Method and device for optimizing intelligent gold bonding wire system based on Internet of Things

Abstract

The invention relates to the technical field of intelligent gold bonding wire systems, in particular to an optimization method and device for achieving an intelligent gold bonding wire system based on the Internet of Things. The optimization method comprises the following steps: acquiring bonding characteristics of bonding points in real time by adopting a multispectral imaging technology; performing feature analysis on the bonding features to obtain feature data; setting a bonding quality scoring system; constructing an ANFIS model, inputting feature data, and outputting a bonding quality score; performing importance analysis on the feature data by using a random forest algorithm to obtain a key feature combination influencing the bonding quality; and formulating an optimization scheme. Through real-time data acquisition of the multispectral imaging technology, the accuracy and the real-time performance of the data are improved. And through feature analysis and fuzzy rule construction, a decision can be made more effectively. And the ANFIS model is introduced, so that the processing capability of complex data is improved, and intelligent evaluation of the bonding quality is also realized.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent bonding wire systems, and particularly to an optimization method for realizing an intelligent bonding wire system based on the Internet of Things and an optimization device for realizing an intelligent bonding wire system based on the Internet of Things. Background Art

[0002] Bonding is a connection technology mainly used for the interconnection between circuit components to ensure the transmission of electrical signals and mechanical connection. It can achieve the encapsulation of microelectronic devices, sensors, radio frequency components, etc. The main bonding technologies include: thermocompression bonding, ultrasonic bonding, and laser bonding. Bonding wires are generally made of gold (Au), aluminum (Al), or copper (Cu). Gold wires made of different materials have different physical and chemical properties. The bonding wire system is mainly applied to semiconductor packaging, optoelectronic devices, or microelectromechanical systems.

[0003] In the gold wire bonding system, during the data acquisition process, the bonding system usually relies on manual monitoring, lacking accuracy and real-time performance. This may lead to incorrect data analysis and decision-making, thus affecting production efficiency. The complex production environment may make it difficult to adjust production parameters. When employees face emergencies, they often have difficulty making effective decisions quickly, which may cause production stagnation or unqualified products. Relying on manual inspection or simple statistical analysis cannot effectively handle complex data, resulting in an incomplete evaluation of bonding quality and affecting the consistency and reliability of products. Summary of the Invention

[0004] The present invention provides an optimization method and device for realizing an intelligent bonding wire system based on the Internet of Things to solve the defects in the prior art.

[0005] On the one hand, the present invention provides an optimization method for realizing an intelligent bonding wire system based on the Internet of Things, including:

[0006] S1. Using multi-spectral imaging technology to collect the bonding characteristics of bonding points in real time, where the bonding characteristics include geometric characteristics, mechanical characteristics, and electrical characteristics.

[0007] S2. Analyzing the bonding characteristics to obtain characteristic data, where the characteristic data includes appearance factor, smooth factor, tensile strength factor, and electrical conductivity factor.

[0008] S3. Constructing a fuzzy rule for bonding quality based on the Internet of Things, and setting a bonding quality scoring system according to historical bonding characteristics.

[0009] S4. Constructing an ANFIS model, inputting the characteristic data, and outputting the bonding quality score.

[0010] S5. Using the random forest algorithm to analyze the importance of the characteristic data and obtain the key characteristic combination affecting the bonding quality.

[0011] S6. Develop an optimization plan based on the combination of key features.

[0012] S7. Implement the optimization plan and verify the optimization effect.

[0013] According to an optimization method for realizing an intelligent bonding wire system based on the Internet of Things provided by the present invention, the process of feature analysis of geometric features includes: performing principal component analysis based on wavelet transform on a multi-spectral image, extracting high-frequency and low-frequency components of geometric features, and performing feature analysis on the high-frequency and low-frequency components to obtain an appearance factor and a smooth factor.

[0014] According to an optimization method for realizing an intelligent bonding wire system based on the Internet of Things provided by the present invention, the process of feature analysis of mechanical features includes: adopting a spectral analysis method based on density functional theory to extract the fine spectrum of a multi-spectral image, adopting a finite element analysis method based on non-local continuum mechanics to construct a mechanical model based on the fine spectrum, and obtaining a tensile strength factor through a constitutive model combined with the microscopic damage evolution law of the bonding wire.

[0015] According to an optimization method for realizing an intelligent bonding wire system based on the Internet of Things provided by the present invention, the process of feature analysis of electrical features includes: adopting a method combining independent component analysis and blind source separation to extract the independent components of resistance, capacitance, and inductance in a multi-spectral image, using impedance spectrum analysis technology based on transmission line theory to represent resistance, capacitance, and inductance in complex form, adopting a generalized Fourier series expansion based on complex variable functions to extract fine electrical parameter features, and introducing a frequency-weighted information entropy metric to obtain a conductivity factor.

[0016] According to an optimization method for realizing an intelligent bonding wire system based on the Internet of Things provided by the present invention, in step S3, the process of setting a bonding quality scoring system includes:

[0017] S31. Collect historical bonding features, set a fuzzy set for each historical feature data, and the fuzzy set includes poor, medium, and excellent.

[0018] S32. Establish fuzzy rules according to the historical data distribution, and the fuzzy rules include different combinations of feature data and corresponding bonding qualities, and the bonding qualities include low and high.

[0019] S33. Set a membership function for each fuzzy set, and the membership function is a Gaussian function.

[0020] S34. According to the historical feature data, use the Maksyms fuzzy inference method to obtain a fuzzy output.

[0021] S35. Use the centroid method to convert the fuzzy output into a bonding quality score.

[0022] According to an optimization method for realizing an intelligent bonding wire system based on the Internet of Things provided by the present invention, in step S4, the process of constructing an ANFIS model includes:

[0023] Construct a training data set, which includes a combination of feature data and the bonding quality score corresponding to the combination of feature data.

[0024] Construct an ANFIS model, including a fuzzification layer, a rule firing strength calculation layer, a normalization layer, an output function calculation layer, and a total output calculation layer. The fuzzification layer is used to convert feature data into fuzzy membership values. The rule firing strength calculation layer is used to calculate the firing strength of each fuzzy rule. The normalization layer is used to normalize the rule firing strength. The output function calculation layer is used to calculate the output function of each rule. The total output calculation layer is used to sum up the outputs of all rules to obtain the bonding quality score.

[0025] Use the training data set to train the ANFIS model and adjust the model parameters, where the model parameters include the parameters of the membership function and the parameters of the output function.

[0026] According to an optimization method for realizing an intelligent bonding wire system based on the Internet of Things provided by the present invention, in step S5, the process of performing importance analysis on feature data using the random forest algorithm includes:

[0027] Randomly sample from the feature data through the Bootstrap method to form multiple subsets.

[0028] Train a decision tree for each subset. At each node, randomly select a feature subset and select the best splitting feature.

[0029] Calculate the Gini index contribution of each feature data in all tree splits for importance evaluation.

[0030] According to an optimization method for realizing an intelligent bonding wire system based on the Internet of Things provided by the present invention, in step S5, the process of obtaining the key feature combination affecting the bonding quality includes: Summarize the Gini index values of all feature data, and use the feature data with a Gini index value higher than the preset threshold as the key feature combination.

[0031] According to an optimization method for realizing an intelligent bonding wire system based on the Internet of Things provided by the present invention, in step S6, the optimization scheme includes adjusting production parameters and introducing a control mechanism. The production parameters include bonding conditions and material quality, and the bonding conditions include bonding temperature, bonding pressure, and bonding time. The control mechanism includes a real-time monitoring system and a feedback mechanism.

[0032] On the other hand, the present invention also provides an optimization device for realizing an intelligent bonding wire system based on the Internet of Things, including:

[0033] A bonding feature acquisition module is used to acquire the bonding features of bonding points in real time. The bonding features include geometric features, mechanical features, and electrical features.

[0034] A feature analysis module analyzes the acquired bonding features to obtain feature data. The feature data includes appearance factor, smoothness factor, tensile strength factor, and electrical conductivity factor.

[0035] A fuzzy rule construction module is used to construct fuzzy rules for bonding quality based on the Internet of Things, and set a bonding quality scoring system according to historical bonding features.

[0036] An ANFIS model construction module is used to construct an ANFIS model, input the feature data, and output the bonding quality score.

[0037] A key feature combination acquisition module is used to perform importance analysis on the feature data using the random forest algorithm to obtain the key feature combination that affects the bonding quality.

[0038] An optimization plan formulation module is used to formulate an optimization plan according to the key feature combination.

[0039] An optimization plan implementation module is used to implement the optimization plan and verify the optimization effect.

[0040] An optimization method and device for an intelligent bonding wire system based on the Internet of Things provided by the present invention improve the accuracy and real-time performance of data through real-time data acquisition of multi-spectral imaging technology, enabling enterprises to promptly master key parameters during the production process. Subsequent feature analysis and fuzzy rule construction enable enterprises to make more effective decisions in the face of complex production environments and reduce losses caused by human errors. The introduction of the ANFIS model not only enhances the processing ability of complex data but also realizes intelligent evaluation of bonding quality. This provides a higher level of intelligent monitoring for production. In addition, in the feature importance analysis, the application of the random forest algorithm makes the identification of key features and the formulation of optimization plans more scientific, thereby improving the utilization rate of resources and the pertinence of production. Finally, through the implementation and verification of the optimization plan, the entire production process will form a data-driven and continuously improving closed-loop system. This method achieves comprehensive coverage from data acquisition to evaluation feedback, significantly improves product quality assurance and production efficiency, and enhances the market competitiveness of enterprises. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] To more clearly illustrate the technical solutions in the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0042] Figure 1 It is a schematic flowchart of an optimization method for an intelligent bonding wire system based on the Internet of Things provided by an embodiment of the present invention;

[0043] Figure 2 It is a schematic structural diagram of an optimization device for an intelligent bonding wire system based on the Internet of Things provided by an embodiment of the present invention. Detailed implementation manners

[0044] To make the objectives, technical solutions and advantages of the present invention clearer, the following will clearly and completely describe the technical solutions in the present invention in conjunction with the drawings in the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts fall within the scope of protection of the present invention.

[0045] The following Figure 1 - Figure 2 describes an optimization method and device for an intelligent bonding wire system based on the Internet of Things of the present invention.

[0046] Figure 1 It is a schematic flowchart of an optimization method for an intelligent bonding wire system based on the Internet of Things provided by an embodiment of the present invention.

[0047] As Figure 1 shown, an optimization method and device for an intelligent bonding wire system based on the Internet of Things provided by an embodiment of the present invention, the execution subject can be an optimization method for an intelligent bonding wire system based on the Internet of Things, including:

[0048] S1. Adopt multi-spectral imaging technology to collect the bonding characteristics of the bonding points in real time. The bonding characteristics include geometric characteristics, mechanical characteristics and electrical characteristics.

[0049] In this embodiment, multi-spectral imaging technology is an imaging technology that utilizes light of different wavelengths. This technology can obtain rich information about the object to be detected by analyzing the spectral information of different bands. Each material has unique reflection, absorption, and transmission characteristics at different spectral wavelengths. These characteristics are called "spectral features", and they can be used to distinguish different materials and states. The multi-spectral imaging system irradiates the object with light sources of multiple bands and receives the reflected light, thereby obtaining multiple sets of spectral data. The multi-spectral imaging system is equipped with multiple sensors, and each sensor is responsible for capturing images within a specific wavelength range. The combination of these images will form a multi-dimensional data set, usually called a "spectral cube". Each pixel in the spectral cube contains information of multiple wavelengths, providing detailed spectral information about the material at that point. The collected multi-spectral data is processed by specific algorithms to extract different characteristic information, including geometric features, mechanical features, and electrical features. These features can be analyzed through data preprocessing, feature extraction, and modeling methods, and finally form data information available for subsequent analysis. For example, the geometric shape of the bonding point is identified through image processing technology, and its mechanical and electrical properties are analyzed using spectral data. Multi-spectral imaging technology can perform real-time monitoring during the production process, quickly feedback data results, and support dynamic decision-making. This enables the production party to promptly understand the characteristics of the bonding point and handle it in a timely manner when abnormalities are found, ensuring production quality.

[0050] Adopting multi-spectral imaging technology to collect the bonding characteristics of the bonding point in real time can obtain more detailed and high-precision data. This technology not only improves the accuracy of data collection but also can simultaneously obtain multi-dimensional features such as geometry, mechanics, and electricity. By integrating these features, a solid foundation is laid for subsequent data analysis. In addition, real-time monitoring can promptly detect potential problems, reduce errors and losses during the production process, and thus improve the overall production efficiency and product quality.

[0051] S2. Conduct feature analysis on the bonding characteristics to obtain feature data, including:

[0052] S21. Conduct feature analysis on the geometric features to obtain the appearance factor and the smoothness factor.

[0053] The process of conducting feature analysis on the geometric features includes: performing principal component analysis based on wavelet transform on the multi-spectral image, extracting the high-frequency and low-frequency components of the geometric features, and conducting feature analysis on the high-frequency and low-frequency components to obtain the appearance factor and the smoothness factor.

[0054] For the multi-spectral image I(x, y), the formula for wavelet transform is expressed as:

[0055]

[0056] In the formula, Let \(\psi\) denote the wavelet basis function, \(a\) denote the scale parameter that controls the stretching and shrinking of the wavelet, and \(b\) denote the translation parameter that controls the position of the wavelet. Through wavelet transform, a multispectral image can be decomposed into wavelet coefficients at different scales and positions, and these coefficients represent the information of the image at different frequencies and spatial positions.

[0057] Perform principal component analysis on the image coefficients after wavelet transform. The purpose is to find the main feature directions of the data, reduce the data dimension while retaining most of the information.

[0058] Let the image coefficient matrix after wavelet transform be \(X\), with size \(n\times m\) (\(n\) is the number of samples, i.e., the number of pixels in the image, and \(m\) is the number of variables, i.e., the number of multispectral bands).

[0059] Calculate the covariance matrix \(G\):

[0060] Then solve for the eigenvalues \(\lambda\) i and eigenvectors \(v\) i (\(i = 1, 2, \ldots, m\)): \(Gv\) i =\(\lambda\) i \(v\) i .

[0061] Arrange the eigenvalues in descending order, and select the eigenvectors \(V\) corresponding to the top \(k\) largest eigenvalues k =\([v_1, v_2, \ldots, v\) k to construct the projection matrix.

[0062] Project the original data \(X\) to obtain the principal component score matrix \(M\): \(M = XV\) k .

[0063] Through wavelet transform and principal component analysis, the high-frequency and low-frequency components of the geometric features are extracted. The high-frequency components mainly contain the detailed information of the image, and the low-frequency components mainly contain the contour information of the image.

[0064] The appearance factor is obtained by analyzing the principal components related to the image appearance in the principal component score matrix \(M\), and is expressed by the formula:

[0065]

[0066] In the formula, \(F\) a denotes the appearance factor, \(A\) denotes the area of the bonding point, and \(P\) denotes the perimeter of the bonding point.

[0067] The smoothness factor is obtained by analyzing the high-frequency components of the image. The high-frequency components reflect the detailed information of the image. The smoother the surface of the bonding point, the lower the energy of the high-frequency components.

[0068] Let the high-frequency coefficient matrix after wavelet transform be \(H\), and the smoothness factor formula is expressed as:

[0069]

[0070] In the formula, F s represents the smoothness factor, and ∑ i，j |H i，j | represents the sum of the absolute values of all elements of the high-frequency coefficient matrix, max(|H|) represents the maximum absolute value in the high-frequency coefficient matrix, and N represents the total number of elements in the high-frequency coefficient matrix.

[0071] The smaller the value of the smoothness factor, the smoother the surface of the bonding point. The larger the value, the rougher the surface.

[0072] S22. Conduct a characteristic analysis of the mechanical characteristics to obtain the tensile strength factor.

[0073] The process of conducting a characteristic analysis of the mechanical characteristics includes: using a spectroscopic analysis method based on density functional theory to extract the fine spectrum of the multispectral image. The formula of density functional theory is expressed as:

[0074] E = E[ρ] + ∫v e (r)ρ(r)dr + F[ρ]

[0075] In the formula, E represents the elastic modulus of the gold wire, E[ρ] represents the functional based on the electron density ρ, v e represents the external potential, and F[ρ] represents the exchange-correlation energy functional.

[0076] Adopt a finite element analysis method based on non-local continuum mechanics to construct a mechanical model based on the fine spectrum, and the formula is expressed as:

[0077] σ = E * θ

[0078] In the formula, σ represents the stress, and θ represents the strain.

[0079] Since the gold wire will exhibit non-linear behavior during the stretching process, a constitutive model and large-scale parallel computing can be adopted, and the formula is expressed as:

[0080] σ = E * θ; (when θ < θ y )

[0081] σ = σ y + S(θ - θ y ); (when θ ≥ θ y )

[0082] In the formula, σ y represents the yield strength, S represents the hardening modulus, and θ y represents the yield strain.

[0083] Combined with the microscopic damage evolution law of the bonding gold wire, the formula is expressed as:

[0084]

[0085] In the formula, D represents the damage variable, D0 represents the initial damage degree, and g(θ′) represents the relationship between the damage development rate and the strain.

[0086] The tensile strength factor is obtained, and the formula is expressed as:

[0087] F l = f(E, θ y , H, D)

[0088] In the formula, F l represents the tensile strength factor, E represents the elastic modulus of the gold wire, θ y represents the yield strain, S represents the hardening modulus, and D represents the damage variable.

[0089] S23. Feature analysis is performed on the electrical characteristics to obtain the conductivity characteristic factor.

[0090] The process of feature analysis of the electrical characteristics includes: using the independent component analysis combined with the blind source separation method to extract the independent components of resistance, capacitance, and inductance in the multi-spectral image, and applying the impedance spectrum analysis technology based on the transmission line theory to represent the resistance, capacitance, and inductance in complex form. The formula is expressed as:

[0091]

[0092] In the formula, R represents the resistance, C represents the capacitance, L represents the inductance, j represents the imaginary unit, and ω represents the angular frequency.

[0093] The generalized Fourier series expansion based on complex variable functions is used to extract the fine electrical parameter characteristics. The formula is expressed as:

[0094]

[0095] Among them, c n represents the Fourier coefficient, and the formula is expressed as:

[0096]

[0097] In the formula, T0 represents the period.

[0098] The frequency-weighted information entropy measure is introduced. The formula is expressed as:

[0099]

[0100] In the formula, p i represents the weight related to the frequency.

[0101] The conductivity characteristic factor is obtained, and the formula is expressed as:

[0102]

[0103] Wherein, F e represents the conductive characteristic factor.

[0104] In this embodiment, using the plural form to represent resistance, capacitance and inductance is to effectively describe the amplitude and phase of the AC signal at the bonding point, which is convenient for calculation and analysis. Using the generalized Fourier series expansion can extract the frequency components of the conductive signal at the bonding point, helping to understand the response of the conductive characteristics of the bonding point at different frequencies. Introducing the information entropy metric can quantify the complexity of the conductive signal at the bonding point, which is helpful for selecting the most important features. These steps work together to ensure the accurate extraction of the conductive characteristic factor and the improvement of system performance.

[0105] By deeply analyzing the collected bonding features, representative feature data can be refined. These data include appearance factor, smoothness factor, tensile strength factor and conductive characteristic factor, which are important indicators for evaluating bonding quality. The results of feature analysis can help identify the key factors affecting bonding quality and provide reliable data support for subsequent steps. At the same time, it also makes the entire data processing process more systematic and ensures the scientific nature of the evaluation process.

[0106] S3. Construct a fuzzy rule for bonding quality based on the Internet of Things, and set up a bonding quality scoring system according to historical bonding features.

[0107] The process of setting up the bonding quality scoring system includes:

[0108] S31. Collect historical bonding features, and set up fuzzy sets for each historical feature data. The fuzzy sets include poor, medium and excellent.

[0109] S32. Establish fuzzy rules according to the historical data distribution. The fuzzy rules include different combinations of feature data and the corresponding bonding quality. The bonding quality includes low and high.

[0110] S33. Set up membership functions for each fuzzy set. The membership function is a Gaussian function.

[0111] S34. According to the historical feature data, use the Maksyms fuzzy inference method to obtain a fuzzy output.

[0112] S35. Use the centroid method to convert the fuzzy output into a bonding quality score.

[0113] In this embodiment, constructing a fuzzy rule for bonding quality based on the Internet of Things can effectively integrate historical data and currently collected data. By setting a reasonable scoring system, the complex bonding quality can be quantified, making it more intuitive during evaluation. This process not only improves the scientific nature of decision-making but also reduces subjective judgment when dealing with complex working conditions, providing a clear basis for formulating optimization plans.

[0114] S4. Construct an ANFIS model, input the feature data, and output the bonding quality score.

[0115] The process of constructing an ANFIS model includes:

[0116] Construct a training data set, which includes combinations of feature data and the corresponding bonding quality scores for the combinations of feature data.

[0117] Construct an ANFIS model, including a fuzzification layer, a rule firing strength calculation layer, a normalization layer, an output function calculation layer, and a total output calculation layer. The fuzzification layer is used to convert the feature data into fuzzy membership values. The rule firing strength calculation layer is used to calculate the firing strength of each fuzzy rule. The normalization layer is used to normalize the rule firing strength. The output function calculation layer is used to calculate the output function of each rule. The total output calculation layer is used to sum up the outputs of all rules to obtain the bonding quality score.

[0118] Use the training data set to train the ANFIS model and adjust the model parameters, where the model parameters include the parameters of the membership function and the parameters of the output function.

[0119] In this embodiment, constructing an Adaptive Neuro-Fuzzy Inference System (ANFIS) model can intelligently process the extracted feature data and generate the corresponding bonding quality scores. This model combines the advantages of fuzzy logic and neural networks, greatly enhancing the ability to process complex and uncertain data. Through the ANFIS model, the real-time data in the production process can be effectively analyzed, thereby realizing the dynamic monitoring of bonding quality, making timely adjustments, and improving the consistency and reliability of products.

[0120] S5. Use the random forest algorithm to perform importance analysis on the feature data to obtain the key feature combinations affecting the bonding quality.

[0121] The process of using the random forest algorithm to perform importance analysis on the feature data includes:

[0122] Randomly sample from the feature data through the Bootstrap method to form multiple subsets.

[0123] Train a decision tree for each subset. At each node, randomly select a feature subset and select the best splitting feature.

[0124] Calculate the Gini index contribution of each feature data in all tree splits for importance assessment. The formula is expressed as:

[0125]

[0126] In the formula, T represents the number of generated trees, and Gini(X i ) represents the Gini index value of the split point related to the feature data X i .

[0127] The process of obtaining the key feature combination that affects the bonding quality includes: summarizing the Gini index values of all feature data, and taking the feature data with Gini index values higher than the preset threshold as the key feature combination.

[0128] In this embodiment, using the random forest algorithm for feature importance analysis can efficiently identify the feature combination that has the greatest impact on the bonding quality. This method not only improves the accuracy of the analysis but also avoids the bias that may occur in traditional single-feature analysis. By clearly identifying the key features, the optimization plan can be more targeted, resulting in higher resource utilization and ultimately a decrease in the overall production cost.

[0129] S6. Develop an optimization plan according to the key feature combination.

[0130] The optimization plan includes adjusting production parameters and introducing control mechanisms. The production parameters include bonding conditions and material quality, and the bonding conditions include bonding temperature, bonding pressure, and bonding time. The control mechanisms include a real-time monitoring system and a feedback mechanism.

[0131] In this embodiment, the optimization plan developed according to the key feature combination is the core of the entire system. By quantifying the impact of key features, the optimization plan can be adjusted in multiple dimensions to ensure the achievement of the expected quality improvement. These plans can not only remedy the current problems but also provide guidelines for long-term production efficiency improvement, helping enterprises maintain their competitiveness.

[0132] S7. Implement the optimization plan and verify the optimization effect.

[0133] In this embodiment, before implementing the optimization plan, it is first necessary to clarify the optimization objectives, such as improving bonding quality, reducing defect rates, or enhancing production efficiency. Determine the required human, material, and technical resources to ensure that all resources can be efficiently invested in the optimization implementation according to the plan. Develop a detailed schedule, clarify the time nodes and responsible persons for each stage, and ensure the orderly progress of the implementation process. According to the formulated optimization plan, deploy the necessary technical means. For example, if the optimization plan mentions changing production parameters or adjusting processes, it is necessary to ensure that the equipment can support it. During the implementation process, monitor and record key parameters in real time to ensure that all measures are implemented. After implementing the optimization plan, adopt a multi-spectral imaging technology similar to that in step S1 to collect the bonding feature data in the new production batches in real time. These data should include appearance factors, smoothness factors, tensile strength factors, and conductive property factors before and after optimization. Use data analysis methods to compare the changes in feature data before and after optimization, and evaluate whether the key indicators have achieved the expected effects. According to the pre-set KPIs, quantitatively and qualitatively evaluate the implementation effects. Such as the decrease in defect rates, the improvement of production efficiency, and the improvement of customer feedback. Use the bonding quality scores output by the ANFIS model for comparison to ensure that the optimization measures can effectively improve the quality level of the product.

[0134] Implementing the optimization plan and verifying the effects are key steps to ensure the effectiveness of the strategy. By continuously monitoring and evaluating the optimization effects, not only can problems be discovered in a timely manner, but also the production process can be continuously improved. The feedback from the verification process provides valuable experience for subsequent improvements, forming a virtuous cycle. The success or failure of this step is directly related to the effectiveness of the entire optimization plan, and helps to enhance the market competitiveness and profitability of the enterprise's products.

[0135] In summary, this embodiment provides an optimization method and device for an intelligent bonding wire system based on the Internet of Things. Through the real-time data collection of multi-spectral imaging technology, the accuracy and real-time nature of the data are improved, enabling the enterprise to promptly grasp key parameters during the production process. Subsequent feature analysis and fuzzy rule construction enable the enterprise to make more effective decisions in the face of a complex production environment and reduce losses caused by human errors. The introduction of the ANFIS model not only enhances the processing ability of complex data but also realizes the intelligent evaluation of bonding quality. This provides a higher level of intelligent monitoring for production. In addition, in the feature importance analysis, the application of the random forest algorithm makes the identification of key features and the formulation of optimization plans more scientific, thereby improving the utilization rate of resources and the pertinence of production. Finally, through the implementation and verification of the optimization plan, the entire production process will form a data-driven and continuously improving closed-loop system. This method achieves comprehensive coverage from data collection to evaluation feedback, significantly improves product quality assurance and production efficiency, and enhances the market competitiveness of the enterprise.

[0136] Based on the same general inventive concept, the present invention also protects an optimization device for an intelligent bonding wire system implemented based on the Internet of Things. The following describes an optimization device for an intelligent bonding wire system implemented based on the Internet of Things provided by the present invention. The optimization device for an intelligent bonding wire system implemented based on the Internet of Things described below can be correspondingly referred to with the optimization method for an intelligent bonding wire system implemented based on the Internet of Things described above.

[0137] Figure 2 It is a schematic structural diagram of an optimization device for an intelligent bonding wire system implemented based on the Internet of Things provided by an embodiment of the present invention.

[0138] As Figure 2 shown, the optimization device for an intelligent bonding wire system implemented based on the Internet of Things includes:

[0139] A bonding feature acquisition module, configured to acquire the bonding features of the bonding points in real time, and the bonding features include geometric features, mechanical features, and electrical features.

[0140] A feature analysis module, which performs feature analysis on the acquired bonding features to obtain feature data, and the feature data includes appearance factor, smooth factor, tensile strength factor, and electrical conductivity factor.

[0141] A fuzzy rule construction module, configured to construct a fuzzy rule for bonding quality based on the Internet of Things, and set a bonding quality scoring system according to historical bonding features.

[0142] An ANFIS model construction module, configured to construct an ANFIS model, input the feature data, and output the bonding quality score.

[0143] A key feature combination acquisition module, configured to perform importance analysis on the feature data using the random forest algorithm to obtain the key feature combination that affects the bonding quality.

[0144] An optimization plan formulation module, configured to formulate an optimization plan according to the key feature combination.

[0145] An optimization plan implementation module, configured to implement the optimization plan and verify the optimization effect.

[0146] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the essence of the above technical solution, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0147] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An optimization method for implementing an intelligent bonding wire system based on the Internet of Things, characterized in that, Including: S1. Using multi-spectral imaging technology to collect the bonding characteristics of bonding points in real time, where the bonding characteristics include geometric characteristics, mechanical characteristics, and electrical characteristics; S2. Conducting feature analysis on the bonding characteristics to obtain feature data, where the feature data includes appearance factor, smooth factor, tensile strength factor, and electrical conductivity factor; S3. Constructing a fuzzy rule for bonding quality based on the Internet of Things, and setting up a bonding quality scoring system according to historical bonding characteristics; S4. Constructing an ANFIS model, inputting the feature data, and outputting the bonding quality score; S5. Using the random forest algorithm to conduct importance analysis on the feature data to obtain the key feature combination affecting the bonding quality; S6. Formulating an optimization plan according to the key feature combination; S7. Implementing the optimization plan and verifying the optimization effect.

2. The optimization method of an intelligent bonding wire system based on the Internet of Things according to claim 1, characterized in that The process of conducting feature analysis on the geometric characteristics includes: conducting principal component analysis based on wavelet transform on the multi-spectral image, extracting the high-frequency and low-frequency components of the geometric characteristics, and conducting feature analysis on the high-frequency and low-frequency components to obtain the appearance factor and smooth factor.

3. An optimization method for an intelligent bonding wire system based on the Internet of Things according to claim 1, characterized in that, The process of conducting feature analysis on the mechanical characteristics includes: using a spectral analysis method based on density functional theory to extract the fine spectrum of the multi-spectral image, using a finite element analysis method based on non-local continuum mechanics to construct a mechanical model based on the fine spectrum, and obtaining the tensile strength factor through the constitutive model combined with the microscopic damage evolution law of the bonding wire.

4. An optimization method for an intelligent bonding wire system based on the Internet of Things according to claim 1, characterized in that, The process of conducting feature analysis on the electrical characteristics includes: using a method combining independent component analysis and blind source separation to extract the independent components of resistance, capacitance, and inductance in the multi-spectral image, using impedance spectrum analysis technology based on transmission line theory to represent the resistance, capacitance, and inductance in complex form, using the generalized Fourier series expansion based on complex variable functions to extract fine electrical parameter characteristics, and introducing a frequency-weighted information entropy metric to obtain the electrical conductivity factor.

5. An optimization method for an intelligent bonding wire system based on the Internet of Things according to claim 1, characterized in that, In step S3, the process of setting up the bonding quality scoring system includes: S31. Collecting historical bonding characteristics, and setting up a fuzzy set for each historical feature data, where the fuzzy set includes poor, medium, and excellent; S32. Establishing fuzzy rules according to the historical data distribution, where the fuzzy rules include different combinations of feature data and the corresponding bonding quality, and the bonding quality includes low and high; S33. Setting up a membership function for each fuzzy set, where the membership function is a Gaussian function; S34. Using the Maksyms fuzzy inference method according to the historical feature data to obtain a fuzzy output; S35. Using the centroid method to convert the fuzzy output into a bonding quality score.

6. The optimization method of an intelligent bonding wire system based on the Internet of Things according to claim 5, characterized in that, In step S4, the process of constructing the ANFIS model includes: Constructing a training data set, where the training data set includes combinations of feature data and the bonding quality scores corresponding to the combinations of feature data; Construct an ANFIS model, including a fuzzification layer, a rule excitation strength calculation layer, a normalization layer, an output function calculation layer, and a total output calculation layer; the fuzzification layer is used to convert the feature data into fuzzy membership values; the rule excitation strength calculation layer is used to calculate the excitation strength of each fuzzy rule; the normalization layer is used to normalize the rule excitation strength; the output function calculation layer is used to calculate the output function of each rule; the total output calculation layer is used to sum up the outputs of all rules to obtain a bonding quality score. Use the training data set to train the ANFIS model and adjust the model parameters, where the model parameters include the parameters of the membership function and the parameters of the output function.

7. An optimization method for an intelligent bonding wire system based on the Internet of Things according to claim 1, characterized in that In step S5, the process of performing importance analysis on the feature data using the random forest algorithm includes: Randomly sample from the feature data through the Bootstrap method to form multiple subsets; Train a decision tree for each subset. At each node, randomly select a feature subset and select the best splitting feature; Calculate the Gini index contribution of each piece of the feature data in all tree splits for importance evaluation.

8. An optimization method for an intelligent bonding wire system based on the Internet of Things according to claim 1, characterized in that, In step S5, the process of obtaining the key feature combination affecting the bonding quality includes: Summarize the Gini index values of all feature data, and use the feature data with a Gini index value higher than the preset threshold as the key feature combination.

9. The optimization method of an intelligent bonding wire system based on the Internet of Things according to claim 1, characterized in that, In step S6, the optimization scheme includes adjusting production parameters and introducing a control mechanism; the production parameters include bonding conditions and material quality, and the bonding conditions include bonding temperature, bonding pressure, and bonding time; the control mechanism includes a real-time monitoring system and a feedback mechanism.

10. An optimization device for implementing an intelligent bonding wire system based on the Internet of Things, adopting the optimization method for implementing an intelligent bonding wire system based on the Internet of Things as described in any one of claims 1 to 9, characterized in that, The optimization device for the intelligent bonding wire system based on the Internet of Things includes: A bonding feature acquisition module, used to collect the bonding features of the bonding points in real time, where the bonding features include geometric features, mechanical features, and electrical features; A feature analysis module, which performs feature analysis on the collected bonding features to obtain feature data, where the feature data includes an appearance factor, a smoothness factor, a tensile strength factor, and a conductivity factor; A fuzzy rule construction module, used to construct fuzzy rules for bonding quality based on the Internet of Things, and set a bonding quality scoring system according to historical bonding features; An ANFIS model construction module, used to construct an ANFIS model, input the feature data, and output a bonding quality score; A key feature combination acquisition module, used to perform importance analysis on the feature data using the random forest algorithm to obtain the key feature combination affecting the bonding quality; An optimization scheme formulation module, used to formulate an optimization scheme according to the key feature combination; An optimization scheme implementation module, used to implement the optimization scheme and verify the optimization effect.

Images