Method for determining the exhaust gas flow of a two-wheel combustion engine

Abstract

The invention relates to a method for determining the exhaust gas flow of a two-wheeled internal combustion engine of a given type. To design such a method in such a way that a sufficiently accurate determination of the exhaust gas flow can be achieved despite limited installation space and resources, it is proposed that engine speed and engine load parameters be recorded in successive time steps and fed as input data to an approximation unit with a type-specific approximation model, which comprises several regression models, each assigned to a parameter range. Based on the regression model assigned to the respective parameter range, an exhaust gas flow value is determined and output as the output data.

Description

[0001] The invention relates to a method for determining the exhaust gas flow of a two-wheel combustion engine of a given type.

[0002] The determination of environmentally harmful pollutants produced as exhaust gases in combustion engines has gained increasing relevance in recent years due to growing regulations. Since direct measurement of the exhaust flow during driving is usually only possible with corresponding attachments or installations that not only require installation space but also affect driving behavior and exhaust flow, there is a need for technical alternatives.

[0003] From US patent 20200355108A1, it is known to use machine learning to determine a vehicle's fuel consumption, whereby the exhaust mass flow and exhaust composition are determined from the fuel consumption using an exhaust gas volume sensor. However, the method shown there is not suitable for two-wheeled combustion engines, as the installation space is significantly smaller, which makes it difficult to incorporate additional sensors. Furthermore, in addition to the vehicle's own diagnostic system, a separate computing unit must be provided, which must be designed to calculate a comprehensive vehicle model.

[0004] The invention is therefore based on the objective of designing a method of the type described above in such a way that, despite limited installation space and resources, a sufficiently accurate determination of the exhaust gas flow can be achieved.

[0005] The invention solves the stated problem by acquiring engine speed and load parameters in successive time steps and feeding them to an approximation unit with a type-specific approximation model. This model comprises several regression models, each assigned to a specific parameter range. Based on the regression model assigned to the respective parameter range, an exhaust gas flow rate is determined and output. The invention is based on the finding that the exhaust gas flow rate, i.e., the exhaust gas mass flow rate and / or the exhaust gas volume flow rate, can surprisingly be derived directly from the engine speed and load parameters without direct measurement. Therefore, thanks to the measures according to the invention, a separate sensor can be dispensed with, thus saving installation space.Because the approximation model comprises several regression models, each assigned to a specific parameter range, that represent a local parameter range and can therefore be implemented simply and resource-efficiently, complex calculations are avoided while still achieving comparatively high accuracy within the local parameter range. An approximation model according to the invention can, for example, be a trained neural network, in particular a robust neural network (RNN), which has several local regression models depending on the parameter range of the input parameters. In this way, overfitting of the neural network during training can be avoided because not all regression models need to cover all variations of the input parameters. The transition between the individual regression models can be achieved by selective switching or by using gating functions.

[0006] As a result of the measures according to the invention, the two-wheel combustion engine can be designed without exhaust gas flow sensors. Consequently, the determination of the exhaust gas flow is achieved exclusively by the approximation model.

[0007] Reliable parameters for estimating exhaust gas flow can be achieved by including a calculated load value and / or a throttle position value in the engine load parameters. Considering these parameters not only enables a reliable estimation of the exhaust gas flow but also simplifies parameter acquisition, as this data is typically available as standard in the vehicle's on-board diagnostics (OBD) system. In a preferred embodiment, the calculated load value from the vehicle's diagnostic system is used as the engine load parameter. Alternatively, the throttle position value can be used directly as the engine load parameter or to calculate the calculated load value.

[0008] In order to better determine the course of the exhaust gas flow values ​​as a function of the engine speed and engine load characteristics despite low resource consumption, it is proposed that the individual regression models include a polynomial function approximation.

[0009] To improve the accuracy of exhaust gas flow determination in transition regions between regression models despite low resource consumption, the approximation model can be a weighted sum of the individual regression models. This avoids abrupt transitions between the regression models without requiring complex gating functions.

[0010] To further improve model accuracy, in a preferred embodiment, lambda values ​​can be acquired as input data and fed to the approximation unit. This allows the oxygen content of the exhaust gas flow to be considered in the modeling and the error of the exhaust gas flow value as output data to be further reduced. In a further preferred embodiment, it is recommended to acquire ambient temperature and pressure, the gear position of the transmission, and / or the temperature of the coolant or engine oil as input data and feed them to the approximation unit.

[0011] The invention also relates to a method for creating an approximation model for determining the exhaust gas flow of a two-wheeled internal combustion engine of a given type, wherein, for different engine speed and load characteristics of a two-wheeled internal combustion engine of the type, the corresponding exhaust gas flow value is recorded as input data and output data, characterized in that a type-specific approximation data set is created from the input and output data, which represents several regression models, each assigned to a characteristic value range. Due to the strong vehicle-specific relationship between the input data and the exhaust gas flow value, the invention proposes to create the approximation model for the approximation unit for each type, in particular for each vehicle type. For this purpose, the input and output data are measured during a test drive, which can be carried out on a real test track or, preferably, on a test bench.Multiple test drives, for example for steady-state measurements or acceleration tests, can be conducted to collect input and output data. Based on this input and output data, an approximation model with multiple local regression models can be created, and in particular, a neural network can be trained. A large number of input and output data points, preferably more than 10,000, and even more preferably more than 20,000, can be collected, covering a broad range of parameters. Therefore, it is recommended to perform a suitably designed test cycle, such as a Real Drive Cycle (RDC) and / or a World Motorcycle Test Cycle (WMTC), to record the input and output data. It is particularly preferred that the vehicle type's idling behavior during downhill driving be covered, including speed ranges exceeding 100% of the maximum rated engine speed.Although robust neural networks (RNNs) with multiple local models are particularly suitable for creating test cycles on the test bench by extrapolating from a few data points, it has been shown according to the invention that such neural networks, such as those of the AVL Cameo 5 validation and verification software package, are suitable for an approximation model if they are trained not on a few data points as intended, but rather on a large number of input and output data points that cover the parameter ranges occurring in operation. For a typical vehicle type, more than 10,000, and more preferably more than 20,000, input and output data points are required for this purpose. Verification drives on real test tracks can be carried out to verify the approximation model.

[0012] In particular, the invention proposes a method for determining the exhaust gas flow of a two-wheel combustion engine of a predetermined type, wherein engine speed and engine load characteristics are recorded in successive time steps and supplied as input data to an approximation unit with a type-specific approximation model, which comprises several regression models, each assigned to a characteristic value range, wherein an exhaust gas flow value is determined and output as output data based on the regression model assigned to the respective characteristic value range, wherein the approximation model is created by a method in which, for different engine speed and engine load characteristics of a two-wheel combustion engine of the type, the corresponding exhaust gas flow value is recorded as output data, and a type-specific approximation data set is created from the input and output data, which comprises several,Each regression model is assigned to a specific range of key figures.

[0013] The invention is illustrated in the drawing as an example. It shows Fig. 1 a schematic block diagram of an approximation unit according to the invention, Fig. 2 a schematic representation of the data points used to determine the approximation model and Fig. 3. The exhaust gas mass flow determined by measurement and the exhaust gas mass flow output by an approximation unit according to the invention during a test cycle are compared.

[0014] A method according to the invention for determining the exhaust gas flow of a two-wheel combustion engine 1 of a predetermined type is described using an approximation unit according to Fig. 1 explains. In successive time steps, both engine speed characteristics 2 and engine load characteristics 3 are recorded and fed as input data 4 to a type-specific approximation model 5. The approximation model 5 comprises several regression models 6, each assigned to a characteristic value range. Based on the input data 4, the approximation model 5 can determine and output exhaust gas flow values ​​7, for example, exhaust gas mass flow values ​​and / or exhaust gas volume flow values.

[0015] To create and / or validate the approximation model 5, measured and / or exhaust gas mass flow values ​​8 determined by the approximation model 5 can be used. Furthermore, other measured and / or determined parameters 9, such as exhaust gas volume flow values, can be incorporated into the approximation model 5.

[0016] To improve the model accuracy, additional characteristic values ​​10, such as lambda values, the measured temperature of the environment or a coolant, or the gear position, can be supplied to the approximation model 5 as input data 4.

[0017] The regression models 6 can include a polynomial function approximation. The approximation model 5 can be a weighted sum 11 of the individual regression models 6.

[0018] Depending on the characteristic range of the input data 4, at least one corresponding regression model 6 is selected in a switching unit 12, and an exhaust gas flow value 7 is determined as output data based on this regression model 6. In the case of several regression models 6, and especially in the transition range between the individual regression models 6, this is preferably done, as mentioned above, via a weighted sum 11.

[0019] In the Fig. Figure 2 schematically shows the recorded data points for a vehicle type, specifically the engine load characteristics 3 as a function of the engine speed characteristics 2, expressed as a percentage of the maximum load or speed. More than 20,000 data points can be used to create an approximation model 5. As shown in the Fig. As can be seen in section 2, speeds above the type-specific maximum engine speed can also be recorded, such as those that are possible during a downhill drive.

[0020] In the Fig. Figure 3 shows a measured exhaust gas mass flow 13, and, shown in dashed lines, an exhaust gas mass flow 14 determined using the method according to the invention. The exhaust gas determination exhibits high accuracy; noticeable differences are only found sporadically at exhaust gas mass peaks 15. QUOTES INCLUDED IN THE DESCRIPTION

[0000] This list of documents cited by the applicant was automatically generated and is included solely for the reader's convenience. The list is not part of the German patent or utility model application. The DPMA accepts no liability for any errors or omissions. Cited patent literature

[0000] US 20200355108A1

[0003]

Claims

[1] Method for determining the exhaust gas flow of a two-wheel internal combustion engine of a given type, characterized by , that in successive time steps engine speed characteristics and engine load characteristics are recorded and fed to an approximation unit with a type-specific approximation model as input data, which includes several regression models, each assigned to a characteristic value range, whereby an exhaust gas flow value is determined and output as output data based on the regression model assigned to the respective characteristic value range. [2] Method according to claim 1, characterized by that the engine load characteristics include a calculated load value and / or a throttle position value. [3] Method according to claim 1 or 2, characterized by that the individual regression models include a polynomial function approximation. [4] Method according to any one of claims 1 to 3, characterized bythat the approximation model is a weighted sum of the individual regression models. [5] Method for creating an approximation model for determining the exhaust gas flow of a two-wheel internal combustion engine of a given type, wherein for different engine speed characteristics and engine load characteristics of a two-wheel internal combustion engine of type as input data the corresponding exhaust gas flow value is recorded as output data, characterized by , that a type-specific approximation dataset is created from the input and output data, which maps several regression models, each assigned to a characteristic value range.

Images