MEMS device modeling and yield prediction method using machine learning
MEMS device modeling and yield prediction are solved through machine learning methods, and the problems of increasing yield iterations and cost improvements in MEMS device manufacturing are achieved, efficient and accurate yield prediction and optimization are achieved, and the performance and reliability of the device are improved.
Patent Information
- Application Number
- CN202510195181.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-02-21
- Publication Date
- 2025-05-16
AI Technical Summary
In the process of large-scale manufacturing of MEMS devices, it is difficult for the prior art to effectively improve yield, resulting in an increase in the number of yield iterations and an increase in cost.
Machine learning methods are used to model MEMS devices and yield prediction. By determining the target performance parameters and device parameters, obtaining sample performance parameters, training performance parameter prediction models, obtaining samples to be predicted based on preset sampling methods, predicting their performance parameters, and determining device yield.
It significantly improves the yield iteration efficiency of MEMS manufacturing, reduces manufacturing costs, optimizes the matching between design and manufacturing, reduces the number of yield iterations, and improves the performance consistency and reliability of the device.
Smart Images
Figure CN120012521A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of micro-electromechanical systems, and in particular to a MEMS device modeling and yield prediction method using machine learning. Background Art
[0002] In recent years, the MEMS (Microelectromechanical Systems) industry has developed rapidly, the market has gradually expanded, and factory shipments have increased significantly. To ensure cost-effectiveness, MEMS device manufacturers / foundries need to increase the yield as much as possible during large-scale manufacturing. Since the improvement of yield requires continuous iteration of design-tapeout-testing, the time and cost of yield improvement gradually increase with the number of iterations. In order to reduce the number and cost of yield iterations and improve the efficiency of yield iterations, a yield prediction and optimization method for MEMS devices is urgently needed. Summary of the invention
[0003] The present invention provides a MEMS device modeling and yield prediction method using machine learning to at least partially solve the above problems.
[0004] A first aspect of the present invention provides a method for MEMS device modeling and yield prediction using machine learning, the method comprising: Determining target performance parameters of a target device and device parameters corresponding to the target performance parameters; Obtaining samples corresponding to the device parameters and sample performance parameters corresponding to the samples; The machine learning model is trained with the sample as input and the corresponding sample performance parameter as output to obtain a performance parameter prediction model; Using a preset sampling method, based on the target deviation, a plurality of samples to be predicted corresponding to the device parameters are obtained, and the performance parameters corresponding to each sample to be predicted are determined based on the performance parameter prediction model; Based on the performance parameters of a plurality of samples to be predicted and preset classification standards, the device yield corresponding to the target deviation of the preset sampling method is determined.
[0005] Optionally, the method further comprises: When the device yield is greater than or equal to a preset device yield, the target deviation is determined to be a qualified deviation.
[0006] Optionally, the preset adoption method is normal distribution sampling, and the preset sampling method is adopted to obtain a plurality of samples to be predicted corresponding to the device parameters based on the target deviation, including: Obtaining a design parameter range of a device parameter of a target device; Based on the normal distribution sampling method and the target deviation, sampling is performed within the design parameter interval to determine a plurality of samples to be predicted.
[0007] Optionally, determining a target performance parameter of a target device and a device parameter corresponding to the target performance parameter includes: determining target performance parameters and multiple device parameters of a target device; Based on the principal component analysis method, main parameters corresponding to the target performance parameters are determined, and the main parameters are used as device parameters.
[0008] Optionally, obtaining a sample corresponding to the device parameter and a sample performance parameter corresponding to the sample includes: Determine the parameter distribution interval corresponding to the device parameter; Sampling within the parameter distribution interval to obtain a sample; Determine the sample performance parameters corresponding to the sample using the finite element method, or For a sample device manufactured based on the sample, sample performance parameters are measured.
[0009] Optionally, obtaining a sample corresponding to the device parameter and a sample performance parameter corresponding to the sample includes: Obtain measurement statistics for an actual device, and determine samples corresponding to device parameters and corresponding sample performance parameters based on the measurement statistics.
[0010] A second aspect of the present invention provides a MEMS device modeling and yield prediction device using machine learning, the MEMS device modeling and yield prediction device using machine learning comprising: A first determination module, used to determine target performance parameters of a target device and device parameters corresponding to the target performance parameters; An acquisition module, used for acquiring samples corresponding to the device parameters and sample performance parameters corresponding to the samples; A training module, used to train a machine learning model using the sample parameters as input and the corresponding sample performance parameters as output to obtain a performance parameter prediction model; A prediction module, configured to obtain a plurality of samples to be predicted corresponding to device parameters by adopting a preset sampling method based on a target deviation, and determine a performance parameter corresponding to each sample to be predicted based on the performance parameter prediction model; The second determination module is used to determine the device yield corresponding to the target deviation of the preset sampling method based on the performance parameters of multiple samples to be predicted and the preset classification standard.
[0011] Optionally, the device further comprises: The third determination module is used to determine that the target deviation is a qualified deviation when the device yield is greater than or equal to a preset device yield.
[0012] Optionally, the prediction module is used to: Obtaining a design parameter range of a device parameter of a target device; Based on the normal distribution sampling method and the target deviation, sampling is performed within the design parameter interval to determine a plurality of samples to be predicted.
[0013] Optionally, the first determining module is used to: determining target performance parameters and multiple device parameters of a target device; Based on the principal component analysis method, main parameters corresponding to the target performance parameters are determined, and the main parameters are used as device parameters.
[0014] Optionally, the acquisition module is used to: Determine the parameter distribution interval corresponding to the device parameter; Sampling within the parameter distribution interval to obtain a sample; Determine the sample performance parameters corresponding to the sample using the finite element method, or For a sample device manufactured based on the sample, sample performance parameters are measured.
[0015] Optionally, the acquisition module is used to: Obtain measurement statistics for an actual device, and determine samples corresponding to device parameters and corresponding sample performance parameters based on the measurement statistics.
[0016] A third aspect of the present invention provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when executed, the processor implements the MEMS device modeling and yield prediction method using machine learning as described in the first aspect of the present invention.
[0017] A fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the MEMS device modeling and yield prediction method using machine learning as described in the first aspect of the present invention.
[0018] A fifth aspect of the present invention provides a computer program product, including a computer program / instruction, which is implemented by a processor to perform the steps in the method for MEMS device modeling and yield prediction using machine learning as described in the first aspect of the present invention.
[0019] In MEMS mass manufacturing, there are generally two aspects: random yield and systematic yield. The embodiments of the present invention mainly analyze the systematic yield, which refers to the yield loss caused by poor uniformity within / between chips due to the "offset" of process conditions. For example, the yield loss caused by uncertainties in the manufacturing process such as chemical mechanical polishing, etching, and photolithography. The gap between design and manufacturing leads to the difference between the ideal and the actual manufactured circuits. For example, the designed physical parameters such as cantilever beam width, length, and functional layer thickness will change in actual situations.
[0020] The modeling and yield prediction of MEMS devices generally adopt empirical formulas, equivalent models, finite element simulation and other methods. However, in the yield prediction of MEMS devices, it is necessary to repeat the iteration of modeling and optimization. The use of the above methods will cause a single iteration to take a long time and require a large number of iterations, which is inefficient. In addition, due to the complex three-dimensional structure of MEMS devices, there are many parameters that affect the performance of MEMS devices, which is reflected in the high dimension of device modeling. When considering higher dimensions, the traditional methods further highlight their shortcomings of low efficiency. The use of machine learning methods for device modeling and yield prediction can perform efficient and accurate calculations on high-dimensional and large sample data, improve the calculation efficiency of a single iteration, reduce the number of iterations of modeling and optimization, and have higher efficiency.
[0021] In the embodiment of the present invention, a preset sampling method can be used to determine multiple samples to be predicted based on the target deviation, and then the corresponding performance parameters can be predicted based on the machine learning algorithm and the samples to be predicted, so that the performance of devices with different geometric dimensions and material properties can be predicted, and the performance parameters of the samples to be predicted can be further analyzed to determine the corresponding device yield prediction results, and finally the device yield results corresponding to the device parameters of different variation parameter distributions can be obtained. In this way, the manufacturing process of MEMS devices can be quickly and accurately simulated and analyzed, and the factors that may lead to low yield can be identified and optimized, so as to achieve quantitative analysis of key process windows and reduce the number of yield iterations in large-scale production; and it can help designers predict the manufacturing defects and failure modes of MEMS devices and make corresponding improvements in the design stage. BRIEF DESCRIPTION OF THE DRAWINGS
[0022] In order to more clearly illustrate the technical solution of the present invention, the accompanying drawings required for use in the description of the present invention will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.
[0023] Figure 1It is a flowchart of the steps of the method for modeling and yield prediction of MEMS devices using machine learning provided by the present invention; Figure 2 It is a structural schematic diagram of a MEMS device modeling and yield prediction device using machine learning provided by the present invention. DETAILED DESCRIPTION
[0024] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.
[0025] The embodiment of the present invention provides a method for MEMS device modeling and yield prediction using machine learning, specifically, Figure 1 As shown, it shows a flowchart of the steps of the method for MEMS device modeling and yield prediction using machine learning provided by an embodiment of the present invention, and the method includes the following steps: S101, determining target performance parameters of a target device and device parameters corresponding to the target performance parameters.
[0026] In an embodiment of the present invention, the target device is a MEMS device that currently needs to be analyzed, which may specifically include: sensors, micromotors, micropumps, filters, oscillators, antennas, micromirror arrays, interferometers, microfluidic chips, biochips, drug delivery systems, energy harvesters, etc.
[0027] In the embodiment of the present invention, after the target device is determined, the industrial background of the device can be determined based on relevant production requirements and industrial application requirements of the target device, and then all performance parameters that affect the performance of the MEMS device can be determined.
[0028] In the embodiment of the present invention, the target performance parameter refers to the main performance parameter concerned in the manufacturing process of the target device. For example, for a microfluidic chip, its target performance parameter may include: fluid control performance and geometric accuracy.
[0029] In the embodiment of the present invention, the target performance parameter may include one or more performance parameters. In actual application, each performance parameter may be analyzed separately.
[0030] In the embodiment of the present invention, the device parameters may include: all parameters that affect the performance of the MEMS device, such as material parameters, structural parameters, morphology parameters, etc.
[0031] S102, obtaining samples corresponding to the device parameters and sample performance parameters corresponding to the samples.
[0032] In the embodiment of the present invention, random sampling can be performed based on the parameter range of the device parameters designed during the manufacturing process of the target device to obtain samples, and then the sample performance parameters corresponding to the samples can be determined based on actual manufacturing or finite element method.
[0033] In the embodiment of the present invention, the data corresponding to the sample includes the values corresponding to the device parameters. In the case where there are multiple device parameters, a sample is obtained by taking the values of each device parameter.
[0034] In the embodiment of the present invention, the sample and the sample performance parameter may also be determined based on the parameter data and corresponding performance parameters of the sample device obtained in the actual manufacturing process.
[0035] S103, training a machine learning model with the sample as input and the corresponding sample performance parameter as output to obtain a performance parameter prediction model.
[0036] In an embodiment of the present invention, the device parameters may be one or more types. When the device parameters of a sample are multiple types, a sample (including the parameter values of multiple device parameters) may be used as an input, and the parameter values of the multiple device parameters corresponding to the sample may be defined as independent variables X1, and the corresponding sample performance parameters may be defined as dependent variables Y1.
[0037] In an embodiment of the present invention, relevant machine learning methods (such as artificial neural networks, support vector machine methods, and decision tree regression methods) can be used to construct a mathematical substitution model, and the sample X1 obtained above is used as the model input, and the sample performance parameter Y1 is used as the model output to train and learn the model.
[0038] In the embodiment of the present invention, sampling can also be performed to obtain test samples of the device parameters, and these test samples are defined as X2. For the sampled test samples, the finite element method or the actual manufacturing measurement method is used to obtain the test performance parameters corresponding to these test samples. The test performance parameters corresponding to each test sample are defined as the dependent variable Y2.
[0039] These test samples X2 are input as test data into the trained model to obtain the prediction results of the model output, and the prediction results are compared with the test performance parameters Y2 corresponding to each test sample to measure the model prediction performance. If the error between the model prediction result and the actual result is small, the model training is confirmed to be complete. If the error of the prediction result is large, the hyperparameter adjustment method can be used to iteratively modify the model hyperparameters until the model prediction accuracy meets the requirements and a performance parameter prediction model is obtained.
[0040] S104, using a preset sampling method, based on the target deviation, obtaining a plurality of samples to be predicted corresponding to the device parameters, and determining the performance parameters corresponding to each sample to be predicted based on the performance parameter prediction model.
[0041] In the embodiment of the present invention, the preset sampling method can be determined according to the actual distribution form of the device parameters. For example, if the statistical characteristics of the device parameters follow the normal distribution form, the preset sampling method can be normal distribution sampling.
[0042] In the embodiment of the present invention, the target deviation is a deviation corresponding to the statistical distribution form of the device parameter. For example, if the statistical characteristics of the device parameter follow the normal distribution form, the corresponding target deviation is the standard deviation.
[0043] In an embodiment of the present invention, a sampling method is determined according to the statistical characteristics of device parameters, and the samples to be predicted are determined by sampling. The performance parameters corresponding to the device parameters in the corresponding distribution form can be predicted, so as to determine whether the device parameters with the statistical characteristics meet the actual production standards (for example, the yield is greater than the preset value), thereby providing guidance for the design process of the target device.
[0044] S105, determining the device yield corresponding to the target deviation of the preset sampling method based on the performance parameters of the plurality of samples to be predicted and the preset classification standard.
[0045] In the embodiment of the present invention, the performance parameters of multiple samples to be predicted can be counted to obtain batch device performance statistics, so that the performance statistics of batch devices with device parameters having target deviations can be determined, and then the device yield of the batch devices can be determined according to a preset classification standard. Then, the device yield of the batch devices corresponding to the target deviation can be obtained.
[0046] In the embodiment of the present invention, by performing corresponding analysis on a variety of performance parameters, quantitative data on device yield of multi-dimensional performance parameters can be obtained, thereby providing decision-making guidance for the actual manufacturing process.
[0047] In an optional embodiment, the method further comprises the following steps: S106, when the device yield is greater than or equal to a preset device yield, determining the target deviation as a qualified deviation.
[0048] In the embodiments of the present invention, the manufacturing process of MEMS devices can be quickly and accurately simulated and analyzed, and factors that may lead to low yield can be identified and optimized, thereby achieving quantitative analysis of key process windows and reducing the number of yield iterations in large-scale production; and it can help designers predict manufacturing defects and failure modes of MEMS devices and make corresponding improvements in the design stage.
[0049] The MEMS device modeling and yield prediction method based on machine learning algorithm proposed in the embodiment of the present invention aims to solve the problem of systematic yield loss in MEMS mass manufacturing. Through this method, the yield iteration efficiency of MEMS manufacturing can be significantly improved, the manufacturing cost can be reduced, and the matching between design and manufacturing can be optimized.
[0050] In terms of industrial application value, the embodiment of the present invention is applied to MEMS manufacturing wafer-level foundries, which can bring technological progress and industrial upgrading to MEMS manufacturing wafer-level foundries. By applying the method of the embodiment of the present invention, the manufacturing yield of typical MEMS products can be increased to more than 95%, while improving the performance consistency and reliability of MEMS products. In addition, the method provided by the embodiment of the present invention is particularly suitable for MEMS devices (such as inertial sensors, RF MEMS devices, etc.) with high manufacturing precision requirements, and can meet the manufacturing needs of high-end MEMS devices. Finally, by applying the method of the embodiment of the present invention, the number of yield iterations of MEMS wafer foundries can be effectively reduced, and the number of iterations required for yield improvement can be reduced by 30%-50%, thereby significantly reducing tape-out and testing costs, shortening the time to market for new products, and enhancing the market competitiveness of MEMS foundry companies.
[0051] In an optional implementation, step S101 includes the following sub-steps: S1011, determining a target performance parameter and multiple device parameters of a target device.
[0052] S1012, determining main parameters corresponding to the target performance parameters based on a principal component analysis method, and using the main parameters as device parameters.
[0053] In the embodiment of the present invention, the device parameters that affect the target performance parameters may include multiple parameters. In order to simplify the calculation process, the embodiment of the present invention proposes that data dimension reduction can be performed, specifically, the principal component analysis method can be used to achieve data dimension reduction, and the main parameters corresponding to the target performance parameters can be obtained as device parameters.
[0054] In the embodiment of the present invention, other alternative methods may be used to achieve data dimensionality reduction. Alternatively, data dimensionality reduction may not be performed, and all device parameters related to the target performance parameters may be directly used as device parameters for subsequent sampling and model training.
[0055] In an optional implementation, the step S102 includes the following sub-steps: S1021, determining a parameter distribution interval corresponding to a device parameter.
[0056] S1022: Sampling within the parameter distribution interval to obtain samples.
[0057] S1023, using a finite element method to determine a sample performance parameter corresponding to the sample, or measuring a sample device manufactured based on the sample to obtain the sample performance parameter.
[0058] In the embodiment of the present invention, the upper and lower bounds of the parameter distribution interval corresponding to the device parameter can be obtained by online data statistics in the manufacturing process, and samples are obtained within the interval by uniform sampling or other sampling methods.
[0059] In an optional implementation, the step S102 includes the following sub-steps: S1021', obtaining measurement statistical data for an actual device, and determining samples corresponding to device parameters and corresponding sample performance parameters based on the measurement statistical data.
[0060] In the embodiment of the present invention, the measurement statistical data of existing actual devices can be analyzed to determine the samples and the corresponding sample performance parameters.
[0061] In the embodiment of the present invention, only the device parameters in the actual measurement are extracted and counted, and the statistical results of the device parameters are directly input into the model, which can avoid the use of sampling and train the model based on real data.
[0062] In an optional implementation, the preset adoption method is normal distribution sampling, and step S104 includes the following sub-steps: S1041, obtaining a design parameter range of a device parameter of a target device.
[0063] S1042, based on the normal distribution sampling method and the target deviation, sampling is performed within the design parameter range to determine a plurality of samples to be predicted.
[0064] In an embodiment of the present invention, during the design process of a target device, a design parameter range of device parameters can be obtained, and based on a normal distribution sampling method and a target deviation, sampling is performed within the design parameter range to determine a plurality of samples to be predicted, and then batch prediction is performed on the plurality of samples to be predicted to determine performance parameters of the corresponding batch devices, and then the yield of the batch devices is determined, thereby determining whether the target deviation can meet actual industrial needs.
[0065] In the embodiment of the present invention, the standard deviation of the sampling can also be changed, and batch input into the trained performance parameter prediction model to obtain device yield quantification data under different standard deviations. In the embodiment of the present invention, the same operation can be performed on each performance parameter dimension in turn to obtain device yield quantification data of multi-dimensional parameters, and quantitative analysis and decision guidance can be provided for the actual manufacturing process based on the above data.
[0066] Based on the same inventive concept, the present invention also provides a MEMS device modeling and yield prediction device using machine learning, such as Figure 2 As shown, it shows a structural schematic diagram of a MEMS device modeling and yield prediction device using machine learning provided in an embodiment of the present invention, wherein the MEMS device modeling and yield prediction device 200 using machine learning includes: A first determination module 201 is used to determine a target performance parameter of a target device and a device parameter corresponding to the target performance parameter; An acquisition module 202, used to acquire samples corresponding to the device parameters and sample performance parameters corresponding to the samples; A training module 203 is used to train a machine learning model using the sample as input and the corresponding sample performance parameter as output to obtain a performance parameter prediction model; The prediction module 204 is used to obtain a plurality of samples to be predicted corresponding to the device parameters based on the target deviation by adopting a preset sampling method, and determine the performance parameters corresponding to each sample to be predicted based on the performance parameter prediction model; The second determination module 205 is used to determine the device yield corresponding to the target deviation of the preset sampling method based on the performance parameters of multiple samples to be predicted and the preset classification standard.
[0067] Optionally, the device further comprises: The third determination module is used to determine that the target deviation is a qualified deviation when the device yield is greater than or equal to a preset device yield.
[0068] Optionally, the prediction module 204 is used to: Obtaining a design parameter range of a device parameter of a target device; Based on the normal distribution sampling method and the target deviation, sampling is performed within the design parameter interval to determine a plurality of samples to be predicted.
[0069] Optionally, the first determining module 201 is configured to: determining target performance parameters and multiple device parameters of a target device; Based on the principal component analysis method, main parameters corresponding to the target performance parameters are determined, and the main parameters are used as device parameters.
[0070] Optionally, the acquisition module 202 is used to: Determine the parameter distribution interval corresponding to the device parameter; Sampling within the parameter distribution interval to obtain a sample; Determine the sample performance parameters corresponding to the sample using the finite element method, or For a sample device manufactured based on the sample, sample performance parameters are measured.
[0071] Optionally, the acquisition module 202 is used to: Obtain measurement statistics for an actual device, and determine samples corresponding to device parameters and corresponding sample performance parameters based on the measurement statistics.
[0072] Based on the same inventive concept, the present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes, the steps in the method for MEMS device modeling and yield prediction using machine learning as described in any of the above embodiments are implemented.
[0073] Based on the same inventive concept, the present invention also provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps in the method for MEMS device modeling and yield prediction using machine learning described in any of the above embodiments are implemented.
[0074] Based on the same inventive concept, the present invention provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the steps in the method for MEMS device modeling and yield prediction using machine learning described in any of the above embodiments.
[0075] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.
[0076] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, devices, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0077] The present invention is described with reference to the flowcharts and / or block diagrams of the method, terminal device (apparatus), and computer program product according to the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of the processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable terminal device to generate a machine, so that the instructions executed by the processor of the computer or other programmable terminal device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1A device that provides the functions specified in a block or multiple blocks.
[0078] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable terminal device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0079] These computer program instructions can also be loaded onto a computer or other programmable terminal device so that a series of operating steps are executed on the computer or other programmable terminal device to produce a computer-implemented process, thereby providing instructions for executing on the computer or other programmable terminal device to implement the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
[0080] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.
[0081] Finally, it should be noted that, in the present invention, relational terms such as first and second, etc., are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or terminal device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or terminal device. In the absence of further restrictions, the elements defined by the sentence "comprise one..." do not exclude the existence of other identical elements in the process, method, article or terminal device including the elements.
[0082] The above is a detailed introduction to a MEMS device modeling and yield prediction method using machine learning provided by the present invention. The present invention uses specific examples to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea; at the same time, for those skilled in the art, according to the idea of the present invention, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.
Claims
1. A method for modeling and yield prediction of MEMS devices using machine learning, characterized in that: The method comprises: Determining target performance parameters of a target device and device parameters corresponding to the target performance parameters; Obtaining samples corresponding to the device parameters and sample performance parameters corresponding to the samples; The machine learning model is trained with the sample as input and the corresponding sample performance parameter as output to obtain a performance parameter prediction model; Using a preset sampling method, based on the target deviation, a plurality of samples to be predicted corresponding to the device parameters are obtained, and the performance parameters corresponding to each sample to be predicted are determined based on the performance parameter prediction model; Based on the performance parameters of a plurality of samples to be predicted and preset classification standards, the device yield corresponding to the target deviation of the preset sampling method is determined.
2. The method for MEMS device modeling and yield prediction using machine learning according to claim 1, characterized in that: The method further comprises: When the device yield is greater than or equal to a preset device yield, the target deviation is determined to be a qualified deviation.
3. The method for MEMS device modeling and yield prediction using machine learning according to claim 1, characterized in that: The preset sampling method is normal distribution sampling, and the preset sampling method is used to obtain multiple samples to be predicted corresponding to the device parameters based on the target deviation, including: Obtaining a design parameter range of a device parameter of a target device; Based on the normal distribution sampling method and the target deviation, sampling is performed within the design parameter interval to determine a plurality of samples to be predicted.
4. The method for MEMS device modeling and yield prediction using machine learning according to claim 1, characterized in that: Determining target performance parameters of a target device and device parameters corresponding to the target performance parameters includes: determining target performance parameters and multiple device parameters of a target device; Based on the principal component analysis method, main parameters corresponding to the target performance parameters are determined, and the main parameters are used as device parameters.
5. The method for MEMS device modeling and yield prediction using machine learning according to any one of claims 1 to 4, characterized in that: Acquiring a sample corresponding to the device parameter and a sample performance parameter corresponding to the sample, including: Determine the parameter distribution interval corresponding to the device parameter; Sampling within the parameter distribution interval to obtain a sample; The sample performance parameters corresponding to the sample are determined by using the finite element method; or, the sample performance parameters are obtained by measuring a sample device manufactured based on the sample.
6. The method for MEMS device modeling and yield prediction using machine learning according to any one of claims 1 to 4, characterized in that: Acquiring a sample corresponding to the device parameter and a sample performance parameter corresponding to the sample, including: Obtain measurement statistics for an actual device, and determine samples corresponding to device parameters and corresponding sample performance parameters based on the measurement statistics.
7. A MEMS device modeling and yield prediction device using machine learning, characterized in that: The MEMS device modeling and yield prediction device using machine learning includes: A first determination module, used to determine target performance parameters of a target device and device parameters corresponding to the target performance parameters; An acquisition module, used for acquiring samples corresponding to the device parameters and sample performance parameters corresponding to the samples; A training module, used to train a machine learning model using the sample as input and the corresponding sample performance parameter as output to obtain a performance parameter prediction model; A prediction module, configured to obtain a plurality of samples to be predicted corresponding to device parameters based on a target deviation by adopting a preset sampling method, and determine a performance parameter corresponding to each sample to be predicted based on the performance parameter prediction model; The second determination module is used to determine the device yield corresponding to the target deviation of the preset sampling method based on the performance parameters of multiple samples to be predicted and the preset classification standard.
8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method for MEMS device modeling and yield prediction using machine learning as described in any one of claims 1 to 6 is implemented.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for MEMS device modeling and yield prediction using machine learning as described in any one of claims 1 to 6 is implemented.
10. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instruction is executed by a processor, the steps in the method for MEMS device modeling and yield prediction using machine learning as described in any one of claims 1 to 6 are implemented.