DE102010031323A1 - Technical system i.e. internal-combustion engine, controlling/regulating method for motor vehicle, involves dividing gross error for components e.g. actuators and/or models, of technical system - Google Patents

Abstract

The method involves determining a gross error (FGes) of a technical system i.e. internal-combustion engine, from a comparison of an output signal of a sensor (S3) and redundantly determined physical system variables (P). The gross error is divided for components e.g. actuators and/or models (M1, M2), of the system, where the models perform redundant determination of the system variables. One of the models is parameterizable by output signals of additional sensors (S1, S2). The gross error is divided depending on fixed preset rules. Independent claims are also included for the following: (1) a computer program comprising a set of instructions to perform a method for controlling/regulating a technical system (2) a control- and regulation device for the technical system.

Description

State of the art

Many technical systems are controlled and / or regulated by control and regulating devices. These control and regulating devices, hereinafter referred to as control unit, usually have one or more sensors which detect relevant variables for the operation of the system and at least one, but as a rule more than one actuator. By means of the actuators, the system can be controlled so that it reaches or retains a predetermined operating state. In this case, functions are implemented in many control units, for example in the control units of internal combustion engines, which serve to detect faults in the technical system and to correct them in a suitable manner. There are various ways to detect a fault in the system. On the one hand, this can be due to the discrepancy between a measured value (output signal of a sensor) and a specified setpoint. Alternatively, it is also possible to detect an error occurring in the system by comparing a plurality of (diversified) redundant measurements or by comparing one or more measurements with one or more model-based variables.

If there is such a total error, the total error can be caused by a single component of the technical system. However, this case is the exception rather than the rule, especially for complex technical systems. As a rule, a total error is made up of the errors of several system components or models.

In the context of the present invention, the term "error" is understood to mean the deviation of a sensor value or model value from the actual, physical value of the system variable. In the case of actuators, the deviation of the system variable actually set by an actuator is understood to be a predefined setpoint value as an error.

In the following, the term "tolerance" is understood to mean the scattering of different copies of identically constructed sensors or actuators. In the context of a model of the technical system, tolerance means the average accuracy of the model.

This is preceded in the prior art, the total error of the system usually only one of the possible causes, for example, the statistically most likely, assigned and possibly corrected.

From the DE 103 31 159 A1 An approach is known in which the corrective intervention takes place in one place, wherein the correction intervention must not exceed a permissible maximum value. Any remaining error will then be corrected elsewhere in the system. The total error is thus "shifted" only when the maximum allowable correction intervention at one point to another point.

Disclosure of the invention

According to the invention, a method for controlling a technical system comprising at least one actuator and at least one sensor for detecting a physical system size and means for redundantly determining the system size is proposed, wherein a total error of the system is determined from a comparison of the output signal of the sensor with the redundantly determined system variables and the total error is always divided into at least two components, preferably all components of the system.

Due to the fact that, according to the invention, the total error is always allocated to all possible causes of the error or the components causing the error (actuators, sensors and / or models), a better estimation of the actual state of the technical system is made possible. As a result, it is possible to optimally distribute the correction to be made to all the faulty components of the system. Since the process according to the invention provides qualitatively better information about the actual state of the system, all subsequent control processes can be carried out with greater accuracy or fewer errors. As a result, the inventively optimized system becomes more economical, efficient and efficient.

In order to minimize the probability of an incorrect breakdown of a total error, it is preferably provided that at least two diverse means for redundant determination of the system size are present. For example, it is possible to check a sensor value using a model. The model can be parametrizable or non-parametrizable. In the following, a parameterizable model is understood to mean a model that contains parameters that are formed such that the latter represent input variables of the model.

A model represents a diverse approach to describing the system size captured by the sensor. Alternatively, other sensors, which preferably operate on a different measuring principle than the first sensor, for Evaluation of the determined by the first sensor value of the system size. Furthermore, other models, in particular parameterizable models, which in turn receive input variables of other system variables in order to parameterize the model, can also be used to evaluate the output signal of a sensor. Two different measuring principles are advantageous, because then the probability is lower that both sensors have the same or similar systematic error.

If one or more sensors fail completely, this usually has to be intercepted by means of suitable diagnostics.

Furthermore, it is advantageous for the same reason, if at least two diverse means, such as another sensor based on a different physical measuring principle, or a second model, one of the models is parametrisierbar and the other model is not parametrisierbar exist ,

It goes without saying that the effort for determining the overall error of a system size depends on the importance of the system size, so that in particularly important system sizes may have three or more redundant means, while at less critical system sizes, the review with a diverse redundant means is sufficient.

If the first means for redundant determination of the system size is designed as a model of a subarea of the system, then it is possible, for example, to adapt this model by the output values of other sensors and thereby to achieve a modeling adapted to the current operating point of the subsystem. This increases the accuracy of modeling. On the other hand, the number of possible error sources increases, since the input variables of the model are also subject to errors. Alternatively, it is therefore also possible to use models that can not be adapted. The imaging quality of these models is a bit less good in some operating points. However, the number of possible sources of error and the implementation effort are lower.

In a further advantageous embodiment of the invention, it is provided that the total error is divided among all models, sensors and actuators used to determine the total error. As a result, only relatively small corrections are made at the individual sensors, models and actuators, so that the stability of the system remains ensured.

The breakdown of the total error on the various system components or models must be carried out so that the actual state of the system is displayed as well as possible. For this purpose, for example, the total error can be distributed by an application defined map depending on the operating point or other sizes on several system components.

Another possibility is to determine from the tolerance of the components of the system according to statistical criteria the most probable distribution for the system errors. In this case, the tolerance of the subcomponents can in turn be predefined or in turn dynamically determined by a model. In addition, it is possible, as already known from the prior art, to set limits for the detected errors of the subcomponents. As a result, it can be prevented, for example, that the detected partial errors lead to an undesirably large system reaction as a result of the correction made. The error portions that exceed the corresponding limit can then optionally be assigned to the other components or ignored.

The functionality according to the invention can also serve to monitor individual system components. If the assigned error of a component exceeds a defined threshold, an error condition can be triggered.

A great advantage of the method according to the invention is the fact that the split and system corrections can better correct the actual system error or the actual partial errors of the subsystems than a correction that is limited to a single or two components of the system. In particular, the actual errors in the components are usually statistically distributed. Thus, the most likely system state for a given total error is one in which the errors are distributed among several system components.

If one takes as an exemplary application of the method according to the invention, the control of an internal combustion engine of a motor vehicle, or as the technical system, the internal combustion engine, then deviates in a variety of identical manufactured in series vehicles, a correction according to the invention of the total error on various individual errors therefore less of the Actual conditions of the individual vehicles than the assignment of the total error to only one component of the internal combustion engine. Moreover, it is possible with the inventive method to correct errors in which the distribution of the total error on several Partial error dynamically, that is, for example, depending on the operating point of the internal combustion engine changes. Thus, the amounts of individual errors can change at different operating points with the same total error.

Another advantage is that the state description of the system obtained by the method according to the invention, ie of all measured variables including their errors, is inherently consistent if the total error is completely distributed among the subcomponents of the system. If, for example, both by measurement with a sensor A and by measurement with a sensor B independently of one another on the same physical system size C closed, an error can be determined from the discrepancy and this can be divided between the two sensors. If the respective output signal of these sensors is corrected with the thus determined partial error before it is used in other user functions, the resulting system behavior is consistent in itself, regardless of which sensor value or which combination of sensor values use further user functions for their calculations.

Another advantage arises for functions of the system which monitor system failure. If this error is composed of several individual errors or partial errors, it is not possible to deduce the cause of a detected error directly. Due to the function according to the invention for distributing the error, the resulting distribution can be used in troubleshooting. As a result, under certain circumstances, the unauthorized exchange of components of the system can be prevented in individual cases. In any case, it is possible to create an improved troubleshooting plan.

Further advantages and advantageous embodiments of the method according to the invention are described below with reference to FIG 1 and 2 exemplified. All features described in the figures, their description and the claims may be essential to the invention both individually and in any combination.

1 a block diagram of the inventive method and

2 the application of the method according to the invention for checking the lambda sensor.

In 1 A sensor 3, which, like all sensors, is faulty and has an error FS3, measures a system variable P. This physical system variable P can also be determined by a model M1 or M2.

A distinction must be made between two groups of models M1 and M2. In the first group of models M2, the models only depend on input quantities which have no errors or whose errors are negligible. In the second group of models M1, the models depend on error-prone inputs.

Since each model M represents a simplified representation of the actual physical conditions, this model M2 also has an error FM2.

Finally, the physical system size P is still determined in a third diverse redundant manner with the aid of the model M1. This model M1 belongs to the first group of models that depends on input quantities that are subject to errors.

As a result, in this example, the system size P has been determined in three diversely redundant ways: firstly by the sensor S3, secondly by the non-parametrizable model M2 and thirdly by the parameterizable model M1.

Due to the errors FS1, FS2 and FS3 of the sensors S1, S2 and S3 and the errors FM1 and FM2 of the models M1 and M2, the values for the physical system variable P determined in the three different ways will differ from one another.

From the discrepancy of these three independently determined values of the system size P, one can conclude that there is a total error in the system. This will now according to the invention with the help of suitable criteria K _i on the errors of the sub-components, which were involved in the determination of the system size P, be divided. Thus one obtains an estimate for the instantaneous errors of the sensors S1, S2 and S3 as well as the models M1 and M2.

The estimation of the sensor errors FS1, FS2 and FS3 can be used to correct the sensor values of the sensors S1, S2 and S3 in order to then make them available to other user functions.

It is possible, the criterion or the criteria K _i , on the basis of which the breakdown of the total error is carried out on the various components and models, either according to fixed rules or online, for example, with the help of an operating point-dependent map split.

It is easier to split the total error according to fixed rules or conditions (offline), since this division is then operating point independent. However, as a consequence of this simplification, it is also to be expected that an optimal distribution of the total error among the various components and models will not take place at every operating point.

Alternatively, it is therefore also possible to split the criteria by means of which the total error is divided between the components and models, depending on operating point (online), for example, with the aid of a suitable characteristic map. As a result, the effort is correspondingly higher. However, this can ensure that a very realistic distribution of the total error on the various components and models takes place in all operating points and therefore the system state is mapped as best as possible.

Based on 2 For example, the method according to the invention will be explained and illustrated using the example of a function, the FMO, which monitors the injected fuel quantity.

An input variable of the function FMO is output signal λ _{s of} a lambda sensor LS. As a diverse redundant means for determining the lambda value λ Lambdamodell LM is implemented. This lambda model LM is a model that can be parameterized in the sense of the invention and has as input variables the output variables M _{air of} an air mass sensor LMS and a desired injection quantity of fuel QK, which is determined by the engine control unit (not shown). Thus, the modeled lambda signal λ _{M is} based on the output M _{air of} the air mass sensor LMS and the target injection amount QK of the fuel. The actual injected fuel quantity QK _is softened because of a possibly existing injection quantity error of the injectors from the target value QC, and the measured air mass is faulty. Further partial errors result from an error FLM in the lambda model LM and an error FLS of the lambda sensor LS itself.

In a functional block V, the deviation between the lambda value λ _s measured by the lambda sensor LS and the lambda value λ _M determined by the lambda model LM is determined and from this a total error F _{Ges is} formed.

In a conventional treatment of this error, the deviation between the measured lambda value λ _s and the lambda value λ _M determined by the lambda model is completely assigned to the injection quantity QK.

In the method according to the invention, however, the occurring total error F _{Ges is divided} into the proportions air mass sensor error FLMS, injection quantity error FQK, lambda sensor error FLS and lambda model error FLM.

In the example according to 2 the division of the total error is made by means of maps / characteristics. This map KF provides depending on the current injection amount QK and the measured air mass M _air factors determining the respective proportion of the air mass error FLMS, injection quantity error FQK, lambda signal error FLS and model errors FLM in the total error. The division of the total error can be determined for the proportions air mass sensor error and injection quantity error from the known tolerances of these components. For example, for injectors, an absolute tolerance of, for example, ± 3 mg / stroke is usually given, while for the air mass sensor LMS a relative tolerance of, for example, ± 7% is given. Based on these assumptions, for each operating point, which is determined by the injection quantity and the air mass, the percentage distribution of the total error can be stored in curves or in the characteristic map KF. For example, the injector tolerance of ± 3 mg / stroke at an injection rate of 43 mg / stroke results in a relative injector tolerance of ± 7%. It follows that at this operating point for the resulting lambda value, the influence of the injection quantity error and the air mass error are the same. An identified total error must therefore be divided equally between an injector error and an air mass error.

The inventive method thus leads in the illustrated embodiment to the fact that at low load (small injection quantity and air mass), the injection quantity is corrected more than the air mass. At high load (large injection quantity and air mass), the distribution of corrections reverses.

The known tolerance of, for example, ± 5% can also be used for the proportion of lambda sensor error FLS and can be determined analogously to the methodology described above via the current lambda value. For the proportion of the lambda model LM, the residual error FLM of the lambdam model LM can be determined in the characteristic diagram KF and be stored as a function of the operating point. Optionally, special dynamic errors and more can be considered here.

QUOTES INCLUDE IN THE DESCRIPTION

This list of the documents listed by the applicant has been generated automatically and is included solely 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.

Cited patent literature

• DE 10331159 A1 [0006]

Claims (13)

Method for controlling / regulating a technical system comprising at least one actuator (A1) and at least one sensor (S3) for detecting a (physical) system variable (P) and means (model M1, model M2) for redundantly determining the system variable (P) in which a total error (F _Ges ) of the system is determined from a comparison of the output signal (P) of the sensor (S3) and the redundantly determined system variables (P _M2 , P _s3 , P _M1 ), and wherein the total error ( F _Ges ) is always divided into at least two components (S3, M1), preferably all components (S3, M1, M2) of the system. Method according to one of the preceding claims, characterized in that at least two diverse means (M1, M2) for redundant determination of the system size (P) are present. A method according to claim 2, characterized in that first means for redundant determination of the system size (P) are formed as a model (M2) of a portion of the system, and that the input variables of this model (M2) have no error or only a negligible error. Method according to one of the preceding claims, characterized in that second means for redundant determination of the system size (P) as a model (M1) of a portion of the system are formed, and that the input variables of this model (M1) may be subject to errors. A method according to claim 3 or 4, characterized in that the model (M1) can be parameterized by the output signals of further sensors (S1, S2). Method according to one of the preceding claims, characterized in that the total error (F _Ges ) is divided among all models (M1, M2) and sensors (S1, S2, S3) used to determine the total error (F _Ges ) (FS1 , FS2, FM1, FM2, FS3). Method according to one of the preceding claims, characterized in that the total error (F _Ges ) is divided in dependence of fixed predetermined rules for the division of the total error (F _Ges ) (FS1, FS2, FM1, FM2, FS3). Method according to one of claims 1 to 6, characterized in that the total error (F _Ges ) is divided depending on the current operating point. A method according to claim 8, characterized in that the operating point-dependent division (FS1, FS2, FM1, FM2, FS3) of the total error (F _Ges ) is stored in a map (KF). Method according to one of the preceding claims, characterized in that the technical system is an internal combustion engine, preferably an internal combustion engine of a motor vehicle. Computer program, characterized in that it operates according to a method according to one of the preceding claims. Computer program according to claim 11, characterized in that it is stored on a storage medium. Control and regulating device for a technical system, characterized in that it operates according to a method according to one of claims 1 to 10.

Images