Microwave assembly quality control method and device based on fusion model, and storage medium

Abstract

The application provides a microwave assembly quality control method and device based on a fusion model and a storage medium. The method is applied to the technical field of electronic manufacturing and testing, and comprises the following steps: based on a microwave assembly physical model, a parameterized reduced-order data model and a base learner, an integrated learning method is used to construct a hybrid model; based on unstructured text data of a microwave assembly production process, a knowledge graph technology is used to construct a microwave assembly knowledge graph model; based on the microwave assembly knowledge graph model, the hybrid model is reviewed and verified to generate a microwave assembly physical-data-knowledge fusion model; and based on the microwave assembly physical-data-knowledge fusion model, microwave assembly quality control is performed. In this way, high-accuracy real-time prediction and online adjustment of microwave assembly manufacturing quality can be realized.

Description

Technical Field

[0001] This invention relates to the field of electronic manufacturing and testing technology, and further to the field of microwave component manufacturing and testing technology, particularly to a microwave component quality control method, equipment, and storage medium based on a fusion model. Background Technology

[0002] In the field of modern microwave component manufacturing and maintenance, performance prediction and fault diagnosis are crucial. Traditional performance prediction and fault diagnosis methods typically rely on a single physical model or data-driven approach. Physical models may not be accurate enough when dealing with complex or unknown environments, while purely data-driven models, although flexible, cannot provide accurate predictions in small sample situations or highly complex systems. Both methods have certain limitations.

[0003] Furthermore, traditional fault diagnosis methods rely heavily on human experience, which not only reduces the efficiency of diagnosis but also increases the subjectivity in the diagnosis process. Therefore, traditional performance prediction and fault diagnosis methods are often time-consuming, inefficient, and have limited accuracy when dealing with fault inversion and diagnosis in complex systems. Summary of the Invention

[0004] In view of this, the present invention provides a microwave component quality control method, device and storage medium based on a fusion model.

[0005] According to a first aspect of the present invention, a microwave component quality control method based on a fusion model is provided. The method includes:

[0006] Based on the equivalent circuits of microstrip line structure, gold wire bonding structure, vertical interconnect structure, assembly parameter equivalent circuits of RF connector, microwave component electrical performance data, and microwave component links, a physical model of microwave component integrating electromechanical coupling effects is constructed.

[0007] Acquire historical data from the intelligent testing unit and preprocess the historical data to obtain valid data;

[0008] Based on valid data, a parameterized reduced-order data model for the shape-property correlation of microwave components is constructed;

[0009] Based on the parameterized reduced-order data model of microwave component shape-property correlation, a base learner for microwave component shape-property correlation is constructed.

[0010] Based on the physical model of microwave components, the parameterized reduced-order data model, and the base learner, an ensemble learning method is used to construct a hybrid model;

[0011] Based on unstructured text data from the microwave component manufacturing process, a knowledge graph model for microwave components is constructed using knowledge graph technology.

[0012] Based on the knowledge graph model of microwave components, the hybrid model is reviewed and verified to generate a physical-data-knowledge fusion model of microwave components;

[0013] Microwave component quality control is carried out based on a physical-data-knowledge fusion model for microwave components.

[0014] Among some feasible implementations of the first aspect, a physical model of a microwave component incorporating electromechanical coupling effects is constructed based on equivalent circuits of microstrip line structures, equivalent circuits of gold wire bonding structures, equivalent circuits of vertical interconnect structures, equivalent circuits of RF connector assembly parameters, microwave component electrical performance data, and microwave component links, including:

[0015] Based on the type of microstrip line structure, the microstrip line structure is characterized as the first equivalent circuit physical model;

[0016] Based on the quasi-static model method and the method of moments, the gold wire bonding structure is characterized as a second equivalent circuit physical model;

[0017] Establish mathematical relationships between lumped distributed parameters and process parameters, and between lumped distributed parameters and electrical parameters; obtain the mathematical relationship between lumped distributed parameters and electrical parameters; characterize the vertical interconnect structure as a third equivalent circuit physical model based on the mathematical relationship between lumped distributed parameters and electrical parameters.

[0018] The relationship between insertion depth and contact impedance, insertion angle and air gap capacitance, and contact impedance and air gap capacitance and electrical parameters of the RF connector are established to obtain the mathematical relationship between insertion depth, angle and electrical performance of the RF connector; based on the mathematical relationship between insertion depth, angle and electrical performance of the RF connector, the assembly parameters of the RF connector are characterized as a fourth equivalent circuit physical model.

[0019] Obtain electrical performance data of microwave components from RF chip manufacturers;

[0020] A physical model of microwave component links is constructed based on the principle of microwave network cascading.

[0021] Based on the first, second, third, and fourth equivalent circuit physical models, the electrical performance data of the microwave components provided by the manufacturer, and the physical model of the microwave component links, a microwave component physical model integrating electromechanical coupling effects is constructed.

[0022] In some possible implementations of the first aspect, historical data from the intelligent testing unit is acquired, and the historical data is preprocessed to obtain valid data, including:

[0023] Acquire historical data from the intelligent testing unit; the historical data includes: manufacturing parameters and measured electrical performance parameters of the microwave components;

[0024] The historical data is cleaned, denoised, reduced in dimensionality, and feature-selected to obtain effective data.

[0025] Among some feasible approaches to the first aspect, a parameterized reduced-order data model for the shape-property correlation of microwave components is constructed based on valid data, including:

[0026] Based on the valid data, the basis functions are solved using the dataset through intrinsic orthogonal decomposition; the expansion coefficients are calculated using the pseudo-inverse algorithm.

[0027] Establish a Gaussian process between the design variables and the expansion coefficients using a Gaussian process;

[0028] By linearly superimposing the expansion coefficients and basis functions, a parameterized reduced-order mathematical model of the shape-property correlation of microwave components is obtained.

[0029] In some possible implementations of the first aspect, a base learner for the shape-property correlation of microwave components is constructed based on a parameterized reduced-order data model of the shape-property correlation of microwave components, including:

[0030] Each set of data in the valid data is used as a sample, and the electrical performance of the microwave component corresponding to each set of data is used as the main label. A training set is generated based on each set of data and its corresponding label.

[0031] Select the type of base learner;

[0032] From the training set, a sub-training set is extracted to train the neural network and obtain the corresponding base learner.

[0033] Among some possible implementations of the first aspect, a hybrid model is constructed using an ensemble learning method based on a microwave component physical model, a parameterized reduced-order data model, and a base learner, including:

[0034] Select the meta-learner based on the characteristics of the microwave component;

[0035] The microwave component physical model, the parameterized reduced-order data model, and the base learner are used as the base model for training.

[0036] Based on the training results of the base model, the meta-learner is updated and tuned until the optimal meta-learner is obtained, and the optimal meta-learner is output to generate a hybrid model.

[0037] In some possible implementations of the first aspect, unstructured text data is collected during the microwave component manufacturing process; the unstructured text data includes: structural knowledge, fault knowledge, expert experience knowledge, and electrical performance knowledge;

[0038] The unstructured text data is preprocessed. 25% of the text data is randomly selected from the unstructured text data, and the entity type label of each selected text data is labeled. The labeled text data is then divided into training set and test set in an 8:2 ratio.

[0039] Utilize graph neural networks to jointly extract entities and relations;

[0040] The extracted knowledge is constructed into knowledge triples according to the "head entity-relationship-tail entity" structure, and all knowledge triples are stored in the Neo4j graph database to build a knowledge graph model for microwave components.

[0041] In some possible implementations of the first aspect, based on the knowledge graph model of microwave components, the hybrid model is reviewed and validated to generate a microwave component physical-data-knowledge fusion model, including:

[0042] The hybrid model is reviewed and validated using a knowledge graph model of microwave components; a fusion model is then generated.

[0043] Collect real-time production data and conduct production control based on the fusion model; optimize and adjust the parameters of the fusion model according to the control results until a microwave component physical-data-knowledge fusion model is generated.

[0044] According to a second aspect of the present invention, an electronic device is provided. The electronic device includes: at least one processor; and a memory communicatively connected to the at least one processor; the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the method described above.

[0045] According to a third aspect of the present invention, a non-transitory computer-readable storage medium is provided storing computer instructions for causing a computer to perform the methods described above.

[0046] In this invention, a physical model of microwave components integrating electromechanical coupling effects is constructed based on the idea of ​​ensemble learning, a hybrid model based on data models is constructed, and a knowledge graph model of microwave components is constructed using knowledge graph technology. Based on the aforementioned model, a physical-data-knowledge fusion model of microwave components is further constructed. This enables high-accuracy real-time prediction of the manufacturing quality of microwave components and online guidance during the commissioning process.

[0047] It should be understood that the description in the Summary of the Invention is not intended to limit the key or essential features of the embodiments of the present invention, nor is it intended to restrict the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0048] The above and other features, advantages, and aspects of the various embodiments of the present invention will become more apparent from the accompanying drawings and the following detailed description. The drawings are provided for a better understanding of the invention and are not intended to limit the invention. In the drawings, the same or similar reference numerals denote the same or similar elements, wherein:

[0049] Figure 1 A flowchart of a microwave component quality control method based on a fusion model provided by an embodiment of the present invention is shown;

[0050] Figure 2 A schematic diagram of a microwave component quality control method based on a fusion model provided by an embodiment of the present invention is shown.

[0051] Figure 3 An exemplary application diagram of a microwave component quality control method based on a fusion model provided by an embodiment of the present invention is shown;

[0052] Figure 4 An exemplary BiLSTM-CRF algorithm structure diagram provided by an embodiment of the present invention is shown;

[0053] Figure 5 An exemplary Bert algorithm structure diagram provided in an embodiment of the present invention is shown;

[0054] Figure 6 An exemplary knowledge graph model provided by an embodiment of the present invention is shown;

[0055] Figure 7 A block diagram of an exemplary electronic device capable of implementing embodiments of the present invention is shown. Detailed Implementation

[0056] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0057] Furthermore, the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0058] In response to the problems mentioned in the background art, the present invention provides a microwave component quality control method, device and storage medium based on a fusion model.

[0059] Specifically, based on the equivalent circuits of microstrip line structures, gold wire bonding structures, vertical interconnect structures, RF connector assembly parameters, microwave component electrical performance data, and microwave component links, a microwave component physical model incorporating electromechanical coupling effects is constructed. Historical data from the intelligent testing unit is acquired and preprocessed to obtain valid data. Based on the valid data, a parameterized reduced-order data model of microwave component shape-property correlation is constructed. Based on the parameterized reduced-order data model of microwave component shape-property correlation, a base learner for microwave component shape-property correlation is constructed. Based on the microwave component physical model, the parameterized reduced-order data model, and the base learner, an ensemble learning method is used to construct a hybrid model. Based on the unstructured text data of the microwave component production process, knowledge graph technology is used to construct a microwave component knowledge graph model. Based on the microwave component knowledge graph model, the hybrid model is reviewed and verified to generate a microwave component physical-data-knowledge fusion model. Based on the microwave component physical-data-knowledge fusion model, microwave component quality control is performed.

[0060] In this way, a physical model of microwave components that integrates electromechanical coupling effects can be constructed based on the idea of ​​ensemble learning, a hybrid model based on data models can be constructed, and a knowledge graph model of microwave components can be constructed using knowledge graph technology. Based on the aforementioned model, a physical-data-knowledge fusion model of microwave components can be further constructed. This enables high-accuracy real-time prediction of the manufacturing quality of microwave components and online guidance during the commissioning process.

[0061] The following description, in conjunction with the accompanying drawings and specific embodiments, provides a more detailed explanation of the microwave component quality control method, device, and storage medium based on a fusion model provided by the present invention.

[0062] Figure 1 The flowchart illustrates a microwave component quality control method based on a fusion model provided by an embodiment of the present invention; as shown. Figure 1 As shown, the microwave component quality control method 100 based on the fusion model may include the following steps:

[0063] S110 constructs a physical model of a microwave component that integrates electromechanical coupling effects, based on the equivalent circuits of microstrip line structure, gold wire bonding structure, vertical interconnect structure, assembly parameter equivalent circuits of RF connectors, microwave component electrical performance data, and microwave component links.

[0064] Specifically, S110 may include the following steps:

[0065] Based on the type of microstrip line structure, the microstrip line structure is characterized as the first equivalent circuit physical model;

[0066] Based on the quasi-static model method and the method of moments, the gold wire bonding structure is characterized as a second equivalent circuit physical model;

[0067] Establish mathematical relationships between lumped distributed parameters and process parameters, and between lumped distributed parameters and electrical parameters; obtain the mathematical relationship between lumped distributed parameters and electrical parameters; characterize the vertical interconnect structure as a third equivalent circuit physical model based on the mathematical relationship between lumped distributed parameters and electrical parameters.

[0068] The relationship between insertion depth and contact impedance, insertion angle and air gap capacitance, and contact impedance and air gap capacitance and electrical parameters of the RF connector are established to obtain the mathematical relationship between insertion depth, angle and electrical performance of the RF connector; based on the mathematical relationship between insertion depth, angle and electrical performance of the RF connector, the assembly parameters of the RF connector are characterized as a fourth equivalent circuit physical model.

[0069] Obtain electrical performance data of microwave components from RF chip manufacturers;

[0070] A physical model of microwave component links is constructed based on the principle of microwave network cascading.

[0071] Based on the first, second, third, and fourth equivalent circuit physical models, the electrical performance data of the microwave components provided by the manufacturer, and the physical model of the microwave component links, a microwave component physical model integrating electromechanical coupling effects is constructed.

[0072] S120: Acquire historical data from the intelligent test unit and preprocess the historical data to obtain valid data.

[0073] Specifically, S120 may include the following steps:

[0074] Acquire historical data from the intelligent testing unit; the historical data includes: manufacturing parameters and measured electrical performance parameters of the microwave components;

[0075] The historical data is cleaned, denoised, reduced in dimensionality, and feature-selected to obtain effective data.

[0076] S130, Based on valid data, construct a parameterized reduced-order data model for the shape-property correlation of microwave components.

[0077] Specifically, S130 may include the following steps:

[0078] Based on the valid data, the basis functions are solved using the dataset through intrinsic orthogonal decomposition; the expansion coefficients are calculated using the pseudo-inverse algorithm.

[0079] Establish a Gaussian process between the design variables and the expansion coefficients using a Gaussian process;

[0080] By linearly superimposing the expansion coefficients and basis functions, a parameterized reduced-order mathematical model of the shape-property correlation of microwave components is obtained.

[0081] S140, based on the parameterized reduced-order data model of microwave component shape-property correlation, constructs a base learner for microwave component shape-property correlation.

[0082] Specifically, S140 may include the following steps:

[0083] Each set of data in the valid data is used as a sample, and the electrical performance of the microwave component corresponding to each set of data is used as the main label. A training set is generated based on each set of data and its corresponding label.

[0084] Select the type of base learner;

[0085] From the training set, a sub-training set is extracted to train the neural network and obtain the corresponding base learner.

[0086] S150 is based on a microwave component physical model, a parameterized reduced-order data model, and a base learner, and employs an ensemble learning method to construct a hybrid model.

[0087] Specifically, S150 may include the following steps:

[0088] Select the meta-learner based on the characteristics of the microwave component;

[0089] The microwave component physical model, the parameterized reduced-order data model, and the base learner are used as the base model for training.

[0090] Based on the training results of the base model, the meta-learner is updated and tuned until the optimal meta-learner is obtained, and the optimal meta-learner is output to generate a hybrid model.

[0091] S160 uses knowledge graph technology to construct a knowledge graph model of microwave components based on unstructured text data from the microwave component manufacturing process.

[0092] Specifically, S160 may include the following steps:

[0093] The unstructured text data is preprocessed. 25% of the text data is randomly selected from the unstructured text data, and the entity type label of each selected text data is labeled. The labeled text data is then divided into training set and test set in an 8:2 ratio.

[0094] Utilize graph neural networks to jointly extract entities and relations;

[0095] The extracted knowledge is constructed into knowledge triples according to the "head entity-relationship-tail entity" structure, and all knowledge triples are stored in the Neo4j graph database to build a knowledge graph model for microwave components.

[0096] S170, based on the knowledge graph model of microwave components, reviews and verifies the hybrid model to generate a physical-data-knowledge fusion model of microwave components.

[0097] Specifically, S170 may include the following steps:

[0098] The hybrid model is reviewed and validated using a knowledge graph model of microwave components; a fusion model is then generated.

[0099] Collect real-time production data and conduct production control based on the fusion model; optimize and adjust the parameters of the fusion model according to the control results until a microwave component physical-data-knowledge fusion model is generated.

[0100] S180 is based on a microwave component physical-data-knowledge fusion model for quality control of microwave components.

[0101] According to embodiments of the present invention, the problems of lack of performance prediction for manufacturing, difficulty in quality prediction under small sample conditions, heavy reliance on human experience for fault diagnosis, and long time for fault inversion and diagnosis debugging are solved.

[0102] To facilitate understanding, more detailed steps are provided below to further explain the above content.

[0103] According to an embodiment of the present invention, a hybrid modeling method integrating a physical model and a data model that integrates electromechanical coupling effects is constructed based on the idea of ​​ensemble learning, and a fault diagnosis technology based on knowledge graphs is integrated to realize the performance prediction and fault diagnosis of microwave components.

[0104] First, the electrical performance of the microwave component is characterized by a physical model of its structural parameters. The interconnection structure of the microwave component is calculated using equivalent circuits to obtain its scattering parameters. The chip scattering parameters of the microwave component are obtained from the parameter manual provided by the manufacturer. Based on field-circuit coupling, an electrical performance characterization model of the microwave component product is established.

[0105] Secondly, the form-property correlation parameters of microwave components are obtained through simulation. Combined with manufacturing parameters and electrical performance data obtained from testing units such as the real-time camera system, vector network analyzer, and spectrum analyzer in the cloud-edge-device collaborative intelligent testing unit during the microwave component manufacturing process, data noise reduction and dimensionality reduction are performed. The data is then randomly segmented to obtain training and testing sets for building the data model. Next, a rapid evolution model from form to property is constructed through parametric degradation modeling, and a base learner is obtained through deep neural network training. Then, the physical model and data model are fused using ensemble learning, and the weights of each model are superimposed by training the meta-learner. Finally, a knowledge graph is constructed using zeroing reports and expert knowledge obtained during the commissioning process to achieve high-accuracy real-time prediction of manufacturing process quality and online guidance during commissioning.

[0106] Furthermore, the specific steps can be as follows:

[0107] 1. Construct a physical model of a microwave component that incorporates electromechanical coupling effects;

[0108] 1.1 Characterize the microstrip line structure as an equivalent circuit physical model;

[0109] The equivalent circuit and parameters of the micro-element of a conventional microstrip line are calculated using classical approximation formulas. The distributed capacitance of the circuit in the form of a microstrip containing sealant is obtained by conformal transformation, and its characteristic impedance is solved by approximation structure. The network parameters of a coupled microstrip line are obtained by parity-mode parameter method.

[0110] 1.2, The gold wire and gold ribbon bonding structure is characterized as an equivalent circuit physical model;

[0111] The performance analysis of gold wire bonding interconnects is carried out by constructing an electrical performance characterization model through a quasi-static model method. The gold wire is divided into several segments, and each micro-element is equivalent to a set of lumped parameters. Then, the inductance, resistance and capacitance of the lumped parameters are obtained by the method of moments. The scattering parameters of the gold wire bonding are calculated based on the equivalent circuit model.

[0112] 1.3, The vertical interconnect structure is characterized as an equivalent circuit physical model;

[0113] The equivalent circuit of a vertical interconnect structure is usually a π-type circuit. A capacitor is connected in parallel across the series resistor and inductor to establish the mathematical relationship between lumped distributed parameters and process parameters. Then, the mathematical relationship between lumped distributed parameters and electrical parameters is established, thus obtaining the mathematical relationship between lumped distributed parameters and electrical parameters.

[0114] 1.4, The assembly parameters of the RF connector are characterized as an equivalent circuit physical model;

[0115] The main assembly parameters of the RF connector are insertion depth and angle. The equivalent circuit is characterized by contact impedance and air gap capacitance. The relationship between the insertion depth and contact impedance of the RF connector, and the relationship between the insertion angle and air gap capacitance are established. Then, the relationship between contact impedance, air gap capacitance and electrical parameters is established. Finally, the mathematical relationship between insertion depth and angle and the electrical performance of the RF connector is obtained.

[0116] 1.5 Obtain the electrical performance data of microwave components provided by the RF chip manufacturer;

[0117] 1.6, Constructing a physical model of microwave component links based on the principle of microwave network cascading

[0118] 2. Generate data;

[0119] 2.1 The manufacturing parameters and measured electrical performance parameters of microwave components are obtained through the visual inspection equipment, AR-assisted equipment, vector network analyzer and spectrum analyzer in the cloud-edge-device collaborative intelligent testing unit, forming the raw data for constructing the data model;

[0120] 2.2 Data cleaning: To ensure data quality, the raw data is analyzed, including partial removal, error correction, handling of missing values, and standardization of data formats, to establish standardized and usable data for the shape-related correlation of microwave components;

[0121] 2.3 Data denoising: For time series data, denoising is performed using filters and wavelet transforms. For non-time series data, denoising is performed using outlier detection and data smoothing. The main purpose is to remove electrical performance interference caused by environmental factors during microwave component testing.

[0122] 2.4 Data dimensionality reduction: Principal component analysis is used to reduce data dimensionality and improve the efficiency of evolutionary computation.

[0123] 2.5 Feature selection: Select the most important features, remove irrelevant or redundant features, and select the electrical performance of microwave components as the main label, finally obtaining 300 sets of samples.

[0124] 3. Construct a parameterized reduced-order data model for the shape-property correlation of microwave components;

[0125] 3.1 Solving for basis functions using the dataset through eigenorthogonal decomposition;

[0126] 3.2, Calculate the expansion coefficients using a pseudo-inverse algorithm;

[0127] 3.3, Establish a Gaussian process between design variables and expansion coefficients using a Gaussian process;

[0128] 3.4 By linearly superimposing the expansion coefficients and basis functions, a parameterized reduced-order mathematical model of the shape-property correlation of microwave components is obtained;

[0129] 4. Construct a base learner for the shape-property correlation of microwave components;

[0130] 4.1. Randomly or intentionally select 10 sub-training sets from the training set, with 100 samples in each sub-training set;

[0131] 4.2 Select the type of base learner;

[0132] 4.3. Train the neural network using the randomly selected sub-training set in step 4.1 to obtain 10 neural network surrogate models with different weights, i.e., 10 base learners.

[0133] 5. Utilize ensemble learning to construct a hybrid model that integrates the physical model and data model of electromechanical coupling effects;

[0134] 5.1 Selecting the meta-learner: Based on the characteristics of the microwave components, a neural network is selected as the meta-learner for the ensemble learning method.

[0135] 5.2 Training the meta-learner: Using the physical model constructed in step 1 and the data model constructed in steps 3 and 4 as the base model for training, the meta-learner is trained on the predicted feature set of the base model, and the hyperparameters of the meta-learner are adjusted to achieve the best combination of the physical model and the data model.

[0136] 5.3 Use cross-validation or test results to verify the predictive performance of the mixture model;

[0137] 6. Construct a knowledge graph model of microwave components using knowledge graph technology;

[0138] 6.1 Collect unstructured text data from the microwave component manufacturing process, including structural knowledge, fault knowledge, expert experience knowledge, and electrical performance knowledge;

[0139] 6.2 Generate training and test sets. Preprocess the unstructured text data, randomly select 25% of the text data from the unstructured text data, label the entity type of each selected text data, and generate training and test sets from the labeled text data in an 8:2 ratio.

[0140] 6.3 Ontology and Relation Extraction: Joint extraction of entities and relations using graph neural networks;

[0141] 6.4 Construct a microwave component knowledge graph, build knowledge triples according to "head entity-relationship-tail entity" of the extracted knowledge, and store all knowledge triples in the Neo4j graph database;

[0142] 7. Construct a physical-data-knowledge fusion model for microwave components that incorporates electromechanical coupling effects;

[0143] 7.1 Construct a question-and-answer system based on a microwave component knowledge graph;

[0144] 7.2 Utilizing knowledge graphs for the review and verification of hybrid model form;

[0145] 8. Production control based on a microwave component physical-data-knowledge fusion model;

[0146] 8.1. Model validation based on data from the production process;

[0147] 8.2 Use a fusion model for quality control in the production process.

[0148] The following is combined with Figures 2-5 The above content will be explained in more detail with a specific embodiment.

[0149] Figure 2 This diagram illustrates a microwave component quality control method based on a fusion model provided by an embodiment of the present invention; as shown. Figure 2 As shown, manufacturing process data, electrical performance data, design data, commissioning data, process data, quality data, and zeroing reports based on intelligent testing units and manufacturing information systems such as MES and PDM serve as the foundational data for constructing the quality control method. The persistence layer extracts this foundational data through data access and performs preliminary cleaning of the foundational data through methods such as preliminary data retrieval, filtering, sorting, adding, deleting, modifying, and handling data anomalies. After further data cleaning, format conversion, and filtering, the data is processed again and stored in the Neo4j database used to construct the fusion model. The fusion model integrating physics, data, and knowledge is constructed using the method described above. The specific steps are as follows:

[0150] 1. Construct a physical model of a microwave component that integrates electromechanical coupling effects, and build a calculation model for microwave component scattering parameters affected by manufacturing factors based on microwave network principles:

[0151]

[0152] Among them, S m*m (·) represents the scattering parameter of the microwave component, d i ,x i This indicates the relevant manufacturing elements.

[0153] 1.1 Characterize the microstrip line structure as an equivalent circuit physical model;

[0154]

[0155] Where W,h,l represent the width, height, and length of the microstrip line.

[0156] 1.2, The gold wire and gold ribbon bonding structure is characterized as an equivalent circuit physical model;

[0157]

[0158] Where D, h, d, and f represent the span, diameter, arch height, and frequency of the gold wire, respectively.

[0159] 1.3, The vertical interconnect structure is characterized as an equivalent circuit physical model;

[0160]

[0161] Where r, R, and h represent the inner and outer diameters of the vertical interconnect structure and the height of the substrate.

[0162] 1.4, The assembly parameters of the RF connector are characterized as an equivalent circuit physical model;

[0163]

[0164] Where x, T, and Q represent the connection displacement, temperature, and angle of the RF connector.

[0165] 1.5 Obtain the electrical performance data of microwave components provided by the RF chip manufacturer;

[0166] Consult the parameter manuals provided by the manufacturers of the RF chips used in the design; their electrical performance is characterized in the form of SNP files.

[0167] 1.6 Construct a physical model of microwave component links based on the principle of microwave network cascading, i.e., the formula in step 1;

[0168] 2. Generate data;

[0169] 2.1 Manufacturing parameters and measured electrical performance parameters of microwave components are acquired through visual inspection equipment, AR-assisted equipment, vector network analyzers, and spectrum analyzers in a cloud-edge-device collaborative intelligent testing unit. These parameters are then combined with simulation data to form the raw data for constructing a data model. Taking the receiving channel as an example, the characteristics of the raw data are the same as the assembly parameters of the physical model described in step 1, namely, the length, width, and height of the microstrip line; the span, diameter, and arch height of the gold wire bonding; the inner and outer diameters and substrate height of the vertical interconnect structure; the connection displacement, temperature, and angle of the RF connector; and the scattering parameters of the RF chip, etc. The labels of the raw data are link gain, gain-band flatness, power capacity, link reflection coefficient, etc. The raw training dataset is as follows:

[0170]

[0171] In the formula, D ori For the original training dataset, x ij Let represent the j-th feature value of the i-th sample.

[0172] 2.2 Data cleaning: To ensure data quality, the raw data is analyzed, including partial removal, error correction, handling of missing values, and standardization of data formats, to establish standardized and usable data for the shape-related correlation of microwave components;

[0173] 2.3 Data denoising: For time series data, denoising is performed using filters and wavelet transforms. For non-time series data, denoising is performed using outlier detection and data smoothing. The main purpose is to remove electrical performance interference caused by environmental factors during microwave component testing.

[0174] 2.4 Data Dimensionality Reduction: Principal component analysis is used for data dimensionality reduction to improve the efficiency of evolutionary computation. Data denoising employs filtering methods, with the specific denoising formula as follows:

[0175]

[0176] 2.5 Feature selection: Select the most important features, remove irrelevant or redundant features, and select the electrical performance of microwave components as the main label. Finally, 300 sets of samples were obtained, with 12 features in each sample.

[0177] 3. Construct a parameterized reduced-order data model for the shape-property correlation of microwave components;

[0178] 3.1 Solving for basis functions using the dataset through eigenorthogonal decomposition

[0179] 3.2 The expansion coefficients are calculated using a pseudo-inverse algorithm, which is as follows:

[0180]

[0181] in, This represents the pseudo-inverse operation of a matrix.

[0182] 3.3, Establish a Gaussian process between design variables and expansion coefficients using a Gaussian process;

[0183] 3.4. By linearly superimposing the expansion coefficients and basis functions, a parameterized reduced-order mathematical model of the microwave component shape-property correlation is obtained. The degradation model is as follows:

[0184]

[0185] 4. Construct a base learner for the shape-property correlation of microwave components;

[0186] 4.1. Randomly or intentionally select 10 sub-training sets from the training set, with 100 samples in each sub-training set;

[0187] 4.2 The type of base learner is selected as a feedforward neural network. All base learners adopt the same neural network structure: the input layer has 12 neurons, the second layer has 32 neurons, the third layer has 64 neurons, the fourth layer has 32 neurons, and the output layer has 4 neurons. The activation function is implemented by the ReLU function.

[0188] 4.3. Train the neural network using the randomly selected sub-training set in step 4.1 to obtain 10 neural network surrogate models with different weights, i.e., 10 base learners.

[0189] 5. Utilize ensemble learning to construct a hybrid model that integrates the physical model and data model of electromechanical coupling effects;

[0190] 5.1 Selection of Meta-learner: Based on the characteristics of microwave components, a simple feedforward neural network is selected as the meta-learner for the ensemble learning method. The neural network structure of the meta-learner is defined as follows: the input layer has 10 neurons, the second layer has 12 neurons, the third layer has 24 neurons, the fourth layer has 12 neurons, and the output layer has 1 neuron. The activation function is implemented by the ReLU function.

[0191] 5.2 Training the meta-learner: Using the physical model constructed in step 1 and the data model constructed in steps 3 and 4 as the base model for training, the meta-learner is trained on the predicted feature set of the base model, and the hyperparameters of the meta-learner are adjusted to achieve the best combination of the physical model and the data model.

[0192] 5.3 Use cross-validation or test results to verify the predictive performance of the mixture model;

[0193] 6. Construct a knowledge graph model of microwave components using knowledge graph technology;

[0194] 6.1 Collect unstructured text data from the microwave component manufacturing process and define the microwave component entity types, as shown in Table 1 below:

[0195] Table 1

[0196]

[0197]

[0198] 6.2 Generate training and test sets. Preprocess the unstructured text data, randomly select 25% of the text data from the unstructured text data, label the entity type of each selected text data, and generate training and test sets from the labeled text data in an 8:2 ratio.

[0199] 6.3, Figure 4An exemplary BiLSTM-CRF algorithm structure diagram provided by an embodiment of the present invention is shown; as follows: Figure 4 As shown, BiLSTM-CRF effectively handles sequence labeling tasks by combining a bidirectional long short-term memory network (BiLSTM) and a conditional random field (CRF). It first converts the text into vectors, uses BiLSTM to capture contextual information, and then optimizes the label sequence through a CRF layer to label each element in the text, thereby achieving entity and relation extraction in microwave components.

[0200] It should be noted that various knowledge graph construction algorithms can be used in the actual construction process, such as CasRel, TPLinker, PRGC, PURE, OneRel, Bert, Bert+CRF, or other algorithms that can achieve the same effect.

[0201] Figure 5 An exemplary BERT algorithm structure diagram provided in an embodiment of the present invention is shown; as follows: Figure 5 As shown, the text is first converted into vector form and then processed through multiple encoder layers to generate the final word vectors. These word vectors are then used to complete pre-training tasks, such as establishing the relationship between manufacturing elements and electrical performance in a zero-point report during the microwave component production process, and determining whether these relationships are logically connected.

[0202] 6.4, Figure 6 An exemplary knowledge graph model provided by an embodiment of the present invention is shown; such as Figure 6 As shown, a microwave component knowledge graph is constructed. The extracted knowledge is organized into knowledge triples according to the "head entity-relationship-tail entity" structure, and all knowledge triples are stored in the Neo4j graph database.

[0203] 7. Construct a physical-data-knowledge fusion model for microwave components that incorporates electromechanical coupling effects;

[0204] 7.1 Construct a question-and-answer system based on a microwave component knowledge graph;

[0205] 7.2 Utilize knowledge graphs to review and verify the form of the hybrid model. During use, verify the accuracy of the model based on actual test data, and improve the accuracy of the model through real-time feedback.

[0206] 8. Production control based on a microwave component physical-data-knowledge fusion model;

[0207] 8.1 The intelligent manufacturing system collects structural parameters and performance parameters of the production process in real time, and also collects design data, zeroing reports, etc.

[0208] 8.2 The fusion model system first uses the detailed and extensive knowledge system embedded in the knowledge graph to judge and compare the input data to ensure that the input data of the model not only matches the expected data type and format, but also conforms to the actual situation and theoretical basis of the microwave component production workshop. It issues warnings for invalid inputs and provides feedback on modification suggestions.

[0209] 8.3 The fusion model processes the legitimate data, uses manufacturing factors as input data to predict the quality of electrical performance, and uses the input data of electrical performance for analysis and judgment.

[0210] 8.4 The fusion model evaluates the validity and accuracy of the output data, identifies and interprets potential biases or errors in the model output, and provides a multi-dimensional, cross-domain information set.

[0211] According to an embodiment of the present invention, a fusion modeling method based on physics-data-knowledge is used to control the performance of the manufacturing process of microwave components. On the one hand, quality prediction is performed during the production process, and on the other hand, fault diagnosis is performed during the commissioning process.

[0212] Figure 3 This diagram illustrates an exemplary application of a microwave component quality control method based on a fusion model, as provided in an embodiment of the present invention; Figure 3 As shown, during the production process, the automatic testing system tests and acquires morphological information on manufacturing elements and electrical performance related to quality in real time, and imports it into the quality control system for data analysis. If the quality prediction is satisfactory, the production task continues. If the quality prediction is unsatisfactory, the quality control system proposes modification suggestions based on the prediction results of the fusion model. It also provides intelligent Q&A and fault statistics functions. Users can make repairs according to the system's suggestions. If the repair is satisfactory, the production task continues. If the repair suggestions provided by the quality control system cannot complete the repair, experts will review and implement the repair work, generating a new quality zero-reset report. This zero-reset report is simultaneously imported into the graph database to update the fusion model and system until the microwave component is produced.

[0213] According to embodiments of the present invention, the following technical effects are achieved:

[0214] 1. Performance control of microwave component manufacturing process can be carried out based on the fusion modeling method of physics-data-knowledge integration. On the one hand, quality prediction is carried out during production, and on the other hand, fault diagnosis is carried out during commissioning and testing.

[0215] 2. Unlike traditional models, the fusion model provided by this invention relies less on large datasets. It can operate efficiently with limited data, especially when collecting large amounts of data is impractical or costly. This feature makes the model more versatile and adaptable to various manufacturing environments.

[0216] 3. The control method of this invention has high accuracy. The high precision of this invention in controlling and predicting problems in microwave component manufacturing is noteworthy. It demonstrates advanced capabilities in diagnosing and resolving manufacturing and testing problems, thereby reducing error rates and improving overall product quality. This precision is crucial for maintaining high standards on the production line.

[0217] 4. This invention is easily integrated with existing systems. The convergence model is designed for easy integration into existing systems. This compatibility factor is crucial because it allows for seamless adoption without requiring significant modifications to current technology settings. It ensures that incorporating this model into the production process is both time-saving and economical.

[0218] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that the present invention is not limited to the described order of actions, because according to the present invention, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily essential to the present invention.

[0219] According to embodiments of the present invention, the present invention also provides an electronic device and a readable storage medium.

[0220] Figure 7 A block diagram of an exemplary electronic device capable of implementing embodiments of the present invention is shown.

[0221] like Figure 7 As shown, electronic device 700 is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. Electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0222] Electronic device 700 includes a computing unit 701, which can perform various appropriate actions and processes according to a computer program stored in ROM 702 or a computer program loaded into RAM 703 from storage unit 708. RAM 703 can also store various programs and data required for the operation of electronic device 700. The computing unit 701, ROM 702, and RAM 703 are interconnected via bus 704. I / O interface 705 is also connected to bus 704.

[0223] Multiple components in electronic device 700 are connected to I / O interface 705, including: input unit 706, such as keyboard, mouse, etc.; output unit 707, such as various types of displays, speakers, etc.; storage unit 708, such as disk, optical disk, etc.; and communication unit 709, such as network card, modem, wireless transceiver, etc. Communication unit 709 allows electronic device 700 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0224] The computing unit 701 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 701 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 701 performs the various methods and processes described above, such as method 100. For example, in some embodiments, method 100 may be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 708. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 700 via ROM 702 and / or communication unit 709. When the computer program is loaded into RAM 703 and executed by the computing unit 701, one or more steps of method 100 described above may be performed. Alternatively, in other embodiments, the computing unit 701 may be configured to perform method 100 by any other suitable means (e.g., by means of firmware).

[0225] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0226] The program code used to implement the methods of the present invention can be written in any combination of one or more programming languages. This program code can be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing device, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code can be executed entirely on the machine, partially on the machine, as a standalone software package partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0227] In the context of this invention, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. Machine-readable media can include, but are not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0228] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including voice input, speech input, or tactile input).

[0229] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with embodiments of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0230] Computer systems can include clients and servers. Clients and servers are generally located far apart and typically interact via communication networks. Client-server relationships are created by computer programs running on the respective computers and having a client-server relationship with each other. Servers can be cloud servers, servers in distributed systems, or servers incorporating blockchain technology.

[0231] It should be understood that the various forms of processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this invention can be achieved, and this is not limited herein.

[0232] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A microwave component quality control method based on a fusion model, characterized in that, The method includes: Based on the equivalent circuits of microstrip line structure, gold wire bonding structure, vertical interconnect structure, assembly parameter equivalent circuits of RF connector, microwave component electrical performance data, and microwave component links, a physical model of microwave component integrating electromechanical coupling effects is constructed. Historical data from the intelligent testing unit is acquired, and the historical data is preprocessed to obtain valid data; Based on the effective data, a parameterized reduced-order data model for the shape-property correlation of microwave components is constructed; Based on the parameterized reduced-order data model of microwave component shape-property correlation, a base learner for microwave component shape-property correlation is constructed. Based on the aforementioned microwave component physical model, parameterized reduced-order data model, and base learner, an ensemble learning method is used to construct a hybrid model; Based on unstructured text data from the microwave component manufacturing process, a knowledge graph model for microwave components is constructed using knowledge graph technology. Based on the knowledge graph model of the microwave component, the hybrid model is reviewed and verified to generate a physical-data-knowledge fusion model of the microwave component. Microwave component quality control is carried out based on the physical-data-knowledge fusion model of microwave components. The step of constructing a parameterized reduced-order data model for the shape-property correlation of microwave components based on the effective data includes: solving the basis functions using the dataset through intrinsic orthogonal decomposition based on the effective data; calculating the expansion coefficients through a pseudo-inverse algorithm; establishing a Gaussian process between the design variables and the expansion coefficients through a Gaussian process; and linearly superimposing the expansion coefficients and the basis functions to obtain the parameterized reduced-order data model for the shape-property correlation of microwave components. The parameterized reduced-order data model based on the form-property correlation of microwave components, and the construction of a base learner for the form-property correlation of microwave components, include: taking each set of data in the effective data as a sample, taking the electrical performance of the microwave component corresponding to each set of data as the main label, generating a training set according to each set of data and its corresponding label; selecting the type of base learner; extracting a sub-training set from the training set to train the neural network, and obtaining the corresponding base learner.

2. The method according to claim 1, characterized in that, The aforementioned method, based on equivalent circuits of microstrip line structures, gold wire bonding structures, vertical interconnect structures, RF connector assembly parameters, microwave component electrical performance data, and microwave component links, constructs a physical model of microwave components that integrates electromechanical coupling effects, including: Based on the type of microstrip line structure, the microstrip line structure is characterized as the first equivalent circuit physical model; Based on the quasi-static model method and the method of moments, the gold wire bonding structure is characterized as a second equivalent circuit physical model; Establish mathematical relationships between lumped distributed parameters and process parameters, and between lumped distributed parameters and electrical parameters; obtain the mathematical relationship between lumped distributed parameters and electrical parameters; characterize the vertical interconnect structure as a third equivalent circuit physical model based on the mathematical relationship between lumped distributed parameters and electrical parameters. The relationships between the insertion depth and contact impedance, the insertion angle and air gap capacitance, and the contact impedance and air gap capacitance and electrical parameters of the RF connector are established to obtain the mathematical relationships between the insertion depth, angle and electrical performance of the RF connector; based on the mathematical relationships between the insertion depth, angle and electrical performance of the RF connector, the assembly parameters of the RF connector are characterized as a fourth equivalent circuit physical model. Obtain electrical performance data of microwave components from RF chip manufacturers; A physical model of microwave component links is constructed based on the principle of microwave network cascading. Based on the first equivalent circuit physical model, the second equivalent circuit physical model, the third equivalent circuit physical model, the fourth equivalent circuit physical model, the electrical performance data of the microwave components provided by the manufacturer, and the physical model of the microwave component link, a microwave component physical model integrating electromechanical coupling effect is constructed.

3. The method according to claim 1, characterized in that, The process of acquiring historical data from the intelligent testing unit and preprocessing the historical data to obtain valid data includes: Acquire historical data from the intelligent testing unit; wherein, the historical data includes: manufacturing parameters and measured electrical performance parameters of the microwave component; The historical data is then cleaned, denoised, reduced in dimensionality, and feature-selected to obtain valid data.

4. The method according to claim 1, characterized in that, The hybrid model, based on the microwave component physical model, parameterized reduced-order data model, and base learner, is constructed using an ensemble learning method, including: Select the meta-learner based on the characteristics of the microwave component; The microwave component physical model, the parameterized reduced-order data model, and the base learner are used as the base model for training. Based on the training results of the base model, the meta-learner is updated and tuned until the optimal meta-learner is obtained, and the optimal meta-learner is output to generate a hybrid model.

5. The method according to claim 1, characterized in that, The unstructured text data based on the microwave component manufacturing process is used to construct a knowledge graph model of the microwave component using knowledge graph technology, including: Unstructured text data is collected during the microwave component manufacturing process; the unstructured text data includes: structural knowledge, fault knowledge, expert experience knowledge, and electrical performance knowledge. The unstructured text data is preprocessed. 25% of the text data is randomly selected from the unstructured text data, and the entity type label of each selected text data is labeled. The labeled text data is then divided into training set and test set in an 8:2 ratio. Utilize graph neural networks to jointly extract entities and relations; The extracted knowledge is constructed into knowledge triples according to the "head entity-relationship-tail entity" structure, and all knowledge triples are stored in the Neo4j graph database to construct the knowledge graph model of the microwave component.

6. The method according to claim 1, characterized in that, The knowledge graph model based on the microwave component reviews and verifies the hybrid model to generate a microwave component physical-data-knowledge fusion model, including: The hybrid model is reviewed and verified using the knowledge graph model of the microwave components; a fusion model is then generated. Real-time production data is collected, and production control is performed based on the fusion model. The parameters of the fusion model are optimized and adjusted according to the control results until the physical-data-knowledge fusion model of the microwave component is generated.

7. An electronic device, characterized in that, include: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method described in any one of claims 1-6.

8. A non-transitory computer-readable storage medium storing computer instructions, characterized in that, The computer instructions are used to cause the computer to perform the method according to any one of claims 1-6.

Images