Airborne equipment product quality risk assessment method based on process parameters
By constructing a quality risk assessment model based on support vector machines and combining factors such as machine, material, method, and environment, real-time and accurate assessment of the quality risk of airborne equipment products was achieved. This solved the problem of lack of process monitoring in existing technologies and improved the ability to identify quality risks in the aviation equipment supply chain.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CHINA AERO POLYTECH ESTAB
- Filing Date
- 2025-12-31
- Publication Date
- 2026-05-08
AI Technical Summary
In the existing technology, the quality risk assessment of airborne equipment products lacks process monitoring, resulting in low inspection efficiency, inability to effectively guarantee product quality, and potential impact on delivery and use. Furthermore, there is a lack of risk assessment of the production process.
By constructing a quality risk assessment model based on support vector machines, combining key factors from four aspects—machinery, materials, methods, and environment—process data is collected and preprocessed. Hyperparameters are optimized using particle swarm optimization, a real-time quality risk assessment method is established, key factors are identified and quantified, and quality risk values and levels are output.
It enables comprehensive, real-time, and accurate monitoring of airborne equipment product quality risks, improves the timeliness and accuracy of quality risk identification in the aviation equipment supply chain, and provides more precise assessment results.
Smart Images
Figure CN121998484A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of product quality risk assessment technology, and specifically to a method for assessing the product quality risk of airborne equipment based on process parameters. Background Technology
[0002] Airborne equipment encompasses all instruments, meters, systems, and subsystems installed on aircraft to perform functions such as flight control, navigation, communication, monitoring flight status, ensuring flight safety, improving flight efficiency, performing specific missions, and providing passenger comfort and crew working conditions. This includes instruments, navigation, automatic control, communication, radar, high-altitude medical services, and weapon systems. The development and production of airborne equipment are characterized by high technical complexity, multiple supporting layers, stringent quality requirements, and complex product composition. The complexity of the development and production process and the diversification of products lead to frequent quality problems with airborne equipment, occurring more frequently than with other equipment. Due to their complex structure and independent functions, quality problems can have more serious consequences for the aircraft than quality problems with individual components or standard parts.
[0003] Airborne equipment product quality risk refers to the possibility of quality problems occurring during the incoming inspection or use of equipment products, which are mostly exposed during assembly and use. However, due to the significant differences in product quality risks that may arise during the incoming inspection or use of different aviation equipment products, current supply chain risk control in the aviation field mainly focuses on the operational capabilities, delivery timeliness, and financial status of the product itself. The control of airborne equipment product quality is basically limited to the strict inspection of various functional performance, technical parameters, safety, environmental adaptability, appearance quality, and documentation at the time of product delivery to ensure the reliability and safety of airborne equipment use. However, there is a lack of risk assessment and monitoring of the research and development and production process of the equipment products.
[0004] This kind of risk control for airborne equipment products, which is mainly based on result-based inspection, is not comprehensive enough and cannot effectively guarantee the quality of airborne equipment products. At the same time, the parameter inspection of airborne products based on results is inefficient because it is carried out at the time of delivery. This may affect product delivery, cause property damage, or delay the time limit for putting airborne products into use. Summary of the Invention
[0005] To address the shortcomings of the existing technology, the present invention aims to provide a method for assessing the quality risk of airborne equipment products based on process parameters. By considering the research and development process control and process inspection and testing results of the product, a scientific quality risk assessment and evaluation method is constructed to conduct real-time, systematic and effective assessment of the quality risk of airborne equipment, thereby ensuring the stability of the aviation equipment supply chain and the reliability of product quality.
[0006] Specifically, the present invention provides a method for assessing the quality risk of airborne equipment products based on process parameters, which includes the following steps: S1. Based on historical testing data, screen out typical quality problems of airborne equipment products and identify key factors affecting the occurrence of quality problems; S2. Based on key factors, construct a product quality risk assessment index system for airborne equipment; S3. Based on the risk assessment index system, construct an airborne product quality risk assessment model and output the quality risk value, including the following sub-steps: S31. Based on the risk assessment indicator system, determine that the collected data will be used as the input value of the assessment model, and the quality risk value will be used as the output value. S32. Construct a quality risk assessment model based on support vector machines, selecting the kernel function and model parameters, including: S321. Determine the evaluation model: ; In the formula, for Lagrange multipliers, corresponding for Optimization of Lagrange multipliers For kernel functions; This is the constant term of the decision function; S322. Determine the kernel function: The discriminant function of the support vector machine constructed using the radial basis kernel function takes the following form: ; Where s is the number of centers, and s support vectors can determine the center location of the radial basis function. and Including input samples, This includes the squared Euclidean distance between two samples. This includes parameters that control the scope of influence; S323. Determine model parameters: Determine model parameters based on radial basis function kernel function: penalty parameter C of loss function, radial basis function, and function parameters. Parameters of the insensitive loss function ; S33. Based on the quality risk assessment model, set the data collection scope, collect data on the product development and production process, and preprocess the process data; S34. Optimize the hyperparameters of the support vector machine based on the particle swarm optimization algorithm to find the optimal parameter combination; S35. Use the preprocessed process data as the training set to train the quality risk assessment model. After training, input the test set data and output the quality risk value of each risk dimension of the product. S4. Based on the set quality risk level threshold, the quality of airborne equipment is divided into several risk levels according to the quality risk value.
[0007] Furthermore, step S3 also includes: S36, after the evaluation is completed, the performance level of the quality risk assessment model is evaluated: the mean squared error (MSE) and correlation coefficient are selected. As an evaluation indicator: ; ; The smaller the MSE, the smaller the model's fitting error to the training data; the closer R is to 1, the closer the model's evaluation result is to the actual value. Based on the assessment results, the quality risk assessment model was optimized by adjusting the evaluation indicators.
[0008] Further: Step S2 includes the following sub-steps: S21. The identified key influencing factors are classified into four types: machine, material, method, and environment. Risk factor identification includes: verification and maintenance of tooling fixtures and molds, status of product production equipment, electronic, information-based and standardized product document management system, inspection equipment capability, periodic calibration and inspection capability of measuring tools, and periodic maintenance of equipment; Risk factor identification includes: packaging material quality, whether the packaging has been verified, product material compliance rate, key component quality level, warehouse location management, annual module production failure rate, and material expiration date management. Risk factors identification includes: the number of scrapped items in the manufacturing process, the accuracy and completeness of standard operating instructions, rework rate, the completeness of the functional inspection checklist, the factory inspection process, and the fault handling process. Environmental risk factor identification includes: compliance of product storage environment and compliance of product transportation protection environment; S22. Based on the research and development process and historical data of airborne equipment products, determine the weight of key factors and quantify the key factors, summarize the risks of historical data, appearance defects and functional performance, and construct a quality risk assessment index system for airborne equipment products.
[0009] Furthermore, the quality risk value in step S31 includes traceability risk, appearance defect risk, and functional defect risk, and the quality risk value is measured by the pass rate.
[0010] Further: The method for determining the evaluation model in step S321 is as follows: Support vector regression is expressed as a set of training samples. The goal of regression is to find the optimal regression function. , Introduction As a risk function: ; Introducing the Lagrange coefficient: ; ; in for Lagrange multipliers, corresponding for The optimization of Lagrange multipliers is obtained by solving the optimization problem. and : ; To ensure stability, the average of the support vectors is used to determine the regression equation: ; in For kernel function, This is the constant term of the decision function.
[0011] Furthermore, the preprocessing method in step S34 includes: standardizing the input data of the model by means of normalization, missing value imputation and outlier handling respectively.
[0012] Further: Step S4 includes the following sub-steps: S41: Determine the quality risk threshold for the quality risk type of airborne equipment products. By assessing the quality risk value, compare the assessment accuracy of the batch of airborne equipment products under different thresholds, and select the threshold with the highest accuracy as the quality risk threshold for this type of airborne equipment products. S42. Establish the correlation between quality risk threshold and quality risk level.
[0013] Furthermore: the quality risk value of resume data risk is the resume data qualification rate, which is obtained by: .
[0014] Furthermore: the quality risk value for appearance defects is the appearance pass rate, which is obtained as follows: .
[0015] Furthermore: the quality risk value for functional defect risk is the functional pass rate, which is obtained by: .
[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention provides a method for quality risk assessment of airborne equipment products. Considering the characteristics of the research and development and production process of airborne equipment products, it identifies key quality risks that affect the occurrence of quality problems, establishes a quality risk assessment model that considers process data characteristics, and realizes comprehensive, real-time and accurate monitoring of quality risks in the delivery and use process of airborne equipment products. The assessment dimensions of airborne equipment products are more comprehensive, and it takes into account the risk factors in the generation process that are missed by traditional result-oriented quality risk assessments, thus providing more accurate assessment results.
[0017] This invention extracts the characteristics of typical quality problems in airborne equipment products, constructs an indicator system representing typical quality risks based on machinery, materials, methods, and environment, and quantifies the indicators. It establishes a quality risk assessment model for airborne equipment products based on the support vector machine algorithm, collects historical data for model training and optimization, assesses the risk values of various quality risks in batches of products, and obtains the risk levels of various risks in batches of products by establishing risk level judgment criteria. This provides strong technical support for the supply chain quality and risk control of aviation equipment products, and significantly improves the timeliness and accuracy of quality risk identification in the overall aviation equipment supply chain. Attached Figure Description
[0018] Figure 1 This is a flowchart of the process for optimizing the hyperparameters of a support vector machine based on the particle swarm optimization algorithm in the risk assessment method disclosed in this invention. Figure 2 This is a trend diagram of MSE change based on the PSO optimization process disclosed in this invention; Figure 3 This is a comparison chart of the actual and assessed risk assessment values of airborne equipment products disclosed in this invention; Figure 4 This is a scatter plot showing the change of risk assessment error of airborne equipment products over time, as disclosed in this invention. Detailed Implementation
[0019] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings.
[0020] Example 1 A method for identifying and assessing quality risks in airborne equipment products includes the following steps: S1: Based on historical testing data, screen out typical quality problems of airborne equipment products and identify key factors affecting the occurrence of quality problems; S2. Based on key factors, construct a product quality risk assessment index system for airborne equipment; S3. Based on the risk assessment index system, construct an airborne product quality risk assessment model and output the quality risk value; S4. Based on the set quality risk level threshold, the quality of airborne equipment is divided into several risk levels according to the quality risk value.
[0021] Furthermore, we collect fault data on airborne equipment products from OEMs, analyze the distribution of quality problems in airborne equipment products, and identify typical quality risks: Typical quality risks for airborne equipment products include risks related to documentation, appearance defects, and functional defects. Risks related to resume materials include incomplete, missing, or version-incorrect resume materials. Appearance defects include missing product markings, out-of-tolerance dimensions, expired product aging, foreign matter, and hole problems. Functional performance risks include poor product contact, loose or detached parts, alarm malfunctions, leaks, and software functionality issues.
[0022] Furthermore, the quality risk assessment index system in step S2 includes four aspects: machine, material, method, and environment.
[0023] S21. Based on typical quality risks of airborne equipment, identify key factors affecting quality risks in the airborne equipment manufacturing process from four aspects: machinery, materials, methods, and environment. S22. Based on the actual characteristics of airborne equipment product development and production, key factors are quantified to form a quality risk assessment index system.
[0024] Furthermore, the airborne equipment product quality risk assessment model in S3 includes an assessment model based on the Support Vector Machine (SVR) algorithm, which mainly includes the following steps: S31. Determine the inputs and outputs of the risk assessment model; Based on the risk assessment indicator system, the collected data will be used as the input value of the assessment model, and the quality risk value will be used as the output value. S32. Based on the quality risk characteristics of airborne equipment products, construct a quality risk assessment theoretical model based on support vector machines, selecting kernel functions and model parameters. The specific steps are as follows: S321. Determine the evaluation model: SVR can be defined as: given training samples The goal of regression is to find the optimal regression function. Introduction As a risk function, i.e., minimization ; Further introduction of the Lagrange coefficient ; ; in the formula for Lagrange multipliers, corresponding for The optimal Lagrange multiplier is only when Time corresponding These are support vectors. Solving the optimization problem yields... and After that, it can be obtained ; For stability considerations, the solution for b uses the average value of the support vectors to ultimately determine the regression equation. ; in the formula For kernel functions; This is the constant term of the decision function.
[0025] S322. Determine the kernel function: Radial basis kernel function (RBF): ; in: and Including input samples, This includes the squared Euclidean distance between two samples. This includes parameters that control the range of influence; the larger the parameter, the smoother the curve.
[0026] The discriminant function of a support vector machine constructed using a radial basis function kernel is as follows: ; Here, s support vectors determine the center location of the radial basis function, and s is the number of centers.
[0027] S323. Determine the model parameters: After determining the RBF radial basis function as the kernel function for evaluating the model, three parameters affect the model: the penalty parameter C of the loss function, and the parameters of the RBF kernel function. That is, function The values of the parameters, and the parameters of the insensitive loss function. .
[0028] S33. Based on the theoretical model, a data collection template for airborne equipment product quality risk is set up to collect data on the product development and production process. The collected process data is preprocessed to eliminate differences between different units of measurement, ensuring all input variables are calculated on the same scale and guaranteeing the stability of the training process. Normalization, missing value imputation, and outlier handling are used to standardize the model's input data. Normalization: Map all input data to the interval [0,1] so that the influence between different features is not unbalanced due to scale differences.
[0029] Missing value imputation: For missing data, imputation is performed using interpolation or other methods to ensure data integrity.
[0030] Outlier handling: By detecting and handling outliers, we can prevent them from negatively impacting model training.
[0031] Furthermore, to ensure the data conforms to the LibSVM data format in Python, it needs to be converted to the LibSVM-required format after data preprocessing. The LibSVM format is typically represented as a sparse matrix, where each row represents a sample data point, in the format: label feature1:value1 feature2:value2 ...featureN:valueN. The label is the target value, the output of the support vector regression machine, which in this project is the value of a certain quality risk of the product; the index is the ordered index, including the feature number, which must be arranged in ascending order; the value is the feature value, the data used for training, usually composed of a set of real numbers. During actual model training, the txt text data conforming to the LibSVM data format needs to be converted into CSV table data. Therefore, data collection can be directly performed in tabular form. For example, when the model's output variable, i.e., the quality risk value, is determined... Input variables, i.e., risk factors corresponding to quality risk. Next, we will collect sample data over a certain time span in CSV format, based on the actual situation. For example, a sample data table of functional failure risks for a certain batch of airborne equipment products is shown below: Each row in this table represents a sample. For each sample in the training dataset, the data is organized in the above format for use in subsequent model training.
[0032] S34. Select a support vector machine kernel function suitable for the airborne equipment product quality risk assessment model; The particle swarm optimization (PSO) algorithm is used to optimize the hyperparameters of a support vector machine (SVM) in order to find the optimal parameter combination. The PSO algorithm simulates the foraging behavior of bird flocks to find the optimal parameter combination. The main parameters optimized include the penalty factor C and the parameter γ of the γ kernel function. The PSO algorithm can effectively avoid local optima, thereby improving the evaluation accuracy of the model. The preprocessed standardized data is used as the training set and input into the risk assessment model for training. The goal of the training process is to fit the training data by minimizing the insensitive loss function, so that the model can assess the quality risk of each product as accurately as possible and obtain the quality risk value of each batch of products in various dimensions.
[0033] S35. After the model training is completed, input the test set data, evaluate the quality level in the test set data, and finally output the risk value of the product in each risk dimension. S36. After the evaluation is completed, assess the performance level of the airborne equipment product quality risk assessment model and optimize the model by adjusting the evaluation indicators. To comprehensively evaluate the model's performance, mean squared error (MSE) and correlation coefficient were selected. The specific calculation formula for this evaluation indicator is as follows: ; ; Where n includes the number of samples, Including actual values, Including the evaluation value.
[0034] A smaller MSE indicates a smaller fitting error of the model to the training data; a closer R-squared is to 1 indicates a high degree of consistency between the model's evaluation results and the actual values, indicating a stronger evaluation capability. These metrics can be used to measure the performance of the established support vector machine regression model in product quality risk assessment.
[0035] Furthermore, the quality risk threshold in S4 includes a dividing point for whether there is a quality risk after the batch delivery of airborne equipment products. The risk level can be obtained by comparing the quality risk value obtained from the assessment with the risk threshold, which can be divided into the following steps: S41. Determine the quality risk threshold for the type of airborne equipment product quality risk. By assessing the risk value, compare the assessment accuracy of the batch of airborne equipment products under different thresholds, and select the threshold with higher accuracy as the quality risk threshold for this type of airborne equipment product. S42. Establish the correlation between quality risk threshold and quality risk level.
[0036] Preferably, the quality risk value of airborne equipment products is represented by the product pass rate, and the quality risk types include traceability risk, appearance quality risk, and functional performance quality risk. Preferably, the quality risk level includes two levels: risky and no risk, wherein: (1) Risk Level: When the quality risk value of a certain type of quality risk of the batch of airborne equipment products is lower than the risk threshold, since the quality risk value is expressed as the product pass rate, that is, it is lower than the risk threshold an, then the quality level of the batch of airborne equipment products in this type of quality risk is poor, and there are problems with quality management. This type of quality risk is high and requires special attention and intervention measures.
[0037] (2) No risk level: When the quality risk value of a certain type of quality risk of the batch of airborne equipment products is higher than the risk threshold, it indicates that the quality level is in a high range and the quality stability is good. This reflects that the batch of products has a good risk control capability for this type of quality risk and is considered to have no quality risk.
[0038] The following example uses airborne equipment products from a certain OEM: For a batch of airborne equipment products with quality risks, the risk points that cause the quality risk value to fall below the risk threshold are identified based on the permutation feature importance method, and process control is carried out.
[0039] S1: Based on historical testing data, screen out typical quality problems of airborne equipment products and identify key factors affecting the occurrence of quality problems; Data on airborne equipment product failures from OEMs were collected, and the distribution of quality problems was analyzed. Table 1 below shows the data on airborne equipment product quality problems collected from a certain OEM from 2019 to 2023. Table 1 Distribution of Quality Problems in Airborne Equipment Products The original problem classifications were inconsistent in granularity, and there was overlap and inclusion among categories. Taking the most frequent categories, "Fault" and "Functional Performance," as examples, much of their content overlapped. "Fault" was further subdivided into "Function Loss," "Function Failure / Abnormality," and "Loosening / Leakage," totaling 87 subcategories. "Functional Performance" covered 13 categories, including alarm malfunctions, abnormal responses, abnormal displays, leaks, mechanical function issues, and abnormal parameters (time, temperature, pressure, speed), all of which overlapped with the 87 subcategories. Furthermore, "Non-Functional Performance" included various specific issues such as historical data, surface quality, assembly dimensions, aging issues, and labeling problems. Therefore, the quality problems in Table 1 were re-statistically analyzed and independently classified to ensure consistent granularity and non-overlapping categories, and presented in descending order of frequency, as detailed in Table 2.
[0040] Table 2. Detailed List of Quality Issues in Airborne Equipment Products The quality risk assessment index system in step S2 includes four aspects: machine, material, method, and environment.
[0041] S21. Based on typical quality risks of airborne equipment, identify key factors affecting quality risks in the airborne equipment manufacturing process from four aspects: machinery, materials, methods, and environment. The identification of risk factors for "machines" mainly includes: verification and maintenance of tooling fixtures and molds, status of product production equipment, electronic, information-based and standardized product document management system, inspection equipment capabilities, periodic calibration and inspection capabilities of measuring tools, and periodic maintenance of equipment. The identification of risk factors related to materials mainly includes: packaging material quality, whether the packaging has been verified, product material compliance rate, quality level of key components, warehouse location management, annual module production failure rate, and material expiration date management. The identification of risk factors in the "law" mainly includes: the number of scrapped items in the manufacturing process (process), the adequacy of the PFMEA / DMEA / quality control plan, the correctness and completeness of the standard operating procedure (SOP), the rework rate, the completeness of the functional inspection checklist, the product factory inspection process, and the product failure handling process, etc. The identification of risk factors in the "ring" mainly includes: whether the product storage environment meets the standards, and whether the product transportation protection environment meets the standards.
[0042] S22. Based on the actual characteristics of airborne equipment product development and production, key factors are quantified to form a quality risk assessment index system, which includes three types of quality risks: historical data risk, appearance defect risk, and functional performance risk. Taking appearance defect risk as an example, the specific index items are shown in Table 3 below: Table 3. Quality Risk Index System for Airborne Equipment Products Furthermore, the airborne equipment product quality risk assessment model in S3 includes an assessment model based on the Support Vector Machine (SVR) algorithm, which mainly includes the following steps: S3-1. Determine the inputs and outputs of the risk assessment model: 1) Input: Refine the data for each indicator in the indicator system, and specify the data collection items as input values for the evaluation model. The specific data to be collected is shown in Table 4 below: Table 4 Data Collection Table of Airborne Equipment Product Quality Risk Indicators Serial Number Data source Example Product Name OEM Product X Product Number OEM ABC123 Batch number OEM EF456 batch OEM 200 Production Year and Month OEM 202x / xx Delivery year and month OEM 202x / xx The corresponding raw material product name from the secondary product product Product X Corresponding raw material batch number from secondary products product EF456 / DK789 Corresponding raw material delivery month from secondary products product 2020 / 03 Before this batch left the factory or in the corresponding production month, were there any changes in the key personnel responsible for the resume data? [Yes / No] product yes *If there have been changes in key personnel positions listed in the resume, please fill in the changed position [text]. product xx position Before this batch left the factory, did the manufacturer connect to a standardized document management system? [Yes / No] product yes Did any process changes occur during the production of this batch or in the corresponding production month? [Yes / No] product yes During the production process of this batch, were there any changes in key production personnel? [Yes / No] product no *If there are changes in key production personnel, please fill in the changed position [text]. product xx position The process capability index Cpk [value] for this batch (or the corresponding production month) product 1.4 During the production of this batch, were any key equipment changes? [Yes / No] product yes Does this batch (or production month) involve a software version update? [Yes / No] product yes The first-piece inspection pass rate for this product is [percentage]. product 95% The scrap rate (percentage) of key processes during this batch of production. product 5% The percentage of batch failing the final inspection before leaving the factory. product 5% Number of modifications and design changes for this batch of products [numerical value] OEM 0 Up to this batch, the number of times problems occurred in nearly 100 resumes [numerical value]. OEM 123 Upon unpacking and inspection, the number of products with issues in the batch's documentation was determined. [Value] OEM 10 Number of products with appearance defects in this batch upon unpacking and inspection [Value] OEM 2 Number of product appearance defects in the past 3 months up to this batch [numerical value] OEM 20 Number of product-related functional and performance issues in the past 3 months [Value] OEM 20 The percentage of items that pass inspection upon unpacking is [percentage]. OEM 90% The pass rate for this batch of tests is [percentage]. OEM 90% The assembly and flight test pass rate for this batch is [percentage]. OEM 90% The total number of products returned for repair due to functional or performance issues in the past three months up to this batch [numerical value]. OEM 20 Up to this batch, the average return period (days) for product functionality and performance issues over the past 3 months is as follows: [Value] OEM 20 As of this batch, the percentage of functional performance issues closed in the past three months. OEM 90% 2) Output: Quality risk values for this batch of airborne equipment products across three dimensions: historical data risk, appearance defect risk, and functional defect risk. Furthermore, the quality risk value of resume documentation risk is the resume documentation qualification rate, calculated using the following formula: ; in, This refers to the number of products with traceable documentation that meet quality standards in this batch. This refers to the total number of products in that batch. A higher compliance rate for the product traceability data indicates better stability and compliance in terms of quality control and information recording.
[0043] Furthermore, the quality risk value for appearance defects is the appearance pass rate, and the formula for calculating the appearance pass rate is: ; in, This indicates the number of products that meet the appearance requirements. This refers to the total quantity of products in this batch. A higher appearance pass rate indicates a more stable quality level in product appearance control, and reduces quality problems caused by appearance defects.
[0044] Furthermore, the quality risk value for the risk of functional defects is the functional pass rate, which is calculated using the following formula: ; in, Including the number of products that meet functional requirements. This refers to the total quantity of products in this batch. A higher functional pass rate indicates a greater proportion of the products meet functional requirements and a lower risk.
[0045] Based on the risk assessment index system, an airborne product quality risk assessment model is constructed to output quality risk values, including the following sub-steps: S32. Based on the quality risk characteristics of airborne equipment products, construct a quality risk assessment theoretical model based on support vector machine, and select kernel function and model parameters; Furthermore, the kernel function selected in S32 is determined to be the radial basis function (RBF). ; The discriminant function of a support vector machine constructed using a radial basis function kernel is as follows: ; Here, *s* support vectors determine the center location of the radial basis function, where *s* is the number of centers. Assuming the sample data is spherically distributed in the input space, then the parameters of the radial basis function kernel... The radius of the hypersphere can be taken.
[0046] Furthermore, after determining the RBF radial basis function as the kernel function for evaluating the model, three parameters affect the model: the penalty parameter C of the loss function, and the parameters of the RBF kernel function. (that is, functions) The values of the parameters, and also the parameters of the insensitive loss function. .
[0047] S33. Based on the quality risk assessment model, set the data collection scope, collect data on the product development and production process, and preprocess the process data; S34. Perform data preprocessing on the collected process data to eliminate differences between different units, in order to ensure that all input values are calculated on a uniform scale. Preferably, the input data of the model can be standardized by means of normalization, missing value imputation, and outlier handling: Normalization: Map all input data to the interval [0,1] so that the influence between different features is not unbalanced due to scale differences.
[0048] Missing value imputation: For missing data, imputation is performed using interpolation or other methods to ensure data integrity.
[0049] Outlier handling: By detecting and handling outliers, we can prevent them from negatively impacting model training.
[0050] Furthermore, to ensure the data meets the LibSVM data format in Python, after data preprocessing, the data is transformed into the format required by LibSVM. The LibSVM format is typically represented as a sparse matrix, where each row represents a sample data point, in the format: label feature1:value1 feature2:value2 ... featureN:valueN. Label is the target value, which is the output of the support vector regression machine; in this model, it represents the value of a certain type of quality risk for airborne equipment products. Index is the ordered index, including feature numbers, which must be arranged in ascending order. Value is the feature value, the data used for training, typically composed of a set of real numbers.
[0051] Select a support vector machine kernel function suitable for the airborne equipment product quality risk assessment model; The hyperparameters of a support vector machine are optimized using the particle swarm optimization algorithm to find the optimal parameter combination. The optimization process is as follows: Figure 1 As shown, the optimization parameters include the penalty factor C and the parameters of the kernel function. .
[0052] The processed standardized data is used as the training set and input into the risk assessment model for training. The goal of the training process is to fit the training data by minimizing the insensitive loss function so that the model can assess the quality risk of each product as accurately as possible and obtain the quality risk values of each batch of products in various dimensions. S35. After the model training is completed, input the test set data, evaluate the quality level in the test set data, and finally output the risk value of the product in each risk dimension. Furthermore, the data is divided into training and testing sets in an 8:2 ratio to ensure that the model can learn fully during the training phase and verify its generalization performance during the testing phase.
[0053] S36. After the evaluation is completed, assess the performance level of the airborne equipment product quality risk assessment model and optimize the model using appropriate evaluation indicators. To comprehensively evaluate the model's performance, mean squared error (MSE) and correlation coefficient were selected. The specific calculation formula for this evaluation indicator is as follows: ; ; A smaller MSE indicates a smaller fitting error of the model to the training data; a closer R-squared is to 1 indicates a high degree of consistency between the model's evaluation results and the actual values, indicating a stronger evaluation capability. These metrics can be used to measure the performance of the established support vector machine regression model in product quality risk assessment.
[0054] like Figure 2 As shown, during the iterative optimization of PSO, the mean squared error (MSE) of the model gradually decreases with the number of iterations, eventually converging to a stable level, indicating good convergence of the parameter optimization process. The final optimal hyperparameters obtained are C=8.7243, Gamma=0.001, and Epsilon=0.0293, with a corresponding minimum cross-validation error of 0.0004. The model's performance on the training set is MSE=0.000378, R0.000378. 2 =0.7874, with an MSE of 0.000375 and R on the test set. 2 =0.7753, indicating that the model can fit the changing trend of the pass rate of design-type products well and has a certain generalization ability.
[0055] The evaluation results show that the actual values and evaluated values of the test set generally follow the same trend. Figures 3 to 4 As shown, the evaluation accuracy is high in the high pass rate range (>0.9), but there is a slight bias in some low pass rate samples, indicating that the model still has certain limitations in capturing sudden fluctuations.
[0056] like Figure 4 The scatter plot results show that most sample points are distributed near the diagonal, indicating a high linear correlation between the estimated and actual values. However, a small number of outliers exist, reflecting the inherent uncertainty and nonlinearity in the design data. The error distribution results further show that the evaluation errors for most samples are concentrated in the range of -0.05 to +0.05, with only a small number of batches showing significant deviations, indicating good overall model stability.
[0057] Furthermore, the quality risk threshold in S4 includes the dividing point for whether there is a risk to the quality of airborne equipment products after batch delivery. The risk level can be obtained by comparing the quality risk value obtained from the evaluation with the risk threshold.
[0058] High-risk batch identification and verification analysis were conducted on this batch of airborne equipment products based on multiple threshold settings (0.85, 0.90, 0.95, 0.97) to test the adaptability and stability of the model under different risk tolerance levels. Due to the non-stationarity of the data and frequent design changes, the model accuracy for this batch of airborne equipment products was slightly lower but still remained within an acceptable range. At thresholds of 0.85, 0.90, 0.95, and 0.97, the overall accuracy rates were 97.5%, 84.0%, 88.0%, and 88.5%, respectively. A higher overall accuracy rate was achieved using a threshold of 0.90. Therefore, a risk value higher than 0.9 is considered risk-free, while a risk value lower than 0.9 is considered risky.
[0059] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made by those skilled in the art to the technical solutions of the present invention without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.
Claims
1. A method for assessing the product quality risk of airborne equipment based on process parameters, characterized in that, It includes the following steps: S1. Based on historical testing data, screen out typical quality problems of airborne equipment products and identify key factors affecting the occurrence of quality problems; S2. Based on key factors, construct a product quality risk assessment index system for airborne equipment; S3. Based on the risk assessment index system, construct an airborne product quality risk assessment model and output the quality risk value, including the following sub-steps: S31. Based on the risk assessment indicator system, determine that the collected data will be used as the input value of the assessment model, and the quality risk value will be used as the output value. S32. Construct a quality risk assessment model based on support vector machines, selecting the kernel function and model parameters, including: S321. Determine the evaluation model: ； In the formula, for Lagrange multipliers, corresponding for Optimization of Lagrange multipliers For kernel functions; This is the constant term of the decision function; S322. Determine the kernel function: The discriminant function of the support vector machine constructed using the radial basis kernel function takes the following form: ； Where s is the number of centers, and s support vectors can determine the center location of the radial basis function. and Including input samples, This includes the squared Euclidean distance between two samples. This includes parameters that control the scope of influence; S323. Determine model parameters: Determine model parameters based on radial basis function kernel function: penalty parameter C of loss function, radial basis function, and function parameters. Parameters of the insensitive loss function ; S33. Based on the quality risk assessment model, set the data collection scope, collect data on the product development and production process, and preprocess the process data; S34. Optimize the hyperparameters of the support vector machine based on the particle swarm optimization algorithm to find the optimal parameter combination; S35. Use the preprocessed process data as the training set to train the quality risk assessment model. After training, input the test set data and output the quality risk value of each risk dimension of the product. S4. Based on the set quality risk level threshold, the quality risk value divides the quality of airborne equipment into several risk levels.
2. The airborne equipment product quality risk assessment method based on process parameters as described in claim 1, characterized in that, Step S3 also includes: S36, After the evaluation is completed, the performance level of the quality risk assessment model is evaluated: the mean squared error (MSE) and correlation coefficient are selected. As an evaluation indicator: ； ； The smaller the MSE, the smaller the model's fitting error to the training data; the closer R is to 1, the closer the model's evaluation result is to the actual value. Based on the assessment results, the quality risk assessment model was optimized by adjusting the evaluation indicators.
3. The airborne equipment product quality risk assessment method based on process parameters as described in claim 1, characterized in that, Step S2 includes the following sub-steps: S21. The identified key influencing factors are classified into four types: machine, material, method, and environment. Risk factor identification includes: verification and maintenance of tooling fixtures and molds, status of product production equipment, electronic, information-based and standardized product document management system, inspection equipment capability, periodic calibration and inspection capability of measuring tools, and periodic maintenance of equipment; Risk factor identification includes: packaging material quality, whether the packaging has been verified, product material compliance rate, key component quality level, warehouse location management, annual module production failure rate, and material expiration date management. Risk factors identification includes: the number of scrapped items in the manufacturing process, the accuracy and completeness of standard operating instructions, rework rate, the completeness of the functional inspection checklist, the factory inspection process, and the fault handling process. Environmental risk factor identification includes: compliance of product storage environment and compliance of product transportation protection environment; S22. Based on the research and development process and historical data of airborne equipment products, determine the weight of key factors and quantify the key factors, summarize the risks of historical data, appearance defects and functional performance, and construct a quality risk assessment index system for airborne equipment products.
4. The airborne equipment product quality risk assessment method based on process parameters as described in claim 1, characterized in that, The quality risk value in step S31 includes risks related to historical data, appearance defects, and functional defects, and the quality risk value is measured by the pass rate.
5. The airborne equipment product quality risk assessment method based on process parameters as described in claim 1, characterized in that, Step S321 determines the method for evaluating the model as follows: Support vector regression is expressed as a set of training samples. The goal of regression is to find the optimal regression function. , Introduction As a risk function: ； Introducing the Lagrange coefficient: ； ； in for Lagrange multipliers, corresponding for The optimization of Lagrange multipliers is obtained by solving the optimization problem. and : ； To ensure stability, the average of the support vectors is used to determine the regression equation: ； in For kernel function, This is the constant term of the decision function.
6. The airborne equipment product quality risk assessment method based on process parameters as described in claim 1, characterized in that, The preprocessing method in step S34 includes: standardizing the input data of the model by means of normalization, missing value imputation and outlier handling.
7. The airborne equipment product quality risk assessment method based on process parameters as described in claim 1, characterized in that, Step S4 includes the following sub-steps: S41: Determine the quality risk threshold for the quality risk type of airborne equipment products. By assessing the quality risk value, compare the assessment accuracy of the batch of airborne equipment products under different thresholds, and select the threshold with the highest accuracy as the quality risk threshold for this type of airborne equipment products. S42. Establish the correlation between quality risk threshold and quality risk level.
8. The airborne equipment product quality risk assessment method based on process parameters as described in claim 4, characterized in that, The quality risk value of resume documentation risk is the resume documentation qualification rate, which is obtained by: 。 9. The airborne equipment product quality risk assessment method based on process parameters as described in claim 4, characterized in that: The quality risk value for appearance defects is the appearance pass rate, which is obtained as follows: 。 10. The airborne equipment product quality risk assessment method based on process parameters as described in claim 4, characterized in that: The quality risk value for the risk of functional defects is the functional pass rate, which is obtained by: 。