KR100654651B1 - Method and device for sensor monitoring, especially for an esp system for motor vehicles - Google Patents

Abstract

The present invention relates to a method of monitoring a sensor for sensing individual reference variables or measurement variables (A, B, C) related to a process in each case, and is particularly suitable for a vehicle electrical stabilization program (ESP). By the multiple process model 31 for a normal mode of operation, analytical redundancy is generated from process reference variables and process measurement variables that are not currently monitored, and the residual 33 is formed with redundancy along with the currently monitored output signal. And in a sequential manner, particularly high reliability is obtained by monitoring the progress of the output signals of the individual sensors. After the residual evaluation 36 and the comparison 35 with the limit value, an error signal F is generated when the residual reaches the limit value.

Description

TECHNICAL AND SENSOR MONITORING, ESPECIALLY FOR AN ESP SYSTEM FOR MOTOR VEHICLES

The present invention relates to a method and apparatus for monitoring a sensor for sensing each reference variable or measurement variable associated with a process in each case, and more particularly to a sensor monitoring method and apparatus for an electronic stabilization program (ESP) for a vehicle.

This type of electronic stabilization program is a driving-dynamics control system for vehicles that is used to assist the driver in dangerous driving situations during braking, acceleration and steering and to intervene where the driver himself has no direct intervention. This control system assists the driver during braking, especially where the vehicle may become uncontrollable or slippery due to the locked wheels on roads with low or variable friction coefficients. Assists the driver in steering during cornering that may oversteer or understeer. Overall, not only comfort but also active safety are significantly improved.

This type of control system is based on a closed loop control circuit that serves typical control purposes during normal vehicle driving and is intended to stabilize the vehicle as quickly as possible in extreme driving situations. Sensors for sensing various driving dynamic variables are particularly important as devices that generate real values. Prerequisites for proper control are an accurate representation of the actual conditions of the system under which the sensor is controlled. This is particularly important in operation stabilization control operation in extreme driving situations where control deviations must be adjusted by control in a very short time. This is why the electronic stabilization program ESP sensors (yaw rate sensor, lateral acceleration sensor, steering angle sensor) require constant monitoring. The purpose of the relevant on-line sensor monitoring is to detect errors in the ESP sensor early, in order to prevent erroneous control which may lead to dangerous vehicle conditions in terms of safety.

In view of the foregoing, it is an object of the present invention, in particular, to provide a method and apparatus for monitoring such types of sensors that provide the reliability required for an electronic stabilization program (ESP) for a vehicle.

This object is, in accordance with claim 1, a process reference variable or process measurement currently monitored from a process reference variable and / or a process measurement variable of a current process that is not currently being monitored based on a multiple process model for a normal mode of operation. Generating an analytical redundancy for the variable, generating a residual by subtracting the generated excess analytical redundancy from the currently monitored process reference variable or process measurement variable, and a residual evaluating function (residua evaluating) evaluating the residuals using a function) and comparing the evaluated residuals with a predetermined limit value, and generating an error message when the residual reaches the limit value for at least one scheduled monitoring time. Of the individual sensors sequentially It is achieved using this type of method characterized by monitoring the trend of the output signal.

This object is furthermore determined in accordance with claim 9, from the process reference variable and / or process measurement variable of the current process not currently monitored by the multiple process model for the normal mode of operation, to the currently monitored process reference variable or process measurement. A first device for calculating analytical redundancy for the variable, a second device for generating a residual by subtracting the excess analytical redundancy calculated from the currently monitored process reference variable or process measurement variable, and a residual using a residual evaluation function A third device for evaluating a second value, a fourth device for generating a limit value, and a comparison of the evaluated residual with a limit value and for generating an error message when the residual reaches the limit value for at least one predetermined monitoring time. An image comprising a fifth device It is accomplished using the type of device.

Dependent claims relate to preferred aspects of the invention.

Further details, features and advantages of the present invention can be further understood from the following description of the preferred embodiments according to the accompanying drawings.

1 is a block diagram of a driving dynamics control system.

2 is a schematic diagram of an ESP system architecture.

3 illustrates the basic principle of an error diagnosis system.

4 is a structure of a model monitoring system for the ESP sensor.

5 is a block diagram of a sensor monitoring arrangement in accordance with the present invention.

6A, 6B and 6C show measurement results in error simulation.

7A, 7B and 7C show measurement results in slalom operation.

According to FIG. 1, the act of driving a motor vehicle can be considered a control circuit in terms of control technology, where the driver 1 represents a controller and the vehicle 2 represents a controlled system. In this arrangement, the reference variable is the driver's personal driving needs FW that the driver creates by constantly monitoring road traffic. The actual value IF is an instantaneous value for the direction and speed of driving that the driver perceives with his eyes or with his sense of driving. The calibration parameter SF is the position of the accelerator pedal and the brake pedal provided by the driver based on the steering wheel angle, the position of the transmission and the mismatch between the nominal and actual values.

This type of control is often adversely affected by disturbances S, such as changes in coefficients of friction, road irregularities, cross winds or other effects, which the driver cannot accurately measure these factors but must take these into account in the control. Because. Thus, the driver 1 can generally handle the task of controlling and monitoring the process of driving a vehicle under normal driving conditions with no difficulty, with the help of training and all the experience gained. However, in the presence and / or extreme circumstances of the aforementioned emergency driving conditions beyond the physical limits of friction between the road and the tire, there is a risk that the driver may react too late or misrepresent the control of the vehicle.

In order to also take into account such a driving situation, the driving dynamic control system is provided with a lower control circuit 3 (ESP) comprising a control algorithm 4, a system monitoring device 5 and an error memory 6. The measured operating condition variable is sent to the system monitoring device 5 and the control algorithm 4. If necessary, the system monitoring device 5 generates an error message F which is transmitted to the error memory 6 and the control algorithm 4. The control algorithm 4 then acts on the vehicle 2 as a function of the calibration parameters produced by the driver 1. Typical control tasks are performed by this control circuit. The vehicle stabilizes as quickly as possible in extreme driving situations.

2 generally shows the structure of such a type of control circuit including an anti-lock system 10, a traction slip control system 11 and a yaw torque control system 12. . In addition, the yaw rate sensor 13, the lateral acceleration sensor 14, the steering angle sensor 15, the pressure sensor 16, and the actual value generator for determining the deviation, and the yaw rate nominal value and various intermediate values Four wheel speed sensors 17 are provided which are used as actual value generators to produce.

The process reference parameters generated by the driver 1 by actuating the accelerator pedal and the brake pedal and the steering wheel are determined by the traction slip control system 11, the lock prevention system 10 and the pressure sensor 16 and / or the steering angle sensor 15. Sent to). Vehicle-related nonlinearities, changes in friction coefficients, effects of side winds, etc. are summed up as obstacles or unknowns 18, which affect the longitudinal and transverse dynamics 19 of the vehicle. The dynamic 19 is further affected by the aforementioned reference variables and the output signals of the engine management unit 20, the wheel speed sensor 17, the yaw rate sensor 13, the lateral acceleration sensor 14 and the pressure. Acts on the sensor 16. The control adjustment 21 to which the output signals of the lock prevention system 10, the traction slip control system 11, the yaw torque control system 12 and the brake intervention algorithm 22 are supplied acts on the engine management unit 20 or It is used for priority distribution of these signals with respect to their action directly on the driving dynamics 19. The brake intervention algorithm 22 is affected by the yaw torque control system 12 and the pressure sensor 16. Also provided is a driving condition detection unit 23 through which signals from the steering angle sensor 15, yaw rate sensor 13, lateral acceleration sensor 14 and wheel speed sensor 17 are transmitted, which unit 23. The output signal of s affects the single track reference model 24 and the yaw torque control system 12 such that the desired nominal yaw rate is generated.

As already explained, wrong sensor signals can lead to dangerous inadequate control. Malfunctions of the yaw rate sensor 13 may have the result that, for example, additional yaw torque occurs and suddenly exert a lateral force on the vehicle despite the driver wishing to move forward. This results in the steering angle being zero while going straight, so the nominal value of yaw-rate becomes zero, but the actual value of yaw-rate is undefined because of sensor malfunction, so the yaw torque control system 12 adds an additional This is due to the fact that it produces yaw torque. Online monitoring of the sensor is very important for this reason. This monitoring should be able to detect early sensor malfunctions, in part or in whole, so that the ESP system can be shut down in a timely manner.

The sensor monitoring concept of the present invention also consists in testing the multi-level functionality of the sensor in which two methods are used. On the one hand, electrical monitoring is performed to check whether the observed sensor signal is within an acceptable error band. On the other hand, analytical redundancy monitoring is performed which allows the signal to be monitored over the full useful range.

In the first step, the sensor supply voltage and wiring are tested by an electrical monitoring configuration. In the second phase, due to its importance, these sensors configured 'intelligent' are always self-tested. If an internal sensor malfunction occurs, the sensor signal falls into the error band. Thus, such sensor malfunction may be detected by an electrical monitoring process.

It is only checked by electrical monitoring that the sensor signal is within the valid range. However, it is not possible with this monitoring to detect other sensor faults such as incorrect or loose installation locations, ground breaks and the like. For this reason, the trend of individual sensor signals within a useful range is monitored periodically and sequentially in the third stage, ie by analytical redundancy calculated from sensor output signals that are not currently monitored based on physical dependence. To this end, a model ESP monitoring and error diagnosis system is provided, the basic structure of which is shown in FIG. 3.

The error diagnosis system 100 basically consists of two parts: the residual generator 30 and the residual evaluation unit 34.

The residual generator 30 is generated by the process 32 that is currently occurring by using multiple process models (G1 to G4; Q1 to Q4; L1 to L4, see below) for the normal mode of operation. A first device 31 for calculating analytical redundancy for the currently monitored process reference variable or process measurement variable C from the process reference variable A and / or the process measurement variable B which is not currently monitored. do. Also provided is a second device 33 for generating a residual r by subtracting the calculated surplus analytical redundancy from the currently monitored process reference variable or process measurement variable C.

The residual evaluation apparatus 34 includes a third apparatus 36 for evaluating the residual r as a residual evaluation function, and a fourth apparatus 35 for generating a threshold value. In order to increase the threshold when the inaccuracy of the multi-process model is relatively high and decrease the threshold when the model is relatively inaccurate, the process reference variables and / or process measurement variables (A, B) that are not currently monitored are also It is also sent to this device 35. In addition, a fifth device 37 is provided for comparing the evaluated residual with the threshold and for generating an error message F when the residual reaches the threshold during at least one predetermined monitoring time.

In order to explain the problem to be solved by the present invention and to understand the solution of the present invention generally shown in FIG. 3, the following background information is given first.

 Using only one process model (instead of multiple process models) to generate the residuals, it is possible to obtain information about the current process conditions and possibly possible malfunctions. However, the effectiveness depends largely on the quality of the process model used. If the inaccuracy of the process model is increased, it is necessary to increase the limit value to prevent false alarms. As a result, many errors remain unrecognized. In contrast, an attempt to increase the accuracy of a process model simultaneously increases model complexity, which, when made, results in high costs incurred from the implementation of the model in on-line calculations and high demands on development and maintenance, resulting in actual manipulation. Often fail. Thus, the trade-offs between model accuracy and adjustment of limit values and the resulting system sensitivity play a large part in the development of a modeled ESP error diagnosis system.

It should also be taken into account that the process of driving a car is known to be greatly affected by many environmental factors which are still unknown. In addition, driving dynamics can be mathematically described to some extent. On the other hand, the limit of feasibility is defined from the beginning by hardware conditions. All of these limit conditions require solutions that are clearly based on the principles of model-based methods, and their use in ESP systems is nevertheless justified.

Modeling The basic concept of error diagnosis is the examination of the physical interrelationships described in the form of mathematical models. It is assumed as in Equation 1 below.                 

는 Equation 1 represents one of the physical correlations, where y represents the output signal of the sensor being monitored, u ₁ , ..., u _m represent the known or measured physical quantities, and f represents a mathematical function. . Analytical redundancy in this case

Is

And the residual r is

Becomes

Residuals are generally zero if there are no errors. If a sensor malfunction occurs, these rules are no longer useful because the residuals differ significantly from zero. The difficulty in realizing this concept is that the model only partially describes the process. This so-called model inaccuracy can be expressed by the extension of the process model as

Δ is an unknown amount depending on the process conditions. It is a prerequisite for reliable modeling error diagnosis that the effect of Δ on the residual is kept as minimal as possible.

In principle, there are two ways to suppress the influence of Δ.

Applying modern robust control theory to increase the robustness of the monitoring system: this is a passive method that generally requires more complex design and increased computational complexity (both offline and online).

2. Acquiring additional information: This can be realized in two ways: on the one hand, by obtaining an offline information and also by improving a model that makes the online computation more complex, on the other hand, by utilizing the additional online information. It is an active way. This approach has been found to be particularly preferred for reaching the solution of the above-mentioned problems according to the present invention.

The use of additional online information permits the creation of multiple (redundant) models for the monitored sensor, as well as signals from different sensors or unmonitored signal sources, and also to reconstruct the behavior and functionality of this sensor. do. On the one hand, this redundant analytical redundancy improves the reliability of the monitoring system and, on the other hand, the robustness to model inaccuracy. One preferred embodiment of the method of the present invention enables the realization of this basic principle, which is presented below.

It should be assumed that a model can be generated for the behavior of the monitored sensor signal using the equation system below.

Where u _ij (i = 1, ..., n; j = 1, ... m) represents signals from various signal sources, f ₁ , ... f _n are for partial models, Δ ₁ ,... _N represent model inaccuracies of individual partial models, PZ relates to process conditions, and GB _i (i = 1, ..., n) represents the range to which the partial model applies.

The usefulness and model inaccuracy of individual partial models depends on process conditions. The remaining problem is to create a residual by the multi-process model, which is sensitive to errors detected on the one hand and robust to model inaccuracy on the other.

The driving situation is divided into two groups for this purpose.

1. Abnormal driving behavior where model inaccuracy is very obvious and only a few partial models are useful.

2. Normal operating behavior with the common characteristics that most partial models are useful and model inaccuracies are low.

For 1: Abnormal driving behavior: The residual r among all possible residuals, since the absolute value of the residual is used as the residual evaluation function.

Is the result of the strongest relative model inaccuracy and at the same time the least sensitive to errors. Thus, for this driving situation the following is determined:

When the number of useful partial models is much smaller than the predetermined number (<< n), the residuals will be evaluated according to the principle of equation (2).

This is called the "minimum of all," and its basic concept implies an increased degree of robustness in an abnormal range where model inaccuracies are very evident.

For 2: Normal driving behavior: meaning that all or almost all partial models are useful

Is present, and therefore, in general, when normal process conditions exist, the residuals can be selected using the following algorithm:

의 생성First step: average value

Generation of

의 최저 편차를 갖는

의 계산 및

의 선택, 즉Second step:

Having the lowest deviation of

Calculation of and

Choice, i.e.

Third step: generation of residual r

가 적용되는 것으로 가정하면,

Assuming that applies,

이다.

to be.

To illustrate the functional principle of this algorithm, two cases are considered.

a) Fault-free operation: In this case, the following applies for the "best case":

This means that model inaccuracy does not affect the residuals. In the "worst case", the application is as follows:

The maximum possible deviation can be limited according to the following equation.

Since the generation of the mean value suppresses model inaccuracy in most cases, the variation caused by model inaccuracy is also minimized.

a) Sensor error: In this case, the following applies for "normal case":

The sensor signal y is markedly different from its normal value, and _hence y _ik , (k = 1,2,3) due to an error. As a result, there is a large difference between y and y _i2 . In the "worst case",

이 적용된다.

This applies.

This means that no error can be detected. However, this case only occurs when the magnitude of the error is within the range of model inaccuracy. It also shows that the acceptance limits of the monitoring concept are substantially determined by model inaccuracy.

As already explained, the concept of residual generation is a prerequisite for testing model usability. This includes testing the reliability of the signals used for generation of the residuals, and testing the model for use in response to driving conditions.

The signal is considered to be reliable when the signal is checked in terms of software or hardware. The reliable signal may be from another partial function of the system or may be from another sensor, which means mutual monitoring. This is online information used to generate a multi-process model.

As already explained, the generated residuals are highly dependent on model inaccuracies, and model inaccuracies are affected by various driving situations. Therefore, it is desirable to develop a residual evaluation unit appropriately adapted to the driving situation.

As is generally known, the driving operation can be described very accurately during normal driving. On the contrary, it is very difficult to mathematically reproduce a very sudden driving operation. Therefore, it is desirable to discern the driving situation according to each case and to adjust the monitoring limit and time appropriately based on this. The application of monitoring limits, on the one hand, means generating error messages in a timely manner when an inappropriate sensor signal appears and, on the other hand, avoiding false error messages that can be caused by inaccuracies in reproduction. This means that in operating situations where sensor signal reproducibility is less accurate, the limit value should be adjusted higher and the monitoring time length adjusted, while in other cases the limit value should be lowered and the monitoring time shortened.

The following describes how to realize the concept of monitoring three important ESP sensors: yaw rate sensor, lateral acceleration sensor and steering wheel sensor, as previously presented.

4 shows the structure of a modeling monitoring system for an ESP sensor, ie yaw rate sensor 13, lateral acceleration sensor 14 and steering angle sensor 15. For the monitoring of each ESP sensor, four redundant modes can be used as long as they are useful. These models are models G1 to G4 for yaw rate sensor 13, models Q1 to Q4 for lateral acceleration sensor 14 and models L1 to L4 for steering angle sensor 15. The mathematical realization of the process models and their usefulness is shown in Table 1. The symbols used in Table 1 are defined as follows.

- 모델 요-레이트.

Model yaw rate.

a _qm -model lateral acceleration.

δ _Lm -Model Steering Wheel Angle.

- 요-레이트.

-Yo-rate.

a _q -lateral acceleration.

δ _L -steering wheel angle.

i _L -Steering Ratio

l-wheel base

S-track width of the vehicle

v _ch -characteristic operating speed

Table 1 Equation Usability Condition Model G1

The two front wheels do not slip, their error flags are not set, and the reproduction is within the useful range. Model G2

The two rear wheels do not slip, their error flags are not set, and the reproduction is within the useful range. Model G3

The operating speed must exceed zero. Model G4

There is no counter steering and no big steering when the vehicle is at high speed. Model Q1

See conditions in model G1 Model Q2

See conditions of model G2 Model Q3

- Model Q4

- Model L1

See conditions of models G1 and G3 Model L2

See conditions of models G2 and G3 Model L3

See conditions of model G3 Model L4

See conditions of model G3

The models are used in the first device 31 and the following signals are used as input quantities to calculate the redundancy and thus determine the residual:

v _vr -wheel speed on the right front wheel

v _vl -wheel speed of the front left wheel

v _hr -wheel speed of the right rear wheel

v _hl -wheel speed of the left rear wheel

v _ref -vehicle speed

These are generated using a partial function of the lock prevention system, yaw rate, lateral acceleration and steering angle, and are from three ESP sensors 13, 14 and / or 15. The calculated redundancy is applied to the third device 36 for generating and evaluating the residuals (which also includes the second device 33), with the sensor signals being monitored respectively. After obtaining the difference between each residual and the threshold value generated by the fourth device 35, when the difference value exceeds the prescribed value, the fifth device 37 is applied to the yaw rate sensor 13. Error message F / UG, F / UQ for lateral acceleration sensor 14 or F / UL for steering angle sensor 15. The third, fourth and fifth devices 36, 35, 37 for each sensor 13, 14, 15 are respectively shown in FIG. 4.

The reliability test of signals generated by the lock prevention system is performed by the monitoring system present therein. The signals are believed to be reliable in the absence of the error message and are not applicable in the presence of the error message.

In the case of the three ESP sensor signals (yaw-rate, lateral acceleration, steering angle) for the sensor monitoring system described here, each signal is determined to be reliable in the absence of an error message, and when the error signal appears Stops working.

As mentioned earlier, the method 'majority principle' is sensitive to sensor failure when the sensor is defective, but the method 'minimum of all' is a system malfunction and transient or extreme operation. More robust in terms of behavior. The aforementioned monitoring concept is realized as follows.

When the number of useful models is less than 3, the residuals are generated according to the 'minimum of all' principle. In all other cases, the residual is generated by the principle of majority vote.

In theory, all these process models apply only to driving dynamics in the normal or linear range. When the driving behavior is no longer in this range, the monitoring limit has to be increased and the monitoring time has to be extended. This is done by detecting the situation in the fourth device (limit calculation 35) and determining the degree of deviation of the driving behavior from the normal or linear range (compare FIG. 4). The signals used for this purpose are: vehicle speed v _ref , wheel speeds of four wheels v _vr , v _vl , v _hl , v _hr , longitudinal accelerations of the vehicle resulting from and investigated here partly function a ₁ And calculated redundancy and other ESP sensor signals.

These monitoring limits and times are set or determined by investigation of vehicle behavior in the presence of various types of errors in various driving situations. When a change in the yaw rate sensor with a large degree of change not related to any of the possible driving operations is detected, the monitoring time is significantly shortened.

The adjustment of monitoring limits and times is briefly summarized in Table 2.

Table 2: Monitoring Limits and Times Driving situation Process Model Accuracy Monitoring limits Monitoring time Error with high slope Etc Normal driving: straight driving and normal round driving Very accurate littleness Very short short Temporary driving where the driving motion is close to the normal or straight range, eg slalom operation inaccuracy greatness Very short 긺 Driving movements very far from the normal or straight range, eg sliding Cannot be expressed Infinitely large (no monitoring) Infinite 긺 (no monitoring)

The structure of the hardware realization is shown in FIG. This structure comprises a microprocessor system 40 having an output signal sent to the unit 41 for brake and engine intervention.

The microprocessor system 40 includes an analog / digital converter 401 for converting analog sensor signals, and a digital / analog converter 403 connected to the analog / digital converter 401 and for generating an analog output signal. A digital control algorithm 402 is connected. The digital sensor signal is again sent to the monitoring system 404 which applies the system quantity generated by the digital control algorithm 402 and sends an error message to this unit 402.

An ESP system comprising both a digital control algorithm and a monitoring system is preferably programmed in the C language and then executed into the microprocessor system 40. The input signal of the microprocessor system 40 is the signal generated by the sensor 43 installed in the vehicle 42. The output signal of the microprocessor system 40 is a calibration parameter sent for the brake and engine management system 41. The monitoring system 404 operates in parallel with the control system and monitors the entire system and thus does not affect control when no error is detected. When an error is detected, the monitoring system 404 sends an error signal to the digital calculation algorithm 402 which will stop the operation of the ESP system.

The monitoring system was tested in several operational tests. 6A, 6B and 6C, the measurement results of the two road test runs are shown as an example, that is, the result of the error simulation of the yaw rate sensor during the straight run.

6A shows the signal of the yaw rate sensor (line 1) and its four representations (lines 2-5). 6B shows the transition and limits (lines 2 and 3) of the residual (line 1). In Fig. 6C, it is shown when an error message has occurred.

It is clear from these figures that the yaw rate can be explained very accurately. The simulated error was detected within 0.25 seconds before yaw torque control applied high pressure to the wheel.

7A, 7B, and 7C show measurement results of monitoring a yaw rate sensor while riding with a slalom maneuver. During this slalom operation, the yaw rate cannot be accurately accounted for due to the phase shift between the sensor signal and the model signal. In this situation, when a model is created, generally something like model inaccuracy cannot be prevented. In order to prevent false alarms from being generated, the monitoring limit is already increased at the beginning of the slalom operation. 7A shows the signal of the yaw rate sensor (line 1) and its four representations (lines 2 to 5). In FIG. 7B, while the trend of the residuals (line 1) and limits (lines 2 and 3) is shown, it can be seen from FIG. 7C that no error message has occurred.

In summary, a method and apparatus for sensor monitoring in an ESP system has been described, where the essence is the generation of residuals based on a multi-process model, the development of which is first performed taking into account driving dynamics and practical availability and applicability. Sensor errors and especially sensor errors with large gradients can be detected during operation by sensor monitoring operations. The monitoring system provides high reliability because it exhibits a high degree of robustness to relative model inaccuracies on the one hand and high sensitivity to sensor errors on the other hand.

Claims (12)

A method of monitoring a vehicle sensor that detects each reference variable or measurement variable in each case related to a driving-dynamics process, Based on the multiple process models 31 (G1 to G4; Q1 to Q4; L1 to L4) for the normal mode of operation, currently unmonitored process reference variables and process measurement variables (A, B) of the current process 32 Analytical redundancy of the currently monitored process reference variable or process measurement variable (C) from one or more of

), From the currently monitored process reference variable or process measurement variable (C), the surplus analytical redundancy provided above (

Generating a residual (r) by subtracting Evaluating the residual (r) by a residual evaluating function, Comparing the estimated residual with the estimated residual and generating an error message F when the residual reaches the threshold for at least one scheduled monitoring time Vehicle monitoring method characterized in that for monitoring the trend of the output signal of each sensor in a periodic and sequential manner. The process of claim 1, wherein the process is part of an on-vehicle electric drive stabilization program (ESP), The process reference variable and the process measurement variable (A, B, C) are yaw rate, lateral acceleration and steering angle. The method according to claim 2, wherein the method is performed in parallel with the electric driving stabilization program (ESP) and deactivates the ESP when an error message (F) is generated. 4. The process reference variable according to claim 1, wherein the multiple process model is formed of a plurality of partial models (G1 to G4; Q1 to Q4; L1 to L4), by which the respective process reference variables are monitored. Or the process measurement variable (C) is the unmonitored process reference variable or process measurement variable (A, B), wheel rotation speed (v _vl , v _vr , v _hl , v _hr ), wheel base (l), track width ( S) and other physical quantities, such as vehicle speed (v _ref ), also determined from the four wheel speeds, can be reproduced based on physical correlations. 4. The method according to any one of claims 1 to 3, wherein the residual evaluation function is a 'minimum of all' where the minimum of all residuals produced is compared to the limit value in transition process conditions distinguished by high model inaccuracy. generated by an algorithm according to the 'all' principle, and under normal process conditions distinguished by low model inaccuracy, an algorithm according to the 'majority principle' is provided, wherein an average value of analytical redundancy is formed, and the average value Analytical redundancy, which is the median of the three redundancies with a minimum deviation from the, is used to generate a residual to be compared with a threshold. 6. The vehicle sensor according to claim 5, wherein when the number of useful models is less than 3, the residuals are generated according to the 'minimum of all' principle, otherwise the residuals are generated according to the 'principle of majority'. Monitoring method. The method according to any one of claims 1 to 3, wherein on the one hand an error message is generated in a timely manner when a wrong sensor signal appears and on the other hand a false error message due to high model inaccuracies is not monitored. Calculate the limit value depending on the process conditions by process reference variables and process measurement variables (A, B), wheel rotation speeds (v _vl , v _vr , v _hl , v _hr ) and vehicle speeds (v _ref ) Vehicle sensor monitoring method comprising the step of applying. 4. A method according to any one of the preceding claims, wherein the monitoring time is adapted to the current process conditions and selected to allow short time malfunction of the sensor. An apparatus for monitoring a vehicle sensor that measures in each case an individual reference variable or measurement variable associated with a driving dynamic process, By means of multiple process models G1 to G4; Q1 to Q4; L1 to L4 for a normal mode of operation, one or more of the currently unmonitored process reference variables and process measurement variables A and B of the current process 32. From the first apparatus 31 for calculating analytical redundancy for the currently monitored process reference variable or process measurement variable C, A second device 33 for generating a residual r by subtracting the calculated surplus analytical redundancy from the currently monitored process reference variable or process measurement variable C; A third device 36 for evaluating the residual r using the residual evaluation function, A fourth device 35 for generating a threshold value, A fifth device 37 for comparing the evaluated residual r with a threshold and for generating an error signal F when the residual reaches the threshold during at least one predetermined monitoring time. Vehicle sensor monitoring device comprising a. 10. The method of claim 9, wherein at least one of the process reference variables and process measurement variables (A, B), which are not currently monitored, increases the threshold value by situation detection and is relatively low in the case of a multiple process model of relatively high inaccuracy. Vehicle sensor monitoring device, characterized in that it is sent to the fourth device (35) in order to reduce the limit value in case of a multi-process model of inaccuracy. 11. A vehicular sensor monitoring device according to claim 9 or 10, characterized in that the first to fifth devices are realized by a microprocessor system (40). In the vehicle electric stabilization program system (ESP), An electric stabilization program system for a vehicle, comprising a device according to claim 9 or 10 for periodic monitoring of yaw rate sensor (13), lateral acceleration sensor (14) and steering angle sensor (15).

Images