WO2022237074A1 - Neural network model-based intake air amount estimation method for secondary inflation model of gasoline engine - Google Patents

Neural network model-based intake air amount estimation method for secondary inflation model of gasoline engine Download PDF

Info

Publication number
WO2022237074A1
WO2022237074A1 PCT/CN2021/125189 CN2021125189W WO2022237074A1 WO 2022237074 A1 WO2022237074 A1 WO 2022237074A1 CN 2021125189 W CN2021125189 W CN 2021125189W WO 2022237074 A1 WO2022237074 A1 WO 2022237074A1
Authority
WO
WIPO (PCT)
Prior art keywords
neural network
network model
intake air
engine
layer
Prior art date
Application number
PCT/CN2021/125189
Other languages
French (fr)
Chinese (zh)
Inventor
郑海亮
闫涛
郝伟
陈立
张文韬
王艳龙
冯朋朋
吴同
郭英俊
祝遵祥
Original Assignee
中国第一汽车股份有限公司
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 中国第一汽车股份有限公司 filed Critical 中国第一汽车股份有限公司
Publication of WO2022237074A1 publication Critical patent/WO2022237074A1/en

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
    • G01M15/00—Testing of engines
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/048—Activation functions
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods

Definitions

  • the invention relates to the field of engine parameter estimation, in particular to a method for estimating the intake air volume of a gasoline engine sub-charge model based on a neural network model.
  • the engine intake air volume is an important parameter in the engine electronic control system, and its accuracy has a great influence on the engine's running stability and economy, and is of great significance to the precise control of the engine electronic control system.
  • the intake air volume cannot be obtained directly through measurement; the current bench calibration method is to obtain the intake air volume through MAP chart interpolation and empirical formula fitting of the sub-charge model in the electronic control system of the gasoline engine. Affected by factors such as engine bench differences, throttle differences, and empirical formulas, the accuracy cannot be guaranteed, and repeated calibration is often required.
  • the invention provides a method for estimating the intake air volume of a gasoline engine sub-charge model based on a neural network model.
  • the estimation method is based on the test data of an engine bench test and uses a constructed neural network model to realize the estimation of the intake air volume of a gasoline engine sub-charge model.
  • the accurate real-time estimation of the engine solves the above-mentioned problems existing in the existing calculation method of the intake air volume of the engine.
  • a method for estimating the intake air volume of a gasoline engine secondary charge model based on a neural network model comprising the following steps:
  • Step 1 Measure and record the data of each stable operating point through the engine bench; the data includes engine speed, air-fuel ratio, fuel consumption, throttle opening, and pressure ratio before and after the throttle;
  • Step 2 Calculate the intake air volume according to the data recorded in step 1;
  • Step 3 normalize the throttle opening, the pressure ratio before and after the throttle, and the intake air volume
  • Step 4 building a neural network model
  • Step 5 Bring the processed data into the neural network model for training to achieve the required accuracy
  • Step 6 using the trained neural network model to estimate the intake air volume of the engine.
  • the concrete method of described step 1 is:
  • the concrete method of described step 2 is as follows:
  • airflow is the actual air intake of the engine, the unit is kg/h; lambda is the air-fuel ratio, the five-dimensional unit; fb_val is the engine fuel consumption, the unit is kg/h.
  • the concrete method of described step 3 is as follows:
  • the normalization processing formula is as follows:
  • the concrete method of described step 4 is as follows:
  • the neural network model includes three layers: the first layer is the input layer, and there are two input parameters: the normalized throttle opening and the pressure ratio before and after the throttle; the second layer is the middle layer containing 10 neurons
  • the third layer is the output layer, i.e. the intake air volume of the engine; the loss function in the neural network model adopts the mean square error function, and the training algorithm adopts the stochastic gradient descent algorithm.
  • i represents the number of sample points
  • y i is the neural network output value of the i-th sample point
  • x i is the i-th quantity
  • y i is the neural network output value of the i-th sample point
  • airflow i is the actual measured value of the i-th sample point.
  • the present invention can save the cost of the flow meter and omit the calibration link of the flow meter;
  • the present invention can be aimed at the engine that does not match the intake air flowmeter, can estimate the intake air volume through the neural network model according to the throttle opening and the front and rear pressure ratio, improves the accuracy of the results, and has a positive effect on the control stability of the engine electronic control system Great help.
  • Fig. 1 is a schematic diagram of the principle of the present invention
  • Fig. 2 is the schematic diagram of test working point
  • Fig. 3 is the schematic diagram of neural network model
  • Fig. 4 is a schematic diagram of the actual air intake of the engine
  • Fig. 5 is a schematic diagram of estimating the intake air volume by the neural network model
  • Fig. 6 is a schematic diagram of neural network model estimation deviation data
  • Fig. 7 is a schematic diagram of the estimated deviation percentage of the neural network model.
  • a method for estimating the intake air volume of a gasoline engine secondary charge model based on a neural network model includes the following steps:
  • Step 1 Measure and record the data of each steady-state operating point through the engine bench; the data includes engine speed, air-fuel ratio, fuel consumption, throttle opening, and pressure ratio data before and after the throttle;
  • Step 2 Calculate the intake air volume according to the data recorded in step 1;
  • the calculation method of intake air volume is as follows:
  • airflow is the actual intake air volume of the engine, the unit is kg/h; lambda is the air-fuel ratio, the unit is dimensionless; fb_val is the engine fuel consumption, the unit is kg/h.
  • Step 3 normalize the throttle opening, the pressure ratio before and after the throttle, and the intake air volume
  • step 4 constructing a neural network model
  • the neural network model includes three layers: the first layer is the input layer, and there are two input parameters: the normalized throttle opening and the pressure ratio before and after the throttle; the second layer is the middle layer containing 10 neurons ; The third layer is the output layer, which is the engine air intake; the transfer function between the first layer and the second layer is a sigmond transfer function. Since the neural network has only one output in the end, it is a regression problem, so no transfer function is needed.
  • the loss function in the neural network model adopts the mean square error function, and the training algorithm adopts the stochastic gradient descent algorithm.
  • i represents the number of sample points
  • y i is the neural network output value of the i-th sample point
  • x i is the i-th quantity
  • y i is the neural network output value of the i-th sample point
  • airflow i is the actual measured value of the i-th sample point.
  • Step 5 Bring the processed data into the neural network model for training to achieve the required accuracy
  • Step 6 using the trained neural network model to estimate the intake air volume of the engine.
  • Fig. 4 shows the actual intake airflow of the engine obtained by back-calculating the fuel consumption of the engine
  • Fig. 5 shows the engine intake airflow estimated by using the trained neural network model
  • Fig. 6 shows is the difference between the estimated value of the engine air intake model and the actual value
  • Figure 7 shows the deviation percentage between the estimated value of the engine air intake model and the actual value, which meets the model accuracy of greater than 95% of the working point deviation of less than 5% Requirements, the actual accuracy rate reaches 98.7%.
  • the present invention can calculate the intake air volume of the engine by collecting the engine speed, air-fuel ratio, and fuel consumption data at different throttle openings and pressure ratios before and after the throttle;
  • the engine air intake data is normalized;
  • the neural network model is constructed; the accuracy of the engine air intake estimated according to the engine throttle opening and the pressure ratio before and after the throttle is high.

Abstract

The present invention relates to the field of engine parameter estimation, and in particular, to a neural network model-based intake air amount estimation method for a secondary inflation model of a gasoline engine. The method comprises the following steps: 1, measuring and recording data of each steady-state operating condition point by means of an engine pedestal, the data comprising an engine speed, an air-fuel ratio, a fuel consumption amount, a throttle opening degree, and the ratio of pressures in front of and behind the throttle; 2, calculating an air intake amount according to the data recorded in step 1; 3, normalizing the throttle opening degree, the ratio of ratio of pressures in front of and behind the throttle, and the intake air amount; 4, constructing a neural network model; 5, substituting the processed data into the neural network model for training; and 6, estimating the intake air amount of an engine by using the trained neural network model. According to the present invention, for an engine not provided with a matched intake air flow meter, the intake air amount can be estimated by means of the neural network model according to the throttle opening degree and the ratio of pressures in front of and behind the throttle, the accuracy of the result is improved, and a great help is provided for the control stability of an engine electronic control system.

Description

一种基于神经网络模型的汽油机次充模型进气量估算方法A Neural Network Model-Based Estimation Method of Gasoline Engine Subcharge Model Air Intake 技术领域technical field
本发明涉及发动机参数估计领域，具体的说涉及一种基于神经网络模型的汽油机次充模型进气量估算方法。The invention relates to the field of engine parameter estimation, in particular to a method for estimating the intake air volume of a gasoline engine sub-charge model based on a neural network model.
背景技术Background technique
发动机进气量在发动机电控系统中是一个重要参数，其准确性对发动机的运行稳定性、经济性有较大影响，对发动机电控系统的精确控制有重要意义。针对未匹配流量计的汽油机，无法直接通过测量得到进气量；当前的台架标定方法是汽油机电控系统中的次充模型通过MAP图表插值结合经验公式拟合得到进气量，这一方法受到发动机台架差异、节气门差异、经验公式等因素影响，无法保证准确性，经常需要重复性标定。The engine intake air volume is an important parameter in the engine electronic control system, and its accuracy has a great influence on the engine's running stability and economy, and is of great significance to the precise control of the engine electronic control system. For gasoline engines that do not match the flowmeter, the intake air volume cannot be obtained directly through measurement; the current bench calibration method is to obtain the intake air volume through MAP chart interpolation and empirical formula fitting of the sub-charge model in the electronic control system of the gasoline engine. Affected by factors such as engine bench differences, throttle differences, and empirical formulas, the accuracy cannot be guaranteed, and repeated calibration is often required.
因此提出一种能够准确快速估算次充模型进气量的方法是有必要的。Therefore, it is necessary to propose a method that can accurately and quickly estimate the intake air volume of the subcharge model.
发明内容Contents of the invention
本发明提供了一种基于神经网络模型的汽油机次充模型进气量估算方法，该估算方法基于一次发动机台架试验的试验数据，运用构建的神经网络模型，实现对汽油机次充模型进气量的精确实时估算，解决了现有发动机进气量计算方法存在的上述问题。The invention provides a method for estimating the intake air volume of a gasoline engine sub-charge model based on a neural network model. The estimation method is based on the test data of an engine bench test and uses a constructed neural network model to realize the estimation of the intake air volume of a gasoline engine sub-charge model. The accurate real-time estimation of the engine solves the above-mentioned problems existing in the existing calculation method of the intake air volume of the engine.
本发明技术方案结合附图说明如下：The technical scheme of the present invention is described as follows in conjunction with accompanying drawing:
一种基于神经网络模型的汽油机次充模型进气量估算方法，包括以下步骤：A method for estimating the intake air volume of a gasoline engine secondary charge model based on a neural network model, comprising the following steps:
步骤一、通过发动机台架测量并记录每个稳定工况点的数据；所述数据包括发动机转速、空燃比、油耗量、节气门开度、节气门前后压力比值； Step 1. Measure and record the data of each stable operating point through the engine bench; the data includes engine speed, air-fuel ratio, fuel consumption, throttle opening, and pressure ratio before and after the throttle;
步骤二、根据步骤一记录的数据计算进气量； Step 2. Calculate the intake air volume according to the data recorded in step 1;
步骤三、对节气门开度、节气门前后压力比值以及进气量进行归一化处理；Step 3, normalize the throttle opening, the pressure ratio before and after the throttle, and the intake air volume;
步骤四、构建神经网络模型；Step 4, building a neural network model;
步骤五、将处理后的数据带入神经网络模型中进行训练，达到满足要求的精度； Step 5. Bring the processed data into the neural network model for training to achieve the required accuracy;
步骤六、利用训练后的神经网络模型估算发动机进气量。Step 6, using the trained neural network model to estimate the intake air volume of the engine.
所述步骤一的具体方法为：The concrete method of described step 1 is:
记录转速从1000转/分钟到5000转/分钟间隔500转/分钟，节气门开度从0％开始间隔2％，调节至节气门前后压力比值达到0.95停止的发动机转速、空燃比、油耗量、节气门开度、节气门前后压力比值数据。Record the engine speed, air-fuel ratio, fuel consumption, Throttle opening, pressure ratio data before and after the throttle.
所述步骤二的具体方法如下：The concrete method of described step 2 is as follows:
airflow＝lambda×fb_val×14.5airflow=lambda×fb_val×14.5
式中，airflow为发动机实际进气量，单位为kg/h；lambda为空燃比，五量纲单位；fb_val为发动机油耗量，单位为kg/h。In the formula, airflow is the actual air intake of the engine, the unit is kg/h; lambda is the air-fuel ratio, the five-dimensional unit; fb_val is the engine fuel consumption, the unit is kg/h.
所述步骤三的具体方法如下：The concrete method of described step 3 is as follows:
归一化处理公式如下：The normalization processing formula is as follows:
Figure PCTCN2021125189-appb-000001
Figure PCTCN2021125189-appb-000001
式中，y i为x i经过归一化处理后的值；x i为第i个量；x max为数据中最大值；x min为数据中最小值。 In the formula, y i is the normalized value of x i ; x i is the i-th quantity; x max is the maximum value in the data; x min is the minimum value in the data.
所述步骤四的具体方法如下：The concrete method of described step 4 is as follows:
所述神经网络模型包含三层：第一层为输入层，输入参数有两个：归一化处理后的节气门开度和节气门前后压力比值；第二层为中间层包含10个神经元；第三层为输出层即发动机进气量；所述神经网络模型中的损失函数采用均方差函数，训练算法采用随机梯度下降算法。The neural network model includes three layers: the first layer is the input layer, and there are two input parameters: the normalized throttle opening and the pressure ratio before and after the throttle; the second layer is the middle layer containing 10 neurons The third layer is the output layer, i.e. the intake air volume of the engine; the loss function in the neural network model adopts the mean square error function, and the training algorithm adopts the stochastic gradient descent algorithm.
所述输出层的计算公式为：The calculation formula of the output layer is:
Figure PCTCN2021125189-appb-000002
Figure PCTCN2021125189-appb-000002
式中，i表示样本点个数；y i为第i个样本点的神经网络输出值；
Figure PCTCN2021125189-appb-000003
为第一层第i个输入与第二层第k个神经元之间的权重；
Figure PCTCN2021125189-appb-000004
为第一层第i个输入与第二层第k个神经元之间的偏置；
Figure PCTCN2021125189-appb-000005
为第二层第k个神经元与输出的权重；
Figure PCTCN2021125189-appb-000006
为第二层第k个神经元与输出的偏置；x i为第i个量；
Figure PCTCN2021125189-appb-000007
为sigmond传递函数，表达式为：
Figure PCTCN2021125189-appb-000008
In the formula, i represents the number of sample points; y i is the neural network output value of the i-th sample point;
Figure PCTCN2021125189-appb-000003
is the weight between the i-th input of the first layer and the k-th neuron of the second layer;
Figure PCTCN2021125189-appb-000004
is the bias between the i-th input of the first layer and the k-th neuron of the second layer;
Figure PCTCN2021125189-appb-000005
is the weight of the kth neuron and output of the second layer;
Figure PCTCN2021125189-appb-000006
is the bias between the kth neuron and the output of the second layer; x i is the i-th quantity;
Figure PCTCN2021125189-appb-000007
For the sigmond transfer function, the expression is:
Figure PCTCN2021125189-appb-000008
所述均方差损失函数计算公式为：The formula for calculating the mean square error loss function is:
Figure PCTCN2021125189-appb-000009
Figure PCTCN2021125189-appb-000009
式中，y i为第i个样本点的神经网络输出值；airflow i为第i个样本点的实际测量值。 In the formula, y i is the neural network output value of the i-th sample point; airflow i is the actual measured value of the i-th sample point.
本发明的有益效果为：The beneficial effects of the present invention are:
1)本发明可以节省流量计费用和省略流量计标定环节；1) The present invention can save the cost of the flow meter and omit the calibration link of the flow meter;
2)本发明可以针对不匹配进气流量计的发动机，能够根据节气门开度和前后压力比值通过神经网络模型估算出进气量，提高了结果准确性，对发动机电控系统控制稳定性有较大帮助。2) The present invention can be aimed at the engine that does not match the intake air flowmeter, can estimate the intake air volume through the neural network model according to the throttle opening and the front and rear pressure ratio, improves the accuracy of the results, and has a positive effect on the control stability of the engine electronic control system Great help.
附图说明Description of drawings
为了更清楚地说明本发明实施例的技术方案，下面将对实施例中所需要使用的附图作简单地介绍，应当理解，以下附图仅示出了本发明的某些实施例，因此不应被看作是对范围的限定，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据这些附图获得其他相关的附图。In order to illustrate the technical solutions of the embodiments of the present invention more clearly, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention, and thus It should be regarded as a limitation on the scope, and those skilled in the art can also obtain other related drawings based on these drawings without creative work.
图1为本发明的原理示意图；Fig. 1 is a schematic diagram of the principle of the present invention;
图2为试验工况点的示意图；Fig. 2 is the schematic diagram of test working point;
图3为神经网络模型的示意图；Fig. 3 is the schematic diagram of neural network model;
图4为发动机实际进气量的示意图；Fig. 4 is a schematic diagram of the actual air intake of the engine;
图5为神经网络模型估算进气量的示意图；Fig. 5 is a schematic diagram of estimating the intake air volume by the neural network model;
图6为神经网络模型估算偏差数据的示意图；Fig. 6 is a schematic diagram of neural network model estimation deviation data;
图7为神经网络模型估算偏差百分比的示意图。Fig. 7 is a schematic diagram of the estimated deviation percentage of the neural network model.
具体实施方式Detailed ways
下面将结合本发明实施例中的附图，对本发明实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅是本发明一部分实施例，而不是全部的实施例。基于本发明中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都属于本发明保护的范围。The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts belong to the protection scope of the present invention.
实施例Example
参阅图1，一种基于神经网络模型的汽油机次充模型进气量估算方法，包括以下步骤：Referring to Figure 1, a method for estimating the intake air volume of a gasoline engine secondary charge model based on a neural network model includes the following steps:
步骤一、通过发动机台架测量并记录每个稳态工况点的数据；所述数据包括发动机转速、空燃比、油耗量、节气门开度、节气门前后压力比值数据； Step 1. Measure and record the data of each steady-state operating point through the engine bench; the data includes engine speed, air-fuel ratio, fuel consumption, throttle opening, and pressure ratio data before and after the throttle;
记录转速从1000转/分钟到5000转/分钟间隔500转/分钟，节气门开度从0％开始间隔2％，调节至节气门前后压力比值达到0.95停止的发动机转速、空燃比、油耗量、节气门开度、节气门前后压力比值。Record the engine speed, air-fuel ratio, fuel consumption, Throttle opening, pressure ratio before and after the throttle.
记录的数值如图2所示；The recorded values are shown in Figure 2;
步骤二、根据步骤一记录的数据计算进气量； Step 2. Calculate the intake air volume according to the data recorded in step 1;
进气量的计算方法具体如下：The calculation method of intake air volume is as follows:
airflow＝lambda×fb_val×14.5airflow=lambda×fb_val×14.5
式中，airflow为发动机实际进气量，单位为kg/h；lambda为空燃比，无量纲单位；fb_val为发动机油耗量，单位为kg/h。In the formula, airflow is the actual intake air volume of the engine, the unit is kg/h; lambda is the air-fuel ratio, the unit is dimensionless; fb_val is the engine fuel consumption, the unit is kg/h.
步骤三、对节气门开度、节气门前后压力比值以及进气量进行归一化处理；Step 3, normalize the throttle opening, the pressure ratio before and after the throttle, and the intake air volume;
归一化处理的具体方法如下：The specific method of normalization processing is as follows:
Figure PCTCN2021125189-appb-000010
Figure PCTCN2021125189-appb-000010
式中，y i为x i经过归一化处理后的值；x i为第i个量；x max为数据中最大值；x min为数据中最小值。 In the formula, y i is the normalized value of x i ; x i is the i-th quantity; x max is the maximum value in the data; x min is the minimum value in the data.
参阅图3，步骤四、构建神经网络模型；Referring to Fig. 3, step 4, constructing a neural network model;
所述神经网络模型包含三层：第一层为输入层，输入参数有两个：归一化处理后的节气门开度和节气门前后压力比值；第二层为中间层包含10个神经元；第三层为输出层即发动机进气量；第一层与第二层之间的传递函数为sigmond传递函数，由于本次神经网络最终只有一个输出，属于回归问题，所以不需要传递函数。所述神经网络模型中的损失函数采用均方差函数，训练算法采用随机梯度下降算法。The neural network model includes three layers: the first layer is the input layer, and there are two input parameters: the normalized throttle opening and the pressure ratio before and after the throttle; the second layer is the middle layer containing 10 neurons ; The third layer is the output layer, which is the engine air intake; the transfer function between the first layer and the second layer is a sigmond transfer function. Since the neural network has only one output in the end, it is a regression problem, so no transfer function is needed. The loss function in the neural network model adopts the mean square error function, and the training algorithm adopts the stochastic gradient descent algorithm.
所述输出层的计算公式为：The calculation formula of the output layer is:
Figure PCTCN2021125189-appb-000011
Figure PCTCN2021125189-appb-000011
式中，i表示样本点个数；y i为第i个样本点的神经网络输出值；
Figure PCTCN2021125189-appb-000012
为第一层第i个输入与第二层第k个神经元之间的权重；
Figure PCTCN2021125189-appb-000013
为第一层第i个输入与第二层第k个神经元之间的偏置；
Figure PCTCN2021125189-appb-000014
为第二层第k个神经元与输出的权重；
Figure PCTCN2021125189-appb-000015
为第二层第k个神经元与输出的偏置；x i为第i个量；
Figure PCTCN2021125189-appb-000016
为sigmond传递函数，表达式为：
Figure PCTCN2021125189-appb-000017
In the formula, i represents the number of sample points; y i is the neural network output value of the i-th sample point;
Figure PCTCN2021125189-appb-000012
is the weight between the i-th input of the first layer and the k-th neuron of the second layer;
Figure PCTCN2021125189-appb-000013
is the bias between the i-th input of the first layer and the k-th neuron of the second layer;
Figure PCTCN2021125189-appb-000014
is the weight of the kth neuron and output of the second layer;
Figure PCTCN2021125189-appb-000015
is the bias between the kth neuron and the output of the second layer; x i is the i-th quantity;
Figure PCTCN2021125189-appb-000016
For the sigmond transfer function, the expression is:
Figure PCTCN2021125189-appb-000017
所述均方差损失函数计算公式为：The formula for calculating the mean square error loss function is:
Figure PCTCN2021125189-appb-000018
Figure PCTCN2021125189-appb-000018
式中，y i为第i个样本点的神经网络输出值；airflow i为第i个样本点的实际测量值。 In the formula, y i is the neural network output value of the i-th sample point; airflow i is the actual measured value of the i-th sample point.
步骤五、将处理后的数据带入神经网络模型中进行训练，达到满足要求要求的精度； Step 5. Bring the processed data into the neural network model for training to achieve the required accuracy;
步骤六、利用训练后的神经网络模型估算发动机进气量。Step 6, using the trained neural network model to estimate the intake air volume of the engine.
参阅图4—图7，图4展示的是根据发动机油耗量反算得到的发动机实际进气量airflow；图5展示的是利用训练完成的神经网络模型估算得到的发动机进气量；图6展示的是发动机进气量模型估算值与实际值的差值；图7展示的是发动机进气量模型估算值与实际值的偏差百分比，满足大于95％的工况点偏差小于5％的模型精度要求，实际准确率达到98.7％。Refer to Fig. 4-Fig. 7. Fig. 4 shows the actual intake airflow of the engine obtained by back-calculating the fuel consumption of the engine; Fig. 5 shows the engine intake airflow estimated by using the trained neural network model; Fig. 6 shows is the difference between the estimated value of the engine air intake model and the actual value; Figure 7 shows the deviation percentage between the estimated value of the engine air intake model and the actual value, which meets the model accuracy of greater than 95% of the working point deviation of less than 5% Requirements, the actual accuracy rate reaches 98.7%.
综上，本发明通过采集不同节气门开度和节气门前后压力比值时的发动机转速、空燃比、油耗量数据,可以计算得到发动机进气量；对节气门开度、节气门前后压力比值和发动机进气量数据进行归一化处理；构建神经网络模型；根据发动机节气门开度和节气门前后压力比值预估的发动机进气量准确率高。To sum up, the present invention can calculate the intake air volume of the engine by collecting the engine speed, air-fuel ratio, and fuel consumption data at different throttle openings and pressure ratios before and after the throttle; The engine air intake data is normalized; the neural network model is constructed; the accuracy of the engine air intake estimated according to the engine throttle opening and the pressure ratio before and after the throttle is high.
以上结合附图详细描述了本发明的优选实施方式，但是，本发明的保护范围并不局限于上述实施方式中的具体细节，在本发明的技术构思范围内，任何熟悉本技术领域的技术人员在本发明揭露的技术范围内，根据本发明的技术方案及其发明构思加以等同替换或改变，这些简单变型均属于本发明的保护范围。The preferred implementation of the present invention has been described in detail above in conjunction with the accompanying drawings, but the protection scope of the present invention is not limited to the specific details of the above-mentioned implementation. Within the scope of the technical concept of the present invention, any person skilled in the art Within the technical scope disclosed in the present invention, equivalent replacements or changes are made according to the technical solutions and the inventive concepts of the present invention, and these simple modifications all belong to the protection scope of the present invention.
另外需要说明的是，在上述具体实施方式中所描述的各个具体技术特征，在不矛盾的情况下，可以通过任何合适的方式进行组合，为了避免不必要的重复，本发明对各种可能的组合方式不再另行说明。In addition, it should be noted that the various specific technical features described in the above specific embodiments can be combined in any suitable way if there is no contradiction. The combination method will not be described separately.
此外，本发明的各种不同的实施方式之间也可以进行任意组合，只要其不违背本发明的思想，其同样应当视为本发明所公开的内容。In addition, various combinations of different embodiments of the present invention can also be combined arbitrarily, as long as they do not violate the idea of the present invention, they should also be regarded as the disclosed content of the present invention.

Claims (6)

  1. 一种基于神经网络模型的汽油机次充模型进气量估算方法，其特征在于，包括以下步骤：A method for estimating the intake air volume of a gasoline engine secondary charge model based on a neural network model is characterized in that it comprises the following steps:
    步骤一、通过发动机台架测量并记录每个稳态工况点的数据；所述数据包括发动机转速、空燃比、油耗量、节气门开度、节气门前后压力比值；Step 1. Measure and record the data of each steady-state operating point through the engine bench; the data includes engine speed, air-fuel ratio, fuel consumption, throttle opening, and pressure ratio before and after the throttle;
    步骤二、根据步骤一记录的数据计算进气量；Step 2. Calculate the intake air volume according to the data recorded in step 1;
    步骤三、对节气门开度、节气门前后压力比值以及进气量进行归一化处理；Step 3, normalize the throttle opening, the pressure ratio before and after the throttle, and the intake air volume;
    步骤四、构建神经网络模型；Step 4, building a neural network model;
    步骤五、将处理后的数据带入神经网络模型中进行训练，达到满足要求的精度；Step 5. Bring the processed data into the neural network model for training to achieve the required accuracy;
    步骤六、利用训练后的神经网络模型估算发动机进气量。Step 6, using the trained neural network model to estimate the intake air volume of the engine.
  2. 根据权利要求1所述的一种基于神经网络模型的汽油机次充模型进气量估算方法，其特征在于，所述步骤一的具体方法为：A method for estimating the intake air volume of a gasoline engine secondary charge model based on a neural network model according to claim 1, wherein the specific method of said step 1 is:
    记录转速从1000转/分钟到5000转/分钟间隔500转/分钟，节气门开度从0％开始间隔2％，调节至节气门前后压力比值达到0.95停止的发动机转速、空燃比、油耗量、节气门开度、节气门前后压力比值数据。Record the engine speed, air-fuel ratio, fuel consumption, Throttle opening, pressure ratio data before and after the throttle.
  3. 根据权利要求1所述的一种基于神经网络模型的汽油机次充模型进气量估算方法，其特征在于，所述步骤二的具体方法如下：A method for estimating the intake air volume of a gasoline engine secondary charge model based on a neural network model according to claim 1, wherein the specific method of said step 2 is as follows:
    airflow＝lambda×fb_val×14.5airflow=lambda×fb_val×14.5
    式中，airflow为发动机实际进气量，单位为kg/h；lambda为空燃比，五量纲单位；fb_val为发动机油耗量，单位为kg/h。In the formula, airflow is the actual air intake of the engine, the unit is kg/h; lambda is the air-fuel ratio, the five-dimensional unit; fb_val is the engine fuel consumption, the unit is kg/h.
  4. 根据权利要求1所述的一种基于神经网络模型的汽油机次充模型进气量估算方法，其特征在于，所述步骤三的具体方法如下：A method for estimating the intake air volume of a gasoline engine secondary charge model based on a neural network model according to claim 1, wherein the specific method of the step 3 is as follows:
    归一化处理公式如下：The normalization processing formula is as follows:
    Figure PCTCN2021125189-appb-100001
    Figure PCTCN2021125189-appb-100001
    式中，y i为x i经过归一化处理后的值；x i为第i个量；x max为数据中最大值；x min为数据中最小值。 In the formula, y i is the normalized value of x i ; x i is the i-th quantity; x max is the maximum value in the data; x min is the minimum value in the data.
  5. 根据权利要求1所述的一种基于神经网络模型的汽油机次充模型进气量估算方法，其特征在于，所述步骤四的具体方法如下：According to claim 1, a method for estimating intake air volume of a gasoline engine secondary charge model based on a neural network model, is characterized in that, the specific method of said step 4 is as follows:
    所述神经网络模型包含三层：第一层为输入层，输入参数有两个：归一化处理后的节气门开度和节气门前后压力比值；第二层为中间层包含10个神经元；第三层为输出层即发动机进气量；所述神经网络模型中的损失函数采用均方差函数，训练算法采用随机梯度下降算法。The neural network model includes three layers: the first layer is the input layer, and there are two input parameters: the normalized throttle opening and the pressure ratio before and after the throttle; the second layer is the middle layer containing 10 neurons The third layer is the output layer, i.e. the intake air volume of the engine; the loss function in the neural network model adopts the mean square error function, and the training algorithm adopts the stochastic gradient descent algorithm.
  6. 根据权利要求5所述的一种基于神经网络模型的汽油机次充模型进气量估算方法，其特征在于，所述输出层的计算公式为：A method for estimating the intake air volume of a gasoline engine secondary charge model based on a neural network model according to claim 5, wherein the calculation formula of the output layer is:
    Figure PCTCN2021125189-appb-100002
    Figure PCTCN2021125189-appb-100002
    式中，i表示样本点个数；y i为第i个样本点的神经网络输出值；
    Figure PCTCN2021125189-appb-100003
    为第一层第i个输入与第二层第k个神经元之间的权重；
    Figure PCTCN2021125189-appb-100004
    为第一层第i个输入与第二层第k个神经元之间的偏置；
    Figure PCTCN2021125189-appb-100005
    为第二层第k个神经元与输出的权重；
    Figure PCTCN2021125189-appb-100006
    为第二层第k个神经元与输出的偏置；x i为第i个量；∮ 1()为sigmond传递函数，表达式为：
    Figure PCTCN2021125189-appb-100007
    In the formula, i represents the number of sample points; y i is the neural network output value of the i-th sample point;
    Figure PCTCN2021125189-appb-100003
    is the weight between the i-th input of the first layer and the k-th neuron of the second layer;
    Figure PCTCN2021125189-appb-100004
    is the bias between the i-th input of the first layer and the k-th neuron of the second layer;
    Figure PCTCN2021125189-appb-100005
    is the weight of the kth neuron and output of the second layer;
    Figure PCTCN2021125189-appb-100006
    is the bias between the kth neuron and the output of the second layer; x i is the i-th quantity; ∮ 1 () is the sigmond transfer function, the expression is:
    Figure PCTCN2021125189-appb-100007
    所述均方差损失函数计算公式为：The formula for calculating the mean square error loss function is:
    Figure PCTCN2021125189-appb-100008
    Figure PCTCN2021125189-appb-100008
    式中，y i为第i个样本点的神经网络输出值；airflow i为第i个样本点的实际测量值。 In the formula, y i is the neural network output value of the i-th sample point; airflow i is the actual measured value of the i-th sample point.
PCT/CN2021/125189 2021-05-12 2021-10-21 Neural network model-based intake air amount estimation method for secondary inflation model of gasoline engine WO2022237074A1 (en)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202110516531.6A CN113392574A (en) 2021-05-12 2021-05-12 Gasoline engine secondary charging model air inflow estimation method based on neural network model
CN202110516531.6 2021-05-12

Publications (1)

Publication Number Publication Date
WO2022237074A1 true WO2022237074A1 (en) 2022-11-17

Family

ID=77616974

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/125189 WO2022237074A1 (en) 2021-05-12 2021-10-21 Neural network model-based intake air amount estimation method for secondary inflation model of gasoline engine

Country Status (2)

Country Link
CN (1) CN113392574A (en)
WO (1) WO2022237074A1 (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113392574A (en) * 2021-05-12 2021-09-14 中国第一汽车股份有限公司 Gasoline engine secondary charging model air inflow estimation method based on neural network model

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20100298095A1 (en) * 2009-05-20 2010-11-25 Gm Global Technology Operations, Inc. Vehicle fuel control based on vacuum-assisted brake component actuation
CN110552804A (en) * 2019-09-05 2019-12-10 重庆邮电大学 method for estimating combustion timing of mixed fuel HCCI (homogeneous charge compression ignition) engine based on LSTM (localized fuel transfer model)
CN211058916U (en) * 2019-06-24 2020-07-21 同济大学 Engine air inlet control device of hybrid electric vehicle
CN113392574A (en) * 2021-05-12 2021-09-14 中国第一汽车股份有限公司 Gasoline engine secondary charging model air inflow estimation method based on neural network model

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20100298095A1 (en) * 2009-05-20 2010-11-25 Gm Global Technology Operations, Inc. Vehicle fuel control based on vacuum-assisted brake component actuation
CN211058916U (en) * 2019-06-24 2020-07-21 同济大学 Engine air inlet control device of hybrid electric vehicle
CN110552804A (en) * 2019-09-05 2019-12-10 重庆邮电大学 method for estimating combustion timing of mixed fuel HCCI (homogeneous charge compression ignition) engine based on LSTM (localized fuel transfer model)
CN113392574A (en) * 2021-05-12 2021-09-14 中国第一汽车股份有限公司 Gasoline engine secondary charging model air inflow estimation method based on neural network model

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
ZHENG HAILIANG; HAO WEI; CHEN LI; YAN TAO: "Simulation of Gasoline Engine Charge Model Based on Neural Network", PROCEEDINGS OF THE 2019 ANNUAL MEETING OF THE CHINESE SOCIETY OF AUTOMOTIVE ENGINEERING, CN, 31 October 2019 (2019-10-31), CN, pages 653 - 656, XP009542989 *

Also Published As

Publication number Publication date
CN113392574A (en) 2021-09-14

Similar Documents

Publication Publication Date Title
CN110118130B (en) Control device for internal combustion engine
DE102013209560B4 (en) Cylinder intake air quantity estimation device for an internal combustion engine
EP2339153B1 (en) Method and apparatus for measuring and controlling the egr rate in a combustion engine
EP2017452A1 (en) Airflow estimation method and apparatus for internal combustion engine
WO2022237074A1 (en) Neural network model-based intake air amount estimation method for secondary inflation model of gasoline engine
JP2816758B2 (en) Apparatus and method for measuring flow rate using fuzzy inference
CN101865710B (en) Method for measuring flow of negative-pressure gas
CN109840364A (en) The evaluation method of fuel cell water content
CN114781245A (en) Transient automobile oil consumption estimation method and system based on high oil consumption related parameter correction
CN109684704B (en) Engine intake flow online calibration method based on velocity density model
CN113267339B (en) Method for calculating pressure behind throttle valve, measuring device, engine and vehicle
CN109781415A (en) A kind of calculating of air input of engine by air and Bench calibration method
DE102020121741A1 (en) PROCEDURE TO PREVENT ENGINE AIRFLOW CALCULATION ERROR AND THE ENGINE SYSTEM THEREOF
CN116595403B (en) Real-time monitoring and managing system for gas state of exhaust pipeline
CN106460711A (en) Method and device for filling level detection in a cylinder of a combustion engine
CN109614722B (en) Modeling method for scroll engine full-state parameters based on fuzzy logic
EP3368758A1 (en) Systems and methods for in-cylinder pressure estimation using pressure wave modeling
CN109146239A (en) Based on FDA and theoretical Seepage Safety integrated evaluating method can be opened up
CN110321646A (en) A kind of multiphase flow rates virtual metrology method promoting regression tree model based on gradient
CN113482785B (en) Engine air inlet flow prediction method and system
CN112883653B (en) Artificial intelligence-based modeling method for real-time engine model
CN109222979A (en) The traffic alignment method of electronics vital capacity measuring device
CN112036079B (en) Diesel engine multi-information fusion diagnosis method based on ANFIS
CN109253884B (en) Turbine exhaust back pressure estimation method based on neural network
CN114254498A (en) Engine dynamic oil consumption prediction method and system based on BP neural network

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 21941639

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE