Method of predicting the performance of a traction motor for a vehicle and optimizing design parameters using AI

Abstract

An artificial intelligence (AI)-based engine development system for optimizing the design of a drive motor for a vehicle according to the present disclosure includes a predictive AI model generation part configured to predict engine performance improvements, including noise / vibration / harshness (NVH), by fitting a polynomial curve to noise peak predictions based on modified engine design variables obtained from the CAD engine drawing. The system also includes a design parameter optimization AI model generation part configured to optimize engine design parameter dimensions from a design parameter optimization proposal AI model obtained by one of reinforcement learning, Q-learning, or particle swarm optimization (PSO) using the predictive AI model as a feature extractor for target engine performance improvements.

Description

BackgroundTechnical area

[0001] The present disclosure relates to the design of a drive motor for a vehicle. In particular, the present disclosure relates to a method of predicting the performance of a drive motor for a vehicle and optimizing an electric motor using artificial intelligence (AI). Description of the technology used

[0002] Generally, computer-aided motor design (CAD) is used to develop electric motors (i.e., drive electric motors) for a vehicle.

[0003] For example, a method for developing an electric motor using CAD includes determining the motor's design parameters (e.g., stator-rotor assembly factors), which may be configurable in CAD motor drawings. The method also includes analyzing all correlation conditions between the determined design parameters on a case-by-case basis. Furthermore, the method may include predicting points for improving the electric motor's performance, including noise / vibration / harshness (NVH), through simulations based on the analysis results. Thus, the electric motor development is completed by applying these findings to the dimensions of the motor design parameters.

[0004] The method used for CAD motor development therefore requires the creation and analysis of a simulation analysis model of the CAD motor drawing. For this purpose, it is very important to determine the motor design parameters, which serve as both analysis and design parameters for predicting the desired performance of the electric motor.

[0005] However, the CAD motor development method uses specific units, which are variably determined based on the experience of an electric motor designer, as design parameters for the motor. When creating an analysis model, combinations of all parameters, which reflect the correlations between the motor design parameters, are often not taken into account.

[0006] This limitation arises from the fact that, due to time and cost constraints, the CAD motor development method cannot create an analysis model that considers all combinations of motor design parameters. Due to such limitations, the influence of certain design parameters, which can be used to modify the design, is inevitably determined based on the experience of the electric motor designer. Brief explanation

[0007] The present disclosure provides an engine design parameter optimization AI model capable of providing engine design parameters for achieving engine performance. To this end, an engine performance prediction AI model is created through AI learning based on engine design parameters from CAD engine drawings and electric motor performance data. The model also applies reinforcement learning to set target engine performance improvements by extracting features from the engine performance prediction AI model.

[0008] A method of predicting performance of a drive motor for a vehicle and optimizing design parameters using artificial intelligence (AI) includes obtaining data based on motor design parameters and performance of the vehicle-mounted drive motor. The method also includes generating an engine performance prediction AI model by an automated machine learning (AutoML) part based on the obtained data about the engine design parameters and engine performance. The method also includes applying an evolutionary algorithm to the engine performance prediction AI model and generating an engine design parameter optimization AI model through reinforcement learning.

[0009] In addition, the target data of the engine performance prediction AI model can be the engine performance and can be obtained by analyzing the engine design parameters.

[0010] In addition, combinations of engine design parameters can be obtained by statistical design of experiments (DOE).

[0011] Furthermore, the input data of the motor performance prediction AI model can be the motor design parameters. The motor design parameters can include one or more of a slot, a tooth, a stator tooth, a bridge, a magnet, and a center column.

[0012] In addition, the output data of the engine performance prediction AI model can be engine performance, and the engine performance can be one or more of noise / vibration / harshness (NVH), torque, torque ripple, and magnetic flux.

[0013] In addition, the engine design parameter optimization AI model may use the engine performance, which is the output data of the engine performance prediction AI model, and have the engine design parameters, which are the input data of the engine performance prediction AI model, as output data.

[0014] In addition, the engine design parameter optimization Kl model can apply one or more of reinforcement learning, Q-learning, and particle swarm optimization (PSO).

[0015] Furthermore, the engine design parameter optimization Cl model can be calculated using a plurality of combinations of the optimized engine design parameters. A priority of the plurality of combinations of the optimized engine design parameters can be set under a constraint condition for engine performance.

[0016] In addition, in the engine design parameter optimization Cl model, if the goal of improving engine performance is to reduce NVH, a minimum torque change can be set as a performance constraint condition.

[0017] In addition, a noise level can be predicted by a machine learning (ML) model for the noise level in an optimization process of the engine design parameter optimization Kl model.

[0018] The method for optimizing the design of a drive motor for a vehicle using the AI-based motor development system according to the present disclosure involves generating the optimized motor design parameter dimension set according to the target motor performance improvement. This is achieved by an engine performance prediction AI model that responds to a change in the design parameters of vehicle motors, i.e., electric vehicle motors, and an engine design parameter optimization suggestion AI model that uses the motor performance prediction AI model as a feature extractor.

[0019] It is possible to train the motor performance prediction AI model for performance prediction by labeling the NVH performance, magnetic force, torque ripple and motor torque performance according to the change in design parameters of the analysis simulation using the experimental radiation noise data and the CAD drawing data of the electric motor.

[0020] By using the performance prediction model according to the change in design parameters, it is possible to propose many combinations of design parameters with high influence probability that can achieve the required target performance.

[0021] The Shapley Additive Explanation Part (SHAP) function, which sorts features in descending order of importance, can be applied to ensure that the electric motor designer's expertise matches the prediction model. In this way, the motor performance prediction AI model can act as the feature extractor of the design parameter model and provide multiple combinations of optimized design parameters to achieve the target performance.

[0022] The experimental and analytical results of the input data are labeled in a two-step process. Specifically, in the first step, the precise performance of the performance prediction model can be predicted with respect to numerous combinations of changes in design parameters. In the second step, the performance prediction AI model is used as one of the feature extractors. Numerous combinations of optimized design parameters for the target performance can be proposed using reinforcement learning.

[0023] It is possible to validate the performance of the combination states of all engine design parameters and to present realistic design parameters. Short description of the drawings

[0024] The above and other objects, features and other advantages of the present disclosure should be more clearly understood from the following detailed description when considered in conjunction with the accompanying drawings. Fig. 1 is a configuration diagram of a method of predicting the performance of a drive motor for a vehicle and optimizing the design according to an embodiment of the present disclosure. Fig. 2 shows a prediction result data difference of a performance prediction AI model for each vehicle segment according to an embodiment of the present disclosure. Fig. 3 is a conceptual diagram of a method of predicting the performance of a drive motor for a vehicle and optimizing the design on a step-by-step basis according to an embodiment of the present disclosure. Fig. 4 is a flowchart of a method of deriving an engine design optimization model after generating a drive motor performance prediction AI model for a vehicle according to an embodiment of the present disclosure. Fig. 5 is a conceptual diagram of a process of generating a drive motor performance prediction AI model for a drive motor for a vehicle according to an embodiment of the present disclosure. Detailed description

[0025] Hereinafter, embodiments of the present disclosure will be described in detail with reference to the accompanying drawings. These embodiments are examples and may be implemented in numerous different forms by those skilled in the art to which the present disclosure belongs, and are therefore not limited to the embodiments disclosed herein.

[0026] When a controller, component, device, element, part, unit, module, or the like of the present disclosure is described as having a purpose or performing an operation, function, or the like, the controller, component, device, element, part, unit, or module should be considered "configured" to fulfill that purpose or perform that operation or function. Each controller, component, device, element, part, unit, module, and the like may separately embody a processor and memory, such as a non-transitory computer-readable medium, or may be provided as part of the device.

[0027] With reference to Fig. 1, a data set used in the present disclosure includes engine design parameters 10a and engine performance 10b. The engine design parameters 10a may be obtained from a CAD engine drawing, and the engine performance may be obtained from analysis results (simulation results) calculated based on the engine design parameters.

[0028] The motor design parameters 10a are main design parameters of a stator / rotor assembly, which is a drive unit of an electric motor for an electric vehicle. In other words, in the present disclosure, the motor design parameters 10a include the dimensions of each of a stator and a rotor as data for learning a drive motor performance prediction AI model for a vehicle. The motor performance 10b includes analysis values ​​such as noise / vibration / harshness (NVH), torque, torque ripple, radial magnetic flux, and tangential magnetic flux as target data.

[0029] Specifically, the motor design parameters 10a apply a slot length, a slot radius, a tooth tip thickness, a tooth tip angle, a tooth width, a bridge thickness of the 1st / 2nd layer, a magnet thickness of the 1st / 2nd layer, a thickness of the center pillars between two magnets, a thickness of the center pillars between two magnets of the 1st / 2nd layer, a magnet angle, a magnet angle of the 1st / 2nd layer, a magnet length of the 1st / 2nd layer, a stator tooth width, and the like.

[0030] The motor design parameters can be determined in a form of a combination of design parameters using a statistical design of experiments (DOE) and can also be applied using data on specifications of competitors' electric motors or by adding new data.

[0031] As the motor design parameters, for example, data on 352 DOE points can be obtained regarding 11 motor design parameters, which include a bridge thickness around a rotor outer diameter (OD), a center column thickness between two magnets, a magnet thickness / width / angle, a stator tooth tip width, a stator tooth thickness, a stator tip angle, and a slot length / width / radius.

[0032] The motor performance 10b is the result of an analysis based on the dimensions assigned to the motor design parameters 10a. These target data represent the motor performance caused by the interaction between the rotor and stator according to the motor design parameters 10a.

[0033] In this case, the engine performance 10b includes NVH, torque ripple, torque ripple harmonic order, radiated power order, maximum stator density, noise level, noise range, peak number, peak torque, shaft speed, and the like.

[0034] The drive motor performance prediction AI model for a vehicle is generated by an AI model learning part 20 in which the engine design parameters 10a are used as input data and the engine performance 10b is used as target data.

[0035] A drive motor performance prediction AI model 30 for a vehicle, trained by the AI ​​model learning part 20, predicts performance result data 31a based on changes in the motor design parameters 10aa. When new dimensions are applied to the motor design, replacing the original motor design parameters 10a, the model generates updated performance predictions.

[0036] The motor design parameters 10aa, which have been changed to the new dimensions, are, for example, a design parameter that represents the main performance of the electric motor, and can be applied by changing the slot length / slot width, the tooth tip thickness / tooth tip angle, the bridge thickness of the 1st / 2nd layer, the magnet thickness / magnet length of the 1st / 2nd layer, the magnet angle, the thickness of the center column between two magnets, the stator tooth tip width, and the like.

[0037] The drive motor performance prediction AI model 30 for a vehicle generated by the AI ​​model learning part 20 is provided as an engine performance prediction AI model whose correlation, which is a change in performance based on changes in the engine design parameters, is set to a value of more than 0.95.

[0038] In addition, a design parameter optimization AI model 40 may be generated from the drive motor performance prediction AI model 30 of the drive motor for a vehicle generated by the AI ​​model learning part 20. Hereinafter, the drive motor performance prediction AI model for a vehicle may be referred to as an engine performance prediction AI model.

[0039] The motor performance prediction AI model is created by setting n points representing an operating range of an electric vehicle motor, obtaining main performance result data from n points, then using them as input data and target data, and performing a learning process to achieve a correlation at which the accuracy of the motor performance on the motor design parameters is 95% or higher.

[0040] By driving an electric power supply by the current driving from the input and target data obtained by a test or calculation, in the engine performance prediction AI model, a process of sequentially calculating a performance of an electromagnetic system by an electromagnetic force, a mechanical system by a speed, an acoustic environment by acoustic sounds, and the like is performed.

[0041] Specifically, the performance of the electromagnetic system is calculated using an FE-based model, combining one or more of an air gap force calculation, a Maxwell stress tensor method, and a Fourier analysis. The performance of the mechanical system is calculated using a combination of one or more analytically based models, a free-motion response, and a force response. The acoustic noise performance is calculated using a combination of one or more analytically based models and sound power levels.

[0042] The engine performance prediction AI model 30 can have improved reliability compared to the conventional model of predicting engine performance depending on the expertise result of the engine designer.

[0043] The AI ​​model generated by the AI ​​model learning part 20 is stored and may be referred to as the engine performance prediction AI model 30 to improve reliability. A location where the engine performance prediction AI model 30 is provided or stored may be referred to as the engine performance prediction AI model generation part.

[0044] The AI ​​model generated from the engine performance prediction AI model 30 is stored and may be referred to as the design parameter optimization AI model 40 for optimizing design parameters. A location where the design parameter optimization AI model 40 is provided or stored may be referred to as the design parameter optimization AI model generation part.

[0045] The design parameter optimization AI model 40 provides a design parameter optimization AI model that can predict optimized engine design parameters and dimensions 41a with respect to the input of a target performance for improving the engine power performance, which has a reduction in NVH 10bb, based on the engine performance prediction AI model 30 with improved reliability.

[0046] In other words, in the design parameter optimization AI model generation part, the target performance of the engine having NVH is the target data, the engine design parameters are the output data, and as an AI recommendation algorithm, reinforcement learning, Q-learning, and particle swarm optimization (PSO) are combined or selectively applied, so that the inputs / outputs of the engine performance prediction AI model are set opposite to the engine performance prediction AI model.

[0047] In other words, the AI ​​recommendation algorithm can provide a design parameter optimization suggestion AI model that includes changes in the design parameters and the output dimensions that can achieve the target performance of improving the engine performance, which has NVH.

[0048] Referring to Fig. 2, engine design parameters and performance data of the engine 200 of a vehicle 300 are obtained for each segment, which is classified based on the overall length (length from a front bumper to a rear bumper) and the price of the vehicle.

[0049] For example, a type A electric motor 200a is an example in which the engine design parameters and the NVH performance characteristics are extracted for a segment A vehicle. A type B electric motor 200b is an example in which the engine design parameters and the NVH performance characteristics are extracted for a segment B vehicle. A type C electric motor 200c is an example in which the engine design parameters and the NVH performance characteristics are extracted for a segment C vehicle. The engine design parameters and the engine performance, including NVH, of the type A, type B, and type C engines can be used as input data and target data, respectively, for modeling, or can be used as input data of an established model.

[0050] Fig. 1 and Fig. 3 is a detailed configuration for generating the engine performance prediction AI model and the design parameter optimization AI model of the AI ​​model learning part 20. Fig. 1 and Fig. 3 shows a conceptual diagram of a method of predicting the performance and optimizing the design of the drive motor for a vehicle on a step-by-step basis.

[0051] Specifically, the engine performance prediction AI model 30 generated by the AI ​​model learning part 20 may be provided by a data acquisition part 32, a data versioning part 33, a model generation part 34, a model test evaluation part 35, and a performance prediction model finishing part 36.

[0052] For example, the data acquisition part 32 of the learning model uses the motor design parameters as input data. The motor design parameters used as input data include the motor design parameters and their dimensions.

[0053] Experimental labeling values ​​or simulation (analysis) labeling values ​​regarding the engine design parameters are obtained as target data, and learning model data from which features have been extracted through a data selection process by outline data analysis of the labeling values ​​are selected.

[0054] The learning model data acquisition part 32 includes a data input part 32a containing massive experimental and analytical characteristic values ​​representing the performance for each engine design parameter. The learning model data acquisition part 32 also includes: a data exploration part 32b for querying and analyzing the acquired data containing the massive experimental and analytical characteristic values ​​and deriving useful information; a data cleaning part 32c for identifying, modifying, and filtering errors, missing, and inaccurate values ​​in the data set; and a feature construction part 32d for optimizing the characteristics of the outline data between the experimental characteristic value and the simulation (analysis).In addition, the feature construction part 32d is configured to extract features as the useful information to finally obtain the learning model data to be used.

[0055] Data versioning 33 manages data versions of the previously acquired data and the newly acquired dataset with respect to the engine design parameters and performance used for learning, and performs systematic model management with an index configuration of the data gain.

[0056] The data versioning 33 provides training data and validation data from the data obtained for the model generation part 34 and provides test data for model evaluation in the model test evaluation part 35.

[0057] The model construction part 34a of the model generation part 34 uses the training data for model generation and creates a machine learning model by automating a process of developing the machine learning model by an automated machine learning part (AutoML part) 34c.

[0058] The created machine learning model is validated by the model evaluation part 34b using validation data provided by the data versioning 33.

[0059] The model generation part 34 is performed based on an ML lifecycle platform 34d in addition to all of the model construction part 34a, the model evaluation part 34b, and the AutoML part 34c. The ML lifecycle platform 34d is a complete solution for supporting the development, deployment, monitoring, and maintenance of the machine learning model.

[0060] The model generation part 34 uses the engine design parameters as input data for the engine performance prediction AI model and has the engine performance as output data, and creates a model that combines the structure of a deep neural network to increase the accuracy for output data of the model compared to the actual performance.

[0061] Specifically, the model construction part 34a and the model evaluation part 34b increase the correlation between the performance of the training model and the validation model through mutual data exchange and validation. In addition, the AutoML part 34c increases the correlation of the performance of the performance prediction model through data exchange and validation between the model construction part 34a and the model evaluation part 34b.

[0062] The model test evaluation part 35 evaluates the performance prediction model of the AutoML part 34c using sensitivity analysis of a change in performance based on a change in the model design parameters (i.e., the dimensions) and visualizes the result as test data. Furthermore, the model test evaluation part 35 compares the test data with the result of a developer's expertise. As a result, the model test evaluation part 35 improves the reliability compared to the performance prediction AI model.

[0063] For this purpose, the model testing part 35a of the model testing evaluation part 35 applies the test data to the performance prediction model and tests the accuracy of the model.

[0064] The Shapley Additive Explanation (SHAP) part 35b of the Model Test Evaluation part 35 is a statistical technique and framework for interpreting and explaining the prediction of a machine learning model. SHAP 35b is used to evaluate how much a predicted value contributes to specific properties or elements to help analyze the model's prediction.

[0065] The performance prediction model finalizing part 36 finalizes the performance prediction model 36a and receives an explainable interface from the SHAP part 35b in an explainable interface part 36b.

[0066] The design parameter optimization AI model generation part 40 extracts the features of the design parameter optimization AI model 40 from the engine performance prediction AI model 30, which is finalized by a final NVH prediction model part 36a. The final design parameter optimization AI model 40 uses the target performance of the electric motor as the target data and has the engine design parameters for the target performance as the output data.

[0067] A reinforcement learning technique is applied to the design parameter optimization AI model, and the features of the engine performance prediction AI model are extracted through optimization. The design parameter optimization AI model generation part 40 is divided into an evolutionary algorithm for the design optimizer development part 42a and a deep reinforcement learning design engine development part 42b. The performance prediction model is finally used as a feature extractor in an optimization method evaluation part 43.

[0068] Fig. Figure 4 shows an example in which NVH is applied as the engine performance, and a training data set including a peak value and peak position of the overall noise level is obtained as the engine performance with respect to the engine design parameters to which the engine development DOE is applied, by the data acquisition part 32 to generate an overall noise level model as the engine performance prediction AI model. The peak noise prediction is output as a curve fit with four polynomials from the overall noise level model.

[0069] Fig.Figure 5 shows a process for deriving the engine design parameter optimization Cl model with respect to NVH as the engine performance. In the engine design parameter optimization Cl model, when the goal of engine performance improvement is to reduce NVH, a minimum torque variation is set as a performance constraint.

[0070] Therefore, if NVH improvement performance is input as a target under the condition that the change in torque is minimal while improving NVH performance, then an optimal engine design parameter for achieving the engine performance improvement target is output. To derive the optimization model capable of achieving this, one or more of Q-learning and Particle Swarm Optimization (PSO) are applied, and an NVH design ML model 44 is optimized to achieve engine performance with a correlation of 95% or more with the engine design parameters. The NVH design ML model is confirmed as the design parameter optimization KL model when the engine performance improvement target including NVH is achieved, and the engine design parameter is output.

[0071] The engine design parameter optimization class model calculates n combinations of design parameters 46a. Furthermore, the priority of the n combinations of engine design parameters is determined under the constraint condition for engine performance.

[0072] In other words, the optimization model proposes n combinations of engine design parameters and dimensions, significantly improving NVH performance. The design parameter at which the torque variation is minimal is set as the priority of the output data.

[0073] Noise levels for n combinations 46b can be predicted by a noise level ML model 45 with respect to noise levels determined by the optimized NVH design ML model 44.

[0074] In other words, by repeatedly executing the NVH design and noise level ML models while deriving the optimized NVH design ML model 44 and the noise level ML model 45 from the optimized design parameter AI model generation part 40, the optimization for the NVH NVH design ML model and the noise level is performed to achieve the targeted NVH performance improvement.

Claims

[1] A method of predicting a performance of a drive motor for a vehicle and optimizing design parameters using artificial intelligence (AI), the method comprising: Obtaining data based on engine design parameters and the performance of the vehicle-mounted drive motor, generating an engine performance prediction AI model (30) by an automated machine learning (AutoML) part (34c) based on the acquired data on the engine design parameters and the engine performance, Applying an evolutionary algorithm to the engine performance prediction AI model (30), and Generating an engine design parameter optimization class model (40) by reinforcement learning. [2] The method according to claim 1, wherein the target data of the engine performance prediction AI model (30) is the engine performance and is obtained by analyzing the engine design parameters. [3] The method of claim 1 or 2, wherein combinations of the engine design parameters are obtained by statistical design of experiments (DOE). [4] The method according to any one of claims 1 to 3, wherein the input data of the engine performance prediction AI model is the engine design parameters, and wherein the engine design parameters comprise one or more of a groove, a tooth, a stator tooth, a bridge, a magnet, and a center column. [5] The method of any one of claims 1 to 4, wherein output data of the engine performance prediction AI model (30) is the engine performance, and wherein the engine performance comprises one or more of noise / vibration / harshness (NVH), torque, torque ripple, and magnetic flux. [6] The method according to any one of claims 1 to 5, wherein the engine design parameter optimization AI model (40) uses the engine performance representing the output data of the engine performance prediction AI model (30) and has the engine design parameters representing the input data of the engine performance prediction AI model (30) as output data. [7] The method of claim 6, wherein the engine design parameter optimization Cl model (40) applies one or more of reinforcement learning, Q-learning, and particle swarm optimization (PSO). [8] The method according to claim 6 or 7, wherein the engine design parameter optimization AI model (40) is calculated by a plurality of combinations of the optimized engine design parameters, and wherein a priority of the plurality of combinations of the optimized engine design parameters is set under a constraint condition for the engine performance. [9] The method according to any one of claims 6 to 8, wherein in the engine design parameter optimization Cl model (40), when an objective of engine performance improvement is to reduce noise / vibration / harshness (NVH), a minimum torque variation is set as a performance constraint condition. [10] The method according to any one of claims 6 to 9, wherein a noise level is predicted by a machine learning (ML) model for the noise level in an optimization process of the engine design parameter optimization Cl model (40).

Images