CN116822058A - Diagnosis and optimization methods, systems, equipment and storage media based on weak sheet metal - Google Patents

Abstract

The invention discloses a diagnosis and optimization method, system, equipment and storage medium based on weak sheet metal. The method includes: identifying the frequency of NVH problems after the vehicle design is changed; and obtaining the NTF/VTF curve after the vehicle design is changed. Data, input the NTF/VTF curve data into the preset NVH problem simulation optimization model, and obtain the performance problem point diagnosis report of NVH problem frequency and the body sheet metal parts optimization plan. The preset NVH problem simulation optimization model is trained based on the SVM algorithm; according to Performance problem point diagnosis report and body sheet metal parts optimization plan deal with weak sheet metal. Compared with the prior art, the mechanism modeling method and analysis process are too complex, resulting in the problem that the real-time calculation process consumes a lot of computing power. However, the present invention obtains the NVH status of the vehicle based on the preset NVH problem simulation optimization model trained by the SVM algorithm. Analyze the results, thereby reducing the computing power consumption during NVH analysis.

Description

Diagnostic optimization method, system, equipment and storage medium based on weak sheet metal

Technical field

The present invention relates to the technical field of vehicle NVH, and in particular to a diagnosis and optimization method, system, equipment and storage medium based on weak sheet metal.

Background technique

Traditional noise, vibration, and harshness NVH (Noise, Vibration, Harshness NVH) simulation requires a lot of time and computing resources. Later, locating the problem structure and optimizing the structure require resubmitting the simulation calculation, which also requires engineers to have considerable experience reserves. Later, with the introduction of parametric models, a problem frequency database was established based on the vibration transfer function (NTF)/noise transfer function (VTF) performance of a single vehicle model, and the working mode (ODS) feature database of the relevant frequencies of the vehicle model was also built. When faced with NVH optimization work on model changes or small design changes, engineers can change the structure setting parameters through empirical design, then perform NTF/VTF performance simulation by changing the parameters, obtain a large number of ODS output results, and then complete the response based on these results. The surface method establishes a continuous variable surface model and automatically predicts the output results after providing variable parameters without simulation. However, this method still requires human selection to construct a response surface. Therefore, how to reduce the computing power consumption during NVH analysis has become an urgent problem to be solved.

The above content is only used to assist in understanding the technical solution of the present invention, and does not represent an admission that the above content is prior art.

Contents of the invention

The main purpose of the present invention is to provide a diagnosis and optimization method, system, equipment and storage medium based on weak sheet metal, aiming to solve the technical problem of how to reduce the computing power consumption during NVH analysis.

In order to achieve the above objectives, the present invention provides a diagnosis and optimization method based on weak sheet metal. The diagnosis and optimization method based on weak sheet metal includes:

Identify the frequency of NVH problems after vehicle design changes, and obtain NTF/VTF curve data after vehicle design changes;

Input the NTF/VTF curve data into a preset NVH problem simulation optimization model to obtain a performance problem point diagnosis report corresponding to the NVH problem frequency and a body sheet metal parts optimization plan. The preset NVH problem simulation optimization model Training based on SVM algorithm;

Weak sheet metal locations are processed according to the performance problem point diagnosis report and the body sheet metal parts optimization plan.

Optionally, before the step of identifying the frequency of NVH problems after the vehicle design changes, the step includes:

Collect the NTF/VTF curve data of the target project model, and correlate the corresponding ODS, TPA and panel contribution of the target project model at different frequency excitation points and response points;

Use the NTF/VTF curve data of the target project vehicle model as input parameters, and use the corresponding ODS, TPA and panel contributions under the different frequency excitation points and response points as output parameters;

Based on the input parameters and the output parameters, an initial network model is trained through the SVM algorithm to obtain a preset NVH problem simulation optimization model.

Optionally, the step of training an initial network model through the SVM algorithm based on the input parameters and the output parameters to obtain a preset NVH problem simulation optimization model includes:

Place the input parameters and the output parameters in a target file;

Read the target file according to the target path, and divide the input parameters and output parameters in the target file into a first list and a second list respectively, where the input parameters are associated with the output parameters;

Determine the training set and the test set according to the input parameters in the first list and the output parameters in the second list according to the preset dividing rules;

Train the initial network model through the SVM algorithm according to the training set to obtain the prediction set;

A preset NVH problem simulation optimization model is determined according to the prediction set and the test set.

Optionally, the step of determining a preset NVH problem simulation optimization model based on the prediction set and the test set includes:

Determine the total sample size and the number of parameter points corresponding to the test set;

Calculate an error rate based on the prediction set, the test set, the total sample size and the number of digits of the parameter points;

Determine the test fitting degree according to the total sample size, the number of digits of the parameter points, the average number of the test set, the prediction set and the test set;

Determine whether the error rate satisfies the preset error condition, and whether the test fitting degree satisfies the preset fitting condition;

When the error rate meets the preset error condition and the test fitting degree meets the preset fitting condition, the trained initial network model is used as a preset NVH problem simulation optimization model.

Optionally, the step of determining the degree of test fitting based on the total sample size, the number of digits of the parameter points, the average number of the test set, the prediction set and the test set includes:

Determine the total sample size, the number of parameter points corresponding to the test set, and the average number of the test set;

The test fitting degree is calculated according to the total sample size, the number of digits of the parameter points, the average number of the test set, the prediction set and the test set.

Optionally, after the step of determining whether the error rate satisfies a preset error condition and whether the testing degree of fitting satisfies the preset fitting condition, the step further includes:

When the error rate meets the preset error condition and the test fitting degree does not meet the preset fitting condition, return to the initial network trained through the SVM algorithm based on the input parameters and the output parameters. Model, the steps to obtain the preset NVH problem simulation optimization model.

Optionally, the step of processing weak sheet metal locations based on the performance problem point diagnosis report and the body sheet metal parts optimization plan includes:

Determine the corresponding ODS, TPA and node contribution amount under the NVH problem frequency after the vehicle design change according to the performance problem point diagnosis report;

Locate the weak sheet metal location based on the corresponding ODS, TPA and node contribution under the NVH problem frequency after the vehicle design change;

The weak sheet metal position is optimized through the vehicle body sheet metal parts optimization solution.

In addition, to achieve the above objectives, the present invention also proposes a diagnosis and optimization system based on weak sheet metal. The diagnosis and optimization system based on weak sheet metal includes:

The acquisition module is used to identify the frequency of NVH problems after the vehicle design changes and obtain the NTF/VTF curve data after the vehicle design changes;

A determination module for inputting the NTF/VTF curve data after the vehicle design change into the preset NVH problem simulation optimization model to obtain a performance problem point diagnosis report and body sheet metal part optimization corresponding to the NVH problem frequency. Solution, the preset NVH problem simulation optimization model is trained based on SVM algorithm;

A processing module, configured to process weak sheet metal locations based on the performance problem point diagnosis report and the vehicle body sheet metal parts optimization plan.

In addition, to achieve the above object, the present invention also proposes a weak sheet metal-based diagnosis and optimization device, which includes: a memory, a processor, and a vulnerability-based diagnosis method stored in the memory and capable of running on the processor. A diagnostic optimization program for sheet metal, the diagnostic optimization program based on weak sheet metal is configured to implement the steps of the diagnostic optimization method based on weak sheet metal as described above.

In addition, in order to achieve the above object, the present invention also proposes a storage medium, which stores a diagnostic optimization program based on weak sheet metal. When the diagnostic optimization program based on weak sheet metal is executed by the processor, the above implementation is implemented. The steps of the diagnostic optimization method based on weak sheet metal.

This invention first identifies the NVH problem frequency after the vehicle design is changed, and obtains the NTF/VTF curve data after the vehicle design is changed, and then inputs the NTF/VTF curve data into the preset NVH problem simulation optimization model to obtain the NVH problem The performance problem point diagnosis report and body sheet metal parts optimization plan corresponding to the frequency are preset NVH problem simulation optimization model based on SVM algorithm training, and then the weak sheet metal is processed according to the performance problem point diagnosis report and body sheet metal parts optimization plan. Compared with the prior art, the mechanism modeling method and analysis process are too complex, resulting in the problem that the real-time calculation process consumes a lot of computing power. However, the present invention obtains the NVH status of the vehicle based on the preset NVH problem simulation optimization model trained by the SVM algorithm. Analyze the results, thereby reducing the computing power consumption during NVH analysis.

Description of the drawings

Figure 1 is a schematic structural diagram of a weak sheet metal-based diagnosis and optimization device for the hardware operating environment involved in the embodiment of the present invention;

Figure 2 is a schematic flow chart of the first embodiment of the present invention's diagnosis and optimization method based on weak sheet metal;

Figure 3 is a simulation flow chart of the first embodiment of the present invention's diagnosis and optimization method based on weak sheet metal;

Figure 4 is a schematic flow chart of the second embodiment of the present invention's diagnosis and optimization method based on weak sheet metal;

Figure 5 is a structural block diagram of the first embodiment of the weak sheet metal-based diagnosis and optimization system of the present invention.

The realization of the purpose, functional features and advantages of the present invention will be further described with reference to the embodiments and the accompanying drawings.

Detailed ways

It should be understood that the specific embodiments described here are only used to explain the present invention and are not intended to limit the present invention.

Referring to FIG. 1 , FIG. 1 is a schematic structural diagram of the weak sheet metal-based diagnosis and optimization equipment of the hardware operating environment involved in the solution of the embodiment of the present invention.

As shown in Figure 1, the weak sheet metal-based diagnosis and optimization device may include: a processor 1001, such as a central processing unit (Central Processing Unit, CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. Among them, the communication bus 1002 is used to realize connection communication between these components. The user interface 1003 may include a display screen (Display) and an input unit such as a keyboard (Keyboard). The optional user interface 1003 may also include a standard wired interface and a wireless interface. The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a Wireless-Fidelity (Wi-Fi) interface). The memory 1005 may be a high-speed random access memory (Random Access Memory, RAM) or a stable non-volatile memory (Non-Volatile Memory, NVM), such as a disk memory. The memory 1005 may optionally be a storage system independent of the aforementioned processor 1001.

Those skilled in the art can understand that the structure shown in Figure 1 does not constitute a limitation to the diagnostic optimization device based on weak sheet metal, and may include more or less components than shown, or combine certain components, or different component layout.

As shown in FIG. 1 , the memory 1005 as a storage medium may include an operating system, a network communication module, a user interface module, and a diagnostic optimization program based on weak sheet metal.

In the diagnosis and optimization equipment based on weak sheet metal shown in Figure 1, the network interface 1004 is mainly used for data communication with the network server; the user interface 1003 is mainly used for data interaction with the user; the present invention is based on diagnosis and optimization of weak sheet metal The processor 1001 and the memory 1005 in the device may be configured in the weak sheet metal-based diagnosis and optimization device. The weak sheet metal-based diagnosis and optimization device calls the weak sheet metal-based diagnosis and optimization program stored in the memory 1005 through the processor 1001. , and execute the diagnosis and optimization method based on weak sheet metal provided by the embodiment of the present invention.

Embodiments of the present invention provide a diagnosis and optimization method based on weak sheet metal. Refer to FIG. 2 , which is a schematic flow chart of a first embodiment of a diagnosis and optimization method based on weak sheet metal of the present invention.

In this embodiment, the diagnosis and optimization method based on weak sheet metal includes the following steps:

Step S10: Identify the NVH problem frequency after the vehicle design change, and obtain the NTF/VTF curve data after the vehicle design change.

It is easy to understand that the execution subject of this embodiment can be a weak sheet metal-based diagnosis and optimization system with functions such as data processing, network communication, and program running, or other computer equipment with similar functions. This embodiment does not No restrictions.

Referring to Figure 3, Figure 3 is a simulation flow chart of the first embodiment of the present invention's diagnosis and optimization method based on weak sheet metal. In this embodiment, the frequency of NHV problems after design changes (design changes) are automatically identified through a database.

It should also be noted that the NTF/VTF curve data after the vehicle design change is all NTF/VTF curve data after the vehicle design change.

Step S20: Input the NTF/VTF curve data into the preset NVH problem simulation optimization model to obtain a performance problem point diagnosis report and a body sheet metal part optimization plan corresponding to the NVH problem frequency. The preset NVH problem The simulation optimization model is trained based on the SVM algorithm.

In specific implementation, before identifying the frequency of NVH problems after vehicle design changes, it is necessary to build a preset NVH problem simulation optimization model.

Furthermore, the processing method of constructing the preset NVH problem simulation optimization model is to collect the NTF/VTF curve data of the target project model, and correlate the corresponding ODS, TPA and panel contribution of the target project model under different frequency excitation points and response points; Use the NTF/VTF curve data of the target project model as input parameters, and use the corresponding ODS, TPA and panel contribution amounts under different frequency excitation points and response points as output parameters; based on the input parameters and the output parameters, use the SVM algorithm to initially train Network model to obtain a preset NVH problem simulation optimization model.

It is understandable that the preliminary work of machine learning requires a large amount of data accumulation. The engineer determines the more important parts of the structure and sets the parameters. The parameters include input and output parameters. The input parameter is to collect the VTF/NTF curve of the corresponding project model. The output parameter is the ODS and TPA of the model under different frequency excitation and response points. , panel contribution, etc.

It should also be noted that the VTF/NTF curve corresponding to the project model should be related to the ODS, TPA, and panel contribution of the model under different frequency excitations and response points, and then a parametric model should be established based on the input parameters and output parameters.

In this embodiment, the parametric model is used to avoid the time and computing resources required for a large number of traditional NVH working condition simulations in the later stage.

Further, the initial network model is trained through the SVM algorithm based on the input parameters and output parameters, and the preset NVH problem simulation optimization model is obtained by placing the input parameters and output parameters in the target file; reading the target file according to the target path, and The parameters in the target file are divided into a first list and a second list, and the input parameters are associated with the output parameters; the training set and the test set are determined according to the preset division rules according to the input parameters in the first list and the output parameters in the second list; The initial network model is trained through the SVM algorithm based on the training set to obtain the prediction set; the preset NVH problem simulation optimization model is determined based on the prediction set and test set.

It should also be understood that the preset division rule can be customized by the user, and can be 7:3, 8:2, etc., which is not limited in this embodiment.

The processing method of determining the training set and the test set according to the input parameters in the first list and the output parameters in the second list according to the preset dividing rules is to adjust the rows and columns of the first list and the rows and columns of the second list respectively, so that the first list and the second list are processed. The second list corresponds to the number of parameter points, and the parameters in the adjusted first list and the adjusted second list are divided into training sets and test sets according to the preset dividing rules.

It should also be understood that the DESIGN OF EXPERIMENT DO E is generated by scattering points, and the working condition simulation is performed to obtain enough data points, and the input parameters and output parameters are placed in the same excel table in the format of xls.

In this embodiment, based on the Python language, the Scikit-learn machine learning library is called, the support vector machine algorithm is combined with simulation to perform simulation predictions, and the results that meet the requirements are screened before output, combined with diagnostic methods such as ODS, TPA, and panel contribution. , that is, it can realize the output diagnosis and optimization plan after inputting the NVH working condition curve (VTF/NT F).

Machine learning replaced the previous response analysis method, improving accuracy and further saving some time. In machine learning Python, select the path and read the Excel file, which is the target file. The first step is to divide the data into two lists (the input is the first list and the output is the second list); the second step is to adjust the rows and columns of the two lists so that the number of parameter points corresponds; the third step is to randomly Classify the parameters into training sets and test sets (input parameters and output parameters at the same time), the number can be modified by yourself, 7:3 is recommended, etc. Then change the parameter format to str (string) so that the sklearn library can calculate decimals. After the change, call the SVM algorithm in the sklearn library, input the training set as the input parameter to the initial network model for training, and obtain the prediction results through training, and put the prediction The result is used as an output parameter.

It should also be noted that the selected path is the target path, which can be a preset path. Support vector machine (SVM) is a generalized linear classification algorithm that performs binary classification of data in a supervised learning manner.

Further, the method of determining the preset NVH problem simulation optimization model based on the prediction set and the test set is to determine the total sample size and the number of parameter points corresponding to the test set; according to the prediction set, test set, total sample size and parameters Calculate the error rate based on the number of digits of the points; determine the test fitting degree based on the total sample size, the number of parameter points, the average number of the test set, the prediction set and the test set; determine whether the error rate meets the preset error conditions, and test the fitting degree Whether the preset fitting conditions are met; when the error rate meets the preset error conditions and the test fitting degree meets the preset fitting conditions, the trained initial network model will be used as the preset NVH problem simulation optimization model.

It should also be noted that the preset error condition is a preset error threshold set by the user, and the error rate needs to be less than the preset error threshold. When the error rate is less than the preset error threshold, the preset error condition is met. The preset fitting conditions can be changed according to own needs, and are not limited in this embodiment.

In the specific implementation, the prediction set (training results) is compared with the test set (real results). This embodiment selects two measurement standards to jointly evaluate the completion of machine learning training, one to check the error and the other to check the fitting, and When the results do not meet the requirements, machine learning will be re-executed, which solves the problem of insufficient accuracy of many machine learning algorithms.

The first type: test error rate.

In the formula, T is the test set, P is the prediction set, n is the total sample size, i is the number of digits of parameter points, and T[i] is the i-th digit in the test set.

Further, the method of determining the test fitting degree based on the total sample size, the number of parameter points, the average number of the test set, the prediction set and the test set is to determine the total sample size, the number of parameter points and the test set corresponding to the test set. Average; calculate the test fitting degree based on the total sample size, the number of parameter points, the average of the test set, the prediction set and the test set.

The second type: test the degree of fit.

In the formula, a is the average of all test sets for this parameter, which is the average of the test set, P is the prediction set, T is the test set, n is the total sample size, i is the number of parameter points, and R square is the test fitting degree. .

In this embodiment, if the calculation result of R square does not meet the set conditions (can be changed according to your own needs), then return to the initial network model trained through the SVM algorithm based on the input parameters and output parameters to obtain the preset NVH problem simulation Optimize the operation of the model, and the machine learning data verification results will be output in the form of images and text.

The image is a comparison of all parameter points between the true value and the predicted value of each output parameter. The text is the specific value of the error rate and R-square of each output parameter. When the error rate and R-square reach the training set target, the project The vehicle model NVH problem data training is completed and can be put into actual use.

In this embodiment, the NTF and VTF curves after subsequent design changes of the research model are input, and the database completed through machine learning training is used to output a diagnosis report of NVH performance problem points (non-standard items) under the design change model of the project, and at the same time output the problem The optimization plan of body sheet metal parts under frequency completes the analysis closed loop.

After the preset NVH problem simulation optimization model is put into actual use, the frequency of NVH problems after the design change is automatically identified through the database, and then the NTF/VTF curve data after the vehicle design change is input into the preset NVH problem simulation optimization model to obtain Diagnostic reports on performance problem points corresponding to the frequency of NVH problems and optimization plans for body sheet metal parts.

Step S30: Process weak sheet metal locations according to the performance problem point diagnosis report and the body sheet metal parts optimization plan.

In this embodiment, the ODS, TPA and node contribution corresponding to the NVH problem frequency after the vehicle design change are determined based on the performance problem point diagnosis report, and then the ODS, TPA corresponding to the NVH problem frequency after the vehicle design change is determined. and node contribution to locate the weak sheet metal position, and then optimize the weak sheet metal through the body sheet metal parts optimization plan based on the weak sheet metal position. In the simulation analysis of actual NVH problems, the analysis process of analysis items such as ODS, TPA, panel contribution, etc., which are highly dependent on the experience of engineers and require multiple verifications, is redefined. This embodiment implements single input (NTF) through machine learning training. /VTF) multi-working condition output (problem frequency, boards, optimization suggestions), thereby improving simulation efficiency.

In this embodiment, the NVH problem frequency after the vehicle design change is first identified, and the NTF/VTF curve data after the vehicle design change is obtained, and then the NTF/VTF curve data is input into the preset NVH problem simulation optimization model. Obtain the performance problem diagnosis report and body sheet metal parts optimization plan corresponding to the NVH problem frequency. The preset NVH problem simulation optimization model is trained based on the SVM algorithm, and then the weak sheet metal is analyzed based on the performance problem point diagnosis report and the body sheet metal parts optimization plan. for processing. Compared with the existing technology, the mechanism modeling method and analysis process are too complex, resulting in the problem that the real-time calculation process consumes a lot of computing power. However, this embodiment obtains the NVH of the vehicle based on the preset NVH problem simulation optimization model trained by the SVM algorithm. Status analysis results, thus reducing the computing power consumption during NVH analysis.

Referring to Figure 4, Figure 4 is a schematic flow chart of a second embodiment of the present invention's diagnosis and optimization method based on weak sheet metal.

Based on the above first embodiment, in this embodiment, before step S10, it also includes:

Step S01: Collect the NTF/VTF curve data of the target project model, and associate the corresponding ODS, TPA and panel contribution of the target project model at different frequency excitation points and response points.

Step S02: Use the NTF/VTF curve data of the target project model as input parameters, and use the corresponding ODS, TPA and panel contribution amounts under the different frequency excitation points and response points as output parameters.

It is understandable that the preliminary work of machine learning requires a large amount of data accumulation. The engineer determines the more important parts of the structure and sets the parameters. The parameters include input and output parameters. The input parameter is to collect the VTF/NTF curve of the corresponding project model. The output parameter is the ODS and TPA of the model under different frequency excitation and response points. , panel contribution, etc.

It should also be noted that the VTF/NTF curve corresponding to the project model should be related to the ODS, TPA, and panel contribution of the model under different frequency excitations and response points, and then a parametric model should be established based on the input parameters and output parameters.

In this embodiment, the parametric model is used to avoid the time and computing resources required for a large number of traditional NVH working condition simulations in the later stage.

Step S03: Train an initial network model through the SVM algorithm based on the input parameters and the output parameters to obtain a preset NVH problem simulation optimization model.

Further, the initial network model is trained through the SVM algorithm based on the input parameters and output parameters, and the preset NVH problem simulation optimization model is obtained by placing the input parameters and output parameters in the target file; reading the target file according to the target path, and The parameters in the target file are divided into a first list and a second list; adjust the rows and columns of the first list and the second list respectively so that the first list and the second list correspond to the number of parameter points; The parameters in the first list and the adjusted second list are divided into a training set and a test set; the initial network model is trained through the SVM algorithm based on the training set to obtain a prediction set; the preset NVH problem simulation optimization model is determined based on the prediction set and the test set.

It should also be understood that the DESIGN OF EXPERIMENT DO E is generated by scattering points, and the working condition simulation is performed to obtain enough data points, and the input parameters and output parameters are placed in the same excel table in the format of xls.

In this embodiment, based on the Python language, the Scikit-learn machine learning library is called, the support vector machine algorithm is combined with simulation to perform simulation predictions, and the results that meet the requirements are screened before output, combined with diagnostic methods such as ODS, TPA, and panel contribution. , that is, it can realize the output diagnosis and optimization plan after inputting the NVH working condition curve (VTF/NT F).

Machine learning replaced the previous response analysis method, improving accuracy and further saving some time. In machine learning Python, select the path and read the Excel file, which is the target file. The first step is to divide the data into two lists (the input is the first list and the output is the second list); the second step is to adjust the rows and columns of the two lists so that the number of parameter points corresponds; the third step is to randomly Classify the parameters into training sets and test sets (input parameters and output parameters at the same time), the number can be modified by yourself, 7:3 is recommended, etc. Then change the parameter format to str (string) so that the sklearn library can calculate decimals. After the change, call the SVM algorithm in the sklearn library, input the training set as the input parameter to the initial network model for training, and obtain the prediction results through training, and put the prediction The result is used as an output parameter.

It should also be noted that the selected path is the target path, which can be a preset path. Support vector machine (SVM) is a generalized linear classification algorithm that performs binary classification of data in a supervised learning manner.

Further, the method of determining the preset NVH problem simulation optimization model based on the prediction set and the test set is to determine the total sample size and the number of parameter points corresponding to the test set; according to the prediction set, test set, total sample size and parameters Calculate the error rate based on the number of digits of the points; determine the test fitting degree based on the total sample size, the number of parameter points, the average number of the test set, the prediction set and the test set; determine whether the error rate meets the preset error conditions, and test the fitting degree Whether the preset fitting conditions are met; when the error rate meets the preset error conditions and the test fitting degree meets the preset fitting conditions, the trained initial network model will be used as the preset NVH problem simulation optimization model.

It should also be noted that the preset error condition is a preset error threshold set by the user, and the error rate needs to be less than the preset error threshold. When the error rate is less than the preset error threshold, the preset error condition is met. The preset fitting conditions can be changed according to own needs, and are not limited in this embodiment.

In the specific implementation, the prediction set (training results) is compared with the test set (real results). This embodiment selects two measurement standards to jointly evaluate the completion of machine learning training, one to check the error and the other to check the fitting, and When the results do not meet the requirements, machine learning will be re-executed, which solves the problem of insufficient accuracy of many machine learning algorithms.

The first type: test error rate.

In the formula, T is the test set, P is the prediction set, n is the total sample size, i is the number of digits of parameter points, and T[i] is the i-th digit in the test set.

Further, the method of determining the test fitting degree based on the total sample size, the number of parameter points, the average number of the test set, the prediction set and the test set is to determine the total sample size, the number of parameter points and the test set corresponding to the test set. Average; calculate the test fitting degree based on the total sample size, the number of parameter points, the average of the test set, the prediction set and the test set.

The second type: test the degree of fit.

In the formula, a is the average of all test sets for this parameter, which is the average of the test set, P is the prediction set, T is the test set, n is the total sample size, i is the number of parameter points, and R square is the test fitting degree. .

In this embodiment, if the calculation result of R square does not meet the set conditions (can be changed according to your own needs), then return to the initial network model trained through the SVM algorithm based on the input parameters and output parameters to obtain the preset NVH problem simulation Optimize the operation of the model, and the machine learning data verification results will be output in the form of images and text.

The image is a comparison of all parameter points between the true value and the predicted value of each output parameter. The text is the specific value of the error rate and R-square of each output parameter. When the error rate and R-square reach the training set target, the project The vehicle model NVH problem data training is completed and can be put into actual use.

In this embodiment, the NTF and VTF curves after subsequent design changes of the research model are input, and the database completed through machine learning training is used to output a diagnosis report of NVH performance problem points (non-standard items) under the design change model of the project, and at the same time output the problem The optimization plan of body sheet metal parts under frequency completes the analysis closed loop.

In this embodiment, the NTF/VTF curve data of the target project model is first collected, and associated with the corresponding ODS, TPA and panel contribution of the target project model under different frequency excitation points and response points, and then the NTF/VTF curve data of the target project model is The VTF curve data is used as input parameters, and the corresponding ODS, TPA and panel contribution under different frequency excitation points and response points are used as output parameters. Then the initial network model is trained through the SVM algorithm based on the input parameters and the output parameters to obtain the preset The NVH problem simulation optimization model realizes a closed loop of simulation analysis and optimization of inputting the NTF/VTF curve after design changes when the design of a certain vehicle model project is changed, and outputting the problem frequency sheet metal diagnosis and optimization plan, thereby improving simulation efficiency.

Referring to Figure 5, Figure 5 is a structural block diagram of the first embodiment of the diagnosis and optimization system based on weak sheet metal of the present invention.

As shown in Figure 5, the diagnosis and optimization system based on weak sheet metal proposed by the embodiment of the present invention includes:

The acquisition module 5001 is used to identify the frequency of NVH problems after the vehicle design changes and obtain the NTF/VTF curve data after the vehicle design changes.

It is easy to understand that the execution subject of this embodiment can be a weak sheet metal-based diagnosis and optimization system with functions such as data processing, network communication, and program running, or other computer equipment with similar functions. This embodiment does not No restrictions.

Referring to Figure 3, Figure 3 is a simulation flow chart of the first embodiment of the present invention's diagnosis and optimization method based on weak sheet metal. In this embodiment, the frequency of NHV problems after design changes (design changes) are automatically identified through a database.

It should also be noted that the NTF/VTF curve data after the vehicle design change is all NTF/VTF curve data after the vehicle design change.

The determination module 5002 is used to input the NTF/VTF curve data after the vehicle design change into the preset NVH problem simulation optimization model to obtain a performance problem point diagnosis report and body sheet metal parts corresponding to the NVH problem frequency. Optimization scheme, the preset NVH problem simulation optimization model is trained based on SVM algorithm.

In specific implementation, before identifying the frequency of NVH problems after vehicle design changes, it is necessary to build a preset NVH problem simulation optimization model.

Furthermore, the processing method of constructing the preset NVH problem simulation optimization model is to collect the NTF/VTF curve data of the target project model, and correlate the corresponding ODS, TPA and panel contribution of the target project model under different frequency excitation points and response points; Use the NTF/VTF curve data of the target project model as input parameters, and use the corresponding ODS, TPA and panel contribution amounts under different frequency excitation points and response points as output parameters; based on the input parameters and the output parameters, use the SVM algorithm to initially train Network model to obtain a preset NVH problem simulation optimization model.

It is understandable that the preliminary work of machine learning requires a large amount of data accumulation. The engineer determines the more important parts of the structure and sets the parameters. The parameters include input and output parameters. The input parameter is to collect the VTF/NTF curve of the corresponding project model. The output parameter is the ODS and TPA of the model under different frequency excitation and response points. , panel contribution, etc.

It should also be noted that the VTF/NTF curve corresponding to the project model should be related to the ODS, TPA, and panel contribution of the model under different frequency excitations and response points, and then a parametric model should be established based on the input parameters and output parameters.

In this embodiment, the parametric model is used to avoid the time and computing resources required for a large number of traditional NVH working condition simulations in the later stage.

Further, the initial network model is trained through the SVM algorithm based on the input parameters and output parameters, and the preset NVH problem simulation optimization model is obtained by placing the input parameters and output parameters in the target file; reading the target file according to the target path, and The parameters in the target file are divided into a first list and a second list, and the input parameters are associated with the output parameters; the training set and the test set are determined according to the preset division rules according to the input parameters in the first list and the output parameters in the second list; The initial network model is trained through the SVM algorithm based on the training set to obtain the prediction set; the preset NVH problem simulation optimization model is determined based on the prediction set and test set.

It should also be understood that the preset division rule can be customized by the user, and can be 7:3, 8:2, etc., which is not limited in this embodiment.

The processing method of determining the training set and the test set according to the input parameters in the first list and the output parameters in the second list according to the preset dividing rules is to adjust the rows and columns of the first list and the rows and columns of the second list respectively, so that the first list and the second list are processed. The second list corresponds to the number of parameter points, and the parameters in the adjusted first list and the adjusted second list are divided into training sets and test sets according to the preset dividing rules.

It should also be understood that the DESIGN OF EXPERIMENT DO E is generated by scattering points, and the working condition simulation is performed to obtain enough data points, and the input parameters and output parameters are placed in the same excel table in the format of xls.

In this embodiment, based on the Python language, the Scikit-learn machine learning library is called, the support vector machine algorithm is combined with simulation to perform simulation predictions, and the results that meet the requirements are screened before output, combined with diagnostic methods such as ODS, TPA, and panel contribution. , that is, it can realize the output diagnosis and optimization plan after inputting the NVH working condition curve (VTF/NT F).

Machine learning replaced the previous response analysis method, improving accuracy and further saving some time. In machine learning Python, select the path and read the Excel file, which is the target file. The first step is to divide the data into two lists (the input is the first list and the output is the second list); the second step is to adjust the rows and columns of the two lists so that the number of parameter points corresponds; the third step is to randomly Classify the parameters into training sets and test sets (input parameters and output parameters at the same time), the number can be modified by yourself, 7:3 is recommended, etc. Then change the parameter format to str (string) so that the sklearn library can calculate decimals. After the change, call the SVM algorithm in the sklearn library, input the training set as the input parameter to the initial network model for training, and obtain the prediction results through training, and put the prediction The result is used as an output parameter.

It should also be noted that the selected path is the target path, which can be a preset path. Support vector machine (SVM) is a generalized linear classification algorithm that performs binary classification of data in a supervised learning manner.

Further, the method of determining the preset NVH problem simulation optimization model based on the prediction set and the test set is to determine the total sample size and the number of parameter points corresponding to the test set; according to the prediction set, test set, total sample size and parameters Calculate the error rate based on the number of digits of the points; determine the test fitting degree based on the total sample size, the number of parameter points, the average number of the test set, the prediction set and the test set; determine whether the error rate meets the preset error conditions, and test the fitting degree Whether the preset fitting conditions are met; when the error rate meets the preset error conditions and the test fitting degree meets the preset fitting conditions, the trained initial network model will be used as the preset NVH problem simulation optimization model.

It should also be noted that the preset error condition is a preset error threshold set by the user, and the error rate needs to be less than the preset error threshold. When the error rate is less than the preset error threshold, the preset error condition is met. The preset fitting conditions can be changed according to own needs, and are not limited in this embodiment.

In the specific implementation, the prediction set (training results) is compared with the test set (real results). This embodiment selects two measurement standards to jointly evaluate the completion of machine learning training, one to check the error and the other to check the fitting, and When the results do not meet the requirements, machine learning will be re-executed, which solves the problem of insufficient accuracy of many machine learning algorithms.

The first type: test error rate.

In the formula, T is the test set, P is the prediction set, n is the total sample size, i is the number of digits of parameter points, and T[i] is the i-th digit in the test set.

Further, the method of determining the test fitting degree based on the total sample size, the number of parameter points, the average number of the test set, the prediction set and the test set is to determine the total sample size, the number of parameter points and the test set corresponding to the test set. Average; calculate the test fitting degree based on the total sample size, the number of parameter points, the average of the test set, the prediction set and the test set.

The second type: test the degree of fit.

In the formula, a is the average of all test sets for this parameter, which is the average of the test set, P is the prediction set, T is the test set, n is the total sample size, i is the number of parameter points, and R square is the test fitting degree. .

In this embodiment, if the calculation result of R square does not meet the set conditions (can be changed according to your own needs), then return to the initial network model trained through the SVM algorithm based on the input parameters and output parameters to obtain the preset NVH problem simulation Optimize the operation of the model, and the machine learning data verification results will be output in the form of images and text.

The image is a comparison of all parameter points between the true value and the predicted value of each output parameter. The text is the specific value of the error rate and R-square of each output parameter. When the error rate and R-square reach the training set target, the project The vehicle model NVH problem data training is completed and can be put into actual use.

In this embodiment, the NTF and VTF curves after subsequent design changes of the research model are input, and the database completed through machine learning training is used to output a diagnosis report of NVH performance problem points (non-standard items) under the design change model of the project, and at the same time output the problem The optimization plan of body sheet metal parts under frequency completes the analysis closed loop.

After the preset NVH problem simulation optimization model is put into actual use, the frequency of NVH problems after the design change is automatically identified through the database, and then the NTF/VTF curve data after the vehicle design change is input into the preset NVH problem simulation optimization model to obtain Diagnostic reports on performance problem points corresponding to the frequency of NVH problems and optimization plans for body sheet metal parts.

The processing module 5003 is used to process weak sheet metal locations based on the performance problem point diagnosis report and the vehicle body sheet metal parts optimization plan.

In this embodiment, the ODS, TPA and node contribution corresponding to the NVH problem frequency after the vehicle design change are determined based on the performance problem point diagnosis report, and then the ODS, TPA corresponding to the NVH problem frequency after the vehicle design change is determined. and node contribution to locate the weak sheet metal position, and then optimize the weak sheet metal through the body sheet metal parts optimization plan based on the weak sheet metal position. In the simulation analysis of actual NVH problems, the analysis process of analysis items such as ODS, TPA, panel contribution, etc., which are highly dependent on the experience of engineers and require multiple verifications, is redefined. This embodiment implements single input (NTF) through machine learning training. /VTF) multi-working condition output (problem frequency, boards, optimization suggestions), thereby improving simulation efficiency.

In this embodiment, the NVH problem frequency after the vehicle design change is first identified, and the NTF/VTF curve data after the vehicle design change is obtained, and then the NTF/VTF curve data is input into the preset NVH problem simulation optimization model. Obtain the performance problem diagnosis report and body sheet metal parts optimization plan corresponding to the NVH problem frequency. The preset NVH problem simulation optimization model is trained based on the SVM algorithm, and then the weak sheet metal is analyzed based on the performance problem point diagnosis report and the body sheet metal parts optimization plan. for processing. Compared with the existing technology, the mechanism modeling method and analysis process are too complex, resulting in the problem that the real-time calculation process consumes a lot of computing power. However, this embodiment obtains the NVH of the vehicle based on the preset NVH problem simulation optimization model trained by the SVM algorithm. Status analysis results, thus reducing the computing power consumption during NVH analysis.

In this embodiment, the NVH problem frequency after the vehicle design change is first identified, and then the NTF/VTF curve data after the vehicle design change is input into the preset NVH problem simulation optimization model to obtain the performance corresponding to the NVH problem frequency. The problem point diagnosis report and body sheet metal parts optimization plan, the preset NVH problem simulation optimization model is trained based on the SVM algorithm, and then the weak sheet metal is processed based on the performance problem point diagnosis report and the body sheet metal parts optimization plan. Compared with the existing technology, the mechanism modeling method and analysis process are too complex, resulting in the problem that the real-time calculation process consumes a lot of computing power. However, this embodiment obtains the NVH of the vehicle based on the preset NVH problem simulation optimization model trained by the SVM algorithm. Status analysis results, thus reducing the computing power consumption during NVH analysis.

For other embodiments or specific implementations of the weak sheet metal-based diagnosis and optimization system of the present invention, reference may be made to the above method embodiments, which will not be described again here.

It should be noted that, as used herein, the terms "include", "comprising" or any other variation thereof are intended to cover a non-exclusive inclusion, such that a process, method, article or system that includes a list of elements not only includes those elements, but It also includes other elements not expressly listed or that are inherent to the process, method, article or system. Without further limitation, an element defined by the statement "comprises a..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element.

The above serial numbers of the embodiments of the present invention are only for description and do not represent the advantages and disadvantages of the embodiments.

Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus the necessary general hardware platform. Of course, it can also be implemented by hardware, but in many cases the former is better. implementation. Based on this understanding, the technical solution of the present invention can be embodied in the form of a software product in nature or in part that contributes to the existing technology. The computer software product is stored in a storage medium (such as read-only memory/random access memory). memory, magnetic disk, optical disk), including a number of instructions to cause a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the method described in various embodiments of the present invention.

The above are only preferred embodiments of the present invention, and do not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the description and drawings of the present invention may be directly or indirectly used in other related technical fields. , are all similarly included in the scope of patent protection of the present invention.

Claims (10)

1. The diagnosis optimization method based on the weak sheet metal is characterized by comprising the following steps of:

identifying NVH problem frequency after the vehicle design is increased, and acquiring NTF/VTF curve data after the vehicle design is increased;

inputting the NTF/VTF curve data into a preset NVH problem simulation optimization model to obtain a performance problem point diagnosis report and a vehicle body sheet metal part optimization scheme corresponding to the NVH problem frequency, wherein the preset NVH problem simulation optimization model is trained based on an SVM algorithm;

and processing the weak sheet metal position according to the performance problem point diagnosis report and the optimization scheme of the sheet metal part of the vehicle body.

2. The method of claim 1, wherein prior to the step of identifying the increased frequency of NVH problems with the vehicle design, comprising:

collecting NTF/VTF curve data of a target project vehicle model, and associating ODS, TPA and panel contribution amounts corresponding to the target project vehicle model under different frequency excitation points and response points;

taking NTF/VTF curve data of the target project vehicle model as input parameters, and taking corresponding ODS, TPA and panel contribution under the excitation points and the response points with different frequencies as output parameters;

And training an initial network model through an SVM algorithm based on the input parameters and the output parameters to obtain a preset NVH problem simulation optimization model.

3. The method of claim 2, wherein the step of training an initial network model by an SVM algorithm based on the input parameters and the output parameters to obtain a preset NVH problem simulation optimization model comprises:

placing the input parameters and the output parameters in a target file;

reading the target file according to a target path, and dividing input parameters and output parameters in the target file into a first list and a second list respectively, wherein the input parameters and the output parameters have an association relation;

determining a training set and a testing set according to a preset dividing rule according to the input parameters in the first list and the output parameters in the second list;

training an initial network model through an SVM algorithm according to the training set to obtain a prediction set;

and determining a preset NVH problem simulation optimization model according to the prediction set and the test set.

4. The method of claim 3, wherein the step of determining a pre-set NVH problem simulation optimization model from the prediction set and the test set comprises:

Determining the total sample size corresponding to the test set and the bit number of the parameter points;

calculating an error rate according to the prediction set, the test set, the total sample size and the bit number of the parameter points;

determining a test fitting degree according to the total sample size, the bit number of the parameter points, the average number of the test sets, the prediction set and the test set;

judging whether the error rate meets a preset error condition or not, and judging whether the test fitting degree meets a preset fitting condition or not;

and when the error rate meets the preset error condition and the test fitting degree meets the preset fitting condition, taking the trained initial network model as a preset NVH problem simulation optimization model.

5. The method of claim 4, wherein the step of determining a test fit from the total sample size, the number of parameter points, a test set average, the prediction set, and the test set comprises:

determining the total sample size, the bit number of the parameter points and the average number of the test sets corresponding to the test sets;

and calculating the test fitting degree according to the total sample size, the bit number of the parameter points, the average number of the test sets, the prediction set and the test set.

6. The method of claim 4, wherein after the step of determining whether the error rate satisfies a predetermined error condition and the test fitness satisfies a predetermined fit condition, further comprising:

and returning to the step of training an initial network model through an SVM algorithm based on the input parameter and the output parameter to obtain a preset NVH problem simulation optimization model when the error rate meets the preset error condition and the test fitting degree does not meet the preset fitting condition.

7. The method of any one of claims 1-6, wherein the step of processing the weak sheet metal location based on the performance issue point diagnostic report and the body sheet metal optimization scheme comprises:

determining corresponding ODS, TPA and node contribution amounts under the NVH problem frequency after the vehicle design is increased according to the performance problem point diagnosis report;

positioning weak sheet metal positions according to corresponding ODS, TPA and node contribution amounts under the NVH problem frequency after the vehicle design is increased;

and optimizing the weak sheet metal position through the optimization scheme of the sheet metal part of the vehicle body.

8. Diagnosis optimizing system based on weak panel beating, characterized by, diagnosis optimizing system based on weak panel beating includes:

The acquisition module is used for identifying NVH problem frequency after the vehicle design is increased and acquiring NTF/VTF curve data after the vehicle design is increased;

the determining module is used for inputting the NTF/VTF curve data after the vehicle design is changed into a preset NVH problem simulation optimization model so as to obtain a performance problem point diagnosis report corresponding to the NVH problem frequency and a vehicle body sheet metal part optimization scheme, and the preset NVH problem simulation optimization model is trained based on an SVM algorithm;

and the processing module is used for processing the weak sheet metal position according to the performance problem point diagnosis report and the optimization scheme of the sheet metal part of the vehicle body.

9. A diagnostic optimizing device based on weak sheet metal, characterized in that it comprises: a memory, a processor and a sheet metal weakness-based diagnostic optimizer stored on the memory and operable on the processor, the sheet metal weakness-based diagnostic optimizer configured to implement the steps of the sheet metal weakness-based diagnostic optimization method of any one of claims 1 to 7.

10. A storage medium, wherein a weak sheet metal based diagnostic optimization program is stored on the storage medium, which when executed by a processor, implements the steps of the weak sheet metal based diagnostic optimization method according to any one of claims 1 to 7.