real-time temperature and humidity data is obtained from the ambient temperature and humidity sensor, and equipment fault codes are parsed from the cutting device log. If the temperature and humidity exceed the preset range, a compensation coefficient is generated. If the fault code appears three times consecutively, the node is marked as abnormal. The compensation coefficient, the fault flag, and the node's CPU usage are combined into an input vector. A support vector machine algorithm is used to train a load prediction model, outputting a load forecast value for the next five minutes.

[0077] Specifically, the task allocation ratio is derived from the results of the ceramic stacking complexity analysis. A weighted round-robin algorithm is used to calculate the allocation weight of each computing node, using the formula 1 / (1 + the node's current load rate). This yields a node task allocation table. Based on this node task allocation table, real-time temperature and humidity data is obtained from the ambient temperature and humidity sensor, and equipment fault codes are parsed from the cutting device log. A compensation coefficient is generated when the temperature and humidity exceed the preset range, and a node is marked as abnormal if the fault code appears three times consecutively. The compensation coefficient, fault flag, and node CPU usage are combined into an input vector. A load prediction model is trained using a support vector machine algorithm, outputting a load forecast for the next five minutes. The load forecast value is weighted and summed with the real-time CPU usage in a ratio of 0.6:0.4 to obtain the expansion factor. If the expansion factor exceeds 1.2, the number of newly added nodes is equal to ceil(current number of nodes × expansion factor / 2). Computing resources are allocated from the resource pool, and the IP address and memory capacity of the newly added node are recorded in the resource configuration status table.

[0078] Based on the resource allocation status table, the task scheduling strategy for the ceramic laminate production system was adjusted to generate an optimized task allocation plan. This new plan was then fed into a weighted round-robin algorithm to update node weights and generate a new node task allocation table. This new node task allocation table was used to monitor the production system's operating status, obtaining real-time CPU usage and fault code data. This data, combined with temperature and humidity compensation coefficients, was then fed into a linear regression algorithm to generate updated load forecasts.

[0079] For example, in a ceramic lamination production system, the allocation weight of each computing node is calculated by a weighted round-robin algorithm. The task allocation can be dynamically adjusted based on the current load rate of the node. The weight formula 1 / (1+current load rate of the node) ensures that the node with higher load allocates less tasks. Assuming that the system has 10 computing nodes, the CPU usage rate of a certain node is 80%, and its weight is 1 / (1+0.8)=0.56, while the weight of the node with a CPU usage rate of 20% is 0.83. The node with higher weight is preferentially allocated tasks, and a node task allocation table is generated to record the IP address and task amount of each node. This way realizes the dynamic balance of tasks and improves the resource utilization efficiency.

[0080] In one possible implementation, when obtaining real-time data from the environment temperature and humidity sensor, a DHT22 sensor can be used to monitor the temperature and humidity of the production workshop. Assuming that the preset temperature range is 20-25°C and the humidity is 40-60%, if the real-time data is 27°C and 65%, a compensation coefficient is generated.

[0081] For example, if the temperature exceeds 2°C, the compensation coefficient is 1.1; if the humidity exceeds 5%, the coefficient is 1.05. The compensation coefficient is used to adjust the input of the load prediction model to ensure that the influence of environmental factors on the demand for computing resources is fully considered.

[0082] Specifically, when analyzing the equipment fault code from the cutting device log, the fault information can be extracted through a log analysis tool. Assuming that the log shows that a certain cutting device has appeared “E101” fault code for 3 times in a row, indicating that the tool is worn out, the node is marked as abnormal. The priority of the task of the abnormal node is reduced, and the task is re-allocated to other nodes. This mechanism quickly identifies equipment problems and ensures production continuity.

[0083] For example, when combining the compensation coefficient, the fault mark, and the CPU usage rate as the input vector, a three-dimensional vector can be constructed.

[0084] For example, if the compensation coefficient of a certain node is 1.1, the fault mark is 0 (no fault), and the CPU usage rate is 70%, the input vector is [1.1, 0, 0.7]. The support vector machine algorithm is trained based on historical data to predict the load value in the next 5 minutes. The predicted value is combined with the real-time CPU usage rate (weight 0.6:0.4) to calculate the expansion factor. Assuming that the predicted load is 85% and the real-time usage rate is 75%, the expansion factor is 0.6×85%+0.4×75%=0.81. If the expansion factor exceeds 1.2, for example, 1.3, the number of new nodes is ceil(10×1.3 / 2)=7. The new nodes are allocated from the resource pool, and the IP address and memory capacity (such as 16GB) are recorded to the resource configuration state table.

[0085] In a possible implementation, when the task scheduling policy is adjusted according to the resource configuration status table, the production tasks may be reallocated through the task scheduler.

[0086] For example, after adding seven new nodes, high-load tasks are prioritized and assigned to the new nodes, generating an optimized task allocation plan. This plan is fed into a weighted round-robin algorithm, which updates node weights and generates a new task allocation table. This new table is used to monitor system operating status and, using a linear regression algorithm, predicts updated load values ​​based on real-time CPU usage and temperature and humidity compensation coefficients.

[0087] For example, the real-time CPU usage of a node is 90%, the compensation coefficient is 1.2, and the predicted load is 92%, which guides subsequent resource adjustments.

[0088] It can be understood that the above method forms a complete technical chain from weight calculation, environmental data collection, fault detection to load prediction and resource expansion, ensuring the efficient operation of the ceramic laminate production system and optimizing the dynamics and stability of task allocation.

[0089] Based on the results of ceramic stacking complexity analysis, an adaptive load balancing algorithm is used to obtain an optimized load distribution plan by dynamically adjusting the task allocation ratio of computing nodes. According to the ambient temperature and humidity control data and the operating status of the high-speed cutting device, the support vector machine algorithm is used to predict the load trend of the ceramic stacking production system and obtain the load prediction value. Based on the load prediction value and the current system utilization, the expansion factor is calculated to determine whether it exceeds the preset threshold. If it exceeds, the elastic expansion mechanism is triggered to determine the optimized ceramic stacking production resource configuration status.

[0090] Historical production data is obtained from the ceramic laminate production database. The number of layers, pattern type, and sintering temperature are extracted as features. A random forest regression model is used to output a ceramic laminate complexity score. Based on the complexity score and a pre-set task duration comparison table, the total computational requirements for each production task are calculated, and an initial allocation table is generated based on the proportion of task types. Based on this initial allocation table, a weighted round-robin algorithm is used to dynamically allocate compute node tasks, updating node weights every 5 minutes. During weight updates, CPU and memory usage are obtained from the compute node monitoring service. If the current node load exceeds 80%, its weight is reduced by 10%. Real-time data is obtained from temperature and humidity sensors at a sampling rate of 1 per second and synchronized with the PLC status register data of the cutting device. The mean temperature and humidity values, cutting motor current, and tool vibration amplitude are combined into a 16-dimensional feature vector. This vector is input into a pre-trained scikit-learn SVC model to output a predicted system load value (R) (tasks / second) for the next 15 minutes.

[0091] The current utilization of the compute node cluster, U (actual percentage), is read and combined with the preset baseline load value, B (based on historical peak loads), to calculate the expansion factor, S = R × (1 + U / 100) / B. When S exceeds the threshold, T = 1.2, the Kubernetes cluster management API is called to add two compute nodes and generate a new resource allocation table. Based on the updated resource allocation table, the weighted round-robin algorithm is rerun to generate a node task allocation plan, and the updated task instructions are distributed to each compute node execution unit.

[0092] Specifically, the ceramic laminate production database stores a large amount of historical production data, including key features such as the number of layers, pattern type, and sintering temperature, which can be used to analyze production complexity.

[0093] For example, a batch of ceramic laminate data is extracted from a database. The number of layers is 32, the pattern type is complex geometry, and the sintering temperature is 1200°C. A random forest regression model is used to process these features. The model is trained based on historical data and outputs a complexity score of 8.5 (out of 10). Comparing with the preset task duration table, a score of 8.5 corresponds to a task duration of approximately 2 hours, and the total required is calculated as 5,000 task units. For different task types, such as printing, cutting, and sintering, an initial allocation ratio table is generated, assuming that printing accounts for 40%, cutting accounts for 35%, and sintering accounts for 25%. This allocation method ensures that tasks are allocated according to demand and balances the resource requirements of each production link.

[0094] In a possible implementation, a weighted round-robin algorithm dynamically allocates tasks according to an initial allocation ratio table.

[0095] For example, a system has eight compute nodes, and the initial allocation ratio table shows that pruning tasks account for 35%. Every five minutes, the CPU and memory usage of each node are obtained from the monitoring service. For example, if a node has a CPU usage of 85% and a memory usage of 70%, its weight is reduced by 10%, from 1.0 to 0.9. Nodes with lower weights are given lower priority, and more tasks are allocated to other less-loaded nodes. This dynamic adjustment optimizes resource allocation efficiency.

[0096] Specifically, temperature and humidity sensor data collection is crucial for load forecasting.

[0097] For example, using an AM2302 sensor to collect data at a frequency of once per second, at a certain moment, the workshop temperature is 26°C and the humidity is 62%, exceeding the preset range (20-25°C, 40-60%). Synchronizing this with the PLC status register data for the cutting device, the motor current is 5A and the tool vibration amplitude is 0.2mm. These are combined into a 16-dimensional feature vector. This input is fed into the trained SVC model, which predicts a load value R of 600 tasks / second for the next 15 minutes. This high-frequency data collection and multi-dimensional feature combination improves prediction accuracy.

[0098] For example, when calculating the scaling factor, assume that the current system utilization U is 75%, the preset reference load value B is 1000 tasks per second, and the predicted load R is 600 tasks per second. The scaling factor S = 600 x (1 + 0.75) / 1000 = 1.05. If S does not exceed the threshold T = 1.2, do not scale up. When S reaches 1.3, call the Kubernetes API to add 2 nodes, generate a resource configuration table, and record the IP and 16 GB memory capacity of the new nodes. After the new nodes are added, the weighted round-robin algorithm is run again to generate a new task allocation scheme.

[0099] For example, high-load cutting tasks are preferentially allocated to new nodes to reduce the pressure on old nodes. This mechanism quickly responds to load changes.

[0100] In one possible implementation, the system monitors the execution state in real time after the task instructions are issued to the computing node execution unit.

[0101] For example, the new allocation scheme issues 40% of the cutting tasks to the new nodes, and the execution unit feedback indicates that the task completion rate reaches 95%, indicating that the allocation scheme is reasonable. The monitoring data, combined with the average temperature and humidity and the motor state, are input into a linear regression model to predict the next period load value, guiding subsequent resource adjustment. This closed-loop mechanism ensures the continuity of production tasks and the dynamic balance of resource utilization.

[0102] Step S104, 4 Through the data encryption protocol and real-time instruction transmission function of remote control, according to the optimized ceramic laminate production resource configuration state and ceramic laminate complexity analysis result, decouple the function modules of hot pressing forming process and high-speed cutting device through micro-service architecture, use application programming interface gateway to uniformly manage data interaction and user permission verification of multi-device collaborative management, use message queue to process asynchronous production requests, if the service response time exceeds the delay threshold T4 determined by the ceramic laminate complexity analysis result, the range is [100, 300] milliseconds, then start the service degradation mechanism, and generate a highly available ceramic laminate production platform.

[0103] The ceramic laminate production resource configuration state is obtained through AES-256-CBC encryption, and the production task allocation table is obtained by analyzing the configuration state. The task instructions of the production task allocation table are distributed to the hot pressing forming module and the cutting device module using the RabbitMQ message queue, and the execution state of the module is obtained. The real-time production request sequence of the module execution state is processed through the Kafka message queue, the request processing time of each module is counted, and if the processing time exceeds the preset delay threshold T4, it is marked as a high-delay module, and a high-delay module list is obtained. The task proportion of the high-delay module is adjusted using the weighted round-robin algorithm to generate an optimized task allocation table.

[0104] Specifically, the data encryption protocol obtains the AES-256-CBC-encrypted ceramic laminate production resource configuration status, which is then transmitted to the application programming interface gateway via a TLS1.3 secure channel. The configuration status is then parsed to obtain a production task allocation table. A microservices architecture is used to decouple the hot pressing and cutting modules. Task instructions for each module are extracted from the production task allocation table and distributed to the corresponding modules using a RabbitMQ message queue to obtain the module execution status. The module execution status triggers the interface gateway to manage and collect data interaction logs. Combined with the permission verification results based on a MAC address whitelist, legitimate device records are filtered to obtain a real-time production request sequence. The real-time production request sequence is processed via a Kafka message queue, and the request processing time of each module is calculated. If the processing time exceeds the preset latency threshold T4, the module is marked as a high-latency module, and a list of high-latency modules is obtained. A weighted round-robin algorithm is used to adjust the task ratio of high-latency modules, generating an optimized task allocation table. Module task instructions are updated from the optimized task allocation table and redistributed via RabbitMQ to obtain the updated module execution status. The updated module execution status is managed and monitored through the interface gateway, and the production platform operation log is generated using AES-256-CBC encryption.

[0105] For example, in ceramic laminate production systems, the AES-256-CBC encryption protocol is used to ensure the security of resource configuration status. AES-256-CBC is a symmetric encryption algorithm that uses a 256-bit key and blocks configuration status data through block encryption. Each block of data is combined with an initialization vector to enhance encryption strength.

[0106] For example, a production task assignment table contains node IP addresses, task quantities, and priorities. This table is encrypted into a ciphertext file, approximately 2MB in size, and transmitted to the application programming interface gateway via a secure TLS 1.3 channel. TLS 1.3 utilizes elliptic curve cryptography to prevent data interception during transmission. Upon receiving the data, the gateway decrypts and parses it, extracting the task instructions for the hot press molding module from the task assignment table, such as "Node 192.168.1.10 executes the 1000 pieces hot press task."

[0107] In one possible implementation, the microservices architecture decouples the hot press forming module and the cutting device module, deploying them in separate containers using Docker containerization technology. The hot press forming module handles temperature control tasks, while the cutting device module manages cutting accuracy. Task allocation tables are distributed via a RabbitMQ message queue, with queue configurations configured for persistent storage to ensure task instructions are not lost.

[0108] For example, the hot pressing module receives the instruction "Set temperature to 230°C for 30 minutes," and the cutting module receives the instruction "Cutting speed to 500mm / s." The module execution status is fed back through the message queue, such as "Hot pressing completion rate 95%."

[0109] Specifically, the interface gateway manages and collects interaction logs and verifies device legitimacy using a MAC address whitelist. The whitelist stores the MAC addresses of legitimate devices, such as "00:1A:2B:3C:4D:5E." If a device with an unknown MAC address is detected, it is recorded as illegal and its task request is blocked. Production request sequences generated by legitimate devices are processed through a Kafka message queue, which stores request data from different modules in partitions.

[0110] For example, the hot pressing module partition records 1000 requests, and the cropping module partition records 800. The request processing time is counted. If the average processing time of the hot pressing module is 2 seconds, which exceeds the preset delay threshold T4 = 1.5 seconds, it is marked as a high-latency module.

[0111] For example, a weighted round-robin algorithm adjusts the proportion of tasks in high-latency modules, reducing the proportion of tasks allocated to the hot-pressing module to 30%, and allocating more tasks to the cutting module. The optimized task allocation table is regenerated to include updated task instructions, such as "The hot-pressing module's task volume is reduced to 800 pieces." Redistributed through RabbitMQ, the hot-pressing module adjusts its operating parameters after receiving the new instructions. The interface gateway monitors the updated execution status, such as "The hot-pressing module completion rate has increased to 98%." It uses AES-256-CBC encryption to generate operation logs. The logs record module runtime, task volume, and other information, and are stored in a distributed database for subsequent auditing.

[0112] In one possible implementation, the high-latency module list is updated regularly, and the resource allocation strategy is dynamically adjusted in combination with real-time production request sequence analysis.

[0113] For example, the processing time of the cutting module increased to 1.8 seconds due to aging equipment, triggering an adjustment of the task ratio to optimize the overall system response speed. Encrypted logs are transmitted to the management platform via TLS 1.3 to ensure data integrity.

[0114] The delay threshold T4 is determined based on the results of the ceramic stacking complexity analysis. If the service response time exceeds this threshold, the service degradation mechanism is activated. The functional modules of the hot pressing forming process and the high-speed cutting device are decoupled through the microservice architecture. A remote control data encryption protocol and real-time command transmission function are adopted to obtain a highly available ceramic stacking production platform.

[0115] The number of layers and sintering temperature parameters are extracted from the ceramic stacking complexity analysis results, input into the linear regression model, and the delay threshold T4 is obtained. According to the high-latency module list, the degradation weight is calculated inversely proportional to the delay multiple, and a JSON format task allocation table containing module name and weight fields is generated. The instructions of the hot pressing molding module and the high-speed cutting device module are extracted from the task allocation table through the gRPC protocol and distributed to the Kafka message queue of the corresponding module. Based on the deviation amplitude of the indicator of the abnormal module, its weight coefficient in the new allocation table is dynamically adjusted, and the updated task allocation table is distributed through the gRPC protocol to obtain the stable state judgment result.

[0116] Specifically, the number of layers and sintering temperature parameters are extracted from the ceramic stacking complexity analysis results and input into a linear regression model to calculate the delay threshold T4. The HTTP request response time of each module is collected through Prometheus, and modules exceeding T4 are marked as high-latency modules and stored in a Redis list. Based on the list of high-latency modules, the degradation weight is calculated inversely proportional to the delay multiple, and a JSON format task allocation table containing module name and weight fields is generated. The instructions for the hot pressing module and the high-speed cutting device module are extracted from the allocation table through the gRPC protocol and distributed to the Kafka message queue of the corresponding module.

[0117] The status data returned by the module is transmitted to the central control unit via TLS 1.3 encryption, and then encrypted using AES-256 to generate a log file. CPU usage and memory usage data are parsed from the logs and compared with preset thresholds to generate a list of abnormal modules. Based on the deviation of the abnormal module's metrics, its weight coefficient in the new allocation table is dynamically adjusted. The updated allocation table is distributed via gRPC. After collecting the encrypted status data, the weighted average of the error rate and latency rate over a five-minute period is calculated. If it is less than 0.1, the platform is considered stable.

[0118] Complexity analysis and resource optimization of ceramic laminate production systems involve multi-dimensional technical implementation. The following analysis and examples are combined with key topics, focusing on the field of ceramic laminate production to ensure rigorous logic and closely related content.

[0119] For example, when analyzing ceramic stack complexity, the number of layers and sintering temperature parameters are extracted. The number of layers reflects the complexity of the stack structure, while the sintering temperature determines process stability. Suppose the analysis results show that a batch of ceramic stacks has 10 layers and the sintering temperature is 1200°C. A linear regression model is used to calculate the delay threshold, T4. The model inputs the number of layers and temperature, and combines historical data. For example, a 5-layer stack corresponds to a T4 of 150 milliseconds, while a 15-layer stack corresponds to a T4 of 250 milliseconds. This infers that a 10-layer stack has a T4 of approximately 200 milliseconds. This approach ensures that the delay threshold is dynamically aligned with process complexity.

[0120] In one possible implementation, the Prometheus collection module responds to the HTTP request time, assuming that the hot press forming module response time is 250 milliseconds, which exceeds the T4 threshold of 200 milliseconds, and is marked as a high latency module, and is stored in the Redis list. Redis stores in the form of key-value pairs, such as the key "high_latency_modules" corresponding to the value "hot_press_module:250ms". Compared with traditional databases, the memory storage of Redis accelerates data access and is suitable for real-time marking of high latency modules.

[0121] Specifically, the degradation weight is inversely proportional to the delay multiple. The hot press module has a delay of 250 milliseconds, which is about 1.25 times the T4 threshold, and the weight is set to 0.8; the cutting module responds in 180 milliseconds, which does not exceed the threshold, and the weight is 1.0. A JSON format task allocation table is generated, containing fields such as "module_name:hot_press,weight:0.8". The task quantity of the module with low weight is reduced to ensure that system resources are tilted towards low latency modules. The gRPC protocol distributes task instructions efficiently and is suitable for scenarios with high real-time requirements.

[0122] For example, the hot press module extracts the instruction "temperature set to 220°C for 25 minutes" from the allocation table and transmits it to the Kafka queue through gRPC. Kafka partition storage, such as hot press module partition storage of 500 instructions and cutting module partition storage of 300 instructions, ensures that instruction distribution is ordered and traceable. State data is transmitted to the central control unit through TLS1.3 encryption, and AES-256 encryption generates log files.

[0123] For example, the log records the hot press module CPU usage of 80% and memory usage of 2GB. The preset threshold is CPU 85% and memory 2.5GB, and the hot press module CPU is close to the threshold, which is marked as an abnormal module. The weight coefficient of the abnormal module is adjusted to 0.7, and after updating the allocation table, the hot press module task quantity is reduced to 600 through gRPC distribution.

[0124] For example, the error rate and delay rate within 5 minutes are collected, assuming that the error rate is 0.03 and the delay rate is 0.05, and the weighted average value is 0.04, which is lower than 0.1, and the platform is determined to be stable. This monitoring mechanism dynamically optimizes resource allocation through real-time data analysis to ensure production efficiency.

[0125] In one possible implementation, the abnormal module list is updated regularly, and the weight is adjusted in combination with the deviation amplitude of CPU and memory indicators.

[0126] For example, the cutting module has aged due to device aging, and the memory usage has risen to 2.8GB, triggering a weight reduction to 0.6 and a task quantity reduction to 400, which is preferentially allocated to modules with better performance. This dynamic adjustment mechanism effectively balances system load.

[0127] Step S105, 5 collects the operating status data of the high-availability ceramic stack production platform and the production indicators of the multi-layer ceramic stack through the real-time monitoring module and the status feedback mechanism, and uses the time series analysis algorithm to perform trend prediction and anomaly detection on the central processing unit utilization rate, memory occupancy rate and environmental temperature and humidity control data, and generates an automated operation and maintenance instruction set, including parameter adjustment and resource reallocation. If any performance indicator deviates from the normal range by more than the preset tolerance T5=10%, the automatic optimization program is triggered to adjust the ceramic stack production parameters, and a stable ceramic capacitor stack digital service capability is generated through remote fault diagnosis and log audit functions.

[0128] The real-time monitoring module collects CPU usage, memory occupancy, and ambient temperature and humidity data from the ceramic laminate production platform and writes them to the InfluxDB database at minute intervals to generate a structured state data set. The ARIMA algorithm is used to perform a 7-day rolling training on the structured state data set to obtain a trend prediction model and output a predicted data set. If any indicator in the predicted data set exceeds the dynamic threshold calculated based on the 3σ principle, the PID controller is triggered to calculate the ceramic laminate production parameter adjustment value and generate a parameter adjustment instruction set containing temperature compensation values ​​and pressure correction values. The parameter adjustment instruction set is distributed to the temperature control execution module and pressure regulation module via the MQTT protocol to obtain the updated operating status data set.

[0129] Specifically, the real-time monitoring module collects CPU usage, memory occupancy, and ambient temperature and humidity data from the ceramic laminate production platform and writes it to the InfluxDB database at minute intervals to form a structured state dataset. The structured state dataset is trained on a seven-day rolling basis using the ARIMA algorithm to generate a trend prediction model and output a predicted data set. When any indicator in the predicted data set exceeds a dynamic threshold calculated based on the 3σ principle, a PID controller is triggered to calculate the ceramic laminate production parameter adjustment value and generate a parameter adjustment instruction set containing temperature compensation and pressure correction values. The remote fault diagnosis module reads the compensation value from the parameter adjustment instruction set and, combined with the actual kiln temperature value in the current operating status data, generates the adjusted ceramic laminate production parameter set.

[0130] The optimized parameter set is distributed to the production platform's temperature control execution module and pressure regulation module via the MQTT protocol. Upon completion, an updated operational status dataset is returned. The log audit module retrieves 30 days of historical operational status data from InfluxDB, compares it with the current updated dataset, and generates a digital service instruction set containing temperature calibration and pressure maintenance instructions. The Kong gateway extracts critical instructions with a priority greater than 90 from the instruction set. After JWT verification, it is sent to the production platform control terminal to retrieve the platform's operational logs and update the stable operational status data.

[0131] For example, in a ceramic laminate production platform, the real-time monitoring module uses sensors to collect data on CPU usage, memory utilization, and ambient temperature and humidity. This data is written to the InfluxDB database once a minute, forming a structured status dataset. As a time series database, InfluxDB is suitable for storing high-frequency time series data, ensuring that data is stored in order by timestamp.

[0132] For example, the CPU usage of a production node is recorded as 75%, memory usage is 60%, ambient temperature is 28°C, and humidity is 45%. This data is stored in JSON format for subsequent analysis.

[0133] It should be noted that the real-time monitoring module communicates with the sensor through the Modbus protocol to ensure the stability and real-time performance of data acquisition.

[0134] In one possible implementation, the ARIMA algorithm performs a rolling 7-day training on a structured state dataset to generate a trend forecasting model. ARIMA analyzes the trends and seasonality of historical data to predict changes in indicators over the next 24 hours.

[0135] For example, the model predicts that the memory utilization of a node may rise to 85% in the next six hours, exceeding the normal range. Based on the 3σ principle, the system calculates a dynamic threshold. Assuming the mean memory utilization is 60% and the standard deviation is 10%, the threshold is 90%. When the predicted value exceeds the threshold, the PID controller activates and calculates temperature compensation and pressure correction values.

[0136] For example, the controller generates instructions: temperature decreases by 2°C, pressure increases by 0.5MPa, forming a parameter adjustment instruction set.

[0137] Specifically, the remote fault diagnosis module reads the compensation values ​​in the instruction set and, combined with the measured furnace temperature (e.g., 230°C), generates an optimized parameter set, such as a temperature of 228°C and a pressure of 5.5 MPa. These parameters are distributed to the temperature control execution module and the pressure regulation module via the MQTT protocol. MQTT, with its lightweight and low-bandwidth characteristics, is well-suited for fast instruction transmission in production environments.

[0138] For example, the temperature control module receives a command to set the temperature to 228°C and adjusts the power to the heating element, while the pressure module regulates the hydraulic system. After execution, the module returns to an operational state, such as "Temperature stabilized at 228°C, pressure reached 5.5 MPa."

[0139] For example, the log audit module extracts 30 days of historical data from InfluxDB, compares it to the current operating status, and generates a set of digital service instructions. Assuming that historical data shows a temperature fluctuation range of ±3°C and current data is ±1°C, the audit module generates a "temperature calibration" instruction to ensure long-term stability. The Kong gateway extracts instructions with a priority greater than 90, such as "Emergency Temperature Calibration," and issues them after verifying the device identity using a JWT. The JWT uses the HMAC-SHA256 algorithm to verify the legitimacy of the control terminal.

[0140] For example, if the terminal MAC address is 00:1A:2B:3C:4D:5E, after verification, the command is received and the operation log is returned, and the stable operation status data is updated.

[0141] In one possible implementation, the extension scheme can further optimize instruction generation by analyzing abnormal patterns in historical data.

[0142] For example, if the audit module detects frequent pressure fluctuations at a certain node, it generates a "pressure maintenance" instruction and recommends checking the hydraulic valve status. This multi-faceted analysis, from real-time monitoring to predictive adjustments, instruction distribution, and audit optimization, forms a complete closed loop to ensure efficient operation of the production platform.

[0143] Based on the operating status data of the high-availability ceramic stack production platform collected by the real-time monitoring module, the central processing unit utilization rate is trend predicted through the time series analysis algorithm. If any performance indicator deviates from the normal range and exceeds the preset tolerance, the automatic optimization program will be triggered to adjust the ceramic stack production parameters to obtain stable ceramic capacitor stack digital service capabilities.

[0144] The real-time monitoring module collects CPU usage data, writes it to the InfluxDB time-series database, extracts the last 24 hours of data from the database, and uses the Prophet algorithm to predict usage trends for the next hour. If the predicted value exceeds the preset threshold of 85%, the decision tree-based rule engine is triggered to calculate the temperature adjustment value and pressure correction value of the ceramic laminate production parameters and write them to the RabbitMQ message queue. The historical operating status data of the past 30 days is extracted from the InfluxDB and merged with the latest collected CPU usage data. The K-means clustering algorithm is used to match similar operating conditions, generate a digital service instruction set, transmit it to the control terminal via the ModbusTCP protocol, and update the production parameters.

[0145] Specifically, the real-time monitoring module collects CPU usage data from the high-availability ceramic laminate production platform and writes it to the InfluxDB time series database. It extracts the last 24 hours of CPU usage data from InfluxDB and uses the Prophet algorithm to predict usage trends for the next hour. If the predicted value exceeds a preset threshold of 85%, a decision tree-based rule engine is triggered to calculate adjustment values ​​for ceramic laminate production parameters. The rule engine outputs temperature adjustment values ​​and pressure correction values, which are written to the RabbitMQ message queue. The temperature control module obtains the temperature adjustment value from RabbitMQ, and the pressure control module obtains the pressure correction value from RabbitMQ to perform parameter updates.

[0146] The updated CPU utilization data is collected by the real-time monitoring module and rewritten to InfluxDB. Historical operating status data from the past 30 days is extracted from InfluxDB and merged with the latest CPU utilization data. A K-means clustering algorithm is used to match similar operating conditions. Based on the clustering results, a digital service instruction set containing optimal parameters is generated. This digital service instruction set is transmitted to the production platform control terminal via the Modbus TCP protocol. After receiving the instructions, the control terminal updates the ceramic laminate production parameters. The real-time monitoring module continuously collects updated CPU utilization data, forming a closed-loop control process.

[0147] The real-time monitoring module continuously collects CPU usage data of the high-availability ceramic laminate production platform through sensors and writes it to the InfluxDB time series database.

[0148] For example, the CPU utilization of a production node is recorded every minute for the past 24 hours, with data points such as 70%, 72%, and 75%, forming an ordered time series. InfluxDB, with its high throughput and timestamp indexing, ensures efficient data storage and fast querying. The data collection process is implemented using the Modbus TCP protocol, ensuring stable communication between sensors and monitoring modules, with data transmission latency below 50 milliseconds.

[0149] It should be noted that the ModbusTCP protocol uses TCP / IP network transmission and is suitable for highly reliable data exchange in industrial environments. The last 24 hours of CPU usage data was extracted from InfluxDB and forecasted using the Prophet algorithm.

[0150] For example, Prophet analyzes cyclical fluctuations in historical data and predicts that utilization may increase from 75% to 86% within the next hour. The Prophet algorithm decomposes trends, seasonality, and noise to generate a smooth forecast curve. If the predicted value exceeds the preset threshold of 85%, the decision tree-based rules engine is triggered. The rules engine generates parameter adjustments based on historical operating conditions and current data.

[0151] For example, the engine infers that the temperature needs to be lowered by 1.5°C or the pressure needs to be increased by 0.3 MPa to reduce CPU load. These adjustments are written to the RabbitMQ message queue. RabbitMQ, with its high concurrency processing capabilities, ensures fast command delivery.

[0152] For example, the queue stores the commands "Set temperature to 27.5°C" and "Set pressure to 5.3 MPa," which are sent to the temperature control module and pressure regulation module, respectively. The temperature control module obtains the temperature adjustment value from RabbitMQ and adjusts the power of the heating element.

[0153] For example, if the current kiln temperature is 230°C, it will be adjusted to 228.5°C upon receiving the command. The pressure regulation module then adjusts the hydraulic system to 5.3 MPa based on the command. After execution, the real-time monitoring module collects the updated CPU utilization data, such as if it has dropped to 78%, and rewrites it to InfluxDB, forming a feedback loop.

[0154] It should be noted that this closed-loop control reduces the risk of system overload through rapid response. Historical data from the past 30 days is extracted from InfluxDB, merged with the latest data, and the K-means clustering algorithm is used to match similar operating conditions.

[0155] For example, the algorithm categorizes data into three operating conditions: normal, high load, and abnormal. A current utilization rate of 78% is classified as normal, matching historical data showing a temperature of 228°C and a pressure of 5.2 MPa. Based on these clustering results, a set of digital service instructions is generated, such as "Maintain current parameters" or "Check cooling system." These instructions are transmitted to the control terminal via the Modbus TCP protocol.

[0156] For example, after receiving the "maintain parameters" command, the terminal retains the existing settings and returns the status log.

[0157] In one possible implementation, the extended solution can generate preventive instructions by analyzing abnormal operating conditions in the clustering results.

[0158] For example, if historical data reveals significant temperature fluctuations at a certain node under high load, a "Check temperature control sensor" command will be generated. This multi-dimensional analysis, from real-time monitoring to predictive adjustments, command distribution, and operating condition matching, forms a complete closed loop to ensure stable operation of the production platform.

[0159] Example 2 This embodiment discloses an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the operation mode of the above-mentioned method for remotely controlling a ceramic capacitor stacking machine is implemented.

[0160] Since the electronic device described in this embodiment is an electronic device used to implement a remote control method for a ceramic capacitor stacking machine in the embodiment of this application, based on the remote control method for a ceramic capacitor stacking machine described in the embodiment of this application, those skilled in the art can understand the specific implementation of the electronic device of this embodiment and its various variations, so how the electronic device implements the method in the embodiment of this application will not be described in detail here. As long as those skilled in the art implement the electronic device used in the remote control method for a ceramic capacitor stacking machine in the embodiment of this application, it falls within the scope of protection of this application.

[0161] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters and thresholds in the formulas are set by technicians in this field according to actual conditions.

[0162] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiment. All technical solutions based on the principles of the present invention are within the scope of protection of the present invention. It should be noted that for users of ordinary skill in the art, various improvements and modifications that do not depart from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A remote control method for a ceramic capacitor lamination machine, characterized in that: include: Acquire micron-level alignment accuracy data and crack and void detection data during the multilayer ceramic stacking process from the automated feeding system and defect detection module of the ceramic capacitor stacking machine through a distributed data acquisition interface; A data cleaning algorithm is used to remove redundant records and outliers. The dataset is verified based on the evaluation criteria of completeness (missing rate lower than expected), consistency (format uniformity higher than expected), and accuracy (error rate lower than expected). If the overall data quality score falls below the preset threshold, a data repair mechanism is triggered to generate a standardized ceramic laminate dataset. The standardized ceramic stackup dataset is sharded and processed in parallel using an in-memory computing engine. The ceramic stackup complexity weight coefficient W is calculated based on the characteristic vectors of the multilayer ceramic stack. If W exceeds a preset threshold, a high-performance computing resource pool consisting of a graphics processing unit cluster and distributed storage is activated to generate the ceramic stackup complexity analysis results. Based on the results of the ceramic stacking complexity analysis, an adaptive load balancing algorithm is used to dynamically adjust the task allocation ratio of computing nodes. A support vector machine algorithm is used to predict the load trend of the ceramic stacking production system based on environmental temperature and humidity control data and the operating status of the high-speed cutting device, generating a load prediction value R in tasks per second. Based on R and the current system utilization U, the expansion factor S is calculated. If S exceeds the preset threshold, the elastic expansion mechanism is triggered to generate an optimized ceramic laminate production resource configuration state.

2. A remote control method for a ceramic capacitor lamination machine according to claim 1, characterized in that: The method of obtaining micron-level alignment accuracy data and crack and gap detection data during the multilayer ceramic stacking process from the automatic feeding system and defect detection module of the ceramic capacitor stacking machine through the distributed data acquisition interface specifically includes: Acquire alignment accuracy data from the feeding system of the ceramic capacitor stacking machine through the OPCUA protocol, obtain crack and gap data from the defect detection module, and merge them to generate the original data set; The drop_duplicates function of Pandas is used to remove duplicate records from the original data set, and the quantile function is used to remove outliers outside the quantiles within the upper and lower scale ranges to generate a cleaned data set; If the missing rate of the cleaned data set is higher than a preset threshold, linear interpolation is used to fill in the missing values ​​to generate a completed data set; According to pre-established format rules, the field types and unit uniformity of the completed data set are verified to determine a format uniformity rate.

3. A remote control method for a ceramic capacitor stacking machine according to claim 2, characterized in that: The sharding and parallel processing of the standardized ceramic stack dataset by the memory computing engine specifically includes: The Apache Spark memory computing framework is used to process the standardized ceramic stack dataset; The data output by Spark is sharded using consistent hashing using the libketama library and stored in A_O data shards in the HDFS distributed file system. Based on the sharding distribution structure, A_R computing nodes are started using the MPI parallel framework. Each node calls OpenCV's SIFT algorithm to extract micron-level alignment coordinate features and calls the ResNet50 pre-trained model to extract crack and void probability features to generate a 256-dimensional feature vector set. Input the feature vector set into the mutual_info_classif function of scikit-learn, calculate the information gain of each feature, and generate a set of feature importance coefficients; MinMaxScaler is used to perform linear normalization on the eigenvector set, compressing all eigenvalues ​​to the range of 0 to 1 to generate a normalized eigenvalue set.

4. A remote control method for a ceramic capacitor lamination machine according to claim 3, characterized in that: Dynamically adjusting the task allocation ratio of computing nodes by using an adaptive load balancing algorithm specifically includes: The allocation weight of each computing node is calculated through the weighted polling algorithm to obtain the node task allocation table; Acquire real-time temperature and humidity data from the ambient temperature and humidity sensor, analyze equipment fault codes from the cutting device log, generate compensation coefficients if the temperature and humidity exceed the preset range, and mark the abnormal node if the fault code appears N times consecutively; The compensation coefficient, the fault mark and the node CPU usage are combined into an input vector, and a load prediction model is trained using a support vector machine algorithm to output a load prediction value for the next M minutes.

5. The remote control method for a ceramic capacitor lamination machine according to claim 4, characterized in that: The ceramic laminate production resource configuration status and ceramic laminate complexity analysis results adopt a remote control data encryption protocol and real-time instruction transmission function, decouple the functional modules of the hot pressing forming process and the high-speed cutting device through a microservice architecture, and use an application programming interface gateway to uniformly manage the data interaction and user authority verification of multi-device collaborative management. Asynchronous production requests are processed through a message queue. If the service response time exceeds the delay threshold determined by the ceramic laminate complexity analysis results, the service degradation mechanism is activated to generate a highly available ceramic laminate production platform.

6. A remote control method for a ceramic capacitor lamination machine according to claim 5, characterized in that: The data encryption protocol and real-time instruction transmission function through remote control specifically include: Obtaining the ceramic laminate production resource configuration status through AES-256-CBC encryption, and parsing the configuration status to obtain a production task allocation table; Using RabbitMQ message queue to distribute the task instructions of the production task allocation table to the hot pressing forming module and the cutting device module, and obtain the execution status of the modules; Processing the real-time production request sequence of the module execution status through the Kafka message queue, counting the request processing time of each module, and marking it as a high-latency module if the processing time exceeds a preset delay threshold, and obtaining a high-latency module list; A weighted round-robin algorithm is used to adjust the task ratio of the high-latency module to generate an optimized task allocation table.

7. A remote control method for a ceramic capacitor lamination machine according to claim 6, characterized in that: The high-availability ceramic stack production platform collects operating status data and production indicators of multi-layer ceramic stacks through a real-time monitoring module and a status feedback mechanism, and uses a time series analysis algorithm to perform trend prediction and anomaly detection on the central processing unit utilization rate, memory occupancy rate and environmental temperature and humidity control data, and generates an automated operation and maintenance instruction set, including parameter adjustment and resource reallocation. If any performance indicator deviates from the normal range and exceeds the preset tolerance, the automatic optimization program is triggered to adjust the ceramic stack production parameters, and a stable ceramic capacitor stack digital service capability is generated through remote fault diagnosis and log audit functions.

8. The remote control method for a ceramic capacitor lamination machine according to claim 7, characterized in that: The operation status data of the high-availability ceramic laminate production platform collected through the real-time monitoring module and the status feedback mechanism specifically includes: The real-time monitoring module collects CPU usage, memory occupancy, and ambient temperature and humidity data from the ceramic laminate production platform and writes them to the InfluxDB database at minute intervals to generate a structured status dataset. Using the ARIMA algorithm to perform X_P-day rolling training on the structured state data set, a trend prediction model is obtained and a prediction data set is output; If any indicator in the predicted data set exceeds a dynamic threshold value calculated based on the 3σ principle, a PID controller is triggered to calculate a ceramic laminate production parameter adjustment value and generate a parameter adjustment instruction set including a temperature compensation value and a pressure correction value; The parameter adjustment instruction set is distributed to the temperature control execution module and the pressure regulation module through the MQTT protocol to obtain the updated operation status data set.

9. The remote control method for a ceramic capacitor lamination machine according to claim 8, characterized in that: The use of a time series analysis algorithm to predict the trend of CPU usage specifically includes: The real-time monitoring module collects CPU usage data, writes it into the InfluxDB time series database, extracts the most recent X_V hours of data from the database, and uses the Prophet algorithm to predict the usage trend for the next X_G hours. If the predicted value exceeds the preset threshold of 85%, the decision tree-based rule engine is triggered to calculate the temperature adjustment value and pressure correction value of the ceramic laminate production parameters and write them to the RabbitMQ message queue; The historical operating status data is extracted from the InfluxDB and merged with the latest collected CPU usage data. The K-means clustering algorithm is used to match similar working conditions, and a digital service instruction set is generated. The set is transmitted to the control terminal via the ModbusTCP protocol to update the production parameters.

10. A remote control system for a ceramic capacitor stacking machine, applied to a remote control method for a ceramic capacitor stacking machine according to any one of claims 1 to 9, characterized in that: include: A distributed data acquisition module is used to acquire micron-level alignment accuracy data and crack and void detection data from the ceramic capacitor stacking machine's automated feeding system and defect detection module during the multilayer ceramic stacking process. A data cleaning algorithm is used to remove redundant records and outliers. The dataset is verified based on the evaluation criteria of completeness (a missing rate lower than expected), consistency (a format uniformity higher than expected), and accuracy (an error rate lower than expected). If the overall data quality score falls below a preset threshold, a data repair mechanism is triggered to generate a standardized ceramic stacking dataset. An in-memory computing module is used to perform sharded storage and parallel processing of a standardized ceramic stack dataset. The module calculates the ceramic stack complexity weight coefficient W based on the eigenvectors of the multilayer ceramic stack. If W exceeds a preset threshold, a high-performance computing resource pool comprising a graphics processing unit cluster and distributed storage is activated to generate ceramic stack complexity analysis results. The load balancing module is used to dynamically adjust the task allocation ratio of computing nodes through an adaptive load balancing algorithm based on the results of the ceramic stacking complexity analysis. It also uses a support vector machine algorithm to predict the load trend of the ceramic stacking production system based on environmental temperature and humidity control data and the operating status of the high-speed cutting device, and generates a load prediction value R in tasks per second; The expansion control module is used to calculate the expansion factor S based on R and the current system utilization U. If S exceeds the preset threshold, the elastic expansion mechanism is triggered to generate an optimized ceramic laminate production resource configuration state.

Images