Deep learning-based model simulation prediction method and system

Abstract

Disclosed in the present invention is a deep learning-based model simulation prediction method, comprising the following steps: determining a plurality of parameters needed by model simulation, and setting an adjustable boundary for each parameter; calling a nanospice simulator to obtain a plurality of groups of data of the plurality of parameters, and normalizing the plurality of parameters on the basis of respective adjustable boundaries; inputting data of the plurality of normalized parameters into a pre-configured deep learning model to train the deep learning model; and obtaining parameters currently needed by model simulation prediction, inputting the parameters into the trained deep learning model, and performing prediction to obtain a prediction result. Resistance and capacitance data of a model can be accurately predicted, and the simulation prediction efficiency is improved through neural network training and prediction of bias-dependent current and capacitance curves under different parameters.

Description

A model simulation prediction method and system based on deep learning Technical Field

[0001] The present invention belongs to the field of chip design technology, and in particular relates to a model simulation prediction method and system based on deep learning. Background Art

[0002] Calculating model parameters for complex circuits has always been a core task in chip design technology. Model simulation involves inputting a set of model parameters and calculating outputs such as current and capacitance. Traditional methods employ complex formulas, such as the quasi-Newton method. However, these methods suffer from slow computation speed, difficulty in deriving partial derivatives of the outputs with respect to the parameters, and difficulty in calculating multiple outputs in parallel across multiple parameter sets. Furthermore, partial derivatives of the simulation with respect to the parameters cannot be directly obtained, forcing the use of differential derivatives.

[0003] With the development of science and technology, integrated circuit technology has made great progress, with increasing integration and decreasing power consumption. At the same time, various complex application scenarios have placed higher demands on the performance and power consumption of integrated circuits. Traditional hardware design methods are no longer able to meet these demands, requiring the use of algorithm design for optimization and simulation. Summary of the Invention

[0004] To solve the above problems, the purpose of the present invention is to provide a model simulation prediction method and system based on deep learning, which can accurately predict the resistance and capacitance data of the model, and train and predict the current and capacitance change curves with bias under different parameters through neural network training.

[0005] The technical solution of the present invention is: a model simulation prediction method based on deep learning, comprising the following steps: determining multiple parameters required for the model simulation, setting an adjustable boundary for each parameter; calling a nanospice simulator to obtain multiple sets of data for the multiple parameters, and normalizing the multiple parameters based on their respective adjustable boundaries; inputting the normalized multiple parameter data into a pre-configured deep learning model to train the deep learning model; obtaining the parameters currently required for the simulation prediction model and inputting them into the trained deep learning model for prediction to obtain a prediction result. Preferably, the machine learning model in this embodiment can be a deconvolutional neural network, a fully connected network, a GAN, an AE network, etc. The core of this technical solution lies in how to obtain parameters and how to preprocess the parameters based on the above machine learning models, so that the model can be effectively trained and a machine learning model that can accurately predict the resistance and capacitance of the model is finally obtained.

[0006] Preferably, determining multiple parameters required for the model simulation and setting adjustable boundaries for each parameter further includes: determining the multiple parameters based on historical real engineering experience, including: k2, u0, ua, eu, uc, vsat, a1, a2, nfactor, eta0, etab, pcm, pdiblc, ndep; taking standard average distribution randomization or logarithmic average distribution randomization to obtain multiple groups of values ​​for the multiple parameters; replacing the corresponding values ​​in the preset model card based on the multiple groups of values ​​of the multiple parameters.

[0007] Preferably, before normalizing the multiple parameters based on their respective adjustable boundaries, the method further includes: writing multiple sets of numerical data of the multiple parameters into the netlist file based on a pre-constructed netlist file, wherein the netlist file is pre-constructed according to the actual measurement points of the corresponding model; dividing according to the measurement points, each size contains multiple sets of different curves.

[0008] Preferably, normalizing the multiple parameters based on their respective adjustable boundaries further includes: when normalizing the parameters using a standard average distribution random method, normalizing is performed based on the upper and lower limits of the adjustable boundaries of each parameter, and the processing result is between 0 and 1; when normalizing the parameters using a random method using an average distribution after taking the logarithm, the parameter value is first logarithmized, and the upper and lower limits of the adjustable boundaries of each parameter are also logarithmed, and then normalizing is performed based on the logarithm of the value, the logarithm of the upper limit, and the logarithm of the lower limit, and the processing result is between 0 and 1.

[0009] Preferably, the method further includes: after randomly selecting multiple sets of values ​​of the multiple parameters using a standard average distribution or a logarithmic average distribution, performing preliminary screening based on set adjustable boundaries and physical facts to prevent the simulator from making mistakes.

[0010] Preferably, before inputting the normalized data of multiple parameters into the pre-configured deep learning model to train the deep learning model, the method further includes: translating the simulation results of the nanospice simulator into tfrecord format for storage through a parsing module to facilitate network training.

[0011] Preferably, inputting the normalized data of multiple parameters into a preconfigured simulator to train the deep learning model further includes: using the fit method of the Model class in the tensorflow.keras library, with a training set to validation set ratio of 6:1 and a number of samples in each batch of 600; using Adam as the optimizer, a learning rate of 1e-3, current and capacitance as outputs, and a loss function of Mean Squared Logarithmic Error.

[0012] Based on the same concept, the present invention also provides a model simulation and prediction system based on deep learning, including: a parameter determination module, used to determine multiple parameters required for the model simulation, and set an adjustable boundary for each parameter; a preprocessing module, used to call a nanospice simulator to obtain multiple groups of data of the multiple parameters, and normalize the multiple parameters based on their respective adjustable boundaries; a model training module, used to input the normalized data of the multiple parameters into a preconfigured deep learning model to train the deep learning model; and a result prediction module, used to obtain the parameters of the current simulation prediction model and input them into the trained deep learning model for prediction to obtain prediction results.

[0013] Based on the same concept, the present invention also provides an electronic device, characterized in that it includes: a memory, the memory is used to store a processing program; and a processor, the processor implements any one of the above-mentioned deep learning-based model simulation prediction methods when executing the processing program.

[0014] Based on the same concept, the present invention also provides a readable storage medium, characterized in that a processing program is stored on the readable storage medium, and when the processing program is executed by the processor, it implements any of the above-mentioned deep learning-based model simulation prediction methods.

[0015] Due to the adoption of the above technical solution, the present invention has the following advantages and positive effects compared with the prior art:

[0016] 1. In the technical solution of the present invention, multiple sets of data of multiple parameters are randomly and automatically generated and preprocessed. The preprocessed data and the resistance and capacitance simulation results obtained by calling the nanospice simulator are input into a deep learning model for training, thereby realizing simulation prediction of resistance and capacitance data based on the deep learning model. The simulation prediction efficiency is improved by training and predicting the current and capacitance change curves with bias under different parameters through neural network training.

[0017] 2. This invention uses a nanospice simulator to define specific circuit conditions and measurement scenarios through a netlist. When generating the simulation netlist, a corresponding netlist file is constructed based on a set of real measurement points from the foundry, giving the machine learning model the physical meaning of the application scenario. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The specific embodiments of the present invention are further described in detail below with reference to the accompanying drawings, wherein:

[0019] FIG1 is a flow chart of a model simulation prediction method based on deep learning of the present invention;

[0020] FIG2 is a schematic diagram of a deep learning neural network model according to an embodiment of the present invention;

[0021] FIG3 is a schematic diagram of the model simulation prediction results based on deep learning of the present invention. DETAILED DESCRIPTION

[0022] The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments. The advantages and features of the present invention will become more apparent from the following description and claims. It should be noted that the drawings are greatly simplified and not to exact ratios, and are intended solely to facilitate and clearly illustrate the embodiments of the present invention.

[0023] It should be noted that all directional indications in the embodiments of the present invention (such as up, down, left, right, front, back, etc.) are only used to explain the relative position relationship, movement status, etc. between the various components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

[0024] First embodiment

[0025] As shown in FIG1 , this embodiment provides a model simulation prediction method based on deep learning, comprising the following steps:

[0026] S100: Determine multiple parameters required for the model simulation, and set an adjustable boundary for each parameter;

[0027] S200: calling a nanospice simulator to obtain multiple sets of data of the multiple parameters, and performing normalization processing on the multiple parameters based on their respective adjustable boundaries;

[0028] S300: Inputting the normalized data of the plurality of parameters into a pre-configured deep learning model to train the deep learning model;

[0029] S400: Obtain parameters of the current simulation prediction model and input them into the deep learning model after training to perform prediction to obtain prediction results.

[0030] In the technical solution of this embodiment, multiple sets of data for multiple parameters are randomly and automatically generated and preprocessed. The preprocessed data and the resistance and capacitance simulation results obtained by calling the nanospice simulator are input into a deep learning model for training, thereby achieving simulation prediction of resistance and capacitance data based on the deep learning model, thereby improving the efficiency of simulation prediction. After determining multiple parameters, adjustable boundaries are set for each parameter. Preferably, in this solution, the parameter size is modified one by one according to a fixed step size, and simulation is performed to obtain the current and capacitance variation curves with bias. The parameters and curves are the data required for training.

[0031] Preferably, determining multiple parameters required for the model simulation and setting adjustable boundaries for each parameter further includes: determining the multiple parameters based on historical real engineering experience, including: k2, u0, ua, eu, uc, vsat, a1, a2, nfactor, eta0, etab, pcm, pdiblc, ndep; taking standard average distribution randomization or logarithmic average distribution randomization to obtain multiple groups of values ​​for the multiple parameters; replacing the corresponding values ​​in the preset model card based on the multiple groups of values ​​of the multiple parameters.

[0032] In combination with the application scenario of the present invention, the parameters determined in this embodiment include k2, u0, ua, eu, uc, vsat, a1, a2, nfactor, eta0, etab, pclm, pdiblc, and ndep. The meanings of these parameters are not redundantly explained here. The above parameters are all commonly used parameters in the semiconductor industry. The model can be different versions such as bsim4, bsim6, and bsimcmg. The above parameters are derived from different versions such as bsim4, bsim6, and bsimcmg.

[0033] For example, see Table 1. To obtain parameters for model simulation training, we first randomly obtained parameter values. To better reflect the actual situation, we used a set of realistic parameter ranges. Based on engineering experience, we also selected a set of more global parameters, totaling 14. Because different parameters have different distribution characteristics, for simplicity, we chose to use either a standard average distribution or a logarithmic average distribution for each parameter based on engineering experience:

[0034] Table 1 Parameters

[0035] A new model card can be obtained by replacing the corresponding parameter values ​​in the pre-configured model card with the random values ​​generated in the previous process. Since the nanospice simulator is used, a netlist is required to define the specific circuit conditions and specific measurement conditions.

[0036] Preferably, before normalizing the multiple parameters based on their respective adjustable boundaries, the method further includes: writing multiple sets of numerical data of the multiple parameters into the netlist file based on a pre-constructed netlist file, wherein the netlist file is pre-constructed according to the actual measurement points of the corresponding model; dividing according to the measurement points, each size contains multiple sets of different curves.

[0037] Since the nanospice simulator is used, a netlist is required to define the specific circuit conditions and the specific measurement conditions. When generating the simulation netlist, the corresponding netlist file is constructed based on a set of real measurement points of the foundry. 9 devices of different sizes are selected:

[0038] Table 2 Size table

[0039] Divided by measuring points, each dimension contains 7 groups of different curves:

[0040] Table 3 Measurement point situation table

[0041] The randomly generated parameters and the simulated resistance and capacitance results based on the Nanospice simulator are written into a pre-configured netlist file. The preferred embodiment of the present invention is based on TensorFlow. Therefore, after the simulation corresponding to a model card is completed, a parsing module is used to translate the simulation results into TFRecord format for storage and network training. The simulator runs on a 3.6GHz CPU.

[0042] Preferably, normalizing the multiple parameters based on their respective adjustable boundaries further includes: when normalizing the parameters using a standard average distribution random method, normalizing is performed based on the upper and lower limits of the adjustable boundaries of each parameter, and the processing result is between 0 and 1; when normalizing the parameters using a random method using an average distribution after taking the logarithm, the parameter value is first logarithmized, and the upper and lower limits of the adjustable boundaries of each parameter are also logarithmed, and then normalizing is performed based on the logarithm of the value, the logarithm of the upper limit, and the logarithm of the lower limit, and the processing result is between 0 and 1.

[0043] The technical solution of this embodiment normalizes each parameter, which can make parameter processing of subsequent models more convenient, avoid the model from additionally processing complex data, improve the accuracy of data processing, and reduce the possibility of model failure.

[0044] Preferably, the method further includes: after randomly selecting multiple sets of values ​​of the multiple parameters using a standard average distribution or a logarithmic average distribution, performing preliminary screening based on set adjustable boundaries and physical facts to prevent the simulator from making mistakes.

[0045] Because the Nanospice simulator is used, a netlist is required to define the specific circuit conditions and measurement scenarios. Random parameter combinations can violate physical constraints, such as the requirement that effective lengths cannot be less than 0. Therefore, a preliminary screening step is necessary. Based on certain criteria, model cards that are inevitably problematic are skipped to reduce simulator errors and shorten simulation data generation time.

[0046] Preferably, before inputting the normalized data of the multiple parameters into the pre-configured deep learning model to train the deep learning model, the method further includes: translating the simulation results of the nanospice simulator into a tfrecord format for storage through a parsing module to facilitate network training.

[0047] Preferably, inputting the normalized data of multiple parameters into a preconfigured simulator to train the deep learning model further includes: using the fit method of the Model class in the tensorflow.keras library, with a training set to validation set ratio of 6:1 and a number of samples in each batch of 600; using Adam as the optimizer, a learning rate of 1e-3, current and capacitance as outputs, and a loss function of Mean Squared Logarithmic Error.

[0048] Preferably, before training the model, a deep learning model must be built first. See Figure 2, which shows a schematic diagram of a deep learning neural network model in one embodiment. The neural network takes 14 model parameters as input; the network adopts a deconvolution network, that is, it is composed of several layers of fully connected layers and several layers of deconvolution layers; then, 6 curve data of 61 points are obtained by deconvolution calculation. The training sample is 500,000 simulation curves. The training uses the fit method of the Model class in the tensorflow.keras library. The number of samples is 450,000, the ratio of training set to validation set is 6:1, and the number of samples in each batch is 600. Adam is used as the optimizer with a learning rate of 1e-3. The current capacitance is used as the output, and the loss function uses Mean Squared Logarithmic Error. After training, see Figure 3, the average fitting error of the simulation curve is 1%.

[0049] Second embodiment

[0050] Based on the same concept, the present invention also provides a model simulation and prediction system based on deep learning, including: a parameter determination module, used to determine multiple parameters required for the model simulation, and set an adjustable boundary for each parameter; a preprocessing module, used to call a nanospice simulator to obtain multiple groups of data of the multiple parameters, and normalize the multiple parameters based on their respective adjustable boundaries; a model training module, used to input the normalized data of the multiple parameters into a preconfigured deep learning model to train the deep learning model; and a result prediction module, used to obtain the parameters of the current simulation prediction model and input them into the trained deep learning model for prediction to obtain prediction results.

[0051] In the technical solution of this embodiment, multiple sets of data of multiple parameters are randomly and automatically generated and preprocessed. The preprocessed data and the resistance and capacitance simulation results obtained by calling the nanospice simulator are input into the deep learning model for training, thereby realizing simulation prediction of the resistance and capacitance data based on the deep learning model. The simulation prediction efficiency is improved by training and predicting the current and capacitance variation curves under different parameters with the bias of the neural network.

[0052] Third embodiment

[0053] Based on the same concept, the present invention also provides an electronic device, characterized in that it includes: a memory, the memory is used to store a processing program; and a processor, the processor implements any one of the above-mentioned deep learning-based model simulation prediction methods when executing the processing program.

[0054] Based on the same concept, the present invention also provides a readable storage medium, characterized in that a processing program is stored on the readable storage medium, and when the processing program is executed by the processor, it implements any of the above-mentioned deep learning-based model simulation prediction methods.

[0055] If the model simulation prediction method based on deep learning is implemented in the form of program instructions and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this embodiment is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of software, and the computer software is stored in a storage medium, including a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present disclosure. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, and other media that can store program code.

[0056] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the identification content specifically executed by the above-described system and device can refer to the corresponding process in the aforementioned method embodiment.

[0057] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings, but the present invention is not limited to the above embodiments. Even if various changes are made to the present invention, if these changes fall within the scope of the claims of the present invention and their equivalents, they still fall within the scope of protection of the present invention.

Claims

1. A model simulation prediction method based on deep learning, characterized in that: The following steps are involved: Determining a plurality of parameters required for simulating the model, and setting an adjustable boundary for each parameter; Calling the nanospice simulator to obtain multiple sets of data of the multiple parameters, and normalizing the multiple parameters based on their respective adjustable boundaries; Inputting the normalized data of the multiple parameters into a pre-configured deep learning model to train the deep learning model; Obtain the parameters of the current simulation prediction model and input them into the trained deep learning model for prediction to obtain the prediction result.

2. The model simulation prediction method based on deep learning according to claim 1, characterized in that: Determining a plurality of parameters required for the model simulation, and setting an adjustable boundary for each parameter further comprises: The multiple parameters are determined based on historical real engineering experience, including: k2, u0, ua, eu, uc, vsat, a1, a2, nfactor, eta0, etab, pclm, pdiblc, ndep; A standard average distribution or a logarithmic average distribution is used for randomization to obtain multiple sets of values ​​of the multiple parameters; The corresponding values ​​in the preset model card are replaced based on the multiple sets of values ​​of the multiple parameters.

3. The model simulation prediction method based on deep learning according to claim 2, characterized in that: Before normalizing the multiple parameters based on their respective adjustable boundaries, the method further includes: Writing multiple sets of numerical data of the multiple parameters into the netlist file based on a pre-constructed netlist file, wherein the netlist file is pre-constructed according to real measurement points of the corresponding model; Divided according to the measuring points, each dimension contains multiple sets of different curves.

4. The model simulation prediction method based on deep learning according to claim 1, characterized in that: Normalizing the multiple parameters based on their respective adjustable boundaries further includes: When the standard average distribution random method is adopted to normalize the parameters, the normalization is performed based on the upper and lower limits of the adjustable boundary of each parameter, and the processing result is between 0 and 1; When the parameters are normalized by taking the average distribution after taking the logarithm, the logarithm of the parameter value is first taken, and the upper and lower limits of the adjustable boundary of each parameter are also taken. Then, normalization is performed based on the logarithm of the value, the logarithm of the upper limit, and the logarithm of the lower limit. The processing result is between 0 and 1.

5. The model simulation prediction method based on deep learning according to claim 2, characterized in that: The method further includes: after randomly selecting multiple sets of values ​​of the multiple parameters using standard average distribution or logarithmized average distribution, performing preliminary screening based on set adjustable boundaries and physical facts to prevent the simulator from making mistakes.

6. The model simulation prediction method based on deep learning according to claim 5, characterized in that: Before inputting the normalized data of the plurality of parameters into the pre-configured deep learning model to train the deep learning model, the method further includes: The simulation results of the nanospice simulator are translated into tfrecord format for storage through the parsing module to facilitate network training.

7. The model simulation prediction method based on deep learning according to claim 6, characterized in that: Inputting the normalized data of the plurality of parameters into a pre-configured simulator to train the deep learning model further comprises: The fit method of the Model class in the tensorflow.keras library was used, the ratio of the training set to the validation set was 6:1, and the number of samples in each batch was 600; Adam is used as the optimizer, the learning rate is 1e-3, current and capacitance are used as output, and Mean Squared Logarithmic Error is used as the loss function.

8. A model simulation prediction system based on deep learning, characterized in that: include: A parameter determination module, used for determining a plurality of parameters required for the model simulation, and setting an adjustable boundary for each parameter; A preprocessing module, used for calling the nanospice simulator to obtain multiple sets of data of the multiple parameters, and performing normalization processing on the multiple parameters based on their respective adjustable boundaries; A model training module, used for inputting the normalized data of multiple parameters into a pre-configured deep learning model to train the deep learning model; The result prediction module is used to obtain the parameters of the current simulation prediction model and input them into the deep learning model after training to perform prediction to obtain the prediction result.

9. An electronic device, characterized in that: include: A memory, the memory being used to store a processing program; A processor, wherein when executing the processing program, the processor implements the model simulation prediction method based on deep learning as described in any one of claims 1 to 7.

10. A readable storage medium, characterized in that: The readable storage medium stores a processing program, and when the processing program is executed by the processor, the model simulation prediction method based on deep learning as described in any one of claims 1 to 7 is implemented.

Images