DE102018114081A1 - Method and device for determining a crankshaft torque in an internal combustion engine - Google Patents

Abstract

The invention relates to a method for determining a crankshaft torque for the operation of an internal combustion engine (10), the crankshaft torque being determined using a statistical data-based function model as a function of an operating point of the internal combustion engine (10).

Description

Technical field

The invention relates to internal combustion engines, and in particular control devices for internal combustion engines. The invention further relates to methods for determining a crankshaft torque of an internal combustion engine, in particular for operating the internal combustion engine.

Technical background

Knowledge of the current crankshaft torque for the execution of various functions is essential for the operation of internal combustion engines. In particular, it is used to monitor and control the driving torque.

In conventional engine systems, however, the crankshaft torque is not measured directly, but is modeled based on other measured or modeled state variables of the engine system, in particular using characteristic maps. This modeling is based on a torque balance that takes into account the individual driving moments and load moments acting on the crankshaft.

The driving torque that is directly relevant for the internal combustion engine corresponds to a high-pressure torque that is usually determined by a e.g. High-pressure torque model applied in the form of a map is modeled. Furthermore, internal engine load moments due to gas exchange losses using a low pressure torque model often applied in the form of a map and friction losses using a friction model are described, which can be modeled according to physical relationships or mapped using a suitable map model. When using map models, the effort for the application is high, since these must be valid over a large operating range of the internal combustion engine or must take them into account. This requires the engine system to be measured over the largest possible operating range, which makes the method very complex.

Control devices are generally known from the prior art which are suitable for calculating state variables in a motor system according to computing models. These control units generally include a microprocessor, which is equipped with suitable software for the model calculation. If such a control device implements a map model, it accesses a corresponding map memory in order to read out the required values.

Furthermore, control devices with integrated control modules with a main computing unit and a separate model calculation unit for calculating data-based function models are known from the prior art. For example, the publication shows DE 10 2010 028 266 A1 a control module with an additional logic circuit as a model calculation unit, which is designed for the purely hardware-based calculation of exponential functions as well as addition and multiplication operations. This makes it possible to support the calculation of Bayesian regression methods, which are required in particular for the calculation of Gaussian process models, in a hardware unit.

The model calculation unit is designed overall to carry out mathematical processes for calculating the data-based function model based on parameters and support points or training data. In particular, the functions of the model calculation unit for the efficient calculation of exponential and sum functions are implemented purely in hardware, so that it is possible to calculate Gaussian process models with a higher computing speed than can be done in the software-controlled main computing unit.

Disclosure of the invention

According to the invention, a method for determining a crankshaft torque of an internal combustion engine for operating the internal combustion engine according to claim 1 and the device and the engine system according to the independent claims are provided.

Further refinements are specified in the dependent claims.

According to a first aspect, a method for determining a crankshaft torque for the operation of an internal combustion engine is provided, the crankshaft torque being determined using a statistical data-based function model as a function of an operating point of the internal combustion engine.

One idea of the above method is to use a statistical function model that takes into account a large number of state variables, each of which represents a state of the engine system, in order to determine a current crankshaft torque. The function model can be used to implement statistical function models and control devices, in particular engine control devices for internal combustion engines.

Furthermore, the statistical data-based function model can correspond to a Gaussian process model. A variant of data-based function models are non-parametric models that can be created from training data, ie a set of training data points, without specific specifications. An example of a data-based function model is the so-called Gauss process model, which is based on a Gauss process regression. Gaussian process regression is a versatile statistical method for data-based modeling of complex physical systems based on usually large amounts of training data.

Furthermore, the crankshaft torque can be determined as the sum of a driving torque and a plurality of load torques, with at least the driving torque being determined as a high-pressure torque using the data-based function model.

The above method provides for the use of a data-based function model for modeling crankshaft torque. The crankshaft torque is usually determined by a balance of a driving torque and load moments. The driving torque corresponds to a high-pressure torque determined by a high-pressure torque model. The load moments include a low-pressure moment determined using a low-pressure moment model, which corresponds to a load moment resulting from charge changes in the internal combustion engine, a frictional moment determined by a friction model, which represents friction losses in the internal combustion engine, and further unit load moments, which are generated by units coupled to the crankshaft, such as, for example, load moments by driving a high pressure pump, an oil pump, a coolant pump, an electric generator and / or an air conditioning system. In particular, the aggregate load moments are dependent on other factors and suitable physical models are available to describe the corresponding aggregate load moments.

However, previous physical models, in particular for modeling the high-pressure torque and the low-pressure torque, are sometimes inaccurate and are generally not available for every operating range of the internal combustion engine that is possible in practice. The estimates made in these operating areas by extrapolations based on the corresponding map models are generally too imprecise to be able to reliably determine the crankshaft torque. By using a statistical function model for modeling the high-pressure moment and / or the low-pressure moment, the effort for supplying information to the map models previously used is significantly reduced. Providing reliable model values, even in previously unmeasured operating ranges, through the properties of statistical function models, can considerably reduce the effort required to model the crankshaft torque and additionally improve the accuracy in the calculation of the crankshaft torque. Overall, by using a statistical function model, the complex behavior of the internal combustion engine can be mapped better than is possible with a simple map model.

In particular, the statistical function model for calculating the high-pressure torque can be calculated based on input variables which can correspond to or are selected from the following input variables: crankshaft speed, injection quantity of a main injection, injection angle of a main injection, rail pressure, air mass, EGR mass, gas temperature of the inlet mass flow, Pre-injection quantity, post-injection quantities, post-injection crankshaft angle, oil temperature and swirl flap position.

According to one embodiment, one of the load torques can correspond to a low-pressure torque, which indicates the gas exchange losses of the internal combustion engine, and is determined using the data-based function model.

In particular, the statistical function model for calculating the low-pressure torque can be calculated based on input variables that can correspond to or are selected from the following input variables: crankshaft speed, exhaust gas back pressure, intake manifold pressure, partial flap position, EGR rate, inlet temperature and oil temperature.

Provision can be made for one or more of the following load torques to be taken into account as a further load torque: friction torque and unit load torques.

According to a further aspect, a control module for an engine system with an internal combustion engine is provided, the control module being designed to carry out the above method.

list of figures

• 1 eine schematische Darstellung eines integrierten Steuerbausteins mit einer hardwarebasierten Modellberechnungseinheit;
• 2 eine schematische Darstellung eines Motorsystems mit einem Verbrennungsmotor, der mit einem integrierten Steuerbaustein betrieben wird;

Embodiments are described below with reference to the accompanying drawings. Show it:

• 1 a schematic representation of an integrated control module with a hardware-based model calculation unit;
• 2 is a schematic representation of an engine system with an internal combustion engine which is operated with an integrated control module;

Description of embodiments

1 shows a schematic representation of a hardware architecture for an integrated control module 1 , e.g. B. in the form of a microcontroller in which a main processing unit is integrated 2 and a model calculation unit 3 are provided for the purely hardware-based calculation of a data-based function model. The main processor 2 and the model calculation unit 3 are via an internal communication link 4 , such as B. a system bus, in communication with each other.

Basically, the model calculation unit 3 essentially hard-wired and therefore not like the main processor 2 trained to execute software code. Alternatively, a solution is possible in which the model calculation unit 3 provides a restricted, highly specialized command set for calculating the data-based function model. In the model calculation unit 3 no processor is provided. This enables a resource-optimized implementation of such a model calculation unit 3 or a space-optimized structure in an integrated design.

The model calculation unit 3 has a computation core 31 on, which implements a calculation of a predetermined algorithm purely in hardware.

The model calculation unit 3 can also have a local SRAM 33 for storing the configuration data. The model calculation unit 3 can also be a local DMA unit 34 (DMA = Direct Memory Access). Using the DMA unit 34 it is possible to access the integrated resources of the control module 1 , especially the internal memory 5 to access.

The control module 1 can have internal memory 5 and another DMA unit 6 (DMA = Direct Memory Access). The internal memory 5 and the other DMA unit 6 stand in a suitable manner, e.g. B. via the internal communication link 4 , in connection with each other. The internal memory 5 can one (for the main processor 2 , the model calculation unit 3 and possibly further units) include common SRAM memory and a flash memory for the configuration data (parameters and reference point data).

The use of non-parametric, data-based function models is based on a Bayesian regression method. The basics of Bayesian regression are described, for example, in C. E. Rasmussen et al., "Gaussian Processes for Machine Learning", MIT Press 2006. Bayesian regression is a data-based method that is based on a model. To create the model, measuring points from training data and associated output data of an output variable to be modeled are required. The model is created on the basis of the use of support point data which correspond to the training data in whole or in part or are generated from these. Furthermore, abstract hyper parameters are determined that parameterize the space of the model functions and effectively weight the influence of the individual measuring points of the training data on the later model prediction.

The abstract hyper parameters are determined by an optimization process. One possibility for such an optimization process is to optimize a marginal likelihood. The marginal likelihood describes the plausibility of the measured initial values of the training data for given model parameters and measuring points of the training data. In model training, the marginal likelihood is maximized by searching for suitable hyper parameters that lead to a course of the model function determined by the hyper parameters and the training data and map the training data as precisely as possible. To simplify the calculation, the logarithm of the marginal likelihood is maximized to simplify the calculation, since the logarithm does not change the continuity of the plausibility function.

aus der sich der Funktionswert z ergibt. Dabei entsprechen D der Dimension des Eingangsdaten-/Trainingsdaten-/ Stützstellendatenraums, v einem Modellwert (Ausgangswert) an einem Testpunkt u (Eingangsgrößenvektor der Dimension D) mit u_1...D, x_i bzw. (x_i)_d einer Stützstelle der Stützstellendaten, N der Anzahl der Stützstellen der Stützstellendaten, sowie I_d , σ_f und der Parameter-Vektor Q_y den Hyperparametern aus dem Modelltraining.The following function is obtained in formula notation for creating the non-parametric, data-based function model: $$z={\displaystyle \underset{i=1}{\overset{N}{\Sigma }}{\left(Q_z\right)}_i{\sigma }_f\text{exp}\left(-\frac{1}{2}{\displaystyle \underset{d=1}{\overset{D}{\Sigma }}\frac{{\left({\left(x_i\right)}_d-u_d\right)}^2}{l_d}}\right)}.$$

from which the function value z results. D corresponds to the dimension of the input data / training data / support point data space, v a model value (output value) at a test point u (input variable vector of dimension D) with u _{1 ... D} , x _i or (x _i ) _{d of} a support point of the Reference point data, N the number of reference points of the reference point data, as well as I _d , σ _f and the parameter vector Q _y the hyperparameters from the model training.

In addition, input and output normalization can be carried out, since the calculation of the Gaussian process model typically takes place in a standardized space.

The computing unit in particular can be used to start a calculation 2 the DMA unit 34 or the further DMA unit 6 instruct the configuration data relating to the functional model to be calculated into the model calculation unit 3 to transfer and start the calculation that is carried out using the configuration data. The configuration data include the hyper parameters of a Gaussian process model as well as interpolation point data, which preferably use an address pointer to that of the model calculation unit 3 assigned address area of the internal memory 5 can be specified. In particular, the SRAM memory can also be used for this 33 for the model calculation unit 3 , which in particular in or on the model calculation unit 3 can be arranged can be used. You can also use the internal memory 5 and the SRAM memory 33 can be used in combination.

The calculation in the model calculation unit 3 takes place in a hardware architecture of the model calculation unit realized by the following pseudo code 3 , which corresponds to the calculation rule above. From the pseudo code it can be seen that calculations are carried out in an inner loop and an outer loop and their partial results are accumulated. At the beginning of a model calculation, a typical value for a counter start variable is Nstart 0 ,

The model data required for calculating a data-based function model thus include parameter vectors and reference point data which are stored in a memory area in the memory unit which is assigned to the relevant data-based function model. In accordance with the above pseudocode, the parameter vectors of data-based function models include the parameter vector Q _y and the length scale vector I, ie I _d for each dimension index d of the input variables of the input variable vector. Furthermore, the number N of support point data points, a start value Nstart of an outer loop when the calculation of the inner loop is resumed (normally = 0) is specified.

In 2 is a schematic representation of an engine system with a control module 1 and an internal combustion engine 10 , which is controlled by the control module. The internal combustion engine 10 is in the form of a conventional internal combustion engine with four-stroke mode and has several cylinders 11 on, which have a movable reciprocating piston with a crankshaft 12 is coupled. The crankshaft 12 delivers a drive torque when operating the internal combustion engine 10 ,

To control the internal combustion engine 10 knowledge of the current crankshaft torque is necessary. The crankshaft moment is the moment on the crankshaft 12 defined that is available for driving a motor vehicle or that is available at the transmission input. The crankshaft torque results from a torque balance of driving moments and load moments that are on the crankshaft 12 Act.

The driving torque corresponds to that of the internal combustion engine 10 Provided high pressure torque, the so-called internal torque, which is caused by combustion and compression processes in the internal combustion engine, in particular by the reciprocating piston operation.

Furthermore, gas changes occur in the internal combustion engine 10 Load moments, which are referred to as low pressure moments and through the movement of gas masses through the internal combustion engine 10 are caused.

In addition, the movement of moving parts in the internal combustion engine results 10 Frictional losses, which also reduce the crankshaft torque provided on the crankshaft as a frictional torque. The friction torque can be determined using a friction model that maps an engine temperature, an engine speed and an engine load to the friction torque.

You can also use the crankshaft 12 Aggregate 14 be coupled, such as an or several of the following aggregates 14 : a high pressure pump, an oil pump, a coolant pump, a power generator and an air conditioner, depending on the operating point of the internal combustion engine 10 can be controlled or activated or deactivated as required. Each of these aggregates 14 provides a corresponding aggregate load torque that is taken into account in the balancing to determine the crankshaft torque according to an aggregate load torque model.

The high-pressure torque model can now be calculated using a statistical function model, based on input variables that can correspond to or are selected from the following input variables: crankshaft speed, injection quantity of the main injection, injection angle of the main injection, rail pressure, air mass, EGR mass, gas temperature of the Inlet mass flow, pre-injection quantities, post-injection quantities, post-injection crankshaft angle, oil temperature and swirl flap position.

Furthermore, the low-pressure torque model for determining gas exchange losses can now be calculated using a statistical function model and take into account the following input variables or select input variables from the following input variables: crankshaft speed, exhaust gas back pressure, intake manifold pressure, partial flap position, EGR rate, intake temperature and oil temperature.

The use of the Gaussian process models for the calculation of the high-pressure moment and the low-pressure moment enables a simple creation of the model by measuring the internal combustion engine 10 in different operating areas.

The data-based function model relating to the determination of the high-pressure moment and / or the low-pressure moment can be created on the basis of training data recorded on a test bench. To do this, the internal combustion engine 10 z. B. operated across different operating areas and the corresponding state variables determined. These are now used to train the corresponding Gaussian process models in a manner known per se.

QUOTES INCLUDE IN THE DESCRIPTION

This list of documents listed by the applicant has been generated automatically and is only included for the better information of the reader. The list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions.

Patent literature cited

• DE 102010028266 A1 [0006]

Claims (11)

Method for determining a crankshaft torque for the operation of an internal combustion engine (10) is provided, the crankshaft torque being determined using a statistical data-based function model depending on an operating point of the internal combustion engine (10). Procedure according to Claim 1 , wherein the statistical data-based function model corresponds to a Gaussian process model. Procedure according to Claim 1 or 2 , wherein the crankshaft torque is determined as the sum of a driving torque and a plurality of load torques, wherein at least the driving torque is determined as a high-pressure torque using the data-based function model. Procedure according to Claim 3 , wherein the statistical function model for calculating the high pressure torque is calculated based on input variables which can correspond to or are selected from the following input variables: crankshaft speed, injection quantity of a main injection, injection angle of a main injection, rail pressure, air mass, EGR mass, gas temperature of the inlet mass flow, Pre-injection quantity, post-injection quantity, post-injection crankshaft angle, oil temperature and swirl flap position. Procedure according to Claim 3 or 4 , wherein one of the load torques corresponds to a low pressure torque, which indicates the gas exchange losses of the internal combustion engine (10), and is determined using the data-based function model. Procedure according to Claim 5 , wherein the statistical function model for calculating the low-pressure torque is calculated based on input variables which can correspond to or are selected from the following input variables: crankshaft speed, exhaust-gas back pressure, intake manifold pressure, partial flap position, EGR rate, inlet temperature and oil temperature. Procedure according to one of the Claims 1 to 6 , whereby one or more of the following load torques is taken into account as a further load torque: friction torque and unit load torques. Control module (1) for an engine system with an internal combustion engine (10), the control module (1) being designed to carry out a method according to one of the Claims 1 to 7 perform. Computer program which is set up to carry out the method according to one of the Claims 1 to 7 perform. Machine-readable storage medium on which a computer program Claim 9 is saved. Electronic control unit, which is an electronic storage medium Claim 10 having.

Images