Design scheme of bonding wire and electronic device, server, storage medium

Abstract

The application provides a bonding wire design scheme and an electronic device, a server and a storage medium, and the scheme comprises the following steps: collecting characteristic data of the bonding wire and the working environment thereof; performing numerical coding on the characteristic data; inputting the numerically coded characteristic data into a design model to obtain a height and a diameter prediction value of the bonding wire, wherein the design model uses an activation function with physical constraints; adjusting the design model by using a reliability weighted loss function, and optimizing the design model by using multi-task learning until the design model meets the process requirement; and outputting the bonding wire design scheme by using the adjusted and optimized design model. The application has the beneficial effects that a systematic, scientific and intelligent solution path is provided for the bonding wire design, which is accurate, reliable, realizable, efficient and sustainable, and is better than the traditional artificial experience method or the simple mathematical optimization method.

Description

Technical Field

[0001] This invention belongs to the field of bonding wire design, and in particular relates to bonding wire design schemes and electronic devices, servers, and storage media. Background Technology

[0002] Currently, aluminum wire bonding is widely used in the manufacturing of low-power devices such as the TO series, DFN series, and IGBT power modules, and plays an indispensable key process and role. During the bonding process, the required bonding quality is achieved through parameter control. In addition, the loop height formed by the aluminum wire between various objects is also a factor that has a significant impact on the bonding quality. During the bonding process, the loop height must also be moderate. If the loop height is too high, the lead may collapse when the lead passes through the subsequent molding process. If the loop height is too low, it may cause a large stress to form inside the bonding point, reducing the reliability of the bonding point.

[0003] In the existing technology, the arc height is defined as the gap between the aluminum wire and the contact surface being greater than twice the wire diameter. However, the loop value is ±500μm, and the maximum loop height is defined in the IGBT industry as not exceeding the potting height. However, the requirements for the loop need to be considered for different objects, connection surfaces, and connection processes. Moreover, each object has different requirements for the thermal cycling and loop path of the loop. Therefore, twice the wire diameter is not an absolute value. Summary of the Invention

[0004] The purpose of this invention is to provide a design scheme for bonding wires and electronic devices, servers, and storage media to solve at least one technical problem existing in the prior art.

[0005] To address the aforementioned technical problems, the first aspect of this invention proposes a design scheme for bonding wires, comprising:

[0006] Collect characteristic data of the bonding wire and its working environment;

[0007] Numerical encoding of feature data;

[0008] The numerically encoded feature data is fed into the design model to obtain the predicted values ​​of the bond wire height and wire diameter. The design model uses an activation function with physical constraints.

[0009] The design model is adjusted using a reliability-weighted loss function and optimized using multi-task learning until the design model meets the process requirements.

[0010] Use the adjusted and optimized design model to output the bond wire design scheme.

[0011] Furthermore, the characteristic data of the bonding wire includes the bonding wire material, current carrying capacity, operating frequency, and package type, and the characteristic data of the operating environment includes thermal cycling and mechanical vibration.

[0012] Furthermore, the process of numerically encoding the feature data includes:

[0013] For bonding wire material and mechanical vibration, they are mapped to low-dimensional numerical vectors; for current carrying capacity, they are mapped to integer labels according to different current requirements; for operating frequency and thermal cycling, they are standardized; for package type, they are conditionally judged and binarized.

[0014] Furthermore, the design model is constructed based on a regression task model;

[0015] The number of neurons in the model's output layer is the same as the number of feature data; the model's hidden layer consists of two fully connected layers, and each fully connected layer has a built-in activation function with physical constraints; the model's output layer has two output nodes corresponding to the height and diameter of the bonding wire, respectively.

[0016] Furthermore, the activation function with physical constraints is specifically as follows:

[0017] Create a monotonically increasing and continuously differentiable function as the initial function;

[0018] Add a normalization constant to the initial function to restrict its output to a finite interval, thus obtaining the activation function;

[0019] The output ranges of the activation functions of the two fully connected layers correspond to the physical constraints of the bond wire height and diameter, respectively.

[0020] Furthermore, the construction process of the reliability-weighted loss function is as follows:

[0021] The weights are calculated based on the reliability information of each sample and combined with the sample prediction error term in the loss function;

[0022] By learning the log-variance of each output, the error penalty intensity is dynamically adjusted for different samples and tasks;

[0023] A physical constraint penalty term is introduced into the loss calculation to apply a smoothing penalty to predicted values ​​that exceed the allowable range of bond wire height or wire diameter.

[0024] Furthermore, the process of optimizing the design model using multi-task learning includes:

[0025] A shared feature extraction network is constructed, with an independent output layer designed for each task; a weighted total loss function is constructed to combine the losses of multiple tasks and incorporate a physical constraint penalty term; the model is trained through an optimizer, and the task weights are dynamically adjusted to balance the importance of tasks.

[0026] A second aspect of the present invention provides an electronic device including a processor and a memory communicatively connected to the processor and used to store processor-executable instructions, the processor being used to execute the bonding wire design scheme described in the first aspect.

[0027] A third aspect of the present invention provides a server comprising at least one processor and a memory communicatively connected to the processor, the memory storing instructions executable by the at least one processor, the instructions being executed by the processor to cause the at least one processor to perform the bonding wire design scheme described in the first aspect.

[0028] The fourth aspect of the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the bonding wire design scheme described in the first aspect.

[0029] Compared with the prior art, the bonding wire design scheme, electronic device, server, and storage medium described in this invention have the following advantages:

[0030] By collecting characteristic data of the bonding wire and working environment (such as material properties, process temperature, stress conditions, etc.), the actual working conditions can be fully reflected. After numerical encoding, the data can be input into the model, enabling the design process to systematically consider multi-dimensional features, improve the prediction accuracy of bonding wire height and diameter, and avoid the errors caused by the reliance on human experience in traditional empirical methods.

[0031] The design model uses activation functions with physical constraints, such as ensuring that wire diameter and height are within allowable process limits, to prevent the model from outputting unreasonable design parameters.

[0032] Using a reliability-weighted loss function to train and adjust the model can optimize the model on key reliability indicators (such as solder joint strength, electrical performance, and fatigue life), avoiding the sacrifice of reliability in the pursuit of geometric or cost optimization, and enhancing the stability of the product in long-term use.

[0033] It can quickly output design solutions, reducing a lot of experimentation and manual calculations. It is especially suitable for multi-batch, multi-specification bonding wire design. It can respond quickly to changes in process parameters and supports batch and automated design. Attached Figure Description

[0034] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0035] Figure 1 This is a schematic diagram illustrating the workflow of the bonding wire design scheme described in an embodiment of the present invention. Detailed Implementation

[0036] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. 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.

[0037] In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0038] Those skilled in the art should understand that the following specific embodiments or implementation methods are a series of optimized configurations listed to further explain the specific content of the invention. These configuration methods can be combined or used in conjunction with each other, unless the invention explicitly states that some or a specific embodiment or implementation method cannot be associated with or used in conjunction with other embodiments or implementation methods. Furthermore, the following specific embodiments or implementation methods are merely optimized configurations and are not intended to limit the scope of protection of the invention.

[0039] The present invention will be further explained below with reference to specific embodiments.

[0040] Example 1:

[0041] like Figure 1 As shown, this embodiment provides a design scheme for bonding wires, including:

[0042] S1. Collect characteristic data of the bonding wire and its working environment;

[0043] S2. Numerically encode the feature data;

[0044] S3. Input the numerically encoded feature data into the design model to obtain the predicted values ​​of the bond wire height and wire diameter. The design model uses an activation function with physical constraints.

[0045] S4. Adjust the design model using a reliability-weighted loss function and optimize the design model using multi-task learning until the design model meets the process requirements.

[0046] S5. Use the adjusted and optimized design model to output the bond wire design scheme.

[0047] The bonding wire characteristic data includes bonding wire material, current carrying capacity, operating frequency, and package type, while the operating environment characteristic data includes thermal cycling and mechanical vibration.

[0048] The process of numerically encoding the feature data includes:

[0049] For bonding wire material and mechanical vibration, they are mapped to low-dimensional numerical vectors; for current carrying capacity, they are mapped to integer labels according to different current requirements; for operating frequency and thermal cycling, they are standardized; for package type, they are conditionally judged and binarized.

[0050] The design model is constructed based on a regression task model;

[0051] The number of neurons in the model's output layer is the same as the number of feature data; the model's hidden layer consists of two fully connected layers, and each fully connected layer has a built-in activation function with physical constraints; the model's output layer has two output nodes corresponding to the height and diameter of the bonding wire, respectively.

[0052] The activation function with physical constraints is specifically as follows:

[0053] Create a monotonically increasing and continuously differentiable function as the initial function;

[0054] Add a normalization constant to the initial function to restrict its output to a finite interval, thus obtaining the activation function;

[0055] The output ranges of the activation functions of the two fully connected layers correspond to the physical constraints of the bond wire height and diameter, respectively.

[0056] In some embodiments, the activation function with physical constraints is specifically as follows:

[0057]

[0058] in, It is the lower limit of the output (such as minimum wire diameter or minimum height). This is the upper limit of the output; It is a normalization constant used to ensure that the output is within the normal range. between;

[0059] when hour, To map the activation function output to a finite interval, a normalization constant needs to be introduced. Define a valid input range ,make: ,Right now At that time, the output is strictly scaled to .

[0060] The derivative of the activation function is ;

[0061] therefore, The fact that the value is always positive indicates that the function is strictly monotonically increasing; nearby, Gradient smoothing; compared to the zero gradient problem of ReLU, this function has a non-zero derivative in the entire domain, thus making training more stable.

[0062] In this specific application scenario, the constrained Softplus activation function ensures that the output wire diameter and height of the model are always within the allowable range of the process, and its smooth boundary guarantees the gradient continuity during backpropagation, preventing learning stagnation due to gradient truncation; when the input features (current, frequency, package type, etc.) change slightly, the output wire diameter and height also change smoothly, which conforms to the laws of real physical systems.

[0063] The process of constructing the reliability-weighted loss function is as follows:

[0064] The weights are calculated based on the reliability information of each sample and combined with the sample prediction error term in the loss function;

[0065] By learning the log-variance of each output, the error penalty intensity is dynamically adjusted for different samples and tasks;

[0066] A physical constraint penalty term is introduced into the loss calculation to apply a smoothing penalty to predicted values ​​that exceed the allowable range of bond wire height or wire diameter.

[0067] In some embodiments, the reliability-weighted loss function is as follows:

[0068] The parameters are defined as follows:

[0069] Sample size ,index The true target vector for each sample The mean of the model predictions The variance predicted by the model simultaneously Sample reliability score Physical upper and lower bounds ; Intensity of constraint and punishment .

[0070] The loss function formula is:

[0071] ;

[0072] in, From the perspective of reliability The obtained sample weights, if the sample has a known measurement variance ,but Normalize it into weights ;

[0073] It is a physical constraint penalty:

[0074] ;

[0075] Softplus() is an activation function with physical constraints.

[0076] The process of optimizing the design model using multi-task learning includes:

[0077] A shared feature extraction network is constructed, with an independent output layer designed for each task; a weighted total loss function is constructed to combine the losses of multiple tasks and incorporate a physical constraint penalty term; the model is trained through an optimizer, and the task weights are dynamically adjusted to balance the importance of tasks.

[0078] Example 2:

[0079] This embodiment provides an electronic device, including a processor and a memory communicatively connected to the processor and used to store executable instructions of the processor, wherein the processor is used to execute the above-described bonding wire design scheme.

[0080] Example 3:

[0081] This embodiment provides a server, including at least one processor and a memory communicatively connected to the processor. The memory stores instructions executable by the at least one processor, which are executed by the processor to cause the at least one processor to execute the bonding wire design scheme.

[0082] Example 4:

[0083] This embodiment provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the design scheme of the bonding wire.

[0084] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing a particular logical function or process, and the scope of the preferred embodiments of this solution includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as will be understood by those skilled in the art to which the embodiments of this solution pertain. The processor performs the various methods and processes described above. For example, the method embodiments of this solution can be implemented as software programs tangibly contained in a machine-readable medium, such as memory. In some embodiments, part or all of the software program can be loaded and / or installed via memory and / or a communication interface. When the software program is loaded into memory and executed by the processor, one or more steps of the methods described above can be performed. Alternatively, in other embodiments, the processor can be configured to perform one of the methods described above by any other suitable means (e.g., by means of firmware).

[0085] The logic and / or steps represented in the flowchart or otherwise described herein may be specifically implemented in any readable storage medium for use by, or in conjunction with, an instruction execution system, apparatus or device (such as a computer-based system, a processor-included system or other system that can fetch and execute instructions from, an instruction execution system, apparatus or device).

[0086] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for outputting bonding wire design schemes, characterized in that, include: Collect characteristic data of the bonding wire and its working environment; Numerical encoding of feature data; The numerically encoded feature data is fed into the design model to obtain the predicted values ​​of the bond wire height and wire diameter. The design model uses an activation function with physical constraints. The design model is adjusted using a reliability-weighted loss function and optimized using multi-task learning until the design model meets the process requirements. Output the bond wire design scheme using the adjusted and optimized design model; The design model is constructed based on a regression task model; The number of neurons in the model's input layer is the same as the number of feature data; the model's hidden layer consists of two fully connected layers, and each fully connected layer has a built-in activation function with physical constraints; the model's output layer has two output nodes corresponding to the height and diameter of the bonding wire, respectively. The activation function with physical constraints is specifically as follows: Create a monotonically increasing and continuously differentiable function as the initial function; Add a normalization constant to the initial function to restrict its output to a finite interval, thus obtaining the activation function; The output ranges of the activation functions of the two fully connected layers correspond to the physical constraints of the bond wire height and diameter, respectively.

2. The method for outputting the bonding wire design scheme according to claim 1, characterized in that: The characteristic data of the bonding wire includes the bonding wire material, current carrying capacity, operating frequency, and package type, while the characteristic data of the operating environment includes thermal cycling and mechanical vibration.

3. The method for outputting a bonding wire design scheme according to claim 2, characterized in that, The process of numerically encoding the feature data includes: For bonding wire material and mechanical vibration, they are mapped to low-dimensional numerical vectors; for current carrying capacity, they are mapped to integer labels according to different current requirements; for operating frequency and thermal cycling, they are standardized; for package type, they are conditionally judged and binarized.

4. The method for outputting a bonding wire design scheme according to claim 1, characterized in that, The process of constructing the reliability-weighted loss function is as follows: The weights are calculated based on the reliability information of each sample and combined with the sample prediction error term in the loss function; By learning the log-variance of each output, the error penalty intensity is dynamically adjusted for different samples and tasks; A physical constraint penalty term is introduced into the loss calculation to apply a smoothing penalty to predicted values ​​that exceed the allowable range of bond wire height or wire diameter.

5. The method for outputting a bonding wire design scheme according to claim 1, characterized in that, The process of optimizing the design model using multi-task learning includes: A shared feature extraction network is constructed, with an independent output layer designed for each task; a weighted total loss function is constructed to combine the losses of multiple tasks and incorporate a physical constraint penalty term; the model is trained through an optimizer, and the task weights are dynamically adjusted to balance the importance of tasks.

6. An electronic device, comprising a processor and a memory communicatively connected to the processor and used for storing processor-executable instructions, characterized in that: The processor is used to execute the bonding wire design output method according to any one of claims 1-5.

7. A server, characterized in that: It includes at least one processor and a memory communicatively connected to the processor, the memory storing instructions executable by the at least one processor, the instructions being executed by the processor to cause the at least one processor to perform the bonding wire design output method as described in any one of claims 1-5.

8. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by the processor, it implements the output method of any of the bonding wire design schemes described in claims 1-5.

Images