DE102019200600B3 - Method for predicting a fault on an actuator in a vehicle and application-specific circuit for implementing such a method - Google Patents

Abstract

The invention relates to a method for predicting a fault in an actuator (3) of a vehicle. A discrete pattern probability density distribution of an energy value is determined for at least one type of disturbance. The energy value required for each actuation of the actuator (3) is continuously determined and stored in a storage unit (4.1). An empirical discrete probability density distribution is continuously determined from the continuously stored energy values, and a similarity measure for the similarity of the empirical discrete probability density distribution with the at least one discrete pattern probability density distribution is continuously determined. A disturbance is predicted if the ascertained similarity measure exceeds a predetermined similarity measure threshold value. The invention further relates to an ASIC for implementing the method.

Description

The invention relates to a method for predicting a malfunction or failure of an electrically operated and intermittently operated actuator in a vehicle. The invention further relates to an application-specific integrated circuit (ASIC) for implementing such a method.

Actuators that are not operated continuously but intermittently are used in the vehicle, for example, to actuate a parking lock. In the case of a parking lock, a pawl engages positively in a parking lock wheel that is coupled to the vehicle wheels. The pawl and the vehicle wheels are blocked by the pawl in the locked state. Such parking locks are in the documents US 6273232 B1 , US 9657836 B2 , US 5807205 A , US 3610004 A and US 6481556 B1 described.

The pawl can be actuated via an electrically operated actuator by means of a motor and optionally a gear. The actuator for unlocking or locking the pawl is operated for a predetermined maximum period and is then not operated until the next actuation process. The sequence of actuation processes in which a predetermined actuation sequence is carried out completely within a predetermined maximum duration and between which there are pauses without actuation is referred to as intermittent actuation.

Another example of intermittently operated actuators on a vehicle are electrically operated door locks.

The electrical energy to be used for a single actuation process at a fault-free parking lock depends on the location, the inclination and the load of the vehicle and is therefore not exactly predictable.

The document US 20020084162 A1 describes an electric parking lock with an emergency release using a Bowden cable. The document W02010104189A1 describes a state machine in which the speed of rotation of a drive gear is evaluated for the correct control of a parking lock. The document US 9233671 B2 describes an electrical parking lock with a redundant blocking mechanism that is automatically activated when a detected movement of the vehicle. The document WO 2017182207 A1 describes a hydraulically operated parking lock comprising a diagnostic unit for a parking brake. Further embodiments of parking locks are in the documents EP 1103744 A2 and EP 2000708 A3 described.

Methods for predicting a fault on an electrically powered actuator from the prior art are out US 9 205 560 B1 , US 2010/0 161 274 A1 and US 2006/0 142 976 A1 known.

The invention is based on the object of specifying a method with which an improved prediction of the failure of an intermittently actuated actuator in a vehicle is possible. The invention is also based on the object of specifying an ASIC for implementing such a method.

With regard to the method, the object is achieved according to the invention by the features of claim 1. With regard to the ASIC, the object is achieved according to the invention by the features of claim 10.

Advantageous embodiments of the invention are the subject of the dependent claims.

In a method for predicting a malfunction on an intermittently actuated electrically operated actuator of a vehicle, which may be affected by a malfunction of at least one type of malfunction, a discrete pattern probability density distribution of the electrical energy value required for a single actuation of the malfunctioning actuator is determined for the at least one malfunction type. This energy value is continuously determined for continuous actuations of the actuator and stored in a storage unit.

An empirical discrete probability density distribution is continuously determined from the continuously stored energy values. An empirical discrete probability distribution can be determined, for example, by quantizing the energy values and by forming a histogram of energy values, which is normalized to the total number of energy values recorded in the histogram.

A similarity measure is continuously determined for the similarity of the empirical discrete probability density distribution with the at least one discrete pattern probability density distribution.

A disturbance is predicted or predicted if the continuously determined similarity measure exceeds a predetermined similarity measure threshold value once or for a certain number of successive actuation processes.

As an alternative or in addition, a gradient can be determined for the particular degree of similarity and a disturbance can be predicted or predicted if the continuously determined degree of similarity exceeds a predetermined similarity measure threshold value once or for a certain number of successive actuation processes or if the gradient exceeds a predetermined number Gradient threshold exceeds.

An advantage of the method is that a malfunction of the actuator can already be recognized in the process of forming the malfunction. This allows the driver to be informed of the risk of such a malfunction and take countermeasures before the malfunction occurs. For example, the driver can be asked to go to a service workshop. In one embodiment of the method, a predicted fault can also be transmitted directly to a service facility, for example via a mobile radio network.

In addition, the early detection of a fault using the method according to the invention can prevent further damage and thus reduce repair costs. It is also possible to avoid any additional costs that may occur, for example for towing a vehicle that has broken down, in the event of a manifest occurrence of the fault.

In one embodiment of the method, a correlation value between the empirical probability density distribution and the at least one pattern probability density distribution is determined as a measure of similarity between the empirical discrete probability density distribution and the at least one discrete pattern probability density distribution. An advantage of this embodiment is that a correlation value is a very easily calculable similarity measure, which is robust against random deviations between the empirical probability density distribution and the at least one pattern probability density distribution. This enables a reliable prediction of a fault even under different operating conditions and when there are fluctuations in the number of actuators.

In one embodiment of the method, a negated Euclidean distance between the empirical probability density distribution and the at least one pattern probability density distribution is determined as a measure of similarity between the empirical discrete probability density distribution and the at least one discrete pattern probability density distribution. An advantage of this embodiment is that a negated Euclidean distance is a very easily calculable similarity measure that is robust against random deviations between the empirical probability density distribution and the at least one pattern probability density distribution. This enables a reliable prediction of a fault even under different operating conditions and when there are fluctuations in the number of actuators.

In one embodiment of the method, a negated Kullback-Leibler divergence between the empirical probability density distribution and the at least one pattern probability density distribution is determined as a measure of similarity between the empirical discrete probability density distribution and the at least one discrete pattern probability density distribution. An advantage of this embodiment is that a negated Kullback-Leibler divergence is a very easily calculable measure of similarity that is robust against random deviations between the empirical probability density distribution and the at least one pattern probability density distribution. This enables a reliable prediction of a fault even under different operating conditions and when there are fluctuations in the number of actuators.

In one embodiment of the method, the empirical discrete probability density distribution is determined from a predetermined number of previously determined energy values. For example, it is possible to use the energy values determined for the previous M = 50 actuation processes of the actuator for the determination of the histogram and, derived from this by standardization to the number of M = 50 actuation processes used, for the determination of the empirical discrete probability density distribution.

An advantage of this embodiment is that the prediction of a newly developing disturbance is also possible if a large number of correctly executed actuation processes occurred, which would otherwise make it difficult or impossible to identify a similarity to the at least one pattern probability density distribution.

In one embodiment of the method, a predicted fault is signaled by means of a display instrument and / or stored in a data memory accessible for on-board diagnosis (OBD) and / or transmitted to a mobile terminal and / or by means of a mobile radio network to a service center. A service center can for example operated by a vehicle manufacturer or by a workshop or by their contractual partners. This advantageously enables the fault to be remedied with little repair work and without impairing the mobility of the vehicle.

In one embodiment of the method, the energy value required in each case for the actuation of a locking device of an electric parking lock by means of the actuator is continuously determined and stored in the storage unit. With this embodiment, it is possible to detect malfunctions or failures of an electric parking lock, which gradually develop over a certain period of time. Such failures can be caused by electrical faults such as loose contacts, leakage currents or short circuits, by liquid or solid contaminants that gradually enter the electric parking lock, by defects or signs of wear on gear wheels of a gearbox for actuating the actuator, by crooked or bent shafts for the storage of such gear wheels , caused by a lateral offset of the end faces of such gearwheels along their supporting shafts or by a failure or fatigue of return springs for disengaging a pawl in an electric parking lock. Such failures can also be caused by increased friction, wear or vibrations or by an increased torque. Damage to the electric parking lock is also possible in that the input-side speed is incorrectly determined on a drive gear and the electric parking lock is inserted into the still rotating locking wheel connected to the drive gear.

Common to these and other malfunctions that can be predicted with the embodiment of the method is that they do not lead to an immediate complete failure of the electric parking lock, but rather that individual actuation processes fail gradually as the sequence increases. Initially, failed operations can be corrected by simply repeating them.

With the embodiment, it is therefore possible to inform a driver of a vehicle with a malfunctioning electrical parking lock that is developing in good time that he can go to a workshop. As a result, more complex repairs and / or additional outlay due to inoperability of the vehicle can be avoided. In addition, operational safety is improved since jamming of the electric parking lock in the unlocking position is avoided. Damage to other components, for example damage to a transmission in the actuator, which can be triggered in the event of incomplete unlocking or locking of the electric parking lock caused by a fault, can also be avoided.

In one embodiment, at least one pattern recognition probability density function is added after the electrical parking lock has been started up. A supplemented pattern recognition probability density function can be found, for example, by evaluating error and fault conditions on an electric parking lock depending on various conditions of use. This makes it possible to continuously improve the robustness and predictive accuracy of the method.

Another advantage is that the method is easy to implement. In particular, the method is easier to implement than methods which are based on a classification using fuzzy logic and / or on a frequency analysis and / or on the use of complex mathematical models.

In particular, the method can be implemented in an Application Specific Integrated Circuit (ASIC) which is already available for the control and diagnosis of the drive stage and the actuator of an electrical parking lock. This eliminates costs for additional hardware components.

In one embodiment of the method, the energy value required in each case for the actuation of a locking device of an electric parking lock is determined by continuously scanning the current through and the voltage on a motor of the actuator over the duration of the actuation. The current through and the voltage on such a motor can be measured easily and reliably.

In further embodiments, the method for diagnosing and / or predicting malfunctions in other electrically or hydraulically operated actuators can be used. For example, the method for predicting failures of critical systems in a vehicle can be used, for example in a lane assistant, in a steering assistant or in brakes.

In one embodiment of the method, in the event of a prediction of a fault, the parking lock is locked and unlocked alternately, the motor being operated with the maximum permissible load current. Certain faults, such as deposits adhering to the pawl, can be eliminated.

Exemplary embodiments of the invention are explained in more detail below with reference to drawings.

• 1 schematisch eine Verriegelungsvorrichtung für eine elektrische Parksperre nach dem Stand der Technik,
• 2 einen Ablaufplan für ein Verfahren zur Vorhersage einer Verklemmung,
• 3 schematisch eine Musterwahrscheinlichkeitsdichteverteilung für einen ersten Verklemmungstyp,
• 4 schematisch eine Musterwahrscheinlichkeitsdichteverteilung für einen zweiten Verklemmungstyp,
• 5 schematisch eine Musterwahrscheinlichkeitsdichteverteilung für einen dritten Verklemmungstyp,
• 6 schematisch eine Musterwahrscheinlichkeitsdichteverteilung für einen vierten Verklemmungstyp sowie
• 7 schematisch eine empirisch geschätzte Wahrscheinlichkeitsdichteverteilung.

In it show:

• 1 schematically shows a locking device for an electric parking lock according to the prior art,
• 2nd a flow chart for a method for predicting a deadlock,
• 3rd schematically a pattern probability density distribution for a first deadlock type,
• 4th schematically a pattern probability density distribution for a second jamming type,
• 5 schematically a pattern probability density distribution for a third jamming type,
• 6 schematically a pattern probability density distribution for a fourth type of jamming as well
• 7 schematically an empirically estimated probability density distribution.

Corresponding parts are provided with the same reference symbols in all figures.

1 shows schematically the structure of a locking device for an electric parking lock according to the prior art. The locking device comprises a ratchet wheel 1 , which is arranged on a drive shaft of a transmission, not shown, in a rotationally fixed manner. The ratchet wheel has on its circumference 1 radial grooves 1.1 on that for the engagement of a pawl 2nd are provided.

Using an actuator 3rd is the pawl 2nd movable between a locking position and an unlocking position. In the in 1 locking position shown engages to match the radial grooves 1.1 pawl shaped end 2nd in a groove 1.1 the ratchet wheel 1 and blocks the ratchet wheel 1 against a rotation. As a result, the vehicle is secured against rolling away.

The pawl is in the unlocked position 2nd out of the groove 1.1 disengaged. The vehicle can roll freely in the unlocked position. The electric parking lock is only activated when the vehicle stops.

The actuator 3rd includes an engine 3.1 and a gear 3.2 that the rotary motion of the motor 3.1 suitable in the release movement of the pawl 2nd translated. The motor 3.1 can be designed as a DC motor with brushes or as a brushless DC motor or as a linear motor. The motor 3.1 can also be replaced by any electromechanical drive that moves the pawl 2nd is formed between the locking position and the unlocking position and the optional further mechanical auxiliary devices, for example a release spring for disengaging the pawl 2nd from a groove 1.1 , having.

The rotation of the engine 3.1 is controlled by a control unit 4th controlled. The control unit 4th comprises a storage unit 4.1 , which is advantageously designed as a first-in-first-out (FIFO) memory. The control unit 4th is for measuring or estimating an electrical current through the motor 3.1 and for measuring or estimating an electrical voltage across the motor 3.1 set up.

2nd schematically shows a flow chart for a method for predicting a jam when releasing an electrical parking lock. The process begins at a starting point S0 . In a first decision step E1 it is checked whether the parking lock can be released. This can be done by checking the signal level of a signal to trigger an unlocking process. Alternatively, the test can be done by running the current through the motor 3.1 and / or a voltage across the motor 3.1 is measured and a request to release the parking lock is recognized when at least one of these measured values lies above a predetermined limit value in each case.

If the parking lock is to be released, the procedure goes to the first step S1 continued. If the parking lock cannot be released, the procedure continues with the second decision step E2 continued.

In the first step S1 , which is executed at the discrete time or time index k, the current voltage U [k] and the current I [k] for the operation of the actuator 3rd when disengaging the pawl 2nd recorded and from this the current electrical power P [k] = U [k] · I [k] is determined. This becomes the cumulative for actuating the actuator 3rd energy E [k] = E [k-1] + P [k] · T _S determined, where T _{S is} the time interval between two successive runs of the first step S1 is.

After the first step S1 the time index k is incremented and the first decision step E1 run through again. In one embodiment, the method can be set up such that the decision step E1 is repeatedly executed until a predetermined actuation time for actuator actuation is reached. The actuation time can be 1.1 seconds, for example.

In the second decision step E2 it is checked whether the calculation of the energy for releasing the parking lock has been completed. If the calculation of the energy for releasing the parking lock the procedure continues with the second step S2 continued. If the calculation of the energy for releasing the parking lock has not been completed, the procedure continues with the first decision step E1 continued.

At the second step S2 becomes the calculated energy value E _d for releasing the parking lock in the storage unit 4.1 filed. After the second step S2 becomes the third step S3 executed.

geschätzt, indem das Histogramm normalisiert wird: $$\widetilde{P_d}\left(E_d\right)=\frac{H_{d,i}\left(E_d\right)}{{\displaystyle {\sum }_{i=1}^{N_b}{H}_{d,i}\left(E_d\right)}}$$

In the third step S3 is a histogram H _{d, i} (E _d ) EN ^{1 × N b} , i = 1 ... N _b over all stored energy values E _d determined, where N _{b is} the number of bins or quantization levels into which the energy values E _d to be grouped. The N _b quantization levels are between 0 and an upper energy limit E _piece , the energy required to release a jammed parking lock. Below is the third step S3 a discrete probability density distribution $\widetilde{P_d}\left(E_d\right)$

estimated by normalizing the histogram: $$\widetilde{P_d}\left(E_d\right)=\frac{H_{d,i}\left(E_d\right)}{{\displaystyle {\sum }_{i=1}^{N_b}{H}_{d,i}\left(E_d\right)}}$$

bestimmt: $$\overline{\widetilde{P_d}}=\frac{1}{N_b}{\displaystyle \sum _{i=1}^{N_b}{P}_{d,i}\left(E_d\right)}$$

Below is the third step S3 the mean probability density value $\overline{\widetilde{P_d}}$

certainly: $$\overline{\widetilde{P_d}}=\frac{1}{N_b}{\displaystyle \sum _{i=1}^{N_b}{P}_{d,i}\left(E_d\right)}$$

und einer Musterwahrscheinlichkeitsdichteverteilung P_X,i(E_d) bestimmt. Eine Musterwahrscheinlichkeitsdichteverteilung P_X,i(E_d) kann für jeweils einen von verschiedenen Verklemmungstypen X={D1, D2, ... } der Parksperre gegeben sein.Below is the third step S3 the normalized correlation between the estimated probability density distribution $\widetilde{P_{d,l}}\left(E_d\right)$

and a pattern probability density distribution P _{X, i} (E _d ) certainly. A pattern probability density distribution P _{X, i} (E _d ) can be given for one of different types of jamming X = {D1, D2, ...} of the parking lock.

bildet sich nach einer gewissen hinreichenden Anzahl von Entriegelungsvorgängen, beispielsweise nach M = 25 Entriegelungsvorgängen, eine typische Form aus, welche vom Zustand der Verriegelungsvorrichtung, aber auch von der Art und Weise beeinflusst wird, in der das Fahrzeug typischerweise geparkt wird.For the estimated probability density distribution $\widetilde{P_{d,l}}\left(E_d\right)$

After a certain sufficient number of unlocking processes, for example after M = 25 unlocking processes, a typical form is formed which is influenced by the state of the locking device but also by the manner in which the vehicle is typically parked.

dann einen Häufungspunkt nahe einer unteren Energie E_d,min aufweisen, wenn das Fahrzeug regelmäßig auf ebener Straße geparkt wird. In diesem Fall ist die Reibung zwischen der Sperrklinke 2 und der Nut 1.1 des Sperrrads 1 minimal. Somit ist nur eine geringe Energie zur Entriegelung erforderlich.If the locking device functions correctly, the estimated probability density distribution becomes $\widetilde{P_{d,l}}\left(E_d\right)$

then a cluster point near a lower energy E _{d, min} if the vehicle is regularly parked on a level street. In this case there is friction between the pawl 2nd and the groove 1.1 the ratchet wheel 1 minimal. This means that only a small amount of energy is required for unlocking.

dann einen Häufungspunkt nahe einer oberen Energie E_d,max aufweisen, wenn das Fahrzeug regelmäßig auf abschüssiger Straße geparkt wird. In diesem Fall ist die Reibung zwischen der Sperrklinke 2 und der Nut 1.1 des Sperrrads 1, abhängig vom Gefälle der Fahrbahn und abhängig vom Gewicht des Fahrzeugs, erhöht, da das Sperrrad 1 ein aus der Hangabtriebskraft des Fahrzeugs resultierendes höheres Drehmoment überträgt. Somit ist eine erhöhte Energie zur Entriegelung der Sperrklinke 2 erforderlich.If the locking device functions correctly, the estimated probability density distribution becomes $\widetilde{P_{d,l}}\left(E_d\right)$

then a cluster point near an upper energy E _{d, max} if the vehicle is regularly parked on a sloping road. In this case there is friction between the pawl 2nd and the groove 1.1 the ratchet wheel 1 , depending on the gradient of the road and depending on the weight of the vehicle, because the ratchet wheel 1 transmits a higher torque resulting from the slope downforce of the vehicle. Thus there is an increased energy to unlock the pawl 2nd required.

heraus, die zwischen der unteren Energie E_d,min und der oberen Energie E_d,max näherungsweise gleichverteilt ist.If the vehicle is parked with the locking device functioning correctly both in a level and in an inclined or sloping position, an estimated probability density distribution is formed $\widetilde{P_{d,l}}\left(E_d\right)$

out that between the lower energy E _{d, min} and the upper energy E _{d, max} is approximately equally distributed.

Abweichungen aufweisen, die für einen bestimmten Typ von Störung, im Folgenden als Verklemmungstyp bezeichnet, charakteristisch sein können.With certain disturbances, the estimated probability density distribution can $\widetilde{P_{d,l}}\left(E_d\right)$

Have deviations that can be characteristic of a certain type of disturbance, hereinafter referred to as jamming type.

und jeder der Musterwahrscheinlichkeitsverteilungen wird gemäß der Vorschrift $${\rho }_{d,X}=\frac{\left({\displaystyle {\sum }_{i=1}^{N_b}\left(\overline{\widetilde{P_d}}-\widetilde{P_{d,l}}\left(E_d\right)\right)}\cdot (\overline{P_X}-P_{X,i}\left(E_d\right)\right)}{\widetilde{{\sigma }_d}\cdot {\sigma }_X}$$

ermittelt, wobei $$\widetilde{{\sigma }_d}=\sqrt{\frac{1}{N_b}{\displaystyle \sum _{i=1}^{N_b}{\left(\overline{\widetilde{P_d}}-\widetilde{P_{d,l}}\left(E_d\right)\right)}^2}}$$

die über die geschätzte Wahrscheinlichkeitsdichteverteilung $\widetilde{P_d}\left(E_d\right)$

Standardabweichung, $$\begin{array}{cc}\overline{P_X}=\frac{1}{N_b}{\displaystyle \sum _{i=1}^{N_b}{P}_{X,i}}\left(E_d\right),& X\in \left\{D\mathrm{1,}D\mathrm{2,}\dots \right\}\end{array}$$

der Mittelwert einer diskreten Musterwahrscheinlichkeitsdichteverteilung für den Verklemmungstyp X sowie $$\begin{array}{cc}{\sigma }_X=\sqrt{\frac{1}{N_b}{\displaystyle \sum _{i=1}^{N_b}{\left(\overline{P_X}-P_{X,i}\left(E_d\right)\right)}^2}},& X\in \left\{D\mathrm{1,}D\mathrm{2,}\dots \right\}\end{array}$$

die Standardabweichung einer diskreten Musterwahrscheinlichkeitsdichteverteilung für den Verklemmungstyp X ist.Examples of deadlock types D1 , D2 , D3 and D4 when releasing the parking lock are subsequently based on the 3rd to 6 explained in more detail. Each of these deadlock types is by an associated discrete pattern probability density distribution P _{D1, i} (E _d ) , P _{D2, i} (E _d ) , P _{D3, i} (E _d ) , P _{D4, i} (E _d ) described. A correlation ρ _{d, X} , X ∈ {D1, D2, ...} between the estimated probability density distribution $\widetilde{P_{d,l}}\left(E_d\right)$

and each of the pattern probability distributions is made according to the regulation $${\rho }_{d,X}=\frac{\left({\displaystyle {\sum }_{i=1}^{N_b}\left(\overline{\widetilde{P_d}}-\widetilde{P_{d,l}}\left(E_d\right)\right)}\cdot (\overline{P_X}-P_{X,i}\left(E_d\right)\right)}{\widetilde{{\sigma }_d}\cdot {\sigma }_X}$$

determined where $$\widetilde{{\sigma }_d}=\sqrt{\frac{1}{N_b}{\displaystyle \sum _{i=1}^{N_b}{\left(\overline{\widetilde{P_d}}-\widetilde{P_{d,l}}\left(E_d\right)\right)}^{\mathrm{2nd}}}}$$

over the estimated probability density distribution $\widetilde{P_d}\left(E_d\right)$

Standard deviation, $$\begin{array}{cc}\overline{P_X}=\frac{1}{N_b}{\displaystyle \sum _{i=1}^{N_b}{P}_{X,i}}\left(E_d\right),& X\in \left\{\mathrm{<figure-callout\; id="1"\; label="D"\; filenames="DE102019200600B3\_0026.png"\; state="\{\{state\}\}">D</figure-callout>}\mathrm{1,}\mathrm{<figure-callout\; id="2"\; label="D"\; filenames="DE102019200600B3\_0026.png"\; state="\{\{state\}\}">D</figure-callout>}\mathrm{2,}...\right\}\end{array}$$

the mean value of a discrete pattern probability density distribution for the deadlock type X and $$\begin{array}{cc}{\sigma }_X=\sqrt{\frac{1}{N_b}{\displaystyle \sum _{i=1}^{N_b}{\left(\overline{P_X}-P_{X,i}\left(E_d\right)\right)}^{\mathrm{2nd}}}},& X\in \left\{\mathrm{<figure-callout\; id="1"\; label="D"\; filenames="DE102019200600B3\_0026.png"\; state="\{\{state\}\}">D</figure-callout>}\mathrm{1,}\mathrm{<figure-callout\; id="2"\; label="D"\; filenames="DE102019200600B3\_0026.png"\; state="\{\{state\}\}">D</figure-callout>}\mathrm{2,}...\right\}\end{array}$$

is the standard deviation of a discrete pattern probability density distribution for deadlock type X.

und einer Musterwahrscheinlichkeitsdichteverteilung P_X,i(E_d) für den Verklemmungstyp X.The correlation ρ _{d, X} , XE {D1, D2, ...} is therefore a measure of the similarity of the empirically determined probability density distribution $\widetilde{P_{d,l}}\left(E_d\right)$

and a pattern probability density distribution P _{X, i} (E _d ) for jamming type X.

und einer Musterwahrscheinlichkeitsdichteverteilung P_X,i(E_d) für den Verklemmungstyp X verwendet werden. Beispielsweise kann eine negative euklidische Distanz $$-\Vert \widetilde{P_{d,l}}\left(E_d\right)-P_{X,i}\left(E_d\right)\Vert =-{{\displaystyle \sum _{i=1}^{N_b}\left(\widetilde{P_{d,l}}\left(E_d\right)-P_{X,i}\left(E_d\right)\right)}}^2$$

als Ähnlichkeitsmaß verwendet werden. Alternativ kann eine negierte Kullback-Leibler Divergenz $$-D\left(\widetilde{P_{d,l}}\left(E_d\right)\Vert P_{X,i}\left(E_d\right)\right)=-{\displaystyle \sum _{n=1}^{N_b}\widetilde{P_{d,l}}\left(E_d\right)\cdot \text{log}\frac{\widetilde{P_{d,l}}\left(E_d\right)}{P_{X,i}\left(E_d\right)}}$$

als Ähnlichkeitsmaß verwendet werden.In embodiments of the method, instead of the correlation ρ _{d, X} , X ∈ {D1, D2, ...}, other measures of similarity between the empirically determined probability density distribution $\widetilde{P_{d,l}}\left(E_d\right)$

and a pattern probability density distribution P _{X, i} (E _d ) can be used for the jamming type X. For example, a negative Euclidean distance $$-\Vert \widetilde{P_{d,l}}\left(E_d\right)-P_{X,i}\left(E_d\right)\Vert =-{{\displaystyle \sum _{i=1}^{N_b}\left(\widetilde{P_{d,l}}\left(E_d\right)-P_{X,i}\left(E_d\right)\right)}}^{\mathrm{2nd}}$$

can be used as a measure of similarity. Alternatively, a negated Kullback-Leibler divergence $$-D\left(\widetilde{P_{d,l}}\left(E_d\right)\Vert P_{X,i}\left(E_d\right)\right)=-{\displaystyle \sum _{n=1}^{N_b}\widetilde{P_{d,l}}\left(E_d\right)\cdot \text{log}\frac{\widetilde{P_{d,l}}\left(E_d\right)}{P_{X,i}\left(E_d\right)}}$$

can be used as a measure of similarity.

On the third step S3 the third decision step follows E3 , in which it is checked for each deadlock type X whether the assigned determined correlation value ρ _{d, X} , X ∈ {D1, D2, ...} is above a predetermined correlation threshold value. If the predetermined correlation threshold value has not been exceeded, in a fourth step S4 the calculated energy value E _d reset to zero and the first decision step E1 run through again. If the predetermined correlation threshold is exceeded, in a fifth step S5 the assigned deadlock type X predicts. With one on the fourth step S4 following endpoint S6 the procedure ends.

eine endliche Folge von M Energiewerten E_d verwendet wird, wobei jeder der M Energiewerte in einer Folge von Durchläufen des ersten Entscheidungsschritts E1 und des ersten Schritts S1 ermittelt wird. Somit wird, mit jedem neu ermittelten Energiewert E_d, ein neues Histogramm H_d,i(E_d) und nachfolgend eine neue empirische Wahrscheinlichkeitsdichteverteilung $\widetilde{P_{d,l}}\left(E_d\right)$

bestimmt, die dann jeweils auch zu neuen Korrelationswerten ρ_d,X,X ∈ {D1, D2, ...} führt. Somit ist aus den Änderungen eines Korrelationswerts jeweils ein Gradient Δρ_d,X, X ∈ {D1, D2, ...} bestimmbar, der einen Trend für die Entwicklung einer Ähnlichkeit zwischen der empirisch bestimmten Wahrscheinlichkeitsdichteverteilung $\widetilde{P_{d,l}}\left(E_d\right)$

und einer Musterwahrscheinlichkeitsdichteverteilung P_X,i(E_d) für den Verklemmungstyp X angibt. Ein Verklemmungstyp X kann prädiziert werden, wenn der zugeordnete Gradient Δρ_d,X,X ∈ {D1,D2, ...} des Korrelationskoeffizienten einen vorbestimmten Gradientenschwellwert überschreitet oder/und der Korrelationskoeffizient einen vorbestimmten Korrelationsschwellwert überschreitet.In addition or as an alternative to the third step S3 a time gradient can be determined for each correlation value ρ _{d, X} , X ∈ {D1, D2, ...}. Such a temporal gradient can be determined, for example, by determining the empirical probability density distribution $\widetilde{P_{d,l}}\left(E_d\right)$

a finite sequence of M energy values E _d is used, each of the M energy values in a series of runs of the first decision step E1 and the first step S1 is determined. Thus, with each newly determined energy value E _d , a new histogram H _{d, i} (E _d ) and subsequently a new empirical probability density distribution $\widetilde{P_{d,l}}\left(E_d\right)$

determined, which then also leads to new correlation values ρ _{d, X} , X ∈ {D1, D2, ...}. Thus, from the changes in a correlation value _, a gradient Δρ _{d, X} , X, {D1, D2, ...} can be determined, which is a trend for developing a similarity between the empirically determined probability density distribution $\widetilde{P_{d,l}}\left(E_d\right)$

and a pattern probability density distribution P _{X, i} (E _d ) for the deadlock type X indicates. A deadlock type X can be predicted if the associated gradient Δρ _{d, X} , X ∈ {D1, D2, ...} of the correlation coefficient exceeds a predetermined gradient threshold value and / or the correlation coefficient exceeds a predetermined correlation threshold value.

The prediction of a jamming type X or a malfunction of the locking device that can be expected with a certain probability can be displayed, for example, by a display on a multifunction instrument of the vehicle. Such a prediction can also be made accessible by an on-board diagnosis (OBD) tool. It is also possible to transmit such prediction information to a driver's mobile device, for example to a tablet personal computer or to a smartphone.

It is also possible to use methods for correcting a malfunction of the locking device. For example, locking and unlocking processes with maximum actuation current at the actuator can be carried out in short succession 3rd the parking lock can be triggered.

3rd shows schematically a pattern probability density distribution P _{D1, i} (E _d ) for a first type of jam D1 , which occurs, for example, when the coefficient of friction increases due to solid or liquid impurities in the parking lock. Such contaminants gradually enter the parking lock and are also deposited in the locking device. As a result, the locking device is sluggish and it requires an increasingly larger electrical energy E _d to unlock the parking lock.

3rd shows the pattern probability density distribution P _{D1, i} (E _d ) for N _b = 25 bins, which are evenly distributed on the horizontal axis for energy values between E _d = 0 and E _d = E _stuck = 400 joules, for a first type of jamming D1 .

The maximum possible energy value E _stuck = 400 joules, which is required when the parking lock is jammed, is from the electrical specification of the actuator 3rd and the maximum allowable actuation period. E _piece can be the product of the maximum permissible load current, the maximum permissible supply voltage and the maximum permissible actuation period for the actuator 3rd be determined.

The normal range [E _{d, min} , E _{d, max} ] = [200 joules, 300 joules] for the energy to be released when the locking device is functioning correctly to unlock the parking lock is also well known. It is E _{d, min} the minimum energy required to release the parking lock when the locking device is functioning according to the specification, for example in a horizontally standing vehicle E _{d, max} the maximum energy to be applied with the locking device functioning according to the specification for unlocking the parking lock, for example in the case of a maximally loaded vehicle which is parked on a steep road with a maximum permissible inclination. E _{d, max} is significantly below the maximum possible energy value E _stuck = 400 Joule.

How out 3rd recognizable, values occur above the normal range [ E _{d, min} , E _{d, max} ] of unlocking energy values. So the deadlock type D1 be connected with a mechanical stiffness of the unlocking device. But it is also possible that increased values of the electrical unlocking energy E _d due to sporadic short circuits in the power electronics to control the motor 3.1 caused that occur with increasing frequency and can turn into permanent short circuits. In the event of a permanent short circuit in the power electronics, it would no longer be possible to unlock the locking device.

4th shows the pattern probability density distribution P _{D2, i} (E _d ) for N _b = 25 bins, which are evenly distributed on the horizontal axis for energy values between E _d = 0 and E _d = E _stuck = 400 joules, for a second type of jamming D2 . This pattern probability density distribution P _{D2, i} (E _d ) differs from that in 3rd shown pattern probability density distribution P _{D1, i} (E _d ) for the first type of jam D1 by the fact that low energy values E _d below the lower limit E _{d, min} of the normal range of energy values. The pattern probability density distribution P _{D2, i} (E _d ) of the deadlock type D2 goes for example with considerable damage to the teeth of the transmission 3.2 which is a free running of the engine 3.1 without disengaging the pawl 2nd can effect. As a result of this damage, the pawl 2nd not out of the groove 1.1 disengaged and thus the unlocking process was not successfully completed in the designated time. By at least partially freewheeling the engine 3.1 only a small amount of electrical energy is used for the unsuccessful unlocking process.

The pattern probability density distribution P _{D2, i} (E _d ) of the deadlock type D2 However, it can also be accompanied by electrical interference, for example with a loose contact or an insufficient electrical connection on the motor 3.1 or on the control unit 4th .

5 shows the pattern probability density distribution P _{D3, i} (E _d ) for N _b = 25 bins, which are evenly distributed on the horizontal axis for energy values between E _d = 0 and E _d = E _stuck = 400 joules, for a third type of jamming D3 . This pattern probability density distribution P _{D3, i} (E _d ) is by a cluster point between the upper limit of the normal range E _{d, max} and the maximum possible energy value E _stuck = 400 joules, which in the present case is approximately 330 joules.

Such a pattern probability density distribution P _{D3, i} (E _d ) of the deadlock type D3 goes hand in hand, for example, with increased stiffness of the unlocking device, caused by increased friction or by torsional stress. Such torsional stress can be caused, for example, by the shaft of at least one gear wheel of the transmission 3.2 is not arranged perpendicular to this gear or misaligned in any other way.

6 shows the pattern probability density distribution P _{D4, i} (E _d ) for N _b = 25 bins, which are evenly distributed on the horizontal axis for energy values between E _d = 0 and E _d = E _stuck = 400 joules, for a fourth type of jamming D4 . This pattern probability density distribution P _{D4, i} (E _d ) is characterized by a first cluster point at energy values close to zero and by a second cluster point at energy values close to the maximum possible energy value E _stuck = 400 joules.

Such a pattern probability density distribution P _{D4, i} (E _d ) of the deadlock type D4 can be associated with a combination of various failures or defects on a locking device. For example, a contact fault in some unlocking processes can result in a very low energy value E _d lead to contamination or bending of a gear shaft in the transmission during other unlocking processes 3.2 to a very high energy value E _d can lead. If such disturbances alternate continuously, the schematically represented pattern probability density distribution is created P _{D4, i} (E _d ) of the deadlock type D4 .

sowie Musterwahrscheinlichkeitsdichteverteilung P_X,i(E_d),X ∈ {D1, D2, D3, D4} gemäß den 3 bis 6. Die empirisch geschätzte Wahrscheinlichkeitsdichteverteilung $\widetilde{P_{d,l}}\left(E_d\right)$

weist einen Häufungspunkt nahe der oberen Grenze E_d,max des Normalbereichs der Energiewerte auf. Der Korrelationswert ρ_d,D3 bezogen auf die Musterwahrscheinlichkeitsdichteverteilung P_D3,i(E_d) für den dritten Verklemmungstyp D3 überschreitet den vorbestimmten Korrelationsschwellwert. Mit dem erfindungsgemäßen Verfahren wird ein bevorstehender Ausfall der Verriegelungsvorrichtung erkannt und angezeigt. Ferner können zusätzliche Hinweise zur Charakteristik des Verklemmungstyps 3 angezeigt werden. 7 shows an empirically estimated probability density distribution superimposed $\widetilde{P_{d,l}}\left(E_d\right)$

and pattern probability density distribution P _{X, i} (E _d ), X ∈ {D1, D2, D3, D4} according to the 3rd to 6 . The empirically estimated probability density distribution $\widetilde{P_{d,l}}\left(E_d\right)$

has a cluster point near the upper limit E _{d, max} the normal range of energy values. The correlation value ρ _{d, D3} related to the pattern probability density distribution P _{D3, i} (E _d ) for the third type of jam D3 exceeds the predetermined correlation threshold. With the method according to the invention, an impending failure of the locking device is recognized and displayed. Furthermore, additional information on the characteristics of the deadlock type can be found 3rd are displayed.

That in the 2nd to 7 The method described by way of example for the detection of a jam during an unlocking process can be transferred to the detection of a jam during a locking process. This is based on other pattern probability density distributions P _{X, i} (E _e ), X ∈ {E1, E2, ...}, which are the distribution of energy values E _e in a locking process in the event of a locking fault E1 , E2 , ... describe. The energy values measured in each case in a single locking process are also advantageous E _e in a storage unit 4.1 filed by the storage unit 4.1 for the storage of energy values E _d is independent in the unlocking processes.

A separation of the prediction of jams during an unlocking process from a prediction of jams during a locking process is advantageous because the typical energy values required for an undisturbed electrical parking lock for unlocking are higher than the required energy values for a locking device and for the loading and inclination of the Depend on the vehicle.

Reference symbol list

1

ratchet wheel

1.1

Groove

2nd

Pawl

3rd

Actuator

3.1

engine

3.2

transmission

4th

Control unit

4.1

Storage unit

E1 to E3

first to third decision steps

S0

Starting point

S1 to S5

first to fifth step

S6

Endpoint

Claims (10)

Method for predicting a malfunction on an intermittently actuated electrically powered actuator (3) of a vehicle, characterized in that - for at least one type of malfunction, a discrete pattern probability density distribution of the energy value required for actuating the malfunctioning actuator (3) is determined, - for each actuation of the actuator (3) required energy value is continuously determined and stored in a storage unit (4.1), - an empirical discrete probability density distribution is continuously determined from the continuously stored energy values, - continuously a similarity measure for the similarity of the empirical discrete probability density distribution with the at least one discrete pattern probability density distribution is determined and - a disturbance is predicted if the determined similarity measure exceeds a predetermined similarity measure threshold. Procedure according to Claim 1 , characterized in that a correlation value between the empirical probability density distribution and the at least one pattern probability density distribution is determined as a similarity measure. Procedure according to Claim 1 , characterized in that a negated Euclidean distance between the empirical probability density distribution and the at least one pattern probability density distribution is determined as a similarity measure. Procedure according to Claim 1 , characterized in that a negated Kullback-Leibler divergence between the empirical probability density distribution and the at least one pattern probability density distribution is determined as a similarity measure. Method according to one of the preceding claims, characterized in that the empirical discrete probability distribution is determined from a predetermined number of previously determined energy values. Method according to one of the preceding claims, characterized in that a predicted fault is signaled by means of a display instrument and / or stored in a data memory accessible for an on-board diagnosis (OBD) instrument and / or on a mobile terminal and / or by means of a Cellular network is transmitted to a service center. Method according to one of the preceding claims, characterized in that the energy value required in each case for actuating a locking device of an electrical parking lock by means of the actuator (3) is continuously determined and stored in the storage unit (4.1). Procedure according to Claim 7 , characterized in that the energy value required in each case for the actuation of a locking device of an electric parking lock is determined by continuously scanning the current and the voltage on a motor (3.1) of the actuator (3) over the duration of the actuation. Procedure according to one of the Claims 7 to 8th , characterized in that if a fault is predicted, the parking lock is locked and unlocked alternately, the motor (3.1) being operated with the maximum permissible load current. Application-specific circuit (Application Specific Integrated Circuit, ASIC), which is set up to control and / or diagnose a drive train and / or to control an electric parking lock in a vehicle, characterized in that a method according to one of the application-specific circuits Claims 7 to 9 is implemented.

Images