Power amplifier production detection system and method combined with big data

Abstract

The invention discloses a power amplifier production detection system and method combined with big data, and relates to the technical field of production detection, and the method comprises the steps: carrying out the sampling inspection of different batches based on the production quality, obtaining historical production data through big data, carrying out the classification and division of the values of all production factors, and forming a sample data set; establishing a qualification rate calculation model; parameters are adjusted based on a gradient descent method, mean square errors under different parameters are compared for iterative optimization, and a parameter model with the minimum mean square error under various classifications is obtained; substituting each batch of products into the model, and dynamically adjusting the sampling inspection rate of each batch according to a confidence interval formula; and finally, feeding back information and optimizing the process. The method has the advantages that products are detected through big data, it is ensured that quality problems can be found and processed in time, real-time product quality feedback can be obtained by dynamically adjusting the spot check rate, and the overall product percent of pass and market competitiveness are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of production detection, and specifically relates to a power amplifier production detection system and method combined with big data. Background Technique

[0002] A power amplifier production detection system and method combined with big data is a method that uses big data technology to analyze and optimize the production and detection processes of power amplifiers to improve production efficiency and product quality. The system establishes a pass rate calculation model based on historical production data, obtains a parameter model with the smallest mean square error through iterative optimization of parameters, calculates the pass rate of each batch, dynamically adjusts the product sampling rate according to the pass rate, and forms a cycle through information feedback to optimize the process.

[0003] The power amplifier production detection systems and methods on the market mainly integrate technologies such as automated detection, performance testing, environmental monitoring, and big data analysis. These systems use machine vision, advanced testing equipment, and sensors to automatically detect welding quality and electrical performance to ensure that products meet design specifications. The performance testing system evaluates key parameters such as the gain, output power, and distortion of power amplifiers through equipment such as network analyzers and oscilloscopes. However, there are deficiencies in the optimization and iteration of product pass rate detection. Redundant repeated detections often waste resources and greatly reduce detection efficiency, while simple detections often produce a large number of undetected products. Therefore, it is very important to balance the relationship between the two. Summary of the Invention

[0004] In order to solve the drawbacks of existing power amplifier production detection, a power amplifier production detection system and method are provided. The method constructs a pass rate calculation model, calculates the pass rate through the similarity with each model, and dynamically adjusts the sampling rate of products in each batch based on the pass rate.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0006] A power amplifier production detection method combined with big data includes:

[0007] Sampling different batches based on the production quality of power amplifiers, and obtaining historical production data of power amplifiers through big data, including production influencing factors and pass rates;

[0008] Based on historical production data, classify and divide the values of each production factor to form a sample data set and establish a pass rate calculation model;

[0009] Based on the gradient descent method, adjust each parameter in the model, compare the mean square errors of the models under different parameters, and continuously iterate and optimize to obtain a parameter model with the smallest mean square error under various classifications;

[0010] Based on the parameter model with the minimum mean square error, each batch of products is brought into the model to calculate the qualification rate. According to the confidence interval formula, the sampling rate of each batch is dynamically adjusted.

[0011] Based on the qualification rates of all batches of power amplifiers, feedback information is provided to form a feedback loop to optimize the process.

[0012] Preferably, based on the production quality of the power amplifier, it is divided into different batches for sampling inspection, and historical production data of the power amplifier is obtained through big data, including production influencing factors and qualification rates, specifically including:

[0013] According to the relevant parameter quality of the power amplifier, it is divided into three grades: excellent products, medium-quality products, and defective products, and sampled in batches.

[0014] Based on the database of the system, historical production data of all power amplifiers is obtained, including the qualification rates of products in different batches and production influencing factors.

[0015] Preferably, based on the historical production data, the values of each production factor are classified and divided to form a sample data set. The establishment of the qualification rate calculation model specifically includes:

[0016] Production factors include: gain linearity and bandwidth, thermal management, power supply stability, power supply efficiency, power supply filtering and decoupling design, component quality, environmental temperature, humidity, electromagnetic interference, production process.

[0017] Each production factor is classified based on the numerical size and divided into different levels.

[0018] Based on the classification results of different levels of different production factors, they are combined with each other to cover all combination situations, and a sample data set table is obtained.

[0019] According to the linear regression model, a qualification rate calculation model is established. Based on the classification of types in the sample data table, the data and qualification rates that meet the requirements are queried from the historical production data and brought into the model, and all types are fully covered. The qualification rate calculation model is:

[0020]

[0021] Among them, f(x) is the product qualification rate, x 1 , x 2 ,..., x n are the values of production factors, and α 1 , α 2 ,..., α n are the coefficients to be estimated.

[0022] Preferably, adjusting each parameter in the model based on the gradient descent method, comparing the mean square errors of the models under different parameters, and continuously iterating and optimizing to obtain the parameter model with the minimum mean square error for each classification specifically includes:

[0023] Adjusting different parameters in the model through the gradient descent method, calculating the mean square error values of a large number of parameter models with estimated coefficients, and the mean square error loss function is defined as:

[0024]

[0025] where, f i (x) is the actual qualified rate value of the i-th model, is the predicted qualified rate value of the i-th model;

[0026] Based on the calculated mean square error values, comparing and obtaining the parameter value with the minimum mean square error as the best parameter with estimated coefficients;

[0027] Collecting and integrating the best parameter models with estimated coefficients for the combinations between different levels of each production factor into a sample parameter model dataset.

[0028] Preferably, the adjusting different parameters in the model through the gradient descent method and calculating a large number of parameter models with estimated coefficients and the corresponding qualified rates specifically includes:

[0029] For each parameter α i , calculating the partial derivative of the objective function, i.e., f(x): Obtaining the gradient;

[0030] According to the calculated gradient, updating the value of each parameter, and the update formula is:

[0031]

[0032] where, is the value of the parameter α i at the k-th iteration, and μ is the learning rate;

[0033] Through multiple iterations, continuously updating different parameters α i until the objective function f(x) converges.

[0034] Preferably, based on the parameter model with the highest qualified rate, bringing the products of each batch into the model to calculate the qualified rate, and dynamically adjusting the sampling rate of each batch according to the confidence interval formula specifically includes:

[0035] Bringing the data of the production factors of all power amplifiers into the best qualified rate calculation model to calculate the qualified rates of products in different batches;

[0036] Calculate the upper and lower limits of the confidence interval based on the calculated pass rate, which are respectively:

[0037]

[0038] Among them, is the pass rate of the power amplifier, Z is the Z value under the standard normal distribution, corresponding to the selected confidence level. For the usual 90% confidence level: Z = 1.645, 95% confidence level: Z = 1.96, 99% confidence level: Z = 2.576, and N is the total number of power amplifiers tested;

[0039] Update the sampling rate for power amplifiers in different quality batches according to the interval relationship between the confidence interval and the target pass rate.

[0040] Preferably, the updating of the sampling rate for power amplifiers in different quality batches according to the interval relationship between the confidence interval and the target pass rate specifically includes:

[0041] If the upper limit of the pass rate confidence interval is lower than the target pass rate, increase the sampling rate;

[0042] If the target pass rate is within the pass rate confidence interval, the sampling rate remains unchanged;

[0043] If the lower limit of the pass rate confidence interval is higher than the target pass rate, reduce the sampling rate.

[0044] Preferably, the feedback of information based on the pass rate of power amplifiers in all batches to form a feedback loop to optimize the process specifically includes:

[0045] By analyzing the unqualified batches, feedback the main factors affecting the pass rate to form a continuously optimized cycle;

[0046] According to the analysis results, adjust the process parameters for the discovered problems, apply the new process settings in small-batch trial production, and observe the effects.

[0047] Furthermore, a power amplifier production detection system combined with big data is proposed, including:

[0048] A processor, which is used to operate on power amplifiers in batches, obtain and call historical data in the processing database, analyze various factors affecting production, construct a pass rate calculation model, and iteratively optimize the parameters based on the gradient descent method, dynamically adjust the sampling rate of each batch based on the confidence interval formula, and feedback various information to form a feedback loop to optimize the process;

[0049] A storage module, which is used to store historical production data, the production factor parameters and pass rates of all power amplifiers;

[0050] An output module, which is used to output the actual detection results of each power amplifier and the adjustment plan for the sampling rate of the detection of each batch of products.

[0051] Optionally, the processor is internally integrated with:

[0052] An analysis unit, which mainly divides different batches based on the quality of the power amplifier;

[0053] A first calculation unit, which is used to calculate the parameters of data at different levels of each production factor;

[0054] A second calculation unit, which iteratively calculates and optimizes the parameters based on the gradient descent method;

[0055] A third calculation unit, which is used to calculate the upper and lower limits of the confidence interval of power amplifiers of different batch qualities, so as to adjust the sampling rate;

[0056] A feedback unit, which is used to feedback information to form a feedback loop to optimize the process;

[0057] Compared with the prior art, the advantages of the present invention are as follows:

[0058] The present invention proposes a power amplifier production detection system and method combined with big data. The key advantages of this algorithm lie in the data-driven decision-making process and the adaptive ability of the model. By continuously optimizing the parameters, the system can dynamically adapt to different production conditions and improve the qualified rate. In addition, this method realizes the rapid transformation and evaluation of production factors for new batches, can timely reflect the production quality, and make adjustments in a timely manner, making the entire production process more efficient and reliable. The strategy of dynamically adjusting the sampling rate can optimize the resource allocation according to real-time data, avoid unnecessary sampling costs, and ensure the effectiveness of detection at the same time. This intelligent detection strategy can also reduce the labor input and improve the production efficiency. In addition, the closed-loop system formed by information feedback uses the real-time detection results for process improvement to achieve continuous optimization. This feedback mechanism enables enterprises to learn from actual production, continuously adjust and improve the process flow, improve product quality, and reduce the defective rate. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 It is a schematic flow chart of the qualified rate detection steps proposed by the present invention;

[0060] Figure 2 It is a schematic flow chart of the product batch sampling inspection proposed by the present invention;

[0061] Figure 3 It is a schematic flow chart of the construction process of the qualified rate calculation model proposed by the present invention;

[0062] Figure 4Schematic diagram of the parameter adjustment model optimization process proposed by the present invention;

[0063] Figure 5 Schematic diagram of the gradient descent method process proposed by the present invention;

[0064] Figure 6 Schematic diagram of the dynamic adjustment process of the sampling rate proposed by the present invention;

[0065] Figure 7 Schematic diagram of the information feedback process proposed by the present invention. Detailed implementation manners

[0066] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments in the following description are only examples, and those skilled in the art can think of other obvious variations.

[0067] A power amplifier production detection system combined with big data, comprising:

[0068] A processor, which is used for batch operations on power amplifiers, obtaining and calling historical data in the processing database, analyzing various factors affecting production, constructing a pass rate calculation model, and iteratively optimizing parameters based on the gradient descent method, dynamically adjusting the sampling rate of each batch based on the confidence interval formula, and feeding back various information to form a feedback loop to optimize the process. The processing unit can be any conventional processor, controller, microcontroller or state machine. The processor can also be implemented by analyzing a combination of systems, such as a digital signal processor and a microprocessor, multiple microprocessors, one or more microprocessors combined with a digital signal processor core, or any other similar configuration.

[0069] A storage module, which is mainly used for storing historical production data, production factor parameters and pass rates of all power amplifiers;

[0070] An output module, which is used for outputting the actual detection results of each power amplifier and the adjustment scheme of the sampling rate for the detection of each batch of products.

[0071] Among them, the processor is internally integrated with:

[0072] An analysis unit, which divides different batches based on the quality of power amplifiers;

[0073] A first calculation unit, which is used for calculating the parameters of data at different levels of each production factor;

[0074] A second calculation unit, which iteratively calculates and optimizes the parameters based on the gradient descent method;

[0075] A third calculation unit, which is used to calculate the upper and lower limits of the confidence interval of power amplifiers of different batches in terms of quality, so as to adjust the sampling rate;

[0076] A feedback unit, which is used to feedback information to form a feedback loop to optimize the process.

[0077] The operation process of the above power amplifier production detection system combined with big data is as follows:

[0078] Step 1: Based on the production quality of power amplifiers, they are divided into different batches for sampling inspection, and historical production data of power amplifiers, including production influencing factors and qualification rates, are obtained through big data;

[0079] Step 2: Based on the historical production data, the values of each production factor are classified and divided to form a sample data set, and a qualification rate calculation model is established;

[0080] Step 3: Based on the gradient descent method, each parameter in the model is adjusted, the mean square error of the model under different parameters is compared, and continuous iteration and optimization are carried out to obtain a parameter model with the minimum mean square error for each classification;

[0081] Step 4: Based on the parameter model with the minimum mean square error, the products of each batch are brought into the model to calculate the qualification rate, and according to the confidence interval formula, the sampling rate of each batch is dynamically adjusted;

[0082] Step 5: Based on the qualification rates of all batches of power amplifiers, feedback information is provided to form a feedback loop to optimize the process.

[0083] Refer to Figure 2 As shown in , based on the production quality of power amplifiers, they are divided into different batches for sampling inspection, and historical production data of power amplifiers, including production influencing factors and qualification rates, specifically include:

[0084] According to the relevant parameter quality of power amplifiers, they are divided into three grades: excellent products, medium-quality products, and defective products, and sampled batch by batch;

[0085] Based on the database of the system, historical production data of all power amplifiers are obtained, including the qualification rates of products in different batches and production influencing factors.

[0086] Specifically, considering the sampling inspection cost and efficiency, each batch should contain a sufficient number of power amplifiers to make the sampling representative. Secondly, when setting the initial sampling rate value, historical data should be collected first, and at the same time, referring to the sampling standards of statistics and combining with the qualification rate target, a reasonable sampling rate should be set.

[0087] Refer to Figure 3As shown, based on historical production data, classifying and dividing the values of each production factor to form a sample data set, and establishing a pass rate calculation model specifically includes:

[0088] Production factors include: gain linearity and bandwidth, thermal management, power supply stability, power supply efficiency, power supply filtering and decoupling design, component quality, environmental temperature, humidity, electromagnetic interference, production process;

[0089] Classify each production factor based on its numerical size and divide it into different levels;

[0090] Based on the classification results of different levels of different production factors, combine them with each other to cover all combination situations and obtain a sample data set table;

[0091] Establish a pass rate calculation model according to the linear regression model. Based on the classification of types in the sample data table, query the corresponding data and pass rate in the historical production data and bring them into the model, and cover all types. The pass rate calculation model is:

[0092]

[0093] Among them, f(x) is the product pass rate, x 1 , x 2 ,..., x n are the values of production factors, and α 1 , α 2 ,..., α n are the coefficients to be estimated.

[0094] Specifically, when calculating the coefficients to be estimated, use the least squares method to minimize the error between the predicted value and the actual value of the model. Calculate the partial derivatives of α 1 , α 2 ,..., α n respectively, and set the partial derivatives to 0 to obtain the normal equation. By solving the normal equation, the estimated values of the parameters can be obtained, which can be expressed in matrix form:

[0095] α = (X T X) -1 X T y

[0096] Among them, X is the design matrix including independent variables and constant terms, and y is the vector of dependent variables;

[0097] It is understandable that in production data, some production factors may be interrelated. For example, there may be a correlation between temperature and humidity. If these features enter the model simultaneously, it may lead to multicollinearity, affecting the stability of the regression coefficients and the predictive ability of the model. In this regard, the method of principal component analysis (PCA) can be used for dimensionality reduction, converting multiple related features into a few unrelated principal components to avoid multicollinearity.

[0098] Refer to Figure 4 As shown, based on the gradient descent method, each parameter in the model is adjusted, the mean squared error of the model under different parameters is compared, and continuous iteration and optimization are carried out to obtain the parameter model with the minimum mean squared error for various classifications. Specifically, it includes:

[0099] By adjusting different parameters in the model through the gradient descent method, the mean squared error values of a large number of parameter models with estimated coefficients are calculated. The mean squared error loss function is defined as:

[0100]

[0101] where, f i (x) is the actual qualified rate value of the i-th model, is the predicted qualified rate value of the i-th model;

[0102] Based on the calculated mean squared error values, compare and obtain the parameter value with the minimum mean squared error among them as the best parameter with estimated coefficients;

[0103] Collect and integrate the best parameter models with estimated coefficients for the combinations between different levels of each production factor into a sample parameter model dataset.

[0104] Refer to Figure 5 As shown, by adjusting different parameters in the model through the gradient descent method, a large number of parameter models with estimated coefficients and the corresponding qualified rates are calculated. Specifically, it includes:

[0105] For each parameter α i , calculate the partial derivative of the objective function, that is, f(x): Obtain the gradient;

[0106] According to the calculated gradient, update the value of each parameter. The update formula is:

[0107]

[0108] where, is the value of the parameter α i at the k-th iteration, and μ is the learning rate;

[0109] Through multiple iterations, continuously update different parameters α i until the objective function f(x) converges.

[0110] Specifically, during the calculation of partial derivatives, the changes of each parameter are closely related to the correlations with other parameters. To reduce the influence of other features on the partial derivative calculation, Lasso regression is used to compress the coefficients of irrelevant parameters to 0 by summing the absolute values of the regression coefficients. The specific Lasso regression loss function is as follows:

[0111]

[0112] where The first term is the ordinary mean square error term, the second term is the Lasso regression term, which sums the absolute values of the parameters. λ is the regularization parameter, and a larger λ will make the regression coefficients smaller.

[0113] To ensure convergence, it is necessary to optimize the learning rate μ to avoid oscillation and slow convergence problems. Usually, an adaptive optimization algorithm is used to dynamically adjust the learning rate, which can achieve faster convergence at a smaller learning rate. And due to the adaptive characteristics of the algorithm, complex learning rate adjustment is usually not required;

[0114] In some preferred embodiments, the adaptive optimization algorithm adopts Adam.

[0115] Refer to Figure 6 As shown, based on the parameter model with the highest qualified rate, the products of each batch are brought into the model to calculate the qualified rate. According to the confidence interval formula, dynamically adjusting the sampling rate of each batch specifically includes:

[0116] Bring the data of all production factors of the power amplifiers into the best qualified rate calculation model to calculate the qualified rates of products in different batches;

[0117] Based on the calculated qualified rate, calculate the upper and lower limits of the confidence interval, which are respectively:

[0118]

[0119] where is the qualified rate of the power amplifier, Z is the Z value under the standard normal distribution, corresponding to the selected confidence level. Usually, for a 90% confidence level: Z = 1.645, for a 95% confidence level: Z = 1.96, for a 99% confidence level: Z = 2.576, and N is the total number of power amplifiers detected;

[0120] According to the interval relationship between the confidence interval and the target qualified rate, update the sampling rate for power amplifiers in different quality batches.

[0121] It can be understood that the confidence interval formula has an issue with the sample size. If the sample size is too small, the confidence interval will be too wide and unable to effectively reflect the overall characteristics. In the early stage of production, the data volume of power amplifiers may be too small to accurately obtain the qualification rate. However, as the sample quantity continuously increases, the obtained qualification rate will gradually approach the actual value.

[0122] Refer to Figure 6 As shown, updating the sampling inspection rate for power amplifiers of different quality batches according to the interval relationship between the confidence interval and the target qualification rate specifically includes:

[0123] If the upper limit of the qualification rate confidence interval is lower than the target qualification rate, increase the sampling inspection rate;

[0124] If the target qualification rate is within the qualification rate confidence interval, the sampling inspection rate remains unchanged;

[0125] If the lower limit of the qualification rate confidence interval is higher than the target qualification rate, reduce the sampling inspection rate.

[0126] Refer to Figure 7 As shown, based on the qualification rates of all batches of power amplifiers, feedback information to form a feedback loop to optimize the process specifically includes:

[0127] By analyzing unqualified batches, feedback the main factors affecting the qualification rate to form a continuously optimized loop;

[0128] According to the analysis results, adjust the process parameters for the discovered problems, apply the new process settings in small-batch trial production, and observe the impacts.

[0129] It can be understood that during the entire detection process, obtaining and analyzing the sampling inspection results requires a certain amount of time, including the time from the sampling sample to obtaining the results, the time for data collation, analysis, and drawing conclusions, and the time from drawing conclusions to implementing adjustment measures. Therefore, time delays will occur, leading to subsequent problems, namely feedback delays. Such delays may cause delays in decision-making, thus affecting product quality and production efficiency. Therefore, for machines, certain requirements should be put forward for the automated equipment designed in the production process. Real-time data collection should be carried out, advanced data analysis software should be adopted, and data mining and machine learning technologies should be used to quickly analyze the sampling inspection results. In terms of human resources, a rapid feedback mechanism should be established, the data reporting process should be simplified to ensure that the results can be transmitted to decision-makers in the shortest possible time, and communication and collaboration between departments should be strengthened to ensure unobstructed information transmission.

[0130] In summary, the advantages of the present invention are as follows: by conducting multi-angle analysis based on historical production big data, comparing the inspected products with the most similar models, making quality control more precise, monitoring the production health status in real time, quickly discovering potential problems, this method improves the flexibility and adaptability of the production process. The strategy of dynamically adjusting the sampling inspection rate can optimize resource allocation according to real-time data, avoid unnecessary sampling inspection costs, and ensure the effectiveness of inspection at the same time. This intelligent inspection strategy can also reduce labor input and improve production efficiency. In addition, the closed-loop system formed by information feedback uses the real-time inspection results for process improvement to achieve continuous optimization. This feedback mechanism enables the enterprise to learn from actual production, continuously adjust and improve the process flow, improve product quality, and reduce the defective rate.

[0131] The foregoing has shown and described the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. What is described in the above embodiments and the specification is only the principle of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed. The scope of protection claimed by the present invention is defined by the appended claims and their equivalents.

Claims

1. A power amplifier production detection method combined with big data, characterized in that: include: Based on the production quality of power amplifiers, they are divided into different batches for random inspection, and the historical production data of power amplifiers, including production influencing factors and qualified rates, are obtained through big data; Based on historical production data, the values ​​of various production factors are classified and divided to form a sample data set, and a qualified rate calculation model is established; Based on the gradient descent method, the parameters in the model are adjusted, the mean square error of the model under different parameters is compared, and the optimization is continuously iterated to obtain the parameter model with the minimum mean square error under various classifications; Based on the parameter model with the minimum mean square error, each batch of products is brought into the model to calculate the qualified rate, and the sampling rate of each batch is dynamically adjusted according to the confidence interval formula; Based on the qualified rate of all batches of power amplifiers, the information is fed back to form a feedback loop to optimize the process.

2. The power amplifier production detection method combined with big data according to claim 1, characterized in that: The power amplifiers are divided into different batches for random inspection based on their production quality, and historical production data of the power amplifiers are obtained through big data, including production influencing factors and qualified rates. Specifically, the following are included: According to the quality of relevant parameters of power amplifiers, they are divided into three grades: high-quality, medium-quality and inferior, and sampled in batches; Based on the system database, the historical production data of all power amplifiers is obtained, including the qualified rates of different batches of products and production influencing factors.

3. The power amplifier production detection method combined with big data according to claim 1, characterized in that: Based on the historical production data, the values ​​of various production factors are classified and divided to form a sample data set, and the qualified rate calculation model is established, which specifically includes: Production factors include: gain linearity and bandwidth, thermal management, power supply stability, power supply efficiency, power supply filtering and decoupling design, component quality, ambient temperature, humidity, electromagnetic interference, and production process; Classify each production factor based on its numerical value and divide it into different levels; The classification results of different levels based on different production factors are combined with each other to cover all combinations and obtain a sample data set table; A qualified rate calculation model is established based on the linear regression model. Based on the classification of categories in the sample data table, the matching data and qualified rates are queried in the historical production data and brought into the model. All categories are fully covered. The qualified rate calculation model is: Among them, f(x) is the product qualification rate, x1, x2, ..., x n are the values ​​of production factors, α1, α2, ..., α n With estimated coefficients.

4. The power amplifier production detection method combined with big data according to claim 1, characterized in that: The method of adjusting each parameter in the model based on the gradient descent method, comparing the mean square error of the model under different parameters, and continuously iterating and optimizing to obtain the parameter model with the minimum mean square error under various classifications specifically includes: The different parameters in the model are adjusted by the gradient descent method, and the mean square error values ​​of a large number of models with estimated coefficient parameters are calculated. The mean square error loss function is defined as: Among them, f i (x) is the actual value of the pass rate of the i-th model, is the predicted value of the pass rate of the ith model; Based on the calculated mean square error values, the parameter value with the smallest mean square error is compared and obtained as the optimal band estimation coefficient parameter; The best parameter models with estimated coefficients for the combinations of various levels of various production factors are collected and integrated into a sample parameter model data set.

5. The power amplifier production detection method combined with big data according to claim 4 is characterized in that: The method of adjusting different parameters in the model by the gradient descent method and calculating a large number of parameter models with estimated coefficients and corresponding qualified rates specifically includes: For each parameter α i , calculate the partial derivative of the objective function, i.e. f(x): Get the gradient; According to the calculated gradient, the value of each parameter is updated. The update formula is: in, is the parameter α i The value at the kth iteration, μ is the learning rate; Through multiple iterations, different parameters α are continuously updated i , until the objective function f(x) converges.

6. The power amplifier production detection method combined with big data according to claim 5, characterized in that: The parameter model with the highest qualified rate is based on which each batch of products is brought into the model to calculate the qualified rate. According to the confidence interval formula, the sampling rate of each batch is dynamically adjusted. Specifically, the following steps are performed: The data of all power amplifier production factors are brought into the optimal qualified rate calculation model to calculate the qualified rates of different batches of products; Based on the calculated qualified rate, the upper and lower limits of the confidence interval are calculated, which are: Upper limit: Lower limit: in, is the qualified rate of power amplifiers, Z is the Z value under the standard normal distribution, corresponding to the selected confidence level, usually 90% confidence level: Z = 1.645, 95% confidence level: Z = 1.96, 99% confidence level: Z = 2.576, N is the total number of power amplifiers tested; According to the interval relationship between the confidence interval and the target qualified rate, the sampling rate is updated for power amplifiers of different quality batches.

7. The power amplifier production detection method combined with big data according to claim 6, characterized in that: The updating of the sampling rate for power amplifiers of different quality batches according to the interval relationship between the confidence interval and the target qualified rate specifically includes: If the upper limit of the confidence interval of the qualified rate is lower than the target qualified rate, the sampling rate is increased; If the target qualified rate is within the qualified rate confidence interval, the sampling rate remains unchanged; If the lower limit of the confidence interval of the qualified rate is higher than the target qualified rate, the sampling rate should be reduced.

8. The power amplifier production detection method combined with big data according to claim 1, characterized in that: The feedback information based on the qualified rate of all batches of power amplifiers to form a feedback loop to optimize the process specifically includes: By analyzing unqualified batches, we can provide feedback on the main factors affecting the pass rate and form a continuous optimization cycle; Based on the analysis results, process parameters are adjusted to address the issues identified, and new process settings are applied in small batch trial production to observe the impact.

9. A power amplifier production detection system combined with big data, used to implement a power amplifier production detection method combined with big data as described in any one of claims 1-8, characterized in that: include: Processor: The processor is used to operate the power amplifier in batches, obtain historical data in the call processing database, analyze various factors affecting production, build a qualified rate calculation model, and iteratively optimize the parameters based on the gradient descent method, dynamically adjust the sampling rate of each batch based on the confidence interval formula, and feed back various information to form a feedback loop to optimize the process; Storage module, the storage module is used to store historical production data, production factor parameters and qualified rates of all power amplifiers; The output module is used to output the actual test results of each power amplifier and the adjustment plan for the sampling rate of each batch of products.

10. The power amplifier production detection system combined with big data according to claim 9, characterized in that: The processor internally integrates: Analytical unit, the analytical unit is divided into different batches based on the quality of the power amplifier; A first calculation unit, the first calculation unit is used to calculate parameters of data at different levels of each production factor; A second computing unit, the second computing unit performs iterative calculation optimization on the parameters based on a gradient descent method; A third calculation unit, the third calculation unit is used to calculate the upper and lower limits of the confidence interval of power amplifiers of different batches of quality, so as to adjust the sampling rate; Feedback unit,The feedback unit is used to feed back information, forming a feedback loop to optimize the process.

Images