CN115828751A - Data-driven prediction and analysis method for NVH diagnosis of vehicle transmission system - Google Patents
Data-driven prediction and analysis method for NVH diagnosis of vehicle transmission system Download PDFInfo
- Publication number
- CN115828751A CN115828751A CN202211561562.4A CN202211561562A CN115828751A CN 115828751 A CN115828751 A CN 115828751A CN 202211561562 A CN202211561562 A CN 202211561562A CN 115828751 A CN115828751 A CN 115828751A
- Authority
- CN
- China
- Prior art keywords
- nvh
- prediction
- vehicle
- data set
- analysis
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T90/00—Enabling technologies or technologies with a potential or indirect contribution to GHG emissions mitigation
Landscapes
- Testing And Monitoring For Control Systems (AREA)
Abstract
The application discloses a data-driven type prediction and analysis method for vehicle transmission system NVH diagnosis, which comprises the following steps: acquiring a collected vehicle state signal; and inputting the vehicle state signal into a pre-trained NVH prediction and diagnosis model for analysis to obtain an NVH state analysis result of the vehicle, wherein the NVH prediction and diagnosis model is constructed by adopting a random forest algorithm. The method solves the problem that the traditional mechanism-based modeling method and analysis process are too complex, so that a large amount of computing power is consumed in the real-time computing process, reduces the computing power consumption in the NVH analysis, and is more suitable for running on the TCU.
Description
Technical Field
The application relates to the technical field of vehicle NVH, in particular to a data driving type prediction and analysis method for vehicle transmission system NVH diagnosis.
Background
Performance issues of sound, vibration and Harshness (NVH) not only affect ride comfort of the vehicle, but also relate to service life and performance of relevant vehicle components and systems, and among them the vehicle driveline is a major source of vehicle NVH problems. NVH phenomena of the vehicle driveline include juder phenomenon and Shuffle phenomenon. The Juder phenomenon is derived from low-frequency longitudinal vibration of a vehicle body with the frequency of 5-20Hz during clutch combination, is characterized by showing a first-order mode of the system, is mostly seen in a vehicle starting state, and can be observed in a gear shifting working condition to be weak. The Shuffle phenomenon is low-frequency longitudinal and pitching vibration of 2-10Hz caused by transient torque of a transmission system generated after clutch engagement is completed in the processes of rapid acceleration and deceleration or starting and gear shifting, and is characterized by a first-order mode of the system. Therefore, under starting and gear shifting conditions, the Judder and Shuffle phenomena can be regarded as low-frequency vibration occurring before and after clutch engagement is completed.
With the increasingly complex urban road traffic conditions, a large number of vehicles need to frequently start and stop and shift gears during the driving process, so that passengers feel strong vibration and impact, and the riding experience of the passengers is influenced. Accordingly, the demand for accurate NVH analysis techniques is in a continuously growing state. Traditional NVH analysis often employs model-based methods to study relevant mechanistic features. Data are generally acquired through sensors and microphones which are arranged, and then the data are processed and analyzed by using methods such as time domain analysis, frequency domain analysis, time-frequency domain analysis, order domain analysis and the like based on interdisciplinary knowledge such as structural dynamics, signal processing and acoustics. The mechanism-based modeling method and the analysis flow are too complex, so that a large amount of computing power is consumed in the real-time computing process, and a large burden is caused on a vehicle TCU.
Therefore, in view of the above problems, it is necessary to provide a solution for reducing the computational power consumption in the NVH analysis.
Disclosure of Invention
The application mainly aims to provide a data-driven type prediction and analysis method for vehicle transmission system NVH diagnosis, and aims to solve the problem that a large amount of computing power is consumed in a real-time operation process due to the fact that a traditional mechanism-based modeling method and an analysis process are too complex, reduce computing power consumption in NVH analysis and be more suitable for running on a TCU.
To achieve the above object, the present application provides a data-driven prediction and analysis method for vehicle driveline NVH diagnosis, including:
acquiring a collected vehicle state signal;
and inputting the vehicle state signal into a pre-trained NVH prediction and diagnosis model for analysis to obtain an NVH state analysis result of the vehicle, wherein the NVH prediction and diagnosis model is constructed by adopting a random forest algorithm.
Optionally, before the step of acquiring the collected vehicle state signal, the method further includes:
and training based on a random forest algorithm to obtain the NVH prediction diagnosis model, wherein the random forest algorithm adopts a CART algorithm.
Optionally, the step of training to obtain the NVH predictive diagnosis model based on the random forest algorithm includes:
acquiring a training data set acquired in advance, wherein the training data set consists of real vehicle experimental data acquired by a sensor;
obtaining a kini index of the data set according to the training data set;
selecting a classification feature according to the training data set;
obtaining the Gini index of the data set under the condition of the classification features according to the classification features and the Gini index of the data set; and returning to the execution step: selecting a classification feature according to the training data set;
and calculating the Gini index of the data set under each classification characteristic condition by the loop to obtain the classification characteristic with the minimum Gini index, and generating the NVH prediction diagnosis model according to the classification characteristic with the minimum Gini index.
Optionally, after the step of generating the NVH predictive diagnostic model according to the classification feature with the minimum kini index, the method further includes:
calculating to obtain a contribution value of the input features of the NVH prediction diagnosis model by an interpretable machine learning SHAP method;
analyzing the contribution value to obtain a contribution analysis result of the input feature to the NVH prediction diagnosis model;
performing feature importance analysis on the input features according to the contribution analysis result to obtain an input and output correlation analysis result of the NVH prediction diagnosis model, wherein the input and output correlation analysis result comprises unimportant non-correlation features;
screening the training data set according to the unimportant non-relevant features to generate a screened training data set; and returning to the execution step: obtaining a kini index of the data set according to the training data set;
and terminating the training until the contribution value of the input feature meets a preset contribution value, so as to obtain the trained NVH prediction diagnosis model.
Optionally, after the step of acquiring a pre-acquired training data set, the method further includes:
preprocessing the training data set.
Optionally, the sensors include an engine speed sensor and a transmission input shaft speed sensor, the real vehicle experimental data includes an engine speed and a transmission input shaft speed, and the step of preprocessing the training data set includes:
acquiring the engine speed acquired by the engine speed sensor, and acquiring the transmission input shaft speed acquired by the transmission input shaft speed sensor;
and carrying out difference processing according to the rotating speed of the engine and the rotating speed of the input shaft of the transmission to obtain the rotating speed difference of the same dimension.
Optionally, after the step of obtaining the trained NVH prediction diagnosis model, the method further includes:
and verifying the prediction accuracy of the NVH prediction diagnosis model through a pre-collected test data set, wherein the test data set is composed of real vehicle experiment data different from the training data set.
An embodiment of the present application further provides a data-driven type prediction and analysis apparatus for vehicle driveline NVH diagnosis, including:
the signal acquisition module is used for acquiring the acquired vehicle state signal;
and the state prediction module is used for inputting the vehicle state signal into a pre-trained NVH prediction and diagnosis model for analysis to obtain an NVH state analysis result of the vehicle, wherein the NVH prediction and diagnosis model is constructed by adopting a random forest algorithm.
The embodiment of the present application also provides an on-board device, which includes a memory, a processor, and a data-driven type prediction and analysis program for vehicle driveline NVH diagnosis stored on the memory and executable on the processor, wherein the data-driven type prediction and analysis program for vehicle driveline NVH diagnosis implements the steps of the data-driven type prediction and analysis method for vehicle driveline NVH diagnosis as described above when executed by the processor.
Embodiments of the present application also provide a computer readable storage medium having stored thereon a data-driven type prognostics and analysis program for vehicle driveline NVH diagnostics, which when executed by a processor, implements the steps of the data-driven type prognostics and analysis method for vehicle driveline NVH diagnostics as described above.
According to the data-driven type prediction and analysis method and device for vehicle transmission system NVH diagnosis, the vehicle-mounted equipment and the storage medium, the acquired vehicle state signals are acquired; and inputting the vehicle state signal into a pre-trained NVH prediction and diagnosis model for analysis to obtain an NVH state analysis result of the vehicle, wherein the NVH prediction and diagnosis model is constructed by adopting a random forest algorithm. The NVH state of the vehicle is analyzed through the data-driven vehicle NVH prediction and diagnosis model based on the random forest, the NVH analysis difficulty can be reduced, and the problem that a large amount of computing power is consumed in the real-time operation process due to the fact that a traditional mechanism-based modeling method and an analysis process are too complex is solved. Based on this application scheme, will gather and relate to the real car experimental data of a plurality of operating modes and make into the data set, train based on this data set and obtain NVH prediction diagnosis model, can be used for accurately judging the vehicle in the operation process, the NVH state that acts on the whole car that vehicle transmission system produced to can the analysis obtain other vehicle states to the influence that NVH state produced, and then provide the direction for the improvement of NVH control strategy. According to the scheme, the off-line model can be obtained by adopting the random forest algorithm, the model only needs to consume a large amount of computing power during training, the branch prediction structure of the trained random forest model is simple, the computing power consumed in actual operation is small, and the model is more suitable for operation on a TCU.
Drawings
FIG. 1 is a schematic functional block diagram of an on-board device to which a data-driven predictive and analytical apparatus for vehicle driveline NVH diagnostics of the present application pertains;
FIG. 2 is a schematic flow chart diagram of a first exemplary embodiment of a data-driven prediction and analysis method for vehicle driveline NVH diagnostics according to the present application;
FIG. 3 is a schematic flow chart diagram of a second exemplary embodiment of a data-driven prediction and analysis method for vehicle driveline NVH diagnostics according to the present application;
FIG. 4 is a schematic flow chart illustrating a training process of an NVH prediction diagnosis model according to an embodiment of the data-driven prediction and analysis method for vehicle driveline NVH diagnosis of the present application;
FIG. 5 is a schematic flow chart diagram illustrating a third exemplary embodiment of a data-driven prediction and analysis method for vehicle driveline NVH diagnostics according to the present application.
The implementation, functional features and advantages of the objectives of the present application will be further explained with reference to the accompanying drawings.
Detailed Description
It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.
The main solution of the embodiment of the application is as follows: training based on a random forest algorithm to obtain an NVH prediction diagnosis model, wherein the random forest algorithm adopts a CART algorithm; acquiring a collected vehicle state signal; and inputting the vehicle state signal into a pre-trained NVH prediction diagnosis model for analysis to obtain an NVH state analysis result of the vehicle. The NVH state of the vehicle is analyzed through the data-driven vehicle NVH prediction and diagnosis model based on the random forest, the NVH analysis difficulty can be reduced, and the problem that a large amount of computing power is consumed in the real-time operation process due to the fact that a traditional mechanism-based modeling method and an analysis process are too complex is solved. Based on this application scheme, will gather and relate to the real car experimental data of a plurality of operating modes and make into the data set, train based on this data set and obtain NVH prediction diagnosis model, can be used for accurately judging the vehicle in the operation process, the NVH state that acts on the whole car that vehicle transmission system produced to can the analysis obtain other vehicle states to the influence that NVH state produced, and then provide the direction for the improvement of NVH control strategy. According to the scheme, the off-line model can be obtained by adopting the random forest algorithm, the model only needs to consume a large amount of computing power during training, the branch prediction structure of the trained random forest model is simple, the computing power consumed in actual operation is small, and the model is more suitable for operation on a TCU.
Specifically, referring to fig. 1, fig. 1 is a functional module schematic diagram of an on-board device to which a data-driven prediction and analysis apparatus for vehicle driveline NVH diagnosis according to the present application belongs. The data-driven type prediction and analysis device for vehicle transmission system NVH diagnosis can be a device which is independent of an on-board device, can perform signal processing and network model training, and can be borne on the on-board device in a hardware or software mode.
In the present embodiment, the on-board device to which the data-driven type prediction and analysis apparatus for vehicle driveline NVH diagnosis belongs includes at least an output module 110, a processor 120, a memory 130, and a communication module 140.
The memory 130 stores therein an operating system and a data-driven prediction and analysis program for vehicle driveline NVH diagnosis, and the data-driven prediction and analysis device for vehicle driveline NVH diagnosis may store in the memory 130 information such as an acquired vehicle state signal, an NVH state analysis result of a vehicle obtained by analysis by an NVH prediction and diagnosis model, and an acquired training data set composed of pre-acquired real vehicle experimental data acquired by a sensor, a kini index of the data set obtained from the training data set, a classification feature selected from the training data set, a kini index of the data set under the classification feature condition obtained from the classification feature and the kini index of the data set, a classification feature with a minimum acquired kini index, a contribution value of an input feature calculated by an interpretable machine learning snap method, an input-output correlation analysis result obtained by analyzing the contribution of the input feature to the NVH prediction and diagnosis model, a screened training data set generated, and a test data set composed of real vehicle experimental data different from the training data set; the output module 110 may be a display screen or the like. The communication module 140 may include a WIFI module, a mobile communication module, a bluetooth module, and the like, and communicates with an external device or a server through the communication module 140.
Wherein the data-driven prognostics and analysis program in memory 130 for vehicle driveline NVH diagnostics when executed by the processor implements the steps of:
acquiring a collected vehicle state signal;
and inputting the vehicle state signal into a pre-trained NVH prediction and diagnosis model for analysis to obtain an NVH state analysis result of the vehicle, wherein the NVH prediction and diagnosis model is constructed by adopting a random forest algorithm.
Further, the data-driven prognostics and analysis program in memory 130 for vehicle driveline NVH diagnostics when executed by the processor further performs the steps of:
and training based on a random forest algorithm to obtain the NVH prediction diagnosis model, wherein the random forest algorithm adopts a CART algorithm.
Further, the data-driven prognostics and analysis program in memory 130 for vehicle driveline NVH diagnostics when executed by the processor further performs the steps of:
acquiring a training data set acquired in advance, wherein the training data set consists of real vehicle experimental data acquired by a sensor;
obtaining a kini index of the data set according to the training data set;
selecting a classification feature according to the training data set;
obtaining the Gini index of the data set under the condition of the classification features according to the classification features and the Gini index of the data set; and returning to the execution step: selecting a classification feature according to the training data set;
and calculating the Gini index of the data set under each classification characteristic condition by the loop to obtain the classification characteristic with the minimum Gini index, and generating the NVH prediction diagnosis model according to the classification characteristic with the minimum Gini index.
Further, the data-driven prognostics and analysis program in memory 130 for vehicle driveline NVH diagnostics when executed by the processor further performs the steps of:
calculating a contribution value of the input features of the NVH prediction diagnosis model by an interpretable machine learning SHAP method;
analyzing the contribution value to obtain a contribution analysis result of the input feature to the NVH prediction diagnosis model;
performing feature importance analysis on the input features according to the contribution analysis result to obtain an input and output correlation analysis result of the NVH prediction diagnosis model, wherein the input and output correlation analysis result comprises unimportant non-correlation features;
screening the training data set according to the unimportant non-relevant features to generate a screened training data set; and returning to the execution step: obtaining a kini index of the data set according to the training data set;
and terminating the training until the contribution value of the input feature meets a preset contribution value, so as to obtain the trained NVH prediction diagnosis model.
Further, the data-driven prognostics and analysis program in memory 130 for vehicle driveline NVH diagnostics when executed by the processor further performs the steps of:
preprocessing the training data set.
Further, the data-driven prognostics and analysis program in memory 130 for vehicle driveline NVH diagnostics when executed by the processor further performs the steps of:
acquiring the engine speed acquired by the engine speed sensor, and acquiring the transmission input shaft speed acquired by the transmission input shaft speed sensor;
and carrying out difference processing according to the rotating speed of the engine and the rotating speed of the input shaft of the transmission to obtain the rotating speed difference of the same dimension.
Further, the data-driven prognostics and analysis program in memory 130 for vehicle driveline NVH diagnostics when executed by the processor further performs the steps of:
and verifying the prediction accuracy of the NVH prediction and diagnosis model through a pre-collected test data set, wherein the test data set is composed of real vehicle experiment data different from the training data set.
According to the scheme, the embodiment specifically comprises the steps of acquiring the acquired vehicle state signal; and inputting the vehicle state signal into a pre-trained NVH prediction and diagnosis model for analysis to obtain an NVH state analysis result of the vehicle, wherein the NVH prediction and diagnosis model is constructed by adopting a random forest algorithm. The NVH state of the vehicle is analyzed through the data-driven vehicle NVH prediction and diagnosis model based on the random forest, the NVH analysis difficulty can be reduced, and the problem that a large amount of computing power is consumed in the real-time operation process due to the fact that a traditional mechanism-based modeling method and an analysis process are too complex is solved. Based on this application scheme, will gather and relate to the real car experimental data of a plurality of operating modes and make into the data set, train based on this data set and obtain NVH prediction diagnosis model, can be used for accurately judging the vehicle in the operation process, the NVH state that acts on the whole car that vehicle transmission system produced to can the analysis obtain other vehicle states to the influence that NVH state produced, and then provide the direction for the improvement of NVH control strategy. According to the scheme, the off-line model can be obtained by adopting the random forest algorithm, the model only needs to consume a large amount of computing power during training, the branch prediction structure of the trained random forest model is simple, the computing power consumed in actual operation is small, and the model is more suitable for operation on a TCU.
Based on the above vehicle-mounted device architecture but not limited to the above architecture, embodiments of the method of the present application are proposed.
Referring to FIG. 2, FIG. 2 is a schematic flow diagram of a first exemplary embodiment of the data-driven prediction and analysis method for vehicle driveline NVH diagnostics of the present application. The data-driven prediction and analysis method for vehicle driveline NVH diagnostics includes:
step S1001, acquiring a collected vehicle state signal;
and S1002, inputting the vehicle state signal into a pre-trained NVH prediction and diagnosis model for analysis to obtain an NVH state analysis result of the vehicle, wherein the NVH prediction and diagnosis model is constructed by adopting a random forest algorithm.
The execution subject of the method of the present embodiment may be a data-driven type predicting and analyzing apparatus for vehicle driveline NVH diagnosis, or may be a data-driven type predicting and analyzing on-board device or server for vehicle driveline NVH diagnosis, and the present embodiment is exemplified by a data-driven type predicting and analyzing apparatus for vehicle driveline NVH diagnosis, which may be integrated on an on-board device having a data processing function.
The scheme of the embodiment mainly realizes the analysis of the NVH state of the vehicle, in particular to the NVH state generated by a vehicle transmission system.
In the present exemplary embodiment, a detected vehicle state signal is initially detected, wherein the vehicle state signal can be detected by a sensor and a microphone arranged in advance for a running vehicle. The vehicle state signal refers to a signal which is generated by each component and each module of the vehicle in the current running environment and reflects the current state of the vehicle, such as a gear signal, an accelerator opening, clutch engagement and the like.
And then inputting the acquired vehicle state signal into a pre-trained NVH prediction diagnosis model for analysis to obtain an NVH state analysis result of the vehicle. The NVH prediction diagnosis model is constructed by adopting a random forest algorithm, and the constructed model is trained by using a large data set formed by sensor data acquired by a real vehicle experiment based on a plurality of working conditions.
According to the scheme, the embodiment specifically comprises the steps of acquiring the acquired vehicle state signal; and inputting the vehicle state signal into a pre-trained NVH prediction and diagnosis model for analysis to obtain an NVH state analysis result of the vehicle, wherein the NVH prediction and diagnosis model is constructed by adopting a random forest algorithm. The NVH state of the vehicle is analyzed through the data-driven vehicle NVH prediction and diagnosis model based on the random forest, the NVH analysis difficulty can be reduced, and the problem that a large amount of computing power is consumed in the real-time operation process due to the fact that a traditional mechanism-based modeling method and an analysis process are too complex is solved. Based on this application scheme, will gather and relate to the real car experimental data of a plurality of operating modes and make into the data set, train based on this data set and obtain NVH prediction diagnosis model, can be used for accurately judging the vehicle in the operation process, the NVH state that acts on the whole car that vehicle transmission system produced to can the analysis obtain other vehicle states to the influence that NVH state produced, and then provide the direction for the improvement of NVH control strategy. According to the scheme, the off-line model can be obtained by adopting a random forest algorithm, the model only needs to consume a large amount of computing power during training, the branch prediction structure of the trained random forest model is simple, the computing power consumed in actual operation is small, and the model is more suitable for operation on a TCU.
Referring to FIG. 3, FIG. 3 is a schematic flow diagram of a second exemplary embodiment of the data-driven prediction and analysis method for vehicle driveline NVH diagnostics of the present application. Based on the embodiment shown in fig. 2, in this embodiment, before the step of acquiring the collected vehicle state signal, the data driving type prediction and analysis method for vehicle transmission system NVH diagnosis further includes:
and S1000, training based on a random forest algorithm to obtain the NVH prediction diagnosis model, wherein the random forest algorithm adopts a CART algorithm. In this embodiment, step S1000 is implemented before step S1001, and in other embodiments, step S1000 may be implemented between step S1001 and step S1002.
In the embodiment, the NVH predictive diagnostic model is trained and obtained based on a random forest algorithm. The basis of the random forest algorithm is a decision tree, and three algorithms for training the decision tree are available. The CART algorithm is selected in this example to train the NVH predictive diagnostic model.
In contrast to the embodiment shown in fig. 2, the embodiment further includes a scheme for training the NVH predictive diagnostic model.
Specifically, referring to fig. 4, fig. 4 is a schematic training flow diagram of an NVH predictive diagnosis model according to an embodiment of the data-driven prediction and analysis method for vehicle driveline NVH diagnosis of the present application, where the step of obtaining the NVH predictive diagnosis model based on random forest algorithm training may include:
step S1100, a training data set collected in advance is obtained, wherein the training data set is composed of real vehicle experiment data collected by a sensor.
Specifically, in the embodiment, a plurality of real vehicle experimental data are collected in advance through the sensor to form a training data set. The real vehicle experimental data is acquired by simulating the working condition that the vehicle frequently starts and shifts on a congested road section in a city. This training data set, which is acquired in advance, is acquired when training the NVH predictive diagnostic model.
And S1200, obtaining the kini index of the data set according to the training data set.
Specifically, the kini index of the data set is calculated according to the acquired training data set. It should be noted that the CART algorithm uses the kini index to perform the handoverAnd (4) selecting. In the classification problem, assuming that there are K classes, the probability that a sample point belongs to the kth class is P k Then, the kini index of the probability distribution is defined as shown in the following equation 1:
thus, for training data set D, the Kernel index of this data set is shown in equation 2 below:
where Ck is the sample subset belonging to the kth class in the training data set D, and K is the number of classes.
And according to the obtained training data set, calculating the Gini index of the data set by the formula 2.
Step S1300, selecting classification characteristics according to the training data set;
step S1400, obtaining the Gini index of the data set under the condition of the classification characteristic according to the classification characteristic and the Gini index of the data set; and returning to the execution step: selecting a classification feature according to the training data set;
specifically, a classification feature is selected from the training data set, and the classification feature can enable the training data set D to be divided into two parts, D1 and D2, according to whether the classification feature takes a certain possible value a, in other words, data that satisfies a classification feature value and data that does not satisfy the classification feature value in the training data set D are divided into two parts, D1 and D2. And then, according to the selected classification characteristic and the calculated Gini index of the data set, obtaining the Gini index of the data set under the condition of the classification characteristic.
For example, if the training data set D is divided into two parts D1 and D2 according to whether the classification feature a takes a certain possible value a, the data sets D1 and D2 can be expressed as shown in the following formula 3:
D 1 ＝{(x，y)∈D|A(x)＝a}，D 2 ＝D-D 1 (3)
then under the condition of the classification feature a, the kini index of the training data set D is as shown in the following formula 4:
wherein the kini index Gini (D) represents the uncertainty of the training data set D, and the kini index Gini (D, a) represents the uncertainty of the training data set D segmented by the classification feature a = a.
It should be noted that the larger the kini index value is, the larger the uncertainty representing the sample set is, and this point is similar to the entropy. The smaller the value of the kini index, the better the classification of the features, which is contrary to the information gain.
After the kini index of the data set under the condition of the classification characteristic A is obtained through calculation, the step S1300 is executed again, another classification characteristic B is selected from the training data set, and then the kini index of the data set under the condition of the classification characteristic B is obtained according to the selected classification characteristic B and the calculated kini index of the data set.
Step S1500, calculating the Gini index of the data set under each classification characteristic condition according to the circulation, then obtaining the classification characteristic with the minimum Gini index, and generating the NVH prediction diagnosis model according to the classification characteristic with the minimum Gini index.
And then analyzing the NVH state of the vehicle through the generated NVH prediction diagnosis model.
According to the scheme, the NVH prediction diagnosis model is obtained through training based on the random forest algorithm, wherein the random forest algorithm adopts a CART algorithm; acquiring a collected vehicle state signal; and inputting the vehicle state signal into a pre-trained NVH prediction diagnosis model for analysis to obtain an NVH state analysis result of the vehicle. The NVH state of the vehicle is analyzed through the data-driven vehicle NVH prediction and diagnosis model based on the random forest, the NVH analysis difficulty can be reduced, and the problem that a large amount of computing power is consumed in the real-time operation process due to the fact that a traditional mechanism-based modeling method and an analysis process are too complex is solved. Based on the scheme, the acquired real vehicle experimental data relating to multiple working conditions are made into a data set, the NVH prediction and diagnosis model is obtained based on the data set training, the NVH state of the whole vehicle, which is generated by a vehicle transmission system, can be accurately judged in the running process of the vehicle, the influence of other vehicle states on the NVH state can be analyzed, and a direction is provided for improvement of an NVH control strategy. According to the scheme, the off-line model can be obtained by adopting the random forest algorithm, the model only needs to consume a large amount of computing power during training, the branch prediction structure of the trained random forest model is simple, the computing power consumed in actual operation is small, and the model is more suitable for operation on a TCU.
Further, referring to fig. 5, fig. 5 is a schematic flow chart of a third exemplary embodiment of the data-driven prediction and analysis method for vehicle driveline NVH diagnosis according to the present application, and based on the embodiments shown in fig. 3 and fig. 4, in this embodiment, after the step of generating the NVH prediction diagnosis model according to the classification feature with the minimum kini index, the data-driven prediction and analysis method for vehicle driveline NVH diagnosis further includes:
and step S1600, calculating and obtaining a contribution value of the input characteristic of the NVH prediction diagnosis model through an interpretable machine learning SHAP method.
Specifically, the contribution value SHAP value of the input feature of the generated NVH predictive diagnostic model is calculated and obtained by an interpretable machine learning method, i.e., a SHAP method. Note that in the game theory, the shield value is used to fairly allocate credits or contribute to each participant. With this heuristic, SHAP value is used to combine optimal credit allocation with local interpretation to propose a method to evaluate the contribution of each feature to the model. The SHAP method can evaluate the importance of the input features of the model to the output of the model, and is a characteristic importance analysis method afterwards.
For example, the method for calculating the SHAP value of the contribution value of the input feature may refer to the following equation 5:
wherein the content of the first and second substances,
is a collection of all feature orderings,
is the set of all features before feature i in rank R, and M is the number of input features to the model.
And S1700, analyzing the contribution value to obtain a contribution analysis result of the input feature to the NVH prediction diagnosis model.
Specifically, the calculated contribution value of the input feature is analyzed, and a contribution analysis result of the input feature to the NVH prediction diagnosis model is obtained.
Step S1800, performing feature importance analysis on the input features according to the contribution analysis result to obtain an input and output correlation analysis result of the NVH predictive diagnostic model, wherein the input and output correlation analysis result comprises unimportant non-correlation features.
Specifically, the input feature is subjected to feature importance analysis according to the obtained contribution analysis result, so as to obtain an input and output correlation analysis result of the NVH predictive diagnostic model, where the input and output correlation analysis result represents a correlation of the input feature to the model output, for example, a certain input feature is positively correlated with the model output in a certain region. Optionally, the input and output correlation analysis result may further include an insignificant non-correlation feature.
Step S1900, screening the training data set according to the unimportant non-relevant features to generate a screened training data set; and returning to the execution step: and obtaining the kini index of the data set according to the training data set.
Specifically, the training data set is screened according to the unimportant uncorrelated features obtained by the SHAP analysis, the unimportant uncorrelated features in the training data set are removed, and then the screened training data set is generated. And then returning to the step S1200, and obtaining the Gini index of the data set according to the screened training data set, so as to repeat the steps of training the NVH prediction diagnosis model.
And S2000, terminating the training until the contribution value of the input feature meets a preset condition, and obtaining the trained NVH prediction diagnosis model.
Specifically, when the contribution value of the input feature calculated by the SHAP method meets a preset condition, namely the input feature has high contribution to the output of the model, the training of the model is terminated, and the trained NVH prediction diagnosis model is obtained.
The present embodiment analyzes the importance of the model input features by combining with the SHAP method, and has better interpretability compared with other methods, such as a method for performing feature selection and dimension reduction based on correlation analysis between input features. According to the SHAP method provided by the embodiment of the application, not only can the contribution of each input characteristic to the model output be analyzed, but also the correlation between each input characteristic and the model output can be obtained through analysis, so that the basis is provided for the screening and dimension reduction of the model input characteristics, and the method is more reliable.
Further, the present embodiment also includes a scheme implemented after the step of acquiring the pre-acquired training data set. In this embodiment, after the step S1100, acquiring the training data set collected in advance, the method may further include:
step S1110, preprocessing the training data set.
Specifically, the acquired training data set is preprocessed to obtain a characteristic labeled training data set. Optionally, the acquired training data set is preprocessed according to different sensors and different acquisition devices.
The acquisition device of the real vehicle experimental data is Rotec, and the voltage signal output by the sensor cannot be directly read by the acquisition device, so the data is converted by a Pulse Former and then input into the acquisition device Rotec.
In this embodiment, for NVH phenomena of a vehicle transmission system, including a Judder phenomenon and a Shuffle phenomenon, a training data set based on real vehicle experimental data and subjected to data preprocessing is used for model training.
As an embodiment, the sensors may include an engine speed sensor and a transmission input shaft speed sensor, the vehicle experimental data may include an engine speed and a transmission input shaft speed, and the preprocessing the training data set at step S1110 may include:
acquiring the engine speed acquired by the engine speed sensor and acquiring the transmission input shaft speed acquired by the transmission input shaft speed sensor;
and carrying out difference processing according to the rotating speed of the engine and the rotating speed of the input shaft of the transmission to obtain the rotating speed difference of the same dimension.
Specifically, the engine speed acquired by the engine speed sensor is acquired, and the transmission input shaft speed acquired by the transmission input shaft speed sensor is acquired. Because the engine rotating speed sensor and the transmission input shaft rotating speed sensor use different adopted frequencies for signal acquisition, in order to obtain the rotating speed difference between the engine rotating speed sensor and the transmission input shaft rotating speed sensor as a characteristic, difference processing is carried out according to the rotating speed of the transmitter and the rotating speed of the transmission input shaft to obtain the rotating speed difference with the same dimension.
According to the embodiment, the effectiveness of NVH analysis can be improved through data preprocessing operation, and the performance and efficiency of the NVH prediction diagnosis model are improved.
Further, the embodiment further includes a scheme implemented after the step of obtaining the trained NVH predictive diagnostic model. In this embodiment, after the step of obtaining the trained NVH prediction diagnosis model, the method may further include:
and verifying the prediction accuracy of the NVH prediction diagnosis model through a pre-collected test data set, wherein the test data set is composed of real vehicle experiment data different from the training data set.
Specifically, a plurality of real vehicle experiment data are collected through a sensor in advance to form a test data set, wherein the data contained in the test data set is different from the data contained in the training data set. And verifying the prediction accuracy of the trained NVH prediction diagnosis model through the pre-collected test set. Optionally, if the prediction accuracy obtained by verification is too low, the NVH prediction diagnosis model is retrained until the prediction accuracy meets the requirement.
According to the scheme, aiming at the NVH phenomena of the vehicle transmission system, including the Juder phenomenon and the Shuffle phenomenon, model training is carried out by adopting the training data set which is based on real vehicle experimental data and is subjected to data preprocessing, the NVH prediction diagnosis model of the data-driven vehicle based on the random forest is established, the NVH prediction diagnosis model obtained by training is verified through the established testing data set, the obtained prediction accuracy is up to 99.6%, more accurate NVH state prediction is realized, and the scheme of the embodiment of the application is further suitable for being applied to the technical field and is more suitable for running on a TCU compared with the prior art.
According to the scheme, the NVH prediction diagnosis model is obtained through training based on the random forest algorithm, wherein the random forest algorithm adopts a CART algorithm; acquiring a collected vehicle state signal; and inputting the vehicle state signal into a pre-trained NVH prediction diagnosis model for analysis to obtain an NVH state analysis result of the vehicle. The NVH state of the vehicle is analyzed through the data-driven vehicle NVH prediction and diagnosis model based on the random forest, the NVH analysis difficulty can be reduced, and the problem that a large amount of computing power is consumed in the real-time operation process due to the fact that a traditional mechanism-based modeling method and an analysis process are too complex is solved. Based on this application scheme, will gather and relate to the real car experimental data of a plurality of operating modes and make into the data set, train based on this data set and obtain NVH prediction diagnosis model, can be used for accurately judging the vehicle in the operation process, the NVH state that acts on the whole car that vehicle transmission system produced to can the analysis obtain other vehicle states to the influence that NVH state produced, and then provide the direction for the improvement of NVH control strategy. According to the scheme, the off-line model can be obtained by adopting the random forest algorithm, the model only needs to consume a large amount of computing power during training, the branch prediction structure of the trained random forest model is simple, the computing power consumed in actual operation is small, and the model is more suitable for operation on a TCU.
In addition, an embodiment of the present application further provides a data-driven type prediction and analysis apparatus for vehicle driveline NVH diagnosis, including:
the signal acquisition module is used for acquiring the acquired vehicle state signal;
and the state prediction module is used for inputting the vehicle state signal into a pre-trained NVH prediction diagnosis model for analysis to obtain an NVH state analysis result of the vehicle, wherein the NVH prediction diagnosis model is constructed by adopting a random forest algorithm.
Further, the data-driven type predicting and analyzing apparatus for vehicle driveline NVH diagnosis further includes:
and the model training module is used for training based on a random forest algorithm to obtain the NVH prediction diagnosis model, wherein the random forest algorithm adopts a CART algorithm.
For the principle and implementation process of implementing NVH state analysis in this embodiment, please refer to the above embodiments, which are not described herein again.
Furthermore, an on-board device is provided in an embodiment of the present application, the on-board device including a memory, a processor, and a data-driven type prediction and analysis program for vehicle driveline NVH diagnosis stored on the memory and executable on the processor, the data-driven type prediction and analysis program for vehicle driveline NVH diagnosis implementing the steps of the data-driven type prediction and analysis method for vehicle driveline NVH diagnosis as described above when executed by the processor.
Since the data-driven prediction and analysis program for vehicle driveline NVH diagnosis is executed by the processor, all technical solutions of all the foregoing embodiments are adopted, so that at least all the beneficial effects brought by all the technical solutions of all the foregoing embodiments are achieved, and detailed description is omitted here.
Furthermore, an embodiment of the present application also provides a computer readable storage medium, which stores thereon a data-driven type prediction and analysis program for vehicle driveline NVH diagnosis, which when executed by a processor implements the steps of the data-driven type prediction and analysis method for vehicle driveline NVH diagnosis as described above.
Since the data-driven prediction and analysis program for vehicle driveline NVH diagnosis is executed by the processor, all technical solutions of all the foregoing embodiments are adopted, so that at least all the beneficial effects brought by all the technical solutions of all the foregoing embodiments are achieved, and detailed description is omitted here.
Compared with the prior art, the data-driven type prediction and analysis method and device for vehicle transmission system NVH diagnosis, the vehicle-mounted equipment and the storage medium provided by the embodiment of the application are based on the acquired vehicle state signals; and inputting the vehicle state signal into a pre-trained NVH prediction and diagnosis model for analysis to obtain an NVH state analysis result of the vehicle, wherein the NVH prediction and diagnosis model is constructed by adopting a random forest algorithm. The NVH state of the vehicle is analyzed through the data-driven vehicle NVH prediction and diagnosis model based on the random forest, the NVH analysis difficulty can be reduced, and the problem that a large amount of computing power is consumed in the real-time operation process due to the fact that a traditional mechanism-based modeling method and an analysis process are too complex is solved. Based on this application scheme, will gather and relate to the real car experimental data of a plurality of operating modes and make into the data set, train based on this data set and obtain NVH prediction diagnosis model, can be used for accurately judging the vehicle in the operation process, the NVH state that acts on the whole car that vehicle transmission system produced to can the analysis obtain other vehicle states to the influence that NVH state produced, and then provide the direction for the improvement of NVH control strategy. According to the scheme, the off-line model can be obtained by adopting the random forest algorithm, the model only needs to consume a large amount of computing power during training, the branch prediction structure of the trained random forest model is simple, the computing power consumed in actual operation is small, and the model is more suitable for operation on a TCU.
It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or system. Without further limitation, an element defined by the phrases "comprising a," "8230," "8230," or "comprising" does not exclude the presence of other like elements in a process, method, article, or system comprising the element.
The above-mentioned serial numbers of the embodiments of the present application are merely for description and do not represent the merits of the embodiments.
Through the above description of the embodiments, those skilled in the art will clearly understand that the method of the above embodiments can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware, but in many cases, the former is a better implementation manner. Based on such understanding, the technical solutions of the present application may be embodied in the form of a software product, which is stored in a storage medium (e.g., ROM/RAM, magnetic disk, optical disk) and includes instructions for enabling a device (e.g., a mobile phone, a computer, a server, a controlled terminal, or a network device) to execute the method of each embodiment of the present application.
The above description is only a preferred embodiment of the present application, and not intended to limit the scope of the present application, and all modifications of equivalent structures and equivalent processes, which are made by the contents of the specification and the drawings of the present application, or which are directly or indirectly applied to other related technical fields, are included in the scope of the present application.
Claims (10)
1. A data-driven prediction and analysis method for vehicle driveline NVH diagnostics, the data-driven prediction and analysis method for vehicle driveline NVH diagnostics comprising:
acquiring a collected vehicle state signal;
and inputting the vehicle state signal into a pre-trained NVH prediction and diagnosis model for analysis to obtain an NVH state analysis result of the vehicle, wherein the NVH prediction and diagnosis model is constructed by adopting a random forest algorithm.
2. The data-driven prognostics and analysis method for vehicle driveline NVH diagnostics of claim 1, wherein the step of acquiring the collected vehicle state signals is preceded by the step of:
and training based on a random forest algorithm to obtain the NVH prediction diagnosis model, wherein the random forest algorithm adopts a CART algorithm.
3. The data-driven predictive and analytical method for vehicle driveline NVH diagnostics of claim 2 wherein the step of deriving the NVH predictive diagnostic model based on random forest algorithm training comprises:
acquiring a training data set acquired in advance, wherein the training data set consists of real vehicle experiment data acquired by a sensor;
obtaining a kini index of the data set according to the training data set;
selecting a classification feature according to the training data set;
obtaining the Gini index of the data set under the condition of the classification features according to the classification features and the Gini index of the data set; and returning to the execution step: selecting a classification feature according to the training data set;
and calculating the Gini index of the data set under each classification characteristic condition by the loop to obtain the classification characteristic with the minimum Gini index, and generating the NVH prediction diagnosis model according to the classification characteristic with the minimum Gini index.
4. The data-driven prognostics and analysis method for vehicle driveline NVH diagnostics of claim 3, wherein the step of generating the NVH prognostic diagnostic model based on the classification characteristic with the minimum kini index is further followed by:
calculating a contribution value of the input features of the NVH prediction diagnosis model by an interpretable machine learning SHAP method;
analyzing the contribution value to obtain a contribution analysis result of the input feature to the NVH prediction diagnosis model;
performing feature importance analysis on the input features according to the contribution analysis result to obtain an input and output correlation analysis result of the NVH prediction diagnosis model, wherein the input and output correlation analysis result comprises unimportant non-correlation features;
screening the training data set according to the unimportant non-relevant features to generate a screened training data set; and returning to the execution step: obtaining a kini index of the data set according to the training data set;
and terminating the training until the contribution value of the input feature meets a preset contribution value, so as to obtain the trained NVH prediction diagnosis model.
5. The data-driven prognostics and analysis method for vehicle driveline NVH diagnostics of claim 4, wherein the step of obtaining a pre-collected training data set is followed by further comprising:
preprocessing the training data set.
6. The data-driven prognostics and analysis method for vehicle driveline NVH diagnostics of claim 5, wherein the sensors include an engine speed sensor and a transmission input shaft speed sensor, the real vehicle experimental data includes engine speed and transmission input shaft speed, and the step of preprocessing the training data set includes:
acquiring the engine speed acquired by the engine speed sensor, and acquiring the transmission input shaft speed acquired by the transmission input shaft speed sensor;
and carrying out difference processing according to the rotating speed of the engine and the rotating speed of the input shaft of the transmission to obtain the rotating speed difference of the same dimension.
7. The data-driven prognostic and analysis method for vehicle driveline NVH diagnostics according to claim 6 wherein the step of deriving the trained NVH prognostic diagnostic model is followed by further comprising:
and verifying the prediction accuracy of the NVH prediction and diagnosis model through a pre-collected test data set, wherein the test data set is composed of real vehicle experiment data different from the training data set.
8. A data-driven type prediction and analysis apparatus for vehicle driveline NVH diagnosis, characterized by comprising:
the signal acquisition module is used for acquiring the acquired vehicle state signal;
and the state prediction module is used for inputting the vehicle state signal into a pre-trained NVH prediction diagnosis model for analysis to obtain an NVH state analysis result of the vehicle, wherein the NVH prediction diagnosis model is constructed by adopting a random forest algorithm.
9. An on-board device comprising a memory, a processor, and a data-driven prediction and analysis program for vehicle driveline NVH diagnostics stored on the memory and executable on the processor, the data-driven prediction and analysis program for vehicle driveline NVH diagnostics when executed by the processor implementing the steps of the data-driven prediction and analysis method for vehicle driveline NVH diagnostics of any one of claims 1-7.
10. A computer readable storage medium having stored thereon a data-driven predictive and analytical program for vehicle driveline NVH diagnostics which when executed by a processor implements the steps of the data-driven predictive and analytical method for vehicle driveline NVH diagnostics of any one of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211561562.4A CN115828751A (en) | 2022-12-06 | 2022-12-06 | Data-driven prediction and analysis method for NVH diagnosis of vehicle transmission system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211561562.4A CN115828751A (en) | 2022-12-06 | 2022-12-06 | Data-driven prediction and analysis method for NVH diagnosis of vehicle transmission system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115828751A true CN115828751A (en) | 2023-03-21 |
Family
ID=85544382
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211561562.4A Pending CN115828751A (en) | 2022-12-06 | 2022-12-06 | Data-driven prediction and analysis method for NVH diagnosis of vehicle transmission system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115828751A (en) |
-
2022
- 2022-12-06 CN CN202211561562.4A patent/CN115828751A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Yu et al. | A bearing fault diagnosis technique based on singular values of EEMD spatial condition matrix and Gath-Geva clustering | |
| US10223842B1 (en) | System for controlling remotely connected vehicle | |
| CN107492251A (en) | It is a kind of to be identified and driving condition supervision method based on the driver identity of machine learning and deep learning | |
| CN107364446B (en) | Method for operating a motor vehicle | |
| CN109547254B (en) | Intrusion detection method and device, electronic equipment and storage medium | |
| CN107953888B (en) | Road type identification method and system | |
| CN111994084B (en) | Method and system for identifying driving style of driver and storage medium | |
| US11741761B2 (en) | State diagnosis apparatus and method of moving system part | |
| US7640148B2 (en) | Method of modeling vehicle parameter cycles | |
| CN112026744B (en) | Series-parallel hybrid power system energy management method based on DQN variants | |
| Massoud et al. | Exploring fuzzy logic and random forest for car drivers’ fuel consumption estimation in IoT-enabled serious games | |
| US11458784B2 (en) | Methods and apparatus to determine tire tread depth | |
| US11556792B2 (en) | Method and apparatus for predicting performance of vehicle NVH system based on deep learning | |
| CN115828751A (en) | Data-driven prediction and analysis method for NVH diagnosis of vehicle transmission system | |
| CN116776208A (en) | Training method of seismic wave classification model, seismic wave selecting method, equipment and medium | |
| CN112937475A (en) | Information sharing platform and method capable of providing bidirectional vehicle state information and system having the same | |
| Reddy | Driving behaviour classification: An eco-driving approach | |
| Zhang | Vehicle Parameters Estimation and Driver Behavior Classification for Adaptive Shift Strategy of Heavy Duty Vehicles | |
| WO2019034233A1 (en) | System and method for vehicle modelling and simulation | |
| CN112508080A (en) | Vehicle model identification method, device, equipment and medium based on experience playback | |
| CN113968234B (en) | Vehicle auxiliary driving control method and device and vehicle-mounted controller | |
| Leite | Convolutional neural networks application for the classification of powertrain booming noise on internal combustion engine passenger cars | |
| CN113610166B (en) | Method for establishing test scene library for intelligent vehicle | |
| CN114861516B (en) | Vehicle dynamic performance determining method, device and storage medium | |
| US20220198282A1 (en) | Decision tree of models: using decision tree model, and replacing the leaves of the tree with other machine learning models |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |