Aluminum ceiling production optimization method and system based on data mining

Abstract

The invention relates to the technical field of data processing, and discloses an aluminum ceiling production optimization method and system based on data mining. The method comprises the steps that aluminum ceiling production parameters are collected and preprocessed to form a standardized data set; calculating feature weights and screening key parameters; dividing production stages and determining positions of differentiation points and decoupling points; performing correlation analysis to construct a parameter-quality mapping matrix; establishing a multi-objective weighting function to generate an optimal parameter scheme; and executing production and recording deviation data for closed-loop updating. According to the method, a full-process parameter optimization mechanism based on data mining is established, multiple intelligent algorithms are fused to deeply mine and analyze mass production data, and scientific configuration and dynamic optimization of aluminum ceiling production parameters are realized, so that the production efficiency is improved, the material waste is reduced, the energy consumption is reduced, and the product quality is ensured.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a method and system for optimizing aluminum ceiling production based on data mining. Background Art

[0002] As consumers' demands for aesthetically pleasing and personalized spaces continue to rise, aluminum ceiling products have become widely used in commercial and residential decoration. Traditional aluminum ceiling production relies primarily on manual experience to set process parameters and control quality. The production process typically includes aluminum sheet pretreatment, cutting and forming, assembly, and surface treatment. In recent years, some advanced companies have begun to introduce automated equipment and digital management systems, which have improved production efficiency to a certain extent by collecting basic production data and conducting simple analysis. Existing aluminum ceiling production optimization methods mainly use traditional quality management methods such as statistical process control (SPC), lean production, and Six Sigma. These methods ensure product quality and control production costs by setting control ranges for key parameters, reducing production waste, and improving process stability.

[0003] However, existing technologies have obvious shortcomings in optimizing the production of aluminum ceilings. First, traditional methods rely too much on expert experience for parameter setting and lack a scientific data-driven decision-making mechanism, resulting in limited and non-replicable parameter optimization effects. Second, production data collection is usually isolated, and there is a lack of effective correlation and integrated analysis between data in various links, making it impossible to explore the potential relationship between parameters and the comprehensive impact on quality. Third, most traditional methods ignore the optimization positioning of product differentiation points and customer order decoupling points, and cannot effectively support customized production of multiple varieties and small batches. Fourth, most existing production parameter optimizations use static configuration methods, and there is no data closed-loop feedback mechanism, which makes it impossible to continuously optimize parameter configuration based on actual production results. These shortcomings have led to widespread problems in the production of aluminum ceilings, such as serious material waste, high energy consumption, unstable product quality, and long delivery cycles, which have seriously restricted the development of the industry and the market competitiveness of enterprises. Summary of the Invention

[0004] The present application provides an aluminum ceiling production optimization method and system based on data mining, which is used to establish a full-process parameter optimization mechanism based on data mining, integrate multiple intelligent algorithms to conduct in-depth mining and analysis of massive production data, and realize scientific configuration and dynamic optimization of aluminum ceiling production parameters, thereby improving production efficiency, reducing material waste, reducing energy consumption and ensuring product quality.

[0005] In the first aspect, the present application provides an aluminum ceiling production optimization method based on data mining, and the aluminum ceiling production optimization method based on data mining includes: collecting cutting parameters, forming parameters, assembly parameters and surface treatment parameters of the aluminum ceiling production line and pre-processing to obtain a standardized production data set; performing feature weight calculation on the standardized production data set, screening key parameters affecting the quality of the aluminum ceiling according to the information gain principle, and obtaining a key feature data set; dividing the production stages according to the key feature data set, calculating the cost curve and time curve of each production stage, and obtaining product differentiation point and customer order decoupling point position data; correlating the product differentiation point and customer order decoupling point position data with historical quality score data, constructing a parameter-quality mapping relationship matrix, and obtaining quality prediction indicators; establishing a multi-objective weight function based on the quality prediction indicators and production cost data, generating an optimal solution set of parameter configuration combinations, and obtaining a production parameter configuration plan; executing production operations according to the production parameter configuration plan, recording the deviation data between actual production indicators and quality evaluation indicators, and performing data closed-loop update through the deviation data.

[0006] In a second aspect, the present application provides an aluminum ceiling production optimization system based on data mining, the aluminum ceiling production optimization system based on data mining comprising:

[0007] The processing module is used to collect the cutting parameters, forming parameters, assembly parameters and surface treatment parameters of the aluminum ceiling production line and pre-process them to obtain a standardized production data set;

[0008] A calculation module is used to calculate feature weights of the standardized production data set, screen key parameters affecting the quality of the aluminum ceiling according to the information gain principle, and obtain a key feature data set;

[0009] a division module, configured to divide the production stages according to the key feature data set, calculate the cost curve and time curve of each production stage, and obtain the location data of the product differentiation point and the customer order decoupling point;

[0010] A correlation module is used to perform correlation analysis on the product differentiation point and customer order decoupling point location data with historical quality score data, construct a parameter-quality mapping relationship matrix, and obtain quality prediction indicators;

[0011] A generation module is used to establish a multi-objective weight function based on the quality prediction index and production cost data, generate an optimal solution set of parameter configuration combinations, and obtain a production parameter configuration plan;

[0012] The updating module is used to execute the production operation according to the production parameter configuration scheme, record the deviation data between the actual production index and the quality evaluation index, and perform data closed-loop update through the deviation data.

[0013] In a third aspect, a computer device is provided, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor calls the instructions in the memory so that the computer device executes the above-mentioned aluminum ceiling production optimization method based on data mining.

[0014] In a fourth aspect, a computer-readable storage medium is provided, wherein instructions are stored in the computer-readable storage medium, which, when executed on a computer, enables the computer to execute the above-mentioned aluminum ceiling production optimization method based on data mining.

[0015] The technical solution provided in this application constructs a high-quality standardized production data set by deploying precision sensors in each link of the aluminum ceiling production line to collect cutting, forming, assembly, and surface treatment parameters. This is combined with an extended Kalman filter to filter outliers, a multivariate interpolation algorithm to fill missing values, information entropy calculation to remove redundant data, and a Z-Score normalization method to unify data dimensions. This lays a solid data foundation for subsequent analysis. Secondly, feature weight calculation and information gain analysis are performed on the standardized data to screen out key features such as aluminum thickness deviation rate, cutting angle accuracy, and forming temperature fluctuation range, significantly reducing data dimensions and improving computational efficiency. In the production stage division link, this solution innovatively combines key feature data to calculate cost and time curves, scientifically determining the location of product differentiation points and customer order decoupling points, achieving optimal configuration of production resources and improving customer responsiveness. By correlating the decoupling point location data with historical quality scores and constructing a parameter-quality mapping relationship matrix, a scientific basis for quality prediction is provided. Based on a comprehensive evaluation system with a multi-objective weight function, the three major goals of quality, cost, and time are taken into account simultaneously to generate an optimal parameter configuration solution that balances multi-dimensional production goals. It is particularly worth emphasizing that this solution utilizes artificial intelligence technologies such as the convolutional neural network-long short-term memory artificial neural network algorithm (C-LSTM) and the support vector machine-red deer optimization algorithm (SVM-RDO). This fully considers the specific needs of the aluminum ceiling production field and deeply integrates these algorithm features with specific production scenarios. For example, the C-LSTM algorithm can effectively capture long-term dependencies when processing production time series data, providing accurate predictions for optimizing decoupling point locations. The SVM-RDO algorithm, through adaptive parameter adjustment, effectively addresses problems such as overfitting and local minima in aluminum ceiling production, significantly improving the accuracy of parameter optimization. Furthermore, the data closed-loop update mechanism established in this solution achieves continuous optimization of production parameters through real-time monitoring, quality inspection, deviation analysis, and parameter correction. This allows the system to continuously improve itself as data accumulates, effectively overcoming the limitations of traditional static configuration methods and ensuring the stability of the production process and the consistency of product quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0017] Figure 1 A schematic diagram of an embodiment of an aluminum ceiling production optimization method based on data mining in an embodiment of the present application;

[0018] Figure 2 This is a schematic diagram of an embodiment of an aluminum ceiling production optimization system based on data mining in an embodiment of the present application;

[0019] Figure 3 It is a schematic block diagram of the structure of a computer device in an embodiment of the present invention. DETAILED DESCRIPTION

[0020] The embodiments of the present application provide a method and system for optimizing the production of aluminum ceilings based on data mining. The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of this application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or that are inherent to these processes, methods, products or devices.

[0021] For ease of understanding, the specific process of the embodiment of the present application is described below. Figure 1 In the embodiment of the present application, an embodiment of the aluminum ceiling production optimization method based on data mining includes:

[0022] Step S101: collecting cutting parameters, forming parameters, assembly parameters and surface treatment parameters of the aluminum ceiling production line and pre-processing them to obtain a standardized production data set;

[0023] Step S102: Calculate feature weights for the standardized production data set, and screen key parameters that affect the quality of aluminum ceilings according to the information gain principle to obtain a key feature data set;

[0024] Step S103: Divide the production stages according to the key feature data set, calculate the cost curve and time curve of each production stage, and obtain the location data of the product differentiation point and the customer order decoupling point;

[0025] Step S104: performing correlation analysis on the product differentiation point and customer order decoupling point location data with the historical quality score data, constructing a parameter-quality mapping relationship matrix, and obtaining quality prediction indicators;

[0026] Step S105: Based on the quality prediction index and production cost data, a multi-objective weight function is established to generate an optimal solution set of parameter configuration combinations to obtain a production parameter configuration plan;

[0027] Step S106: Execute production operations according to the production parameter configuration plan, record the deviation data between the actual production indicators and the quality evaluation indicators, and perform data closed-loop update based on the deviation data.

[0028] It is understandable that the execution subject of this application can be an aluminum ceiling production optimization system based on data mining, or a terminal or a server, which is not limited here. The embodiment of this application is described by taking the server as the execution subject as an example.

[0029] Specifically, data collection is achieved by installing multiple sensors on the aluminum ceiling production line. During the cutting process, sensors collect parameters such as cutting speed, cutting angle, cutting temperature, and equipment vibration frequency; during the forming process, they collect parameters such as forming pressure, forming temperature, and forming speed; during the assembly process, they collect parameters such as assembly accuracy, assembly time, and tightening torque; and during the surface treatment process, they collect parameters such as spray thickness, drying temperature, and drying time. This raw data often contains outliers and missing values, necessitating outlier filtering using an extended Kalman filter. The extended Kalman filter establishes a data state equation and an observation equation to identify and filter data points that deviate from the normal distribution. For example, if the cutting temperature data suddenly increases to an abnormal value of 800°C, this value significantly deviates from the normal operating temperature range (typically 200-400°C). The extended Kalman filter identifies and filters this outlier. Subsequently, a multivariate interpolation algorithm is used to fill in the missing values, appropriately filling in the gaps based on the historical data trends of the relevant parameters. Information entropy calculation was performed on the complete data, and redundant data with an entropy contribution rate below a set threshold was removed. Finally, the Z-Score normalization method was used to convert the parameters of different dimensions into a unified standard normal distribution interval to form a standardized production dataset. Feature weights were calculated and screened for the standardized production dataset. First, a preliminary feature screening was performed on the cutting speed, cutting angle, cutting temperature, and equipment vibration frequency parameters to form candidate features for the cutting process. Similarly, preliminary screening was performed on the parameters for the forming, assembly, and surface treatment processes to obtain corresponding candidate features. These candidate features were then integrated into a feature pool. Information gain calculation was performed on each feature in the feature pool to rank the features by importance. Information gain calculation was determined by analyzing the degree to which the features affect the final quality of the aluminum ceiling products. Specifically, the range of each feature's value was divided into several intervals. The product quality distribution within each interval was statistically analyzed, and the reduction in information entropy was calculated. After calculation, key indicators such as aluminum thickness deviation rate, cutting angle accuracy, forming temperature fluctuation range, assembly gap parameters, and surface treatment uniformity were selected. The interrelationships between these parameters were analyzed, and a feature association network was constructed to form a key feature dataset.

[0030] Based on the key feature data set, the production stage is divided. According to the distribution characteristics of the key features, the aluminum ceiling production process is divided into a general production stage and a special production stage. The general production stage refers to the basic processing procedures applicable to a variety of product specifications, such as aluminum plate pretreatment and basic cutting; the special production stage refers to the customized processes for specific product models, such as special shaping and surface treatment. The time of each process node in these two stages is measured, and the process time distribution diagram is drawn to obtain the time curve. At the same time, the unit time cost of each process is calculated to obtain the cost function of each stage. The time curve and cost function are combined to construct a time-cost comprehensive evaluation model. The comprehensive cost data at different decoupling points are analyzed to find the cost jump point and determine the optimal location of the product differentiation point and the customer order decoupling point.

[0031] A correlation analysis was conducted between the location data of product differentiation points and customer order decoupling points and historical quality score data. First, historical quality score data for aluminum ceiling products was extracted from the company's production management system and labeled as "high-quality," "qualified," and "defective." This data was then matched with the location data of product differentiation points and customer order decoupling points for the corresponding production batches. Statistical analysis was then performed to calculate the distribution of quality grades at different decoupling point locations. This analysis identified areas of stable and fluctuating quality, assessed quality risks at different decoupling points, and extracted production parameter configuration data at each decoupling point location. Correlation analysis was performed between these parameters and the quality scores, and correlation coefficients were calculated. A parameter-quality mapping matrix was established to generate quality prediction indicators.

[0032] A multi-objective weighting function is established based on quality prediction indicators and production cost data. First, the quality prediction indicators are converted into numerical quality scores. Production cost data is extracted from the enterprise resource management system to establish a parameter-cost function. Delivery time data is extracted from historical orders, and the production cycle under different production parameter configurations is analyzed to form a time objective function. The quality objective function, cost objective function, and time objective function are linearly combined to construct a multi-objective weighting function, which determines the parameter optimization space. Parameter combinations are traversed within the parameter optimization space, and the comprehensive evaluation score for each parameter group is calculated. The parameter combination with the highest comprehensive score is selected to form the optimal parameter configuration. Production operations are executed according to the production parameter configuration plan. The parameter configuration plan is converted into a detailed production execution plan and distributed to each process operation station. While the aluminum ceiling production operation is in progress, a full-process monitoring system is activated to record production parameter data in real time and compare it with the preset parameters. Warning signals are triggered when the parameters deviate from the set range. After production is completed, the aluminum ceiling products are quality inspected, and the actual quality data is compared and analyzed with the quality prediction indicators to calculate the prediction deviation. Perform correlation analysis on parameter deviation records and quality prediction deviation data to reveal the law of how parameter changes affect quality, update the weight coefficients and influencing factors in the parameter-quality mapping relationship matrix, form updated quality prediction rules, feed back to the data acquisition system, integrate with the original data, update the production parameter optimization database, and complete the data closed-loop update.

[0033] In the embodiment of the present application, by deploying precision sensors in each link of the aluminum ceiling production line to collect cutting, forming, assembly and surface treatment parameters, combining the extended Kalman filter to filter outliers, the multivariate interpolation algorithm to fill missing values, the information entropy calculation to remove redundant data, and the Z-Score standardization method to unify the data dimension, a high-quality standardized production data set is constructed, which lays a solid data foundation for subsequent analysis. Secondly, feature weight calculation and information gain analysis are performed on the standardized data to screen out key features such as aluminum thickness deviation rate, cutting angle accuracy, and forming temperature fluctuation range, which greatly reduces the data dimension and improves the calculation efficiency. In the production stage division link, this solution innovatively combines key feature data to calculate cost and time curves, scientifically determines the product differentiation point and customer order decoupling point location, and achieves the optimal configuration of production resources and the improvement of customer responsiveness. By correlating the decoupling point location data with the historical quality score, a parameter-quality mapping relationship matrix is constructed, which provides a scientific basis for quality prediction. Based on the comprehensive evaluation system of the multi-objective weight function, the three major goals of quality, cost and time are taken into account at the same time, and the optimal parameter configuration scheme is generated to balance the multi-dimensional production goals. It is particularly worth emphasizing that this solution utilizes artificial intelligence technologies such as the convolutional neural network-long short-term memory artificial neural network algorithm (C-LSTM) and the support vector machine-red deer optimization algorithm (SVM-RDO). This fully considers the specific needs of the aluminum ceiling production field and deeply integrates these algorithm features with specific production scenarios. For example, the C-LSTM algorithm can effectively capture long-term dependencies when processing production time series data, providing accurate predictions for optimizing decoupling point locations. The SVM-RDO algorithm, through adaptive parameter adjustment, effectively addresses problems such as overfitting and local minima in aluminum ceiling production, significantly improving the accuracy of parameter optimization. Furthermore, the data closed-loop update mechanism established in this solution achieves continuous optimization of production parameters through real-time monitoring, quality inspection, deviation analysis, and parameter correction. This allows the system to continuously improve itself as data accumulates, effectively overcoming the limitations of traditional static configuration methods and ensuring the stability of the production process and the consistency of product quality.

[0034] In a specific embodiment, the process of executing step S101 may specifically include the following steps:

[0035] (1) Sensors are set in the cutting process of the aluminum ceiling production line to collect cutting speed, cutting angle, cutting temperature, and equipment vibration frequency parameters to obtain original cutting parameter data; sensors are set in the aluminum ceiling forming process to collect forming pressure, forming temperature, and forming speed parameters to obtain original forming parameter data;

[0036] (2) Sensors are set in the aluminum ceiling assembly process to collect assembly accuracy, assembly time, and tightening torque parameters to obtain original assembly parameter data; sensors are set in the aluminum ceiling surface treatment process to collect spray thickness, drying temperature, and drying time parameters to obtain original surface treatment parameter data;

[0037] (3) filtering outliers on the original cutting parameter data, original molding parameter data, original assembly parameter data, and original surface treatment parameter data using an extended Kalman filter to obtain filtered data;

[0038] (4) Based on the filtered data, the missing values are filled using a multivariate interpolation algorithm to obtain complete data, and the information entropy of the complete data is calculated to remove redundant data whose information entropy contribution rate is lower than the set threshold to obtain streamlined data;

[0039] (5) The streamlined data is converted to the standard normal distribution interval through the Z-Score standardization method to obtain a standardized production data set.

[0040] Specifically, this is achieved by deploying a sensor network on the production line. During the cutting process, a laser speed sensor measures the cutting speed, an angle sensor records the cutting angle, an infrared temperature sensor monitors the cutting temperature, and a piezoelectric vibration sensor detects the vibration frequency of the equipment. These sensors continuously collect data at a sampling frequency of 10Hz and transmit it to the data acquisition unit via the industrial control network. Corresponding sensors are also deployed in the molding process: pressure sensors record molding pressure in real time, thermocouples monitor molding temperature, and displacement sensors measure molding speed. During the assembly process, a laser ranging sensor measures assembly accuracy, a time counter records assembly time, and a torque sensor measures tightening torque. Finally, in the surface treatment process, a coating thickness gauge measures the spray thickness, an intelligent temperature sensor monitors the drying temperature, and an intelligent timer records the drying time. These raw data are subject to noise interference, occasional outliers, and inevitable omissions, and require rigorous preprocessing before they can be used for subsequent analysis.

[0041] The collected raw data is first filtered for outliers using an extended Kalman filter (EKF). The EKF is a recursive filtering algorithm particularly well-suited for nonlinear systems in aluminum ceiling production. The EKF predicts the current state by establishing state and observation equations, and then corrects the prediction based on the actual observed values. In specific implementation, reasonable value ranges are first set for each parameter. For example, the normal range for cutting temperature is 200-400°C. If the measured value at a certain moment is 650°C, significantly deviating from the normal range, the EKF calculates the predicted value (approximately 350°C) and its uncertainty based on the temperature trends of the previous moments. By comparing the predicted value with the actual observed value, the Kalman gain is calculated to determine the state estimate. The clearly abnormal value of 650°C is adjusted by the EKF to be closer to the predicted value, effectively filtering out outliers. The filtered data may still contain missing values due to temporary sensor failures or communication interruptions, requiring data interpolation. A multivariate interpolation algorithm fills in missing values based on the correlations between parameters. First, a correlation matrix is calculated between the parameters to identify other parameters that are highly correlated with the missing parameter. The values of these correlated parameters are then used to estimate the missing value using multiple linear regression or spline interpolation. For example, if cutting speed data is missing, but it is known that cutting speed is highly correlated with cutting temperature and equipment vibration frequency, the values of these two parameters can be used to infer a reasonable cutting speed value. This method is more accurate than simple mean filling or linear interpolation because it accounts for the complex interactions between parameters.

[0042] After completing missing value filling, the dataset may contain a large amount of redundant information, which increases the computational burden and may introduce noise. The information entropy calculation method is used to identify and remove redundant data. Information entropy is an indicator of data uncertainty, and the contribution rate of each parameter to the overall information entropy is calculated. First, each parameter data is divided into multiple intervals, and its information entropy is calculated. Then, the degree of change in the overall information entropy after removing a certain parameter is analyzed. Parameters whose information entropy contribution rate is lower than the set threshold (usually 0.05) are considered redundant data and removed. For example, if the correlation coefficient between drying temperature and drying time is found to be as high as 0.95, and the contribution rate of drying temperature to the overall information entropy is only 0.03, the drying temperature can be removed from the dataset to reduce the dimensionality without significantly losing information.

[0043] The final step is data standardization, which addresses the issue of inconsistent dimensionality across parameters. The Z-Score standardization method converts each parameter to a standard normal distribution with a mean of 0 and a standard deviation of 1. For each parameter value, the mean is subtracted and divided by the standard deviation to obtain the standardized value. The advantage of Z-Score standardization is that it preserves the distribution characteristics and outlier information of the original data while enabling direct comparison of parameters of different dimensionality. Standardized data facilitates subsequent feature extraction and model building.

[0044] For example, during a day of data collection on an aluminum ceiling production line, sensors in the cutting process recorded data every 5 seconds, resulting in a total of 17,280 raw records. Among these records, the cutting temperature showed 15 abnormally high temperatures (over 500°C) and 23 abnormally low temperatures (below 100°C). These outliers were adjusted to more reasonable values after EKF processing. Approximately 200 records were found to contain missing cutting speed values. These missing values were filled using a multivariate interpolation algorithm based on the correlation between cutting angle, cutting temperature, and equipment vibration frequency. Information entropy analysis revealed that the information entropy contribution rates of three of the 11 monitored parameters in the cutting process (auxiliary air pressure, cooling water flow rate, and backplane temperature) were only 0.02, 0.03, and 0.04, respectively. Therefore, these parameters were removed from the dataset. The remaining parameters were Z-score normalized. For example, if the original cutting speed had a mean of 45 m / min and a standard deviation of 5 m / min, and a recorded value of 50 m / min at a certain moment was 50 m / min, the normalized value would be (50 - 45) / 5 = 1.0. Through processing, a standardized production data set is formed.

[0045] In a specific embodiment, the process of executing step S102 may specifically include the following steps:

[0046] (1) Preliminary feature screening of cutting speed, cutting angle, cutting temperature, and equipment vibration frequency parameters in the standardized production data set was performed to obtain candidate features of the cutting process;

[0047] (2) Perform preliminary feature screening on the molding pressure, molding temperature, and molding speed parameters in the standardized production data set to obtain candidate features of the molding process;

[0048] (3) Perform preliminary feature screening on the assembly accuracy, assembly time, and tightening torque parameters in the standardized production data set to obtain candidate features for the assembly process;

[0049] (4) Preliminary feature screening of spray thickness, drying temperature, and drying time parameters in the standardized production data set was performed to obtain candidate features for the surface treatment process;

[0050] (5) Integrate the candidate features of the cutting process, the candidate features of the forming process, the candidate features of the assembly process, and the candidate features of the surface treatment process to form a feature pool and obtain a feature selection scheme;

[0051] (6) Calculate the information gain of each feature in the feature pool based on the feature selection scheme, sort the features by importance, and obtain the feature importance sequence;

[0052] (7) According to the feature importance sequence, the aluminum thickness deviation rate, cutting angle accuracy, forming temperature fluctuation range, assembly gap parameters, and surface treatment uniformity indicators are screened out to obtain the key indicator set;

[0053] (8) Analyze the relationship between the parameters in the key indicator set, construct a feature association network, combine the key indicators with the corresponding weight values, and obtain the key feature data set.

[0054] Specifically, a preliminary feature screening was conducted on parameters involved in the cutting process, including cutting speed, cutting angle, cutting temperature, and equipment vibration frequency. This screening process employed analysis of variance (ANOVA) to calculate the variance ratio of each parameter across products of different quality grades. A larger variance ratio indicates a more significant impact of that parameter on quality. The cutting speed parameter exhibited significant mean differences between high-quality and low-quality product groups, indicating a significant impact on final product quality. Therefore, it was retained as a candidate feature. Cutting angle accuracy also demonstrated a high correlation with product quality, particularly in the production of complex aluminum ceilings, where angle deviation directly impacts splicing quality. Cutting temperature and equipment vibration frequency were also preliminarily screened to form a candidate feature set for the cutting process. A similar analysis was conducted on the forming pressure, forming temperature, and forming speed parameters involved in the forming process. Forming pressure has a direct impact on the degree of deformation of the aluminum sheet, and a one-way ANOVA confirmed its strong correlation with the product deformation rate. Forming temperature, in turn, influences the aluminum's plastic deformation capacity; excessively high or low temperatures can lead to quality issues. The forming speed affects the metal flow state and internal stress distribution. By comparing the product measurement data at different speeds, the optimal speed range is screened out, and these three parameters are used as candidate features of the forming process.

[0055] The assembly process also underwent preliminary screening for assembly accuracy, assembly time, and tightening torque. Assembly accuracy is directly related to the flatness and aesthetics of the aluminum ceiling, and its importance was confirmed by measuring the correlation coefficient between assembly accuracy and customer satisfaction scores. Assembly time reflects operational difficulty and process stability, and its impact on product consistency was determined through regression analysis. Tightening torque affects the strength of the connection. By designing controlled experiments and analyzing the connection strength test results at different torque values, these three parameters were selected as candidate characteristics for the assembly process.

[0056] The surface treatment process also underwent preliminary screening for spray coating thickness, drying temperature, and drying time. Spray coating thickness affects the corrosion resistance and appearance quality of the aluminum ceiling. The optimal thickness range was determined by comparing salt spray test results and appearance scores at different thicknesses. Drying temperature and time jointly determine the degree of coating cure. Orthogonal experiments were conducted to identify the optimal parameter combinations and generate a candidate feature set for the surface treatment process.

[0057] The candidate features from the four steps are integrated into a feature pool to construct a feature selection scheme. At this point, the feature pool contains multiple parameters, which require further screening to reduce model complexity. Based on the feature selection scheme, the information gain value is calculated for each parameter in the feature pool to measure the contribution of each parameter to product quality. The information gain calculation process is as follows:

[0058] IG(Y,X α )=H(Y)-H(Y|X α )

[0059] Among them, IG(Y,X α ) represents feature X α The information gain of the target variable Y; H(Y) represents the entropy of the target variable Y; H(Y|X α ) indicates that given feature X α Conditional entropy of the target variable Y. The entropy calculation formula is:

[0060] H(Y)=-∑ y∈Y P(y)log2P(y)

[0061]

[0062] Among them, P(y) represents the probability that the target variable Y takes the value y; P(x α ) represents feature X α The value is x α The probability of H(Y|X α =x α ) indicates that in feature X α The value is x α The entropy of the target variable Y under the condition.

[0063] By calculating the information gain value for each feature and ranking all features by importance, we generated a feature importance sequence. For example, the information gain for cutting angle accuracy was 0.85, the information gain for forming temperature fluctuation range was 0.78, and the information gain for assembly gap parameter was 0.72. Based on this feature importance sequence, we screened out the parameters with the highest information gain, including aluminum thickness deviation rate, cutting angle accuracy, forming temperature fluctuation range, assembly gap parameter, and surface treatment uniformity, to form a key indicator set.

[0064] Analyze the interrelationships between parameters in the key indicator set and construct a feature association network. This network is implemented by calculating the correlation coefficient matrix between parameters, identifying the mutual influence of parameters. For example, a negative correlation was found between aluminum thickness deviation and cutting angle accuracy, indicating that the thicker the aluminum sheet, the more difficult it is to ensure cutting angle accuracy. A positive correlation was also found between the forming temperature fluctuation range and surface treatment uniformity, indicating that temperature control stability has a positive impact on coating uniformity. Based on these correlation analyses, key indicators are combined with corresponding weights to form a key feature dataset.

[0065] In a specific embodiment, the process of executing step S103 may specifically include the following steps:

[0066] (1) The aluminum ceiling production process is divided into two stages based on the aluminum thickness deviation rate, cutting angle accuracy, forming temperature fluctuation range, assembly gap parameters, and surface treatment uniformity indicators in the key feature data set, and the general production stage and the special production stage are obtained;

[0067] (2) Measure the time of each process node in the general production stage, draw a process time distribution diagram, and obtain the general stage time curve; measure the time of each process node in the special production stage, draw a process time distribution diagram, and obtain the special stage time curve;

[0068] (3) Based on the general stage time curve and the special stage time curve, the unit time cost of each process is calculated to obtain the cost function of each stage;

[0069] (4) Combining the cost functions of each stage with the product delivery time requirements, constructing a time-cost comprehensive evaluation model, and obtaining comprehensive cost data at different decoupling points;

[0070] (5) Analyze the comprehensive cost data, find the cost transition point, take the inflection point of the cost curve as the candidate decoupling point, and obtain the candidate set of decoupling points;

[0071] (6) Based on the decoupling point candidate set and the customer order history data, calculate the customer satisfaction scores corresponding to different decoupling point positions to obtain the decoupling point evaluation results;

[0072] (7) Comprehensively analyze the decoupling point evaluation results and cost data to determine the optimal positions of product differentiation points and customer order decoupling points, and obtain the position data of product differentiation points and customer order decoupling points.

[0073] Specifically, the production process was divided into stages based on five key indicators in the key feature dataset. The distribution characteristics and variation patterns of aluminum thickness deviation rate, cutting angle accuracy, forming temperature fluctuation range, assembly gap parameters, and surface treatment uniformity reveal the inherent structure of the production process. Through feature cluster analysis, the aluminum ceiling production process was divided into general production stages and specialized production stages. The general production stage refers to basic processing procedures applicable to a variety of product specifications, such as aluminum sheet pretreatment and basic cutting. The specialized production stage refers to customized processes for specific product models, such as special shaping and surface treatment.

[0074] Time measurements are performed on the divided production stages, recording the average processing time, time fluctuation range, and influencing factors for each process node. Time measurements for the general production stage primarily include key nodes such as raw material preparation time, aluminum plate pretreatment time, and basic cutting time. At least 30 sample data points are recorded for each node to ensure statistical significance of the time distribution. The same method is used to measure the time of each process node in the specialized production stage, including customized cutting time, special molding time, and surface treatment time. Using this time data, a process time distribution diagram is plotted, resulting in time curves for the general and specialized stages, visually displaying the time characteristics of each production stage.

[0075] Based on the time curve, combined with factors such as labor cost, material cost, equipment operating cost and energy consumption, the unit time cost of each process is calculated. The cost function of each stage can be expressed as:

[0076]

[0077] Among them, C s (t) represents the total cost of stage s at time point t; P represents the number of processes included in stage s; L p (t) represents the labor cost per unit time of process p at time point t; M p (t) represents the unit time material cost of process p at time point t; E p (t) represents the equipment operating cost per unit time of process p at time point t; U p (t) represents the energy cost per unit time of process p at time point t; T p (t) represents the processing time of process p at time point t.

[0078] This cost function takes into account the impact of time-varying costs, particularly at the interface between the general and specialized stages, where cost variations often exhibit nonlinear characteristics. For example, labor costs per unit time are often higher in the specialized stage than in the general stage, as specialized processes require operators with higher skills. Material costs also vary significantly across stages, with material waste rates typically lower in the general stage than in the specialized stage.

[0079] Combining the cost functions of each stage with the product delivery time requirements, a time-cost comprehensive evaluation model is constructed. This model comprehensively considers production costs, production cycles, and delivery time differences, and calculates comprehensive cost data at different decoupling points. The comprehensive cost can be expressed as:

[0080]

[0081] Where TC(d) represents the total comprehensive cost when the decoupling point is at d; S represents the total number of production stages; C s(d) represents the cost of stage s at the decoupling point d; δ early (d) represents the amount of time delivered in advance at the decoupling point position d; P early The profit coefficient of unit lead time; δ late (d) represents the amount of time the delivery is delayed at the decoupling point position d; P late It represents the penalty coefficient for unit delayed delivery time.

[0082] Analyze the comprehensive cost data to identify cost inflection points. Cost inflection points are locations where the slope of the comprehensive cost curve changes significantly, typically manifesting as inflection points. Numerical differentiation is used to calculate the slope change of the cost curve, and points where the slope change exceeds a threshold are marked as candidate decoupling points. The candidate decoupling point set contains multiple possible decoupling locations, each corresponding to a different production organization method and customer response strategy.

[0083] Based on the candidate set of decoupling points and combined with the customer order history data, the customer satisfaction scores corresponding to different decoupling point locations are calculated. The formula for calculating the customer satisfaction score is:

[0084]

[0085] Where CS(d) represents the customer satisfaction score when the decoupling point is at d; T(d) represents the average delivery time when the decoupling point is at d; T max and T min They represent the maximum and minimum values of historical delivery time respectively; F(d) represents the order fulfillment rate deviation when the decoupling point is at d; F max and F min They represent the maximum and minimum values of the historical order fulfillment rate deviation respectively; Q(d) represents the product quality score when the decoupling point is at d; Q max and Q min They represent the maximum and minimum values of the historical product quality scores respectively; ω1, ω2 and ω3 represent the weight coefficients of delivery time, order fulfillment rate and product quality in customer satisfaction evaluation respectively, and satisfy ω1+ω2+ω3=1.

[0086] The decoupling point evaluation results are combined with cost data for analysis, and a weighted summation method is used to determine the optimal locations for the product differentiation point (PDP) and customer order decoupling point (CODP). The PDP is where a product begins to be customized for specific needs, while the CODP is where production transitions from forecast-driven to order-driven. Once these two points are determined, data on the PDP and CODP locations is generated, providing a basis for subsequent production optimization.

[0087] For example, an aluminum ceiling manufacturer applied this method to optimize its production process. By analyzing five key indicators, including aluminum thickness deviation rate and cutting angle accuracy, the production process was divided into a general production phase (comprising three processes: raw material preparation, aluminum sheet pretreatment, and basic cutting) and a specialized production phase (comprising four processes: custom cutting, special forming, surface treatment, and packaging). Time measurements for each process revealed that the aluminum sheet pretreatment process in the general phase had minimal time fluctuation, while the custom cutting process in the specialized phase varied significantly with product complexity. Cost functions for each phase were calculated by incorporating factors such as labor and material costs. A comprehensive time-cost evaluation model was constructed, revealing a clear cost transition point between basic cutting and custom cutting. Furthermore, as customer service requirements increased, the optimal decoupling point shifted toward the early stages of production. Analyzing customer satisfaction based on historical order data ultimately determined that the product differentiation point lies after the basic cutting process, while the customer order decoupling point lies after the aluminum sheet pretreatment process. This arrangement ensures rapid response to customer needs while maintaining production efficiency and cost control, validating the practicality of this method for optimizing aluminum ceiling production.

[0088] In a specific embodiment, the process of executing step S104 may specifically include the following steps:

[0089] (1) The quality score data of historical aluminum ceiling products were extracted from the enterprise production management system and labeled according to three categories: high-quality, qualified, and defective, to obtain a historical quality data set;

[0090] (2) Match the historical quality data set with the product differentiation points and customer order decoupling point location data of the corresponding production batches to obtain location-quality correlation data;

[0091] (3) Statistical analysis is performed on the position-quality correlation data to calculate the distribution ratio of each quality level at different decoupling point positions and obtain a position-quality distribution table;

[0092] (4) Identify the quality stable area and quality fluctuating area according to the position-quality distribution table, divide the quality risk level, and obtain the decoupling point quality risk assessment data;

[0093] (5) Based on the decoupling point quality risk assessment data, the production parameter configuration records at each decoupling point are extracted to obtain a parameter configuration-quality correspondence table;

[0094] (6) Perform correlation analysis on each parameter and quality score in the parameter configuration-quality correspondence table, calculate the correlation coefficient and significance level, and obtain the parameter importance ranking;

[0095] (7) According to the parameter importance ranking, the corresponding relationship between key parameters and quality scores is fitted, and the numerical corresponding relationship is established to obtain the parameter-quality mapping table;

[0096] (8) The parameter-quality mapping table is integrated into a matrix form, which includes the influence weight and variation range of each parameter combination on the quality, and the quality prediction index is obtained.

[0097] Specifically, historical quality score data for aluminum ceiling products was extracted from the company's production management system. Based on industry standards and company quality control requirements, products were categorized into three categories: "high-quality," "qualified," and "defective." High-quality products are defined as those with a quality score of 90-100, characterized by aesthetic appeal, precise dimensions, and easy installation. Acceptable products are defined as those with a quality score of 75-89, meeting basic functional requirements but exhibiting minor defects. Defective products are defined as those with a quality score below 75, exhibiting significant defects or functional issues. This labeling process is based on a comprehensive assessment of the quality inspection department's inspection records and customer feedback, resulting in a historical quality dataset containing product ID, production date, quality score, and quality grade. This historical quality dataset is then matched with product differentiation points and customer order decoupling point location data for the corresponding production batches. This matching process uses product ID and production batch number as key keys to link quality data with production process data. Each record contains fields such as product ID, quality grade, PDP location number, and CODP location number, generating location-quality association data. This dataset clearly presents the distribution of product quality under different decoupling point location combinations, providing a data basis for analyzing the relationship between quality and decoupling point location.

[0098] Perform statistical analysis on the position-quality correlation data and calculate the distribution ratio of each quality level at different decoupling point locations. The distribution ratio calculation formula is as follows:

[0099]

[0100] Among them, R qd Indicates the proportion of products with quality grade q under the decoupling point position combination d; N qd represents the number of products with quality grade q under the decoupling point position combination d; Q represents the total number of quality grades. In this solution, Q = 3, corresponding to the three grades of high quality, qualified, and defective. Represents the total quantity of products of all quality grades under the decoupling point location combination d.

[0101] Furthermore, in order to more comprehensively analyze the impact of the decoupling point location on quality, a quality weighted score is introduced:

[0102]

[0103] Among them, WS d represents the quality weighted score of the decoupling point position combination d; W qThe weight coefficient represents the quality grade q, with weights for high-quality, acceptable, and defective products set to 1.0, 0.6, and 0.2, respectively. By calculating the quality grade distribution ratio and quality weighted score for each decoupling point location combination, a location-quality distribution table is generated, providing a visual basis for quality risk assessment. This table is used to identify stable and fluctuating quality areas. Stable quality areas are regions with a high proportion of high-quality products and minimal fluctuation, typically characterized by a high-quality product ratio exceeding 85% for several consecutive decoupling point locations. Fluctuating quality areas are regions with significant variation in product quality grade distribution, typically characterized by a difference in quality weighted scores exceeding 20% between adjacent decoupling point locations. By setting risk level assessment criteria, all decoupling point location combinations are classified into three risk levels: low, medium, and high. Low-risk areas typically have a high-quality product ratio exceeding 85% and a defective ratio below 5%. High-risk areas typically have a defective ratio exceeding 15%. This generates decoupling point quality risk assessment data, providing a basis for subsequent parameter optimization decisions.

[0104] Based on the decoupling point quality risk assessment data, production parameter configuration records for each decoupling point are extracted. These parameter configuration records include the actual set values for key process parameters such as cutting speed, cutting angle, and molding temperature. By screening the parameter configurations corresponding to decoupling points located in low-risk areas, a parameter configuration-quality correspondence table is generated. This table contains key information such as decoupling point location combinations, quality levels, and parameter configurations, providing data support for parameter optimization.

[0105] Correlation analysis was performed between each parameter in the parameter configuration-quality correspondence table and the quality score, and the correlation coefficient and significance level were calculated. The correlation coefficient was calculated using the Pearson correlation coefficient, and the significance level was determined using a t-test. Parameters were ranked from highest to lowest based on the absolute value of the correlation coefficient to form a parameter importance ranking table, clarifying the degree of impact of each parameter on product quality. This ranking helps production managers identify key quality factors and provides guidance for process parameter optimization.

[0106] Parameters are ranked by importance and selected as key parameters with absolute correlation coefficients greater than 0.5 and significance levels less than 0.05. Data fitting is performed to establish numerical relationships between these key parameters and quality scores. The fitting method is selected based on the data distribution characteristics and can include linear regression, polynomial regression, or support vector regression. The fitting results form a parameter-quality mapping table, which clearly demonstrates the quantitative relationship between each parameter value and the quality score.

[0107] Finally, the parameter-quality mapping table is integrated into a matrix to construct a parameter-quality mapping relationship matrix. This matrix has parameters as rows and quality grades as columns, with matrix elements representing parameter value ranges or parameter influence weights. This matrix allows production managers to intuitively understand the impact of each parameter on product quality, providing a direct basis for parameter optimization decisions. The calculation of matrix elements takes into account the parameter influence weights and variation ranges, comprehensively reflecting the sensitivity of parameter changes to quality. This forms a quality prediction indicator, providing important input for subsequent multi-objective optimization.

[0108] For example, this method was applied to optimize production processes at an aluminum ceiling manufacturer. The company extracted nearly 12 months of product quality rating data from its production management system, encompassing over 3,000 records. By labeling these records with quality grades, a historical quality dataset was generated, containing 1,560 high-quality products, 1,120 qualified products, and 320 defective products. This data was then matched with the corresponding PDP and CODP position data to generate position-quality correlation data. Statistical analysis revealed that when the PDP was positioned after basic cutting and the CODP after aluminum sheet pretreatment, the proportion of high-quality products reached 92%, while the proportion of defective products was only 2%. The weighted quality score was 0.93, placing the company in the low-risk range for stable quality. However, when the PDP was moved to the post-special forming stage and the CODP position remained unchanged, the proportion of high-quality products dropped to 75%, the proportion of defective products increased to 10%, and the weighted quality score dropped to 0.81, placing the company in the medium-risk range for quality fluctuation. By extracting parameter configurations from low-risk areas, we found that the correlation coefficients between cutting angle accuracy, molding temperature stability, and assembly accuracy and quality scores were 0.82, 0.76, and 0.69, respectively. These correlations were all strong and significant at a level less than 0.01. Polynomial regression fitting was performed on these parameters to establish a parameter-quality mapping relationship, ultimately forming a quality prediction indicator matrix. This provided data support for the company's parameter optimization decisions, achieving quality improvement and cost control in aluminum ceiling production.

[0109] In a specific embodiment, the process of executing step S105 may specifically include the following steps:

[0110] (1) Convert the quality prediction index into a numerical quality score index, determine the quality score target value, and obtain the quality objective function;

[0111] (2) Extract production cost data from the enterprise resource management system, including material cost, energy cost, labor cost, and equipment depreciation cost, to obtain a cost data set;

[0112] (3) Establish a corresponding relationship between the cost data set and the production parameters, construct a parameter-cost function, and obtain the cost target function;

[0113] (4) Extract delivery time data from historical orders, analyze the production cycle under different production parameter configurations, and obtain the time objective function;

[0114] (5) Linearly combine the quality objective function, cost objective function, and time objective function to construct a multi-objective weight function and obtain a comprehensive evaluation function;

[0115] (6) Divide the feasible domain of production parameters according to the comprehensive evaluation function, determine the search space boundary, and obtain the parameter optimization space;

[0116] (7) Traverse the parameter combinations in the parameter optimization space, calculate the comprehensive evaluation score of each group of parameters, and obtain a score ranking table;

[0117] (8) Select the parameter combination with the highest comprehensive score from the score ranking table to form the optimal parameter configuration plan and obtain the production parameter configuration plan.

[0118] Specifically, the quality prediction index is converted into a numerical quality score index. This conversion process is achieved by constructing a quality mapping function, which quantifies the qualitative relationship in the parameter-quality mapping relationship matrix into a clear numerical index. The quality score index calculation formula is as follows:

[0119]

[0120] Where QF(V) represents the quality score of parameter combination V; V = (v1, v2, ..., v A ) represents the production parameter combination vector; A represents the total number of parameters; K represents the number of key parameters; β k represents the weight coefficient of the key parameter k; f k (v k ) represents the parameter v k Individual contribution functions to mass; γ ab represents the interaction weight of parameter a and parameter b; h ab (v a ,v b ) represents the parameter v a and v b The interaction function.

[0121] This quality score indicator comprehensively considers the independent influence of a single parameter and the interaction between parameters, reflecting the complex relationship between parameter configuration and product quality in aluminum ceiling production. According to the company's quality goals and market demand, the quality score target value QF is set. target , construct the quality objective function:

[0122]

[0123] Quality objective function G q(V) represents the ratio of the actual quality score to the target value. The larger the value, the closer the quality is to or exceeds the target requirement.

[0124] Production cost data, including material costs, energy costs, labor costs, and equipment depreciation, is extracted from the enterprise resource planning system. Material costs primarily include aluminum sheet raw material costs, auxiliary material costs, and scrap recycling costs; energy costs include electricity, gas, and other energy consumption costs; labor costs include direct and indirect labor costs; and equipment depreciation is calculated based on equipment usage time and value. This cost data is aggregated by batch or product model to form a structured cost data set.

[0125] Based on the cost data set, a corresponding relationship is established with the production parameters to construct a parameter-cost function. Different production parameter configurations will directly affect material utilization, energy consumption, and production efficiency, which in turn affects the total cost. Through regression analysis methods, a functional relationship between parameters and various costs is established to obtain the cost target function:

[0126]

[0127] Among them, G c (V) represents the cost objective function value; C(V) represents the total production cost under the parameter combination V; C min Indicates the lowest cost level in history. The higher the cost objective function value, the better the cost control effect.

[0128] Extract delivery time data from historical orders and analyze the production cycle under different production parameter configurations. Through statistical analysis, establish the corresponding relationship between parameter configuration and production cycle, and obtain the time target function:

[0129]

[0130] Among them, G t (V) represents the time objective function value; T(V) represents the production cycle time under the parameter combination V; T min Indicates the shortest production cycle in history. The higher the time objective function value, the higher the production efficiency.

[0131] The quality objective function, cost objective function and time objective function are linearly combined to construct a multi-objective weight function and obtain a comprehensive evaluation function:

[0132] F(V)=α q ·G q (V)+α c ·G c (V)+α t ·G t (V)

[0133] Among them, F(V) represents the comprehensive evaluation score of the parameter combination V; α q , α c and α t Respectively represent the weight coefficients of quality, cost and time in the comprehensive evaluation. The setting of weight coefficients reflects the strategic focus of the enterprise. Quality-oriented enterprises usually set a higher α q value, while companies in highly competitive markets may pay more attention to cost and delivery time, and increase α accordingly. c and α t The value of .

[0134] The feasible domain of production parameters is divided based on a comprehensive evaluation function to determine the search space boundaries. The determination of the feasible domain requires consideration of multiple factors, including equipment technical limitations, production safety requirements, and product quality standards. For example, the cutting speed parameter may be limited by the equipment's maximum cutting capacity, while the molding temperature parameter must consider material properties and production safety requirements. By setting upper and lower limits for each parameter, a multidimensional parameter space is constructed, forming a search space for parameter optimization. Parameter combinations are traversed within the parameter optimization space, and a comprehensive evaluation score is calculated for each parameter group. Given the large number of parameters involved in aluminum ceiling production, a complete traversal of all possible parameter combinations would be computationally prohibitive. Therefore, a grid search method is employed for efficient traversal. The grid search method divides the range of each parameter into a number of discrete points and then calculates the comprehensive evaluation scores for these discrete point combinations. For example, if there are five key parameters and 10 discrete points are selected for each parameter, the evaluation scores for a total of 10^5 parameter combinations must be calculated. This calculation yields a comprehensive evaluation score for each parameter combination, which is then used to form a ranking table.

[0135] The parameter combination with the highest overall score is selected from the ranking table to form the optimal parameter configuration solution. To enhance the robustness of the solution, not only is the single highest-scoring combination selected, but multiple alternative solutions with similar scores are also examined to analyze their sensitivity to environmental changes and process fluctuations. Ultimately, the parameter combination with the highest score and good stability is selected as the production parameter configuration solution.

[0136] In a specific embodiment, the process of executing step S106 may specifically include the following steps:

[0137] (1) Convert the production parameter configuration plan into a target production execution plan, distribute it to each process operation station, and obtain the production operation instructions;

[0138] (2) Carry out aluminum ceiling production operations according to the production operation instructions, and start the whole process monitoring system at the same time to obtain real-time production parameter data;

[0139] (3) Compare the real-time production parameter data with the production parameter configuration plan, trigger an early warning signal when the parameter deviates from the set range, and obtain a parameter deviation record;

[0140] (4) Test the finished aluminum ceiling products according to the quality inspection standards, measure the values of various quality indicators, and obtain the actual quality data;

[0141] (5) Compare and analyze the actual quality data with the quality evaluation index, calculate the prediction accuracy and deviation direction, and obtain the quality deviation data;

[0142] (6) Conduct correlation analysis on parameter deviation records and quality deviation data, explore the influence of parameter changes on quality, and obtain parameter-quality relationship correction data;

[0143] (7) Based on the parameter-quality relationship correction data, the weight coefficients and influencing factors in the parameter-quality mapping relationship matrix are updated to obtain the updated quality prediction rules;

[0144] (8) The updated quality prediction rules are fed back to the data acquisition system and integrated with the original data to obtain the updated production parameter optimization database, completing the data closed-loop update process.

[0145] Specifically, the production parameter configuration plan is converted into a detailed production execution plan. This is accomplished through two key steps: parameter decomposition and process mapping. Parameter decomposition involves breaking down the optimized parameter configuration plan into specific operating parameters for each process. For example, cutting angle accuracy can be broken down into specific parameters such as the cutting equipment's feed rate, angle compensation, and clamping pressure. Process mapping then maps these refined parameters to specific processes and equipment on the production line. The converted execution plan is distributed to each process operator via the enterprise's digital work instruction system, generating a production operation manual containing operational steps, parameter settings, and key quality control points. The production operation manual uses a combination of graphics and text to clearly present key parameter settings and precautions for each process, facilitating accurate execution. Aluminum ceiling production operations are executed according to the production operation manual, while a full-process monitoring system is activated. The monitoring system uses a sensor network to collect real-time production parameter data, including speed, angle, and temperature for cutting equipment; pressure, temperature, and speed for forming equipment; accuracy, duration, and torque for assembly equipment; and spray thickness, drying temperature, and time for surface treatment equipment. These sensors continuously collect data at a preset sampling frequency (typically 1-10 times per second) and transmit it to a data processing unit via the industrial control network, forming a real-time production parameter data stream. This data stream undergoes preliminary filtering and formatting to remove obvious communication interference and outliers, ensuring basic data accuracy.

[0146] Real-time production parameter data is compared against the production parameter configuration plan in real time, triggering an early warning signal when a parameter deviates from the set range. This comparison utilizes the parameter control limit method, setting upper and lower control limits for each key parameter. When the actual parameter value exceeds the limit, an early warning signal of the corresponding level is triggered. Early warning levels are categorized into three levels: minor deviations result in a prompt message, moderate deviations lead to adjustment suggestions, and severe deviations result in a shutdown command. For example, if the cutting temperature setting is 350°C, with upper and lower control limits of 370°C and 330°C, respectively, if the actual monitored value reaches 375°C, a level 2 early warning is triggered, prompting a temperature adjustment suggestion. If the monitored value exceeds 400°C, a level 3 early warning is triggered, prompting a shutdown command to prevent accidents. All parameter deviation events and their handling process are recorded in a structured log, containing fields such as parameter name, standard value, actual value, deviation value, time of occurrence, duration, and action taken. Completed aluminum ceiling products are inspected according to quality inspection standards, measuring various quality indicators. Quality inspection is conducted in both online and offline stages. Online testing uses on-line testing equipment to monitor product dimensions, appearance, connection strength, and other parameters in real time. Offline testing, after the product is finished, uses specialized testing equipment to measure more comprehensive quality indicators, including dimensional accuracy, surface flatness, coating adhesion, corrosion resistance, and other indicators. Test data is entered and organized into actual quality data, including product ID, test time, and the values of various quality indicators.

[0147] Actual quality data is compared and analyzed against quality prediction indicators to calculate prediction accuracy and deviation direction. Prediction accuracy is calculated as the percentage deviation between the actual and predicted quality scores, while the deviation direction indicates whether the actual value is higher or lower than the predicted value. This comparative analysis not only focuses on the overall quality score but also breaks down into specific quality indicators, such as the actual versus predicted surface flatness value and the difference between the actual and expected coating uniformity performance. By calculating the prediction accuracy and overall accuracy of each quality indicator, quality deviation data is generated, providing a foundation for subsequent analysis.

[0148] Correlation analysis is performed on parameter deviation records and quality deviation data to uncover patterns in the impact of parameter changes on quality. Correlation analysis uses a combination of time series correlation and parameter sensitivity analysis. Time series correlation matches the time points of parameter deviation with the time points of quality change to identify parameter-quality pairs that may have a causal relationship. Parameter sensitivity analysis determines the weight of each parameter's impact on quality by calculating the responsiveness of quality indicators to parameter changes. For example, analysis may reveal a high correlation between cutting angle deviation and product surface flatness, with the flatness score decreasing by approximately 2 points for every 0.1-degree increase in angle deviation. Conversely, the correlation between molding temperature fluctuation and coating uniformity is weak, with temperature fluctuations within 5°C having little effect on coating uniformity. Through these analyses, parameter-quality relationship correction data is generated to clarify the actual impact of each parameter on quality.

[0149] Based on the parameter-quality relationship correction data, the weight coefficients and influencing factors in the parameter-quality mapping relationship matrix are updated. This update process utilizes an incremental learning approach, weighting and integrating newly discovered relationship patterns with the original mapping relationship, preserving historical experience while incorporating new discoveries. The updated weight coefficients are based on a weighted average of the original and newly calculated values. Weights are set based on data volume and reliability, typically with new data receiving a weight between 30% and 50%. The updated mapping relationship incorporates more accurate parameter-quality correspondence rules, forming updated quality prediction rules that provide a more precise basis for subsequent parameter optimization.

[0150] The updated quality prediction rules are fed back into the data acquisition system and integrated with the original data to form an updated production parameter optimization database. This integration process isn't a simple data overlay; rather, it utilizes version control and data fusion technologies to ensure data consistency and integrity. The updated database includes both parameter-quality mappings and retains original production records and quality data, forming a complete data closed loop. This closed-loop update mechanism enables continuous improvement in aluminum ceiling production optimization methods, and parameter configuration solutions become increasingly precise as production practice accumulates.

[0151] The above describes the aluminum ceiling production optimization method based on data mining in the embodiment of the present application. The following describes the aluminum ceiling production optimization system based on data mining in the embodiment of the present application. Figure 2 In the embodiment of the present application, an embodiment of the aluminum ceiling production optimization system based on data mining includes:

[0152] The processing module is used to collect the cutting parameters, forming parameters, assembly parameters and surface treatment parameters of the aluminum ceiling production line and pre-process them to obtain a standardized production data set;

[0153] The calculation module is used to calculate the feature weights of the standardized production data set, screen the key parameters affecting the quality of the aluminum ceiling according to the information gain principle, and obtain the key feature data set;

[0154] The division module is used to divide the production stages according to the key feature data set, calculate the cost curve and time curve of each production stage, and obtain the location data of product differentiation points and customer order decoupling points;

[0155] The correlation module is used to correlate the location data of product differentiation points and customer order decoupling points with historical quality score data, construct a parameter-quality mapping relationship matrix, and obtain quality prediction indicators;

[0156] A generation module is used to establish a multi-objective weight function based on the quality prediction index and production cost data, generate an optimal solution set of parameter configuration combinations, and obtain a production parameter configuration plan;

[0157] The updating module is used to execute the production operation according to the production parameter configuration scheme, record the deviation data between the actual production index and the quality evaluation index, and perform data closed-loop update through the deviation data.

[0158] Through the collaborative efforts of these components, precision sensors deployed throughout the aluminum ceiling production line collect cutting, forming, assembly, and surface treatment parameters. This approach, combined with an extended Kalman filter for outlier filtering, a multivariate interpolation algorithm for missing value filling, information entropy calculation for redundant data removal, and a Z-Score standardization method for data dimensionality unification, creates a high-quality, standardized production dataset, laying a solid foundation for subsequent analysis. Next, feature weight calculation and information gain analysis are performed on the standardized data to identify key features such as aluminum thickness deviation rate, cutting angle accuracy, and forming temperature fluctuation range, significantly reducing data dimensionality and improving computational efficiency. Within the production stage segmentation phase, this solution innovatively combines key feature data to calculate cost and time curves, scientifically identifying product differentiation points and customer order decoupling points, achieving optimal allocation of production resources and improving customer responsiveness. By correlating decoupling point location data with historical quality scores, a parameter-quality mapping matrix is constructed, providing a scientific basis for quality prediction. A comprehensive evaluation system based on a multi-objective weighting function simultaneously addresses the three objectives of quality, cost, and time, generating an optimal parameter configuration that balances multidimensional production goals. It is particularly worth emphasizing that this solution utilizes artificial intelligence technologies such as the convolutional neural network-long short-term memory artificial neural network algorithm (C-LSTM) and the support vector machine-red deer optimization algorithm (SVM-RDO). This fully considers the specific needs of the aluminum ceiling production field and deeply integrates these algorithm features with specific production scenarios. For example, the C-LSTM algorithm can effectively capture long-term dependencies when processing production time series data, providing accurate predictions for optimizing decoupling point locations. The SVM-RDO algorithm, through adaptive parameter adjustment, effectively addresses problems such as overfitting and local minima in aluminum ceiling production, significantly improving the accuracy of parameter optimization. Furthermore, the data closed-loop update mechanism established in this solution achieves continuous optimization of production parameters through real-time monitoring, quality inspection, deviation analysis, and parameter correction. This allows the system to continuously improve itself as data accumulates, effectively overcoming the limitations of traditional static configuration methods and ensuring the stability of the production process and the consistency of product quality.

[0159] Reference Figure 3 In an embodiment of the present invention, a computer device is also provided. The computer device may be a server, and its internal structure may be as follows: Figure 3As shown. The computer device includes a processor, memory, display screen, input device, network interface and database connected via a system bus. The processor of the computer design is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The database of the computer device is used to store the corresponding data in this embodiment. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, the above method is implemented.

[0160] Those skilled in the art will understand that Figure 3 The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present invention and does not constitute a limitation on the computer device to which the solution of the present invention is applied.

[0161] An embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, which implements the above-described method when executed by a processor. It is understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0162] Those skilled in the art will appreciate that all or part of the processes in the above-described method embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the above-described method embodiments. Among them, any reference to memory, storage, database, or other media provided by the present invention and used in the embodiments can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double-speed SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct RAM bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM.

[0163] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0164] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), magnetic disk or optical disk, etc., various media that can store program code.

[0165] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for optimizing aluminum ceiling production based on data mining, characterized in that: The aluminum ceiling production optimization method based on data mining includes: Collect cutting parameters, forming parameters, assembly parameters and surface treatment parameters of the aluminum ceiling production line and pre-process them to obtain a standardized production data set; Calculating feature weights on the standardized production data set, screening key parameters affecting the quality of the aluminum ceiling according to the information gain principle, and obtaining a key feature data set; Divide the production stages according to the key feature data set, calculate the cost curve and time curve of each production stage, and obtain the location data of the product differentiation point and the customer order decoupling point; Correlation analysis is performed on the product differentiation point and customer order decoupling point location data with historical quality score data to construct a parameter-quality mapping relationship matrix to obtain quality prediction indicators; Based on the quality prediction index and production cost data, a multi-objective weight function is established to generate an optimal solution set of parameter configuration combinations and obtain a production parameter configuration plan; Execute production operations according to the production parameter configuration plan, record deviation data between actual production indicators and quality evaluation indicators, and perform data closed-loop update based on the deviation data.

2. The aluminum ceiling production optimization method based on data mining according to claim 1 is characterized in that: The cutting parameters, forming parameters, assembly parameters and surface treatment parameters of the aluminum ceiling production line are collected and pre-processed to obtain a standardized production data set, including: Sensors are set up in the cutting process of the aluminum ceiling production line to collect cutting speed, cutting angle, cutting temperature, and equipment vibration frequency parameters to obtain original cutting parameter data; sensors are set up in the aluminum ceiling forming process to collect forming pressure, forming temperature, and forming speed parameters to obtain original forming parameter data; During the aluminum ceiling assembly process, sensors are set to collect assembly accuracy, assembly time, and tightening torque parameters to obtain original assembly parameter data; during the aluminum ceiling surface treatment process, sensors are set to collect spray thickness, drying temperature, and drying time parameters to obtain original surface treatment parameter data; Performing outlier filtering on the original cutting parameter data, the original molding parameter data, the original assembly parameter data, and the original surface treatment parameter data through an extended Kalman filter to obtain filtered data; Filling missing values with a multivariate interpolation algorithm based on the filtered data to obtain complete data, performing information entropy calculation on the complete data, removing redundant data whose information entropy contribution rate is lower than a set threshold, and obtaining streamlined data; The simplified data were converted to a standard normal distribution interval using a Z-Score standardization method to obtain a standardized production data set.

3. The aluminum ceiling production optimization method based on data mining according to claim 2 is characterized in that: The feature weight calculation is performed on the standardized production data set, and the key parameters affecting the quality of the aluminum ceiling are screened according to the information gain principle to obtain a key feature data set, including: Performing preliminary feature screening on the cutting speed, cutting angle, cutting temperature, and equipment vibration frequency parameters in the standardized production data set to obtain candidate features for the cutting process; Performing preliminary feature screening on the molding pressure, molding temperature, and molding speed parameters in the standardized production data set to obtain candidate features of the molding process; Performing preliminary feature screening on assembly accuracy, assembly time, and tightening torque parameters in the standardized production data set to obtain candidate features for the assembly process; Performing preliminary feature screening on the spraying thickness, drying temperature, and drying time parameters in the standardized production data set to obtain candidate features for the surface treatment process; Integrate the candidate features of the cutting stage, the candidate features of the forming stage, the candidate features of the assembly stage, and the candidate features of the surface treatment stage to form a feature pool and obtain a feature selection scheme; Calculating information gain for each feature in the feature pool based on the feature selection scheme, sorting the features by importance, and obtaining a feature importance sequence; According to the feature importance sequence, the aluminum material thickness deviation rate, cutting angle accuracy, forming temperature fluctuation range, assembly gap parameter, and surface treatment uniformity index are screened to obtain a key index set; The mutual relationship between the parameters in the key indicator set is analyzed, a feature association network is constructed, and the key indicators are combined with corresponding weight values to obtain a key feature data set.

4. The aluminum ceiling production optimization method based on data mining according to claim 1 is characterized in that: The production stages are divided according to the key feature data set, the cost curve and time curve of each production stage are calculated, and the location data of the product differentiation point and the customer order decoupling point are obtained, including: The aluminum ceiling production process is divided into stages based on the aluminum thickness deviation rate, cutting angle accuracy, forming temperature fluctuation range, assembly gap parameters, and surface treatment uniformity indicators in the key feature data set, and a general production stage and a special production stage are obtained; Measuring the time of each process node of the general production stage, drawing a process time distribution diagram, and obtaining a general stage time curve; measuring the time of each process node of the special production stage, drawing a process time distribution diagram, and obtaining a special stage time curve; Based on the general stage time curve and the special stage time curve, the unit time cost of each process is calculated to obtain the cost function of each stage; Combining the cost functions of each stage with the product delivery time requirements, a time-cost comprehensive evaluation model is constructed to obtain comprehensive cost data at different decoupling points; Analyze the comprehensive cost data to find the cost transition point, use the inflection point of the cost curve as a candidate decoupling point, and obtain a candidate set of decoupling points; Calculate the customer satisfaction scores corresponding to different decoupling point positions based on the decoupling point candidate set and customer order history data to obtain a decoupling point evaluation result; The decoupling point evaluation results and cost data are comprehensively analyzed to determine the optimal positions of the product differentiation point and the customer order decoupling point, and obtain the product differentiation point and customer order decoupling point position data.

5. The aluminum ceiling production optimization method based on data mining according to claim 1 is characterized in that: The correlation analysis of the product differentiation point and customer order decoupling point location data with historical quality score data is performed to construct a parameter-quality mapping relationship matrix to obtain quality prediction indicators, including: The quality score data of historical aluminum ceiling products were extracted from the enterprise production management system and labeled into three categories: high-quality, qualified, and defective, to obtain a historical quality data set. Matching the historical quality data set with the product differentiation points and customer order decoupling point location data of the corresponding production batches to obtain location-quality correlation data; Performing statistical analysis on the position-quality correlation data, calculating the distribution ratio of each quality level at different decoupling point positions, and obtaining a position-quality distribution table; Identify quality stable areas and quality fluctuating areas according to the position-quality distribution table, divide the quality risk levels, and obtain decoupling point quality risk assessment data; Based on the decoupling point quality risk assessment data, extract the production parameter configuration records at each decoupling point location to obtain a parameter configuration-quality correspondence table; Perform correlation analysis on each parameter in the parameter configuration-quality correspondence table and the quality score, calculate the correlation coefficient and significance level, and obtain the parameter importance ranking; According to the parameter importance ranking, data fitting is performed on the correspondence between key parameters and quality scores, a numerical correspondence is established, and a parameter-quality mapping table is obtained; The parameter-quality mapping table is integrated into a matrix form, which includes the influence weight and variation range of each parameter combination on the quality, to obtain a quality prediction index.

6. The aluminum ceiling production optimization method based on data mining according to claim 1 is characterized in that: The method of establishing a multi-objective weight function based on the quality prediction index and production cost data, generating an optimal solution set of parameter configuration combinations, and obtaining a production parameter configuration scheme includes: Converting the quality prediction index into a numerical quality score index, and determining a quality score target value to obtain a quality objective function; Extract production cost data from the enterprise resource management system, including material cost, energy cost, labor cost and equipment depreciation cost, to obtain a cost data set; Establishing a corresponding relationship between the cost data set and the production parameters, constructing a parameter-cost function, and obtaining a cost target function; Extract delivery time data from historical orders, analyze the production cycle under different production parameter configurations, and obtain the time target function; The quality objective function, the cost objective function and the time objective function are linearly combined to construct a multi-objective weight function to obtain a comprehensive evaluation function; Dividing the feasible region of the production parameters according to the comprehensive evaluation function, determining the search space boundary, and obtaining the parameter optimization space; Perform parameter combination traversal in the parameter optimization space, calculate the comprehensive evaluation score of each group of parameters, and obtain a score ranking table; The parameter combination with the highest comprehensive score is selected from the score ranking table to form an optimal parameter configuration scheme, thereby obtaining a production parameter configuration scheme.

7. The aluminum ceiling production optimization method based on data mining according to claim 1 is characterized in that: The step of executing production operations according to the production parameter configuration plan, recording deviation data between actual production indicators and quality evaluation indicators, and performing data closed-loop updating based on the deviation data includes: Convert the production parameter configuration plan into a target production execution plan, distribute it to each process operation station, and obtain a production operation instruction; Perform aluminum ceiling production operations according to the production operation instructions, and simultaneously start the full-process monitoring system to obtain real-time production parameter data; Comparing the real-time production parameter data with the production parameter configuration plan, triggering an early warning signal when the parameter deviates from the set range, and obtaining a parameter deviation record; Test the finished aluminum ceiling products according to the quality inspection standards, measure the quality index values and obtain the actual quality data; Comparing and analyzing the actual quality data with the quality evaluation index, calculating the prediction accuracy and deviation direction, and obtaining quality deviation data; Perform correlation analysis on the parameter deviation records and quality deviation data, explore the influence of parameter changes on quality, and obtain parameter-quality relationship correction data; Based on the parameter-quality relationship correction data, the weight coefficients and influencing factors in the parameter-quality mapping relationship matrix are updated to obtain an updated quality prediction rule; The updated quality prediction rules are fed back to the data acquisition system and integrated with the original data to obtain an updated production parameter optimization database, completing the data closed-loop update process.

8. A data mining-based aluminum ceiling production optimization system, used to implement the data mining-based aluminum ceiling production optimization method according to any one of claims 1 to 7, characterized in that: The aluminum ceiling production optimization system based on data mining includes: The processing module is used to collect the cutting parameters, forming parameters, assembly parameters and surface treatment parameters of the aluminum ceiling production line and pre-process them to obtain a standardized production data set; A calculation module is used to calculate feature weights of the standardized production data set, screen key parameters affecting the quality of the aluminum ceiling according to the information gain principle, and obtain a key feature data set; a division module, configured to divide the production stages according to the key feature data set, calculate the cost curve and time curve of each production stage, and obtain the location data of the product differentiation point and the customer order decoupling point; A correlation module is used to perform correlation analysis on the product differentiation point and customer order decoupling point location data with historical quality score data, construct a parameter-quality mapping relationship matrix, and obtain quality prediction indicators; A generation module is used to establish a multi-objective weight function based on the quality prediction index and production cost data, generate an optimal solution set of parameter configuration combinations, and obtain a production parameter configuration plan; The updating module is used to execute the production operation according to the production parameter configuration scheme, record the deviation data between the actual production index and the quality evaluation index, and perform data closed-loop update through the deviation data.

9. A computer device, characterized in that: It comprises a memory and a processor, the memory stores a computer program that can be run on the processor, and is characterized in that when the processor executes the computer program, it implements the aluminum ceiling production optimization method based on data mining as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the processor is enabled to execute the aluminum ceiling production optimization method based on data mining according to any one of claims 1 to 7.

Images