EP1431927A1 - Method for estimating the remaining lifetime of an apparatus - Google Patents

Abstract

A relationship is determined between number of flights (A), temperature increase (T) and wear (V), which represents operating lifetime of the machine under test. A limit value (G) is set for the temperature parameter (T) which corresponds to the maximum permissible wear. A characteristic field (KF) is created which identifies a relationship between the temperature (T), number of flights (A) and wear (V). An actual value of the temperature parameter (T) is determined dependent on the number of flights (A) using measurement data. An existing wear (V) is determined based on the actual temperature value (T) and the characteristic field (KF). Using the actual temperature (T) and its limit value (G), a target value of number of flights (A), which corresponds to the maximal permissible wear, is obtained by extrapolation. A remaining service life is estimated by comparing the target value with a given number of flights corresponding to existing wear, An Independent claim is also included for planning servicing of one or more airplanes.

Description

The invention relates to a method for estimating the remaining service life of a Device that is subject to wear during operation. The The invention further relates to the use of such a method for Maintenance planning.

For many devices, such as turbines or jet engines of aircraft, is the wear of or on essential components of the Main reason for the limited lifespan, with "lifespan" usually the time span or the operating time between two ordinary revisions is meant. Wear can become apparent Changes in the mechanical properties of components manifest, for example, to friction, heat or fatigue is due. These changes to the components also have an effect a change in behavior of the operating conditions entire device. It is therefore necessary at regular intervals To carry out revisions or maintenance work on the device.

Methods known today for planning such maintenance work are based Usually at purely time-defined maintenance intervals, that is, it is a fixed time interval for the lifespan between two revisions stated. For security reasons, very conservative approximations for assumed the lifetime already used. It is clear that such is pure time-based process does not lead to optimal use of the Possibilities of the device lead because of the actually occurred Wear, which is also due to the concrete operating and Ambient conditions is influenced, is not taken into account. To the Efficiency of use of such devices, for example Aircraft engines, it is therefore desirable to increase one to have a more realistic estimate of the remaining lifespan at which also the actual operating and environmental conditions be taken into account in order to make it more efficient and economical Allow maintenance planning.

The present invention is dedicated to this task. So it should be a Method for estimating the remaining life of a device be proposed in which the specific operating conditions and thus the effective or real life of the device is taken into account.

The method that solves this problem is characterized by the features of characterized independent process claim.

a) für mindestens einen charakteristischen Parameter, der sensitiv für den Verschleiss ist, wird ein Zusammenhang mit einer Zeitgrösse ermittelt, welche repräsentativ für die Betriebsdauer ist;

b) für den charakteristischen Parameter wird ein Grenzwert festgelegt, der den maximal zulässigen Verschleiss angibt;

c) es wird ein Kennfeld erstellt, das einen Zusammenhang zwischen dem charakteristischen Parameter, der Zeitgrösse und dem Verschleiss angibt;

d) mit Hilfe messtechnisch erfasster Daten werden Ist-Werte für den charakteristischen Parameter in Abhängigkeit von der Zeitgrösse ermittelt;

e) aus den Ist-Werten wird jeweils anhand des Kennfelds der momentan vorhandene Verschleiss ermittelt;

f) ausgehend von dem momentanen Ist-Wert des charakteristischen Parameters wird mittels Extrapolation auf den Grenzwert bestimmt, für welchen Endwert der Zeitgrösse der maximal zulässige Verschleiss erreicht wird;

g) durch Vergleich dieses Endwerts mit dem Wert für die Zeitgrösse, der zu dem momentan vorhandenen Verschleiss gehört, wird die Restlebensdauer geschätzt.

According to the invention, therefore, a method for estimating the remaining service life of a device which is subject to wear during operation is proposed, with the following steps:

a) for at least one characteristic parameter that is sensitive to wear, a relationship is determined with a time variable that is representative of the operating time;

b) a limit value is specified for the characteristic parameter, which indicates the maximum permissible wear;

c) a map is created which indicates a relationship between the characteristic parameter, the time size and the wear;

d) using measured data, actual values for the characteristic parameter are determined as a function of the time variable;

e) the current wear is determined from the actual values on the basis of the characteristic diagram;

f) on the basis of the current actual value of the characteristic parameter, extrapolation to the limit value is used to determine the end value of the time variable for which the maximum permissible wear is achieved;

g) the remaining service life is estimated by comparing this end value with the value for the amount of time associated with the wear currently present.

In the method according to the invention, the wear is based on quantifies at least one characteristic parameter that is sensitive regarding wear. A map is created for the device, which is the relationship between the characteristic parameter, the amount of time and the wear. In the case of only one characteristic parameter, this map can be viewed as a surface in a three-dimensional space, which space through the characteristic parameters, the time size and the wear is spanned.

Data is recorded on the device from which an actual value can be obtained can be determined for the parameter depending on the time. The map can then be used to determine how far the wear has progressed quantitatively, for example in percent. By a Extrapolation to the maximum permissible limit for the characteristic parameters can then be estimated the remaining life become.

The method according to the invention thus actually takes into account the effectively already used lifetime, based on that Estimating remaining life. So it's not just the time size are taken into account, but also the operating and, if necessary takes into account the environmental conditions under which the device was operated at the current time. This is operational Estimation of the remaining life, which is the history of the device taken into account, enables a much more efficient use of the Device because maintenance only needs to be carried out when they are actually necessary. The reliable prediction of the Wear development thus enables a condition-based Maintenance planning.

Preferably, the remaining life for different stages in the Lifecycle determined, that is, with increasing consumption of Lifetime - measured by the time size - becomes the remaining lifespan re-estimated several times.

In a particularly preferred embodiment, the map is also Created with a priori knowledge of wear behavior. There Devices such as aircraft engines (jet engines) very much represent complex systems, it is usually - if at all - only with greater effort possible a sufficiently precise physical or perform deterministic modeling of the device. Hence it will preferred, a priori know-how for the creation of the map consulted. The experiences, observations or measurements that you have collected on the same or similar devices used to determine the qualitative and / or quantitative behavior of the characteristic parameters with increasing wear describe. Simplified by using such a priori knowledge the creation of the map significantly. Furthermore, the use of the Prior knowledge usually means that the map of the device describes better or more precisely.

The a priori knowledge preferably includes the qualitative and / or quantitative course of wear curves showing the relationship specify between the characteristic parameter and the time variable.

Because the a priori know-how is usually available in verbal form includes, the map is particularly preferably by means of a linguistic Fuzzy model created. For example, this can be a purely qualitative one Model for determining the wear that is then generated the basis of measured life cycles regarding its quantitative Properties is sufficiently optimized.

It has also proven to be advantageous in practice if the map based on measurement data or based on plausibility considerations is modified. Such plausibility considerations have, for example proven to be very useful to map certain edge areas of the map model that correspond to conditions that the real device rarely reached.

It is further preferred if the data recorded by measurement technology is used for Determination of the actual values for the characteristic parameter one each Filtering or averaging. Often they are metrologically recorded data from a noise or other Interference size superimposed, so that their direct use, especially in Fuzzy models, no robust statement regarding the degree of wear allowed.

To determine the actual values, it is preferable to use several sentences A model is created from the data recorded using the measurement, with which an actual value is determined for the characteristic parameter.

The inventive method is particularly suitable for Estimating the remaining life of an engine, particularly an aircraft engine.

The inventive method is particularly well suited for Maintenance planning, especially of an aircraft or a variety of Aircraft (fleet management).

Further advantageous measures and preferred configurations of the Invention result from the dependent claims.

Fig. 1:

typische Verschleisskurven für ein Flugzeugtriebwerk,

Fig.2:

Zugehörigkeitsfunktion (membership function) eines Fuzzy Sets,

Fig. 3:

ein Kennfeld in einem dreidimensionalen Raum, der durch die A-Achse, die T-Achse und die V-Achse aufgespannt wird,

Fig. 4:

Eine Projektion von historischen Daten in die Ebene, die von der A-Achse und der V-Achse aufgespannt wird,

Fig. 5:

eine Darstellung der A-T-Ebene zur Illustration einer Extrapolation in einem Ausführungsbeispiel eines erfindungsgemässen Verfahrens,

Fig. 6:

wie Fig. 4, jedoch zusätzlich mit der Projektion einer Extrapolation; und

Fig. 7:

mehrere Darstellungen jeweils wie in Fig. 6, jedoch zu unterschiedlichen Zeiten des Lebenszyklus, zur Verdeutlichung der jeweiligen Schätzung der Restlebensdauer nach dem Ausführungsbeispiel des erfindungsgemässen Verfahrens.

The invention is explained in more detail below using an exemplary embodiment and the drawing. The schematic drawing shows:

Fig. 1:

typical wear curves for an aircraft engine,

Figure 2:

Membership function of a fuzzy set,

Fig. 3:

a map in a three-dimensional space spanned by the A-axis, the T-axis and the V-axis,

Fig. 4:

A projection of historical data into the plane spanned by the A-axis and the V-axis,

Fig. 5:

2 shows a representation of the AT level to illustrate extrapolation in an exemplary embodiment of a method according to the invention,

Fig. 6:

like FIG. 4, but additionally with the projection of an extrapolation; and

Fig. 7:

several representations each as in FIG. 6, but at different times in the life cycle, to clarify the respective estimate of the remaining life according to the exemplary embodiment of the method according to the invention.

The method according to the invention is described below with an example Character with reference to an aircraft jet engine explained. The engine is a representative example of one Device which is subject to wear during operation and for which should be used to estimate the remaining life. It goes without saying the invention is not limited to this application, but is suitable in similarly also for other devices such as land-based turbines, turbomachines or other mechanical Systems that are subject to wear during operation and therefore need to be serviced.

With the term "lifespan" is the section between two ordinary ones Revisions or maintenance meant, that means immediately after the Maintenance, the degree of wear is assessed as "new".

As a characteristic parameter the sensitive to the wear of the Aircraft engine is, the change in the gas outlet temperature of the Engine selected. In the following, this change in temperature is indicated by T designated.

As a time variable, which is representative of the operating time or the Is the number of flights selected in the following with A referred to as. Of course, other times for example, the number of operating hours.

The wear V is a number from the interval from zero to one described, where V = 0 means 'new', there is no wear yet occurred and V = 1 'used up', i.e. the permissible wear limit is reached. A wear of V = 0.6 means, for example, that the Degree of wear is 60% of the maximum permissible wear.

As a limit value G for the temperature change T, which means that the indicates the maximum permissible wear, a temperature increase of 40 K selected. This value is based on experience as a priori know-how be introduced.

Fig. 1 shows several typical wear curves K1-K4, each one possible relationship between the temperature change T as characteristic parameters and the number A of flights as a time variable specify. Such wear curves represent a priori knowledge, which in the present embodiment for creating the map is used. The wear curves are based on experience or metrologically recorded data that can be seen on the same or similar Engines.

1 is the limit value G = 40K for the temperature increase plotted as well as an interval L, which is a typical life cycle for a indicates such an engine.

It can be seen that the curves K1-K4 are all different look, but show the same quality behavior. At the beginning of the Life cycle that starts at A = 0 and T = 0 is initially a stronger increase the wear curves to recognize, there follows a plateau on which the temperature increase hardly changes and is towards the end of the life cycle again a stronger increase in the temperature increase can be observed. All Wear curves end at T = 40K, i.e. when the limit G is reached.

One way to quantify wear is to: the path integral normalized to one over the entire wear curve. Then defines the path length for each point on the wear curve appropriate wear. For example, if you've completed A1 flights, the quantified wear results from the path integral over the Wear curve from 0 to A1.

The problem, however, is that for a given engine that metrological wear curve only at the end of its life is known. In other words: for a new, i.e. freshly serviced Engine is not known, along which of any number Wear curves it will move during its life cycle.

In a somewhat simplified way, the one described here is used Embodiment of the inventive method in the course of progressive consumption of life estimate the Remaining lifespan improved by the fact that the prediction constantly (the means, for example, after every flight) on the correct wear curve is adjusted.

From the point of view of an input / output model, this is the characteristic Parameters and the size of the time, which is the lifetime used up to now measures as inputs and wear as output as necessary and sufficient information to estimate the remaining life use.

In the method according to the invention, a map KF is first (see Fig. 3) created that a relationship between the characteristic parameters (here the temperature increase T), the Time size (here the number A of flights) and the wear V.

This map is preferably created with the aid of a priori know-how, So for example using wear curves K1-K4 as they are shown in Fig. 1.

A is particularly preferred for the implementation of this a priori knowledge linguistic fuzzy model (MAMDANI model) used. Because such fuzzy Models or the fuzzy logic per se are sufficiently known to the specialist are only briefly explained here, as in the embodiment of The inventive method, the map KF with the help of a linguistic fuzzy model is created.

The linguistic fuzzy model stands for the one described here Embodiment following information as a priori know-how Disposition: Knowledge of the qualitative and quantitative course of Wear curves (see Fig. 1); additional requirements for the Forecast in peripheral areas.

The temperature increase T and so far serve as input variables Life spent, measured by the number A of flights, as Output size of wear and tear V. Each of the input and output sizes is characterized by a fuzzy set. As an example, Fig. 2 shows the Membership function of the fuzzy set for the Input variable temperature increase T. For the linguistic variable T are linguistic values "small", "medium" and "large" are provided.

For input quantity A "number of flights", the linguistic values are "little", "much", "limit" and for the output size "wear" V die linguistic values "new", "as good as new", "used", "used".

IF (T is klein) AND (A is wenig) then (V is neu)

IF (T is klein) AND (A is viel) then (V is neuwertig)

IF (T is mittel) AND (A is viel) then (V is gebraucht)

IF (T is gross) AND (A is limit) then (V is verbraucht)

The set of rules is then defined using the fuzzy sets, which in the specific case comprises the following four rules

IF (T is small) AND (A is little) then (V is new)

IF (T is small) AND (A is much) then (V is as new)

IF (T is medium) AND (A is much) then (V is used)

IF (T is large) AND (A is limit) then (V is used)

From this information, a map KF can then be created as in Fig. 3 is shown as a mesh. Of course it is possible and usually also common that first a "draft" for the Map KF is created and this is then refined. such Refinements can be made, for example, by experimental and Error strategies (trial and error), consideration of measurement data, manual re-modeling based on plausibility considerations (e.g. that the edge of the map lies on the line with T = 40K and V = 1) or other optimization or calibration procedures. The result resulting map KF for the embodiment described here is in Fig. 3 shown.

With the help of this map KF, which shows the relationship between Temperature increase T, number of flights A and wear V describes, the estimate of the Remaining lifetime made, which is explained below.

During the life cycle of the engine, data is measured of the engine - for example during or after each flight - from which are then actual values for the characteristic parameter (here the Temperature increase T) depending on the amount of time (here the number A of flights) can be determined. Of course it depends on the specific Application also possible and possibly also advantageous, as actual values to use the data recorded directly by measurement, in However, the present exemplary embodiment has proven to be advantageous the metrologically recorded data for filtering or averaging undergo. The reason for this is that the change in Gas outlet temperature T due to the high absolute value of the A strong noise is superimposed on the gas outlet temperature. A typical noise amplitude can be half or even more than that measured data signals.

In principle, of course, many methods and methods known per se are suitable Methods to measure the data of an averaging or a Undergo filtering. With an exemplary character it becomes a possibility explains that uses a method which is described in EP-A-0 895 197 (P.6822).

The change in the gas outlet temperature T is recorded at regular intervals, for example during each flight. This leads to a data record of the form [T _i ; A _i] with i = 1, ..., n, where T denotes _i the temperature change for the flight with the number A _i. Now a part of this data set, for example the data with i = 1,2, ..., 20; the data with i = 21.22, ..., 40; etc. used to create a model. This leads to a variety of models. Each of these models is then evaluated for the same state, the so-called nominal state, in order to determine the actual values for the characteristic parameter. For further details, reference is made to EP-A-0 895 197.

This measure averages the metrologically recorded Data that significantly reduces the noise amplitude. In this way actual values for the temperature increase T as a function of the Determine number A of flights.

The historical measurement data or the actual values for the temperature increase T up to the present are available for the engine at the time at which the remaining service life is to be estimated. Each pair of values (T _k , A _k ) from an actual value for the temperature increase and the associated service life, measured by the number of flights A _k, corresponds to a point in the plane spanned by the A axis and the T axis (see Fig 3). For reasons of a better overview, only one such point with the coordinates (T _k , A _k ) is drawn in the TA plane as a cross in FIG. 3. This point is now projected onto the map KF, that is to say upward onto the map KF represented by the network. Then the wear V _k belonging to (T _k , A _k ) can be read off directly.

As already mentioned, all pairs of values (T _k , A _k ) are known up to the present. If each pair of values is mapped onto the map KF and the result is projected onto the plane that is spanned by the A-axis (number of flights) and the V-axis (wear), the curve shown in FIG. 4 results, for example. For all historical data, it describes the quantified wear V as a function of the lifetime already used, measured by the number A of flights. In the concrete example, about 1300 flights have been completed at the moment and the wear is around 55%.

For the prediction of the future course of wear V and thus for the remaining service life is now extrapolated to the Limit G for the temperature increase T. Of course there are numerous extrapolation options. Choosing a suitable one Extrapolation depends on the specific application. It must in particular, how conservative the estimate can be or should, that means high demands on the security of the forecast must be asked. Accordingly, extrapolation considered more or less pessimistic developments and can be calculated by extrapolation.

In very sensitive applications, such as in the present case of an aircraft engine, one becomes with more pessimistic forecasts work to have enough safety reserves and to achieve the Wear limit before the estimated remaining life excluded.

In the exemplary embodiment described here, a pessimistic one Development of the temperature increase derived from the number of flights that are normally required at least around the limit G = 40K for the To achieve temperature increase T. For example, you can such wear curves K1-K4, as shown in Fig. 1, Estimate the minimum number of flights A until reaching the Limit is. In Fig. 1, this minimum number is given by the point where the wear curve K1 reaches the limit.

3 results from the data on which the characteristic diagram KF is based minimum number of flights to 2000 flights.

The procedure for extrapolation is now as follows. Fig. 5 shows the plane spanned by the A-axis and the T-axis. As already shown in FIG. 3, only one point (A _k , T _k ) is shown. This is the most current point, i.e. the one that corresponds to the present, in which the estimate is made. From this point, as the dash-dotted line E shows, linear extrapolation is made to the limit value G = 40K for the temperature increase T. The gradient with which the line E is "appended" to the point (T _k , A _k ) has been determined as follows. Assuming that the minimum number of flights to reach the limit value G of the temperature increase T, as mentioned above, is A = 2000 flights, the associated incline is determined, which results if the points T = 0, A = 0 (new condition) and T = 40K, A = 2000 flights connected by a straight line. The resulting gradient is then reduced for safety reasons, in the present example it has been halved, that is, it has been assumed that the limit is reached after half the number of flights. The resulting extrapolation is line E shown in FIG. 5.

In the next step, the line E is mapped into the map KF, whereby it generally becomes a curved line. Now project that Map KF again - analogous to the representation in Fig. 4- on the of the V-axis and the plane spanned by the A axis, the result is that in FIG. 6 shown representation. In addition to the historical values (which of course are identical with those in Fig. 4) the projection E 'of the extrapolation is now too see that is shown in dashed lines. The projection E 'of extrapolation is reached the wear limit V = 1 after a number of flights, the final value with AG is designated. The estimate of the remaining service life is now given by Difference formation from the final value AG and the current value for the Number of flights, the most recent of the historical values. In Fig. 6 this is the last point of the solid curve.

This method of estimating the remaining life will now be used during the Lifetime of the engine continuously or at predefinable intervals repeated. The forecast horizon is essentially through the choice of the gradient is determined. Choosing a reasonable gradient has application-specific.

7 illustrates the estimate of the Remaining life for seven stages in the life cycle of an engine. In each of the figures in Fig. 7 is in the same way as in Fig. 4 and 6 shows the plane spanned by the V-axis and the A-axis, which results from the projection of the map. The historical data, the means those data that are based on measurement data, are each shown as solid lines, the extrapolations that the Underlying the estimate are shown in dashed lines. The transition from the solid to the dashed line indicates the respective present, in who made the estimate. To the right of the illustrations is the Estimated remaining lifespan RL in number of flights. In the top one The presentation in which no historical data are yet available is Residual life essentially determined by the selected gradient.

In the second to fifth figure seen from above there is always one The remaining lifespan of around 500 flights is forecast. That is, the For example, aircraft operators get after 650 flights (third Figure) the information that there are still at least 500 flights to the next Maintenance are possible. After about 1000 flights (fourth illustration) gets the operator the information that there are still at least 480 flights are feasible. After just over 1300 flights (fifth illustration) the operator receives the information that there are still at least 480 flights are feasible. These predictions also in the second to fifth Representation are still essentially through the forecast horizon and thus determined by the choice of the gradient.

Only in the sixth representation (sixth estimate) does the end appear the life cycle, so that the operator can now plan the revision can start. Based on the chosen gradient for extrapolation the looming end of the life cycle is about 500 flights away Detect when the wear limit is reached. If one relates this to the entire life cycle of about 2000 flights, the operator remains for the maintenance planning a considerable scope of action of about 25% of the Lifecycle.

The forecast horizon can be selected by choosing a smaller one Extend or extend gradients for extrapolation. This would in practice, however, does not lead to a significantly greater benefit, however increase the risk of an overly optimistic estimate of the remaining life.

By estimating the remaining life of the Aircraft engine that takes quantified wear into account possible to use the potential of the engines much more efficiently without that concessions to operational security are required. The entire maintenance or revision planning can be optimized, thereby a economically significantly cheaper operation is made possible. So you can the method according to the invention in particular for maintenance planning Advantageously use aircraft engines. This also enables a clearly More efficient planning of maintenance work on a large number of aircraft. The method according to the invention consequently enables an extremely efficient fleet management of an entire fleet of aircraft.

The type of extrapolation described above is of course only exemplary understand. Other types of extrapolation, e.g. B. non-linear or based on qualitatively known developments. The special choice of one adapted to the respective application Extrapolation poses no problems for the person skilled in the art.

Even if in the specifically described preferred embodiment a linguistic fuzzy model for determining the map Processing of a priori know-how is used, that is the invention not limited to such modeling.

It is also not necessarily the case that the map uses a priori know-how must be created. Other methods are possible to to determine a map which shows the relationship between the characteristic parameters, the time size and the wear indicates. So can, for example, depending on the application, use ab initio calculations be carried out, or design calculations, Dimensioning calculations, physical modeling, data-based Modeling, system behavior calculations. It is also possible that Map in the form of polynomials, look-up tables, multilayer perceptrons (neural networks), radial basis functions, Singleton and Takadi-Sugeno Describe fuzzy models as well as hanging hyperplanes.

Of course, it is also possible to have more than one characteristic for to use the wear-sensitive parameters, which makes the map is defined in a higher dimensional space.

Claims (10)

A method for estimating the remaining life of a device that is subject to wear during operation, comprising the following steps: a) for at least one characteristic parameter (T) that is sensitive to wear (V), a relationship is determined with a time variable (A) that is representative of the operating time; b) a limit value (G) is specified for the characteristic parameter (T), which indicates the maximum permissible wear; c) a map (KF) is created, which indicates a relationship between the characteristic parameter (T), the time variable (A) and the wear (V); d) using measured data, actual values for the characteristic parameter (T) are determined as a function of the time variable (A); e) the current wear (V) is determined from the actual values on the basis of the characteristic diagram (KF); f) on the basis of the current actual value of the characteristic parameter (T), extrapolation to the limit value (G) determines for which end value of the time variable (A) the maximum permissible wear is achieved; g) the remaining service life (RL) is estimated by comparing this end value with the value for the time quantity that belongs to the wear currently present. The method of claim 1, wherein the map (KF) with the help is created from a priori knowledge of wear behavior. The method of claim 2, wherein the a priori knowledge qualitative and / or quantitative course of wear curves (K1, K2, K3, K4), which covers the relationship between the Specify characteristic parameters and the time size. Method according to one of the preceding claims, in which the Map (KF) is created using a linguistic fuzzy model. Method according to one of the preceding claims, in which the Map (KF) based on measurement data or based on Plausibility considerations are modified. Method according to one of the preceding claims, in which the Map (KF) represents a surface in a three-dimensional space, what space by the characteristic parameter (T) that Time size (A) and the wear (V) is spanned. Method according to one of the preceding claims, in which the metrologically recorded data to determine the actual values for the characteristic parameters of filtering or averaging be subjected. Method according to one of the preceding claims, in which with A model using several sets of metrologically recorded data is created with an actual value for the characteristic parameter is determined. Method according to one of the preceding claims, in which the Device is an engine, in particular an aircraft engine. Use of a method according to one of claims 1 to 9 for maintenance planning, especially of an aircraft or a large number of airplanes.

Images