A Real-time Observation Method for Injection Quantity under Multiple Injections of a High-pressure Fuel System Based on Model Feature Transfer Learning

Abstract

The object of the present invention is to provide a method for real-time observation of injection quantity under multiple injections of a high-pressure fuel system based on model feature transfer learning, belonging to the field of fuel injection. The present invention collects excitation current signals, injector inlet pressure and injection rate data, establishes single-injection and multiple-injection data sets and performs batch normalization; constructs a single-injection rate prediction model based on a double-layer bidirectional long short-term memory neural network, and optimizes its performance using a quantum particle swarm optimization algorithm; transfers the characteristic parameters of the single-injection model to the multiple-injection model, and freezes the transferred parameters, and trains with a relatively small multiple-injection data set; based on the real-time measured excitation current signal and injector inlet pressure, uses the multiple-injection rate prediction model for real-time prediction to obtain the injection quantity at each injection stage. Compared with the present invention, the dependence on large data sets is significantly reduced, and real-time online accurate observation of the injection quantity at each stage under multiple injections can be realized.

Description

Technical Field

[0001] The present invention relates to a fuel system control method, specifically a fuel injection observation method. Background Art

[0002] The multiple injection technology is an important technology for modern high-pressure fuel systems. However, inaccurate injection can lead to incomplete combustion and low efficiency, resulting in unstable exhaust gas temperatures in each cylinder and reducing the conversion efficiency of the catalyst in the exhaust gas after-treatment device. The current fuel injection control method is mainly based on open-loop control embedded with an injection quantity calibration map, and determines the injection pulse width of the required injection quantity through the current engine speed and load. However, there are differences between the actual working environment of multiple cylinders of the engine and the single-cylinder measurement environment during the calibration map production, resulting in inconsistencies between the common rail pressure and the fuel pressure during the injector injection process. This makes it difficult to maintain consistency between the actual injection quantity and the desired target value.

[0003] In addition, the influence of the pressure fluctuations generated during the multiple injection process on the subsequent injection characteristics has not been fully considered. During the previous injection, the rapid opening and closing of the injector needle valve generates high-frequency expansion waves and compression waves. These pressure waves are reflected and superimposed in the high-pressure pipeline of the system, thus affecting the injection pressure in the subsequent injection stage and changing the actual injection rate. Therefore, the prediction of the injection rate for multiple injections is more complex than that for single injection. Especially under high-pressure and high-speed conditions, how to accurately predict the injection rate at each stage and perform feedback control is crucial for the overall performance and emission level of the engine. However, the existing single injection rate prediction methods based on mathematical model derivation often cannot meet the accurate prediction requirements for multiple injections. In addition, compared with single injection, the data-driven multiple injection rate prediction model requires a much larger dataset for training, which is usually an exponential multiple of the data volume required for single injection, requiring a large amount of measurement time and resources and increasing the R & D cost. Summary of the Invention

[0004] The purpose of the present invention is to provide a real-time observation method for the injection quantity under multiple injections of a high-pressure fuel system based on model feature transfer learning, which can solve problems such as the inability to accurately and real-time predict the injection law and injection quantity under multiple injections.

[0005] The purpose of the present invention is achieved as follows:

[0006] A real-time observation method for the injection quantity under multiple injections of a high-pressure fuel system based on model feature transfer learning according to the present invention is characterized by including the following steps:

[0007] (1) Clamp a current clamp on the wire of the injector high-speed solenoid valve, install a pressure sensor at the connection between the injector and the high-pressure fuel pipe, collect the excitation current signal of the high-speed solenoid valve and the pressure signal at the injector inlet, use a single injection instrument to collect the injection rate of the injector, obtain the data sets of the excitation current, injector inlet pressure, and injection rate for single injection, pre-main injection, and pre-main-post injection under different working conditions, divide the data sets into two categories, one is the single injection data set, and the other is the multi-injection data set, and perform batch processing normalization on the data within all data sets;

[0008] (2) Build a single injection rate prediction model based on the double-layer bidirectional long short-term memory neural network DL-BiLSTM, use the single injection data set to train the model, and at the same time use the quantum particle swarm optimization algorithm to optimize the model to improve the prediction performance;

[0009] (3) Build a multi-injection rate prediction model based on the double-layer bidirectional long short-term memory neural network, transfer the model feature parameters such as the first-layer weights and biases of the single injection rate prediction model constructed in step (2) to the first layer of the multi-injection rate prediction model, and freeze the transferred model feature parameters. On this basis, use the multi-injection data set to train the multi-injection rate prediction model;

[0010] (4) Based on the excitation current signal and the pressure signal at the injector inlet measured in real time by the high-pressure fuel system, use the multi-injection rate prediction model established in step (3) to perform real-time prediction of the injection rate under multi-injection, and obtain the real-time injection volume of each injection stage under multi-injection through segmented time integration.

[0011] The present invention may further include:

[0012] 1. In the above step (1):

[0013] In the single injection data set, the working conditions include pressure conditions and injection pulse width conditions. The pressure conditions include all pressure conditions at intervals of 10 MPa between the idle pressure and the rated pressure, and the injection pulse width conditions include all injection pulse width conditions at intervals of 0.2 ms between the minimum injection start pulse width and the rated pulse width;

[0014] In the multi-injection dataset, the pre-main injection operating conditions include pressure conditions, pre-injection pulse width conditions, pre-main injection interval conditions, and main injection pulse width conditions. The pre-main-post injection operating conditions include pressure conditions, pre-injection pulse width conditions, pre-main injection interval conditions, main injection pulse width conditions, main-post injection interval conditions, and post-injection pulse width conditions. The pressure conditions include all pressure conditions at intervals of 20 MPa between the idle pressure and the rated pressure. The pre-injection pulse width conditions include all pre-injection pulse width conditions at intervals of 0.05 ms between the minimum start injection pulse width and the maximum pre-injection pulse width. The main injection pulse width conditions include all main injection pulse width conditions at intervals of 0.2 ms between the minimum start injection pulse width and the rated pulse width. The post-injection pulse width conditions include all post-injection pulse width conditions at intervals of 0.05 ms between the minimum start injection pulse width and the maximum post-injection pulse width. The pre-main injection interval conditions include all interval conditions at intervals of 1 ms between the minimum pre-main injection interval and the maximum pre-main injection interval. The main-post injection interval conditions include all interval conditions at intervals of 1 ms between the minimum main-post injection interval and the maximum main-post injection interval.

[0015] 2. The data batch processing normalization in step (1) is specifically as follows:

[0016]

[0017] In the formula, x * is the processed data; represents the average value of the unnormalized data; represents the data standard deviation; ε is a margin used to prevent division by zero.

[0018] 3. Step (2) specifically includes:

[0019] (21) Establish the DL-BiLSTM single-injection rate prediction model framework, including an input layer, a first BiLSTM layer, a second BiLSTM layer, a fully connected layer, and an output layer; randomly discard a set proportion of neurons between the first BiLSTM layer and the second BiLSTM layer to reduce the model's dependence on specific neurons;

[0020] (22) Define the data format of the input layer as the drive current x a with n in the time dimension and the pressure x b at the injector inlet. Define that there are m groups of data under different operating conditions, and the input matrix of the model is Define the data format of the output layer as the injection rate y with n in the time dimension, and the output matrix is The input matrix and the output matrix are expressed as:

[0021]

[0022] (23) The model is trained using the single injection dataset. After the input data passes through the first - layer BiLSTM and the second - layer BiLSTM, dimensionality reduction and feature extraction are performed through the fully - connected layer, and the predicted injection rate matrix is obtained at the output layer.

[0023] (24) The quantum particle swarm optimization algorithm is used to iteratively optimize the model hyperparameters. The hyperparameters to be optimized are set as the number of units in the first - layer BiLSTM and the second - layer BiLSTM, the maximum number of iterations, and the initial learning rate. The root - mean - square error of the prediction results on the validation set is used as the fitness function, and the maximum consecutive number of iterations without improving the global best fitness value is used as the termination condition for optimization. The global optimal quantum particle position is obtained, and the corresponding optimal hyperparameters are used as the settings of the model.

[0024] 4. The specific functional form of the parameter transfer in step (3) is as follows:

[0025] θ T =argm θ inL T (f(X T ;θ frozen ,θ train ),Y T )

[0026] In the formula, θ T represents the model parameters trained on the multi - injection dataset, L T is the loss function of the multi - injection rate prediction model, θ frozen are the model feature parameters such as the weights and biases of the first - layer BiLSTM transferred and frozen from the single - injection rate prediction model, θ train are the remaining parameters to be trained, X T and Y T are the input data and output data of the multi - injection rate prediction model, respectively.

[0027] The advantages of the present invention are as follows:

[0028] (1) The construction process of the multi - injection rate prediction model of the present invention does not require a large amount of data sets. Only by transferring the model features and training with a relatively small number of multi - injection data sets can a high - precision prediction model be obtained.

[0029] (2) The present invention only needs to be based on the injector inlet pressure and the excitation current to real - time and online predict the injection rate and injection volume of each injection stage under multi - injection, so as to realize the online real - time and accurate observation of the injection information under multi - injection. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 is the flow chart of the present invention;

[0031] Figure 2 Schematic diagram of single injection rate prediction model structure and optimization;

[0032] Figure 3 Schematic diagram of multi-injection rate prediction model structure and migration process. Specific implementation mode

[0033] The present invention will be described in more detail with reference to the accompanying drawings as follows:

[0034] Combined with Figures 1 - 3 , a real-time observation method for injection quantity under multi-injection of a high-pressure fuel system based on model feature transfer learning of the present invention includes the following steps:

[0035] Step S1, clamp a current clamp on the wire of the high-speed solenoid valve of the injector, install a pressure sensor at the connection between the injector and the high-pressure fuel pipe, collect the excitation current signal of the high-speed solenoid valve and the pressure signal at the injector inlet. In addition, use a single injection instrument to collect the injection rate of the injector, obtain the data sets of the excitation current, injector inlet pressure, and injection rate of single injection, pre-main injection, and pre-main-post injection under different working conditions, divide the data sets into two categories, one is the single injection data set, and the other is the multi-injection data set, and perform batch normalization on the data in all data sets;

[0036] Step S2, construct a single injection rate prediction model based on a double-layer bidirectional long short-term memory neural network (DL-BiLSTM), use the single injection data set to train the model, and at the same time use the quantum particle swarm optimization algorithm to optimize the model to improve the prediction performance;

[0037] Step S3, as Figure 3 shown, construct a multi-injection rate prediction model based on a double-layer bidirectional long short-term memory neural network, transfer the model feature parameters such as the first-layer weights and biases of the single injection rate prediction model constructed in step S2 to the first layer of the multi-injection rate prediction model, and freeze the transferred model feature parameters. On this basis, use the multi-injection data set to train the multi-injection rate prediction model;

[0038] Step S4, based on the excitation current signal and the pressure signal at the injector inlet measured in real time by the high-pressure fuel system, use the multi-injection rate prediction model established in step S3 to perform real-time prediction of the injection rate under multi-injection, and obtain the real-time injection quantity of each injection stage under multi-injection through segmented time integration.

[0039] The working condition of the data set in step S1 is specifically:

[0040] In the single injection dataset, the operating conditions include pressure and injection pulse width. Among them, the pressure conditions include all pressure conditions from the idle pressure to the rated pressure at intervals of 10 MPa, and the injection pulse width conditions include all injection pulse width conditions from the minimum injection start pulse width to the rated pulse width at intervals of 0.2 ms;

[0041] In the multi-injection dataset, the pre-main injection operating conditions include pressure, pre-injection pulse width, pre-main injection interval, and main injection pulse width. The pre-main-post injection operating conditions include pressure, pre-injection pulse width, pre-main injection interval, main injection pulse width, main-post injection interval, and post-injection pulse width. Among them, the pressure conditions include all pressure conditions from the idle pressure to the rated pressure at intervals of 20 MPa, the pre-injection pulse width conditions include all pre-injection pulse width conditions from the minimum injection start pulse width to the maximum pre-injection pulse width at intervals of 0.05 ms, the main injection pulse width conditions include all main injection pulse width conditions from the minimum injection start pulse width to the rated pulse width at intervals of 0.2 ms, the post-injection pulse width conditions include all post-injection pulse width conditions from the minimum injection start pulse width to the maximum post-injection pulse width at intervals of 0.05 ms, the pre-main injection interval conditions include all interval conditions from the minimum pre-main injection interval to the maximum pre-main injection interval at intervals of 1 ms, and the main-post injection interval conditions include all interval conditions from the minimum main-post injection interval to the maximum main-post injection interval at intervals of 1 ms.

[0042] The data batch processing normalization in step S1 is specifically as follows:

[0043]

[0044] In the formula, x * is the processed data; represents the average value of the unnormalized data; represents the data standard deviation; ε is a margin used to prevent division by zero.

[0045] As Figure 2 shown, step S2 specifically includes:

[0046] Step S201, establish the DL-BiLSTM single injection rate prediction model framework, including an input layer, a first BiLSTM layer, a second BiLSTM layer, a fully connected layer, and an output layer; randomly discard a specific proportion of neurons between the first BiLSTM layer and the second BiLSTM layer to reduce the model's dependence on specific neurons;

[0047] Step S202, define the data format of the input layer as having n drive currents x a in the time dimension and the pressure x b at the injector inlet. Define that there are m groups of data under different operating conditions, and the input matrix of the model is Define the data format of the output layer as the injection rate y with n in the time dimension, and the output matrix is The input matrix and the output matrix are expressed as:

[0048]

[0049] In step S203, use the single-injection dataset to train the model. After the input data passes through the first-layer BiLSTM and the second-layer BiLSTM, dimensionality reduction and feature extraction are performed through the fully connected layer, and the predicted injection rate matrix is obtained in the output layer;

[0050] In step S204, use the quantum particle swarm algorithm to iteratively optimize the model hyperparameters. Set the hyperparameters to be optimized as the number of units in the first-layer BiLSTM and the second-layer BiLSTM, the maximum number of iterations, and the initial learning rate; use the root mean square error of the prediction results of the validation set as the fitness function, and use the maximum number of consecutive iterations without improving the global best fitness value as the termination condition for optimization to obtain the global optimal quantum particle position, and use the corresponding optimal hyperparameters as the settings of the model.

[0051] The specific functional form of the parameter migration in step S3 is:

[0052]

[0053] In the formula, θ T represents the model parameters trained on the multi-injection dataset, and L T is the loss function of the multi-injection rate prediction model, θ frozen is the model feature parameters such as the weights and biases of the first-layer BiLSTM migrated and frozen from the single-injection rate prediction model, θ train is the remaining parameters to be trained, X T and Y T are the input data and output data of the multi-injection rate prediction model respectively.

Claims

1. A method for real-time observation of injection quantity under multiple injections in a high-pressure fuel system based on model feature transfer learning, characterized by: The following steps are involved: (1) A current clamp is clamped on the wire of the high-speed solenoid valve of the injector, and a pressure sensor is installed at the connection between the injector and the high-pressure fuel pipe to collect the excitation current signal of the high-speed solenoid valve and the pressure signal at the injector inlet. A single injection instrument is used to collect the injection rate of the injector, and the excitation current, injector inlet pressure and injection rate data sets of single injection, pre-main injection, pre-main-post injection under different working conditions are obtained. The data sets are divided into two categories, one is a single injection data set, and the other is a multiple injection data set. The data in all the data sets are batch processed and normalized; (2) A single injection rate prediction model based on a double-layer bidirectional long short-term memory neural network (DL-BiLSTM) was constructed, the model was trained using a single injection dataset, and the quantum particle swarm algorithm was used to optimize the model to improve the prediction performance. (3) constructing a multiple injection rate prediction model based on a two-layer bidirectional long short-term memory neural network, migrating the first-layer weight and bias model feature parameters of the single injection rate prediction model constructed in step (2) to the first layer in the multiple injection rate prediction model, and freezing the migrated model feature parameters. On this basis, the multiple injection rate prediction model is trained using a multiple injection data set; (4) Based on the excitation current signal and the pressure signal at the injector inlet measured in real time by the high-pressure fuel system, the multiple injection rate prediction model established in step (3) is used to perform real-time prediction of the injection rate under multiple injections, and the real-time injection amount of each injection stage under multiple injections is obtained by segmented time integration.

2. The method for real-time observation of injection quantity under multiple injections of a high-pressure fuel system based on model feature transfer learning according to claim 1 is characterized by: In the step (1): In the single injection data set, the operating conditions include pressure conditions and injection pulse width conditions. The pressure conditions include all pressure conditions with an interval of 10 MPa from the idle pressure to the rated pressure, and the injection pulse width conditions include all injection pulse width conditions with an interval of 0.2 ms from the minimum start-up pulse width to the rated pulse width. In the multiple injection data set, the pre-main injection working conditions include pressure conditions, pre-injection pulse width conditions, pre-main injection interval conditions and main injection pulse width conditions, and the pre-main-post-injection working conditions include pressure conditions, pre-injection pulse width conditions, pre-main injection interval conditions, main injection pulse width conditions, main-post-injection interval conditions and post-injection pulse width conditions. The pressure conditions include all pressure conditions between the idle pressure and the rated pressure with an interval of 20MPa, and the pre-injection pulse width conditions include all pressure conditions between the minimum start-injection pulse width and the maximum pre-injection pulse width with an interval of 0.05ms. There are pre-injection pulse width conditions, the main injection pulse width conditions include all main injection pulse width conditions with an interval of 0.2ms between the minimum start-injection pulse width and the rated pulse width, the post-injection pulse width conditions include all post-injection pulse width conditions with an interval of 0.05ms between the minimum start-injection pulse width and the maximum post-injection pulse width, the pre-main injection interval conditions include all interval conditions with an interval of 1ms between the minimum pre-main injection interval and the maximum pre-main injection interval, and the main-post-injection interval conditions include all interval conditions with an interval of 1ms between the minimum main-post-injection interval and the maximum main-post-injection interval.

3. The method for real-time observation of injection quantity under multiple injections of a high-pressure fuel system based on model feature transfer learning according to claim 1 is characterized by: The data batch normalization in step (1) is specifically as follows: In the formula, x * is the processed data; represents the mean of unnormalized data; represents the standard deviation of the data; ε is the remainder used to prevent division by zero.

4. The method for real-time observation of injection quantity under multiple injections of a high-pressure fuel system based on model feature transfer learning according to claim 1 is characterized by: Step (2) specifically includes: (21) Establish a DL-BiLSTM single injection rate prediction model framework, including an input layer, a first BiLSTM layer, a second BiLSTM layer, a fully connected layer, and an output layer; randomly discard a set proportion of neurons between the first BiLSTM layer and the second BiLSTM layer to reduce the model's dependence on specific neurons; (22) The data format of the input layer is defined as n driving currents x in the time dimension. a and the pressure x at the ejector inlet b , it is defined that there are m groups of data under different working conditions, and the input matrix of the model is The data format of the output layer is defined as having n injection rates y in the time dimension, and the output matrix is The input matrix and output matrix are expressed as: (23) The model was trained using a single injection dataset. The input data passed through the first and second BiLSTM layers and then through a fully connected layer for dimensionality reduction and feature extraction, and the predicted injection rate matrix was obtained at the output layer; (24) The quantum particle swarm algorithm is used to iteratively optimize the model hyperparameters. The hyperparameters to be optimized are set as the number of units in the first layer BiLSTM and the second layer BiLSTM, the maximum number of iterations, and the initial learning rate. The root mean square error of the prediction results of the validation set is used as the fitness function, and the maximum number of consecutive iterations without improving the global optimal fitness value is used as the termination condition of the optimization. The globally optimal quantum particle position is obtained, and the corresponding optimal hyperparameters are used as the model settings.

5. The method for real-time observation of injection quantity under multiple injections of a high-pressure fuel system based on model feature transfer learning according to claim 1 is characterized by: The function form of parameter migration in step (3) is specifically: In the formula, θ T represents the model parameters trained on the multiple injection dataset, L T is the loss function of the multiple injection rate prediction model, θ frozen The first-layer BiLSTM weights and bias model feature parameters migrated and frozen from the single injection rate prediction model, θ train are the remaining parameters to be trained, X T and Y T They are the input data and output data of the multiple injection rate prediction model respectively.

Images