Engine knock energy prediction method and device, vehicle, storage medium and product
By constructing a knock energy prediction model based on the Transformer layer and optimizing the parameters using an optimizer, the problem of insufficient knock energy prediction accuracy caused by power coupling in hybrid engines is solved, and high-precision knock energy prediction is achieved.
Patent Information
- Application Number
- CN202511555716.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-10-28
- Publication Date
- 2026-02-17
AI Technical Summary
Hybrid engines operate under different conditions due to power coupling, resulting in poor accuracy in predicting knock energy.
By acquiring operational data and corresponding actual knock energy from multiple power-operating stages of the engine, a training dataset and a validation dataset are constructed. A knock energy prediction model is then developed using a stacked encoder-decoder structure of Transformer layers and a fully connected network layer. Combined with an optimizer, the model parameters are optimized to achieve high-precision prediction of knock energy.
It improves the accuracy of knock energy prediction for hybrid engines under complex operating conditions, enhances the model's anti-interference ability, adapts to the special operating conditions of hybrid engines, and achieves low-cost, high-precision knock prediction.
Smart Images
Figure CN121542601A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of engine knock detection and control, and particularly relates to an engine knock energy prediction method and device, a vehicle, a storage medium and a product. BACKGROUND
[0002] The hybrid engine is affected by the power coupling characteristics, and the operating conditions are significantly different from those of the traditional fuel engine. In particular, during the process of frequent start-stop and rapid motor dragging of the engine into the high speed load area, the pressure gradient and the turbulence intensity of the combustion chamber increase significantly, and the probability of knock occurrence increases. SUMMARY
[0003] The present application provides an engine knock energy prediction method, device, vehicle, storage medium and product to solve the problem of poor knock energy prediction accuracy of the hybrid engine due to differential operating conditions caused by power coupling in related technologies.
[0004] The first aspect of the present application provides an engine knock energy prediction method, comprising the following steps: obtaining an operating data set of an engine test, wherein the operating data set comprises operating data of the engine in multiple work stages and knock true energy corresponding to the operating data of each work stage; generating a training data set and a verification set according to the operating data set, wherein the verification set and the training data set both comprise data samples and true labels of the data samples, the data sample comprises operating data of the engine in the current work stage, and the true label is the knock true energy of the data sample; training a pre-constructed knock energy prediction model using the training data set, verifying the knock energy prediction model using the verification set, and predicting the knock prediction energy of the engine using the currently collected operating data and the knock energy prediction model.
[0005] Optionally, the knock energy prediction model comprises a multi-Transformer layer stacked encoder-decoder, a pooling layer and a fully connected network layer, wherein the encoder-decoder extracts knock-related features from the operating data through a local attention mechanism and a global attention mechanism, and outputs the knock prediction energy after the knock-related features pass through the pooling layer and the fully connected network layer.
[0006] Optionally, the knock energy prediction model further comprises an optimizer, which optimizes the model parameters of the knock energy prediction model in a reverse direction according to the deviation between the knock prediction energy and the knock true energy corresponding to the true label.
[0007] Optionally, obtaining the operating data set of the engine test comprises: dividing the speed of the hybrid engine into multiple speed intervals; dividing the load of the hybrid engine into multiple load intervals; generating multiple speed-load intervals according to the multiple speed intervals and the multiple load intervals; and collecting operating data and corresponding knock true energy of the current operating stage in each speed-load interval.
[0008] Optionally, the operation data comprises at least one of ignition timing, ignition accumulation time, fuel injection timing, fuel injection pulse width, engine speed, load, coolant temperature, intake temperature, air-fuel ratio, and differential voltage signal.
[0009] Optionally, the knock true energy is an explosion energy obtained by processing a cylinder pressure sensor.
[0010] The second aspect embodiment of the present application provides an engine knock energy prediction device, comprising: an acquisition module configured to acquire an engine test operation data set, wherein the operation data set comprises operation data of the engine in a plurality of working stages and a knock true energy corresponding to the operation data of each working stage; a generation module configured to generate a training data set and a verification set according to the operation data set, wherein the verification set and the training data set both comprise a data sample and a true label of the data sample, the data sample comprises operation data of the engine in a current working stage, and the true label is a knock true energy of the data sample; a training module configured to train a pre-constructed knock energy prediction model by using the training data set, to verify the knock energy prediction model by using the verification set, and to predict a knock prediction energy of the engine by using the currently collected operation data and the knock energy prediction model.
[0011] The third aspect embodiment of the present application provides a vehicle, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the engine knock energy prediction method of the above-mentioned embodiments.
[0012] The fourth aspect embodiment of the present application provides a computer program product, wherein the computer program is executed to implement the engine knock energy prediction method of the above-mentioned embodiments.
[0013] Therefore, the present application has at least the following beneficial effects: The embodiments of the present application can acquire an engine test operation data set comprising a plurality of working stage operation data and corresponding knock true energy, generate a training data set and a verification set according to the data set, the data sample is the operation data of the current working stage, the true label is the knock true energy thereof, train a pre-constructed knock energy prediction model by using the training data set, verify the model by using the verification set, and predict a knock prediction energy of the engine based on the currently collected operation data and the model, so as to fuse multi-source heterogeneous data to enhance the anti-interference ability of the model, effectively adapt to the complex working conditions of the hybrid engine, and realize high-precision prediction of the knock energy. Therefore, the problems such as poor knock energy prediction accuracy of the hybrid engine due to the differential operation conditions caused by power coupling in the related art are solved.
[0014] Additional aspects and advantages of the present application will be made apparent by the following description and the accompanying drawings. BRIEF DESCRIPTION OF DRAWINGS
[0015] The above and / or additional aspects and advantages of the present application will become apparent and be readily appreciated from the following description, including the accompanying drawings. Figure 1 A flow chart of an engine knock energy prediction method according to an embodiment of the present application is provided; Figure 2 A flow chart of a knock energy prediction model according to an embodiment of the present application is provided; Figure 3 An offline training flow chart of a knock energy prediction model according to an embodiment of the present application is provided; Figure 4 A structure diagram of an engine knock energy prediction device according to an embodiment of the present application is provided; Figure 5 A structure diagram of a vehicle according to an embodiment of the present application is provided. DETAILED DESCRIPTION
[0016] Embodiments of the present application are described in detail below with reference to the accompanying drawings, in which examples of embodiments are shown, wherein the same or similar notations are used to denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by reference to the accompanying drawings are exemplary and are intended to explain the present application, and cannot be understood as limiting the present application.
[0017] In the related art, the knock energy prediction method uses models such as spectrum analysis, support vector machine, or CNN (Convolutional Neural Network), and usually directly maps high-fidelity knock sensor data to low-fidelity data without considering the strong correlation between knock energy and engine operating state history time series. The deep learning models used are mostly standard convolutional neural networks and other general network architectures, which have the problems of high model complexity and poor generalization ability, making it difficult to improve the prediction accuracy.
[0018] At the same time, it is not adapted to the special working conditions of hybrid engines. The power coupling characteristics result in complex operating conditions and frequent start-stop, and when accelerating, the motor quickly drags the engine into the high-speed load area, which easily causes the end mixture to self-ignite. The engine intervention rate is less than 15% when the torque demand is small (<30%), and the sudden intervention under large torque conditions (≥70%) increases the turbulence intensity and in-cylinder pressure gradient in the combustion chamber by 40% and 65%, respectively, resulting in a significantly higher knock risk than traditional gasoline vehicles and higher prediction accuracy requirements.
[0019] Therefore, the application obtains multiple work phases running data and corresponding knock true energy, constructs a training set and a verification set, takes current running data as input and takes true energy as label, trains and verifies a knock energy prediction model, fuses multiple source heterogeneous data to improve anti-interference ability, adapts to complex working conditions of the hybrid power, and realizes low-cost and high-precision knock prediction.
[0020] The engine knock energy prediction method, device, vehicle, storage medium and product provided in the embodiments of the application are described below with reference to the drawings. In view of the problem of insufficient knock energy prediction accuracy of the hybrid engine due to differential working conditions caused by power coupling in the related art, the application provides an engine knock energy prediction method. In the method, an engine test running data set containing multiple work phase running data and corresponding knock true energy is obtained, a training data set and a verification set are generated according to the data set, the data sample is the running data of the current work phase, the true label is the knock true energy thereof, a pre-constructed knock energy prediction model is trained by using the training data set, the model is verified by using the verification set, and the knock prediction energy of the engine is predicted based on the currently collected running data and the model, so that multiple source heterogeneous data can be fused to enhance the anti-interference ability of the model, and the complex working conditions of the hybrid engine can be effectively adapted to, and high-precision knock energy prediction is realized. Thus, the problem of poor knock energy prediction accuracy of the hybrid engine due to differential working conditions caused by power coupling in the related art is solved.
[0021] Specifically, Figure 1 A flowchart of an engine knock energy prediction method provided in the embodiments of the application is shown in FIG. 1.
[0022] As Figure 1 shown, the engine knock energy prediction method includes the following steps: In step S101, a running data set of an engine test is obtained, and the running data set includes running data of the engine in multiple work phases and knock true energy corresponding to the running data of each work phase.
[0023] It can be understood that the running data set refers to a data set composed of multiple work phase running data and corresponding knock true energy synchronously collected under the working conditions of the hybrid engine; the work phase refers to a stroke in which the spark plug ignites and pushes the piston to do work in one working cycle of the engine, and the application collects the running data and the knock true energy based on this basic unit to construct a time series data set to support dynamic prediction of the knock energy.
[0024] In some embodiments, the operation data set of the engine test is acquired, including: dividing the rotating speed of the hybrid engine into a plurality of rotating speed intervals; dividing the load of the hybrid engine into a plurality of load intervals; generating a plurality of rotating speed-load intervals according to the plurality of rotating speed intervals and the plurality of load intervals, and collecting operation data of a current operation stage and corresponding knock true energy in each rotating speed-load interval.
[0025] wherein the load of the engine is a percentage of the current output torque of the engine to the rated maximum torque.
[0026] It can be understood that, by dividing the rotating speed-load intervals and collecting the operation data in each interval, the embodiment of the present application can realize fine coverage of the working conditions of the hybrid engine and improve the representativeness and distribution balance of the training data.
[0027] Specifically, the rotating speed of the hybrid engine is divided into commonly used intervals 1600 RPM to 4000 RPM of the hybrid engine, with an interval of 200 RPM, and a total of 12 intervals. The load is divided into commonly used intervals 30% to 100% of the hybrid engine, with an interval of 5% for each load interval, and a total of 14 intervals, totaling 168 rotating speed-load intervals. In the 168 rotating speed-load intervals, the engine is controlled to run at a fixed rotating speed-load point by an engine test bench dynamometer.
[0028] In some embodiments, as shown in FIG. 1, Figure 2 the operation data includes at least one of the ignition time, the ignition energy storage time, the fuel injection time, the fuel injection pulse width, the engine rotating speed, the load, the coolant temperature, the intake temperature, the air-fuel ratio, and the differential voltage signal.
[0029] It can be understood that the ignition energy storage time refers to the charging duration of the ignition coil, indicating the ignition energy storage state; the air-fuel ratio refers to the mass ratio of the actual inhaled air amount to the fuel amount in the engine combustion process; and the differential voltage signal is the output signal of the cylinder vibration signal collected by the knock sensor and processed by the differential amplification circuit.
[0030] Specifically, the operation data output by the engine control unit, i.e., the ignition time, the ignition energy storage time, the fuel injection time, the fuel injection pulse width, the engine rotating speed, the load, the coolant temperature, the intake temperature, the air-fuel ratio, and the differential voltage signal, are recorded in real time by a CAN (Controller Area Network) communication device as inputs of the knock energy prediction model.
[0031] In some embodiments, the knock true energy is the knock energy obtained by processing the cylinder pressure sensor.
[0032] Specifically, the knock true energy is obtained by processing the cylinder pressure signal collected by the cylinder pressure sensor through the combustion analyzer, and is taken as the training label P_Label of the knock energy prediction model.
[0033] In step S102, a training data set and a validation set are generated according to the operation data set, the validation set and the training data set each include data samples and true labels of the data samples, the data samples include operation data of the engine in a current working stage, and the true labels are knock true energies of the data samples.
[0034] It can be understood that the true label refers to the knock true energy corresponding to the current working stage of the engine; the training data set refers to a set composed of a plurality of data samples and corresponding true labels randomly selected or selected according to the working condition coverage principle from the engine test operation data set; and the validation set refers to a set composed of a plurality of independent data samples and corresponding true labels reserved from the same operation data set and not participating in model training.
[0035] Specifically, the input and label data of T_Re working stages are recorded in each speed-load interval respectively and stored in T_Train, a total of T_Memory segments of data are stored, T_Var segments of data are extracted as the validation set for offline training and testing, and the remaining T_Train segments of data are taken as the training data set.
[0036] In the formula, T_Re is the number of working stages collected at each working condition point; T_Memory is the total number of segments of working stage data collected under all working conditions; T_Var is the number of segments of working stage data of the validation set; T_Train is the memory unit, that is, the number of segments of working stage data of the training set; and T_Train is a buffer area for storing working stage data.
[0037] In step S103, the knock energy prediction model is trained by using the training data set, the knock energy prediction model is verified by using the validation set, and the knock prediction energy of the engine is predicted by using the currently collected operation data and the knock energy prediction model.
[0038] It can be understood that the knock energy prediction model can be trained by using the training data set and evaluated in combination with the validation set, so as to learn the nonlinear mapping relationship between the operation data and the knock true energy, improve the generalization ability of the model under complex working conditions of the hybrid engine, and realize high-precision prediction of the knock prediction energy based on the currently collected operation data.
[0039] In some embodiments, the knock energy prediction model comprises a multi-Transformer layer stacked encoder-decoder, a pooling layer and a fully connected network layer, wherein the encoder-decoder extracts knock-related features from the operation data through a local attention mechanism and a global attention mechanism, and the knock-related features are output as knock prediction energy after passing through the pooling layer and the fully connected network layer.
[0040] The knock-related features refer to time series features related to knock intensity extracted from operation data of multiple work stages of the engine by the encoder-decoder through the local attention mechanism and the global attention mechanism.
[0041] It can be understood that the Transformer is a neural network architecture based on an attention mechanism, which can model the dependency between time steps in the input sequence through local and global attention mechanisms, and has stronger advantages in time series modeling and model generalization. It can analyze all time points of engine sensor data in parallel through self-attention mechanism, capture the global dependency between knock signals and complex background noise, and thus realize more accurate and robust detection of knock.
[0042] In some embodiments, the knock energy prediction model further comprises an optimizer, which propagates back the model parameters of the knock energy prediction model according to the deviation between the knock prediction energy and the knock real energy corresponding to the real label.
[0043] It can be understood that the embodiments of the present application can use the optimizer to propagate back according to the deviation between the knock prediction energy and the knock real energy, which can dynamically adjust the model parameters and reduce the prediction error, and improve the convergence speed and prediction accuracy of the knock energy prediction model.
[0044] It should be noted that the knock energy prediction model can optimize the model parameters through offline training and online learning, so that the knock prediction energy output by the knock energy prediction model is approximately equal to the knock real energy obtained by processing the data collected by the cylinder pressure sensor.
[0045] Specifically, as shown in Figure 3 the process of offline training of the knock energy prediction model comprises the following steps: In step 301, T_Size data segments are extracted from T_Train each time, and each segment contains data of the current work stage and the previous P_PS work stages.
[0046] In step 302, the input data of the total T_PS work stages are input to the knock energy prediction model.
[0047] In step 303, the T_Size combustion analyzers record the actual data.
[0048] In step 304, the knock prediction energy is generated according to the knock energy prediction model.
[0049] In step 305, the combustion analyzer calculates the knock true energy.
[0050] In step 306, the knock prediction energy is compared with the knock true energy and the deviation is calculated.
[0051] In step 307, the deviation obtained in step S306 is taken as the training prediction deviation.
[0052] In step 308, the knock energy prediction model parameters are optimized according to the training prediction deviation of step S307.
[0053] In step 309, the optimized parameters are returned according to step S308 to realize iterative update.
[0054] Wherein, T_Size is the number of data segments extracted from the data set each time the training is performed; P_PS is the number of historical work stages on which the model input depends; T_PS is the total number of input work stages.
[0055] It should be noted that the knock energy prediction model completes one training each time T_Train is completely traversed, and then the output result is verified using the verification set and the verification error is recorded. If the verification error does not decrease for 5 consecutive times, the offline training is ended.
[0056] Further, after the knock energy prediction model completes offline training, the online learning method is used to improve the accuracy of the output of the knock energy prediction model. In the online learning stage, the hybrid engine is set to work in the commonly used speed and load range, and the knock energy prediction model records the ignition time, ignition energy storage time, fuel injection time, fuel injection pulse width, engine speed, load, coolant temperature, intake temperature, air-fuel ratio and differential voltage signal of the current work stage and the previous P_PS work stages through the controller area network communication device as the input of the model. The knock model continuously optimizes the model parameters through the stochastic gradient descent method to reduce the deviation between the knock prediction energy and the knock true energy P_Label obtained by processing the cylinder pressure signal collected by the combustion analyzer.
[0057] After completing offline training and online learning, the parameters of the knock energy prediction model are determined, and the label data P_Label calculated by using the combustion analyzer is no longer used. By inputting the ignition time, ignition energy storage time, fuel injection time, fuel injection pulse width, engine speed, load, coolant temperature, intake temperature, air-fuel ratio and differential voltage signal of the current work stage and the previous P_PS work stages recorded by the controller area network communication device into the knock energy prediction model, the knock energy prediction model realizes real-time output of the knock prediction energy. According to the engine knock energy prediction method provided in the embodiments of the present application, an engine test operation data set containing operation data in multiple working stages and corresponding knock true energy is obtained, a training data set and a verification set are generated according to the data set, the data sample is the operation data in the current working stage, the true label is the knock true energy thereof, the pre-constructed knock energy prediction model is trained by using the training data set, the model is verified by using the verification set, and the knock prediction energy of the engine is predicted based on the currently collected operation data and the model, so that the multi-source heterogeneous data can be fused to enhance the anti-interference capability of the model, the complex working conditions of the hybrid engine can be effectively adapted, and high-precision knock energy prediction is realized. Therefore, the problems in the related art, such as poor knock energy prediction precision of the hybrid engine due to the differential operation conditions caused by power coupling, are solved.
[0058] Secondly, the engine knock energy prediction device provided in the embodiments of the present application is described with reference to the accompanying drawings.
[0059] Figure 4 is a block schematic diagram of the engine knock energy prediction device in the embodiments of the present application.
[0060] As shown in Figure 4 , the engine knock energy prediction device 10 comprises an acquisition module 100, a generation module 200 and a training module 300.
[0061] The acquisition module 100 is configured to acquire an operation data set of an engine test, wherein the operation data set comprises operation data of the engine in multiple working stages and knock true energy corresponding to the operation data in each working stage. The generation module 200 is configured to generate a training data set and a verification set according to the operation data set, wherein the verification set and the training data set both comprise data samples and true labels of the data samples, the data sample comprises operation data of the engine in the current working stage, and the true label is knock true energy of the data sample. The training module 300 is configured to train a pre-constructed knock energy prediction model by using the training data set, verify the knock energy prediction model by using the verification set, and predict knock prediction energy of the engine by using currently collected operation data and the knock energy prediction model.
[0062] In some embodiments, the knock energy prediction model comprises a multi-Transformer layer stacked encoder-decoder, a pooling layer and a fully connected network layer, wherein the encoder-decoder extracts knock-related features from the operation data through a local attention mechanism and a global attention mechanism, and outputs the knock prediction energy after the knock-related features pass through the pooling layer and the fully connected network layer.
[0063] In some embodiments, the knock energy prediction model further comprises an optimizer, which back-propagates to optimize the model parameters of the knock energy prediction model according to the deviation of the knock prediction energy and the knock real energy corresponding to the real label.
[0064] In some embodiments, the acquisition module 100 is configured to: divide the rotation speed of the hybrid engine into a plurality of rotation speed intervals; divide the load of the hybrid engine into a plurality of load intervals; generate a plurality of rotation speed-load intervals according to the plurality of rotation speed intervals and the plurality of load intervals, and collect the operation data of the current operation stage and the corresponding knock real energy in each rotation speed-load interval.
[0065] In some embodiments, the operation data comprises at least one of the ignition time, the ignition accumulation time, the fuel injection time, the fuel injection pulse width, the engine rotation speed, the load, the coolant temperature, the intake temperature, the air-fuel ratio, and the differential voltage signal.
[0066] In some embodiments, the knock real energy is the knock energy obtained by processing the data collected by the cylinder pressure sensor.
[0067] It should be noted that the foregoing explanation of the engine knock energy prediction method embodiments is also applicable to the engine knock energy prediction device of the embodiments, which will not be described here again.
[0068] According to the engine knock energy prediction device provided in the embodiments of the present application, an engine test operation data set containing a plurality of operation data of working stages and corresponding knock real energies is acquired, a training data set and a verification set are generated according to the data set, the data sample is the operation data of the current working stage, the real label is the knock real energy thereof, a pre-constructed knock energy prediction model is trained by using the training data set, the model is verified by using the verification set, and the knock prediction energy of the engine is predicted based on the currently collected operation data and the model, so that the multi-source heterogeneous data can be fused to enhance the anti-interference capability of the model, the complex working conditions of the hybrid engine can be effectively adapted, and high-precision prediction of the knock energy is realized. Thus, the problems in the related art, such as the poor knock energy prediction precision of the hybrid engine due to the differential operation conditions caused by power coupling, are solved.
[0069] Figure 5 A structural schematic diagram of a vehicle is provided for the embodiments of the present application. The vehicle can include: The memory 501, the processor 502, and the computer program stored in the memory 501 and executable on the processor 502.
[0070] The processor 502 implements the engine knock energy prediction method provided in the above embodiments when executing the program.
[0071] Further, the vehicle further includes: The communication interface 503 is configured to communicate between the memory 501 and the processor 502.
[0072] The memory 501 is configured to store a computer program executable on the processor 502.
[0073] The memory 501 can include a high-speed RAM (Random Access Memory) memory, and can further include a nonvolatile memory, for example, at least one disk memory.
[0074] If the memory 501, the processor 502 and the communication interface 503 are independently implemented, the communication interface 503, the memory 501 and the processor 502 can be connected through a bus and communicate with each other. The bus can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component) bus or an EISA (Extended Industry Standard Architecture) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For the convenience of representation, Figure 5 Only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus.
[0075] Optionally, in a specific implementation, if the memory 501, the processor 502 and the communication interface 503 are integrated on a chip, the memory 501, the processor 502 and the communication interface 503 can communicate with each other through an internal interface.
[0076] The processor 502 can be a CPU (Central Processing Unit) or an ASIC (Application Specific Integrated Circuit) or one or more integrated circuits configured to implement the embodiments of the present application.
[0077] The embodiments of the present application further provide a computer readable storage medium, which stores a computer program, and the program is executed by the processor to implement the engine knock energy prediction method.
[0078] In the description of the application, the description of the terms "one embodiment", "some embodiments", "an example", "a specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the application. In the description of the application, the illustrative description of the above terms is not necessarily directed to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any appropriate manner in any one or N embodiments or examples. In addition, different embodiments or examples described in the description of the application and the features of different embodiments or examples can be combined and combined by those skilled in the art without contradiction.
[0079] In addition, the terms "first", "second" are only for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined with "first", "second" can explicitly or implicitly include at least one of the features. In the description of the application, the meaning of "N" is at least two, for example, two, three, etc., unless otherwise specifically limited.
[0080] Any process or method descriptions in flow charts or otherwise described herein can be understood as representing code modules, segments, or portions of code that include one or more executable instructions for implementing the specified logical functions or steps, and the various preferred embodiments of the application include additional or fewer steps or processes in addition to those shown in the figures and described herein, and the order in which the steps are presented is not necessarily the order in which the functions are performed, as those skilled in the art will appreciate. Additionally, the described embodiments can be implemented in hardware, software, firmware, or combinations thereof, and can be implemented in a computer program product.
[0081] It should be understood that parts of the application can be implemented in hardware, software, firmware, or a combination thereof. In the above-described embodiments, the steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. As in another embodiment implemented in hardware, any one or a combination of the following technologies known in the art can be used: discrete logic circuit with logic gate circuit for implementing logical functions on data signals, application specific integrated circuit with suitable combination logic gate circuit, programmable gate array, field programmable gate array, etc.
[0082] Those skilled in the art of the art can understand that the method of implementing the above-mentioned embodiments carries out all or part of the steps, which can be instructed by the program to complete the relevant hardware, and the above-mentioned program can be stored in a computer readable storage medium. The program includes one of the steps of the method embodiment or its combination when executed.
[0083] Although the embodiments of the present application have been shown and described above, it is understood that the above embodiments are exemplary, and are not to be interpreted as limiting the present application, and those skilled in the art can make changes, modifications, replacements and variations to the above embodiments within the scope of the present application.
Claims
1. An engine knock energy prediction method characterized by, The method comprises the following steps: obtaining an operation data set of engine test, wherein the operation data set comprises operation data of the engine in multiple working stages and knock true energy corresponding to the operation data of each working stage; generating a training data set and a verification set according to the operation data set, wherein the verification set and the training data set both comprise data samples and true labels of the data samples, the data samples comprise operation data of the engine in a current working stage, and the true labels are knock true energy of the data samples; training a pre-constructed knock energy prediction model by using the training data set, verifying the knock energy prediction model by using the verification set, and predicting knock prediction energy of the engine by using currently collected operation data and the knock energy prediction model.
2. The engine knock energy prediction method according to claim 1, characterized by, The knock energy prediction model comprises a coding decoder stacked by multiple Transformer layers, a pooling layer and a fully connected network layer, wherein the coding decoder extracts knock-related features from the operation data through a local attention mechanism and a global attention mechanism, and the knock-related features are output as knock prediction energy after passing through the pooling layer and the fully connected network layer.
3. The engine knock energy prediction method according to claim 2, characterized by, The knock energy prediction model further comprises an optimizer, which optimizes model parameters of the knock energy prediction model in a reverse direction according to a deviation between the knock prediction energy and knock true energy corresponding to the true label.
4. The engine knock energy prediction method according to claim 1, characterized by, The operation data set of the engine test comprises: dividing a rotating speed of the hybrid engine into multiple rotating speed intervals; dividing a load of the hybrid engine into multiple load intervals; generating multiple rotating speed and load intervals according to the multiple rotating speed intervals and the multiple load intervals, and collecting operation data of a current working stage and corresponding knock true energy in each rotating speed and load interval.
5. The engine knock energy prediction method according to any one of claims 1 to 4, characterized by, The operation data comprises at least one of ignition time, ignition accumulation time, fuel injection time, fuel injection pulse width, engine rotating speed, load, coolant temperature, intake temperature, air-fuel ratio and differential voltage signal.
6. The engine knock energy prediction method according to any one of claims 1 to 4, characterized by, The knock true energy is knock energy obtained by collecting and processing a cylinder pressure sensor.
7. An engine knock energy prediction device characterized by comprising: The method comprises: an obtaining module, configured to obtain an operation data set of engine test, wherein the operation data set comprises operation data of the engine in multiple working stages and knock true energy corresponding to the operation data of each working stage; a generating module, configured to generate a training data set and a verification set according to the operation data set, wherein the verification set and the training data set both comprise data samples and true labels of the data samples, the data samples comprise operation data of the engine in a current working stage, and the true labels are knock true energy of the data samples; a training module, configured to train a pre-constructed knock energy prediction model by using the training data set, verify the knock energy prediction model by using the verification set, and predict knock prediction energy of the engine by using currently collected operation data and the knock energy prediction model.
8. A vehicle characterized by comprising: The method comprises: A memory, a processor, and a computer program stored on the memory and executable on the processor, the processor executing the program to implement the engine knock energy prediction method of any one of claims 1-6.
9. A computer readable storage medium having stored thereon a computer program or instructions, characterized in that, The computer program or instructions, when executed, implement the engine knock energy prediction method of any one of claims 1-6.
10. A computer program product comprising computer programs or instructions, characterized in that, The computer program or instructions, when executed, implement the engine knock energy prediction method of any one of claims 1-6.